基于DKNPE方法半导体蚀刻过程健康状态监视

Health status monitoring based on DKNPE for semiconductor etch processes

针对传统方法对非线性或多模态间歇过程的故障检测率低的问题,提出一种基于K近邻邻域保持嵌入得分差分(difference of K nearest neighbors score associated with neighborhood preserving embedding,DKNPE)的健康状态监视方法。首先,通过NPE方法计算训练数据集的得分矩阵,称其为样本的本质得分。然后,在训练数据集计算每个样本的K近邻均值,并将其投影到低维空间以获得样本的估计得分。接下来,在差分子空间(diffe-rence subspaces,DS)和差分残差子空间(difference residual subspaces,DRS)中分别建立两个新的统计量对样本进行过程监控。将本方法在两个模拟数值例子和半导体蚀刻过程中进行测试,并与PCA、FD-KNN和NPE等传统方法进行对比分析,测试结果验证了该方法的有效性。

Aiming at the fault detection in batch processes with nonlinear or multimodal characteristics,this paper proposed a health status monitoring based on DKNPE.First,it calculated the score matrix of the training data set using NPE method,which was named real scores.Then,it calculated the K-nearest neighbor mean of each sample in the training data set and projected the mean into the low-dimensional space to obtain the estimated score of the sample.Next,it established two new statistics to monitor of samples in DS and DRS respectively.Compared with traditional methods such as principal component analysis(PCA),fault detection based on K nearest neighbors(FD-KNN)and NPE,the results indicate the effectiveness of the proposed method in two numerical cases and semiconductor etch processes.