Optimization of Kirigami‐Based Displacement Sensors Using Machine Learning Techniques

ABSTRACT This study presents a machine learning‐driven framework for the optimal design of kirigami‐based piezoelectric displacement sensors. We generated a comprehensive dataset using finite element analysis (FEA) by systematically varying key design parameters of the polyvinylidene fluoride (PVDF) film, including the number of cuts, joint configurations, and joint widths. An artificial neural network (ANN) with optimized hyperparameters is then trained to function as an inverse design model. This model accurately recommends the design parameters required to achieve user‐specified performance targets for two critical metrics: maximum displacement and displacement sensitivity. Experimental validation confirmed the model's high predictive accuracy, yielding coefficients of determination (R 2 ) of 0.980 for displacement and 0.950 for sensitivity. To compensate for the limited amount of data, k‐fold cross‐validation is conducted, allowing us to evaluate the model's robustness and generalization capability. Furthermore, as a practical demonstration, we fabricated composite sensors integrating elements with dual sensitivity settings, enabling sophisticated applications such as multi‐resolution motor control. Ultimately, this approach enables the rapid design of input devices with performance tailored to specific user objectives, offering significant advantages including improved operational efficiency and enhanced precision control.