Performance prediction of magnetorheological fluid‐based liquid gating membrane by kriging machine learning method

Abstract

Abstract A smart liquid gating membrane is a responsive structural material as a pressure‐driven system that consists of solid membrane and dynamic liquid, responding to the external field. An accurate prediction of rheological and mechanical properties is important for the designs of liquid gating membranes for various applications. However, high predicted accuracy by the traditional sequential method requires a large amount of experimental data, which is not practical in some situations. To conquer these problems, artificial intelligence has promoted the rapid development of material science in recent years, bringing hope to solve these challenges. Here we propose a Kriging machine learning model with an active candidate region, which can be smartly updated by an expected improvement probability method to increase the local accuracy near the most sensitive search region, to predict the mechanical and rheological performance of liquid gating system with an active minimal size of experimental data. Besides this, this new machine learning model can instruct our experiments with optimal size. The methods are then verified by liquid gating membrane with magnetorheological fluids, which would be of wide interest for the design of potential liquid gating applications in drug release, microfluidic logic, dynamic fluid control, and beyond.

Keywords

• active candidate region techniques
• artificial intelligence
• Kriging machine learning method
• magnetorheological fluid‐based liquid gating membrane
• rheological and mechanical model