Research On Fault Detection Technology And Application System Based On Principal Component Analysis

The function of fault detection is required to improve the safety and reliability of complex equipment system. Thus it is theoretically significant and practically valuable to the following work, including logical functions and architecture of fault detection system study, fault detection algorithm developing,and equipment fault detection platform design.On the basis of characteristics of the equipment’s running status data, this thesis studied principal components analysis method which can eliminate data dependency and redundancy, and designed a fault detection algorithm based on kernel principal components analysis. In order to reduce the false alarm rate,wavelet denoising was added to the fault detection algorithm based on kernel principal components analysis. Thus a fault detection algorithm based on wavelet denoising and kernel principal components analysis was designed. The algorithms were tested by using the simulation data of the Tennessee Eastman process. And the results show that the fault detection algorithm based on principal components analysis has a poor performance and high false alarm rate to handle same nonlinear problem, while the fault detection algorithm based on kernel principal components analysis works well. Further, it can be found that the fault detection algorithm based on wavelet denoising and kernel principal components analysis has a lower false alarm rate.Then a system of data-collection for a kind of typical PLC-controlled hydraulic press was designed. To do this, we assumed that such hydraulic press’ s running status data has been collected by the sensors equipped in PLC-controlled system. And the data were transferred to the data-collection platform by serial port, then the fault detection program running on the platform raised the alarm in necessity, meanwhile the data were transferred to a extended system by short distance wireless channel. For convenience, the platform was developed using embedded technology, and the fault detection subsystem was designed based on TQ2440 development board. The experimental results show effectiveness of design.