Dynamic reconfigurations of matrix assembly layouts

Traditional assembly lines have become less efficient due to increasing customization and changing demand (e.g., the trend in e-vehicles). Matrix assembly systems, in which automated guided vehicles move products between the workstations laid out on a grid, are gaining popularity. One advantage of such systems is that they are easier to reconfigure compared to traditional assembly lines. In this work, we develop a methodology for the configuration and reconfiguration of matrix assembly layouts under changing demand. The decisions consist of selecting active stations, task assignments, and product flows for each period of a multi-period planning horizon. The three objective functions minimize the number of active stations, the number of reconfigurations, and the total flow distance. We formulate a lexicographic multi-objective mixed-integer linear programming model for this problem. We develop an exact solution approach using period-based, layout-based, and Benders decompositions. For our numerical tests, we adapt standard instances from the literature. In terms of computational performance, our approach is, on average, 53.3% faster than the original MIP solved with a commercial solver for practice-size instances. Our insights reveal that matrix layouts with dynamic reconfigurations enhance the active number of stations by 31.3% and reduce flow distances by 12.4% on average, compared to static layouts.