A knowledge-based system approach to work shift selection for multilane highway reconstruction and maintenance projects

MISSING IMAGE

Material Information

Title:
A knowledge-based system approach to work shift selection for multilane highway reconstruction and maintenance projects
Physical Description:
xi, 221 leaves : ill. ; 29 cm.
Language:
English
Creator:
Ahmed, Q. Amin
Publication Date:

Subjects

Subjects / Keywords:
Civil Engineering thesis Ph.D
Dissertations, Academic -- Civil Engineering -- UF
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1993.
Bibliography:
Includes bibliographical references (leaves 218-220).
Statement of Responsibility:
by Q. Amin Ahmed.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001922331
oclc - 30495384
notis - AJZ8143
System ID:
AA00004731:00001

Full Text












A KNOWLEDGE-BASED SYSTEM APPROACH TO WORK SHIFT SELECTION
FOR MULTILANE HIGHWAY RECONSTRUCTION AND
MAINTENANCE PROJECTS













By

Q. AMIN AHMED


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF
THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1993





























To my parents













ACKNOWLEDGEMENTS


I would like to extend my deep gratitude to Dr. Ralph D. Ellis,

my supervisory committee chairman, for his guidance and dedicated

support, which made this study possible. A special thank you is due to

Dr. Zohar Herbsman for his encouragement and valuable recommendations

throughout my doctoral studies. My sincere appreciation goes to Dr.

Fazil T. Najafi, Dr. Paul Y. Thompson, and Dr. Ajay Shanker for serving

on my supervisory committee. Additionally, I am grateful to Dr. Ronald

A. Cook for serving as an observer in my committee.

I wish to thank Mr. Ananth Prasad and Mr. Henry Haggerty of the

Florida Department of Transportation for their personal efforts in

providing valuable information for this research. I am grateful to Mr.

Ashish Kumar, graduate assistant in construction engineering, for his

support and advice--particularly for the statistical part of this study.

Finally, much gratitude is owed to my parents for their love,

continuous encouragement, and support during the entire course of my

studies.














TABLE OF CONTENTS

page

ACKNOWLEDGEMENTS ........................................... iii

LIST OF TABLES ......................................... vii

LIST OF FIGURES .......................................... ix

ABSTRACT .................. ................................ x

CHAPTERS

1 INTRODUCTION AND PROBLEM STATEMENT ................ 1

Work Shift Selection in Highway Construction ....... 1
Need to Identify Alternate Shift Times ............ 2
Day vs. Night Shift .............................. 3
Decision Factors ...................... ........... 6
Need for Decision Model ......................... 6
Research Objectives ................... ............ 9

2 REVIEW OF LITERATURE ............................ 11

Introduction ....................................... 11
Overview of Factors Attributable to Day and Night
Shift Work ....................................... 12
Traffic Volume ..................... .. ............ 13
Work Zone Safety ................................. 21
Public Relations ................................... 29
Noise Level ........................................ 30
Cost .......................................... 32
Quality of Work ................................ 41
Job Types in Highway Construction ................ 46
Human Factors ................................... 50
Productivity ........................................ 54
Other Factors ................................... 55
Summary ........................................... 57

3 KNOWLEDGE-BASED EXPERT SYSTEM TECHNOLOGY .......... 58

Introduction ..................................... 58
Architecture of an Expert System ................. 59








Knowledge Representation ........................ 63
Inference Engine .................................. 69
Knowledge Acquisition ........................... 70
Expert System Development Shells ................. 72
Current Applications in Construction Engineering .. 74
Summary ........................................ 75

4 PROJECT COST AND PRODUCTIVITY DATA FOR DAY AND
NIGHT SHIFTS ................................... 77

Introduction .................................... 77
Project Cost Comparison ......................... 78
Productivity Comparison ......................... 86
Summary ........................................ 93

5 MODEL APPROACH FOR WORK SHIFT SELECTION ........... 94

Introduction .................................. 94
Lane Closure Traffic Analysis ................... 96
Accident Analysis ............................... 99
Project Cost Analysis ........................... 100
User Cost Analysis ............................ 102
Summary ........................................ 105

6 DEVELOPMENT OF KNOWLEDGE-BASED SYSTEM MODEL ....... 107

Introduction ....................................... 107
Knowledge Acquisition ............................ 108
Knowledge Base Objective ........................ 110
Formulation of Rules ............................ 120
External Database Interface ..................... 123
Solution Search Technique ....................... 125
Explanation/Help Facility ....................... 125
Summary ........................................ 127

7 MODEL TESTING--CASE STUDY ....................... 128

Introduction .................................. 128
Case Study I .................................... 128
Case Study II .................................... 132

8 CONCLUSIONS AND RECOMMENDATIONS ................. 135

Summary and Conclusions ......................... 135
Recommendations ................................. 138








APPENDICES

A


B

C

D

E

F

G


PROGRAM INTRODUCTION, QUALIFIERS, CHOICES, AND
VARIABLES ......................................

KNOWLEDGE BASE RULES ..............................

EXAMPLE OF EXTERNAL TRAFFIC COUNT DATABASE ........

PROGRAM HELP FILES ................................

EXAMPLE OF PROJECT DATA COLLECTION FORM ...........

PROGRAM INPUT DATA AND RESULTS ....................

INTERVIEW RESULTS .................................


REFERENCES ...................................................

BIOGRAPHICAL SKETCH ........................................














LIST OF TABLES


Table page

1.1 Work Hour Restrictions (In Accordance to Traffic
Plan Developed by Maryland & Virginia)............... 4

1.2 Highway Construction Work Shift Selection--
Factors To Be Considered .......................... 7

2.1 Lane Capacities for Selected Metropolitan Areas ..... 15

2.2 Number of Accidents Before and During Construction
in Seven States ...................................... 23

2.3 Summary of Work Zone Accidents at Night ............. 24

2.4 Fatal Accidents on Urban Freeways by Time of Day
and Day of Week ................................... 28

2.5 Estimated Increase in Cost Differential
of Night Work .......................................... 34

2.6 1-70 Night Paving Project vs. 1-25 Day Paving
-- Percent Cost Differential ...................... 35

2.7 Costs to Purchase, Install, and Remove Light Posts .. 39

2.8 Paving Quality Test Results from Field Samples ...... 43

2.9 Quality Rating of Nighttime Operations .............. 45

2.10 Florida DOT Projects Using Night Work .............. 48

2.11 Construction Activities Performed Better At Night ... 49

3.1 Difference Between Expert System and
Conventional Program ............................... 60

4.1 Statistical Summary of Unit Costs for Selected Work
Items for All Daytime FDOT Projects in 1990 ......... 79

4.2 Statistical Summary of Unit Costs for Selected Work
Items for All Nighttime FDOT Projects in 1990 ....... 80








4.3 Difference Between Day and Night Unit Costs for
Selected Work Items for All FDOT Projects in 1990 ... 81

4.4 Quantities of Work Items for Eight Selected FDOT
Night Projects ...................................... 83

4.5 Effect of Quantity on Project Costs for Eight
Selected Work Items ............................... 84

4.6 Summary of Productivity Rates for FDOT Construction
Projects ................................... ... 87

4.7 Summary of Productivity Rates for FDOT Nighttime
Construction Project on I-95 in St. Johns County .... 88

4.8 Guidelines for Estimating Production Rates for
FDOT Projects ..................................... 89

4.9 Statistical Test for Day & Night Production Rate
of Plant Mixed Surface ............................. 91

4.10 Statistical Test for Day & Night Production Rate
of Milling Existing Surface ........................ 92

5.1 Decision Factors That Influence Shift Selection ..... 95

5.2 1-75 (North Bound) Average Weekday Hourly Traffic
Volumes. Location: Hamilton County ................ 98

5.3 Personal Cost of Time Delay for Queuing ............. 104


viii













LIST OF FIGURES


Figure page
1.1 Research Development Flowchart ...................... 10

2.1 Traffic Flow on a 4-Lane Freeway ..................... 17

2.2 Traffic Volume/Capacity Relations .................... 19

2.3 Effect of Lane Closure at Different Hours ............ 20

2.4 California Urban Freeway Fatal Accident Rates ........ 26

2.5 Triad of Shift Work Coping Factors ................... 51

3.1 The Expert System Architecture ...................... 61

3.2 A Frame Representation of Knowledge .................. 64

3.3 Knowledge Representation Using Semantic Network ...... 66

6.1 Decision Tree Segment for Work Shift Selection:
Congestion Analysis ................................. 114

6.2 Decision Tree Segment (Closed) for Work Shift
Selection: Accident Analysis ....................... 115

6.3 Decision Tree Segment (Open) for Work Shift
Selection: Accident Analysis ....................... 116

6.4 Decision Tree Segment (Open) for Work Shift
Selection:Noise, Quality, Productivity Considerations.. 117

6.5 Decision Tree Segment (Open) for Work Shift Selection:
Experience, Supply and Temperature Considerations ..... 118

6.6 Decision Tree Segment (Closed) for Work Shift Selection:
Supervision/Communication and Human Factors ........... 119













Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

A KNOWLEDGE-BASED SYSTEM APPROACH TO WORK SHIFT SELECTION
FOR MULTILANE HIGHWAY RECONSTRUCTION AND
MAINTENANCE PROJECTS

By

Q. Amin Ahmed

August 1993

Chairman: Dr. Ralph D. Ellis
Major Department: Civil Engineering

The timely, efficient, and quality completion of highway projects

partly depends on the selection of the appropriate work shift. Certainly

there are advantages and disadvantages for both day and night shifts.

Daytime operations are generally considered to be safer for both workers

and motorists because of better visibility and a higher state of

alertness. However, to accommodate the flow of higher daytime traffic

volumes, work zone lane closures may not be possible--and the option to

work during the night becomes a serious consideration. A highway

project's characteristics may favor a particular work shift. Type of

work, lane capacity, average daily traffic (ADT), work zone accidents,

project duration, and project costs are some of the issues that may

dictate shift times.

This research is aimed towards the development of a decision model

that will incorporate all the factors that may influence the selection of

either a day or a night work shift. The process is based on a decision

x








tree representing qualitative and quantitative factors, ranked in the

order of importance. The shift selection methodology is similar to the

human reasoning process and that is why a knowledge-based system approach

has been chosen for this decision model. An expert system shell has been

utilized to develop the knowledge base, which consists of IF-THEN logic

rules. The rule-based knowledge structure also has the option to

interface with an external traffic count database for the purpose of lane

closure analysis.

The model approach to work shift selection includes mathematical

reasoning in the analysis of traffic congestion, vehicle accident numbers,

motorist (user) costs, and project (owner) costs. The final solution

offered by the knowledge-based system model consists of the recommended

work shift, number of daytime lane closures allowed, user cost savings for

a night shift option, and percent change in total owner cost for a

nighttime alternative.













CHAPTER 1
INTRODUCTION AND PROBLEM STATEMENT

Work Shift Selection in Highway Construction


In recent years, in Florida, as in many other states, the emphasis

in highway construction has primarily been on rehabilitating existing

facilities. Resurfacing, widening, multilane reconstruction, and bridge

repair are typical activities currently taking place in various states.

A major concern for state transportation agencies during construction and

rehabilitation of the nation's highway system is to minimize public

inconvenience. Work zone traffic accidents, congestion, noise, and cost

to the delayed motorists are issues that may lead to adverse public

reaction. In the urban areas, required lane closures during

rehabilitation work are resulting in heavy traffic congestion. This

problem is not limited to roads in urban locations, but also includes

freeways that become crowded during peak travel periods. For example, in

Florida, Interstate 75 is so crowded during holiday periods that it

resembles an urban main street. I-4 near Orlando often becomes a "local"

street due to congestion during tourist season.

Agency experience or mathematical reasoning can be utilized to

evaluate the acceptability of this congestion level. A recent survey of

ten state agencies done by researchers indicates a cutoff volume of

approximately 1500 vehicles per hour per lane at which backups are

expected (Shepard and Cottrell, 1985). As the traffic volume exceeds the










cutoff point, the agency is obliged to consider alternate working hours--

namely, night shift. However, noise ordinances in some areas may prevent

usage of heavy and noisy equipment during night hours. The selected shift

hours must be such that traffic congestion and public disturbance are kept

minimal and that overall safety requirements are adequately met during

those periods. After these preliminary criteria are satisfied, the state

agency will have to consider or weigh a variety of other factors (positive

and negative) that are associated with different shift times. In addition

to the many site-related variables, there are human and sociological

factors that can be attributed to a work shift and which in turn may

influence the timely, efficient, and quality completion of highway

projects. Different work shifts have different effects on a worker's

health, attitudes, and performance. Thus, the selection of appropriate

work shift hours is an important project management decision for the state

highway agency.


Need to Identify Alternate Shift Times

Unlike the manufacturing industry where production takes place in an

enclosed, controlled surrounding, highway construction takes place in a

dynamic environment, influenced by many external factors. Although the

highway work shift may be broadly classified by "day" and "night" shifts,

it is necessary also to identify the specific hours that provide the ideal

work environment. These shift times will no doubt vary from project to

project, when all the factors are taken into account. In some situations,

the work shift may even be a combination day and evening hours. When

working in major metropolitan areas, highway agencies typically restrict








3

their roadway construction and maintenance activities to hours of off-peak

traffic and weekends.

In certain areas, highway construction and maintenance work can

proceed unhindered during "off-peak" hours, i.e., 9 a.m. to 3 p.m.

However, there are situations where lanes cannot be closed during the day

at all due to high traffic volumes. Traffic congestion in some areas may

last up to 12 or 14 hours a day. Nighttime operation is now being

considered as a viable and necessary option, not only from the point of

view of project management, but also from the public relations

perspective. Table 1.1 shows work hour restrictions for varying traffic

conditions in accordance with the traffic control plan developed by

Maryland and Virginia highway agencies (Shepard and Cottrell, 1985). Each

traffic condition represents different lane closure situations.


