Frontiers of Engineering: Reports on Leading-Edge Engineering from the 2001 NAE Symposium on Frontiers of Engineering (2002)

FENIOSKY PEÑA-MORA

Civil and Environmental Engineering Department

Massachusetts Institute of Technology

Cambridge, Massachusetts

Construction processes for civil infrastructure involve inherently complex interactions among variables, including but not limited to physical attributes, logistics, resource availability, budget restrictions, and management techniques. Poorly planned interactions among these variables lead to inefficiencies and uncertainties in project execution, a deterioration of planned construction sequences, schedule delays, and increased costs. The impact of such unbalanced interactions is greater in concurrent construction, a technique widely used in modern construction projects and a critical capability for returning infrastructure to service after a major disaster. In this paper, I will present a strategy for simulation-based reliability buffering that can contribute to more robust construction plans, reduce uncertainties, and mitigate the impact of changes.

The purpose of a reliability buffer is to avoid disruptions in a project schedule by failures in individual activities. Reliability buffering involves the systematic pooling, relocating, resizing, and recharacterizing of contingency buffers (Figure 1). Reliability buffering begins by eliminating contingency buffers that are explicitly or implicitly based entirely on an individual activity. In this way, these activities are subject to appropriate schedule pressure and a rubber band effect is avoided. In establishing precedences, the reliability buffer, which is inserted before the downstream activity, can be characterized as a time to identify problems or finish upstream activities and ramp up resources for completing downstream activities. Because the reliability buffer is put at the beginning of an activity instead of at the end, it addresses the issue of poorly defined tasks that

FIGURE 1 Steps in reliability buffering.

require time for better definition. Reliability buffering makes it possible to focus on activities with problems before they activate a domino effect, as might happen with traditional buffers.

Because reliability buffering is based on a simulation approach, it provides a systematic way of sizing a buffer. The buffer must be long enough to ensure that downstream activities will be performed according to plan (i.e., “reliable”). If a buffer is too long, it can create unproductive, idle time. Therefore, a buffer should be sized in a systematic way (based on simulation and analysis) rather than in an ad hoc way based only on individual experience. Moreover, with a dynamic buffering process, the initial size and location of a reliability buffer can be changed at any time during the construction, which might also change the initial precedence relationships.

Successful reliability buffering requires paying attention to feedback processes, which contribute to indirect and/or unanticipated events during a project and make the construction process dynamic and unstable. These inherent instabilities cannot be captured with traditional planning tools. Sometimes steps taken to reduce variations from the planned performance can fix problems and improve performance but, at the same time, worsen performance in another area as a result of unanticipated side effects. For example, when a construction project is behind schedule, one possible way to meet the original schedule is to replace

current equipment with high-performance equipment. At first glance, this change should facilitate the construction process. However, it may take time for workers to learn how to operate the new equipment, or it may be difficult to coordinate use of the new equipment with subsequent processes. If the “fix” reduces productivity and increases coordination problems, changing equipment can actually increase the delay in the construction schedule. These feedbacks must be identified and analyzed before a change is made. Once major feedbacks have been identified, construction processes can be simulated more realistically before actual resources are committed.

Dynamics should be captured in the construction schedule. Factors that trigger feedbacks in construction are changes, dependencies among activities, construction characteristics, and human responses to the work environment and policies. Normally, changes refer to work state, processes, or methods that deviate from the original plan or specification. Changes are major contributors to dynamics and instabilities; changes also create non-value-adding iterations.

As Figure 2 shows, changes are usually made to improve the quality of work or working conditions or to accommodate changes in scope. In addition, changes that have already been made can lead to other necessary changes in concurrent, succeeding, or preceding tasks. For example, changes in design that have been made by mistake can cause subsequent changes in construction. In this case,

FIGURE 2 Changes as triggers of iteration.

even though the designers are responsible for the design changes, those changes can require changes by the construction crew. Changes can be categorized as unintended changes (such as the changes just described) and managerial changes, which are intentional decisions made during quality management or project monitoring and control. Both kinds of changes can lead to subsequent changes or require rework.

