Previsão de parâmetros para máquina de pintura industrial usando machine learning

Abstract

This final project presents the development and implementation of an artificial intelligence-based system for predicting parameters of an industrial painting machine. The main objective is to optimize the painting process through the accurate prediction of critical variables, such as paint flow rates and equipment speeds, based on input characteristics of the material and the process. The methodology employed involved data preprocessing steps and the division of the dataset into training and testing sets. Subsequently, a Random Forest Regressor model was trained to handle the multioutput nature of the problem. Model performance evaluation was carried out using metrics such as R² (coefficient of determination), MAE (Mean Absolute Error), MSE (Mean Squared Error), and RMSE (Root Mean Squared Error), both for each individual output parameter and in an aggregated manner. The results demonstrate the model’s ability to satisfactorily predict various parameters with accuracy, highlighting the potential of artificial intelligence in improving efficiency and quality in industrial environments. To facilitate the practical application of the model, an interactive Streamlit interface was developed, which integrates the predictions with their respective confidence metrics and provides all detailed information for analysis. Thus, this work not only demonstrates the applicability of supervised learning algorithms for the prediction and control of critical parameters in painting lines, but also provides a practical and effective tool to support decision-making and increase efficiency and quality in industrial processes.