The Cumulative sum (CUSUM) chart is a general method in statistical control engineering to monitor controlled variables. By comparing the cumulated error between the process variable and the set point with predefined criteria, it is able to detect faults in the system that lead to a small but consistent deviation from the set point. CUSUM chart has been applied to VAV terminal units by NIST for fault detection, known as VAV Box Performance Assessment Control Charts (VPACC) method. This paper develops a causal network based method, which can extend the functionality of the CUSUM method to fault diagnostics, and tests it on the whole secondary systems (including air handling units and terminal units). To use this method, the user needs to specify the causal relationship between all the control variables, and all the physical components that are part of the control loop for each variable. Based on the input, an automatic fault diagnostics algorithm could be applied that determines the possible faulty components in the system. It is realized that the CUSUM method can only detect faults that will lead to control problems, therefore the capability of this method is limited. But the CUSUM method has a strong advantage over other methods, it requires only building control related sensor information, which are the sensors in any building automation control system, and it can be easily adapted if thesystem is changed or upgraded, which frequently happens in renovations.