Depth feature and statistical feature fused inverter fault diagnosis method

Abstract

The invention discloses a depth feature and statistical feature fused inverter fault diagnosis method, which comprises the following steps: firstly, extracting depth features of three-phase output current signals of an inverter by using an SE-DenseNet method, extracting statistical features of current signal samples by using a Hilbert-Huang transform (HHT) method, and combining the depth features and the statistical features; secondly, performing dimension reduction on the combined high-dimensional features by using a local Fisher discriminant analysis algorithm LFDA to obtain low-dimensional features capable of expressing inverter fault features, and realizing fusion of deep features and statistical features; and finally, adopting an extreme learning machine (ELM) classifier, and taking the low-dimensional features as input to realize fault state identification of the three-level inverter. Compared with a traditional diagnosis method, the diagnosis method provided by the invention is higher in fault recognition accuracy, can obtain ideal performance in fault diagnosis under different working conditions, and has higher adaptive capacity and generalization capacity for actual industrial scenes.

Description

Technical field [0001] The present invention belongs to the field of fault diagnosis, and in particular, in particular, inverter fault diagnosis methods of depth features and statistical fusion. Background technique [0002] The inverter is a very important and complex device for realizing motor frequency conversion speed, and it is also easy to fail. The main components of the inverter have rectifiers, inverters and control circuits. Motor control system in mine hoist equipment generally assembles multi-level inverters, neutral Point Clamped, NPC) three-level inverter is one of the commonly used multi-level inverters, phase Compared to the two-level inverter, the NPC three-level inverter can reduce the rate of voltage and current, smaller for the impact of the motor, can effectively improve the life of the motor; however, due to the three-level inverter The desired IGBT power switch tube is more, the structure is more complicated, resulting in fault frequency, in all faults of t...

AI Technical Summary

Problems solved by technology

However, in the feature extraction process, the extracted features may have incomplete fault information, and the extraction process is complex, which not only increases the complexity of data processing, but also increases the computational burden of the computer and affects the classification accuracy.

Images