Data-driven monitoring and validation of experiments on automotive engine test beds

作者：Haghani Adel^*; Jeinsch Torsten; Roepke Mathias; Ding Steven X; Weinhold Nick

来源：Control Engineering Practice, 2016, 54: 27-33.

DOI：10.1016/j.conengprac.2016.05.011

• Data-driven
• Fault detection
• Fault isolation
• Engine test bed
• Multimode systems

Engine test beds are widely used to estimate automotive engine parameters and design controllers in the preliminary development phase. The controller parameters are optimized to fulfill emission, fuel consumption and driving comfort requirements and they will be further validated on chassis dynamometer and road driving experiments. It is common that the results of two experiments deviate, due to some external disturbances or faults. The main purpose of this paper is to demonstrate the application of data driven fault diagnosis techniques to detect the deviations in the experiments and analyze their root causes to reduce the costs and time of the engine design and its control concept. To this end, two different methods are introduced for detection of the problems in the experiment. Based on the results of the detection step, a fault isolation technique has been proposed to support test engineers in finding the cause of the deviations. The results have been demonstrated on an industrial engine test bed and the effectiveness of the methods is discussed.

• 出版日期2016-9