Grouping sparse filtering: a novel down-sampling approach toward rotating machine intelligent diagnosis in 1D-convolutional neural networks

Abstract

Convolutional neural networks (CNNs) has weight sharing and feature learning ability, which can efficiently and effectively complete the health monitoring of industrial equipment. However, the pooling operation in a typical CNN can cause a loss of valuable impulse features during data down-sampling. We propose grouping sparse filtering (GSF) to overcome this problem. Instead of using a pooling operation, the GSF splits the channels of feature obtained after convolution into equal-length groups. A feature selector with a feature aggregation function based on the channel importance factors and a lasso constraint is used to filter the groups to perform down-sampling. The GSF method preserves the impulse features because of the block sparsity of the vibration signal. Theoretical analysis demonstrates that the GSF has a similar computational complexity as using a pooling layer in a CNN for the same number of layers. Two experimental studies were conducted using data from a laboratory test and industrial environments. The experimental results show that the 1D-CNN with GSF provides better performance for retaining the impulse features of the rotating machinery signals and higher fault identification accuracy than a CNN with a pooling layer.

In-Wheel Motor Fault Diagnosis Using Affinity Propagation Minimum-Distance Discriminant Projection and Weibull-Kernel-Function-Based SVDD

“…Moreover, it is essential to select a suitable and effective classifier for an on-time fault diagnosis system of an in-wheel motor. In recent years, there are many classification algorithms that have been widely used in the field of fault diagnosis, such as the support vector machine [23,24], artificial neural networks [25], convolutional neural network (CNN) [26,27], multi-scale CNN [28], artificial hydrocarbon network [29], generative adversarial network [30], dynamic Bayesian network [31], graph attention network [32], segmentation framework [33], Gaussian process regression [34] and hidden Markov model [35]. These methods are constructive to the fault diagnosis system of an in-wheel motor.…”

Identification of engine faults based on acoustic emission signals using a 1DCNN-ViT ensemble model

“…Intelligent fault diagnosis methods based on neural networks are becoming increasingly prevalent, following the popularity of deep learning. When using neural networks to intelligently diagnose engine fault AE signals, the two main steps are feature extraction and fault identification [25][26][27]. Feature extraction refers to the extraction of meaningful feature parameters from signals.…”

Fault diagnosis of industrial robot gearbox using joint interclass and intraclass mappings

“…This can achieve the purpose of maximizing intraclass distance. Moreover, since high-precision sensors have been basically used in fault diagnostics [33,34], the corresponding costs were relatively expensive. Considering the practicability and economy of the fault diagnostics, low-cost attitude sensors are used to collect data on the motion trajectory of industrial robots.…”