A new wavelet based diagnosis and protection of faults in induction motor drives

Abstract

In this paper, a novel diagnostic and protection algorithm based on wavelet neural network is developed and implemented in real-time for inverter faults in the vector controlled induction motor drives. The phase currents of an induction motor drive for different faulted and unfaulted conditions are preprocessed by wavelet packet transform in order to minimize the structure and timing of the proposed diagnostic technique. The wavelet packet transformed coefficients of line currents are used as inputs of a three-layer wavelet neural network. The performance of the proposed diagnosis and protection scheme is evaluated by simulation and experimental results. The proposed technique is evaluated and tested on-line for a laboratory 1-hp induction motor drive using the ds1102 digital signal processor board. In all the tests carried out, the type of fault is identified promptly and properly, and the tripping action is initiated almost at the instant or within one cycle of the fault occurrence.Index Terms-Digital signal processor, fault diagnosis, induction motor protection, inverter faults, wavelet neural network, wavelet packet transform.

Mechanical Failures Detection by Vibration Analysis in Rotary Machines Using Wavelet and Artificial Neural Network in the Gera Maranhão Plant.

“…According to ANHESINE, 1999, the main methods used for induction motors are the following: ultrasound, vibration, electrical signature (ESA), thermography, infrared, among others [2]. In combination with wavelets, which have, as one of the advantages in their favor, the ability to observe transient phenomena [3] and decompose the signal into various scales, making it possible not only to locate a phenomenon in time but to examine the signal in frequency, such a tool gives the user the possibility to collect a signal from nature and visualize it in all its subtleties. The wavelet transform has shown a wide applicability in the field of power systems, from the protection of three-phase generators [4][5][6] or the monitoring of other types of rotating machines through vibration analysis, to the protection of power system elements such as transmission lines [7] and distribution networks [7].…”

A Hybrid System Model-Based Open-Circuit Fault Diagnosis Method of Three-Phase Voltage-Source Inverters for PMSM Drive Systems

“…The diagnosis speeds of the above three voltage-based methods are high and the fault diagnosis can be completed within 1 ms. Some data-driven methods are also applied to OC fault diagnosis of inverters, such as neural networks [24][25][26], Bayesian networks [27], fuzzy logic [28,29], wavelet analysis [30], support vector machines [31] and machine learning [32]. However, these methods are difficult to be implemented due to the complex algorithms and large amounts of calculations, plus the diagnosis time is generally longer than one current fundamental wave period.…”