Day vs. Night Shift

The obvious advantage of night work is that there is less

interference from heavy traffic, allowing efficient project scheduling.

According to Lee (1969), a concrete paving operation in California

conducted at night was finished in 16 working days, whereas the same

project would have taken a minimum of 35 days for completion during the

daytime. A day shift would have provided fewer working hours and more

interference from heavy traffic, resulting in delay of material delivery.

Quality of construction may or may not differ during day and night

shifts. Price, in his 1985 report on nighttime paving, has stated that

overall quality on a night paving job (1-70) in Denver, Colorado, was very

similar to a well done day job (1-25 north of Denver). Compaction did not














Table 1.1 Work Hour Restrictions (In Accordance with the Traffic Control
Plan Developed by Maryland & Virginia)


Traffic Condition
Day

I II III

Monday 6:00 am 9:00 am 9:00 am 3:30 pm 12:00 pm 6:00 am
through
Thursday 3:30 pm 8:00 pm 8:00 pm 12:00 pm

6:00 am 9:00 am 9:00 am 3:00 pm 12:00 pm 6:00 am
Friday
3:30 pm 8:00 pm 9:30 pm 12:00 pm

12:00 pm 6:00 am
Saturday N/A 7:30 am- 8:00 pm
8:00 pm 12:00 pm

10:30 am 10 pm* 12:00 pm 10:30 am
Sunday N/A 10:00 pm 12:00 pm*
10:30 am 8 pm** 8:00 pm 12:00 pm**


Traffic Condition I : No lane closures permitted, all lanes open
to traffic.
Traffic Condition II: Single lane closure (median, center, right)
permitted.
Traffic Condition III: Three, two or one lane closure permitted.

* After Easter weekend through second Sunday in September.
** Third Sunday in September through weekend before Easter.

Source: Shepard and Cottrell, 1985








5

suffer despite cooler night temperatures. The only aspect of quality that

deteriorated during night work was aesthetics.

There is a cost advantage in night operation in terms of user costs,

congestion and delay being less during night hours. A user cost analysis

done by Price revealed a total savings of $1,154,900 for the motorists in

an 1-70 night project compared to an equivalent 1-25 daytime project.

However, construction costs tend to be higher during the night shift.

Price (1985), in comparing the 1-70 and 1-25 projects, has stated that the

average cost per item was approximately 40% higher for the night paving

job (1-70). Material acquisition, extra personnel, and lighting fixtures

are some of the items that have additional costs during night work.

The issue of safety, for both construction workers and vehicular

traffic, is an important consideration in deciding shift times. Safety

records in highway work zones can indicate the factors that cause

accidents and whether they can be attributable to certain shift hours.

For a meaningful comparison of safety and accident characteristics of

night and day shifts, it is necessary to consider many variables, such as

duration of lane closure, proportion of closed lanes, job type, closure

length, traffic control devices, etc. At present, there are no general

conclusions regarding safety on various work shifts. According to Lum

(1980), in his study of 7 states, construction during night shift did not

increase the percentage of total night accidents to overall accidents

(constant at 30%). However, Lum's research did reveal that there was a

9.4% increase for night accidents as a result of construction, which means

that there is a similar increase for day accidents due to construction.








6

Another issue in project management, which may be of concern during

a particular work shift, is communication. Communication between the

highway agency and the contractor may be difficult during night work

hours. If the proper authority cannot be reached for consultation at a

certain time, construction work could come to a virtual halt. A similar

concern arises for equipment maintenance. Equipment failure would result

in production delay if parts or mechanics were unavailable during a work

shift.

Decision Factors

The highway agency, as a decision maker, will have to consider both

qualitative and quantitative factors in the work shift selection process.

Examples of qualitative factors are agency policy, public relations,

noise, job type, communication, worker morale, work quality, and

temperature (see Table 1.2). Highway capacity, traffic volume, motorist

delay resulting from traffic congestion, and accident rate are typical

quantitative variables. The cost differential factor, if it can be

quantified, may be categorized as quantitative; otherwise, it can be

qualitative in nature, being higher or lower.

Need for Decision Model

Currently there is no decision model available for state highway

agencies that will enable them to select work shifts quickly and

efficiently, taking all possible factors into consideration. For most

highway projects, state agency engineers select the work shift hours based

on judgment and experience, with some usage of mathematical models to


















Table 1.2 Highway Construction Work Shift Selection--Factors To
Be Considered


Quantitative

Traffic Volume vs. Roadway
Capacity

User Costs Due to Delay at
Work Zone
Vehicle
Fuel
Personal

Owner Costs
Project Item Cost
Administrative Cost

Work Zone Accidents


Qualitative

Agency Policy

Public Relations

Noise

Worker Morale

Work Quality

Temperature

Job Type

Communication








8
calculate expected traffic volumes and delay times. But there is no

systematic methodology to "optimize" the shift selection process. A

project may be done during the night unnecessarily, resulting in increased

project costs. Work shift selection by the state agency is aimed towards

public satisfaction mainly; owner costs are largely ignored. Some

projects remain behind schedule because of delays caused during a

particular work shift. For example, at night, materials or spare parts

may be available on a limited basis; during the day, work progress may

slow down due to traffic interference. An interactive consulting tool,

such as a knowledge-based expert system, would aid the highway agency in

optimizing the selection of a particular work shift and identify the

reasons for the selection. For instance, if daytime traffic volume

allowed lane closures and night work were determined to be unsafe, then

certainly the appropriate work period choice would be day shift, even if

cooler night temperature is preferred.

The knowledge of the experienced decision makers in various state

agencies is scattered in the form of literature in reports, journals,

survey results, and guidelines. Although various state agencies have

guidelines that determine allowable daytime lane closures based on traffic

analysis, there are many other qualitative factors that are not considered

in a formal manner. This information needs to be compiled and made

available to state highway engineers in the form of a knowledge base. A

rule-based expert system approach is ideally suited for the type of

decision-making problem discussed above. Its interactive feature and ease

of use makes it a suitable substitute for a human consultant.










Research Objectives

A systematic, focused research in a narrow domain will enable the

practical development of a prototype expert system model for work shift

selection in highway construction. Figure 1.1 illustrates the research

development flowchart. To this end, the goals of the research study are:

1. To visit several district offices of the Florida Department of

Transportation (FDOT) to investigate the decision-making

process used in selecting work shifts for highway construction

and interview experts in highway construction.

2. To review current procedures used by the FDOT and other state

agencies nationwide for construction projects during day and

night shifts. Cost differentials, accident and safety, human

performance, and case studies of successful night operations

are some of the areas for review.

3. To perform a statistical analysis/comparison to determine if

there are any significant differences in project related costs

and productivity rates between day and night shifts.

4. To identify the various parameters necessary for the

development of knowledge-base rules for an expert system model

for highway work shift selection.

5. To utilize PC-based shell EXSYS Professional in designing a

knowledge-base structure.

6. To evaluate the possible uses of the model by using a sample

case study approach.

7. To summarize the final outcomes and state the conclusions

drawn forth from this research.














INTRODUCTION &
PROBLEM STATEMENT


LITERATURE REVIEW
LITERATURE REVIEW


KNOWLEDGE-BASED SYSTEM
DEVELOPMENT


SKNOWLEDE-BASED SYSTEM
MODEL TESTING
Case Studies


SUMMARY AND
RECOMMENDATIONS





Figure 1.1 Research Development Flowchart


KNOWLEDGE ACQUISITION
. Interview Experts
. Review Current Procedures
. Review Research Studies
. Data Analysis


KNOWLEDGE BASE STRUCTURE
SDecision Tree
Usage of ES Shell
SFormulation of Rules


I I


I













CHAPTER 2
REVIEW OF LITERATURE

Introduction

Recent literature on highway construction emphasizes the need to

perform rehabilitation work during shifts other than the typical day shift

utilized in most industries. In urban areas the problem of congestion is

of particular concern. To accommodate rehabilitation and improvement

activities on the freeways, state agencies are required to close traffic

lanes, creating heavy congestion on roads already loaded to capacity. The

congestion has created problems for state agencies and contractors in

terms of safety and project scheduling. The situation is also hazardous,

costly, and inconvenient for the traveling public. As a result, more

construction and rehabilitation work is being performed during the night

shift, when traffic flow is minimum.

This literature search was done to identify the principal factors

involved in deciding shift times for highway construction projects.

Qualitative and quantitative attributes that are linked to activities

during different work periods were obtained from various highway

construction literature. Specifically, unique aspects of day and night

operations were targeted for identification. A computer database search

through the Southern Technology Applications Center (STAC) in Gainesville,

Florida, revealed over 25 published articles related directly to the

subject of nighttime highway construction.








12
In the following section, an overview of the factors attributable to

day or night shift work in highway construction is presented. The

literature review for each of the workshift selection factors is presented

in detail section by section in this chapter.

Overview of Factors Attributable to Day and Night Shift Work

Several published reports in the transportation area have provided

information on issues relating to the planning, safety, and traffic

control aspects of nighttime maintenance and construction work and their

advantages and disadvantages. Shepard and Cottrell (1985) conducted a

study that compiled information on current practices in nighttime highway

maintenance and construction operations. This information was used to

develop guidelines for determining when night work should be done and what

traffic control devices should be utilized. The authors have researched

the many variables, problems, decisions, and issues involved in nighttime

highway rehabilitation. Results from visits and discussions with 11 state

departments of transportation revealed the different strategies and

philosophies involved in nighttime operations. The major areas covered in

their nighttime construction feasibility study are scheduling of lane and

road closures, work zone costs, safety, public relations/user costs, and

traffic control. Hinze and Carlisle (1990) have done further evaluation

of the important variables in nighttime construction. The authors have

focused their research on rehabilitation and maintenance activities of

major metropolitan highways. Qualitative and quantitative factors related

to nighttime construction have been detailed in their study. Advantages

and disadvantages of a nighttime construction schedule are also discussed.








13
The following sections review information obtained on traffic

congestion, work-zone safety, public relations, noise, cost, quality

control, job type, human and sociological factors, and productivity. All

these are factors that need to be considered in the shift selection

process.


Traffic Volume

Typically, daytime traffic volume in metropolitan areas demands a

particular lane capacity to keep congestion low. Lane closures during

construction work are dictated by the comparison of observed traffic

volume and specified lane capacities. The determination of lane

capacities and actual traffic volume is necessary to estimate the level of

congestion created by lane or road closures. If this congestion is

unacceptably high during a certain time period, lane closures for

construction may be limited or not permitted at all, requiring agencies to

decide on alternate work shifts. Low volume periods should be identified

for scheduling work shifts. Night shift offers the advantage of low

traffic volume.

Shepard and Cottrell (1985) have discussed in some detail how

traffic volume and lane capacities are determined and how traffic flow can

be synchronized with road construction activities. Agencies rely on

mathematical models as well as past experience to make traffic capacity

forecasts. According to Dudek and Richards (1985), the State of Texas

expects a lane capacity of 1050 vehicles per hour (VPH) on a 3-lane urban

highway with 1 lane open, when resurfacing or removing asphalt. A study

of cumulative distributions and average capacities of similar work zones








14
allowed the Texas officials to make this capacity forecast. Based on

policies and experience, work location, job type, etc., some state

agencies have developed guidelines for lane capacities. Agency policy may

dictate that nighttime construction is necessary if the traffic volume

during the daytime hours for a certain location exceeds a "cutoff" value,

after which resulting delays are "unacceptable." Table 2.1 shows a

listing of cutoff values adopted by ten agencies surveyed by Shepard and

Cottrell. The Highway Capacity Manual (National Research Council, 1985)

provides one of the empirical formulas to estimate roadway capacity.

VPH (per lane) = Z f, fhv fp

where,

Z = capacity under ideal conditions at design speed.

f, = lane width / lateral clearance adjustment factor.

fh, = heavy vehicle adjustment factor.

fp = driver population factor.

The values for these adjustment factors may be obtained from tables

included in the manual.

To ensure good volume estimates, agencies supplement existing data
with special spot counts. In major cities such as Chicago, Los Angeles,

and Detroit, agencies have supplemented traffic volume data with real-time

traffic flow data from their freeway surveillance system. The estimation

of traffic volumes is also influenced by diversion. Motorists are often

induced to take alternate routes if prior publicity is made by the agency.

According to Shepard and Cottrell (1985), information from California

indicated that a traffic volume diversion up to 30% was possible for

local traffic when prior publicity on alternate routes was done.













Lane Capacities for Selected Metropolitan Areas


Area

Los Angeles



Atlanta


Chicago

Detroit


Raleigh

Philadelphia

Long Island

Dallas


Houston


VPH per Lane

1500-1800



1200-1500


1300-1500

1200-1500


1200-1600


1500

1300-1500


1200-1500


Comments

a) 1500--usually no congestion unless
more than one lane is closed.
b) Sometimes use 1800 (with some backups
to give contractor more time).

a) No daytime closure. Depending on the
area, no daytime closures if 2 or more
lanes need to be closed.

a) Depends on location, number of ramps.

a) 1200--volume before backup starts.
b) 1500--expect serious backups.

a) Depends on area and experience.

a) Closures based on experience.

a) If closure is 2 or more lanes, have to
detour traffic and work at night.

a) In many cases will accept daytime
backups rather than work at night.

a) >1200--start worrying.
b) =1500--requires detailed analysis.
c) >1500--only with special traffic control.


