An auto-deployed model-based fault detection and diagnosis approach for Air Handling Units using BIM and Modelica

• Published in: Automation in construction Vol. 96; pp. 508 - 526
• Main Authors: Andriamamonjy, Ando, Saelens, Dirk, Klein, Ralf
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.12.2018; Elsevier BV
• Subjects: Automation; Building Information Model (BIM); Building information modeling; Building management systems; Chains; Computer simulation; Dampers; Efficiency; Energy consumption; Error detection; Fault detection; Fault Detection and Diagnosis (FDD); Fault diagnosis; First principles; Industry Foundation Class (IFC); Mathematical models; Model based method; Modelica; Multiple objective analysis; Robustness (mathematics); Stability; Windows (intervals); Building Information Model (BIM); Model based method; Modelica; Industry Foundation Class (IFC); Fault Detection and Diagnosis (FDD)
• ISSN: 0926-5805; 1872-7891
• DOI: 10.1016/j.autcon.2018.09.016

The Air Handling Unit (AHU) is one of the most energy consuming devices in building systems. Fault Detection and Diagnosis (FDD) methods integrated into AHUs can help to ensure that they comply with the intended design, and their efficiency is maintained throughout the entire operational stage of the building. Nonetheless, the implementation and deployment of FDDs at the operational stage require an extensive effort. Especially, FDD approaches that rely on first principle models (model-based FDD) need to be manually implemented, and the information necessary for this process is scattered between several exchange formats and files, thus making it time-consuming, error-prone and subject to modellers' poor judgment. This study aims at facilitating and partially automating the implementation and deployment of model-based FDD. An automated tool-chain that combines a BIM (Building Information Model)-to-BEPS (Building Energy Performance Simulation) tool with a model-based FDD approach is developed. The contribution of this paper lies in the extension of an existing BIM to Modelica BEPS method with an automated calibration approach and a novel model-based FDD. These three elements are integrated in a framework (implemented using Python) to reduce experts' involvement in FDD implementation and deployment. The developed model-based FDD combines a parity relation procedure for fault detection and profile identification for fault diagnosis. The latter uses the robust multi-objective optimisation algorithm NSGA-2. An error is detected when the difference between prediction and measured data over a specific time window is superior to a predefined threshold. The origin of the error is subsequently identified by estimating the profile of the different controllable components' control signal. The developed tool-chain was applied to an actual AHU as well as on several numerical scenarios to identify typical AHU faults such as faulty dampers, valves and sensors. This study shows that the developed model-based FDD approach can identify some of the most common faults in AHUs, but more importantly that BIM can facilitate the deployment of model-based FDD in building systems. •Propose a BIM workflow to automate the implementation and deployment of FDD.•Use an existing IFC to BEPS interface and propose an automated calibration method.•Use a combined parity relation–parameter estimation FDD approach.•Identify typical AHU faults and demonstrate BIM relevance for FDD implementation.