数控装备运行可靠性的多源时变耦合建模与实时评估研究

Field reliability assessment and prediction for complex numerical control (NC) equipment is the key to promote core competition of manufacturing industry. It is also a hot topic in the area of reliability engineering. Modern NC equipment is characterized as high complex system structure and subjected to time-dynamic working conditions. As a result, classical methods for field reliability assessment and prediction are challenged with the issues of information integrating, time-dynamic modeling, and real-time computing. On the background of field reliability assessment and prediction for a type of high-grade CNC machine tool, this proposal aims to advance the state of the art of field reliability assessment and prediction for complex NC equipment. It is based on the in-depth investigation of degradation modeling and condition monitoring. The methods of multiple degradation modeling with random effects and covariates, methods of multiple information integration, and methods of approximate Bayesian computation are incorporated coherently into the methods of field reliability assessment and prediction for NC equipment. Special focuses are laid on the following topics: (1) Develop an integrated framework for field reliability assessment and prediction to take multiple information integration and time-dynamic working conditions into account; (2) Investigate a comprehensive modeling method for the dealing with multiple source information and the consideration of individual difference and time-dynamic covariates; (3) Develop a real-time algorithm for reliability assessment and prediction by modifying the approximate Bayesian computation method. Finally, based on the techniques and methods developed above, a systematic methodology for real-time field reliability assessment and prediction for NC equipment is constructed. It can serve as the support for the improvement of field reliability of NC equipment, for the cutting down of total cost of ownership for the users of NC equipment, and ultimately for the continual promotion of self-innovation capability and core competitiveness of manufacturing industry.

对数控装备这类复杂的机电产品进行使用可靠性评估，是我国装备制造业核心竞争力提升的关键因素，也是可靠性工程领域的研究热点之一。数控装备复杂的系统结构和时变的工况应力使得目前的使用可靠性评估方法受到多源信息融合、时变耦合建模和实时评估算法的挑战。本项目以某型号高档数控机床这一典型数控装备为工程背景，以性能退化建模和基于状态监控信息的使用可靠性评估研究为基础，将多退化、协变量和随机影响建模方法、多源信息融合方法和近似贝叶斯计算方法引入数控装备使用可靠性评估中，深入研究数控装备考虑多时变因素的多源信息融合评估框架、融合退化试验、状态监测和任务工况信息的使用可靠性建模、基于近似贝叶斯计算方法的实时评估算法。在此基础上，建立系统的、实时的装备运行可靠性评估方法，为数控装备使用可靠性的提高、运行使用费用的降低、自主创新能力和核心竞争力的提升提供技术支持。

高档数控机床的实时可靠性评估和预测对于保证其长期稳定的运行和实施健康管理至关重要。针对这类产品的现场可靠性评估和预测的经典方法面临着昂贵的可靠性测试，小样本量和个体非均匀性的问题。退化分析是对长寿命和高可靠性产品进行可靠性分析的常用方法。然而，对于新开发的产品，尤其是针对样本量较小的高度定制产品而言，个体差异性及少量退化观测值的挑战仍然是一个值得深入研究的问题。本项目针对数控机床典型个体差异性特征，运用伽马过程对其性能退化过程进行可靠性建模，以贝叶斯方法对模型进行数据更新和计算，并用实测性能数据对模型及算法进行有效验证。考虑到小样本的特征，为提高评估结果的准确性，在上述性能退化模型基础上，结合了基于贝叶斯定理的信息融合技术，充分利用数控机床在寿命周期不同阶段不同来源的退化信息，提高了参数估计和退化分析的精度。通过与数控机床主机厂商的合作，初步验证了本方法的有效性。项目研究结果对高档数控机床的可靠性设计与健康管理提供了一定的技术支持。