Feature-level interpretability in transfer learning-based chiller fault diagnosis

Chen, Zhen, Rezgui, Yacine ORCID: https://orcid.org/0000-0002-5711-8400, Zhang, Ran, Zhang, Xingxing, Zhao, Wanqing and Li, Yu 2025. Feature-level interpretability in transfer learning-based chiller fault diagnosis. Building and Environment , 113527. 10.1016/j.buildenv.2025.113527 Item availability restricted.

PDF - Accepted Post-Print Version Restricted to Repository staff only until 8 August 2026 due to copyright restrictions. Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (3MB)

Abstract

Accurate fault diagnosis in chillers is essential for maintaining optimal energy efficiency and operational reliability in building Heating, Ventilation, and Air Conditioning (HVAC) systems. However, chillers often operate under diverse conditions, which can cause data-driven models trained on specific operational conditions to suffer significant performance deterioration when deployed in new conditions. Transfer learning offers a promising solution by leveraging knowledge from source domains, but its "black-box" nature raises concerns about model interpretability, hindering practical application. To address this challenge, this study developed an evaluation framework integrating Layer-wise Relevance Propagation (LRP), Integrated Gradients (IG), and SHapley Additive exPlanations (SHAP) to assess feature-level interpretability in transfer learning-based FDD model. We modified the 1D CNN architecture specifically for compatibility with interpretation methods. The model achieves an overall accuracy of 97.01%, with component-level faults showing higher accuracy than system-level faults. All interpretation methods consistently identify physically meaningful discriminative features, and the robustness of the methods is validated through 10 randomized trials. Meanwhile, data volume critically impacts the clarity of interpretation results—larger datasets yield higher feature importance scores, though even 1% of training data is capable of identifying discriminative features. To simulate real-world chiller operation, multiple cross-condition transfer learning tasks were designed, covering a wide range of operating scenarios. Results demonstrate that the Domain-Adversarial Neural Network (DANN) and Fine-Tuning (FT) improve target-domain accuracy by 25% over baseline models while preserving physically meaningful discriminative features from the source domain.

Item Type:	Article
Date Type:	Published Online
Status:	In Press
Schools:	Schools > Engineering
Publisher:	Elsevier
ISSN:	0360-1323
Date of First Compliant Deposit:	10 August 2025
Date of Acceptance:	4 August 2025
Last Modified:	12 Aug 2025 16:30
URI:	https://orca.cardiff.ac.uk/id/eprint/180319

Actions (repository staff only)

Edit Item