Abstract

Uncertainty is an entrenched characteristic of most construction projects. Typically, probability distributions are utilized to accommodate uncertainty when estimating duration of project’s activities. Distributions are fitted, based on the collected data from construction projects, to estimate activity durations, to assess productivity and cost, and to identify resource bottlenecks using simulation. The subjectivity in selecting these fitted probability distributions is an imprecise process and may significantly affect simulation outputs. Most research works in simulating construction operations has focused predominantly on modeling and has neglected to study the effect of subjective variables on simulation process. Therefore, there is a need for a system, which: (1) handles uncertainty, fuzziness, missing data, and outliers in input data, (2) effectively utilizes historical data, (3) models the effect of qualitative and quantitative variables on the simulation process, (4) enhances simulation modeling capabilities, and (5) optimizes simulation system output(s).
The main objective of this research is to develop a knowledge discovery based simulation system for construction operations, which achieves the abovementioned necessities. This system comprises three stages: (i) a Knowledge Discovery Stage (KDS), (ii) a Simulation Stage (SS), and (iii) an Optimization Stage (OS). In the KDS, raw data are prepared for the SS where patterns, which represent knowledge implicitly stored or captured in large databases, are extracted using Fuzzy K-means technique. During the KDS, the effect of qualitative and quantitative variables on construction operation(s) is modelled using Fuzzy Clustering technique. This stage improves the efficiency of data modeling by 10% closer to real data. The movement of units is modeled in the SS where the interaction between flow units and idle times of resources can be examined to discover any bottlenecks and estimate the operation's productivity and cost. The OS, using Pareto ranking technique, assists in selecting and ranking feasible productivity-cost solution(s) for diverse resource combinations under different conditions.
An automated general purpose construction simulation language (KEYSTONE) is developed using C#. The developed system is validated and verified using several case studies with sound and satisfactory results, i.e. 4% - 11% digression. The developed research/system benefits both researchers and practitioners because it provides robust simulation modeling tool(s) and optimum resources allocation for construction operations.