Optimization study of production decision based on Monte Carlo simulation and particle swarm optimization algorithm

Zhaoyang Wu, Bowen Bai, Lin LiuHighlights 2025 · DOI

TL;DR

This paper tackles the complex decision-making challenges in production processes, specifically focusing on optimizing spare parts inspection and overall production flow. It proposes a novel approach combining Monte Carlo simulation for efficient sampling strategies and Particle Swarm Optimization (PSO) for multi-stage production decisions, aiming to minimize costs and enhance product quality.

What Problem Does This Paper Address?

Modern production lines face intricate decisions, from accepting incoming spare parts to managing the entire production lifecycle. Traditional methods, often relying on empirical or simple statistical approaches, struggle with the inherent complexity and uncertainty of these processes. This paper addresses two key issues: 1) optimizing the decision-making for incoming spare parts inspection to balance accuracy and testing costs, and 2) developing a comprehensive strategy for multi-stage production decisions (including inspection, assembly, and handling of defective products) to maximize profitability and product quality.

Methods

The paper employs a two-pronged methodological approach:

  • Monte Carlo Simulation for Spare Parts Inspection: For the incoming inspection of spare parts, a sampling and testing methodology is developed. This involves using Monte Carlo simulation to evaluate inspection accuracy and costs across various sample sizes. By simulating a large number of random samples, the optimal sampling scheme is determined. This method is designed to meet specific confidence requirements for accepting or rejecting batches based on their defective rates.
  • Particle Swarm Optimization (PSO) for Multi-Stage Production: For the broader production process, an Integer Linear Programming (ILP) model is constructed. This model is then optimized using the Particle Swarm Optimization algorithm. PSO, inspired by the foraging behavior of bird flocks, is used to explore a vast number of decision combinations to find optimal strategies for detection, dismantling, and processing, thereby minimizing total costs. The paper also hints at the potential integration of machine learning for continuous optimization.

Datasets

The paper utilizes pre-processed data for spare parts and finished products, including defective rates, purchase prices, testing costs, assembly costs, market selling prices, and costs associated with non-conforming products. This data appears to be derived from an open-source website and is presented in several tables (Tables 1-4). No new benchmarks or datasets are introduced.

Key Insights and Ideas

  • Bridging Simulation and Optimization: The paper effectively demonstrates how Monte Carlo simulation can be used to validate and optimize sampling strategies, while PSO can tackle complex, multi-stage discrete decision problems in production.
  • Balancing Cost and Quality: The core of the proposed solutions lies in finding an optimal balance between the cost of inspection/processing and the desired level of product quality and confidence.
  • Data-Driven Decision Support: By modeling production processes and using simulation/optimization, the paper aims to provide enterprises with scientific and reasonable decision support, moving beyond purely empirical methods.
  • The Power of Heuristics for Complex Problems: PSO's ability to explore a large search space and converge towards optimal solutions for discrete decision problems is highlighted as a key advantage.

Key Results

The paper presents expected costs for six different scenarios of spare parts and finished product states. The calculated expected costs range from 91.20 to 103.20, with the lowest cost observed in scenario 2. The Monte Carlo simulation results are visualized, showing the change in rejection probability with sample size for 95% and 90% confidence levels. The PSO results are visualized in a stereoscopic scatter plot, illustrating the clustering of particles around an optimal position as iterations progress, leading to the minimum expected cost.

Limitations and Future Work

  • Model Complexity: While the paper presents a comprehensive model, the practical implementation and scalability for extremely large and dynamic production environments might be a consideration.
  • Data Specificity: The presented results are based on specific cost and defect rate data. The generalizability to other product types or industries would require re-calibration.
  • Machine Learning Integration: The paper mentions the potential of machine learning (reinforcement learning and deep learning) for continuous optimization, but this aspect is not deeply explored or demonstrated in the current work, suggesting a clear avenue for future research.
  • Further Validation: While the paper verifies the effectiveness of the proposed methods, further real-world case studies or extensive comparative analyses with other optimization techniques could strengthen the findings.

Summary updated

Abstract

In this paper, a solution based on Monte Carlo simulation and particle swarm optimization algorithm is proposed for the problems of spare parts monitoring and production process optimization in the production process of enterprises. A Monte Carlo simulation-based sampling and testing method is designed for spare parts incoming inspection decision, which evaluates the inspection accuracy and cost under different sample sizes by simulating a large number of random samples through a large number of random simulations. Thus, the optimal sampling scheme is determined. For the multi-stage decision optimization in the whole production process, an integer linear programming model is constructed and optimized and solved using particle swarm optimization algorithm. A large number of decision combinations and the foraging behavior of bird flocks are simulated to find the optimal detection, dismantling and processing strategies to minimize the total cost and improve the product quality. The final analysis verifies the effectiveness of the proposed method and provides scientific and reasonable decision support for enterprises in complex production environments.
BibTeX
@inproceedings{wu2025optimization,
  author = {Zhaoyang Wu and Bowen Bai and Lin Liu},
  title = {{Optimization study of production decision based on Monte Carlo simulation and particle swarm optimization algorithm}},
  booktitle = {Highlights},
  year = {2025},
  doi = {10.54097/dgw4ce43},
}
Figure