Increasing the Accuracy of a Prefab Building Design Process Simulation Using Simulated Annealing

Monte-Carlo simulation analysis has been discussed in project management literature as tool for proactive scheduling and to gain better insights into projects which are characterized by a high level of complexity and uncertainty, such as the design phase of prefab building projects. The application of simulation as proactive scheduling tool in construction projects is hampered by limited accessibility of proper input data, though, because of long project duration, the often temporary organization and multidisciplinary nature of such projects. In this study we use simulated annealing to adjust parameters of a simulation model for which the simulation outcome is sensitive to data perturbation by making use of data from related parameters which is easier to estimate. The applicability of the approach was demonstrated on a real life project, the construction of a 1100 m2 residential building in Sweden. More precisely, we used Design Structure Matrix simulation, i.e. an activity network based Monte-Carlo simulation technique with which stochastic project evolution (deviations from the planned activity sequence due to unexpected iteration of sub-processes) can be simulated, to model the workflow of the design process of the observed project. Then, by means of the simulated annealing approach, we adjusted the rework probabilities (model parameter) such that the frequencies of executed activities in simulated activity sequences fitted the frequencies as observed in the real project. Adjusting input data by using prior knowledge of the dependencies of the project activities and cross analysis with related data that is easy to estimate would help to increase the accuracy of simulations when access to statistical data of the input variable in question is limited. The suggested approach is interesting for practitioners who work with standardized design processes (e.g. as part of standardized building systems) and continuous improvement.