Figure - available from: Advances in Mechanical Engineering
This content is subject to copyright.
Source publication
Ensemble local mean decomposition has been gradually introduced into mechanical vibration signal processing due to its excellent performance in electroencephalogram signal analysis. However, an unsatisfactory problem is that ensemble local mean decomposition cannot effectively process vibration signals of complex mechanical system due to the constr...
Citations
... It is one of the most widely used methods in vibration signal processing. The time-frequency domain signal processing techniques include short-time Fourier transform (STFT) [15], wavelet transform (WT) [16], wavelet packet decomposition (WPD) [17], empirical mode decomposition (EMD) [18] and their variants [19,20]. These methods extract vibration signal characteristics by analyzing the generation times of different frequency signals, and the frequency components contained in different times. ...
Early detection of faults is of importance to avoid catastrophic accidents and ensure safe operation of machinery. A novel graph neural network-based fault detection method is proposed to build a bridge between AI and real-world running mechanical systems. First, the vibration signals, which are Euclidean structured data, are converted into graph (non-Euclidean structured data), so that the vibration signals, which are originally independent of each other, are correlated with each other. Second, inputs the dataset together with its corresponding graph into the GNN for training, which contains graphs in each hidden layer of the network, enabling the graph neural network to learn the feature values of itself and its neighbors, and the obtained early features have stronger discriminability. Finally, determines the top-n objects that are difficult to reconstruct in the output layer of the GNN as fault objects. A public datasets of bearings have been used to verify the effectiveness of the proposed method. We find that the proposed method can successfully detect faulty objects that are mixed in the normal object region.
Early fault features of large-scale and low-speed mechanical equipment with heavy duty are weak and exhibit strong non-stationary characteristics. The adaptive extraction and identification of highly relevant important features from such signals has attracted significant attention. In this study, a novel empirical variational mode decomposition and exact Teager energy operator are proposed to explore valuable information. To highlight the fault impact signal representation, we use the exact energy operator to enhance the weak-impact components in the early fault signal. The proposed binary mechanism effectively distinguishes irrelevant features based on the adaptive decomposition parameter construction strategy. Therefore, interference features are easily removed from similar mixed signals, and the independent mode features are determined. The experimental results of the simulation and collected data are compared with those obtained with existing signal decomposition methods, and the superiority of the proposed method, owing to its better modal distinction and less time consumption, is verified.
Mortar void is a hidden defect in ballastless slab track difficult to be efficiently identified by traditional detection methods. This paper is dedicated to proposing a new detection method to identify the mortar void position and length using the vehicle response combined with the hybrid convolutional neural network-support vector machine (CNN-SVM) classifier. The vertical wheelset accelerations with different mortar void conditions are collected from a vehicle-track coupled dynamics simulation model. The first components decomposed from wheel-set accelerations by local mean decomposition and their envelopes are utilized as the training data due to their sensitivity to mortar void. To improve the identification precision, the scope descent method is proposed to determine the range influenced by mortar void (IMVR) and samples are labeled according to IMVR. Meanwhile, identification results are post processed based on the mortar void characteristics. The results show that over 90 % mortar void conditions with the length of 0.65 m are detected correctly and the identification has a higher precision with the mortar void length greater than 0.95 m. The proposed technology of mortar void detection using the wheelset accelerations with the hybrid CNN-SVM classifier provides reference for engineering application, which is of great significance to relieve the pressure of health monitoring of railway track.
