Application of Improved Genetic Algorithm in Motor Production Line

According to the motor production line of M Company, this study addresses variations in processing sequences and times across multiple products. A mathematical model is established with objectives to minimize workstation count and maximize load uniformity. An improved genetic algorithm is proposed to resolve issues including unbalanced workstation loads and bottleneck drift in mixed-model assembly lines. First, all processing times are standardized into a virtual single product by taking the maximum processing time for each operation. Second, to ensure initial population diversity given numerous operations, both Breadth-First Search （BFS） and Depth-First Search （DFS） strategies are employed to generate feasible operation sequences. These sequences are then selected via roulette wheel selection to form paternal and maternal populations. Finally, two crossover algorithms （Fixed-Point Variant Crossover and Random Two-Point Crossover） are designed for early and late iterations, respectively. Results show that optimized workstations decreased from 9 to 7, line balance rate improved from 66% to 86%, and smoothness index reduced to 0.9. The concentration of operation time distribution and balanced workload across workstations demonstrate the effectiveness of the algorithm. In addition, the Arena simulation platform was utilized to precisely quantify and dynamically validate the personnel bottleneck at the 5th key workstation. Furthermore, by implementing flexible staffing and cross-training, the optimization results were further consolidated.