【作者】 张静；

【导师】 姜斌；

【作者基本信息】 南京航空航天大学 ， 控制理论与控制工程， 2013， 硕士

【摘要】 高频电源以其体积小、效率高、功耗低的特点逐步取代工频电源应用到静电除尘器中，效力于节能减排工作。但是，由于高频电源电路集成度、复杂度高，使其在运行过程中由一个小的故障而引发灾难性的事故，因此，能够在故障发生早期检测出来并作相应处理是保证系统安全、高效运行的必要条件。本文针对高频电源故障提出两种故障检测方法：其一，基于自适应PCA的高频电源故障检测方法。在分析高频电源相关参数——母线电压、高频高压变压器相电流、原边谐振电流、副边电流以及副边电压的基础上，考虑到这些参数的变化以及各参数相互之间的影响，基于传统主元分析理论，提出了自适应PCA的故障检测新方法。首先，建立正常工况下的历史数据矩阵，对实时采集的无故障数据进行相似度计算，决定模型的局部更新，结合历史数据和更新数据进行协方差矩阵的权重分配，采用改进的迭代算法建立新的主元模型，最后使用SPE和T2结合的统计量进行实时故障检测。其二，基于相似度匹配的故障检测算法。选取高频电源原边谐振电流为研究变量，对具有噪声的波形进行均值滤波，结合灰度投影算法粗匹配和混合相似度精匹配来判断待测波形是否与库波形匹配成功，若成功则表示无故障，反之，则表示故障发生，作报警处理。最后，在MATLAB中对上述两种方法进行了仿真，验证了方法的可行性和正确性。进一步，为验证所提方法的实际工程应用性，将上述提出的基于自适应PCA的故障检测方法在嵌入式平台上进行实时仿真验证，平台采用基于Linux的ARM嵌入式系统。首先，对实验平台的硬件部分进行设计及搭建，并在硬件基础上配置了相关的软件开发环境，完成蜂鸣器的驱动程序编写，开发相应的上位机和下位机应用程序。最后，在所搭建的平台上，对现场采集的数据进行模拟的实时故障检测实验。实验结果表明，本文所提出方法具有一定的工程应用性，设计的ARM嵌入式平台具备良好的故障检测能力和人机交互性。更多还原

【Abstract】 The high-frequency power supply gradually replaces the industrial frequency power supply appliedin the ESP for its small size, high efficiency and low power consumption; it can work on energyconservation and emission reduction well. However, due to high circuit integration and complexity,and easily a small fault will lead to collapse of the whole system while in the process of the running ofthe high-frequency power supply, the early detection of the fault and the corresponding treatment arethe necessary conditions to ensure the safety and high efficiency of the system.In this paper, two fault detection methods are proposed for the high-frequency power supply. Onemethod is based on adaptive-PCA which focused on parameters variation and the mutual effectsbetween each other. The parameters include bus voltage, phase current, primary current, secondarycurrent and secondary voltage. Based on the traditional PCA, the real-time model updating isintroduced. It’s first to establish the historical data matrix of normal operating conditions anddetermine the updated data by calculating the similarity of the real-time data with no fault. Then, theimproved iterative algorithm is used to create the real-time PCA model by the combination of thehistorical data and the updated data. Last, the combination of the SPE and T2is applied to achievefault detection. The other method is based on the similarity matching. The primary current ofhigh-frequency power supply is selected to be the study variable and the mean filter is used for thewaveform denoising, and then the similarity between the test data and the library data is calculated bygray-scale projection combined with mix-similarity algorithm. If the match is successful, that is to sayno fault occurs. The simulation of the data collected at the scene in Matlab environment hasdemonstrated the feasibility, accuracy and rapidity of the proposed methods.In order to verify the engineering application value of the proposed method, the ARM embeddedLinux-based system is built to realize the fault detection system of the high-frequency power supplybased on the adaptive-PCA method. Through setting up a hardware platform, configuring an integralsoftware development environment, completing the programming of the buzzer driver and thenprogramming the upper and the lower machine, and at last, simulating the online fault detection of thedata collected at the scene, the experiment result indicates that the method proposed has a certainengineering application and the embedded system has good fault detection ability and interaction.更多还原