A material is like a person: if you stretch it for a long time, it remembers and behaves differently. Engineers have spent years fitting formulas to describe such memory. Now a neural network does it on its own.
The new architecture, iCKAN, takes test data and converts it into equations. It captures not only elasticity but also viscosity — how the material flows under load. The method was tested on synthetic data and on the real polymer VHB 4910. The results matched laboratory measurements.
The key strength of iCKAN is that it doesn't just predict numbers; it outputs formulas. It's like a height measurer that derives the rule: "height = age × nutrition." The formulas can be read and verified.
In the future, the method could extend beyond mechanics. By adding data on temperature or humidity, the neural network could discover how environmental conditions alter material properties. This is a step toward materials designed on a computer rather than selected by trial and error.
VHB 4910 is so sticky that it's used as double-sided tape in industry.