Source: Shepard and Cottrell (1985)


Table 2.1








16
A common method used by agencies to evaluate congestion is simply to
plot the hourly volumes for the work shift time period. Figure 2.1
indicates the volume distribution (curve) on a 4-lane highway during a
probable work period, along with estimated capacities (horizontal lines)
for 3 lane, 2 lane and 1 lane closures. From the graph it is apparent
that for 2 lane closure the traffic volume from 10:00 a.m. to 3:30 p.m.

and from 8:00 p.m. to 6;30 a.m. does not exceed the given capacity.
However, these work periods are very short in duration to allow major

repair activities. Statistics have shown that under normal daytime
operations, there are two peak traffic loads that actually cut the work
period to an inefficient 5-1/2 hour day (Abbott, 1978). Figure 2.1

indicates a lengthy work period during the night when 3 lanes are closed,

with only 1 lane required to remain open to handle traffic. By noting
when the traffic demand and the capacity are in the same range, it is
possible to determine the times at which the lanes can be closed and
reopened to traffic.

Vehicle delay costs is an important quantitative factor in the
workshift selection process. Highway construction literature has

indicated that nighttime construction greatly reduces user costs related

to vehicle delay. A research report by Price (1985) comparing two similar
paving projects (one nighttime, the other daytime) in Colorado, reveals a
reduction of vehicle cost from $119,100 (daytime) to $10,100 (nighttime).
The author obtained the dollar estimates from a 1977 report titled A
Manual on User Benefit Analysis of Highway and Bus Transit Improvement.
The values were estimated showing excess cost of vehicle speed change
cycles above cost of continuing at initial speed. For the daytime cost,













Volume VPH (thousands)


12 1 2 3 4 5 6 7 8 9101112 1 2 3 4 5 6 7 8 9101112
A.M.<--Time of Day-->P.M.

-Volume Curve 1-LANE CLOSURE
2-LANE CLOSURE --- 3-LANE CLOSURE
Source: Hinze & Carlisle (1990)
Figure 2.1 Traffic Flow on a 4-Lane Freeway








18

the author assumed an initial vehicle speed of 55 mph and a 10 minute

delay due to traffic stoppage, followed by resumption of speed. For the

daytime paving project analyzed by Price, a cost of $64.68 per 1000

vehicles was estimated. By using daytime traffic counts and considering

a project duration of 37 days, a cost of $119,100 was calculated (Price,

1985). For the nighttime paving project, traffic speed was assumed to

slow down to 30 mph, and the estimated vehicle cost was $10,100 for the

change in speed over the project duration.

Vehicle delay, length and size of vehicle queue can be calculated

from a mathematical model, as shown graphically in Figure 2.2 (Shepard &

Cottrell, 1985). The graph shows the arrival curve and the departure

curve indicating the upstream traffic volume and bottleneck capacity,

respectfully. At time t2 the number of vehicles in queue, Q, equals to the

corresponding ordinate on the arrival curve minus the corresponding

ordinate on the departure curve, the ordinate value being the total number

of vehicles. The queue begins to form when the arrival curve slope is

greater than the departure curve slope. The queue is at its maximum when

the two slopes are equal, and it starts to dissipate after that point

(Mannering & Kilareski, 1988). As shown in Figure 2.2, the longest

vehicle delay occurs at time t3. The length of this delay is given by the

horizontal line D, joining the two points where the two curves have equal

slopes. Agency planners can utilize the quantitative information from

this model to determine whether the traffic characteristics of a

particular workshift is acceptable or not. Figure 2.3 illustrates an

example usage of this model, based on a study done in California (Gillis,

1969).


















Bottleneck Capacity
DEPARTURE CURVE

V


Upstream Traf
ARRIVAL CU



No. of

Vehicles




Vehicles in
Queue /














St









Source: Shepard & Cottrell (1985)


fic Volume
RVE







Q

I/


Delay Time


t2 t3 t4


Time _


Figure 2.2 Traffic Volume / Capacity Relations















mu mum. I -,


10 AM Lane


No Delay A


Closure


AllM












9 AM L


Del


CAPACITY












4--CAPAC


witl


y (minutes)


Ei



ITY 3080


S2 lanes cl


mne Closure --Begin Delay


JI/ Cum. Vehicle a Using Rojad


S AM


(thouaandO)
Hour of Day









Source: Gillis (1969)

Figure 2.3 Effect of Lane Closure at Different Hours


Cum.


Vehicles


12 Noon
id Delay



vph


ised


1 PM








21
Vehicle delay in hours can also be measured by the following

equation noted by Hinze and Carlisle (1990):

Delay = D, / S, D/S

where,

D, = work zone route (detour) distance in miles.

S, = average work zone route (detour) speed in m.p.h.

D = normal route distance in miles.

S = average normal route speed in m.p.h.

This equation takes into consideration the fact that vehicle delay

can occur by speed changes through the work zone and also by distance

changes resulting from a detour. During highway construction work shift

selection, this type of detailed analysis allows planners to evaluate the

traffic impact objectively.


Work Zone Safety

One of the most important factors that state agencies consider when

deciding highway work shift periods is safety. The issue of safety during

nighttime construction has been addressed in several research reports.

These reports note the controversy over which work shift is safer than the

other. The obvious hazards during the night shift are reduced visibility,

higher speeds, and a higher frequency of drunk or inattentive drivers on

the roadway. But the night shift also offers a reduction of traffic

volume, which in turn results in safer working conditions. From reviewing

existing literature, it is difficult to compare the safety aspects of

night shift as opposed to those for day shift simply because of the lack

of comparable data.










Motorist Safety

The subject of highway work-zone accidents is covered by several

published reports. A report prepared by Graham, Paulsen, and Glennon

(1977) for the Federal Highway Administration includes results of a study

of construction zone traffic accidents. The study involved analysis of

traffic accidents occurring in 79 zones in seven states. In two states

the total number of accidents actually decreased during construction.

However, when data from all seven states were combined, results indicated

an overall "before work" to "during work" average accident rate increase

of 7.5%, with a corresponding 9.4% increase for night accidents (Graham et

al., 1977). Lum (1980) has shown these results in a tabular format (Table

2.2) in a paper published in the December issue of Public Roads.

Breakdowns by accident type, severity, light condition, roadway type, area

type, work area, job type and location are also included in the results.
The authors concluded that the number of traffic related night accidents

increased during construction, but the proportion of night accidents to

total accidents remained the same at 30% for before and during

construction.

Nemeth and Rathi (1983) published a work-zone accident report in the

Transportation Quarterly that contains accident data collected through a

28 month study by the Ohio Turnpike Commission. The report includes the

number and percent of total accidents in different portions of the work

zone that occurred during the night shift. Table 2.3 summarizes the

percent of night accidents in the work zone. The study concluded that the

crossover zone, where traffic is forced to shift over to other lanes, has

the highest accident rates.













Table 2.2 -


Number
States


of Accidents Before and During Construction in Seven


State and Before Construction During Construction % Change
Time


State 1:
Day
Night


State 2:
Day
Night


State 3:
Day
Night


State 4:
Day
Night


State 5:
Day
Night


State 6:
Day
Night


State 7:
Day
Night


All States
Day
Night


* = Statistically


971
561
1,532


N/A
N/A
1,023


1,075
744
1,819


708
271
979


1,540
659
2,199


340
183
523


61
32
93


4,695
2,454
7,149


892
592
1,484


733
280
1,013


1,216
778
1,994


723
353
1,076


1,726
729
2,455


439
196
635


91
37
128


5,087
2,685
7,772


-8.1
4.8
-3.1


N/A
N/A
-1.0


13.1*
4.6
9.6


2.1
30.3*
9.9


12.1*
10.6
11.6*


29.1*
7.1
21.4**


49.2*
15.6*
37.6*


8.3
9.4
8.7


L _______________ I


significant at the 5 % level


Source: Lum (1980)



















Table 2.3 Summary of Work Zone Accidents at Night
(Ohio Turnpike Commission)

Zone Total Number Number of Percent
of Accidents Accidents Night
at Night Accident

Advance 12 1 8.3
Taper 17 6 35.3
Single Lane 43 13 30.2
Crossover:
First Curve 49 34 69.4
Total 63 39 61.9
Bi-direction 41 18 43.9
Other Work 9 3
Zone Total 185 80 41.6
All Turnpike 3,429 1,431 41.7

Source: Nemeth and Rathi (1983)








25

A 1985 California accident study reveals that although the lightest

traffic congestion occurred between the hours of 11:00 p.m. and 6:00 a.m.,

the fatal accident rate (per 100 MVMT, million vehicle miles of travel)

during this period was very high. Figure 2.4 (Highway Maintenance

Activities During Low Volume Traffic Hours, Report to the Legislature,

1988) illustrates the California fatal accident rate expressed as a

function of hours of the day. The graph shows how the rate increases

sharply at night, peaking at about 3:00 a.m. A work zone vehicle accident

report by Hargroves and Martin (1980) indicates similar results in a 1977

study in Virginia, where for all accident categories the lowest number of

accidents occurred during the late night hours.

Shepard and Cottrell conducted interviews on the subject of safety

at night work zones with state agencies from Texas, California, Illinois,

Michigan, New York, and Pennsylvania. Texas is taking extra measures to

ensure safety during night operations; for example, in Houston a special

crew was formed to manage traffic on high volume freeways. Another urban

area in Texas is more ready to accept long daytime traffic backups rather

than accept the high risk of night construction. The California

Department of Transportation (DOT) is prepared to accept daytime vehicle

delay up to a certain cutoff point, beyond which night work is considered.

In the Los Angeles area, the hours between 2:00 a.m. and 3:00 a.m. are

considered unsafe for workers due to the presence of irresponsible drivers

on the road. Safety is given the most priority in Chicago when the agency

selects work shifts for road maintenance operations. Illinois DOT

officials reported that nighttime accidents were more severe because of

higher vehicular speeds and more frequent encounters with drunk drivers.









URBAN FREEWAY FATAL ACCIDENT RATES PER
100 MILL. VEH. MILES OF TRAVEL (1985)


FATAL ACCIDENTS PER 100 MVMT


121 2 3 4 5 6 7 8 91011121 2 3 4 5 6 7 8 9101112
A.M.(--TIME OF DAY --> P.M.


Accident Rate

Source: California DOT (1988)
Figure 2.4 California Urban Freeway Fatal Accident Rates








27
In Chicago, diversion of traffic volume by detours during highway work is

not considered very often because of the potential safety hazard in "bad"

neighborhoods. However, in Detroit, providing detours is not a problem,

due to the availability of convenient alternate routes such as express

lanes and service roads. As a result, Michigan DOT has been able to

close freeways entirely for nighttime maintenance, diverting traffic

through detours successfully. In New York State, agency officials realize

the importance of providing nighttime safety and the need for additional
spending on safety measures. Pennsylvania DOT, although initially

skeptical of the safety of nighttime operations, discovered that night

shift work could be less accident-prone if vehicular speeds are reduced by

the presence of a police car with flashing lights at the beginning of the

lane closure. Also, advance publicity through news media increased safety

for both day and night shift work.

Worker Safety

According to the California DOT report "Highway Maintenance
Activities During Low Volume Traffic Hours," nighttime construction

accounted for 60 percent of the injuries/fatalities sustained in highway

work zones. Table 2.4 shows the percentage of fatal accidents on urban

freeways by the time of day and day of week. A high percentage of fatal

accidents were noted during nightime and early morning hours. Most worker

fatalities (and major injuries) are related to errant drivers causing
these accidents. The California report concludes that the high fatal
accident rate during night hours creates a very hazardous environment for

the workers on site. The report cautions that a large scale move from day

maintenance to night maintenance could mean a significant increase in

worker injury and fatality rates.














Table 2.4 -


Fatal Accidents on Urban Freeways
and Day of Week (1986)


by Time of Day


Time No. of Percent Day of No. of Percent
of Day Accidents Week Accidents


Sunday *

Monday

Tuesday

Wednesday

Thursday

Friday

Saturday *

Total


15.8

12.2

10.8

13.8

12.1

14.4

20.9

100.0 %


12 mid
1 a.m.*
2 a.m.*
3 a.m.
4 a.m.
5 a.m.
6 a.m.
7 a.m.
8 a.m
9 a.m.
10 a.m.
11 a.m.
12 noon
1 p.m.
2 p.m.
3 p.m.
4 p.m.
5 p.m.
6 p.m.
7 p.m.
8 p.m
9 p.m.
10 p.m.*
11 p.m.*
Unknown

Total


24
40
46
21
21
22
11
9
12
10
18
12
22
18
23
23
17
25
27
27
26
27
45
31
7

564


4.3
7.1
8.2
3.7
3.7
3.9
2.0
1.6
2.1
1.8
3.2
2.1
3.9
3.2
4.1
4.1
3.0
4.4
4.8
4.8
4.6
4.8
8.0
5.5
1.2

100.0 %


* High Percentage of Freeway Fatal Accidents

Source: Report to the Legislature, California DOT (March 1988)








29
The most effective way to protect workers is, of course, to divert

vehicular traffic away from the work zones through detours provided by

city streets. Although this will cause some inconvenience to the

motorists (5 to 10 minutes added to their travel time), it is not a high

price to pay to eliminate daytime congestion caused by maintenance

activities. Another protection method is to construct a physical barrier

separating the workers from the traffic. Better illumination of work

areas, highly visible clothing for workers, reflective cones/barriers,

police presence, etc., are additional ways to increase worker safety.


Public Relations

Public relations is an important element in the workshift selection

process for highway construction. Public inconvenience during road

construction operations is a major concern for the state agency. Public

acceptance of a particular work shift (day or night) means it warrants

high consideration by the decision makers. Good public relations can

facilitate the success of an operation in terms of reduced congestion,

increased safety, and goodwill (Shepard and Cottrell, 1985). Shepard and

Cottrell discuss various means of informing the public about nighttime

operations. The size of the project, location, traffic volume,

experience, and so forth dictate how extensive the coverage should be.

