Uma abordagem interativa para auxiliar no diagnóstico automotivo

Abstract

The goal of this project is to develop an approach to assist automakers in obtaining information about the problems and failures that occur in the vehicles during their use by drivers, relating the data obtained by reading the telemetry of the vehicle, as well as the perceptions of the driver about a particular failure. The exploration of techniques to optimize automotive diagnosis is a problem to be discussed in Engineering. Diagnostic techniques should support the wide range of systems that will emerge in the coming years. Thus, it is also possible to use the connectivity that mobile phones and vehicles have today, in order to evolve the product offered and improve the customer experience while driving. The customer inclusion in the diagnostic process aims to allow engineers identifies points to be improved in the car, even if they dont present an apparent error. The customer opinion should be considered, since he/she is included in the process as a new "sensor"(the most intelligent and important of all) able to report his/her perceptions. Considering this fact, the motivation of this research is: (i) search for alternatives to efficiently apply vehicle connectivity in the diagnostic process, (ii) enable automakers to obtain more concrete information on the vehicles they sell. The object is provide an interface between the user and the car helping them to perform a more complete preventive diagnoses than only using telemetry. In this project it is sought that there is a greater interaction with the driver, and the experience of the automakers could be included the diagnostic process. Even if the customer isnt so accustomed with the operation of the vehicle, the system must guide the driver by the process. The customer input data is performed by text or voice command. The analysis of this customer input data, a data mining study was performed based on a base that automaker already had. These bases are telephone interviews in which the automaker asks the customer for perceived problems in the car, and then the answers are transcribed to the base. The data mining study was performed to create one classifier which was responsible for receiving the input data from the customer and sorting it into a particular problem related with the multimídia or related to the motor of the vehicle. The software contained in the mobile phone will be responsible asking to the vehicle sensor system for the readings necessary for the customer can to obtain the information he/she wants. An experiment was carried out in which diagnostic analysts answered a questionnaire that sought to analyze whether the proposed system actually influenced the diagnostic process, so the solution of the problems in the vehicle was performed on less steps compared to the current process. The results showed the approach reduced the number of steps required to obtain the correct diagnosis. Of the three case studies performed, the steps required to arrive at appropriate diagnosis decreased in one step to the first case study, one step to the second case study decreased, and three less steps to the third case study . With these results it was possible to demonstrate that the proposed approach influenced the diagnostic process currently applied by the automakers and concessionaires analyzed.