Development of a smart system for gasoline car emissions diagnosis using Bayesian Network

Abstract

A vehicle exhaust emissions test is an activity carried out to determine the content of the remaining combustion products that occur in the fuel in the vehicle engine. Many people do not understand exhaust gas content from emission tests, so to make this easier, this study aims to create a smart application that can diagnose vehicle emissions quickly and accurately using the Bayesian Network (BN) algorithm. Application development begins with BN modeling using the MSBNx application until the appropriate results are achieved. Validation of the BN structure that has been designed with various inputs is carried out to ensure that the BN modeling is correct. The next step is to compile the BN modeling algorithm in the MATLAB application so that it becomes a system that can process input in the form of measurement results for Toyota car emissions. The new BN model for vehicle emission gas diagnosis has been successfully constructed. The results of the system reading when there is an HC content of 217 ppm, the probability value of bad emissions increases to 63.5%. Of the 10 tests performed, the system was able to diagnose them all correctly.