Certainly, the amount of publicity for a night project should be more

extensive for safety reasons. Survey results indicate that personal

contact (special signs, door-to-door contact, letters) is thought to be

the most effective method for publicity. The city of Chicago has an

elaborate plan to provide the public with information on upcoming freeway








30
operations. Local radio, television, and newspapers are effectively

utilized to increase public awareness. Detroit has successfully used

changeable message signs, permanently installed along freeway sections, as

a method of informing the public of lane closures.


Noise Level

A high level of noise, resulting from operation of heavy equipment

during road construction activities, may adversely affect public opinion.

Thus, noise is one of the important qualitative factors attributable to

shift work. During the night shift, the public might be particularly

sensitive to construction noise. Little information on the subject of

noise levels during road construction was found in the literature review.

According to Shepard and Cottrell, many states have noise ordinances that

limit noise levels in urban areas. Agency officials who were interviewed

by the authors reported that if equipment such as a jackhammer is

necessary during work near a residential area, the night shift is avoided.

Another cause for concern for the agency is that diversion of traffic

during freeway construction may increase noise levels on the detour

routes.

A research study (Hinze and Carlisle, 1990) included a survey of

project planners of various state highway agencies that found that noise

is of high importance next to congestion and safety. However, the same

study also concluded that resident engineers of the agencies do not

consider noise an important factor compared to other factors when planning

partial or full closure of freeways during construction.








31
The primary noise sources in road construction and maintenance can

be identified as (1) diesel engines, powering drive trains and cutting

heads; (2) the vacuum/blower system; and (3) grinding heads in contact

with pavement. According results obtained by the Arizona DOT (Kay, 1985),

these sources combine to produce a noise in the range of 82 to 95 dB (A)

at 50 feet from construction. Typically, noise generated during concrete

pavement construction is higher than noise during flexible pavement

construction. Many highways in the metropolitan areas are of rigid

pavement, thus the potential for construction noise in those areas appear

to be significantly high. The study done by the Arizona DOT (Kay, 1985)

notes an upper limit of 86 dB (A) for an acceptable noise level at a

distance of 50 feet. Other states also supported this finding, verifying

the acceptability of the limit. By using a full complement of silencers

it was estimated that noise levels could be reduced to 80 dB. Noise

reduction is also possible by the insertion of a portable sound barrier

between the source and any sensitive receivers.

Kay discusses about the problems encountered in setting and

enforcing noise-level limits for nighttime construction operations. In

the state of Arizona, a noise-abatement incentive was established to

encourage potential bidders to silence noisy construction equipment,

specifically the grinding machines. This noise-abatement incentive plan

was considered to be an alternate approach to including a relevant

specification in the contract. In some occasions, contractors have

threatened not to bid if specifications on noise levels were maintained or

tightened. The objective of the incentive plan was to provide

compensation for the additional cost incurred by the contractor. The








32

Arizona DOT set the ceiling for the incentive amount at $ 50,000 -- where

as the actual amount turned out to be $ 12,684 (for the first contract

using this plan) to maintain an average of 82 dB (A) noise level

throughout the project.

Some of the important findings in the research done by the Arizona

DOT, as mentioned by Kay in his paper published in the 1985 issue of the

Transportation Research Record, are as follows:

1. Fast completion of project results in less noise complaints

from public.

2. Accessory equipment, such as jackhammers, have a noise impact due

to their higher frequencies. Retrofitting on machines can result

in significant dampening of noise.

3. For large scale projects, it may be too expensive to

retrofit major components to maintain an overall

construction noise level to 86 dB(A). In such cases, public

relations can play a major role.

4. Typically, high productive equipment make more noise than

low productive equipment. Therefore, strict noise

restrictions may cause decrease in productivity and longer

project duration. A "trade-off" between project duration

and acceptable noise level can be considered.

Cost

Like any decision making process, the selection of day shift over

night shift or vice versa is influenced by the economic result of the

alternatives. Literature review indicated a lack of cost information for

an effective comparison between daytime and nighttime construction. In








33
highway construction, the costs attributable to a particular workshift can

be categorized into the following: owner/construction costs, user costs,

accident costs, and pavement maintenance costs. The latter cost refers to

post construction maintenance costs which depends on quality control

during a shift.

Owner Cost

Owner costs are essentially the costs borne by the state highway
agency, resulting from the construction of the specified facility The

cost of contract (labor, material, equipment and contractor) and agency

costs (planning, evaluating, monitoring) are included in this category.

Construction costs can vary from shift to shift. Additional construction

costs that can be attributed to night work include: lighting, additional

traffic control, inspection, labor premiums, overtime, and increased

material costs. A recent survey of agencies and contractors helped

identify the cost difference by project element for night construction

(Hinze & Carlisle, 1990). The results are shown in Table 2.5.

Material costs may be higher during the night shift due to batch

plants charging higher rates. Material prices are increased because of

hiring night employees at the plant and also because of shift differential

and overtime. Price (1985) in his report entitled Niqhttime Paving,

investigated a cost comparison between two similar overlay projects in the

Denver area. One job was done during the day on 1-25 north of Denver, and

the other was done during the night on 1-70. Table 2.6 (Price, 1985)

shows the material cost percentage that the night (1-70) job was over that

of the compared day (1-25) job. Although asphalt prices are subject to

variation, this table gives an idea of expected material cost increases.























Table 2.5 Estimated Increase in Cost Differential of Night Work

Project Element Increase in Cost

Lighting 63 %
Traffic Control 28 %
Engineering Inspection 22 %
Labor (Shift Premiums) 18 %
Overtime (Agency Personnel) 16 %
Material 5 %
Total Contract Amount 9 %


Source: Hinze and Carlisle (1990)





















Table 2.6 1-70 Night Paving Project vs. 1-25 Day Paving
Percent Cost Differential
Cost Item 1-70 Night Paving

HBP (Patching) (H-Asphalt) / ton 23 % Higher
HBP (GR EX) (H-Asphalt) / ton 23 % Higher
EMUL ASPH (CSS-1 H) / gallon 42 % Higher
Flagging / hour 71 % Higher


Total Costs
Average / Item


= 159 % higher at night
= 40 % higher at night


Source: Price (1985)








36

The table shows that for the night project, the price for hot bituminous

pavement (HBP) (patching) (haul and asphalt) per ton was 23% higher than

that of the day project. The price for emulsified asphalt (CSS-1H) per

gallon was 42% higher. From a survey of various state highway agencies,

Hinze and Carlisle (1990) concluded that the average percent cost

differential for materials during nighttime construction is approximately

5% higher. Special materials required for night work may also contribute

to greater cost. Conditions of the night, such as cooler temperature may

require a higher cost specific material.

A cost comparison study of asphalt and base widening roadway work

for day shift versus night shift was done by R. G. Layfield, resident

engineer of the Florida Department of Transportation (FDOT). The study

indicated a 2-3% increase in material costs for asphalt roadway work

during the night (Layfield, 1988). The daytime unit price for limerock

base (4 inch) was $4.50 per square yards compared to its nighttime unit

price of $6.34 per square yards, indicating a 3% increase in night cost.

The nighttime unit cost for asphalt was $33.07 per ton compared to $31.50

per ton for its daytime unit cost, resulting in an increase of 2% in night

cost.

For scheduling reasons, special material (e.g. rapid curing

concrete) may be utilized (Fallin, 1990). On the other hand, if the

contractor owns the batch plant, material acquisition costs may be lower

during the night shift due to less traffic interference on the road. When

materials can be stockpiled during the day, the only additional cost may

be the cost to double handle the material. Thus, material costs may or

may not be a factor during night operations.








37
Labor cost differential is another item that may increase during the

night shift. Literature review did not reveal detailed information on

this area. According to Willis (1982), labor is considered to be one of

the premium costs for nighttime paving. It could be expected that some

incentive should be used to influence the performance of nighttime

overlay. For example, if the contractor works 8 hours per night, the

contractor's workforce could be paid for 9 hours. Willis states that

there should not be any across-the-board pay increase for nighttime work.

If the contractor works 6 nights a week, there will be overtime on the

sixth night. The contractor may be subjected to specific union rules for

its use of labor force during nighttime work. This possibility should be

thoroughly investigated before preparing labor cost estimates (Willis,

1982). Labor and inspection costs are additional cost items for night

construction. Shift premiums accounted for an increase of 18% in direct

labor costs, while overtime costs for agency personnel required an

additional 16% (Fallin, 1990). A Florida contractor agreed to pay $0.50

per hour extra for all of his personnel involved in nighttime operations

(Layfield, 1988). The shift differential can vary depending upon the

specific time period. According to the 1988 report, "Highway Maintenance

Activities During Low Traffic Hours", prepared by the California DOT,

there is a 25 cent per hour differential if four or more hours are worked

in the swing shift (6:00 p.m. to 12 midnight). The pay differential for

four or more hours worked in the graveyard shift (midnight to 6:00 a.m.)

is 35 cent per hour.

The survey done by Hinze & Carlisle indicates that artificial

lighting is one of the most significant items in terms of differential








38

cost between day and night construction. Table 2.5 shows the estimated

increase (63% for lighting) in the cost differential of night work. Since

the cost of artificial lighting is unique to nighttime work, it can be

regarded as a project specific cost, when rented or leased lighting

equipment are used on an "as needed" basis. To minimize setup costs and

maximize efficiency, equipment mounted lighting is recommended. For

contractors involved regularly in night work, it is best to purchase their

own lighting systems. Fallin (1990) mentions the potential use of laser

technology to reduce the need for artificial lighting. Although currently

this alternative is very expensive, it may be a viable option in the near

future. Lum (1980), in his report published in Public Roads, investigates

the economic feasibility of lighting an entire construction zone to reduce

night traffic accidents. Table 2.7 (Lum, 1980) shows the costs to

purchase, install, and remove light posts. The cost information was

obtained from a number of utility companies and municipalities, and hence

not representative of cost data nationwide. Lum concludes from his

analysis that the benefit/cost ratio for lighting a construction project

is, at all situations, less than 1.

Another major cost factor in night work zone construction is traffic

control. According to Shepard and Cottrell (1985), the cost involved in

traffic control can approach an upper limit of the cost of the permanent

construction work itself. For nighttime construction, there is an added

cost for traffic control, due to the need for additional signs in a low

visibility environment. Additional signs for night shift work may include

changeable message signs, arrow boards, warning signs, and channelizing

devices. Homburger (1989) has written an investigative report detailing

















Install, and Remove Light Posts


Item


1. Unit cost of purchasing and
installing wooden pole,
foundation and bracket.
...........$ 598/each

2. Unit cost of luminaire.
...........$ 188/each

3. Unit cost of purchasing and
installing wiring.
...........$ I/ lin. ft.

4. Unit cost of energy and
maintenance.
...........$ 0.50/day

5. Cost of poles, foundations,
brackets, & luminaires.
...........$ 41,658

6. Cost of wiring.
......... ..$ 5,280

7. Cost of energy and
maintenance.
...........$ 115

8. Cost of removing poles.
...........$ 6,625

9. Total cost of lighting 1 mile
of roadway for 230 days.
...........$ 53,700


Information Source / Assumptions


Mounting height of 40 ft.; poles
located on one side of roadway.



High pressure sodium, 400 watt.


Overhead as opposed to underground
wiring; costs about twice as much.


Includes energy & minor replacement
or repair of lamps.


100 ft. spacing for 1 mile and one
side only, for a total of 53 units.


1 mile of wiring.


For 230 days.



Estimated at $ 125 per pole


Sum of items 5-8. Figure rounded to
the nearest $ 100.


Source: Lum (1980)


~ ~I~I


Table 2.7 Costs to Purchase,








40
traffic control measures in construction zones with nighttime activities.

Apart from signs, other traffic control measures are emergency/enforcement

controls (police and ambulance), and deterrent controls (barricades,

vehicle-mounted barriers). Literature review did not specifically reveal

cost differentials attributable to traffic control during the night.

However, Price (1985) in his analysis of two (night and day) paving

projects in Colorado, mentions that the flagging per hour cost was 71%

higher for the I-70 night paving project.

User Cost

The cost incurred by travelling motorists due to ongoing

construction and maintenance work on the roadway, can be classified as
"user" costs. Vehicle operating costs, personal costs, and accident costs

are the primary components of this category. Literature review indicates

the availability of detailed information on the first two components, but

very little on accident costs. It is difficult to estimate a cost for the

difference in accidents for night operations versus day operations. There

is a lack of nighttime accident data and considerable variability in

accidents between the types and locations of night projects. Vehicle

delay costs investigated by Price have been discussed previously in the

traffic volume section of the literature review. The calculation of unit

user costs is presented in A Manual on User Benefit Analysis of Highway

and Bus Transit Improvements (AASHTO, 1977). Tables for operating costs

and time value are utilized to determine unit user costs. Traffic volume

and work-zone capacity are used to calculate the vehicle operating cost

and the queuing cost due to time delays. This methodology allows the

comparison of alternate work shifts to evaluate the work zone cost impact

and the optimal shift time to affect a lane closure.








41

A method which combines user operating and time costs has been

discussed by Lytton and McFarland (1975). Using workzone location as a

parameter, tables are provided to calculate the user costs. The

limitation of their cost model is that it is applicable only for an

overlay type of road work. Dudek and Richards have developed a computer

model, using data from 14 work zone sites in Texas. Hourly traffic

volumes, capacity, traffic control method, and other parameters can be

entered into the QUEWZ model to determine user costs with precision and

efficiency (Dudek and Richards, 1985).