Sometimes, by adopting managerial changes, rework on problematic tasks that would require more resources can be avoided. However, even these changes can lead to subsequent changes that might have more of a negative impact on construction performance than the rework option. For example, if some piles have not been correctly positioned, it may be possible to proceed with the super-structure without correcting the position of the piles by changing the position of columns. However, this change option may require unplanned cantilever construction to preserve the original floor layout. The impact of this option will have to be compared to the impact of redriving the piles. The decision must be based on a good understanding of how changes evolve to non-value-adding iterations, which can have unanticipated, indirect side effects. This is particularly important for concurrent construction. In the case of reconstruction after a disaster, for example, a rushed schedule with limited resources can lead to managerial changes that may actually delay putting infrastructure back in service. In short, because construction changes combined with other factors have different impacts on the construction system, it is important to understand how they can affect the planning and control of construction projects.

The impact of changes on construction performance can vary depending on whether changes are managerial (intentional) or unintentional. As Figure 3 shows, managerial changes can create subsequent non-value-adding iterations both at the point of change (C_up) and in downstream activities (C_dn). Thus, managerial changes might have more of an impact on construction performance than rework, depending on the sensitivity of associated tasks to the change and how much work has been done by the time the change is introduced. The impact of a managerial change on the upstream activity (C_up) is in proportion to the sensitivity of the upstream work to internal changes and the progress of the upstream work. The impact of the change on the downstream activity (C_dn) can be measured as a function of sensitivities to changes in the upstream work and progress in the downstream work at the time the change is made.

Unintended changes have more complex impact patterns. Normally, the impact of a change increases with the length of time before discovery and the distance of the discovery from the location of the original change. Unintended, undiscovered changes can create a ripple effect that affects all subsequent work. The impact of an unintended change can vary depending on whether the change

was the result of poor workmanship or the result of undiscovered changes in upstream work. The work quality of an activity is in proportion to its reliability, while the effect of an upstream change on work quality is a nonlinear function. By incorporating all of these determinants into the plan, the change impact on downstream activities can be categorized in terms of type, path, and magnitude.

If reliability buffering is used in the cases described above, the impact of subsequent changes in downstream activity can be absorbed by the systematical assigning of time buffers. Systematically and strategically located and sized time buffers can help reduce the domino effect of changes on downstream work by effectively controlling the start time and progress of the downstream work.

Reliability buffering is an effective technique for more robust construction planning and for addressing inherent uncertainties and the impact of intended and unintended changes. Research results thus far have shown that appropriately pooled, located, sized, and characterized reliability buffers could reduce the impacts of change on construction processes. In addition, case studies on bridge construction projects have demonstrated the applicability of reliability buffering for critical infrastructure construction. More research will have to be done to analyze fully the dynamics of construction projects in chaotic environments, such as disaster recoveries, before reliability buffering will be widely accepted by the construction industry.

I would like to thank my graduate students for making this research possible: Moonseo Park, Margaret Fulenwider, and Sanghyun Lee. I would also like to acknowledge the contribution to this research by David N. Ford, assistant professor at Texas A&M University; and Joseph Peck, corporate planning and scheduling manager, William Lemoine, vice president, and John Foster, senior project manager, at Modern Continental Co.; and Philip Helmes, vice president at InteCap, Inc. Finally, I would like to acknowledge the support for this research received from InteCap, Inc., the National Science Foundation CAREER Award, and the White House PECASE Award CMS-9875557.

Fazio, P., O. Moselhi, P. Theberge, and S. Revay. 1988. Design impact of construction fast-track. Construction Management and Economics 6(2):195–208.

Goldratt, E.M. 1997. Critical Chain. Great Barrington, Mass.: North River Press.

Huovila, P., L. Koskela, and M. Lautanala. 1994. Fast or concurrent: the art of getting construction improved. Pp. 143–158 in Proceedings of the 2nd Workshop on Lean Construction, Santiago, Chile, L. Alarcón, ed. Rotterdam, Netherlands: A. A. Balkema.

Park, M. 2001. Dynamic Planning and Control Methodology for Large-Scale Concurrent Construction Projects. Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge.

Peña-Mora, F., and M. Park. 2001. Dynamic planning for fast-tracking building construction projects. Journal of Construction Engineering and Management 127(6):445–456.

Russell, A., and M. Ranasinghe. 1991. Decision framework for fast-track construction: a deterministic analysis. Construction Management and Economics 9(5):467–479.

Tighe, J. 1991. Benefits of fast tracking are a myth. International Journal of Project Management 9(1):49–51.

Williams, G. 1995. Fast-track pros and cons: considerations for industrial projects. Journal of Management in Engineering 11(5):24–32.