2605.05032

Total: 1

#1 Quantized Probabilistic AI for Gear Fault Diagnosis in Motor Drives [PDF] [Copy] [Kimi] [REL]

Authors: Subham Sahoo, Huai Wang, Frede Blaabjerg

Deploying large artificial intelligence (AI) models in power electronics often demands high computational resources. Driven by the quantization paradigm, this digest proposes a quantization-aware training (QAT) principle to substantially minimize the number of bits required and simultaneously maximize the accuracy of computations in pre-trained AI models. Considering a pre-trained probabilistic Bayesian Neural Network (BNN) for gear fault diagnosis in motor drives as an example, we quantize its weights and activation functions from floating-point FP32 to low-precision INT8 values, which enhances the computational efficiency by a significant margin of 30-45% (for different model versions) without any compromise in the accuracy and uncertainty estimates. This substantiates a sustainable mechanism of deploying most quantized light-weight AI models into low-cost edge processors for power electronic applications.

Subject: Systems and Control

Publish: 2026-05-06 15:27:15 UTC