STRATEGIES IN MACHINE LEARNING

Abstract

Artificial intelligence has achieved great potential in technological development, especially in the optimization of internal combustion engines. This research seeks to forecast the performance of diesel engines using regression strategies in machine learning. The study, with a quantitative and applied approach, collects data from a 30-liter, 1200 HP Komatsu diesel engine through dynamometric tests. Brake power, torque and fuel consumption are measured, monitoring various operating parameters. Using the data, a forecasting model was developed using multiple linear regression in Python. The results show a high correlation between the input and output parameters, highlighting the intake manifold pressure as the most relevant. The predictions reach high R² values: torque (0.96), brake power (0.97) and instantaneous consumption (0.98). The coefficients of the regression model applicable to the input parameters are also determined. In conclusion, machine learning algorithms, specifically multiple linear regression, are effective in predicting the behavior of diesel engines in dynamometric tests.