船舶电力推进系统状态监测与故障诊断的信息融合方法

This project concentrates on national vital demand for operational security of high tech marine engineering equipment and ultimate goal “from ‘made in Zhejiang’ to ‘Zhejiang create’” formulated in policy of the deep integration of information and industrialization. In condition monitoring and fault diagnosis of high tech marine electric propulsion system, in order to deal with uncertainties in fault states, fault features, mapping relation between features and faults caused by complexities of work environments and running states, we study on information fusion theory and key technology for high-end equipment condition monitoring and fault diagnosis. These research contents are suitable for all stages of running life cycle including optimal allocation of monitoring variables, dynamic mining and extraction of fault feature from monitoring data, fault belief functions acquisition from multi-source fault features, fusing fault belief functions in spatio-temporal domain and diagnosing incipient faults, fault state prediction based on belief data and expert’s knowledge. The research objective is to improve accuracy and reliability of fault decision-making under uncertain environments by fault feature information fusion in spatio-temporal domain, so as to provide the novel theoretical basis and key technical support for enhancing intelligent self-diagnosis level, which are universal and different from traditional framework or methodology. In the course of research, a talent team will be built.

围绕国家对高技术船舶工程装备运行安全性技术的重大需求，结合两化深度融合中从“浙江制造”到“浙江创造”的终极目标，针对高技术船舶电力推进系统状态监测与故障诊断中，由于工作环境及运行状态复杂性所导致的故障状态、故障特征、特征与故障之间映射关系中的不确定性问题，开展高端装备运行状态监测与故障诊断信息融合理论与关键技术研究。主要包括：与运行全生命周期各阶段相适应的监测变量优化配置、面向海量监测数据的故障特征动态挖掘与提取、基于多源故障特征的故障信度函数获取、多源故障信度函数的“时空域”融合与微小故障诊断、基于故障信度数据与专家知识的故障状态预测等。以期通过“时空域”多源故障特征信息的融合过程，增强不确定环境下故障决策的精度和可靠性，为提升高端装备的智能化自诊断水平提供理论依据和关键的技术支撑，取得一批具有普适性且有别于传统框架的故障诊断新理论与新方法。培养一支高水平的人才队伍。

1）项目背景：针对高技术装备安全运行方面的迫切需求，以船舶电力推进系统为对象，重点解决故障信息挖掘与综合利用、早期微小故障在线诊断、故障预测等关键科学问题，建立了一套面向故障决策的信息融合理论与方法。. 2）主要研究内容：监测变量优化配置与故障特征提取，基于多源故障特征的故障信度获取，基于故障信度融合的故障诊断，基于置信规则库推理的故障预测。. 3）重要成果：. （1）电推系统运行全命阶段划分及监测变量优化配置：基于信度区间的监测变量约减与优化配置，基于深度学习网络的监测变量故障特征提取，基于信度多属性推理决策的系统全生命周期评价与可靠性分析。. （2）基于证据推理（ER）与置信规则库（BRB）推理的分类器建模。基于证据推理规则和粗糙集的分类器设计，基于属性向量化与证据融合的分类器设计，ER模型/BRB模型参数和结构的联合优化方法，基于向量式置信规则库（BRB）推理的非线性因果关系建模方法。. （3）基于证据推理与证据更新的微小故障（报警）检测：证据更新滤波器在多阶上的推广及报警器设计，基于置信规则推理的多阶证据更新滤波报警器设计，基于多源证据可靠性度量的多变量报警器设计。.（4）基于证据推理（故障信度融合）的故障诊断方法：基于故障特征向量化与ER融合的故障诊断方法，完备/不完备故障特征样本条件下的ER融合故障诊断方法，基于ER规则的多分类器融合故障诊断方法。.（5）基于置信规则库(BRB)优化模型的故障预测方法：基于主导从属框架的BRB多目标优化方法，基于粒子滤波的BRB动态更新方法，基于动态BRB和ER规则的故障预测方法。. 在项目组研制的电推系统试验台和仿真平台上，对所研究的信息融合故障诊断与故障预测算法进行实验验证，并搭建电力推进系统安全监控平台，对相关算法及软件的效果进行应用验证。. 项目组共发表SCI检索的学术论文59篇，出版学术专著2本，授权发明专利35件，培养博士生7名、硕士生30余名。