Detecting Transistor Defects in Medical Systems Using a Multi Model Ensemble of Convolutional Neural Networks

Abstract

The correct and flawless functioning of medical devices is of utmost importance for patients and the medical sector. Although medical device manufacturers utilize various protocols and procedures during the development and production of medical devices, the integrated circuits are usually manufactured by other parties. During the manufacture of the semiconductors utilized in the integrated circuits, the performance of the components can vary from batch to batch and even samples within the same batch can have different performance characteristics. For applications in life critical applications the selection of the most fit for purpose components is vital. In this paper, we propose a novel method for the semiconductor industry to verify the quality of transistors. The I-V graphs of the components are evaluated by multiple Convolutional Neural Networks (CNN) using visual data similar to expert evaluation, then these Machine Learning (ML) architectures utilize a multi model ensemble technique where one architecture providing a negative output will overrule the vote of the other architectures is utilized to assure very stringent quality control. Our method is tested on CMOS transistors and the results are comparable to those of experts with 10 years of experience in the industry.