Feasibility Examination of Machine Learning-based Process Monitoring Approach for Pyroprocessing Safeguards

Abstract

During the pyroprocessing operation, various signals can be obtained from process monitoring (PM): Physical, electrochemical, and spectroscopic analytical. These signals can provide information about the state of process operations. To manage massive data effectively, and utilize it productively, Artificial Intelligence (AI) can be applied. Machine Learning (ML) is an application of AI. Based on existing data, the computer learns how to conduct a given task by using statistical inference without explicit programing. ML makes it possible to diagnose an operational state in real time. Depending on the nature of the signal, the type of machine learning and the appropriate algorithm could take different forms. This study is conducted to examine feasibility of ML based PM approach with an electrorefining process which is a one of main unit processes in pyroprocessing. The tasks performed include: 1) Suggesting signals suitable for PM from among the process signals that can be measured during the process operation; 2) Identifying operating conditions that can be considered ‘normal’ or ‘off-normal’ based on using the process signals; 3) Accumulating data for ML under various conditions, including normal operation and off-normal operation. Data development was implemented by simulating normal/off-normal operation through a computer model. Based on this investigation, a ML based pattern recognition system will be developed to utilize suggested signals to examine the operational state of the process for pyroprocessing safeguards applications.