Quality of Work

In some situations, the quality of work during the night shift may

be lower than that on the day shift. Poor visibility during the night may

result in less than perfect finishing of paving jobs. Worker morale,

unsafe working conditions, and quality control for materials are factors

that are shift dependent and these may affect construction quality.

According to Shepard and Cottrell, the key word is "acceptable quality",

when considering the quality of work during night shift. Even if quality

is suspect during night operations, the product may be accepted under the

specifications.

From the literature review it is not possible to conclude that

quality is always adversely affected during the night shift. Some

resident engineers of the Arizona DOT have found better quality control

during the night for certain projects requiring cooler temperatures for

material and equipment (Kay, 1985). One state official noted that quality

during the night was better because the night hours allowed total road

closure, which provided safer working conditions.








42
The comparison of day and night paving jobs in Colorado done by

Price indicates that the overall quality of the nighttime work was very

similar to that of the daytime project. Test results from the 1-70 night

paving job and from a similar daytime job on 1-25 are compared in Table

2.8 (Price, 1985). The results indicate that compaction on the night job

was as high as that of the day job, and in some cases better. Quality was

affected mostly in aesthetics, such as roller marks being more apparent on

the finished product. The Colorado study was an indication of the success

of night paving operations from a quality standpoint regarding smoothness.

However, pavement densities during nighttime work were more difficult to

attain, possibly due to cooler temperatures. Because of limited

visibility during the night, crack filling and rolling operations were

identified as being more challenging tasks. High levels of illumination

were recommended for these activity types. One other point mentioned by

Price in his Colorado study was the difficulty in detecting approaching

rain storms which could affect the pavement material quality.

The night shift provides cooler temperatures which permits easy

placement of cement concrete or paving in certain locations during summer

time. High temperatures during the day may cause increased evaporation

and faster rate of set (Hinze & Carlisle, 1990). During the placement of

26 miles of single lane concrete roadway in California, it was realized

that the quality of concrete placed at night was better. The cooler night

temperatures allowed the concrete to be placed closer to its ideal

temperature of 70 degrees Fahrenheit. This resulted in good strength for

concrete and minimized any surface evaporation (Fallin, 1990).




















Paving Quality Test Results from Field Samples


Test Night Paving (1-70) Day Paving (1-25)


No.of Samples = 62 No.of Samples = 62
Compaction Mean = 96.1 Mean = 95.6
Specified = 95.0 Specified = 95.0
STD.* = 0.810 STD.* = 0.735

Asphalt No.of Samples = 47 No.of Samples = 47
Content Mean = 5.49 Mean = 5.81
Specified = 5.0-6. Specified = 5.3-6.3
STD.* = 0.267 STD.* = 0.222

Field No.of Samples = 62 No.of Samples = 62
Specific Mean = 2.22 Mean = 2.23
Gravity Specified = 2.31 Specified = 2.33
STD.* = 0.019 STD.* = 0.018

Standard Deviation of Samples


Source : Price (1985)


Tabl e 2.8 -








44
A recent survey of state highway agencies revealed that general

results of paving operations conducted during the night shift produced

less quality results (Hinze & Carlisle, 1990). Both portland cement

concrete and asphalt had apparent finish defects when placed during the

night. However, the results were acceptable to the state agencies and

satisfied the specification standards. The quality rating (on a scale of

1 to 7) of certain tasks performed during the night, is shown in Table 2.9

(Hinze & Carlisle, 1990). Compaction of subgrade is the highest rated

night construction task and all the crack sealing related activities are

rated lowest.

Literature study indicated the following problem areas in paving

quality during the night shift :

roughness of paving surface

S inconsistency in the mix

S poor compaction

S cold joints

S inspection of work

S less control on tack spread

S alignment

S repairing roller marks

Quality of work is highly dependent on visibility. Almost all of
the above quality related problems result from inadequate lighting at the

project site. In the Colorado project, shadows were reported as a problem

in the rolling and crack filling operations. Proper illumination with

additional portable flood lights can reduce many of these defects.





















Table 2.9 Quality Rating of Nighttime Operations

Activity Performed At Night Quality Rating

Compaction of Subgrade 4.88
Surface Compaction of Asphalt Concrete 4.77
Installing Guard Rails 4.73
Placing PCC Pavement 4.54
Pavement Marking 4.36
Crack-Sealing on PCC Pavement 4.19
Crack-Sealing on AC Pavement 4.10


Source: Hinze and Carlisle (1990)










Job Types in Highway Construction

The type of highway work to be done often dictates the selection of

workshift hours. Highway construction and rehabilitation activities can

be classified into two types: fixed-location projects and moving projects

(Homburger, 1989). Fixed-location projects are more commonly found --

they involve work on highway and freeway structures, new paving which

proceeds relatively slowly (e.g. adding lanes, completely rebuilding

existing pavement), revising interchanges, etc. Moving projects are

typically pavement overlay projects which move relatively quickly from one

end of the job to the other.

Moving projects require special alertness from both motorist and

worker since the control conditions change quickly. If traffic control is

not a factor, then moving projects are undoubtedly preferred to be

undertaken during daytime hours. At a fixed-location site, the control

change can be programmed in advance and signs can be posted alerting

motorists to the changes ahead of time. Thus, fixed-location projects are

relatively safer and suited for night shifts.

The following fixed-location and moving job types were identified

from literature review as being more appropriate for the night shift

because of comfortable working temperatures, scheduling, and ease of

traffic control:

(1) Widening and replacement of concrete deck (using precast slabs) for
multi-lane divided bridge in a metropolitan location. Usage of

polymer concrete -- which sets to 4000 psi in one hour at

temperature range of 20 to 100 degrees F (Fahrenheit) -- can allow

the bridge to remain fully open during the day (Pasko, 1985).








47

(2) Interstate widening and reconstruction (e.g. from 4 lane to 6 lane)
through a metropolitan city.

(3) Concrete median barriers safety upgrading, pavement (concrete joint)

patching, shoulder work, on multi-lane divided urban highway, with

one or two lanes closed.

(4) Resurfacing operation on a multi-lane divided urban freeway, with
all lanes closed, and detour set up to follow parallel arterial

routes (Strakovits, 1974).

(5) Installation of raised pavement markers on freeways.

(6) Grooving of concrete and asphalt pavement to improve friction
factor.

(7) Placing of D-mix asphalt on an urban interstate highway. D-mix

placement requires minimum temperature of 60 degrees F (Hudson,

1991). During cold seasons this operation may need to be done

during the day shift.

(8) Pavement overlay on multi-lane freeway (approximately 10 mile

stretch) involving an asphalt overlay on existing portland cement

concrete.

Table 2.10 contains a list of highway projects undertaken by the

Florida DOT for night schedule at various counties as of March, 1991.

Approximately half of these projects involved resurfacing operations.

Table 2.11 shows a list of tasks which have the potential of better

performance during the night shift. These activities are shown in

descending order of preference (Hinze and Carlisle, 1990).

Shifting all maintenance work in a given area to a night schedule

may not be considered practical. Certain types of roadway repair

activities which are temperature sensitive such as asphalt paving,
















Table 2.10 : Florida DOT Projects Using Night Work

Type of Work Road County


Resurfacing
Resurfacing
Repair 18 Bridges
Skid Hazard Resurfacing
Resurfacing
Add Turn Lanes
Milling and Resurfacing
Milling and Resurfacing
Resurfacing
Paving Shoulders and Resurfacing
Safety
Resurfacing
Intersection (Minor)
Resurfacing
Widening and Resurfacing
Multi-lane Reconstruction
Resurfacing
Add Lanes and Reconstruct
Multi-lane Reconstruction
Construct Grade Separation
Add Lanes and Resurface
Add Lanes and Resurface
Replace Low Level Bridge
Intersection (Major)
Add Lanes and Resurface
Multi-lane Reconstruction
Multi-lane Reconstruction
Bridge Rehabilitation
Resurfacing
Widen Bridge
Skid Hazard Resurfacing
Interchange (Major)
Multi-lane Reconstruction


US-90
I-4
1-75
US-90
SR-21
US-90
SR-60
SR-688
1-95
US-441
US-90
1-95
US-90
I-95
US-301
1-95
1-95
US-41
1-95
1-4
Turnpk
US-19
SR-934
US-92
I-95
US-98
US-98
SR-A1A
SR-933
I-95
SR-580
Turnpk
US-319


Duval
Orange
Marion
Duval
Clay
Duval
Hillsborough
Pinellas
Duval
Alachua
Duval
St. Johns
Leon
St. Johns
Hillsborough
Nassau
St. Johns
Dade
Duval
Seminole
Dade
Pasco
Dade
Hillsborough
Dade
Bay
Bay
St. Johns
Dade
Dade
Hillsborough
Broward
Leon


Source : F.D.O.T. (1991)


1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.



















Table 2.11 : Construction Activities Performed Better At Night

No. of Respondents
Construction Activity Naming The Activity

Asphalt Concrete Paving 6
Bridge Deck Rehabilitation 5
Pavement Marking 5
PCC Paving 4
Demolition 3
Setting Girders 2
Concrete Joint Repairs 2
PCC Pavement Spall Repairs 2
PCC Pavement Slab Repair 2
Latex Modified Concrete 2
Grooving PCC Concrete 1
Large Concrete Pours 1
Asphalt Concrete Removal 1
Loading and Hauling Dirt 1
Preparing for Structural Pours 1
Backfilling Structures 1
Shifting Traffic 1
Installing Guard Rails 1
Signals 1
Street Lighting 1
Grading/Crushing Aggregate 1


Source : Hinze and Carlisle (1990)








50
chip seals, and slurry seals may still need to be done by day shift crews.

Other activities, such as landscape work, can continue to be performed

during the day without causing traffic congestion. And certainly, daytime

emergency response to repair needs must continue. Work activities on the

pavement and shoulders are the obvious candidates for night schedule.


Human Factors

Worker behavior during a particular work shift is an important

consideration for project management. The human factor has a high

potential to adversely affect both productivity and safety. Worker

performance is significantly affected by the choice of work shift.

Several studies on human performance (related to shift work), particularly

in the manufacturing industry, were found during the literature search.

Based on this review, three principal factors were identified as having

the most impact on worker performance: 1) sleep; 2) human circadian

rhythms; and 3) social/domestic issues.

Monk uses a model for coping with shift work to illustrate the

relationship between the three human factors. The worker's internal

system and the condition of his external surroundings dictate his ability

to sleep. The human biological clock depends on the degree of

fragmentation of sleep and the daily social demands. Finally, the

worker's social harmony is dependent upon the degree of sleep deprivation

(Monk, 1989). Monk stresses the importance of understanding the

relationship between sleep, the worker's biological clock, and the

domestic aspects of the worker's life. Figure 2.5 (Monk, 1989) presents

a triad of factors for coping with shift work. The three interrelated





























Shift

Work

Coping


Social/
Domestic
Factors


Source: Monk (1989)

Figure 2.5 Triad of Shift Work Coping Factors


Clock

Factors


Sleep

Factors








52
factors determine the individual's ability to cope with shift work. All

three components complement each other, a failure in one can negate

advances made in the other two. During an atypical working environment,

such as nighttime construction, the project management should try to

maintain a proper balance of these factors among the workers.

Night shift workers have faced the problem of insufficient sleep

during the day because of the difficulty in physiological adjustments and

also because of the heat, light, and noise distruptions of daytime hours

(Delisle, 1990). Lack of adequate sleep can lead to physical and mental

disorders, which can lead to on-the-job accidents (Finn, 1981). A recent

survey has indicated that over 20% of night shift workers in the U.S.

suffer from insomnia, fatigue, digestive disorders, loss of alertness, and

other sleep-related problems. From the point of view of safety and

productivity, it is necessary for the project management to identify those

individuals who are more adaptable to the stresses of night shift work.

According to a study by Vidacek (1986), night productivity rises with

successive night shifts, but eventually declines, while afternoon shift

productivity remains high and consistent. In construction, safety is

dependent on the work habits of the crew. Rotating shift workers often

resort to usage of drugs and alcohol to overcome their sleep deprivation.

Compared to other industries, construction has produced higher accident

results, and thus, the need to have a well rested and properly adjusted

night worker cannot be over-emphasized.

Research indicates that most workers placed on a night shift,

eventually becomes adjusted, as long as the same pattern is followed every

day. Delisle (1990) has examined two basic issues regarding a rotating








53
shift system: the length of time that workers should be kept on a

particular shift before rotating, and, the steps that management should

follow when implementing a shift schedule to meet their needs. Opinions

are divided between rapid and slow rotating shift schedules. Rapid

rotation means that a person works two or three nights in a row before

going back to standard hours, while slow rotation involves working many

more shifts (20 or greater, with intervening rest days). Scheduling the

shift rotations should depend on the type of work to be performed. A

rapidly rotating shift is best for tasks requiring complex cognitive

performance, involving a high short-term memory load (Delisle, 1990). For

highway construction, however, slow rotating schedule is more appropriate,

since it involves the performance of more simple, non-memory type tasks.

This allows the worker's circadian rhythm (biological clock) adequate time

to adjust to different shift periods, and, as a result, increase work

performance. Delisle emphasizes that, in order to successfully implement

a slow rotating schedule, the project management should encourage the

workers to try to keep on a reversed schedule on their days off. A

complete internal readjustment, at the start of each new work week, will

present added stresses on the human body and negate the benefits of a slow

rotating scheduling system.

Management should consider the type of specialization workers are
trained for, when assigning them to certain shifts. Members of an asphalt

crew, who are accustomed to lengthy work periods, would probably be more

adaptable to atypical schedules such as night shifts. Also the night

shift provides a more comfortable temperature for the crew placing hot

asphalt.








