Sensor Fault Detection And Diagnosis Research In VAV System

Effective fault detection and diagnosis can reduce energy costs, maintaining and improving comfort in occupancy, reducing wear on HVAC (Heating, ventilation and air-conditioning) system and reducing emission of greenhouse gas. Hence Fault detection and diagnosis is important for HVAC system. But it's beyond operator's power to detect and diagnose fault in HVAC system timely and effectively due to the HVAC system getting more and more complicated. Hence, researchers in HVAC attach importance to automation fault detection and diagnosis.Principal component analysis (PCA) is presented in this paper for sensor bias faults detection and diagnosis in HVAC system. PCA models the system by measurement data under normal operation condition, so it don't need to build an analytic model of the system directly. To choose the reasonable number of principal component is a key when build a system model by PCA. The optimal number of principal component is decided by minimizing the total unreconstructed variance in this paper.PCA approach partitions the measurement space into principal component subspace (PCS) and residual subspace (RS). The normal data is included in the PCS and the fault or noise is included in the RS. So, fault can be detected by detecting the projection of the measurement data in RS.The essence of fault reconstruction is a process of seeking an estimation value for correct value corresponding to fault measurement data. And iteration is used for fault reconstruction in this paper. And the nature of iterative reconstruction is a process of sliding the measure to PCS along the direction of fault. For one fault sensor, we can identify the fault by senor validity index (SVI) which is the ratio of squared prediction error (SPE) of its reconstruction data and SPE of its measurement data under normal conditions. Examples show that SVI can identify the fault when there is no collinearity between fault sensors and can not identify the fault when there is collinearity between fault sensors.Finally, PCA approach is proved by examples in this paper. Simulation data is obtained via simulating the VAV system of a building under certain conditions. And then the PCA model can be built by the simulation data. The PCA model can be used to detect, identify and reconstruct the bias fault of sensors in the VAV system.7Results show that the PCA approach is valid.