A multi-stage geometric approach for sensor fault isolation on engine test beds

Abstract The paper presents a new model based method for sensor fault isolation on engine test beds. The proposed multi-stage geometric approach is based on the geometrical analysis of error functions, which are obtained by Taylor series approximation. The geometric analysis includes intersection analysis and distance evaluation resulting in fault probability values for individual sensors. The method is able to process any type of static models, whether they are physics-base or databased, and whether they are available in the form of an equation or in the form of an inequality. Thus the method has very high general validity and is not limited to any specific application or test engine. Validation by a simulation study with measured data from a single-cylinder research engine shows good isolation results, especially with regard to the robustness against model errors.