Chance-constrained stochastic assembly line balancing with branch, bound and remember algorithm

Abstract

Assembly lines are widely used mass production techniques applied in various industries from electronics to automotive and aerospace. A branch, bound, and remember (BBR) algorithm is presented in this research to tackle the chance-constrained stochastic assembly line balancing problem (ALBP). In this problem variation, the processing times are stochastic, while the cycle time must be respected for a given probability. The proposed BBR method stores all the searched partial solutions in memory and utilizes the cyclic best-first search strategy to quickly achieve high-quality complete solutions. Meanwhile, this study also develops several new lower bounds and dominance rules by taking the stochastic task times into account. To evaluate the performance of the developed method, a large set of 1614 instances is generated and solved. The performance of the BBR algorithm is compared with two mixed-integer programming models and twenty re-implemented heuristics and metaheuristics, including the well-known genetic algorithm, ant colony optimization algorithm and simulated annealing algorithm. The comparative study demonstrates that the mathematical models cannot achieve high-quality solutions when solving large-size instances, for which the BBR algorithm shows clear superiority over the mathematical models. The developed BBR outperforms all the compared heuristic and metaheuristic methods and is the new state-of-the-art methodology for the stochastic ALBP.