Optimization of test procedures based on OBD system specific field data

This paper shows the results of the research work with regard to optimization of test procedures based on systematic analysis of field behavior. The presented approach outlines all important steps of the overall process, the statistical methods and algorithms used with the help of an example from the automotive industry. The example is based on the data gathered during various rides with four different vehicles (BMW, Chevrolet, Opel and Peugeot) and stored with a data logger connected to an OBD II port. An important part of the approach is the mathematical and statistical analysis of stored signals, which are divided into two groups with regard to the characteristics of their dynamics. Furthermore, the importance and additional benefit of bi- and multivariate analysis of signals is summarized. The results of the signal analysis with all important information are prepared for further analysis of the distinctive functions of car fleets with the rule learning algorithms. The analysis is performed with three different algorithms (NNge, JRip and J48) which provide slightly different results. The performance of all three algorithms to the applied problem is discussed in detail. Nevertheless, the main result shows that the predefined rule (with 5, 10 and 15 attributes) can be found easily with the NNge algorithm. This means that the specific set of operating conditions (OCs) at which the claimed car operates can be distinguished from the rest of the fleet. The assumption for the set of OCs is that the specific, searched information is forwarded to the OBD II interface and stored by the data logger. The obtained set of OCs forms the basis for the optimization of the test procedures and update of the operating conditions of a certain product. The major steps of the optimization of the testing with regard to the present work are as follows: 1. Collect the adequate OBD data 2. Perform the analysis of all OBD signals in order to obtain the operating conditions of the product 3. Compare the gathered information of claimed and not-claimed vehicles to detect the distinctive features 4. Optimize the test conditions with regard to the obtained operating (test) conditions.