The Advancements of Fault Detection in Microelectronics Based on Machine Learning Models

Abstract

Fault detection in microelectronics constitutes an indispensable component of microelectronic applications, routinely employed to ensure the normal operation of electronic components. However, due to the inherent characteristics of microelectronic devices, such as their miniature size and high precision, conventional detection methods often struggle to meet the stringent requirements for their inspection. Machine Learning (ML) recently has gained prominence as a result of the development of Artificial Intelligence (AI) technologies, and it offers new methods for microelectronics. Convolutional structures are specifically used in Convolutional Neural Networks (CNNs), a type of deep neural network that helps mitigate model overfitting problems and lowers the memory footprint of deep networks. Recurrent Neural Networks (RNNs) are an important class of artificial neural networks that are specifically made to process sequential data. They can handle any length of sequence. Transformers are characterized by the introduction of the self-attention mechanism, enabling exceptional performance when dealing with sequential data. The application of models trained on these various architectures in the context of microelectronics inspection is the main focus of this paper, with an emphasis on how they affect fault detection accuracy. Furthermore, the concluding section of this paper discusses the current challenges associated with employing machine learning for microelectronics fault detection and explores potential future solutions. The process of addressing these challenges will not only drive the advancement of ML itself but is also expected to lead to more efficient and diverse methodologies for fault detection in microelectronics.