A rolling mill fault diagnosis method with enhanced extended deep confidence network for sample imbalance

Abstract

The invention relates to a rolling mill fault diagnosis method for unbalanced samples. The method is established in the field of rolling mills, and uses vibration signal data and an enhanced extended depth confidence network to carry out fault diagnosis of the rolling mill; After the vibration signal data is obtained, the fast Fourier transform (FFT) can be used on the PC side to realize the rapid transformation of the signal time domain to the frequency domain, so as to realize the extraction of vibration signal features, and then use the extracted fault vibration signals of all categories. Train a Augmented Extended Deep Belief Network for fault diagnosis of subsequent rolling mills. In this method, the visible layer unit of the previous RBM is added to the visible layer unit of each RBM to form an RSRBM; the RSDBN composed of RSRBM can extract the missing useful information, and to a certain extent, it can reduce the sample imbalance caused by Problems with a low diagnosis rate can achieve a high diagnosis rate and can speed up the diagnosis.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis of rolling mills, in particular to a fault diagnosis method for an enhanced extended deep confidence network aiming at sample imbalance. Background technique [0002] With the rapid development of my country's manufacturing industry, the demand for steel continues to increase, and the health of the rolling mill has also received widespread attention. Among them, the rolling mill is one of the important equipments in the iron smelting industry. Due to its complex structure and harsh working environment, its (gearbox and roller bearings) are prone to failure. Economic losses. Therefore, it is very practical to detect the working state of the rolling mill (gear box and roll bearing) in time, to study and solve the diagnosis technology and method of the rolling mill, and to ensure the normal operation of the rolling mill. [0003] With the progress and development of science and technology, ...

AI Technical Summary

Problems solved by technology

[0005] Training the DBN network requires a large amount of training data, but in actual production, the rolling mill is in a normal state for a long time and the data about different types of faults in the rolling mill (gearbox and roll bearing) are not easy to obtain, which makes the rolling mill vulnerable to damage and makes the The data collection process is time-consuming and labor-intensive

The imbalance of samples will also cause the DBN model to overfit, which will make the diagnosis effect poor.

Images