基于多域特征的海流机故障诊断方法研究

As a clean and renewable energy source, marine current power generation has become an important supplement to the existing power generation methods. Therefore, it is extremely important to ensure the reliability and efficiency of marine current turbine system. Due to the harsh marine environment such as cavitation, surge turbulence, biofouling and corrosion, the existing research methods fail to efficiently solve the problems like the variety of fault features under various ocean current velocities, the inconspicuous of fault features under strong interference and the fault coupling. This project intends to investigate the diagnosis methods for marine current turbine system, which are based on diversity measure in statistics and combinatorial reasoning in knowledge discovery，by using working conditions’ partition and multi-domain projection, and by combining multi-domain features. The research contents include the following steps. 1) Design the partition of the working conditions of marine current turbine system and build multi-condition based fault diagnosis framework. 2) Establish a series of multi-domain feature extraction methods and fault diagnosis methods corresponding to each domain features. 3) Establish a set of online diagnosis methods of marine current turbine system and multi-domain diagnosis combinatorial reasoning methods. 4) Build the evaluation mechanism of the fault diagnosis methods. The key scientific issues are the difference measurement of maritime current, the multi-domain feature extraction, the perfect feature representation and the confidence selection of combinatorial reasoning. This research will provide practical research approaches and theoretical basis on fault diagnosis methods of marine current turbine system, and give some idea to fault diagnosis methods of other deep-sea equipment.

海流发电作为一种清洁、可再生能源发电，成为现有发电方式的重要补充。因此，保证海流机可靠而有效地运行极为重要。由于受气蚀、浪涌、附着物、腐蚀等海洋复杂环境的影响，现有研究方法用于解决不同海流流速下故障特征多样、强干扰下故障特征不明显、故障耦合等问题效果不佳。本项目以统计方法中的差异性度量、知识发现方法中的组合推理等理论为基础，以工况划分、多域投影为手段，结合多域特征开展海流机多工况多域诊断组合推理的诊断方法研究：设计海流机工况划分方法并构建多工况故障诊断框架；建立多域特征提取方法以及适合各域特征的故障诊断方法；建立工况在线辨识方法并提出多域诊断组合推理方法；建立海流机故障诊断方法的评价机制。重点解决所遇到的海流流速差异性度量、多域特征提取、完美特征表示、组合推理的置信度选取等关键科学问题。本项目的研究将为海流机故障诊断提供研究途径和理论依据，也可为其他深海装备的故障诊断方法研究提供借鉴。

海流发电作为一种清洁、可再生能源发电，成为现有发电方式的重要补充。由于受气蚀、浪涌、附着物、腐蚀等海洋复杂环境的影响，其维修维护成本非常高，制约了海流发电的发展，为此我国在《能源技术革命创新行动计划（2016~2030年）》中强调“突破发电机组水下密封等关键技术，研究海上机组的新型状态监测系统装备技术及智能故障预估的维护技术”。因此，有效的诊断技术对于保证海流机安全高效的运行以及减少维护费用等方面有着非常重要的实际意义。本项目针对海流机这一类深海机电一体化装备故障诊断中存在的不同流速下故障特征差异大、故障特征不明显、随机干扰与故障耦合等问题，建立了海流流速与海流机工况的模型、提出了多种特征提取方法、多域特征表示和诊断方法，提高了海流机故障诊断的精度。重点研究了海流机多工况分区的标准及相应特征描述，综述了海流机的故障类型、产生机理、发生方式等，建立一套针对海流机定子电流、定子电压、水下摄像头、水下振动信号、水听器等多种传感器在不同流速、随机浪涌湍流等各种工况下的海流机多域故障特征库，提出一系列故障检测动态限的自适应设置方案，构建了海流机不同工况的故障诊断框架，提出了一系列基于多域特征、多域诊断组合推理的故障诊断策略，还搭建了海流机远程监控系统实验平台。在海流机叶片冲击故障、叶片附着物识别、海流发电机绝缘系统故障、逆变系统故障等方面提出了一系列的解决方案。为海流机故障诊断的应用奠定坚实的理论基础以及关键技术储备，也为其他深海装备的故障诊断提供解决方案。本项目的研究成果已发表学术论文50篇，其中SCI检索21篇，EI检索21篇，ESI高被引论文1篇；申请27项中国发明专利、2项美国专利；其中授权15项中国发明专利、授权1项美国专利，出版学术专著3本。培养博士生4名，硕士生20名。王天真负责开发的海流机“多电平逆变器嵌入式故障监控系统”2018年获第二十届中国国际工业博览会高校展区优秀展品奖一等奖。超额完成各项指标。