Diagnosing faulty gearbox using pre-processing (PCA and neural network analysis).

The fault diagnosis of urban transit gearboxes has the characteristics of complex vibration signals and large amounts of data. The daily scheduled maintenance cannot meet the needs of gearbox maintenance.The purpose of this research is to diagnose and ensure the health of the gearbox or gear box with the help of PCA pre-processing and neural network analysis. First, the mentioned gear box is installed on the test table, then by defining the test conditions in the ۳rd gear and the half-load mode with a speed of ۲۵۰۰ rpm, by means of the test table equipment, the vibration signals related to two healthy and defective states are extracted. And then the vibration signals from ۰ to ۵۰۰ seconds are taken every ۵۰ seconds and then based on the distance or height of the three primary waves in each time interval, the data related to the input in the neural network is pre-processed and mostly the excel file to the network neural network is given, and then the pattern recognition operation is performed by SVR, multi-layer-perceptron, perceptron neural networks, and then the most optimal possible mode is selected among them.