You are on page 1of 5

Henry R.

Block, CCE Senior Project Engineer, Cost & Schedule

F-2

Richard A. Mazzini, CCE

Supervisor, Project Cost & Schedule

Pennsylvania Allentown,

Power & Light Co.

Pennsylvania Power & Light Co. Two North Ninth Street Allentown, PA 18101

Two North Ninth Street PA 18101

A RETROSPECTIVE

EVALUATION OF A PROBABlLlST,lC SCHEDULE ANALYSIS


Construction began in November 1973 and is now 65% complete. At the time of the study, the project was 40% complete. The fuel load date for Unit 1 was then scheduled for May 1980 and for Unit 2, November 1981, with in-service dates of February 1981 and May 1982, respectively. Recently the schedule was revised to reflect Commercial Operation of Unit 1 in January 1982 and for Unit 2, January 1983. The most recent cost estimate, completed in October 1979, is $2.7 billion. Objectives and Technique

Introduction The environment of uncertainty associated with all "Super-Projects" of the influences on today's causes great difficulty in long-term schedule Current transportation and major energy assessment. projects are expecting construction durations in I excess of ten years and in most cases these esti-' Nuclear power mates far exceed earlier projections: plant construction is typical,of this generation of construction projects and is plagued by all of the factors that make long term schedule analysis most Multi-year procurement lead times, difficult. changing regulatory requirements, labor productivity and shortage of skilled management variability, personnel contribute to the uncertainty facing long-term planners. As a result of this situation, those responsible for providing accurate assessments of project,schedy;;i",i;=;;~+Jys, '!y~~ f5p '"&U,s;ro,'a;;;;~;f'approach permits one to understand better the overall schedule exposure on a project and forces analysis of potentially critical activities which might otherwise have been dismissed as non-critical. In 1977 PP&L initiated an analysis of this type in order 'to understand better the exposure to delay in construction of Susquehanna Steam Electric Station. This paper is an examination of that study in light of subsequent events in order to evaluate the effectiveness of this technique and emphasize that in addition to successfully achieving the desired goals, many additional benefits were realized. Throughout this narrative, emphasis will be. on the various qualitative benef,its obtained from this approach. The Project The Susquehanna Steam Electric Station,has two boiling water reactors rated at 1100 m electric each and is located along the Susquehanna River Luzerne County, Pennsylvania, near the city of Berwick. The station is owned by Pennsylvania Power and Light Company and Allegheny Electric Cooperative. Major engineering was.begun by Bechtel in 1970.

Prob,abilistic scheduling starts with the premise that an activity can be better represented by a range of durations.rather than a single value. An appropriate probability distribution is selected defining the'range of values. Monte Carlo (random walk) techniques are then used to select durations and determine the critical path,. This process is repeated many times and a probability distribution for the total project duration is generated. The probabilities that various paths through the logic will'eventually prove critical can also be deterThe theoretical justification and, the stamined. tistical basis for this technique is described in detail elsewhere(l* 6). When such a probabilistic schedule for completion hanna SES was requested, four major objectives: 1. 2. 3. 4. ,Identify senting Identify ties.on assessment of the current of the Units' at SusquePP&L proceeded to identify repreactiviacex-

the components of the schedule the highest risk; the ultimate impact the Project Schedule; of these

Determine the nature in which the various tivities influence the schedule; Determine posure. methods of mitigating schedule

in

To ,perform this analysis, PP&L developed a simplitilized fied project construction logic model an a network simulation code called GERTSIMt7Y which combined a conventional PERT type analytical program with a Monte Carlo generator.

F.2.1

1I

Schedule

Model and Data Generation --

A significant benefit in performing an analysis of this type in this discussion is derived from the insights gained during development of the schedule model. The logic used is shown in Figure 1. A closer examination reveals that this network is actually constructed from summaries of six major These were chosen to facilicategories of work. tate the probabilistic analysis and activities having no real chance of delaying work were omitted. The major categories are: 1. 2. 3. 4. 5. 6. Licensing Reactor building civil work

Most likely, optimistic and pessimistic durations for each activity were then developed and were Any probability distribuinput to the simulation. tion can be used that is consistent with these In PP&L's analysis, a Beta distribujudgments. tion was used because its shape could be modified conveniently to reflect almost all conditions To arrive iat these suggested by the data. parameters, projections were made on the basis of information available from similar projects, jobsite history, if available, assessments of the degree of difficulty of the to-go work, status of design and procurement support, etc. An example of the type of data considered is shown in Figure 2. A sheet of this type was prepared for every activity in the network and served as a reference for subsequent reevaluation.

