vehicle tests with EGR leakage; first results
- May 9, 2014
- Marcel Romijn
BRACE Automotive is cooperating with the HAN University of Applied Science. In this project the effects of an EGR (Exhaust Gas Regeneration) leakage of a Diesel engine are measured. For BRACE this data is used to validate and further improve data analysis methods as used in customer projects on the topic of OBD. For the HAN this data is used to improve the knowledge level on emission control systems and the failure mode effects in these systems. See also the earlier blog posts on this topic.
Based on the EGR leakage as defined in previous emissions test (see also previous blog-posts) a number of vehicle tests have been performed. During these tests data was collected while driving on a closed test track to verify that the effects of EGR leakage are visible and can be detected during normal and frequently occurring driving conditions.
Data was collected using the Citroën workshop tool. This tool communicated with the car using an OBD protocol. As this car is already somewhat older the communication protocol isn’t very modern as it still uses a K-line instead of a CAN-bus for diagnostic communication. The most noticeable effect was the limited sample rate and number of logged parameters.
The tests that we wanted to perform consisted of steady vehicle speed runs of 30s to 1min and of typical accelerations such as 50-80km/h. The accelerations were performed with several accelerator pedal positions. For the driver to be able to do so in a repeatable way an accelerator pedal position display was created. To do so the CAN-bus between engine and transmission was tapped into and with some reverse engineering the throttle pedal signal was isolated. Using a script in the CAN bus analysis program, developed by a student from the HAN, a visual display was created.
The data collected was analyzed. As an example the 30km/h steady speed driving is used. Plotting against time shows that the EGR valve command (front and rear valve are controlled by the same command) increases when driving with the leakage. It is also visible that the driver had a slight change in throttle pedal (due to a bump in the road) that caused a drastic effect. One could say that this car is “nervous” in throttle response.
Clearly the vehicle’s EGR control system senses that the proper required EGR flow is not achieved when leakage is present and decides to open both EGR valves more to compensate. In order to get a good estimate on whether the data could be used to design a diagnostic for EGR leakage, the data is checked for being normally distributed.
The data of the “Good” system seems not to be fully normally distributed, although further investigation in different mathematical checks might prove it is but just on the borderline. Some noise is occurring here as well based on the nervous response of very small throttle variations.
The data on the “Fail” system is a bit more conclusive in being normally distributed although the clear drop in EGR command that is visible in the “time” based plot also shows up in the normal distribution plot. By improving data sampling enable conditions proper normally distributed data is what is left and with that a diagnostic could be developed. Driving on Cruise Control could limit the noise generated from the throttle pedal. Adapting the enabling conditions is a difficult decision typically as the data for future possible diagnostics design must be collected frequently and for all types of drivers. Any restriction in reducing the amount of data that can be collected opens the possibility that certain combinations of drivers and environments are completely excluded. This balance is a typical struggle in the OBD field.
In a follow-up of this section of the project more analysis will be performed.