عنوان مقاله [English]
In some process control applications, the quality of a product or process can be characterized by a relationship between two or more variables, which is typically referred to as a profile. In recent years, profile monitoring of attribute quality characteristics, such as Bernoulli, Poisson and multinomial distributed response variables using generalized linear models, has received a great deal of attention by academic researchers as well as practitioners. In spite of their ability to detect the state of a process, traditional control charts for monitoring profiles are not sufficiently capable to determine the real time of change when a process enters the state of out-of-control condition. In this paper, we investigate the use of likelihood ratio test (LRT) method to monitor Poisson profiles and compare its performance with that of the traditional approach which employs maximum likelihood estimator (MLE) of change point, once an out- of- control signal appears in Hotteling T 2 control chart for monitoring profile parameters. It should be added that this study considers only a single step shift in parameters of the profile. Simulation results indicate that the proposed method outperforms the traditional approach in terms of both average run length (ARL) of detecting out-of-control conditions as well as accuracy of change point estimation. The relatively high dispersion of estimation of small shifts of the parameters is the main consideration of the proposed method which can be attributed to the higher detection speed manifested by shorter average run length of control chart resulting in fewer data points accessible to identify the exact change point. Integrating the LRT method with exponentially weighted moving average (EWMA) chart has proved to be more powerful in detecting small process shifts as well as dealing with other types of process shifts, such as linear trend, which are the main areas proposed in the concluding remarks of the paper.