AI gives racing-shoe foam a mechanical fingerprint

A September study tested two ASICS midsole foams in tension, compression and shear, using five samples per foam for each loading mode. Similar compression stiffness concealed a much larger difference when the materials were pushed sideways.[1]

An equation that connects the tests

A material law describes how stress changes as a solid deforms. It gives a simulation a way to calculate the forces associated with stretching, squashing or shearing a material. Choosing that law usually involves deciding which mathematical form fits the measurements.[2]

Constitutive artificial neural networks make the form itself learnable. In the method’s foundational study, the network learns an energy function; stress follows from how that energy changes with deformation. The inputs can encode deformation without making the answer depend on the observer’s viewing direction. Physical requirements therefore shape what the network can learn.[2]

This connects measurements taken under different loads through one mathematical description. An equation that fits compression must also account for what happens when the material stretches or shears. The resulting parameters have mechanical meaning, making the fit something researchers can inspect.[2]

For the foams, combining deformation measures improved the tension fits. Penalising unnecessary terms then reduced the fitted equations to three terms with little loss of fit. These were material experiments, without tests of runners.[1]

Report an error or suggest a correction