Artificial neural network modelling of laser micro drilling process

Abstract

Laser micromachining is gaining popularity in precision manufacturing of bioimplants, owing to its ability to create microstructures from difficult-to-machine materials like nitinol. This superalloy, known for its shape memory effect and super elasticity, presents significant challenges during conventional machining because of its low thermal conductivity and tendency to harden during processing. Using input parameters including sheet thickness, laser spot diameter and scanning speed, the study aims to model and predict hole quality, specifically circularity, and heat-affected zone thickness using an Artificial Neural Network Model. Using a nanosecond pulsed Nd:YAG laser, micro drilling was conducted on Nitinol sheets and the hole diameters are measured using a confocal microscope. A feedforward Artificial Neural Networks(ANN) model using Levenberg Marquardt Agorithm was trained for the performance measures of Heat Affected Zone thickness and circularity. With R2 values above 0.98 and Mean Squared Error (MSE) values below 0.01, the model demonstrated high accuracy. This study helps to achieve predictive control in precision production by combining of Artificial Neural Networks(ANN) with laser micromachining. This hybrid strategy could greatly enhance the manufacturing results for biomedical applications like orthopedic instruments, stents, and other implants whose performance is directly impacted by hole geometry.