线性压缩机运行状态在线监测及故障诊断方法研究

Linear compressors are receiving a great deal of attention for high efficiency, compacted structure and a variable cooling capacity. However, for its unique free-piston structure, the linear compressor is hypersensitive to the gas load changing in refrigeration system. The dynamic characteristics will be apt to cause a temporary failure. The temporary failure will develop a serious breakdown if the voltage exerted on the linear motor is not adjusted in time. Thus a system which can online monitor and diagnose the failure is needed for the linear compressor efficiency operating. The fault characteristics of the characteristic parameters current used may not obvious in some operating condition. Besides a comprehensive relationship between the breakdown characteristics and failure still requires more study. As the kinetic parameters of the compressor directly responses to the operational status and it is highly sensitive to the failure，However there is no feasible method to calculate the stiffness and damping of the compressor as it directs at a characteristic of time-varying and nonlinear. This project proposes a method for the stiffness and damping calculation in frequency domain, and provides new characteristic parameters including the stiffness and damping of the compressor and the time-frequency characteristic of current for fault diagnosis. The wavelet decomposition and reconstruction is used to extract fault characteristics. By collecting and analyzing the stiffness, damping, time-frequency characteristic extracted from current and etc. in theoretical and experimental way, the fault mode library will be established. And then the fault mechanism and the causes lead to such failure characteristics are discussed. The mathematics model of identifying the failure of the linear compressor is established using neural networks and pattern recognition technology. Based on artificial neural networks theory, the operational status of the linear compressor are monitored intelligently and the failures are diagnosed via collecting and processing voltage and current signals to pick up characteristic parameters. Therefore, the research will contribute to a deep understand the fault process and to identify the failure which will enrich the control strategy of the linear compressor.

线性压缩机采用自由活塞式结构设计而具有高效节能的优势，但也使得活塞运动特性对环境参数变化敏感，从而易发生暂时性故障。为防止暂时性故障损坏压缩机，需要对压缩机运行状态进行在线监测。目前线性压缩机运行状态监测对象的特征提取通用性不强，故障特征与故障对应关系不明确，严重制约了线性压缩机产业化应用。动力学参数刚度和阻尼能直接反映压缩机运行状态,对故障敏感，但由于其非线性时变的特点，目前尚无准确的计算方法。本课题提出了一种线性压缩机刚度与阻尼的识别方法，并将所识别参数与小波分析提取的电流时频特征作为故障诊断新的分析对象，通过理论分析与实验研究，建立刚度及阻尼等动力学参数与电流时频特征的故障特征库，构建基于模式识别的故障诊断方法。通过本课题的研究，从理论上揭示线性压缩机故障动态演化机理及故障特征成因，完善线性压缩机故障特征提取的理论和方法；同时为线性压缩机运行状态在线监测及故障诊断提供一种有效的手段。

线性压缩机采用直线电机驱动，具有效率高、结构紧凑、体积小的特点，已成为小型制冷装置用高效压缩机的一个重要发展方向。但自由活塞式的结构设计也使得活塞的运动对系统参数变化敏感，易发生暂时性故障。为防止暂时性故障损坏压缩机，需要对压缩机运行状态进行实时监测。但由于现有的线性压缩机运行状态监测对象的特征提取通用性不强，故障特征与故障对应关系不明确，严重制约了线性压缩机的产业化应用。本项目提出了一种线性压缩机刚度与阻尼的识别方法，通过对动力学参数特征进行实时在线提取，作为压缩机故障诊断的分析对象，通过理论分析与实验研究，建立了刚度及阻尼等动力学参数的故障特征库，构建了基于模式识别的故障诊断方法。.(1)通过建立基于矢量模型的线性压缩机动力学模型，系统且全面地研究了线性压缩机动力学参数的在线提取方法，为压缩机故障特征的提取奠定了基础。.(2)在线性压缩机动力学参数在线提取的基础上，对常见故障状态时的动力学特征进行了理论和实验研究，揭示了线性压缩机故障动态演化机理及故障特征成因，完善了线性压缩机故障特征提取的理论和方法。.(3)根据故障特征监测数据建立了线性压缩机故障特征库，并对故障特征进行了建模，通过神经网络自主学习对压缩机动力学参数的计算以及运行状态进行判断，为压缩机的安全运行和故障识别提供了参考。.(4)项目实施过程中，参与本项目的硕士研究生毕业2人，另有1名硕士研究生在读。.(5)发表高水平研究论文11篇，其中SCI/EI收录5篇；申请发明专利4项，已授权1项。.(6)本项目的研究成果作为核心技术在中国科学院科技服务网络计划（STS计划）“电冰箱用直线压缩机产业化研究”（2018～2019）、企业合作横向项目“直线压缩机的冰箱匹配技术研究”（ 2018～2020）等相关项目中得到了推广应用。