Enhancing production planning in metal-mechanical industries: statistical analysis and machine learning approach for predicting machine failures

Metal-mechanical industries have been implementing Internet of Things (IoT) sensors on machines to generate data that can support decision-making. However, challenges in analyzing the big data produced for predicting failures that may impact production planning are still observed. This work explores the application of statistics and machine learning in the analysis of big data produced by a metal-mechanical industry to explain the interactions between input variables and to predict, based on these variables, the different types of machine failures that negatively affect production planning. It is concluded that torque generates increase in temperature is being the main cause of tool breakage leading to unexpected machine stops. Thus, this study innovates when analyzing big data from a metal-mechanical industry using statistical analysis and machine learning combined with the SHAP (SHapley Additive exPlanations) technique for predicting machine failures. Additionally, this study guides managers on how to use computational techniques to support decision-making in operations and encourages investments in implementing IoT sensors and Artificial Intelligence.