A novel strategy of correntropy-based iterative neural networks for data reconciliation and gross error estimation in semiconductor industry

Data from sensors in the semiconductor industry are often contaminated by random or gross errors. To improve the precision and reliability of process data, data reconciliation and gross error detection (DRGED) carries out corrections to these measurements to alleviate the impact of the errors. Conventionally, data reconciliation methods rely heavily on specific process constraints, which are not readily available in some complex semiconductor processes. Thus, simultaneously capturing process correlations among the measurements and conducting DRGED are important for such an application. In this work, a novel correntropy estimator based iterative neural network (C-INN) for DRGED is developed. The robust correntropy estimator is incorporated to reduce the effect of gross errors. Even without given constraints, this strategy can effectively solve the DRGED problem using the process correlation extracted by C-INN. The performance of C-INN is demonstrated via a numerical example and semiconductor process data.

Automated summary

Data reconciliation and gross error detection (DRGED) is important for improving the precision and reliability of process data in the semiconductor industry. This study develops a novel correntropy estimator based iterative neural network (C-INN) for DRGED, which effectively solves the problem in complex semiconductor processes.