基于BN凸模型和混合群智能的多态复杂液压系统可靠性分析及优化研究

Reliable operation is of great practical significance for hydraulic systems, however the multi-states, complexities and uncertainties characteristics of hydraulic systems, cause the existing reliability analysis and optimization methods difficult to meet the engineering requirements. Aiming at the above problems, reliability analysis and swarm intelligence optimization for hydraulic systems will be studied in this project. Firstly, in order to solve the multi-state complex reasoning and uncertain fault information, Bayesian network(BN) convex model method for reliability analysis will be proposed, definition of importance and sensitivity indices will be studied, and reliability optimization models combined with resource constrains and engineering practice will be constructed; Then, reliability particle swarm optimization methods will be studied, to overcome the shortages of the existing particle swarm optimization algorithms which considered one kind of attraction-repulsion rule and one kind of population topology, hybrid attraction-repulsion and hybrid topology will be proposed, and in order to simulate the biological evolution mechanism, dynamic topology particle swarm optimization algorithm with notes' birth-death variation and edges' changes will be proposed, to improve the global search ability and optimization accuracy; Finally, to overcome the inherent optimization performance and applicability defects caused by single population limitations in the particle swarm and ant colony, reliability hybrid swarm intelligence optimization methods will be studied, particle swarm and ant colony hybrid swarm intelligence optimization algorithm will be proposed, to further improve the optimization performance and applicability. Combining with theory, simulation and experiment research, new methods of reliability analysis and optimization for multi-state complex hydraulic systems will be established, to provide supports for engineering applications.

液压系统的可靠运行具有重要的现实意义，但液压系统具有的多态性、复杂性、不确定性等特征，导致现有的可靠性分析、优化方法难以满足工程要求。本项目针对上述问题开展液压系统可靠性分析及群智能优化研究。首先，针对多态复杂推理计算和故障不确定问题，提出贝叶斯网络(BN)凸模型可靠性分析方法，给出重要度和灵敏度指标，结合资源约束要求构建符合工程实际的可靠性优化模型；然后，研究可靠性微粒群优化方法，针对微粒群单种引斥力和拓扑结构的不足，提出混合引斥力、混合拓扑微粒群算法，模拟生物进化机制，提出节点生灭和边变化的动态拓扑微粒群算法，提高全局搜索和寻优精度；最后，针对微粒群、蚁群因单种生物种群局限性而存在优化性能和适用性的缺陷，研究可靠性混合群智能优化方法，提出微粒群-蚁群混合群智能优化算法，进一步提高优化性能和适用性。结合理论、仿真及实验研究，建立多态复杂液压系统可靠性分析及优化新方法，为工程应用提供支撑。

实际工程中液压系统具有的多态性、复杂性、不确定性等特征及用户对液压系统日益增长的高可靠性需求，使得现有的可靠性分析、优化方法在适用性和精准性方面面临了更大的挑战。本项目针对多态复杂液压系统，提出了基于贝叶斯网络(BN)凸模型和混合群智能的可靠性分析及优化方法。首先，针对多态复杂系统建模和故障概率不确定问题，提出了贝叶斯网络(BN)凸模型可靠性分析方法，建立了重要度和灵敏度指标公式，结合资源约束要求和可靠度指标构建出符合工程实际的可靠性优化模型；其次，针对微粒群单种引斥力和拓扑结构的不足，提出了多作用力微粒群算法，模拟生物演化机制，提出了节点生灭和边变化的动态拓扑多作用力微粒群算法，提高了全局搜索和寻优精度；然后，针对微粒群、蚁群因单种生物种群演化局限性而存在适用性和优化性能的缺陷，研究可靠性混合群智能优化方法，提出了融合微粒群多作用力规则和蚁群自适应信息素更新机制的混合群智能优化算法，进一步提高了优化性能和适用性。最后，基于电液伺服技术设计了12通道液压软管总成脉冲试验台，基于双气液泵复合增压技术设计了软管耐压爆破试验台。基于试验数据进行失效分布拟合优度检验及分布鉴别，得到脉冲、爆破试验数据分别服从对数正态分布、威布尔分布的结论。综合理论、仿真及实验研究，建立了更贴近工程实际、可靠性优化精度更高多态复杂液压系统可靠性分析及优化新方法，有力地提高了工程液压系统可靠性分析和优化水平。