Neuro-fuzzy system dynamics technique for modeling construction systems

Abstract

The performance of construction systems (e.g., activities, operations, projects) is commonly measured using different indicators, such as productivity or production rate. The accurate prediction of performance, which is an important concern of construction researchers and practitioners, requires effective techniques for construction modeling. However, the complexity of construction systems creates three challenges for construction modeling: (1) construction systems are affected by numerous interacting factors, (2) the factors that affect construction systems often exhibit both probabilistic and non-probabilistic uncertainty, and (3) construction systems are dynamic. Fuzzy system dynamics (FSD) is a simulation technique that can be used for modeling construction systems with the potential to address these three challenges. However, the application of FSD in construction is still limited due to its low accuracy for modeling the non-linear, complex, and highly-dimensional relationships that exist between the system variables. Currently, these system relationships are most often defined in FSD by linear regression, due its computational simplicity. This paper introduces a new hybrid technique – neuro-fuzzy system dynamics (N-FSD) – by integrating FSD with hybrid neuro-fuzzy systems. In N-FSD, hybrid neuro-fuzzy systems are used to define the non-linear, complex and high-dimensional relationships between the system variables, which improves the accuracy of FSD models in construction applications. The applicability of the N-FSD technique is tested through a construction case study by modeling the production rate of earthmoving operations.