54
Additional guidelines for management regarding rotating shifts are

outlined by Benjamin (1984). Night schedules should be planned in

advance, so the workers can prepare themselves adequately. The shift

workers require uninterrupted weekends off as often as possible, for

social reasons. Shift duration should be limited according to the type of

tasks performed. It is recommended that tasks involving intensive

physical labor be limited to 9-hour shifts. Management should avoid

scheduling peak work loads on the job site between 3 a.m. and 6 a.m for

safety reasons. Construction scheduling should be such that the critical

activities occur at the start of the night shift, when workers are at

their peak mental alertness.


Productivity

The unique aspects of night construction can have both negative and
positive effect on productivity. Productivity during a particular work

shift is impacted by several factors, such as traffic volume, job type,

material delivery, lighting, supervision, communication, worker morale,

etc. During typical daytime construction operations there are two peak

traffic loads that actually reduce a work day to a 5 1/2 hour work shift.

Whereas, during the night, the work shift and actual daily working hours

are extended (Layfield, 1988). Availability and supply of material,

equipment spare parts at night also has an effect on productivity.

Artificial lighting, which may vary with the type of project, has a

potential impact on the output of night construction workforce. Certain

human factors, such as the biological clock, also govern the crew

productivity during the night shift. Productivity also depends on the job








55
type. Projects allowing total road closure may save more than 50% of the

construction time. A state transportation official notes that for a job

requiring large number of road patches, daytime patching (closing one lane

at a time) would take much longer than nighttime patching, when total road

closure is permitted (Shepard & Cottrell, 1985).

In a recent ENR article "Barriers Remain to Safe Work Zones" a

contractor in Tampa, Florida stated that "productivity is 28 to 30% less

at night due to set up and take-down time" (Stussman, 1988). On the

positive side, night shift was reported to have allowed a certain project

in Long Beach, California to be completed in 16 working days, whereas it

would have taken at least 35 working days to complete the same project

during the day shift (Lee, 1969). During a Florida road project, asphalt

was laid down during the night shift at a rate of 147.03 tons per hour,

compared to 98.09 tons per hour for another daytime project (Layfield,

1988). In his feature ENR article "Paving After Dark Turns Profitable",

McConville stated that in a Indiana road project, the contractor reported

10% improvement in speed of hauling cycles and tonnage during night shift.

Another contractor in a Pennsylvania Interstate project reported that

nighttime production increases have been logged as high as 30%

(McConville, 1991). Cooler temperatures and less traffic interference

during the night shift allowed for this increased productivity.


Other Factors

When deciding work shift periods, other miscellaneous factors may

also need to be considered by the state agency. Since each construction

or maintenance task differs in some respect from others, these "other"








56
factors vary in importance. The areas covered in this section are

supervision/communication, labor unions, parts availability, and liability

(Shepard & Cottrell, 1985).

The ability to supervise and communicate effectively during night

shifts depends largely on the agency's experience. Communication between

on site personnel and higher authority is vital during late shifts. A

person capable of making decisions should either be present at the site or

be "on call", responding as needed. Another problem mentioned by Shepard

and Cottrel is the difficulty in communication after the transition

between night and day personnel. The night engineer-in-charge may be fast

asleep during the daylight hours when project management may require his

feedback.

Based on information obtained from literature review, labor unions

do not present big problems for night operations. However, if there are

any differences, they should be negotiated and any stipulations should

reflect in the contract bid price.

Availability of spare parts for construction equipment is necessary

in order to maintain steady production at the site. During the night

shift it may be difficult to obtain spare mechanical parts because of

equipment dealers being closed after normal business hours. Shepard and

Cottrell suggest that the highway agencies have extra parts and equipment

available at the project site to ensure continuity of work.

Work-zone safety is a matter of concern during both day and night

shifts. The possibility of lawsuits, resulting from accidents due to road

construction, is an important consideration for agency planners.

Selecting the "safe" working period for both the worker and motorist

reduces the owner's liability.










Summary

The number of available references relating directly to highway

nighttime construction, as a whole, are limited. Only a few studies

provide a comprehensive approach towards night shift operations and the

agency shift selection process. Most of the literature related to one

specific aspect of night work, such as traffic control, safety, lighting,

human factors, etc. Numerous research studies pertaining to shift work

have been conducted in the manufacturing industry, only a few are

applicable to the construction field. Several published reports in the

highway construction area have provided information on issues relating to

the planning, safety, and traffic maintenance aspects of nighttime

operations.

It can be concluded from the literature review that the work shift

selection by the agency is mostly based on personal experience of traffic

data and congestion analysis. There is no formal step-by-step methodology

that takes all the factors into consideration in order to arrive at a

particular decision. Most of the information obtained from the literature

study is based on opinions, and is not quantitative. Information on shift

work for the manufacturing sector is readily available, but other than the

data on human factors, it cannot be related to highway construction.

Research has been done in specific areas of highway nighttime construction

which can be used as a guideline for state transportation agencies.

However, a more comprehensive study of all the factors involved in day and

night operations needs to be conducted.














CHAPTER 3
KNOWLEDGE-BASED EXPERT SYSTEM TECHNOLOGY

Introduction

Knowledge-based expert system technology is a branch of artificial

intelligence (AI) that has seen rapid advancement in recent years. An

expert system is essentially a computer program using AI techniques to

assist users in solving complex problems involving knowledge, heuristics

(rules of thumb), and decision-making. The knowledge-based system

approach has received broad attention in construction engineering

literature. In a typical construction engineering environment there are

decision problems that simply cannot be solved by procedural, algorithmic

computer models. In construction, knowledge and experience are used more

often than complex mathematical formulas, making this field ideal for

expert system application.

A knowledge-based system has the ability to provide explanations of

its reasoning making it useful as a management decision making tool.

The objective of this approach is to create intelligent behavior on the

computer through stored knowledge acquired from human experts. The

knowledge-based system or expert system is an "interactive" program,

playing the role of a human expert by utilizing heuristic knowledge.

Heuristics allows the system to make educated guesses, recognize promising

approaches, and narrow down the search process in a solution space.








59

The goal during the development of an expert system is to capture

specialized knowledge. This knowledge must be within a narrow, well

defined domain, and it should simulate the expert's reasoning process to

provide consultation about a difficult task.

An important difference between an expert system (ES) and a

conventional program is the representation of knowledge or data.

Knowledge in an expert system is usually divided into separate entities or

rules, shielded from the application methodology. Conclusions are reached

from the knowledge-base by invoking inference reasoning techniques. In a

conventional program, data is typically stored in a data base and the data

is manipulated by usage of algorithms, giving numerical results. Table

3.1 shows the difference between expert systems and conventional programs.

Architecture of an Expert System

The typical expert system (ES) structure consists of four primary

components: a knowledge-base, an inference mechanism, a working memory,

and an input/output interface with explanation and help facility (Adeli,

1988). In addition, features to facilitate knowledge acquisition,

debugging, editing, and intelligent interfacing may also be included. The

structure is schematically shown in Figure 3.1.

The knowledge-base is a repository of information available in a

particular domain. It consists of well-established and documented

definitions, facts, rules, as well as heuristics or judgmental information

associated with the problem domain. Knowledge acquisition is the process

by which expert knowledge is obtained from various sources for the

representation in a knowledge-base. The structuring and development of

the knowledge-base is aided by the knowledge acquisition facility.

















Table 3.1 Difference Between Expert System and Conventional Program


Expert System


L. Representation and use of
knowledge

. Knowledge-base and control
strategy integrated

1. Inferential (heuristic)
process

i. Effective manipulation of
large knowledge-bases

i. Developed by knowledge
engineer with or without
programming expertise

i. Midrun explanation desirable
and possible

'. Modifications, additions
relatively easy

1. Oriented toward symbolic
processing


Conventional Program


Representation and use of data


Data and control
strategy separated

Repetitive (algorithmic)
process

Effective manipulation of
databases

Developed by programmer


Midrun explanation not
possible

Not as flexible to changes


Oriented toward numerical
processing


Source: Expert Systems for Civil Engineers, ASCE (1987)





































KNOWLEDGE BASE


Rules



Facts


INFERENCE

MECHANISM


Forward chaining

Backward chaining


Figure 3.1 The Expert System Architecture








62
The inference mechanism (or reasoning mechanism) controls the

reasoning strategy of the ES by attempting to match the input data with

the information available in the knowledge base, and subsequently draw

conclusions and produce explanations. In a rule-based ES, the inference

mechanism determines the order in which the rules should be fired, and

resolves any conflict among rules when several rules are satisfied. The

mechanism seeks to solve the problem by chaining the rules together, and

eventually provide a conclusion.

Working memory (or context database) is a temporary storage of

information pertaining to the state of the specific problem currently

being solved. It is a flexible database, its contents changing

dynamically. Included in the working memory is the problem information

provided by the user as well as data derived by the system. The context

database contains all the intermediate results of the problem solving

process as well as the solution upon completion of the ES processing.

The user interface provides a link between the user and the expert

system. The user may create or modify a knowledge-base through the

interface by using a editor facility. Access to the knowledge-base for

information utilization is governed by the user interface. The

explanation facility and the help facility are both attached with the

input-output interface. The former provides answers to questions and

justifies answers. The latter guides the user to use the system

effectively and easily. The intelligent interface is a feature that

allows the user to interact with the ES and query the ES. This may

include natural language processors, menus, multiple windows, icons or

graphics.










Knowledge Representation

The development of the knowledge-base is an important step in the

creation of an expert system. The developer must decide by which method

the knowledge is to be represented in the knowledge base. Procedural and

declarative representations are two different ways to represent knowledge

(Adeli, 1988). In procedural representation, the knowledge is context

dependent, unintelligible and difficult to modify--it is commonly used in

traditional algorithmic programming. Declarative representation permits

knowledge to be context independent, more understandable, and easily

accessible for modifications. For these reasons, expert systems usually

use declarative knowledge representation. The three most widely used

declarative knowledge representation approaches used in current expert

systems are rule-based, frame-based, and logic-based. Other schemes

include semantic networks, and object-oriented methods. The choice of

representation will depend on the type of problem to be solved and the

inference methods available.

The frame system is a network data structure that represents the

relations between concepts, objects, or events, and their attributes. A

frame consists of a number of attributes, called slots, in which different

characteristics of an object or a piece of information are described.

Slots may contain default values, pointers to other frames, or procedures.

A procedure consists of a set of instructions for determining the value of

the slot, which is known as procedural attachment. The frame structure

has advantages in representing sequences of events, and for knowledge

acquisition and modification (Adeli, 1988). An example of a frame

representation is shown in Figure 3.2, where the object represented is:



















SSlots Entries

Location Alachua County
Highway 1-75 North
Number of Lanes 2
Project Length 5 miles
Project Duration 250 days


Material Default: Asphalt Conc.


Traffic Analysis : If needed, determine
roadway capacity and
compare with traffic
count database


Figure 3.2 A Frame Representation of Knowledge








65
"resurfacing project". The properties (location, highway name, number of

lanes, project length, duration, material used, & traffic analysis) of

this object are contained in the slots. The material slot allows for a

default value. Default values are typically used when representing

knowledge in domains where exceptions are rare. The last slot (traffic

analysis) in the frame in Figure 3.2 illustrate a procedural attachment,

in which instructions for determining an entry are contained.

For representation of concepts, objects, events, etc., semantic

networks can be used. The network consists of a collection of nodes and

connecting links. Like frames, this system also has flexibility for

modification, allowing addition of new nodes and links. In this type of

representation, each node can "inherit" the characteristics of its

connected nodes. Figure 3.3 illustrates an example of a semantic network.

For example the statements "Resurfacing is_a Highway Construction

Activity" and "Vehicle Delay caused_by Lane Closures" can be represented

in a simple semantic net by using the is_a or caused_by relations.

Knowledge can also be represented by logic. The two most common

forms are: propositional logic and predicate calculus (Harmon and King,

1985). Each basic element of the proposition can be either true or false.

Propositions can be connected to each other by the connectives AND, OR,

NOT, EQUIVALENT, IMPLIES, etc. This type of logic is concerned with the

truthfulness of compound sentences. For example, if proposition A is true

and proposition B is false, then "A AND B" is false, but "A OR B" is true.

Predicate calculus is a special subset or extension of propositional logic

in which propositions can contain variables. In this scheme, the

knowledge is represented through a programming language (e.g. PROLOG) to

















is a


caused by


66





HIGHWAY
CONSTRUCTION
ACTIVITY
is a




) (RESURFACING




caused by




LANE CLOSURES




caused by




(VEHICLE DELAY


Figure 3.3 Knowledge Representation Using Semantic Network








67
describe the facts and relationships in the problem domain (Chen, 1987).

In PROLOG, a fact statement such as "Highway rehabilitation/construction

activity is a state agency responsibility" can be expressed as:

"is a (rehabilitation/construction activity, state agency
responsibility)".
This is an example of a predication that is presented by a predicate name

(i.e., "is a") followed by a list of arguments. A set of clauses

represents this rule.

The following form depicts each clause:

consequent: [antecedent-1, antecedent-2,...antecedent-n

Logic-based knowledge representation is shown in the example below:

is a (resurfacing, state agency task) :
[ is a (resurfacing, transit construction activity),
is a (transit construction activity, state agency task) ].

The antecedents and consequent in each clause are predictions. The

consequent is true if the antecedents are true.

The rule-based (or production rules) system has been the most

popular representation approach for developing expert systems. It is the

chosen scheme for the knowledge base structure of the ES decision model

described in this thesis. The knowledge base is a collection of rules

which consist of IF-(antecedent)-THEN-(consequent) statements. The

general form for this type of representation is as follows (Adeli, 1988):

Rule N

IF [(antecedent 1) ......................(antecedent n)]

THEN [(consequent 1 with certainty ci)...................

