Gearbox Fault Diagnosis Based on Two-Class NMF Network Under Variable Working Conditions

Abstract

The gearbox is an important part of the wind turbine, and it is also a part prone to failure. Most of the fault diagnosis methods for gearboxes lack the ability to adapt to changing operating conditions. Once the operating conditions change, it is necessary to relearn the object characteristics, otherwise it is prone to misjudgment. Therefore, this paper proposes a method based on a two-class non-negative matrix factorization (NMF) network to realize gearbox fault diagnosis under variable operating conditions without additional data. The method is divided into two parts: model training and fault diagnosis. The former takes the basic operating conditions as the benchmark. It extracts the local static characteristics of the fault samples, and trains them into classifier models with different diagnostic functions to construct a network. The latter uses the static distance obtained from the data input network of the new operating condition as an important indicator for detecting the occurrence of a fault and identifying the type of the fault. The experimental results based on the QPZZ-II rotating machinery vibration test stand show that compared with other methods, the algorithm proposed in this paper has a good application effect on the diagnosis under variable operating conditions.

Diagnosis of Loose Core Fault in Saturable Reactor of Thyristor Valve Based on Vibration Signal Time-Frequency Analysis and CNN

“…At present, there is little research on the fault diagnosis of saturable reactors. However, there are abundant research achievements in fault diagnosis based on vibration signals [3][4][5][6][7]. In practical engineering, thyristor valves operate under different conditions, which can affect the vibration characteristics of saturable reactors.…”

A Machine Learning Approach for Gearbox System Fault Diagnosis

“…A novel deep neural network which combines EMD, LSTM, and particle swarm optimization was presented in the study [50]. A method based on the usage of a two-class nonnegative matrix factorization network was proposed in [51]. There are many more applications of deep learning techniques in gearbox fault diagnosis that were published recently, which we are aware of but not mentioning here due to the scope of this article.…”