With the improvement of connectivity technologies, and the increase of data exchange capacity and cloud technologies, the usage of connected vehicle data by automakers is growing fast, and it represents an exciting, multi-faceted and high-growth area in the data analytics development – enabling a myriad of new possibilities, by including but not limited to: predicting failures in parts, accelerating issue detection and resolution, improve design validation, calibration updates over the air, advanced navigation assistance, passenger entertainment. The aim of this work is to present the process, methodologies and tools adopted in the implementation of an unsupervised machine learning solution based on data collected from connected vehicles whose main objective will be support the analysis of usage severity of the engine and transmission mounts parts of the vehicle. Since there is no specific signal or parameter collected directly from mounts parts made available in databases, the method intended to collect relevant powertrain data from connected vehicles, clean and organize it and then use those data as input parameters in associations studies through principal component analysis and clusters identifications to unsupervised machine learning models. Those results can be used to define filters, generate KPIs indicators and relevant data visualizations through dashboards and graphs that can help the analysis of usage severity, maturation and aging of the mounts parts and also in failure mode analysis refinement, improvement actions with suppliers and provide valuable data to the quality and product development engineers.