Predictive Modeling of Part Quality in the WAAM Process Using PCA Model Reduction and Machine Learning

The Wire Arc Additive Manufacturing (WAAM) technology is a directed energy deposition process that uses an arc welding equipment to melt and continuously deposit material in wire form. This work focuses on the study of the WAAM process using an approach based on machine learning and statistical data analysis to solve the problem of weld quality classification. In this study, experiments were carried out to explore the essential parameters of this process, such as welding speed, wire feed rate and electric current, with the aim of producing high-quality weld layers in terms of geometry. The resulting data from the experiments were rich in information. In order to efficiently manage the data collected during these experiments, Principal Component Analysis (PCA) was applied to reduce the dimensionality of the data while preserving its relevance. Then, several machine learning models, including Random Forest, Logistic Regression, Decision Tree, Naive Bayes Gaussian and SVM, were used to predict welding quality based on parameter data generated by PCA. To evaluate and compare the performance of these models, evaluation metrics such as precision, recall, F1-score and confusion matrix were used. Following this comparative evaluation, we identified that the Random Forest model is the most accurate, capable of reliably predicting part quality based on various combinations of parameters. This approach was implemented in python and was also successfully applied to accurately predict the results of the WAAM process.