非线性系统故障可诊断性量化评价与优化设计方法研究

The research shows that the low fault diagnosability is one of the major reasons of high system failure rate, namely the measurable information cannot support the fast and reliable system fault detection. As a result, conducting diagnosability quantitative evaluation and optimization design toward fault of nonlinear system is an effective way to solve above problem. Based on the research basis of fault diagnosability quantitative evaluation and dynamic system sensor optimization, this project overcomes the uncertainties of nonlinear system, and employs the effective methods of K-L divergence as basics and the effective methods of maximum entropy density function estimation and adaptive threshold as supplement so as to realize a relatively accurate fault diagnosability quantitative evaluation. And then, based on system characteristics, the methods of dynamic programming and soft sensor are employed to design measured points and so as to solve the problem of insufficient measured points information which restrict the fault diagnosability, to put the after-system affair fault diagnosability before the system design stage and to enable the system has relatively higher diagnosability during design stage. In addition, the improved NSGA-II algorithm is to be employed to conduct optimal allocation with the core of quantitative index and support of stability, economy and other goals toward the sensor so as to realize the purpose of stable, high-effective and saving operation of nonlinear system, thus providing new ideal and space for theory and application research of fault diagnosability of nonlinear system.

研究表明，故障可诊断性低是导致系统故障率居高不下的主要原因之一，而可测信息不足则是难以支撑迅速、可靠检测系统故障的本征所在。因此，针对非线性系统，从故障可诊断性量化评价着眼，并进行优化提升设计是解决上述问题的有效途径。本项目结合故障可诊断性量化评价和动态系统传感器优化配置的研究基础，为克服非线性系统不确定因素的影响，以K-L散度为基础，最大熵密度函数估计和自适应阈值为补充的有效方法，实现较为准确的故障可诊断性量化评价。基于此，切实考虑系统特性，运用动态规划和软测量的方法对测点进行设计，解决测点信息不足这一制约故障可诊断性的症结，并将故障可诊断性提升这一后系统事件前置于系统设计阶段。还拟对传感器采用改进的NSGA-II算法进行以量化指标为主，兼顾可靠性、经济性等多目标的优化配置，实现非线性系统可靠、高效、节约运行的目标。项目的实施可为非线性故障可诊断性研究的理论与应用拓展新的思路与空间。

在现代工程系统中，如何采用有效手段提高系统的故障可诊断性和故障预测的准确率，从而降低事故风险是一个迫切需要解决的问题，特别是在非线性系统，由于其构造的复杂性和噪声、外部扰动等不确定因素的存在，始终是开展相关研究所面临的难题。.因此，本项目旨在提高系统可诊断性和故障预测效率，针对非线性系统，从故障可诊断性的量化评价方法、故障可诊断性提升设计方案、基于故障可诊断性量化评价的传感器优化配置等方面开展研究。通过对制约故障可诊断性的关键因素进行分析，建立了基于K-L散度思想的系统故障可诊断性量化评价框架，利用统计方法构建自适应阈值以提升评价的准确性。根据量化评价结果，在动态规划法的基础上构建了增设最优测点的解法，形成故障可诊断性提升设计方案，解决了系统在设计阶段即可着手进行诊断性设计的操作性问题。改进贪心算法的测点传感器配置方法，结合软传感器技术和基于故障特征的可分离性设计方法，研制集节约系统成本和简化设计复杂度的多目标优化配置模型，为切实提高系统的故障诊断能力，进而使得系统具有更高的安全性、可靠性和经济性提供了新的思路和技术支撑。.区别于系统的故障可诊断性设计作为后事件分析的非线性系统故障诊断方法，本课题另辟蹊径，从诊断性设计前移至系统设计之初，提出对可诊断性量化评价方法，并考虑结构、成本等约束进行系统可诊断性优化设计。改进的最大熵概率密度函数估计和自适应阈值技术对于解决在不确定性因素影响下难以准确进行故障可诊断性评价是一突破。从量化角度改进NSGA-II算法使传感器多目标优化配置模型更为高效，为现实中高品质低成本的系统运行提供可借鉴的方法和参考依据。本项目已实施应用于车辆电源的实际运用中，取得了一些理论和技术成果，对于促进现代工程稳定健康发展具有重要的科学意义和现实意义。

内容获取失败，请点击重试