A learning-based metaheuristic for a multi-objective agile inspection planning model under uncertainty

In this paper, we present an agile integrated inspection-operation planning model wherein inspection actions are planned alongside the machining operations to make the production process agile. Such an agile integrated plan can respond quickly to inspection-machining needs while still controlling costs and quality. A tri-objective mixed-integer nonlinear programming (TMINLP) model is developed for planning the integrated process in a serial multi-stage production (MSP) system. This model addresses several inter-related decisions; (1) what is the most appropriate inspection process for a quality characteristic, (2) at which stage the inspection of these quality characteristics should be performed, (3) how these inspections should be performed, (4) which inspection tools should be used, and (5) which machine should operate on products. The three objectives are: (1) minimizing the total manufacturing cost, (2) minimizing the number of nonconforming products shipped, and (3) minimizing the total manufacturing time for each product. We also address the uncertainty of manufacturing parameters and equipment disruptions. To solve the model, a novel learning-based metaheuristic is developed based on Multi-Objective Differential Evolution (MODE) algorithm, k-Means clustering method, and an Iterated Local Search (ILS) algorithm. The proposed learning-based metaheuristic algorithm is then integrated with the Taguchi Loss Function and Monte Carlo methods to address the input parameters’ uncertainty. The proposed model and solution algorithm are validated through a set of experiments against optimal solutions, and benchmarked against four existing well-known approaches, i.e. NSGA-II, MODE and two learning-based metaheuristics. The proposed approach is applied to a real industrial case and insights are provided.