不确定条件下基于分群策略的柔性Flow Shop调度问题研究

Scheduling in flexible manufacturing environment is characterized by high complexity, inherent uncertainty, and multi-constraint. As an NP-hard combinatorial optimization problem, it has attracted considerable attention of the researchers in both academia and industry. By incorporating operational research and artificial intelligence techniques, this project aims to develop a holonic scheduling model to solve flexible flow shop (FFS) problems under uncertainties. This project mainly covers the following research topics: (1) establishing a holonic scheduling architecture for FFS problems under uncertainties; (2) developing a cluster-based self-organization mechanism to generate holon clusters with different stochastic nature; (3) developing a procedure of approach assignment based on machine learning and simulation techniques to measure the stochastic nature of each holon cluster, and accordingly assigning a suitable approach for schedule generation; (4) establishing a data-mining-based chromosome generation mechanism to improve the performance of population-based meta-heuristic; (5) integrating Q-learning with Contract Net Protocol to provide better adaptability and responsiveness in the face of disturbances. The proposed holonic scheduling model of this project provides a promising methodology to solve FFS scheduling problems under uncertainties, and hence offers great academic and practical potential in real-world production scheduling.

柔性制造环境下的车间生产调度问题具有复杂性、不确定性、多约束等特点，是近年来生产管理和组合优化领域的重点和难点课题。本项目旨在应用运筹学、人工智能和Holonic制造系统（HMS)等学科相关理论和方法，采用分群策略研究不确定条件下的柔性Flow Shop生产调度问题，具有前沿性和探索性。主要研究内容是：分析HMS体系结构和运行机制，建立柔性Flow Shop的Holonic调度模型；依据随机特性分群的思想，研究基于聚类算法的资源Holon（加工机器）自组织机制；采用机器学习和仿真方法，建立Holon群随机特性的预测模型，实现不同生产环境下调度方法的自适应选择；引入基于数据挖掘的种群更新策略，探讨元启发式优化算法对资源Holon自治调度问题的求解效率；采用基于Q-学习和合同网协议的协商调度方法，提高资源Holon动态环境下的协调能力。研究成果将为解决不确定条件下的生产调度问题提供创新方法。

柔性制造环境下的车间生产调度问题是近年来生产管理和组合优化领域的重点和难点课题，具有复杂性、不确定性、多约束等特点。本项目综合运用运筹学、人工智能和Holonic制造系统（HMS)等学科相关理论和方法，采用分群策略求解不确定条件下的柔性Flow Shop生产调度问题。主要研究成果包括：构建了柔性Flow Shop的Holonic调度模型；根据随机特性分群的思想，建立了基于聚类算法的资源Holon（加工机器）自组织机制；采用机器学习和仿真方法，提出了Holon群随机特性的预测模型，该模型可实现不同生产环境下调度方法的自适应选择；引入基于数据挖掘的种群更新策略，提高了元启发式优化算法对资源Holon自治调度问题的求解效率；采用基于Q-学习和合同网协议的协商调度方法，加强了资源Holon动态环境下的协调能力。本项目的研究成果为不确定条件下生产调度问题的求解提供了创新方法，具有较高的理论意义和实际应用价值。