基于贝叶斯模型平均的重型数控机床可靠性精细建模与综合评估

Reliability modeling and assessment has long been a critical method for the achievement of high reliability of heavy-duty numerical control (NC) machine tool, which has also been recognized as one of the key generic technologies that are indispensable for the breakthrough of key sectors highlighted by “Made in China 2025” strategy. The heavy-duty NC machine tools are characterized with enormous system structure, complicated drive control, onerous operating missions, and multiple influencing factors. These characteristics have made the reliability assessment of heavy-duty NC machine tool challenged by the difficulties of the partially relevant data modeling, heterogeneous reliability model coexisting, and multiple assessment results synthesizing. To overcome these difficulties, this proposal introduces a new type of reliability assessment technology under the background of investigating one type of heavy-duty machine tools. The research is designed to be based on the deep investigation of reliability modeling of failure time data under multiple complex factors interacting. Various models and methods, including covariate model and random effect model, Bayesian model averaging method, and stochastic simulation based Bayesian inference method, are implemented and incorporated into the investigation of reliability assessment of heavy-duty NC machine tool. Major efforts are laid on the following research contents: (1) Construct a new type of Bayesian model averaging based reliability assessment framework for complex systems under multiple factors interacting; (2) Introduce a refined reliability modeling method by considering the effect of partially relevant information and heterogeneous maintenance methods; and (3) Develop a reliability assessment by leveraging stochastic simulation method for the implementing of Bayesian model averaging. Finally, based on the framework and methods introduced above, a coherent methodology for reliability assessment of heavy-duty NC machine tool is constructed. This methodology can be further served as the fundamental technique for quality improvement and technique innovation of heavy-duty NC machine tool.

可靠性评估是支撑重型数控机床高可靠性达成的关键基础方法，也是助力高端装备行业制造能力提升的重要共性技术。重型数控机床由于体积结构大、任务载荷重和影响因素多等特性，使得其可靠性评估受到故障信息部分相关、多类模型表征复杂、评估结果集成融合等共性问题的挑战。本项目以某重型数控机床为研究背景，以故障时间数据的建模分析为研究切入点，将协变量和随机影响模型、贝叶斯模型平均方法以及贝叶斯推理的随机模拟方法引入到可靠性评估中，重点研究重型数控机床可靠性评估的贝叶斯多模型综合评估框架、计及故障关联程度和维修方式混合的可靠性精细建模方法、基于随机模拟方法的贝叶斯多模型综合评估方法。在此基础上，致力取得可靠性评估方法在评估框架、精细建模和综合评估等方面的突破，构建贝叶斯多模型综合的可靠性评估新方法，为重型数控机床性能水平的提升和共性问题的突破提供方法支撑。

重型数控机床可靠性评估作为关键的可靠性共性基础方法，是对重型数控机床进行定量控制、研制考核和健康管理的关键所在；同时也是为机床设计的改进、工艺的优化和使用的保证提供信息和方法支撑的重要保障。本项目围绕重型数控机床可靠性评估中面临的故障信息部分相关、数据模型复杂多样、评估结果集成融合等共性问题，系统地研究了重型数控机床可靠性精细建模与综合评估方法：1）研究了复杂耦合因素定性识别和定量描述的分析框架、考虑故障信息部分相关和维修方式随机混合的建模框架，构建了重型数控机床可靠性评估的贝叶斯多模型综合评估框架；2）研究了似然函数权重因子在故障数据建模中的引入、不同维修情形下故障序列建模的扩展，提出了同时计及故障关联程度和维修方式混合的可靠性精细建模方法；3）研究了多种模型的贝叶斯参数估计与贝叶斯模型平均的整合、多模型评估结果的集成融合，提出了基于随机模拟方法的贝叶斯多模型综合可靠性评估方法。项目发表英文专著1章节，SCI收录国际期刊论文7篇，并有1篇论文入选ESI“高被引论文”（ESI Highly Cited Paper）。本项目从理论研究的角度来看，是对复杂系统可靠性评估关键共性问题的深入研究，可为复杂系统可靠性评估理论方法的拓展提供关键方法的补充；从实际应用的角度来看，是对重型数控机床可靠性评估关键问题的专题突破，可为重型数控机床可靠性的量化控制、研制考核和健康管理提供基础方法的支撑。