(546d) A Hybrid Modeling Approach for Fault Detection and Isolation of HVAC Systems

Authors: 
Hassanpour, H. - Presenter, McMaster University
Mhaskar, P., McMaster University
House, J. M., Johnson Controls Inc
Salsbury, T. I., Johnson Controls Inc
The occurrence of various faults in heating, ventilation and air conditioning (HVAC) systems in buildings may result not only in compromised comfort levels, but also lead to increased energy consumption. This has motivated research on the design of fault detection and isolation (FDI) strategies for HVAC units. One of the approaches for design of FDI strategies utilizes first principles models to generate residuals (in order to capture the difference between observed behavior and expected behavior). In [1], the problem of sensor and actuator fault isolation is addressed by designing a state observer to exploit the analytical redundancy in the system and in [2], this problem is considered for control affine nonlinear uncertain systems. In HVAC applications, first principles models have also been utilized to achieve FDI [3]. Developing and maintaining first principles models in general, and even more so with HVAC systems, remains challenging. This has led to efforts to devise purely data driven fault detection and isolation filters.

The use of data-driven approaches such as artificial neural network, principal component analysis, wavelet analysis, etc., has been increasing significantly in the area of FDI for HVAC systems. In [4], PCA and joint angle analysis (JAA) are utilized to detect and diagnose multiple faults in variable air volume (VAV) systems. In [5], the problem of flow sensor fault detection and validation of VAV systems is considered by employing PCA models. In [6], the PCA and wavelet transform are combined to detect and isolate the faults in HVAC systems. In [7], a causal FDI framework is proposed to diagnose multiple faults in VAV boxes of HVAC systems. Then the results are compared with PCA and joint angle analysis. In [8], The problem of distributed FDI of HVAC system is considered by designing local FDI framework for each unit. In [9], The combination of basic and auxiliary neural networks along with clustering analysis are used for FDI of HVAC systems. The notion of applying hybrid approaches, those that utilize both, a rudimentary first principles model (relatively easy to develop), along with data driven approaches has been gaining significant attention lately [10,11], and could bring potential value to the problem of FDI for HVAC systems.

Motivated by the above, this work addresses the problem of fault detection and isolation by utilizing a hybrid FDI framework. The proposed hybrid FDI framework uses a combination of first principles models along with data driven model to first model the process dynamics. These hybrid models are then used as part of an FDI framework to design residuals dedicated to detecting and isolating faults. The framework is implemented on an HVAC simulation testbed, and results compared with purely data driven approaches as well as first principles model-based approaches.

References

[1] Du, M., Scott, J., & Mhaskar, P. (2013). Actuator and sensor fault isolation of nonlinear process systems. Chemical Engineering Science, 104, 294-303.

[2] Shahnazari, H., & Mhaskar, P. (2018). Actuator and sensor fault detection and isolation for nonlinear systems subject to uncertainty. International Journal of Robust and Nonlinear Control, 28(6), 1996-2013.

[3] Seem, J. E., & House, J. M. (2009). Integrated control and fault detection of air-handling units. HVAC&R Research, 15(1), 25-55.

[4] Du, Z., & Jin, X. (2007). Detection and diagnosis for multiple faults in VAV systems. Energy and Buildings, 39(8), 923-934.

[5] Wang, S., & Qin, J. (2005). Sensor fault detection and validation of VAV terminals in air conditioning systems. Energy Conversion and Management, 46(15-16), 2482-2500.

[6] Li, S., & Wen, J. (2014). A model-based fault detection and diagnostic methodology based on PCA method and wavelet transform. Energy and Buildings, 68, 63-71.

[7] Shahnazari, H., Mhaskar, P., House, J. M., & Salsbury, T. I. (2018). Heating, ventilation and air conditioning systems: Fault detection and isolation and safe parking. Computers & Chemical Engineering, 108, 139-151.

[8] Shahnazari, H., Mhaskar, P., House, J. M., & Salsbury, T. I. (2019). Distributed fault diagnosis of heating, ventilation, and air conditioning systems. AIChE Journal, 65(2), 640-651.

[9] Du, Z., Fan, B., Jin, X., & Chi, J. (2014). Fault detection and diagnosis for buildings and HVAC systems using combined neural networks and subtractive clustering analysis. Building and Environment, 73, 1-11.

[10] Wilson, J. A., & Zorzetto, L. F. M. (1997). A generalised approach to process state estimation using hybrid artificial neural network/mechanistic models. Computers & chemical engineering, 21(9), 951-963.

[11] Ghosh, D., Hermonat, E., Mhaskar, P., Goel, R., & S. Snowling. (submitted 2019). A hybrid modeling approach integrating first principles models with subspace identification, Industrial & engineering chemistry research.