Integration artificial intelligence with conventional X-Ray inspection for improved solder void detection in SSDC SiC modules.

Mohd Sa'at, Muhammad Haikal and Jamaluddin, Muhammad Herman and Mohamed Ameerdin, Muhammad Irshat and Shukor, Ahmad Zaki and Ahmad Izzuddin, Tarmizi and Mat Ibrahim, Masrullizam and Sukiman, Muhamad Shafiq (2024) Integration artificial intelligence with conventional X-Ray inspection for improved solder void detection in SSDC SiC modules. In: 2024 IEEE International Conference on Artificial Intelligence in Engineering and Technology (IICAIET).

Text Integration Artificial Intelligence with Conventional X-Ray Inspection for Improved Solder Void Detection in SSDC SiC Modules.pdf Download (508kB)

Abstract

In the context of electric vehicles (EVs), the demand for advanced power modules has grown significantly. Single Side Direct Cooling Silicon Carbide (SSDC SiC) is a key platform in the Onsemi VE-TracTM Family, designed for EV- traction inverters. This platform features advanced cooling and semiconductor technologies to ensure high performance and reliability. However, one of the critical challenges in manufacturing SSDC SiC modules is the detection of solder voids. Solder voids can significantly compromise the reliability and performance of power modules by reducing the effectiveness of heat dissipation and leading to potential failures. Conventional inspection methods relying solely on X-ray machines are often time-consuming and prone to human error, highlighting the need for a more efficient and accurate approach. The primary aim of this research is to develop an Artificial Intelligence (AI) based system to serve as a secondary validation layer for void detection in SSDC SiC modules. This system works in conjunction with X-ray machines, providing independent and accurate results for solder void inspections. This study leverages semantic segmentation techniques by using deep learning approach which is Convolutional Neural Networks (CNN), specifically U-Net, to classify void pixels from the background in X-ray images. The system processes output images from X-ray machines, differentiating void pixels from the background with high precision. Preliminary results indicate that the AI-based system can accurately classify solder voids, significantly reducing false positives and negatives compared to conventional inspection methods. The integration of this system with existing X-ray inspection processes promises enhanced detection, contributing to the improved solder void detection of SSDC SiC modules.

Item Type:	Conference or Workshop Item (Paper)
Divisions:	Faculty Of Electrical Technology And Engineering
Depositing User:	NURHASHIRAH BORHAN
Date Deposited:	27 Jul 2026 04:05
Last Modified:	27 Jul 2026 04:05
URI:	http://eprints.utem.edu.my/id/eprint/30092
Statistic Details:	View Download Statistic

Actions (login required)

View Item