Development of Customized Data Preprocessing Techniques and Artificial Intelligence Model for Predicting Extraction Current in Ion Implanters

In this paper, we developed an artificial intelligence(AI)-based model for predicting the extraction current of ion implanter by analyzing various operational conditions. We identified the device conditions that have a high correlation with the extraction current and applied an outlier removal method based on the distribution of these conditions during the preprocessing stage, which improved the learning performance. We performed a customized data preprocessing process based on Linear regression and DBSCAN. Using artificial neural network(ANN) techniques and ensemble methods with data-specific preprocessing techniques, we achieved high accuracy in predicting the extraction current of ion implanter. When applying preprocessing techniques, the R² for the predicted N-ion result was 0.904, whereas the R² (0.724) without applying these techniques was lower. Also, we validated the model’s applicability across various ion species. The value of the extraction current influences the number of ions injected into the ion source per second, thereby affecting the entire ion current. By predicting the extraction current, potential issues with the ion implanter can be identified early, and the operational efficiency of the equipment can be maximized. Through these results, it was confirmed that the possibility of using an ion implanter by incorporating artificial intelligence was confirmed.