Abstract
New technologies and calibration to lower engine emissions affect the vibration profile and may lead to premature damage. Therefore, this effect is a key factor to be evaluated during calibration and durability testing to ensure system reliability. Vibration profile monitoring during durability testing allows anomalous behavior to be detected before critical breakage is caused to the test sample, permitting thus an analysis of the root causes of the failures in order to implement countermeasures. The challenge is to develop a method that is able to objectively determine the acceptance criteria with no need for subjective human determination. In this study a six-cylinder diesel engine was used as a test device for the definition of metrics and objective damage acceptance thresholds during calibration and durability. First the engine was tested under normal conditions and then abnormal conditions were induced. Test results allowed the development of different learning algorithms, damage metrics and damage acceptance thresholds based on artificial intelligence, statistical analysis and signal energy analysis. The suitability of each proposed method is analyzed in terms of detection capability, avoidance of false detections and objectiveness of the criteria.