A multi-objective optimization framework for aerosol jet customized line width printing via small data set and prediction uncertainty

Abstract

Abstract Aerosol jet printing (AJP) is a promising non-contact writing technology to fabricate customized and conformal microelectronics devices on flexible substrates. However, in recent years, the printed line quality is highlighted as a limitation in the applications of AJP technology. According to previous researches, a line printed with high edge roughness and low cross-sectional area will reduce the resistance homogeneity and current carrying capacity, respectively. Despite a high line thickness is beneficial to increase the cross-sectional area, it will be in contradiction with a customized line width under a certain mass flow rate, and may lead to an increase in the line edge roughness. Therefore, it is necessary to minimize the inherent contradictions between different printed line features in a design space. In this research, a multi-objective optimization framework is proposed to optimize the overall printing quality of customized line width. In the proposed framework, Latin hyper sampling is utilized for initial experimental design as it could maximize uniformity in a design space with small dataset. Gaussian process regression (GPR) is then adopted for rapid modeling of the printed line morphology due to its capability of providing prediction uncertainty. Following that, GPR models are driven with an efficient multi-objective genetic algorithm to minimize the inherent contradictions of the AJP process. Thus, the optimal process parameters for customized line width printing can be identified systematically and cost-efficiently in a design space. Experimental results indicate the validity of the proposed framework for customized line width printing. Till now, there are few systematic researches on the optimization of printed line morphology, which is an essential component for AJP. This research attempts to contribute to enriching the body of knowledge on printing process optimization.