An innovative methodology for Predictive Maintenance applied to DTC signals in heavy commercial vehicles leveraging telematic data

Knowledge Discovery from Data is one of the major challenge in the Digital Revolution, carried forward within the Industry 4.0 transformation. Role of new technologies is to centralize information and derive data-driven insights to enhance business performance. Derive real value from collected data is the main purpose of Data Analytics, achieved leveraging state-of-the-art techniques of Machine learning. This thesis deals with a relevant issue for automotive industries: Predictive Maintenance. The aim consists in identify in advance eventual failures in order to improve workshop repairs management. This work considers error signals on trucks as subject of the predictive analysis. Primary role is represented by Telematic, a complex system embedded in vehicles that, continuously monitoring their status thanks to numerous sensors, is able to create an increasingly connected fleet. After a preliminary discussion of procedures from the theoretical point of view, the attention is focused on the study of a new data-driven methodology to reach project goals. It starts with data cleaning and pre-processing steps to end with mathematical models performance evaluation, through crucial passes of feature engineering. The entire methodology is then applied to real-life case study, chosen depending on major needs of involved industries: Accenture, consulting partner, and IVECO, commercial trucks leader. Thesis ends with a discussion of obtained results and relative improvements and, in last, a business case proposal is presented as future development of the project.