An integration of risk assessment and random-fuzzy network scheduling for complex construction operations

Abstract

The uncertainty may be broadly classified into aleatory and epistemic uncertainty. A large or complex construction project is not only fraught with the aleatory uncertainties, but also influenced by risk factors, such as adverse weather condition, unavailable quantities of materials, operator experience, and poor performance of machines. These risk factors contribute to stochastic phenomenon. Insufficient data or lack of site productivity data compound aleatory uncertainty in the risk assessment and project scheduling by leading to imprecise, vague, and subjective data causing epistemic uncertainty. Although distributions of risk variables and temporal variable are propagated through a simulation model by introducing randomness into the analysis of construction processes, the establishment of probabilistic distributions of random variables is generally provided without a consideration about the epistemic uncertainty. Inaccurate probability distribution can bring about misleading outputs. This study attempts to provide an accurate duration estimate by examining these two types of uncertainty. A risk assessment integrated within random – fuzzy network scheduling method (RAIRFNET) is developed for modeling uncertainties associated with risk assessment and construction scheduling and estimating activity duration and project completion time. The proposed method uses historical data, subjective data from professional experience and judgment, and simulation data to produce information associated with risk affecting activity duration. Risk variables and temporal variable are considered as a random – fuzzy variable and represented by different shapes of membership functions (i.e., nil, rectangular, trapezoidal). The membership function is developed by inserting the internal membership function representing the fuzzy effect into the external membership function representing the random effect which is transformed from the corresponding probability distribution by using the Salicone's method and neurofuzzy metamodel. Mathematics for random-fuzzy variables are used in the network calculation to propagate each type of uncertainty. The proposed method provides results that cover actual duration. The results can reflect reality of a construction project by addressing every uncertainty. The estimated duration varies based on the assigned value of epistemic uncertainty and dependencies between the considered risk factors. An application of nil internal membership functions provides results close to the simulation results as both of them determine only uncertainty due to the random contribution, but these two methods cannot provide results covering the actual duration.