Advancing HVAC fault detection using integrated dynamic system modeling, fault simulation, and deep learning framework
- Publication Type:
- Thesis
- Issue Date:
- 2025
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Heating, ventilation, and air conditioning (HVAC) systems are frequently affected by operational faults, which can increase energy consumption and degrade occupant comfort.
Despite various fault detection and diagnosis (FDD) strategies having been developed, there remains a gap in comprehensive approaches that integrate HVAC system modeling, fault simulation, and detection with advanced deep learning methods, particularly in overcoming the challenge of obtaining diverse real-world operational data.
This study addresses these gaps by developing an integrated framework that combines system modeling, fault simulation, and deep learning-based detection methods. Firstly, the dynamic simulation model is developed to investigate across various severity levels of operations for both fault-free and faulty conditions. Secondly, a novel hybrid approach for fault detection and diagnosis (FDD) is proposed by integrating random forest (RF) and support vector machine (SVM) classifiers. The proposed approach significantly reduces the number of sensors necessary for efficient fault identification across the entire HVAC system, thus enhancing practical application.
Additionally, a one-dimensional convolutional neural network (1DCNN) based FDD system is developed, enhancing fault detection by eliminating the need for manual feature extraction.
This proposed approach not only streamlines the fault detection process but also reduces the complexity associated with manual feature engineering, thereby enhancing classification accuracy, efficiency, and reliability. The 1DCNN-based FDD system demonstrates its effectiveness in accurately categorizing nine major HVAC faults and normal conditions without the need for additional data processing or feature engineering.
Finally, the advanced HVAC fault detection framework is developed using simulated HVAC fault operational scenarios with Gramian angular field (GAF) and two-dimensional convolutional neural networks (GAF-2DCNNs) to provide a robust and proactive solution. The GAF transforms time series data into visual representations by encoding pairwise angles between data points, which are then input into a CNN for further analysis and fault detection.
In summary, this research demonstrates the effectiveness of the proposed framework through rigorous validation and evaluation. The integrated approach achieves high accuracy in fault detection, highlighting its potential for practical application in HVAC systems.
Future research could focus on enhancing model robustness under diverse environmental conditions and hybridization with complementary fault detection methods to further improve performance.
Please use this identifier to cite or link to this item: