Fault detection and isolation for a secondary loop refrigeration system. (5th June 2023)

• Record Type: Journal Article
• Authors: Poks, Agnes; Fallmann, Markus; Fink, Lorenz; Rinnofner, Lukas; Kozek, Martin
• Is Part Of: Applied thermal engineering. Volume 227(2023)
• Publication Date: 2023-06-05
• Subjects: Fault detection -- Fault isolation -- Secondary loop refrigeration -- Model-based method -- Extended Kalman filter -- Unknown input observer -- Experimental validation; Heat engineering -- Periodicals; Heating -- Equipment and supplies -- Periodicals; Periodicals; 621.40205
• DOI: 10.1016/j.applthermaleng.2023.120277
• Languages: English
• ISSNs: 1359-4311

Abstract

Abstract: Secondary loop refrigeration systems offer advantages over conventional system architectures: refrigerant charge reduction, maintenance simplification, and reduced safety hazards. Currently, no model-based fault detection for a secondary loop system has been reported. This article presents a fault detection and isolation scheme for a secondary loop refrigeration system where parametric faults (dirt/ice-buildup and secondary loop pump failure) and sensor faults (four temperature sensors in the secondary loop) are detected and isolated. The fault detection and isolation system is based on unknown input observers and on extended Kalman filters. The models are obtained through physical relations combined with experimental parameter determination (gray-box model) and are cross-validated against measurement data. The complete fault detection and isolation architecture is experimentally validated on a secondary loop refrigeration system mounted on a cooling box inside a climate chamber. Experimental results show that important parametric faults can be reliably detected and isolated, and the fault detection and isolation of sensor faults are effective. Highlights: Fault detection and isolation are applied to a secondary loop refrigeration system. Experimental validation of effective fault detection for that system is presented. Both parametric and sensor faults can be reliably detected and isolated. The physical system model can be well parametrized from a small experimental data set.