Machine learning-based optimization of the design of composite pillars for dry adhesives

A machine learning-based optimization strategy was used to find optimal designs for adhesive composite pillars with either high adhesion strength or tunability. Neural networks were trained data generated by finite element analysis predict the of different designs; an average prediction error less than 1% achieved. Through a sensitivity study neural networks, it is found that geometry stiff core above critical cut off height has no effect on interfacial stress distribution and pillar. randomly initialized constrained algorithm then applied pillar design. design enlarged tapered flat end realizing robust adhesion, since can achieve under loading contact conditions. The optimized normal detachment force nearly 11 times homogenous 1.7 simple wide rectangular loading. thin shows highest effective difference between being loaded moment.