(consequent m with certainty cm)

( c = certainty factor )








68
Each rule is unique by its rule number. The order of rule application is

not specified by the rule number. A rule represents an independent piece

of knowledge. The antecedent can be regarded as a pattern and the

consequent as a conclusion reached or action to be taken. The antecedent

part of the rule is linked to the working memory of the ES. The rule is

fired when all the conditions of the antecedent part are satisfied. Since

the antecedent-consequent or IF-THEN rules can easily be transformed into

questions, this type of representation can facilitate the generation of

explanations. Certainty or confidence factors can be attached to rules in

a rule-based system. Each rule may be assigned a certainty factor

typically in the range of 0 to 10 or 0 to 100. These factors simply

indicate the level of confidence in a piece of information.

The following is an example of a production rule:

IF: Roadway type is 4-lane freeway
and location is urban
and construction activity type is resurfacing
and number of required lane closures >2
and expected day traffic VPH per lane > 1500

THEN: Work should be performed during 7 pm to 5 am.

The consequent part is executed when the condition part provides a match

with the available facts in the working memory.

The rule-based knowledge representation allows the ES to provide

explanations to its conclusions rather easily. This scheme also permits

a natural way for the developer to describe complex knowledge. However,

one drawback of the rule-based approach is that the addition of new rules

or modification of existing rules may introduce contradictions.










Inference Engine

The inference engine or mechanism is the heart of an expert system.

It is a built-in reasoning process that determines which rules are to be

fired to reach a conclusion. The three common search techniques used in

an inference mechanism are: forward chaining, backward chaining, and the

hybrid approach. The selection of a particular search strategy depends on

the application area. The details of these techniques will be discussed

in this section.

In the forward chaining technique, the rules are scanned until one

is found whose antecedents (IF-parts) match the information entered in the

working memory. The rule is then fired, updating the working memory. The

process is repeated until a goal state is reached. This search strategy

is recommended when the goal state is unknown and has to be constructed or

the number of possible outcomes is large. Complex planning problems,

particularly in construction management, are well suited for this

application method.

Backward-chaining is a goal-driven strategy, in which the rules are

scanned for those whose consequent (THEN-parts) actions lead to the goal

state. These rules are then checked to determine whether their

antecedents (IF-parts) match the information in the working memory. When

a match is obtained, the rule is applied and the solution is reached. For

an unmatched antecedent, a new subgoal is defined as "arrange conditions

to match that antecedent" (Adeli, 1988). The process is applied

recursively. This strategy is particularly efficient when the values of

the goal state are known and its number of possible outcomes is small.

Backward chaining strategy is appropriate for diagnostic expert systems.








70
The hybrid approach combines both forward-chaining and backward-

chaining to yield conclusions. The "blackboard" environment, as described

by Harmon and King (1985), utilizes this combined approach. The

blackboard model is essentially a central global database maintaining a

two-way communication with independent rule groups (knowledge sources).

An agenda-based control system continually examines all of the possible

pending actions and chooses the one to try next. Processing in the

blackboard model is based concept of independent cooperating experts.

This type of model is appropriate for structuring complex, problem-solving

tasks that require multiple experts.


Knowledge Acquisition

Knowledge acquisition is the process by which the ES developer

collects knowledge from expert sources. Personal contact with domain

experts, reviewing literature documenting the experiences of the domain

experts, are ways to acquire the required knowledge. This process may be

divided into five steps, as described by Hayes-Roth (1983).

The initial step is identification and characterizing the important

aspects of the problem: the problem domain, knowledge sources, and goals.

The next step is conceptualization, during which the key properties and

relations are made explicit. Knowledge representation ideas and tools are

considered at this stage. The formalization process is the third step.

The model of the task is mapped, its key properties and relations are

expressed into some representation schemes. The fourth step is

implementation, which involves mapping the knowledge into the

representation method associated with the tool (ES shell) chosen for the








71
model. The testing and revision of the prototype system is the final

step. The knowledge base may be refined, or the knowledge representation

scheme may be redesigned, depending on the results of the sample run.

Personal interview with the domain expert is a direct and efficient

way to extract information from the domain expert. The interview can

be in the form of questions and answers and example problem-solving

sessions. This process entirely depends on the amount of time and effort

the expert is willing to spend. Also, the domain expert may not know how

to describe the decision process, or may simply misunderstand the

questions. To minimize this difficulty, it is best to narrow the focus of

the problem domain.

Protocol analysis is another way to obtain knowledge (Hart, 1985).

Instead of making direct contact with the domain expert, the knowledge

engineer allows the domain expert to perform and record problem solving

procedures. The recorded transcript is then analyzed by the knowledge

engineer with assistance from the domain expert. The advantage of this

technique is that it gives the domain expert more freedom to express his

knowledge. However, in this process, a communication gap between the

developer and the expert will most likely exist.

Automated knowledge acquisition methods, such as the induction

method, may reduce the gaps between the domain expert and the ES developer

(Chen, 1987). This process uses algorithms to obtain the knowledge. The

system typically uses a spreadsheet interface to get examples (describing

a particular problem) in tabular format from the knowledge-base builder.

In this method, the expert provides a set of examples of different types

of decisions and the corresponding attributes influencing the decisions.








72
Rules are induced by the algorithm using the examples. By this method, it

is not necessary for the domain expert to detail the decision making

process himself.


Expert System Development Shells

Expert system programming environments or shells are recent

commercial developments, intended to facilitate the building of knowledge-

based expert systems. Programming languages, system-building aids can

also be used as ES development tools. Languages such as LISP, PROLOG, C,

PASCAL, or FORTRAN are typically used in AI applications. The ES

developer is expected to be proficient in computer programming in this

situation. Although LISP and PROLOG have been the most chosen languages

in ES development, currently the usage of C language is gaining popularity

since it reduces operational times for the ES (Barber, 1987). ES-building

aids have not seen much usage to date. Usually, these are knowledge

acquisition systems geared toward assisting the developer in obtaining and

structuring knowledge. Examples of existing knowledge acquisition systems

are ETS, TIMM, and RULEMASTER (Adeli, 1988).

The shell is a commercial ES development tool which is intended to

provide an opportunity for non-programmers to build prototype expert

systems. With shells, the developer can concentrate on the knowledge

representation and not be concerned about mastering a programming

language, or deal with complex inference strategies. The ES shell

consists of a domain-independent inference engine, an empty knowledge

base, and an user-friendly interface. Shells contain specific

representation methods and inference mechanisms, thus they are less








73

flexible than an AI language such as LISP or PROLOG. Adeli has provided

a detailed evaluation of a variety of shells (both personal computer-based

and main frame-based) currently in use.

ES shells are usually developed by taking out the original knowledge

content from domain-dependent expert systems. As a result, these shells

are suitable only for problems similar to that of the original ES. The

developer must bear this in mind when he or she selects a shell for a

particular application. For the decision problem described in this

dissertation, a PC-based ES shell, EXSYS-Professional, was selected as the

frame for the prototype model development. The software is available at

the University of Florida, Department of Civil Engineering.

EXSYS Professional is a generalized expert system development

package. The shell developed by EXSYS Inc., is written in C programming

language. The principal development tool in Professional is the Rule

Editor. The Rule Editor can be utilized to construct simple rule-based

systems or complex modular blackboard systems. The editor provides menus,

prompts and help -- eliminating the need to memorize complex rule syntax.

All input is in the form of normal English text or algebraic expression.

A command language can be used to control the rule execution, allowing the

developer increased flexibility and control for complex applications.

Hypertext, rule compiler, blackboarding, agenda managers, table lookup,

and interface with external databases (Lotus or dBase) are some of the

useful features of the EXSYS shell. EXSYS programs are able to respond to

an end user's query with a full explanation of the logic used to arrive at

a particular conclusion. The knowledge base of this shell can accommodate

upto 2500 rules in a personal computer. Programs written with EXSYS are

directly compatible between the IBM PC/XT/AT, VAX/VMS and UNIX computers.










Current Applications in Construction Engineering

Several successful ES prototypes have been developed in the area of

civil/construction engineering in recent years. Experts systems are most

effective in non-theoretical applications where judgment and experience

plays an important role and in which there are no single solutions. Some

of the areas which are appropriate for ES applications in construction

are: 1) diagnosis, 2) fault detection, 3) prediction, 4) interpretation,

5) monitoring, 6) instruction, 7) planning, and 8) design.

Levitt (1986) has developed an expert system, "Howsafe", that

evaluates the safety-related aspect of a construction contractor's

organization and operating procedures. A personal computer-based ES

shell, Deciding Factor, was used for this system. The knowledge base

contains documented results of construction safety studies obtained from

various technical reports and journals. By using Howsafe, a construction

manager can evaluate a project's or company's safety practices.

An expert system for the selection of materials handling equipment

for construction of concrete frame buildings is described by Wijesundera

and Harris (1985). The system suggests suitable categories of equipment

for materials handling based on factors such as ground type, soil

condition, structural features, site accessibility, etc. The ES utilizes

an external database containing supplemental equipment information, which

is able to provide more specific recommendations. The ES shell SAVOIR has

been used to develop this system.

The area of construction project monitoring has received some

attention by ES developers. McGartland and Hendrickson (1985) have done

research on cost/time control, and purchasing/inventory control. An








75
expert system would analyze and verify weekly input to a database of

activity schedules and estimates, recognize cost overruns, time slippage

problems, and diagnose probable causes and offer solutions such as

activity duration/cost adjustment. For purchasing and inventory control,

an ES would be able to minimize the overall materials cost and assist a

project manager to determine the most economical inventory levels. A

forward chaining inference mechanism has been suggested by McGartland and

Hendrickson for this type of application.

Stone & Webster have developed expert systems to solve a variety of

problems in the area of welded construction (Hathaway and Finn, 1986).

A PC-based ES available to field engineers on site allows the appropriate

selection of welding procedures, reducing potential construction delays

and problems. Other ES applications include welder qualification test

selection, weld estimating, and weld defect diagnosis.


Summary

The current ES technology has not yet reached the stage where expert

systems are able to completely substitute human experts. Only a few

systems have performed rather close to a human expert. MYCIN, a medical

diagnostic program, is considered to be the first major ES to perform at

a human expert's level. The term "knowledge-based system" may be more

appropriate for present day technology.

The architecture of an expert system differs from that of a

conventional computer program. The knowledge base of an ES is separate

from the methods of applying the knowledge to the problem, contained in

the inference engine. In a conventional program, the problem related








76

knowledge and the methods for using the knowledge are inter-mixed, which

makes it difficult to modify the program for changes and additions to the

knowledge. Processing in a conventional program is algorithmic in nature,

symbols are used to represent numbers, arithmetic properties, and

mathematical operations. In a knowledge-based system, the inference

engine governs the sequence of rules that are fired to lead to multiple

actions or to no action at all. The knowledge rules may include

heuristics as well as mathematical reasoning.

There are three ways knowledge can be represented in an expert

system: rules, frames, logic, and semantic network. Each method is suited

for a particular type of problem. The frame method is appropriate for a

complex knowledge system. The rule-based system is the most popular

method for knowledge representation, more applicable to a decision problem

with a narrow domain.

Due to availability of micro computer-based ES shells, knowledge

engineers are able to develop decision models without the usage of

programming languages such as PROLOG and LISP. However, shells are only

geared toward specific representation methods and are less flexible than

AI languages. There have been several successful knowledge-based systems

developed in the construction engineering field. The knowledge-based

approach represents engineering expertise, theory, and judgment in an

integrated manner and also offers flexibility in data modifications.













CHAPTER 4
PROJECT COST AND PRODUCTIVITY DATA FOR DAY AND NIGHT SHIFTS

Introduction

The selection of day shift over night shift or vice versa may be

influenced by the project costs of the alternatives. Literature review

indicated a lack of project cost information for an effective comparison

between daytime and nighttime construction. Project related cost is

essentially the contract cost (total work item cost) plus agency

administrative costs (planning, evaluating and monitoring). Most highway

projects are unique, and usually consist of different sets of work items.

This makes it difficult to compare the construction costs for day and

night projects.

Productivity may also influence shift selection. Shift productivity

is affected by several factors, which include traffic volume, type of

work, material delivery, lighting, supervision, communication and worker

morale. High daytime traffic volume affects productivity negatively.

During the day, the work shift is reduced to a 5 or 6 hour period due to

the morning and evening rush hours. While during the night, the actual

working hours are extended. However, poor lighting and low worker morale

during night can decrease crew productivity. In the following sections,

unit project costs and productivity rates are compared between day and

night shifts, based on data obtained from the Florida Department of

Transportation (FDOT).









Project Cost Comparison

Since no two highway projects are exactly the same, work items

may differ accordingly. To make an effective cost comparison, a set of

typical work items have been selected for this study. These work items
were selected based upon: a) their usage during a typical day as well as

night project, b) the significance of their contribution in project cost,

and c) their large quantities. The 8 common items are listed as follows:

1) Removal of existing pavement (unit of measure = square yard, SY)

2) Regular excavation (unit = cubic yard, CY)
3) Bituminous material-prime coat (unit = gallon, GA)

4) Bituminous material-tack coat (unit = GA)

5) Milling existing asphalt pavement-2" depth (unit = square yard, SY)

6) Class I concrete-miscellaneous (unit = CY)

7) Type S asphalt concrete-including bitumen (unit = ton, TN)
8) Asphalt concrete friction course-including bitumen (unit = SY)

The unit prices for the above work items were obtained from the FDOT

official cost estimate for road projects done in 1990. Table 4.1 shows

the statistical summary of the rates for these selected work items

performed during daytime (Ellis, Herbsman, & Kumar 1991). A similar

statistical summary for the eight items done during nighttime is shown in
Table 4.2. In both tables, for each work item, columns 4 to 8 contain:

1) number of samples, 2) mean unit cost, 3) standard deviation of unit
cost, 4) highest unit cost, and 5) lowest unit cost, respectively.

