Fault prediction method based on fast integrated convolutional neural network

Abstract

The invention belongs to the technical field of neural network fault prediction, and discloses a fault prediction method based on a fast integrated convolutional neural network. The method comprises the following steps: (a) collecting an output signal of a to-be-predicted object, and carrying out time-frequency analysis on the output collected signal by adopting S transformation to obtain a time-frequency signal; (b) constructing a convolutional neural network structure based on LeNet-5, providing a maximum and minimum cosine cyclic learning rate scheduler based on a self-weight starting cosine cyclic learning rate, and setting a learning rate range of the convolutional neural network; and (c) training a time-frequency signal of the prediction data by using the convolutional neural network, and taking a prediction result of the SECNN as a final prediction result. According to the invention, fault prediction with high prediction speed and high prediction precision can be realized.

Description

technical field [0001] The invention belongs to the technical field of neural network fault prediction, and more specifically relates to a fault prediction method based on fast integrated convolutional neural network. Background technique [0002] Fault diagnosis plays a vital role in modern complex systems. Data-driven fault diagnosis methods have been increasingly studied by many researchers from academic and engineering fields. Advanced data-based methods to effectively mine data features and accurately predict fault modes have become the key research directions of data-driven fault diagnosis. These are based on Data-based methods can use historical data to build failure models without relying on previous physical models, making them well suited for modern complex industrial processes. [0003] Deep learning (DL) uses data representation learning instead of explicit feature extraction to perform learning tasks, and enables features to be learned automatically, showing gr...

AI Technical Summary

Problems solved by technology

[0004] Almost all ensemble learning methods need to train multiple base learners, and the training process is very time-consuming, so a new ensemble learning is proposed, called Fast Ensemble Learning (SEL), which uses Cosine Cyclic Learning Rate (CCLR) scheduling , and divide the entire training process into M cycles. Since the training model will reach local minima at the end of each cycle, SEL combines these local minima and shows strong performance. SEL trains the model once Obtaining M different base learners overcomes the above disadvantages. However, how to choose the appropriate learning rate range of the cosine cyclic learning rate (CCLR) is very important. In the cosine cyclic learning rate (CCLR), the upper learning rate Bounds are pre-defined empirically or experimentally; the Linear Learning Rate Test (ILR Test) method explores the optimal learning rate by increasing the learning rate from low to high and observing when the test accuracy starts to increase and when the accuracy starts to vibrate , the ILR test can provide extremely valuable information about the learning rate when faced with a new data set, but when the learning rate range is large, the ILR test has low resolution

Images