Crankshaft speed measurements and analysis for control and diagnostics of diesel engines
2001 (English)Licentiate thesis, comprehensive summary (Other academic)
The increasing demands from governments on the engine manufactures to lower the fuel consumption, lower the exhaust emissions and to reduce the noise have lead to an intensive research in the combustion process. Measurement of the combustion process inside the cylinder is only suitable in laboratory environment due to a number of limitations; the pressure transducer needed to measure the pressure is expensive, difficult to mount in the cylinder and has a limited lifetime that is much shorter than the engine's lifetime. Demands of on-board diagnostics where the combustion process is continuously monitored, on production vehicles have created a need for a method to indirectly measure the combustion process. The two main indirect methods are vibration measurement based reconstruction and crankshaft angular speed measurement reconstruction. The combustion process give rise to vibrations in the engine body that in the former method is measured with an accelerometer and the pressure can be reconstructed by using inverted transfer functions. The idea behind the latter method, the crankshaft angular speed reconstruction method, is that when one cylinder fires the produced torque is higher than the load torque and the crankshaft accelerates. As next cylinder goes into compression the total load torque increases and the crankshaft speed will decrease. This is repeated when the next cylinder fires and the produced crankshaft speed fluctuations will then contain information about the combustion and compression that caused it. In this thesis an indirect method to predict the maximum cylinder pressure is developed based on the crankshaft speed fluctuations combined with neural networks. The speed fluctuations were measured on a 6-cylinder inline diesel engine at 9 speed-load-combinations. A two layer (one hidden and one output layer) feedforward neural network was trained with the backpropagation algorithm. The prediction accuracy for pmax was found to be better than ±5 % at 95%-confidence interval for the validation set. Another important parameter for the engine control and for optimising the fuel efficiency at the same time as the exhaust emissions are kept to a minimum, is the position of the pistons most upper position, TDC (top dead centre). The TDC position is normally measured mechanically with means that need access to the cylinders (the cylinder head has to be removed). This method is time consuming and therefore expensive and because of that not used on production engines. Several indirect methods to measure the TDC- positions have been suggested. Either based on measured cylinder pressures, that again need a pressure transducer mounted in the cylinder, or on the crankshaft speed fluctuations. An indirect method based on the speed fluctuations, that are measured when the starter motor rotates the engine with turned off ignition, is developed. From the measured crankshaft speed fluctuations the TDC-positions can be determined either by curve fitting or with neural networks. The TDC position determined by curve fitting has a bias error, due to the out-of-phase acceleration component in the crankshaft that are induced by the starter motor, but also caused by heat exchange between the compressed gas and the cylinder walls and gas. The results from the neural network were found to be better and the TDC-position for all 6 cylinders was determined within ±0.1 degree crank angle at 95%-confidence interval.
Place, publisher, year, edition, pages
Luleå: Luleå tekniska universitet, 2001. , 50 p.
Licentiate thesis / Luleå University of Technology, ISSN 1402-1757 ; 2001:09
Research subject Engineering Acoustics
IdentifiersURN: urn:nbn:se:ltu:diva-18687Local ID: 9d3b65a0-d3f7-11db-b6e3-000ea68e967bOAI: oai:DiVA.org:ltu-18687DiVA: diva2:991698
Godkänd; 2001; 20070110 (biem)2016-09-292016-09-29Bibliographically approved