Abstract:In the high-end electronics manufacturing sector, annual production planning at the beginning of the year is frequently challenged by global resource conflicts and delivery risks due to the concurrent execution of multi-variety, variable-batch, and highly complex projects. Traditional manual block-based scheduling approaches suffer from low efficiency and poor feasibility. To address these issues, this paper proposes an intelligent scheduling methodology tailored for multi-variety, variable-batch production of electronic assemblies. Centered on a genetic algorithm, the method establishes a closed-loop framework encompassing static scheduling, bottleneck identification, dynamic capacity expansion, and rescheduling. Initially, it minimizes the total weighted tardiness across all projects, aligning with delivery-oriented management objectives. The model explicitly captures the synchronized scheduling logic of batch-split flows by incorporating operations splitting, inter-batch intervals, and full-kit batch constraints, thereby faithfully reflecting real-world production dynamics. A two-segment encoding scheme—integrating process sequence and cycle-time coefficients—is employed to generate feasible solutions, complemented by a subspace optimization strategy to mitigate premature convergence in large-scale instances. Furthermore, the approach automatically identifies overdue projects and bottleneck resources, and—within enterprise-defined limits—dynamically expands eligible capacities before triggering iterative rescheduling to ensure on-time delivery. Validation using real-world data at an electronics manufacturing enterprise demonstrates that the proposed method reduces the number of overdue projects, lowering the overdue rate from 50.47% to 24.30%,and compresses the overall production cycle by 30.9%, effectively eliminating resource conflicts and significantly enhancing delivery precision.