PROJECT TITLE :

Multi-Switches Fault Diagnosis Based on Small Low Frequency Data for Voltage-Source Inverters of PMSM Drives

ABSTRACT:

Using small low-frequency data for inverter failure diagnosis of permanent magnet synchronous machine drives requires additional hardware and has a negative impact on diagnostic accuracy. There is a lot of superfluous work and corresponding system memory wasted, which raises the hardware requirements and implementation difficulties as well as affects practicability. A unique multi-switches fault diagnosis algorithm is given to improve this problem, to make the design, debugging, and implementation of fault diagnostic more convenient and at cheap cost, and further to increase the practicability of the algorithm. Small low-frequency data can be obtained from the controller's feedback signals using a second low-frequency processing method, as shown in Figure 1. The most important information about switch states can be found in these low-frequency data points. These low-frequency data are also effectively adjusted by the single extremum normalisation approach, which is based on the symmetry properties of various state data. The asymmetry of an inverter's fault state can be well preserved using these processed data. For the third step, the primary fault components and features are recovered from the processed low-frequency data's distortion and envelope change. It is then utilised in conjunction with the retrieved features to create an intelligent categorization system. The design of the structure of the hidden layer network is simplified, and the network training time is extremely quick, unlike the standard neural network. Since this network is so easy to troubleshoot, it is a great benefit. The suggested fault diagnosis algorithm is more straightforward and less expensive to implement for multi-switch fault diagnosis than the existing algorithms, and the usefulness of the fault diagnosis algorithm is demonstrated by an experiment.