Yield analysis of microsystems is increasingly becoming important due to their decreasing feature sizes. Despite many ways and methods for statistical yield analysis, the contribution of worstcase methods in the yield analysis of microsystems is significant. Worst-case methods, in general, offer advantages like reduced number of simulations and a quantifiable yield metric in yield analysis of a system. However, the common yield analysis problems in MEMS and other microsystems have been identified to be falling into the category of non-linear specification boundaries. In this paper we present an enhancement to the yield analysis of microsystems using worst-case methods. The enhancement has been achieved in terms of accuracy of the parametric yield calculation process by considering non-linear metamodels in place of linear main-effect models. A function suite (WCAS) has been developed in order to implement the worst-case analysis and compare the results of the pre and post enhancements. The demonstrator used for comparing the results of pre and post enhanced worst-case analysis was a U-Shaped thermal actuator.