Material Classification in Laser Cutting using Deep Learning

Industrial laser cutting offers numerous advantages, including high speed, precision, and the ability to process a wide range of materials. However, the versatility of this technology presents a significant challenge when determining the appropriate power and speed for cutting materials. As the laser cutting machine operator selects the cutting parameters manually, confusion can quickly arise when cutting visually similar materials. This confusion can directly affect the cutting quality and result in errors and wasted materials. This study presents an automated material classification approach based on deep learning that utilizes convolutional neural networks (CNN) and laser speckle sensing. A comprehensive dataset of speckle pattern images for the most commonly used laser-cutting materials is proposed in this paper. The dataset is created by capturing the scattered speckle patterns of the material being cut using a USB camera and includes 6000 speckle pattern images. A deep-learning model based on the pre-trained VGG16 model is then trained on the proposed dataset. The performance of the model is experimentally evaluated on a test set of 200 images with varying orientations and lighting conditions for each sample. The F1-score, recall, and precision metrics validate the robustness of the model in distinguishing between the used materials. The results demonstrate the ability of the model to classify the materials, achieving a notably high classification accuracy of 98.1%. This approach provides an automated and cost-effective solution for organizing laser-cutting materials, enabling operators to select the appropriate power and speed for the cutting process based on the type of material being cut.