Machine-Learning Guided Strain-Relief Patterns for Maximizing Stretchability of Printed Conductors

There has been an explosion in the application of flexible and wearable electronics in a wide range of fields in the past few decades. When in use, flexible electronics are subjected to a variety of mechanical loadings, including stretching, bending, and twisting. These loadings will introduce damage in the printed materials and will eventually result in their failure. Therefore, it is beneficial to minimize the damage in the printed elements by reducing the mechanical strains without compromising the functionality of flexible electronics. Serpentine structures have been often used to mitigate the maximum strain during stretching, and in a prior publication, we have demonstrated that combining serpentine structure with strain-relief patterns will further reduce strain generated under the same stretching level. Strain-relief patterns represent a novel approach in the field of stretchable electronics. These patterns are cut-outs on the substrate that, when stretched, alter the strain distribution within the critical elements on the substrate. By strategically defining their geometry, dimensions, and locations, it is possible to significantly minimize the maximum strain experienced by these critical elements during stretching. As a result, the overall stretchability of the device is substantially improved. To achieve the optimal strain-relief pattern, machine-learning models have an effective way of identifying the effect of such parameters and help design and locate the strain-relief patterns. As a continuation of our previous paper, this paper explores a wider range of strain-relief patterns and examines their effect under stretching. Further, the circular strain-relief pattern is optimized to ensure the maximum reduction of the strain generated in the conductive ink. This optimization is achieved by conducting a parametric analysis of each strain-relief pattern and strain level via numerical simulation and developing the machine-learning models that correspond to these patterns. Other geometric designs of the strain-relief patterns are also considered as an initiation of future study.