Novel Method for Determining Internal Combustion Engine Dysfunctions on Platform as a Service

Abstract

This article deals with a unique, new powertrain diagnostics platform at the level of a large number of EU25 inspection stations. Implemented method uses emission measurement data and additional data from significant sample of vehicles. An original technique using machine learning that uses 9 static testing points (defined by constant engine load and constant engine speed), volume of engine combustion chamber, EURO emission standard category, engine condition state coefficient and actual mileage is applied. An example for dysfunction detection using exhaust emission analyses is described in detail. The test setup is also described, along with the procedure for data collection using a Mindsphere cloud data processing platform. Mindsphere is a core of the new Platform as a Service (Paas) for data processing from multiple testing facilities. An evaluation on a fleet level which used quantile regression method is implemented. In this phase of the research, real data was used, as well as data defined on the basis of knowledge of the manifestation of internal combustion engine defects. As a result of the application of the platform and the evaluation method, it is possible to classify combustion engine dysfunctions. These are defects that cannot be detected by self-diagnostic procedures for cars up to the EURO 6 level.

Different Induction Motor Faults by New Proposed Random Forest Method

“…The results were compared with different classification algorithms such as regression and SVM. Harach et al, proposed a model for the EU25 inspection station to detect emission measurements [14]. In this model, different vehicle samples were tried to be determined with 9 static test points, constant engine load and engine speed using machine learning algorithms.…”