Abstract:
Water treatment plants (WTPs) encompass complex processes, presenting challenges for both control algorithms and human operators. Traditional anomaly detection often requires human intervention, and both parties exhibit limitations when handling system anomalies. This study introduces a "digital twin" model of water systems, enhanced with an extended reality (XR) interface, designed to capture and replicate operators' anomaly detection strategies. The objective is to overcome the difficulties associated with tracing and interpreting operators' behaviors, which are often due to inadequate process pattern mining methods. The proposed approach is fourfold: (1) a water system digital twin equipped with simulation models and a virtual reality (VR) interface, (2) process capture for inspection and operation, (3) visual trajectory pattern mining for knowledge discovery, and (4) an augmented reality interface to guide workers. Experimental results indicate that for simulated anomaly inspection and detection, the computer's recall was 0.712, whereas human operators achieved a recall of 0.851. A combination of both yielded a higher classification recall of 0.885. Knowledge transfer predicated on specific observations could address pressing WTP challenges, including a shortage of experienced operators and unreliable fault recovery systems. Β© 2024 Computing in Civil Engineering 2023: Visualization, Information Modeling, and Simulation - Selected Papers from the ASCE International Conference on Computing in Civil Engineering 2023. All rights reserved.
