Multi-resolution deep neural network intelligence diagnosis method of mechanical transmission fault

Abstract

The invention discloses a multi-resolution deep neural network intelligence diagnosis method of a mechanical transmission fault. The method comprises the following steps of for a sample set, acquiring an initial parameter of a deep neural network; successively inputting the deep neural network for all the samples in the sample set and acquiring a network output result; comparing the output result and a sample label so as to acquiring a corresponding total classification error; using the acquired total classification error to calculate an error of each layer of the neural network; using a deep neural network error to update a network parameter; using an acquired new deep neural network to calculate a network output result of each sample; and successively carrying out iteration till that classification precision or an iteration frequency reaches a preset requirement, and acquiring and outputting a classification result and a network parameter. In the invention, mechanical signals acquired through collection are taken as a network input so that manpower and material resources spent on characteristic extraction optimization are omitted; and through adaptive characteristic learning of a multilayer nerve network, classification precision and network anti-noise performance are increased.

Description

technical field [0001] The invention relates to a mechanical equipment failure intelligent diagnosis technology, in particular to a multi-resolution deep neural network intelligent diagnosis method for a mechanical transmission system failure. Background technique [0002] At present, the traditional intelligent fault diagnosis technology of mechanical equipment needs to extract and optimize the features of the collected signals first, in order to carry out the corresponding classification task learning, such as extracting the root mean square value, skewness, kurtosis, information entropy, etc. of the signal. With the increase of data volume and the significant increase of computational complexity, traditional mechanical fault intelligent diagnosis algorithms need to consume a lot of manpower and material resources to calculate sensitive features and perform feature optimization, which is no longer competent for intelligent diagnosis tasks under complex and large-scale data....

AI Technical Summary

Problems solved by technology

With the increase in the amount of data and the significant increase in computational complexity, traditional mechanical fault intelligent diagnosis algorithms need to consume a lot of manpower and material resources to calculate sensitive features and perform feature optimization, which is no longer competent for intelligent diagnosis tasks under complex and large-scale data.

Traditional methods are difficult to meet the current needs of learning and intelligent classification of massive data

Images