混合不确定性与时变相关故障下的重型数控机床动态可靠性建模与评估

Heavy-duty CNC machine tools (HDMTs) are the basic equipment for critical parts processing in nations’ key industries, such as aerospace, military, power generation equipment, etc. The research on the generic technologies, breakthrough of key technologies and convergence of emerging technologies are the critical points to realize the self-dependent innovation of whole machine. Reliability modeling and evaluation technology is the key reliability basic generic technology of HDMTs. The challenges about aleatory and epistemic uncertainty (hybrid uncertainty), time-varying failure correlation, and the multi-source monitoring information still cannot be solved by the existing technology. Aiming at appropriately treating these issues to achieve a better and more accurate assessment for such a complex system, this proposal focus on the following important research topics. Firstly, build a framework for system reliability assessment under multiple influence factors. Secondly, develop system reliability modeling method with consideration of hybrid uncertainty and time-varying failure correlation by using evidence theory and dynamic Bayesian network, mainly focus on the unified quantification method of hybrid uncertainty considering variable correlation. Finally, study comprehensive reliability assessment methods to fuse multivariate information and multi-source information based on hierarchical mixed Bayes information fusion method. The research outcomes to be developed will provide a suit of practical tools for the high reliability and safety assurance of HDMTs, and also provide novel theoretical technical approaches to support improvement of complex system reliability assessment.

重型数控机床基础共性技术的研究、关键技术的突破及新兴技术的融合，是实现整机自主创新的关键所在。可靠性建模与评估技术是重型数控机床关键的可靠性基础共性技术，传统技术面临着混合不确定性量化、故障时变相关性分析、多源异类数据融合等难题。本项目基于混合层次贝叶斯多源信息融合技术，研究混合不确定性下与故障时变相关性下的可靠性建模与评估方法。首先构建多因素综合作用下的系统可靠性评估框架；在此框架下基于证据理论及动态贝叶斯网络，开展计及混合不确定性及故障相关性影响的系统可靠性建模研究，重点实现考虑变量相关性的混合不确定性统一量化方法的突破；最后基于动态Copula相关理论及混合层次贝叶斯信息融合方法，研究融合多变量信息及多源异类层次数据的系统综合可靠性评估方法及算法实现。本项目将为保障重型数控机床高可靠及安全运行提供理论技术支持，同时为复杂系统可靠性评估框架的完善提供理论依据及关键技术支撑。

针对重型数控机床可靠性研究面临的混合不确定性量化、故障相关性建模与分析、多源异类数据融合等问题，本项目重点研究融合混合不确定性与复杂失效特征的复杂系统可靠性建模与评估方法。首先构建了考虑多因素耦合作用下的复杂系统可靠性综合评估框架。在此框架下采用随机变量、证据变量或模糊变量等多变量形式对系统混合不确定性因素进行综合描述，基于概率盒与证据结构体相互转化规则，将精确概率分布、证据变量、随机数集、主观信息等转化为概率盒及证据结构体，实现了基于概率盒的多源不确定性(随机和认知不确定性)量化与统一。对于复杂失效特征，考虑相关失效中的共因失效、确定性和非确定性级联失效等情况，采用多因子参数模型、Copula理论、仿射算法等对相关失效关系进行建模，提出了多类系统可靠性评估模型。基于证据网络，提出了考虑多变量信息及混合不确定性下基于动态证据网络的复杂系统可靠性分析方法、基于证据贝叶斯网络层次模型的系统可靠性综合评估方法等。在此基础上，对现有重要度指标进行了改进，实现了考虑混合不确定性下的系统部件重要度分析；并对系统变量敏感性进行了分析，实现了系统输入不确定性对系统可靠性及不确定性影响的量化。最后，将本文所提框架，应用于重型数控机床关键子系统的可靠性分析与评估中。本项目的实施，为保障重型数控机床高可靠及安全运行提供理论技术支持，同时进一步完善了现有复杂系统可靠性评估框架。