Quantum-enhanced wafer failure prediction in semiconductor fabrication

In recent years, rapid advancements in semiconductor manufacturing have underscored the critical need for robust quality control to maintain high yields and reliability in wafer production. In this context, accurately predicting wafer failures plays a pivotal role across the industry in minimizing production losses and enhancing cost efficiency. However, despite the use of traditional machine learning and deep learning methods to predict wafer failures, they have failed to achieve the desired levels of accuracy. Consequently, the present study investigates the application of Quantum Machine Learning (QML) methods on the SECOM dataset. This dataset was balanced using the Synthetic Minority Oversampling Technique (SMOTE), wherein the QML models were implemented using the PennyLane library. In this study, comprehensive data pre-processing was performed on the SECOM dataset; subsequently, various QML algorithms such as Quantum Support Vector Machines (QSVM) and Quantum Random Forests (QRF) were implemented to classify wafers as 'good' or 'bad.' Although initial improvements in performance compared to traditional methods were modest, leveraging PennyLane's capabilities and fundamental principles of QML, this study highlights their transformative potential for industrial applications.