The results of an item-by-item comparison of unit prices are

tabulated in Table 4.3, so that the variation in means between day and

night rates can be determined. Columns 4 and 5 of Table 4.3 show mean


















Table 4.1 Statistical Summary of Unit Costs for
All Daytime FDOT Projects in 1990


Selected Work Items for


Pay Item Number Mean Std.Dev. High Low
Number Name of Item Unit of
Samples $/unit $/unit $/unit $/unit

110-4 Rem. exist. pavt. SY 104 10.52 10.99 100.0 0.39

120-1 Regular excavation CY 151 7.41 7.71 60.0 0.42

300-1-1 Bit. mat'l-prime GA 55 2.30 1.70 6.5 0.01

300-1-3 Bit. mat'l-tack GA 190 1.36 1.36 12.6 0.01

327-70-5 Milling existing SY 23 0.68 0.26 1.5 0.32
asphalt pavt.

400-1-15 Class I concrete CY 70 348.38 234.42 1050 10

5331-2 Type S asph. conc. TN 188 45.88 34.14 382 19

5337-1-2 Asph. conc. fric. SY 102 1.26 0.82 5.54 0.65

Source: Ellis, Herbsman, & Kumar, 1991




















Table 4.2 Statistical Summary of Unit Costs for
All Nighttime FDOT Projects in 1990


Selected Work Items for


Pay Item Number Mean Std.Dev. High Low
Number Name of Item Unit of
Samples $/unit $/unit $/unit $/unit
110-4 Rem. exist. pavt. SY 22 9.54 9.48 50.0 1.84

120-1 Regular excavation CY 20 4.59 2.83 14.1 1.13

300-1-1 Bit. mat'l-prime GA 12 5.19 3.89 15.0 1.00

300-1-3 Bit. mat'l-tack GA 26 1.00 0.33 2.1 0.68

327-70-5 Milling existing SY 19 0.81 0.49 1.65 0.27
asphalt pavt.

400-1-15 Class I concrete CY 17 401.48 153.11 800 125

5331-2 Type S asph. conc. TN 25 34.06 11.93 75 22

5337-1-2 Asph. cone. fric. SY 23 1.27 0.50 2.65 0.75


Source: Ellis, Herbsman, & Kumar,


1991

















Table 4.3 Difference Between Day and Night Unit Costs for Selected
Work Items for All FDOT Projects in 1990.

Difference
Pay Item Name of Item Unit Mean Mean Test
Number (night) (day) Amount Per- Result
($/unit) ($/unit) cent
$/unit (%)

110-4 Rem. exist. pavt. SY 9.54 10.52 -0.98 -9.3

120-1 Regular excavation CY 4.59 7.41 -2.82 -38.1 *

300-1-1 Bit. mat'l-prime GA 5.19 2.30 2.89 125.6 *

300-1-3 Bit. mat'l-tack GA 1.00 1.36 -0.36 -26.5

327-70-5 Milling existing SY 0.81 0.68 0.13 19.1
asphalt pavt.
400-1-15 Class I concrete CY 401.48 348.38 53.1 15.2

5331-2 Type S asph. cone. TN 34.06 45.88 -11.82 -25.8 *

5337-1-2 Asph. conc. frict. SY 1.27 1.26 0.01 0.8


NOTE: An
in


asterisk (*) in the
the hypothesis test


last
for


column indicates significant difference
that work item at 95% confidence level.


Source: Ellis, Herbsman, & Kumar, 1991








82

rates for nighttime and daytime projects respectively. Column 6 of the

table contains the difference in amount for means, where a negative value

indicates lower nighttime costs. For item, bituminous material-prime

coat, percent difference is as high as 125.6%, as shown in column 7. For

item, regular excavation, percent difference is -38.1%, indicating a lower

nighttime unit cost. Although the percent differences have high

variations, these variations are not necessarily conclusive from a

statistical point of view. The significance of these differences can be

tested by performing statistical t-tests for the eight work items at a 95%

confidence level. Many of these differences appear to be inconclusive

because of high standard deviations. The null hypothesis was rejected for

only three items: regular excavation, bituminous material-prime coat, and

type S asphalt concrete.

The total project item cost depends on what items are involved and

on the quantities of those work items. For this reason, further study of

the impact of shift work on project costs was done. Quantity data from

eight selected projects, utilizing most of the above mentioned work items,

was obtained from the FDOT. Table 4.4 lists the corresponding quantities

of work items of the selected projects. Respective item costs for the

eight projects were determined by multiplying the unit costs from columns

4 and 5 of Table 4.3 with the quantities from Table 4.4. The probable

difference in project costs for night and day operations due to the eight

work items, is given by the summation of such products. The total costs

of the work items for each of the selected eight projects is listed in

columns 2 and 3 of Table 4.5. Column 4 shows the difference of day and

night total costs, while the last column shows the percentage difference

with respect to daytime total cost.















Table 4.4 Quantities of Work Items for Eight Selected FDOT Night Projects

Name of Item Unit Projects

#1 #2 #3 #4 #5 #6 #7 #8

Rem exist. pavt SY 416 50 1156 532 96 263 2200

Reg. excavation CY 13398 17909 9952 8279 9306 8595 8137 110041

Bit. mat'l prime GA 86 8888 50 50 50 50 660 -

Bit. mat'l tack GA 18205 44551 41197 13874 19211 31116 8748 74469

Milling asphalt SY 3228 553530 20613 20620 15523 7563 63601 84307

Class I concrete CY 5 1.4 4.58 0.8 9.6 3.2 5 50

Type S asph.conc TN 4403 23284 50010 22889 33935 40972 9275 2912

Asph. conc. fric SY 84008 465686 250955 99472 220267 230100 88005 348080

Source: Florida Department of Transportation, 1990


















Table 4.5


- Effect of Quantity on Project Costs for Eight
Selected Work Items


Project Total Cost of Eight Work Items Difference of Percentage
Night & Day Difference
# Night Cost ($) Day Cost ($) Costs of Day Cost

1 345,399 440,413 -95,014 -21.6

2 2,006,764 2,246,199 -239,435 -10.6

3 2,138,746 2,768,313 -629,567 -22.7

4 984,170 1,279,185 -295,015 -23.1

5 1,727,683 2,228,934 -501,251 -22.5

6 1,768,486 2,284,877 -516,391 -22.6

7 530,719 655,124 -124,405 -19.0

8 1,230,145 1,586,748 -356,603 -22.5

Source: Ellis, Herbsman, & Kumar, 1991








85

According to the study described above, the degree of variation in

unit costs for the eight selected work items, for both daytime and

nighttime construction, is very high. Tables 4.1 and 4.2 reveal that the

standard deviation is in most cases nearly 100% of the mean. This

confirms that unit costs in highway construction are highly project-

oriented and are influenced more by project-related conditions rather than

on type of work shift (day or night). The study also rules out the

speculation that nighttime costs are exceptionally higher than daytime

costs. As seen in Tables 4.1 and 4.2, for most work items the maximum

daytime unit cost is higher than the maximum nighttime unit cost, and

conversely the minimum daytime costs are lower than the minimum nighttime

costs.

Table 4.3 shows that four items (remove existing pavement, regular

excavation, bituminous material-tack coat, and type S asphalt concrete)

have higher daytime mean unit cost. The other four items have a higher

nighttime mean unit cost. T-tests have confirmed significant differences

for only three work items. Two of these items (regular excavation, and

asphalt concrete) appear to have significantly lower mean unit costs

during nighttime. The third item (bituminous material-prime coat) has a

significantly higher nighttime unit cost. It is reasonable to conclude

that the work item characteristics are responsible for such a manner of

differences.

In this study, it is seen that the percentage difference is negative

for all eight of the selected FDOT projects. When the total costs of the

eight items for selected projects are compared, nighttime costs are

observed to be lower than the corresponding daytime costs in the range of








86
10 to 20%. Although the participation of these item costs to the total

contract cost varies, the pattern allows a prediction of a probable

nighttime cost. This prediction theory is utilized in the cost model of

the proposed expert system as described in a later chapter.


Productivity Comparison

A comparison of day and nighttime productivity rates for typical

highway construction activities was possible by obtaining information from

the Florida Department of Transportation (FDOT). Daytime production rates

were collected from a 1988 report, "Establishing Contract Duration Based

on Production Rates for FDOT Construction Projects", prepared by the

University of Florida Civil Engineering Department. The information

includes: 1) number of observations, 2) mean production rate, 3)

standard deviation, 4) maximum and 5) minimum production rates for each

operation, which are further categorized by project type, local

conditions, traffic conditions. Table 4.6 summarizes this data.

The FDOT provided nighttime production data, in form of daily

reports, for a highway project located on 1-95 in St. Johns County,

Florida. A summary of this data is presented in Table 4.7. Plant mixed

surface and milling existing pavement are the two work items for which

data was collected. The mean and standard deviation of production rates

for these work items are shown in columns 3 and 4 of Table 4.7. The

combined results of all the projects, as listed in this table, are used

for nighttime productivity values. Only limited access facility (e.g.

interstates) observations from Table 4.6 are utilized for daytime

productivity values, in order to make an accurate comparison with the 1-95

nighttime productivity values.











Table 4.6 Summary of Productivity Rates for FDOT Construction Projects

PLANT MIXED SURFACES: STRUCTURAL COURSE

Number of Mean Standard High Low
Category Observa- (Tons/ Deviation (Tons/ (Tons/
tions Day) (Tons/Day) Day) Day)
Project Type
Reconstruction 147 833 533 2,359 6
Construction 27 623 639 2,863 114
Intersection 15 122 111 356 10
Bridge 9 178 70 274 84

Local Condition
Rural 111 855 616 2,863 6
Urban 72 436 387 1,638 17
Limited 15 1,090 157 1,247 582

Traffic Condition
Light 20 1,189 761 2,359 119
Medium 81 822 562 2,863 14
Heavy 97 539 426 14 6

Total Combined 198 720 565 2,863 6


MILLING EXISTING PAVEMENT

Number of Mean Standard High Low
Category Observa- (SY/ Deviation (SY/ (SY/
tions Day) (SY/Day) Day) Day)
Project Type
Reconstruction 94 12,350 7,429 32,028 444
Construction 1 2,274 0 2,274 2,274
Intersection 0 0 0 0 0
Bridge 0 0 0 0 0

Local Condition
Rural 48 14,850 8,108 32,028 444
Urban 32 8,987 4,847 20,533 2,351
Limited 15 10,854 6,765 26,422 3,833

Traffic Condition
Light 14 20,306 8,159 32,028 5,488
Medium 32 12,137 7,680 29,376 444
Heavy 49 10,011 5,180 26,422 2,274

Total Combined 95 12,244 7,461 32,028 444

Source: Herbsman & Ellis, 1988













Table 4.7 Summary of Productivity Rates for FDOT Nighttime
Project on 1-95 in St. Johns County


Construction


PLANT MIXED SURFACES: STRUCTURAL COURSE

Project Number Mean Standard High Low
Number of (Tons/ Deviation (Tons/ (Tons/
Samples Day) (Tons/Day) Day) Day)

78080-3420 14 950.68 348.49 1,428.15 320.86

78080-3421 32 1,110.31 327.06 1,602.68 327.08

78080-3422 29 1,093.52 422.97 1,871.21 196.59

78080-3424 20 1,043.45 379.43 1,644.97 325.92

Total Combined 95 1,067.59 378.57 1,871.21 196.59





MILLING EXISTING PAVEMENT

Project Number Mean Standard High Low
Number of (SY/ Deviation (SY/ (SY/
Samples Day) (SY/Day) Day) Day)

78080-3420 7 11,246.42 5,582.80 16,840 2,766

78080-3421 10 7,379.40 1,061.80 9,080 5,333

78080-3424 12 8,256.40 2,957.20 13,864 4,170

Total Combined 29 8,675.70 3,711.90 16,840 2,766

Source: FDOT (1990)












Table 4.8 Guidelines for Estimating Production Rates for FDOT Projects

WORK ITEM DAILY COMMENTS
PRODUCTION


General Time
(Move in)

Clear & Grub

Excavation
(Regular)

Excavation
(Truck Haul)

Stabilized
Roadbed

Bases:
Sand-Clay
Limerock stabil.
Soil Cement

Surface Treatment

Cement Concrete


Milling Existing
Pavement

Plant Mixed
Surfaces

Guardrails


Compression Seal
Replacement

Breaking and
Compacting Exist-
ing Concrete
Pavement


15 days


3 acre


1400
5600
11200

900
3800
7500


4500 SY


900 SY

1800 SY

400 CY

2000 SY
4000 SY

6000 SY


500 TN
1200 TN

300 LF
1500 LF

100 LF

5000 SY


Normally for all projects unless specific
circumstances justify additional time.

Medium clearing (50 to 100 acres)

Small quantity jobs under 100,000 CY
Medium quantity jobs 100,000 300,000 CY
Large quantity jobs over 300,000 CY

Small quantity jobs under 100,000 CY
Medium quantity jobs 100,000 300,000 CY
Large quantity jobs over 300,000 CY

Normal.


Double lift installation.

Single lift installation.

Normal.

Average quantity jobs (under 25,000 CY)
Large quantity jobs (over 25,000 CY)

Average jobs.


Average jobs (less than 50,000 Tns)
Large jobs (over 50,000 Tns)


(less than 5000 LF)
(over 5000 LF)


Small jobs
Large jobs


Normal.


Normal


Source: Herbsman &


Ellis (1988)