Compressive sensing-based feature extraction for bearing fault diagnosis using a heuristic neural network

Abstract

Bearing fault diagnosis collects massive amounts of vibration data about a rotating machinery system, whose fault classification largely depends on feature extraction. Features reflecting bearing work states are directly extracted using time-frequency analysis of vibration signals, which leads to high dimensional feature data. To address the problem of feature dimension reduction, a compressive sensing-based feature extraction algorithm is developed to construct a concise fault feature set. Next, a heuristic PSO-BP neural network, whose learning process perfectly combines particle swarm optimization and the Levenberg–Marquardt algorithm, is constructed for fault classification. Numerical simulation experiments are conducted on four datasets sampled under different severity levels and load conditions, which verify that the proposed fault diagnosis method achieves efficient feature extraction and high classification accuracy.

Research on Medical Data Feature Extraction and Intelligent Recognition Technology Based on Convolutional Neural Network

“…At the same time, the experimental results on multiple datasets also verify the versatility of the model. Highrisk assessment of cerebral infarction was performed on structured dataset P1 using three machine learning methods, including naive Bayesian model [5], K nearest neighbor model [7], and decision tree model (Decision Tree) [9].…”

“…Calculate the error between the predicted value of the sample and the true target value of the sample, and perform layer-by-layer propagation from the output layer to the input layer to optimize the connection parameters between the layers in the structure, in order to minimize the cost function value. This process is backward widely used convolutional neural networks [7]- [9], using supervised learning methods to learn the characteristic representation of raw input data. Local connections between adjacent layers in the CNN are a variation of the neural network.…”

Improved SNR to detect the unknown characteristic frequency by SR

“…Therefore, the SR method has significant applications in bearing fault diagnosis. This method is different from the traditional methods such as wavelet analysis [2,3], neural network [4,5] and empirical mode decomposition [6,7]. The unique feature of SR is the noise energy transfers to the weak signal which contains feature information [8] rather than suppressing the noise in the raw signals.…”

An intelligent fault diagnosis method of rolling bearings based on Welch power spectrum transformation with radial basis function neural network

“…With the big data age arriving, deep learning technology has achieved excellent results in most mechanical fault diagnosis. Common fault diagnosis methods use the technology of signal processing to extract features in frequency-domain signal, time-domain signal, and time–frequency-domain signal, such as, time domain statistical analysis (Yan and Jia, 2018), Fourier spectrum analysis (Feng et al, 2018), wavelet transform (Wang et al, 2018), feature drop dimension (Yuan et al, 2017), etc. These signal processing methods need to remove useless and unimportant information during feature extraction and principal component analysis (PCA) (Gu et al, 2018), and then some of intelligent classifiers, including support vector machine (SVM) (Santos et al, 2015), multi-layer perceptron neural network (MLP) (Jedliski and Jonak, 2015), k-nearest neighbor (Guo et al, 2018), deep neural network (DNN) (Duong and Kim, 2018), and other classifiers, are used for auxiliary diagnosis.…”