论文部分内容阅读
固态介质击穿寿命特性通常用威布尔分布来描述,形状参数β反应了固态介质的失效特征,因而需要精确估计β值。提出了在小样本情况下基于最小二乘支持向量机(LSSVM)的参数评估方法,并给出了LSSVM在MOS电容与时间有关的击穿寿命分布评估中的应用实例,并与常规的最小二乘评估方法相比,得到的结果表明LSSVM的评估精度更高(均方误差更小)、鲁棒性更好,在小样本情况下能更精确地确定威布尔分布的形状参数。
The breakdown life characteristics of solid-state dielectrics are usually described by the Weibull distribution. The shape parameter β reflects the failure characteristics of the solid-state medium, and thus the β-value needs to be accurately estimated. A method of parameter estimation based on least square support vector machine (LSSVM) is proposed in the case of small samples. The application example of LSSVM in time-dependent breakdown lifetime distribution of MOS capacitors is given. Compared with the conventional minimum two Compared with the evaluation method, the obtained results show that LSSVM has better estimation accuracy (smaller mean square error) and better robustness, and the shape parameters of Weibull distribution can be more accurately determined in the small sample case.