Title: An efficient fault diagnosis model using Lappet Falco optimisation based on a deep neural network for the VSI under varying load conditions

Authors: Vaishali Baste; Dipali Shende; Seema Idhate; Arya Deshpande

Addresses: Department of Electronics and Telecommunication, SKNCOE, Vadgaon Pune, India ' Department of Electronics and Telecommunication, PCCOER, Ravet, India ' School of Computer Engineering and Technology, MIT-WPU, Pune, India ' Department of Electronics and Telecommunication, SKNCOE, Vadgaon Pune, India

Abstract: Numerous industrial applications employ three-phase converters that are based on insulated-gate bipolar transistors (IGBTs). However, the functioning and safety of power electronic devices and loads can be considerably impacted by IGBT faults. Maintaining high-power quality and system availability requires timely and accurate detection of power inverter failures. Constantly monitoring the failures in three-phase voltage source inverter (VSI) has greatly improved maintenance efficiency and stability. Hence, the developed research employs the discrete wavelet transform (DWT) and Lappet Falco optimised deep neural network (LFO-DNN) model to create an open circuit fault detection model for the VSI circuit. Data collection involves extracting features such as three-phase voltage, current, speed, and torque from erroneous data. The DNN classifier trained on these features uses the average three-phase current value to identify faulty switches. The VSI acting as a load with variable frequency reference is connected to a three-phase induction motor. The proposed Lappet Falco optimisation accurately yields impressive results in terms of prediction accuracy of 96.34%, precision of 96.34%, recall of 96.24%, F1 measure of 94.23%, MSE of 3.66, and specificity of 95.28%, demonstrating high efficiency for both 90% training and a k-fold value of 10.

Keywords: voltage source inverter; VSI; insulated-gate bipolar transistors; IGBTs; open circuit fault; deep neural network; DNN; Lappet Falco optimisation; LFO; discrete wavelet transform; DWT.

DOI: 10.1504/IJPELEC.2025.145951

International Journal of Power Electronics, 2025 Vol.21 No.3, pp.247 - 278

Received: 08 Feb 2024
Accepted: 10 Aug 2024
Published online: 30 Apr 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article