Simulation-based analytics for fabrication quality-associated decision support

Abstract: Computer-based quality management systems have been widely implemented throughout the construction industry as per the requirement of the International Organization for Standardization (ISO) 9000. Although these systems have facilitated the collection of vast amounts of quality management data, conversion of this data into useable information remains challenging for many practitioners. Automated, data-driven quality management systems, which facilitate the transformation of data into useable information, are often implemented to enhance decision-making processes. However, for a data-driven quality management system to be successful, it must accurately estimate process uncertainty. Integration of accurate, reliable, and straightforward approaches that measure uncertainty of inspection processes are instrumental for the successful implementation of automated, data-driven quality management systems. This research has addressed these limitations by exploring and adapting Bayesian statistics-based analytical solution and Markov Chain Monte Carlo (MCMC)-based numerical solution for fraction nonconforming posterior distribution derivation purposes. Using these accurate and reliable inputs, this research further develops novel, analytically-based approaches to improve the practical function of traditional pipe welding quality management systems. Multiple descriptive and predictive analytical functionalities are developed to support and augment quality-associated decision-making processes. These include (1) operator quality performance measurement, (2) project quality performance forecast, (3) product complexity measurement, and (4) rework cost estimation and control. Multi-relational databases (e.g., quality management system, engineering design system, and cost management system) from an industrial company in Edmonton, Canada, are investigated and mapped to implement the proposed novel approaches, and case studies are conducted to demonstrate their feasibility and applicability. This research has contributed to the academic literature by: (1) providing a novel Bayesian-based approach for fraction nonconforming uncertainty modelling to address hard issues in simulation input model updating; (2) creating an MCMC-based numerical solution for complex probability distribution approximation; (3) developing a dynamic simulation environment that utilizes real-time data to enhance simulation predictability; (4) advancing uncertain data clustering techniques using Hellinger distance-based similarity measurement; (5) providing a systematic approach for analyzing product complexity using the indicator of product quality performance; and (6) creating a novel absorbing Markov chain model for simulating construction product fabrication processes associated with rework. The industrial contributions of this research are identified as: (1) developing a simulation-based analytics decision-support system to enhance quality-associated decision-support processes; (2) creating reliable and interpretable decision-support metrics for quality performance measurement, complexity analysis, and rework cost management to reduce the data interpretation load of practitioners and to uncover valuable knowledge and information from available data sources; and (3) generating meaningful simulation results to assist practitioners in performing quality and rework cost risk analysis during both the project planning and execution phases of a project.