Design and development of a fault monitoring system for automotive vehicles

The work described in this thesis is the entire development process, from design to actual code implementation, of a fault monitoring system for automotive vehicles, based on a cloud solution. In the last years, data produced across the world have been grown by 90%: nowadays, companies produce a huge amount of data in a really short time and for this reason they need solutions able to manage huge moles of data, while scalability and performance efficiency must be guaranteed. For what concern big data and automotive industries, connected vehicle technologies have emerged as one of the trend topics across automotive industries. There is an increasing global growth of vehicles equipped with on-board diagnostic sensors, radars or sophisticated telematics services, communicating more and more often with the surrounding world. This project is born from this necessity of collecting telematics data, coming from a certain category of vehicles, elaborating it and producing advanced insights on a flexible and versatile platform. The aim of these advanced insights is to produce as output the current overall status of a single vehicle, monitoring its latest movements and activities or detecting and managing eventual faults. Raw data, coming directly from telematics box, initially go through a complex ingestion process, which includes some data quality techniques and the identification of different data flows (depending on the type of information raw messages contain) . At the end of this process, data are stored into tables, one for each identified data flow, on a SQL database. Considering the huge amount of data to deal with and limit size of SQL databases, the need of moving data on cloud has arisen quickly and this has been done by means of Databricks services. Ingested data are the input dataset given to the main core of the project, which is the rule engine; rule engine is the application that combines this input data with set of rules, producing as output every information related to a certain vehicle status, including possible faults. Rules are set of conditions based on data parameters; if a rule is triggered, so if all expressions defined in that are satisfied from input data of a certain vehicle, it means that a fault is detected for that vehicle. It is also possible that more than one rule is triggered for the same vehicle. So, the rule engine is the system component which makes an association between input vehicles data and rules, evaluating rules expressions for each vehicle and eventually associate to it a fault of different severity. The output generated by rule engine is finally shown in a web interface, through which an end user is able to verify and monitor current vehicles status, in terms of geographical location, faults and movements.