Modelling and visualisation of laser micromachining of biopolymers using Fibonacci-inspired neural networks

This work explores a data-driven modeling approach for laser micromachining of biopolymers [polylactic acid (PLA)] using Fibonacci-inspired neural networks (FINN). The main goal is to predict key machining outcomes, specifically the Machining Width (MW), Heat Affected Zone (HAZ), and ablation volume from laser parameters, providing a pre-estimate of the appearance of machining patterns. The experimental data collected by randomly varying laser energy and repetition rate, with the laser position, are used as inputs (energy, position, repetition rate) to the FINN. The FINN directly predicts MW and HAZ; ablation volume is subsequently derived analytically from these predictions using a geometric triangular approximation. To understand the network behavior, the predictions are examined under six different activation functions. The results show that the FINN can effectively model the nonlinear relationship between laser settings and machining results, enabling accurate prediction of feature dimensions. The visualization of laser-machining patterns with different hyperparameters and activation functions provides a more useful and potentially applicable application of predictive neural networks. This predictive modeling can aid in planning and optimizing laser processing on biopolymers by anticipating the resulting geometry from given laser parameters.