Containment Installation commodities Auxiliary Start-up of bulk structures piping and electrical equipment Each iteration of the program included a critical path calculation based on durations randomly selected from the probability distributions. Several thousand iterations were performed and the results summarized into a total project probability distribution.

and associated

SUSQUEHANNA S .E .S. TO-GO ACTIVITY SEQUENCE [CUT-OFF DATE: 5-1-77)

Figure 1

ff.a.2

ACTIVITY

NUMBER:

150-290

SPRAY POND PIPE

SOURCES:

Milestone Report

Summary

Schedule;

Field

Subcontract

Status

TO-GOPROJECT DURATION TO FUELLOAD CUMIJLATIYE PROBABILITY FOR 5000 REALIZATIONS 100


*

i '\

SCOPE:

Includes installation of all of pipe in trenches, pouring around risers and installation flooding of spray pond.

spray pond of concrete of spray

piping, missile heads,

backfill barriers up to

90
CURRENT SCHEDULE: -we l/1/78 to 12 Months l/1/79

80 70 60 50

COMMENTS:

Milestone work. 4/l/78 work. including likely schedule

Sumnary Schedule allows full 12 months to complete Field subcontract stat,,s report starts work on and completes by 7/l/78. Bechcel performs all pipe Twelve months is more than adequate for all work Improvement by two months is contingencies. Field subcontract and a third month possible. is overly optimistic.

CONCLUSION:

Optimistic Most likely

- 9 months - 10 months - 12 months

40 30 20 10
TYPICAL ACTIVITK DATA SHEET

Pessimistic

Figure 2

40

41

42

43

44

45

46

47

48

TO-GO DURATION, YONTHS (FROM S-1-77)


Figure 3 In order to develop the schedule logic and activity duration distributions, numerous reviews were conducted with key PP&L and selected Contractor personnel. This was an extremely important aspect of the analysis because it required the attention of representatives of many project functional groups including engineering, procurement, construction, start-up and management. Furthermore, since the analysis demanded a consensus for selection of the model, the various parties had to exchange considerably more information than might occur in the normal day-to-day workings of the project. In addition, the process of making these determinations was extremely valuable in establishing for management the relationship of Susquehanna's current plan to previous performance and reasonably attainable goals. Analyses of this type also provided an opportuni'ty for management to gain insight into the status of the construction and forced judgments to be made that put the various categories of work into proper prospective. Managers who might ordinarily be preoccupied with short I term-problems were compelled to decide on the long term direction of the project and actually quantify the results of their speculation. Results The results of the analysis are illustrated in Figure 3. The major path producing these results (although others 'contributed) is shown in Figure

This path is through electrical bulks followed by a chain of start-up activities. Further analysis revealed that the influence of the start-up activities was small. It was concluded that the single activity which was most likely to delay the project was installation of electrical bulks. In retrospect, this result was not surprising since delays in prior civil and piping activities had compressed the time available for electrical installations. At the time, however, it was considered a revelation since there were more visible short term problems at hand. Two other paths closely approached criticality. These were the control room installation and wire and cable connections. The control room exposure resulted from the potential for delayed material delivery and massive testing requirements, while terminations required extremely ambitious installation rates because of schedule compression. Small pipe was revealed as potentially cause of schedule compression, status and craft availability. critical of design be-

4.

These results 'of this type

illustrate of analysis.

the particular At the time

importance this assess-

F.2.3

ment was performed, most resources were being deCritivoted to support of the piping operations. cal craft shortages and material delays seemed to be the most urgent path to address. The analysis electrical bulks were to showed that, in fact, become far more critical and suggested that planning should begin promptly to minimize the impact This raises an interesting question. of this path. In view of immediate resource-demanding problems, how can engineering and construction management be convinced to address a problem that is revealed by this type of analysis, months before the earliest phases of the activity is to begin? Certain activities initially anticipated cal were not confirmed by the analysis. included: 'Drywell equipment and bulks *Wetwell modification -Licensing Each of these was given side of the simulation. In two years following supplemental the initial analysis study, outas critiThese

