Identification and reconstruction of anomalous sensing data for combustion analysis of marine diesel engines

• Published in: Measurement : journal of the International Measurement Confederation Vol. 193; p. 110960
• Main Authors: Ou, Shunhua, Yu, Yonghua, Yang, Jianguo
• Format: Journal Article
• Language: English
• Published: London Elsevier Ltd 01.04.2022; Elsevier Science Ltd
• Subjects: Algorithms; Combustion; Combustion analysis system; Cylinder pressure; Cylinders; Data reconstruction; Diesel engine; Diesel engines; Measurement; Noise measurement; Pressure; Pressure measurement; Reconstruction; Sensors; Signal anomaly identification; Thermodynamic models; Diesel engine; Data reconstruction; Cylinder pressure; Combustion analysis system; Signal anomaly identification
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2022.110960

•Identification and reconstruction method of anomalous sensing data was proposed.•This method is designed based on trigger-sample principle and thermodynamic model.•The anomaly cylinder pressure can be automatically located on a crank-angle basis.•The cylinder pressure is reconstructed without losing valuable sensing information.•The proposed algorithm can be implementable in real-time. The combustion analysis system (CAS) is a highly sophisticated tool for engine combustion supervision, control and diagnosis. The analytic quality of the combustion process is strongly influenced by the accuracy of sensing cylinder pressure. At present, cylinder pressure measurement still faces the challenge of signal contamination by various noises. To improve the analysis quality of CAS, an anomaly identification and reconstruction method for CAS is proposed by combining the trigger-sample principle and the 0D-thermodynamic model. Anomaly identification globally detects sensing data patterns that significantly deviate from the expected behavior, and finely decomposes the raw data into normal and abnormal data. Then, the data is reconstructed by the anomaly reconstruction algorithm. The experimental results show the effectiveness of the anomaly identification and reconstruction algorithm in locating the abnormal cylinder pressure on a crank-angle basis, and in reconstructing the cylinder pressure by rejecting measurement noise without losing valuable sensing information.