چكيده به لاتين
Punctual and reliable diagnosis, especially in key and vital industries with high performance, have a crucial importance. Rotational machines, and among them brushless DC motors (BLDCM), have an essential role in the industry. As a result, diagnosis approaches for the detection of defects and increase of reliability have a high degree of importance in these machines in stationary and non-stationary operation conditions. The goal of this research is to diagnose BLDCM by using micro electro-mechanical systems (MEMS) low-cost acceleration sensors. The main subjects of this research, considering the setup that has been designed and made, are the most prevalent causes of faults and machinery breakdown in these machines such as bearing inner, outer and inner-outer race and ball bearing defects and rotor unbalance. Fault diagnose was first performed by using MEMS based on vibration analysis and then by combining vibration- current analysis under stationary and non-stationary conditions under variable loads in no-load, half-load, and full-load conditions. In the following study, by using signal-processing techniques in the time-frequency domain and use of discrete wavelet transform (DWT), Wigner-Ville distribution (WVD), and Choi-Williams distribution (CWD) methods, feature extraction has been done on a primary dataset. Thereafter, dimensionality reduction is performed on the dataset, using the principal component analysis (PCA) algorithm. Then, by using target vector and the reduced dataset, the multi-layer perceptron (MLP) neural network is modeled to classify and diagnose the faults in the motor's health condition. The performance of the network in diagnosis and identification of the flaws is more than 98%. Due to the result of this study, the presented approach is reliable.
