Optimizing the quality characteristics of glass composite vias for RF-MEMS using central composite design, metaheuristics, and bayesian regularization-based machine learning

Abstract

Technological improvement in micro devices has accentuated the demand for glass and its composites. The μ-ECDM is emerging as an evolutionary technique for glass composite micro drilling, required for glass vias in the packaging of Radio Frequency Micro Electromechanical Systems (RF-MEMS). A response surface-based central composite design and metaheuristic algorithms have been employed to optimize the quality characteristics (in terms of deviation and smoothness) of drilled micro holes in glass epoxy composite. Subsequently, a Bayesian regularization-based Machine Learning (ML) algorithm has been deployed to substantiate the reliability of optimal outcomes from metaheuristic algorithms. At optimal point, process parameters were obtained around 48 V, 48 °C, 3.56 ms, 550 rpm with corresponding optimal outcomes of around 35 % smoothness and 20 % deviation of micro holes. The Bayesian regularization-based ML model validated the optimal outcomes with an insignificant deviation ranging from 0.45 to 2.87 %. Consequently, improvement in quality characteristics at optimal conditions has been conjectured for the industrial feasibility of the process in micro drilling the glass composite vias.