Fault diagnosis of rolling element bearing by using multinomial logistic regression and wavelet packet transform

Bearing fault identification of three-phase induction motors bases on two current sensor strategy

“…Hence, as p = 0.016 for a significance level α = 0.05, it is concluded that MLP is more accurate than the RF method. Table 9 presents a comparative summary of the results of this study with recent research in the literature (Zarei et al 2014;Prieto et al 2013;Vakharia et al 2015;Godoy et al 2014;Pandya et al 2014). This table shows that recent studies of the identification of bearing faults address issues of power quality and also variations in load torque.…”

“…As the MLP algorithm has been proven to be an encouraging tool for bearing fault diagnosis in electrical motors (Zarei et al 2014;Prieto et al 2013;Vakharia et al 2015;Godoy et al 2014;Pandya et al 2014), this paper uses this technique to identify bearing faults in three-phase induction motors using two current sensors and signals in the time domain.…”

“…Four types of faults were recreated in a controlled manner, such as an outer race fault (Zarei et al 2014;Prieto et al 2013;Vakharia et al 2015;Pandya et al 2014;Palácios et al 2015), inner race fault (Zarei et al 2014;Prieto et al 2013;Vakharia et al 2015;Pandya et al 2014), short circuit and excessive wear (Prieto et al 2013).…”

“…The work of Pandya et al (2014) focuses primarily on the selection of real wavelet for feature extraction, using wavelet packet transform (WPT)-based on maximum energy/entropy ratio criteria-in order to select better fault classifier techniques. The multinomial logistic regression (MLR) classifier is found to be the most effective technique, as it is 100 % accurate when compared to an MLP ANN, or sequential minimal optimization (SMO), using the latest SVM algorithm.…”

Research on a Signal Analysis Method based on Wavelet Theory and Approximate Entropy Algorithm

“…Fault feature extraction is the key to fault diagnosis, which is related to the accuracy of fault diagnosis and the reliability of early prediction. So people analyze the signal form various angles including the time domain, frequency domain and time-frequency domain in order to propose a lot of feature parameters and achieve a certain effect [2][3][4].…”