Method and system for fault diagnosis of transmission system of engineering equipment

Abstract

The invention discloses an engineering equipment transmission system fault diagnosis method and a system thereof. The method comprises the steps of converting a time domain signal of an objective dataset and an auxiliary data set to an RF signal; by means of a deep automatic encoder, extracting an RF signal characteristic, and using the RF signal characteristic as input of migration learning; bymeans of an instance-based migration boosting algorithm, changing the weights of the objective data set and the auxiliary data set in each iteration, and changing the weight of fault data in the objective data set. and using a classifier which is obtained through each iteration of the migration boosting algorithm and the weight thereof as a final classifier. The system comprises a time frequency conversion module, a fault signal characteristic extracting module and a fault diagnosis module. The engineering equipment transmission system fault diagnosis method and the system thereof have advantages of effectively extracting the signal characteristic, sufficiently using the data set, diagnosing the fault equipment as possible, and improving safety.

Description

technical field [0001] The invention relates to the technical field of engineering equipment transmission systems, in particular to a fault diagnosis method and system for engineering equipment transmission systems based on characteristic signal analysis, deep learning and transfer learning. Background technique [0002] For the vibration data of transmission system of engineering equipment, expressing its characteristics well is the key to fault diagnosis. It is time-consuming and labor-intensive to extract features manually, and different extraction methods are required for different problems, so it is difficult to find a unified feature extraction method. [0003] The fault diagnosis of the current engineering equipment transmission system is based on the assumption that the training data and the test data have the same distribution, but in the actual situation, due to the variable working environment, the amount of new fault data is small, and it is difficult to train a ...

AI Technical Summary

Problems solved by technology

[0006] However, the above-mentioned patent has the following disadvantages: although the above-mentioned patent uses the Adaboost algorithm to integrate multiple SVM classifiers, it does not make full use of the data under other conditions

If the current amount of data is not enough to train the SVM classifier, that is, when the fitting effect of the trained SVM classifiers is not ideal, a good model will not be obtained in the end. Using such a model for fault prediction is still not safe. Guaranteed, unable to meet engineering equipment transmission system fault diagnosis requirements

Images