Incremental Learning-Based Algorithm for Anomaly Detection Using Computed Tomography Data

Abstract

In a nuclear power plant (NPP), the used tools are visually inspected to ensure their integrity before and after their use in the nuclear reactor. The manual inspection is usually performed by qualified technicians and takes a large amount of time (weeks up to months). In this work, we propose an automated tool inspection that uses a classification model for anomaly detection. The deep learning model classifies the computed tomography (CT) images as defective (with missing components) or defect-free. Moreover, the proposed algorithm enables incremental learning (IL) using a proposed thresholding technique to ensure a high prediction confidence by continuous online training of the deployed online anomaly detection model. The proposed algorithm is tested with existing state-of-the-art IL methods showing that it helps the model quickly learn the anomaly patterns. In addition, it enhances the classification model confidence while preserving a desired minimal performance.

One or two things we know about concept drift—a survey on monitoring in evolving environments. Part A: detecting concept drift

“…The constantly changing world presents challenges for automated systems, for example, those involved in critical infrastructure, manufacturing, and quality control. Reliable functioning of automated processes and monitoring algorithms requires the ability to detect, respond, and adapt to these changes (Ditzler et al, 2015 ; Reppa et al, 2016 ; Chen and Boning, 2017 ; Vrachimis et al, 2022 ; Gabbar et al, 2023 ).…”

One or two things we know about concept drift—a survey on monitoring in evolving environments. Part B: locating and explaining concept drift

“…While humans are capable of navigating an ever-changing environment, these changes pose challenges to many automated systems (Ditzler et al, 2015 ). Considering monitoring and control tasks, e.g., in critical infrastructure (Vrachimis et al, 2022 ), manufacturing (Chen and Boning, 2017 ), and quality control (Gabbar et al, 2023 ), in order to work reliably, automatized processes and supervision algorithms need to be able to detect, react, and adapt to changes (Reppa et al, 2016 ).…”