Automated reset controller design for optimal machine performance

Abstract

In industrial settings that necessitate high precision and speed, particularly within the semiconductor manufacturing industry, linear control solutions are frequently employed due to their intuitive tuning methods, namely through frequency-domain analysis, and their compatibility with tuning based on experimental measurements of the systems, Frequency Response Function (FRF). Nonetheless, inherent limitations exist, such as Bode’s phase-gain relationship and the waterbed effect, which constrain their achievable performance. Consequently, non-linear control solutions must be considered. Reset control presents an interesting solution as it has been proven to outperform linear controllers and also allows for frequency-domain analysis based on FRFs. Nonetheless, this kind of analysis is subject to a set of assumptions. The most recent reset control element developed for broadband phase compensation, the Parallel Constant in Gain Lead in Phase ((P)CGLP), presents several benefits towards meeting the required assumptions. However, tuning a control architecture that includes this element can be a challenging and time-consuming task, as no tuning rules have currently been established. This research focuses on the development of an automatic tuning framework to tune not only control architectures that contain the (P)CGLP element but also any reset control structure. The tuning framework was validated in an industrial wire bonding machine against an optimized linear control solution developed by the manufacturer. An average improvement in settling time of 7.1% was achieved and an average improvement of 13.5% in the RMS value of the error signal across different operational scenarios. The proposed framework proved to be capable of tuning reset controllers that outperform equally optimized linear solutions while maintaining a similar tuning workflow and not requiring additional expertise.