An Intelligent Fault Diagnosis Method Based on Multi-task Feature Sharing Neural Network

Abstract

The invention discloses an intelligent fault diagnosis method based on a multi-task feature sharing neural network, comprising the steps of: (1) separately collecting the original vibration acceleration signals of rotating machinery under different experimental working conditions, and constructing a method by intercepting a certain length of signal data (2) Build a multi-task feature sharing neural network, including: input layer, feature extractor, classification model and prediction model; (3) Use multi-task joint training to train classification and prediction models at the same time; (4 ) Input the vibration acceleration signal collected in the actual industrial environment into the trained model to obtain the multi-task diagnosis result. The invention can realize the classification of fault types and the prediction of fault degree at the same time, and has high practical application value.

Description

technical field [0001] The invention belongs to the field of mechanical fault diagnosis, and in particular relates to an intelligent fault diagnosis method based on a multi-task feature sharing neural network. Background technique [0002] With the rapid development of science and technology and the deployment of national strategies such as "Made in China 2025" and "New Generation Artificial Intelligence Development Plan", my country's machinery manufacturing industry is gradually entering the era of digital intelligent manufacturing. Mechanical equipment systems in all walks of life are constantly developing in the direction of complexity, digitization and intelligence. However, mechanical equipment operating under high load, high speed and high operating rate for a long time is prone to fatigue failure, which will lead to equipment shutdown, and even lead to major safety accidents and huge economic losses. Therefore, intelligent fault diagnosis and predictive maintenance ...

AI Technical Summary

Problems solved by technology

However, the general intelligent fault diagnosis method still has the following limitations: 1) The same type of faults with different degradation degrees are regarded as multiple different fault modes, and the classification method is used to identify the degradation degree

However, in the actual industrial environment, the parameters for evaluating the degree of equipment degradation are mostly continuously changing physical quantities, and it is not in line with the actual industrial situation to evaluate equipment degradation by classification; 2) When the working conditions (such as speed and load) change, Poor generalization ability of diagnostic algorithms

Traditional methods generally improve the generalization ability of the model by enlarging the sample size of the data set, but collecting data of each fault type under all working conditions is not only economically costly, but also time-consuming and laborious, and the feasibility of implementation is poor; 3) Existing methods Most of them are limited to specific diagnostic tasks, such as equipment fault classification or performance degradation prediction. The scalability and portability of the algorithm are poor, and it is difficult to meet the increasingly diverse and flexible diagnostic needs in the actual industrial environment.

Images