Optimization of Multi-Stage Production Decision-Making Based on Dynamic Block-Accelerated Enumeration Algorithm

Abstract

This paper introduces a Dynamic Blocking Accelerated Enumeration Algorithm (D-BES) for high-dimensional production decision-making, employing recursive space partitioning and sensitivity-based pruning to reduce a 20-dimensional search space from 2²⁰ to 2⁵ sub-blocks, achieving a 71.3% computational speedup over exhaustive methods. By integrating Analytic Hierarchy Process (AHP) and Bayesian sampling, D-BES ensures precise defect rate estimation (errors ≤0.03%) and adaptive weight adjustment (±30%), while its multi-objective optimization model lowers the resource-defect conflict coefficient from 0.82 to 0.47. Benchmark tests demonstrate D-BES’s superiority over Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) in convergence and efficiency, achieving 98.3% defect detection accuracy, an optimized defect rate of 0.8%, and a computation time of 0.5 seconds on the MIT PCB dataset. Under ±20% weight fluctuations and 10dB Gaussian noise, D-BES maintains resource consumption stability (σ=0.23) with only a 0.1% increase in defect rates, showcasing its robustness and real-time capabilities for Industry 4.0 smart production systems.