Neural Network Based Fault Isolation in the Air Handling Unit

Various faults may result in poor indoor air quality or more energy consumption of the heating, ventilation and air conditioning systems. The healthy measure, control and operation are essential for the real systems. Generally, two fault detection and diagnosis methods have been developed and widely applied in HVAC systems. One is the model-based, and the other is the data-driven. Each of these two methods has its own characteristic and application condition. The model-based method deeply relies on the accuracy of the mathematic model built. And the data-driven approach requires excellent training data. First of all, the characteristic of the faults in the air handling unit is discussed. And the influences of various faults for the energy consumption are also discussed in this paper. Secondly, A data-driven method based on neural network is presented to detect and isolate the faults occurred in the air handling unit. As a pattern recognition method, neural network is capable of identifying different operation conditions. In addition, the efficiency of the neural network based fault isolation is validated using the simulator based on the TRNSYS platform. Different faults including temperature, flow rate and pressure sensors are tested. Various faults affect the indoor air quality and energy consumption. The healthy measure and control environments are essential for the systems. Actually, the small faults including measure and control problems are difficult to discover. The neural network is the method of pattern recognition. After well training, the neural network can isolate the faults successfully.