Equipment fault diagnosis method based on multi-source signals and deep learning

Abstract

The invention relates to an equipment fault diagnosis method based on multi-source signals and deep learning. The method is characterized in that the multi-source signals are employed to carry out equipment fault diagnosis, a plurality of state monitoring data of equipment is fully and effectively utilized, more important fault information for fault diagnosis is provided, and the diagnosis still keeps satisfactory robustness and accuracy in complex environments such as multiple working conditions. According to the method, a deep neural network model is constructed based on deep learning, and original data and feature extraction do not need to be manually processed; besides, a thought of a residual network is used for reference, jump connection is added on the basis of the convolutional neural network model to expand the network depth, and each layer of network only learns the residual represented by the feature, so that not only the model convergence process is accelerated, and the training efficiency of the model is improved, but also the fitting capability and diagnosis performance of the model can be improved; a global average pooling is introduced to replace a part of a full connection layer, the training process of the network is accelerated while the over-fitting risk of the classifier is reduced by reducing network structure parameters, and therefore, the diagnosis performance and generalization ability are effectively improved.

Description

technical field [0001] The invention relates to a device fault diagnosis technology, in particular to a device fault diagnosis method based on multi-source signals and deep learning. Background technique [0002] The failure of equipment or parts has a great impact on the operation of the system, which may cause huge economic losses and even seriously endanger personal safety. Timely diagnosis and recovery of faulty equipment is of great significance to production safety and production efficiency. Modern industrial systems have large scale, large number of components, complex structures, and high component coupling, which makes system fault diagnosis difficult. Therefore, research on accurate, fast and stable fault diagnosis methods is the focus of current research. [0003] From the aspect of signal source selection, most studies in the field of rolling bearing fault diagnosis currently use a single signal as the analysis object for feature learning, such as vibration sig...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the insufficient robustness of single-signal fault diagnosis, a device fault diagnosis method based on multi-source signals and deep learning is proposed, using multi-source sensor signals to provide more device status information, so that it still meets the diagnostic performance in complex environments Require

Images