The present work shows the procedure realized for the training and validation of an algorithm that predict, the damage in the internal combustion engine’s different components and whose diagnosis uses to generate excessive times of maintenance. Several stages, which show up for the generation of the algorithm music consists in the classification of which ones the data that better they interpret and that tell a failure from other, these failures are related to the calibration between electrodes of the spark plug, the percentage of opening of the injector and the pressure of the fuel pump. Straightaway has trained a machine of classification than, on the basis of saying process of learning, help to predict of adequate way the existence and the position of the damage in point. Finally there appears an analysis of the percentages of reliability of the algorithm realized by means of the use of the SVM on having applied it in the diagnosis of flaws in the engine, where it is observed that it is possible to obtain a reliability of 96.5 %, with a percentage error of 3,448 % corresponding to only one it fails badly classified.
|Translated title of the contribution||Applications of Support Vector Machines in the Diagnosis of Combustion Engines|
|Original language||Spanish (Ecuador)|
|Title of host publication||Desarrollo Tecnológico en Ingeniería Automotriz|
|State||Published - 31 Dec 2017|
CACES Knowledge Areas
- 8145A Logistics and Transportation