基于数据的故障模式识别方法研究

Due to hard obtained physical models from the first principal, especially for modern complex industrial processes, the recent research on data based abnormality pattern recognition, diagnosis and approaches for safety operation have received considerable attention both from academic and practical points of view. Motivated by these observations, our major objective is to develop an integrated scheme for safety operation of complex industrial processes, including directly design key parameters for fault pattern recognition only relying on available process measurements and based on it, further integration of data based fault tolerant, fault prognosis as well as optimal maintenance strategy in order to ensure the overall safety, reliability and economic performance of the underlying processes. The proposed techniques will be firstly evaluated on experimental-scale setups and finally applied on real systems to show their effectiveness and superior performance, which might make potential contributions on efficient yet reliable operation of modern industrial processes.

针对现代复杂系统机理模型难以获取的情况，本课题主要研究基于可测数据的故障模式识别、诊断及系统安全运行新方法。在不依靠机理模型的情况下，利用海量历史数据及在线信息，直接完成对故障模式识别系统关键参数的设计，并以此为基础，进一步探讨针对典型非正常工况在线容错策略研究。以故障模式识别为基础，建立非正常工况预报体系亦是本项目研究的重点内容之一，这将为关键设备适时维护和系统主动安全策略的建立，提供重要的理论基础。本项目采用基础理论分析、实验室平台仿真验证与工业实际应用相结合的方式，验证所提出理论方法的有效性，并力争在数据意义下，为故障模式识别理论及其在复杂系统安全运行应用方面做出贡献。

本项目针对现代复杂系统机理模型难以获取的情况主要研究了基于可测数据的故障模式识别、诊断及系统安全运行的新方法。首先，在不依靠机理模型的情况下，利用海量历史数据及在线信息，直接完成对故障模式识别系统关键参数的设计；其次，以此为基础，进一步探讨了针对典型非正常工况在线容错策略研究。最终以故障模式识别为基础，建立了非正常工况预报体系，并为关键设备适时维护和系统主动安全策略的建立，提供重要的理论基础。本项目采用基础理论分析、实验室平台仿真验证与工业实际应用相结合的方式，验证了所提出理论方法的有效性。在数据意义下，为故障模式识别理论及其在复杂系统安全运行应用方面做出贡献。