Machine Learning for Diesel After-Treatment Modeling

The massive computerization of our society and the increasingly powerful data collection and storage technology lead to an explosive growth of available data. A big challenge is to extract useful information inside the collections of data and use it to support and improve business processes or to increase knowledge in different human activities. In this context, Data Mining represents an important and promising field. An important application domain is represented by the automotive industry. A current problem is the modelling of emission control component ageing. The after-treatment system is a very complex system, in which each component behaviour is influenced by the status of other components and by other external factors. But also the ageing of one component could be influenced by an anomalous behaviour of another part of the AFT pipe or by external working conditions, such as the climate temperature as well as driving operational conditions. In this scenario, Data Mining is applied in order to develop a classification/regression system able to detect the emission level of the system, starting from data collected by the ECU, modelling the after-treatment exhaust gases system.