【摘要】 对ARMA过程均值变点检测的广义似然比检验GeneralizedLikelihoodRatioTest(GLRT)方法进行改进.改进的GLRT(IGLRT)方法考虑到,即使变点后的开始1(2,3,…)个残差均值的实现不符合故障特征,但后面的实现较符合故障特征,也有一定的把握认为变点已发生.模拟比较了GLRT和改进的GLRT的平均运行长度.最后,在已知一定先验分布的情形下,给出了GLRT的贝叶斯方法.更多还原

【Abstract】 The Generalized Likelihood Ratio Test (GLRT) method for the detection of the change-point of the mean in ARMA process was improved. In the improved GLRT, even if the first 1 (2,3,...) generalization of the means of residual errors are not in accordance with the fault signature, if the remaining generalization are there is some evidence to claim that the change-point has happened. By the simulation, this paper compared the average run length of the GLRT with that of the improved GLRT. And gave the Bayes GLRT under some prior information.更多还原