[5K2-OS-38a] OS-38

12:00 PM - 12:15 PM JST(3:00 AM - 3:15 AM UTC)

[5K2-OS-38a-01] Fault Detection and Cause Estimation System Using Digital Twins

〇Shinji Orihara¹, Kota Hiroshima¹, Yohei Nawaki¹, Kentaro Nomoto¹, Kazuyuki Tsuruoka¹ (1. USHIO INC)

Keywords:

Fault Detection,Cause Estimation,Digital Twin

In recent years, numerous studies have investigated fault detection using digital twins. However, only limited work has addressed fault cause estimation. Furthermore, the cost of constructing high-fidelity digital twins can pose a practical barrier to implementation. This study proposes a fault detection and cause estimation system that explicitly allows for modeling errors and can be easily updated as operational data are accumulated. As inputs to the estimator, we use a small number of physical parameters. These are obtained by fitting a physical model to time-series data from inspection operations, together with the RMSE of the residuals between model outputs and measurements. The diagnostic algorithm is implemented as a pluggable module. Multiple model candidates can be flexibly selected or replaced depending on the available data and operational requirements. This presentation describes a framework in which models are constructed already at the design stage, before the actual equipment is available. These models are subsequently updated as operational data are accumulated. It also reports preliminary evaluations using virtual faults and a small amount of real fault data.