Robust fault tolerant application for HVAC system based on combination of online SVM and ANN black box model

Efficient heating, ventilation, and air-conditioning (HVAC) systems are one of the big challenges today around the world. The fault detection and isolation (FDI) play a significant role in the monitoring, repairing and maintaining of technical systems for the final destination of cost reduction. FDI makes it possible to reduce total cost effective of maintenance and thus increase the capacity utilization rates of equipment. Reduction of energy wasting in the system by on time fault detection is another goal. Therefore, this work proposes a new fault detector based on a black box Artificial Neural Network (ANN) model and online support vector machines (SVM) classifier which integrates a dimension reduction scheme to analyze the failure of air fan supply and dampers fault. The key advantage of this algorithm is to make robustness for SVM to recognize a faulty condition with unexpected sensors values. The ANN generates a high accurate model which is based reference for SVM classifier. Now by using this black box model we make possibility of robustness for SVM to increase detection probability. Finally, a series of faulty experimental data are applied to evaluate the effectiveness of the robust classifier. Final results show that online SVM can detect accurately the air supply fan fault and damper fault of a HVAC system with minimum usage data. It is also outperforms offline SVM on such energy systems for classification.