Data-driven analysis of the effects of microtextured surfaces on friction reduction for plastic syringe applications

Abstract This study explored the use of machine learning to optimize low-friction microstructures for plastic syringe applications, eliminating the need for silicone oil. Machine learning was employed to analyze available experimental data collected from the literature and identify key microstructure features affecting the coefficient of friction (COF) reduction. An artificial neural network (ANN) was used to analyze how the features affect COF reduction. The contact pressure primarily influenced the magnitude of % COF reduction, with higher contact pressure leading to a decrease in % COF reduction. A lower pitch increased % COF reduction due to a smaller contact area. Microdimples were generally more effective at reducing friction than micropillars or protruded structures. Two-photon polymerization (TPP) was employed to fabricate microdimpled prototypes, and friction tests validated the ANN predictions. Experimental validation demonstrated up to 57% friction reduction on microdimpled surfaces, with pitch and aspect ratio identified as the most critical factors. While some discrepancies were observed between ANN predictions and experimental outcomes, the machine learning model effectively highlighted the relative significance of different factors. This study demonstrates the potential of combining machine learning with advanced manufacturing techniques to enhance the performance of microtextured surfaces for friction reduction.