非全数据EV模型的参数辨识及可靠性分析

Parameter estimating and reliability analyzing for EV model with incomplete data

对于在测量系统(I)下分别测到n个数据值的n件产品,在测量系统(Ⅱ)下只测到其中的n1件产品,未测到的n2(n1+n2=n)件产品在使用了t1h后,需要在测量系统(Ⅱ)下做无替换定期检测,要求对这n2件产品作出可靠性分类,分类标准是与使用前相比其改变量不超过一个额定值.论文使用期望最大(EM)算法,对这种非全数据的变量含误差(EV)模型的参数给出了一种辨识方法.并给出了估计标准、错分概率及相应的错分代价.用此方法解决了某飞机发动机公司的高压涡轮叶片的寿命分析问题.

For the total number of the products which are surveyed and determined by measuring system (I) and parts of the products are measured by measuring system (Ⅱ), the remains, which are not measured by measuring system (Ⅱ), need to be surveyed and determined in measuring system (Ⅱ) after using certain hours. It is required that reliable classification of the products should be worked out. The criterion is that their increments compared with the products before putting into use is respectively not to exceed a specific value. An estimating parameters method for this incomplete data errors-in-variables(EV) model is given by using expectation-maximum(EM) algorithm. A new identification criterion of reliability for the products is proposed. The probability of wrong classification and the cost of wrong classification are obtained. We apply these results to solving the problem of life measuring of high pressure turbo blade for an aero-engine company.