Bibliographic info
Building Simulation, 2007, Beijing, China
VAV system is a very complicated one in airconditionging systems, thus automatic control become the key of such a system. As necessary components in automatic control system, sensor has failure risk. It is so expensive that detect sensor fault by hardware redundancy in comfortable air-conditioning system. This paper presents an approach, Principal Component Analysis (PCA), to detect and identify sensor fault in VAV system. The PCA model partitions the measurement space into a principal component subspace (PCS) where normal variation occurs, and a residual rubspace (RS) that faults may occupy. When the actual fault is assumed, the maximum reduction in the squared prediction error (SPE) is achieved. A fault-identification index was defined in terms of SPE. Some examples were provided to prove this method is feasible. This paper also presents a fault reconstruction algorithm to reconstruct the identified faulty data.
