非对称损失下威布尔分布总体的参数设计

田口玄一的参数设计思想只是针对对称的二次损失函数所做,有一定的局限性.讨论了非对称的二次损失函数,定义了损失系数比.说明了在非对称的二次损失函数下,也可以采用田口玄一减小质量损失的思想:先进行稳健性设计减小波动,再进行灵敏度设计减小偏差.并在指标服从威布尔分布的情形下定义了调整参数,给出了参数设计的方法和步骤.指出了调整参数在参数设计中的特殊位置和重要特性,列出了部分最优调整参数的值.

Parameter Design with Asymmetric Loss Function under Weibull Distribution

Mr.Taguchi used parabola to describe the quality loss in his parameter design theory. Asymmetric loss function is discussed instead of parabola, and loss ratio is defined to describe the degree of asymmetry. It's proved that using the above asymmetric loss function, Taguchi's ideas of parameter design still works: first reduce the variance, and then reduce the bias. Under Weibull population, 4 steps are given to decrease the total quality loss, and adjustment parameter is also defined and used. Some of the best adjustment values are calculated.