An application example of CUSUM (Cumulative sum) control chart to condition monitoring tasks for a gas turbine installed in a co-generation plant is presented. In order to suppress unwanted alarm caused by slow drift of variables, an improved CUSUM with variable target is employed. The proposed method requires parameter determination that greatly affects fault detection ability of the algorithm. From the viewpoint of ARL (averaged run length), a parameter setting method is proposed. It is also presented that implementation of regression model that estimates values under normal operation conditions along with the CUSUM technique is effective to take into account external variables dependence such as inlet air temperature that affects greatly on the gas turbine performance.