Machine learning-driven approach for optimization of the fabrication process with additive manufacturing

Additive manufacturing or 3D printing technology reduces production time and costs by integrating with artificial intelligence systems. However, this technology has some shortcomings. This thesis proposes a new machine learning (ML) platform that predicts optimized parameters for the 3D production process using artificial intelligence methods. This platform aims to save time and costs regardless of the product geometry. Additionally, it enables obtaining products with high mechanical properties. The dataset was created using the slicing program of a Fused Deposition Modeling (FDM) based printer. In experiments conducted on 80 models, input parameters such as layer thickness, print speed, infill rate, and print temperature were used, and output parameters such as print time, material consumption, and filament length were measured. A total of 6480 data points were generated with 81 combinations. 75% of the data was used for training and 25% for testing. Four different ML algorithms, including Gaussian Process Regression (GPR), Random Forest Regression (RFR), Artificial Neural Networks (ANN), and Linear Regression (LR), were used to predict the production process. Models trained with MATLAB software were evaluated using performance criteria such as R-squared (R2), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Square Error (MSE). The most successful model was RFR, showing the highest accuracy in predicting print time, material consumption, and filament length parameters. This study provides significant advancements in 3D printing technology by enabling quick and accurate predictions of key parameters in traditional processes. The proposed method can operate automatically without considering the shape, size, and material of the printed object.