the schedule model has been actualized. Wire and cable installation has indeed become the major restraint to initiating start-up activities. Small pipe and control room verfication are also significant bottlenecks, all of which were anticipated by the analysis. Major piping modifications required in the drywell and the Three Mile Island mishap have caused the drywell and licensing to be more urgent than anpaths, respectively, ticipated by the study. Improvements After the initial results were reviewed and analyzed, it was determined that certain features of the simulation procedure needed improvement. there seemed to be a mismatch Most significantly, between the method of collecting judgments on activity durations and the statistical approximations made within the simulation processor. calculation included Simply put, the approximation essentially all of the potential data points for a given activity between the optimistic and pessimistic durations and constructed a probability distribution around the most likely duration

much of

F.2.4

However, most judgments made in,prepsnecified. aration of the possible range of activity were The selection of a probability not so absolute. distribution around the most likely specified. However, most judgments made in,preparation of the possible range of activity were not so absoThe selection of a probability distributios lute. should have recognized that when an individual was asked to specify an optimis:ic or pessimistic duration for an activrsy-,-he/She did not necessarily mean that there is zero probability of an acIn fact, these tivity taking more or less time. judgments are probably made with consideration that durations outside those extremities may have some small chance of occurring (say 5%). Also, determination of the hierarchy of critical paths was cumbers me in the initial study. A tabulating criticality index '76, for each activity the.relative frequency that an activity is This realized on the critical path was required. would define how often an activity is on the critillustraFor example, in the logic ical path. tions in Figure 1, Nuclear Steam Supply System start-up activities (activities 330-350 and'350370) are critical 91% of the time but installation of large pipe (activity 40-140) only 2% of the, time. This is a very precise way of directing management's attention to the most significant problems. Finally, a better method to initiate subsequent activities when a bulk installation was partially The earlier analysis recomplete was needed. quired that the bulk installation duration be divided at the initiation point.of the subsequent activity. The random selection process might conceivably realize radically different parameters ' for the two parts of the bulk installation, thereby cancelling each other. Statistically, this is not acceptable since these parts are not mutually exclusive. Conclusion Analyses of the type described in this paper have become standard tools on major.construction projects. Many packaged programs are now available from various vendors that are similar to the effortdeveloped by PP&L for internal ,use. In our work, this process identified problem areas that were not obvious from the conventional deterministic evaluation of schedule direction. Perhaps the,least appreciated, but most important benefit of such an analysis is .that it forces' managers to' participate in the analytical process such as creating schedule models, determining

potential influences on individual schedule tivities, and observing the manner in which uncertainties may impact the schedule.

acthese

PP&L has updated its schedule analysis,twice since the original results were obtained; The first revision took place three months after the first, in August of 1977, because changes in certain field conditions impacting electrical installations had occurred. The results were consistent with the,earlier study. Early 'in 1978 the imrpovements referenced in an earlier section were made by incorporation o new, more flexible precessor called Q-GERT(5f.a At this include time, the schedule model was expanded to Unit 2 and the restraints between units.

Because of the status of construction, current plans are to eliminate Unit 1 logic from the analysis and initiate a periodic cycle for a revised Unit 2 analysis. References 1. An;, A.H.S., Abdelnom, J., and Chaker, A.A., "Analysis of Activity Networks Under Uncertainty", Journal of the Engineering Mechanics Division, ASCE, Vol; 101, No, EM4, Proc. Paper'11514, August 1975, pp. 373-387. Campbell, D.W., "Schedule Risk Analysis", Proceedings, 4th Annual Seminar Symposium, Project Management Institute, Drexel Hill, October 1972. Crandall, Journal of Vol. 102, ber 1976,

2.

PA,

3.

K.C., "Probabilistic Time Scheduling" the Construction Division, ASCE, No. C03, Proc. Paper 12398, Septempp. 415-423.

4.

Klingel, A.R., Jr., "Bias in PERT Project Completion Time Calculations for a Real Network", Management Science, Vol. 13, No. 4, December 1966, pp. 194-201. Pritsker, A.A.B., Q-GERT Networks, Modeling and Analysis Using Wiley, Newxrk, 1977.

5. 6.

Van Slyke, R.M., "Monte Carlo Methods and the PERT Problem", Operations Research, Vol. 11, No. 5, September 1963, pp. 839-860. Whitehouse, G.E., Systems Analysis and Design Using Network Techniques, ,Prentice-gl, Inc., Englewood Cliffs, NJ, 1973.

7.

F.2.5

You might also like