Neuronal network for flexible process control in diesel engine

• Autorzy: Sitnik L.
• Konferencja: Kongres Silników Spalinowych (20-23.05.2007; Kraków, Poland)
• Języki publikacji: PL

Abstrakty

EN

Technology of internal combustion engines is still researched and developed. Potential of engine manufacturing is very high and there are a lot of new engines in operation. Conventional fuels supply them first of all. Classical engines will replace with new ones, alternative power sources but the process of exchanging needs time. Combustion engine is equipped with peripheries goods and together are complex control systems. The control system is still developed. Many of them employed neuronal networks. Systems like that are going to be used for control in combustion engines. Neuronal networks have been used to model of engines, to control combustion process as well as diagnostics of engine. The introduced work is a continuation of earlier the author's works connected with implementations of neuronal nets to control the run of internal combustion engines. An application of neuronal nets to control of work of internal combustion engines requires a new approach to philosophy of that control. Results presented in this paper confirm earlier observations, that on the one hand it is possible to map of nowadays engine control unit, but on the other control processes in not satisfy because it is not able to flexible adapt operational conditions of engine changing very fast in addition. It could happen when engine is fuelled by alternative mixture of fuel as well as because of changing degradation level of engine components. The implantation of neuronal nets finds difficulties connected with their design, too. Present paper shows, that using of neuronal nets in diesel engine control is possible on the example of fuel dose control as well as angle of injection advance in each of cylinder separately. It was proved that for engine run control it is possible to use both Multi Layer Perceptron (MLP) and Radial Basis Function (RBF) net. Linear nets are useful in little scale. Particularly useful in control of engine setting, mentioned above, seems to be the RBF nets. Apply of these nets is easily. The optimization their structure is also simply (it restrains practically to selection of quantity of neurons in indirect layer). It does not have also the necessity of selection of function transformations. It is big advantage. It was showed that the RBF net is suitable both to modeling the advance angle of fuel injection and the fuel charge. In both applications the predictions of values of object responses are with high efficiencies. Essential is, that is possible to control operational processes in each cylinder independently. It is one of the requirements of flexible control. It was simultaneously noticed that it is possible to create MLP net, which gives better modeling of angle of injection advance than application of RBF net. On the another way this net gives worse model of size of fuel drop than RBF. It means that modeling of the different processes in engine needs different types of neuronal nets. These nets will be work simultaneously. Mentioned above observations have been found during modeling investigations. The experimental part of research was executed on test bench equipped with model of common rail fuel injection system. The simulations of behaviours of nets were conducted in computer. These conditions are different from natural exploitation ones and because of it is imperative to verify results. On the other hand the results of research are so encouraging to get publish and to continuation of hard working of this project.

Słowa kluczowe

PL

silnik Diesla; silnik wysokoprężny; aplikacja sieci neuronowej; elastyczne sterowanie

EN

diesel engine; flexible control; neural network application

• Wydawca: Polskie Towarzystwo Naukowe Silników Spalinowych
• Czasopismo: Silniki Spalinowe
• Rocznik: 2007
• Tom: R. 46, nr SC3
• Strony: 354--360
• Opis fizyczny: Bibliogr. 14 poz.

Twórcy

autor

Sitnik L.

• Wrocław University of Technology. Poland

Bibliografia

• [1]. Bauer M., Bredenbeck J., Krause F-L., Pucher H., Raubold W.: Online-Prozessoptimierung für aufgeladene Dieselmotoren MTZ Motortechnische Zeitschrift 57 (1996) 6.
• [2]. Kessel J-A., Schmidt M., Isermann R.: Modelbasierte Motorsteuerung, -regelung und -überwachung. MTZ Motortechnische Zeitschrift 59 (1998) 4.
• [3]. Golcu M., Sekmen Y., Erduranli P., Sahir S.; Artificial neural-network based modeling of variable valve-timing in a spark-ignition engine, Applied energy (Appl. energy) ISSN 0306-2619.
• [4]. Manzie C., Palaniswami M., Watson H.; Gaussian networks for fuel injection control Proceedings of the Institution of Mechanical Engineers. Part D, Journal of automobile engineering (Proc. Inst. Mech. Eng., D J. automob. eng.) ISSN 0954-4070.
• [5]. Kovalenko O., Derong L., Javahenan H.; Neural network modeling and adaptive critic control of automotive fuel-injection systems. Proceedings of the 2004 IEEE International Symposium on Intelligent Control 2004, pp. 368-373.
• [6]. Cuiping Z., Qiugfo Y.; Study On Injection And Ignition Control Of Gasoline Engine Based On Bp Neural Network. China Taiyuan 030024.
• [7]. Isermann R., Mueller N.; Modeling And Adaptive Control of Combustion Engines With Fast Neural Networks. Darmstadt University of Technology, Institute of Automatic Control, Laboratory of Control Systems and Process Automation, Darmstadt, Germany 2001.
• [8]. Beuschel M.; Neuronale Netze zur Diagnose und Tilgung von Drehmomentschwingungen am Verbrennungsmotor. Diss. Lehrstuhl für Elektrische Antriebssysteme Technische Universität Munchen, 2000.
• [9]. Bayir R., Bay O. F.; Serial Wound Starter Motor Faults Diagnosis Using Artificial Neural Network. Department of Electronics and Computer, Faculty of Technical Education, Gazi University 06500 Teknikokullar, Ankara, TURKEY, 2007.
• [10]. Sitnik L.: Koncepcja wykorzystania sieci neuronowych do sterowania nastawami silnika. wielopaliwowego Ogólnopolska Konferencja Naukowa KONSSPAL'96, Wrocław 1996.
• [11]. Sitnik L. -Neuronal network for diesel engine control. PTNSS KONGRES - 2005, September 25th - 28th, 2005 Bielsko Biała/Szczyrk Poland, PTNSS P05-C091.
• [12]. Rawski F.; Metody wyznaczania czasowych charakterystyk transportu czynnika roboczego w silniku spalinowym. Rozprawa habilitacyjna. Politechnika Warszawska. Warszawa 1999
• [13]. Zespół sterowania silnikiem ZS typu Common-Rail. Instrukcja do stanowiska modelowego. Wrocław 2003.
• [14]. Mikołajczyk S.; Koncepcja elastycznego sterowania pracą silnika spalinowego. Politechnika Wrocławska, Wydział Mechaniczny, praca dyplomowa. Wrocław 2007.