Swarm Intelligence Parametric Yield Optimization of Microwave Devices Using the Non-Linear Partial Least-Squares Polynomial Chaos Expansion Technique

The operation of high frequency devices can be severely deteriorated due to fabrication tolerances, especially when more than one manufacturing process is required. The effects of parameter geometry deviations can be mitigated at the design phase through optimization of suitably selected statistical figures of merit, such as yield. Optimization of yield is normally carried out with the use of the Monte Carlo method. This is a computationally expensive procedure due to the need of multiple full wave electromagnetic (EM) simulations. To alleviate this issue, surrogate modelling techniques are commonly used. However, common surrogate modelling techniques suffer from the curse of dimensionality. This work proposes a technique for optimization of high frequency devices, which relies on the NLPLS-PCE technique. NLPLS-PCE is incorporated in a swarm intelligence framework for the adaptive adjustment of design parameters as well as a convergence safeguard. The yield optimization method is demonstrated using a KA band Ridge waveguide filter terminated with a Vivaldi antenna. The yield specification for this problem includes that the reflection coefficient must be below -15 dB across the passband. The number of statistically independent dimensional variables are 18. Yield improved from 61% to 72.18% after 38 optimization iterations. Reliability and performance of the yield optimization algorithm is verified through EM-driven Monte Carlo simulations.