Yan, Bingxin, Sun, Qiuzhuang, Ma, Xiaobing, and Shen, Lijuan (2025) Anomaly Detection in Complex Systems Using a Hybrid Framework. In: Proceedings of the 16th International Conference on Reliability, Maintainability and Safety. From: ICRMS 2025: 16th International Conference on Reliability, Maintainability and Safety, 27-30 July 2025, Shanghai, China.

PDF (Published Version) - Published Version Restricted to Repository staff only

Abstract

In complex systems such as high-speed train bogies, supervisory control and data acquisition (SCADA) systems continuously collect data reflecting both operational states and environmental conditions. These data are well-known as process variables and covariates, respectively. Effective anomaly detection using SCADA data is challenging due to intricate dependencies between process variables and covariates. We propose a hybrid framework that embeds physical domain knowledge into a statistical modeling pipeline to improve anomaly detection accuracy and interpretability. A constrained spline-based regression model is developed to characterize the intricate dependencies between process variables and covariates. Domain-specific physical insights are incorporated to guide parameter estimation and ensure interpretability. The proposed method is validated through case studies in high-speed train power bogies, illustrating its ability in anomaly detection.

Item ID:	93850
Item Type:	Conference Item (Research - E1)
ISBN:	979-8-3315-3513-1
Copyright Information:	© 2025 IEEE
Date Deposited:	02 Sep 2026 02:57
FoR Codes:	49 MATHEMATICAL SCIENCES > 4905 Statistics > 490508 Statistical data science @ 70% 40 ENGINEERING > 4010 Engineering practice and education > 401006 Systems engineering @ 30%
SEO Codes:	24 MANUFACTURING > 2415 Transport equipment > 241504 Rail equipment @ 80% 27 TRANSPORT > 2703 Ground transport > 270306 Rail safety @ 20%
Downloads:	Total: 1 Last 12 Months: 1
	More Statistics