Development of a Neuro-Fuzzy Diagnostic System Mathematical Model for Internal Combustion Engines

Author Name(s): Lenar A. Galiullin*, Rustam A. Valiev, Lejsan B. Mingaleeva
Author Email: galilenar@yandex.ru

Abstract

The methods are given for an adaptive engine control system creation in this paper for tests and diagnostics based on a neuro-fuzzy system, which made it possible to simplify the elements of an automated system structure for the testing and the diagnosing of internal combustion engines. The system created provides a more accurate method of controlling certain parameters of the engine, which improved the quality of control when it was diagnosed and allowed to reduce the volume of the knowledge base.

In order to control the engine during its testing and diagnostics it is proposed to use the methods of neuro-fuzzy inference, widely applied in the development of intelligent systems. In order to describe an engine management, expert knowledge is used instead of mathematical models.

The control of internal combustion engines (ICE) in ASID based on a neural-fuzzy system is founded on a knowledge base containing a number of fuzzy rules automatically generated by a neural network. The filling in of this knowledge base is also possible by direct measurement method, but it is more appropriate to fill in the knowledge base with a neural network.

The use of a neural-fuzzy system in ASID allowed the introduction of the initial information – the knowledge base of fuzzy control rules into the control system. The proximity of the form of control rule forms to natural language makes it possible to work with the knowledge base for test technologists.

Introduction

Fuzzy logic systems and neural networks are the most important application of intelligent systems. The functioning of such systems differs from the functioning of classical systems [1].

The main goal of the neuro-fuzzy system is to control the external impact, in which an object is controlled by fuzzy rules [2].

The control system of an internal combustion engine based on the neuro-fuzzy system can be represented as a set of modules (Fig. 1). The ASID of internal combustion engines includes the following modules: “Program’s History Module”, “Learning Module”, “Knowledge Base”, “Testing program’s creation module”, “Control Unit”, “Execution Units Module”, “Testing Programs Database”, “Engine”.

A technologist creates a knowledge base for fuzzy control rules, on the basis of which a control program is created using ASID ICE [3].

A user creates the sequences to change the parameters of interest in the “Testing program’s creation Module” that are recorded from time to time in the “Testing Programs Database” unit. In this multi-level database the input images are transformed sequentially into control actions through a fuzzy output system, which is located in the “Testing program’s creation Module”. In order to obtain control values, the fuzzy system applies fuzzy control rules and linguistic variables that are in the “Knowledge Base” unit. Eventually, the sequence of control actions is created, which is stored in the Testing Programs Database. Based on the received data, the “Control Unit” controls the ACID ICE directly, for example, changes the amount of fuel supplied to an engine or external load. During the tests with the “Execution Units Module” information is generated about the current status of a control object, for example, about the crankshaft speed, torque, fuel consumption, etc. The listed data is stored in the “Program’s History Module”. The incoming data are also directly transferred to the “Control Unit” from the sensors, which is necessary to prevent emergencies. The ASDIC ICE stops its operation in the event of an engine speed sudden increase, oil temperature increase or in other cases.

In “Learning Module” feedback is received from the “Program’s History Module”. This feedback is necessary to implement the principle of adaptive management, as well as to fill in the knowledge base in which the management rules are located. These rules can also be created or adjusted on the basis of direct measurement method by the analysis of the information coming from the “Program’s History Module” in the “Learning Module”.

Summary

Thus, the application of a neuro-fuzzy system will shorten the time of the knowledge base development and will increase the adequacy of an engine control at different modes with the increase in the number of carried out engine tests. The use of the direct measurement method will also make it possible to apply the already available knowledge base of the tested engine to another engine of the same series.

The result of the work carried out is the software development for ASID engines.

Acknowledgements

The work is performed according to the Russian Government Program of Competitive Growth of Kazan Federal University.

References

  1. M. Tian. Fuzzy neural network diagnose expert system of engine // ICICIP          2012 – 2012 3rd International Conference on Intelligent Control and Information  Processing, 2012. – pp. 154-156.
  2. Z.T. Yao, H.X. Pan. Engine fault diagnosis based on improved BP neural network with conjugate gradient // Applied Mechanics and Materials, 2014. – №536, pp. 296-299.
  3. Valiev R.A., Galiullin L.A. Automated system of engine tests on the basis of Bosch controllers // International Journal of Applied Engineering Research, 2015. – 10 (24). – pp. 44737-44742.
  4. Galiullin L.A. Automated test system of internal combustion engines // (2015) IOP Conference Series: Materials Science and Engineering, 86 (1), art. no. 012018.
  5. Valiev R.A., Galiullin L.A. Modeling of internal combustion engines test conditions based on neural network // International Journal of Pharmacy and Technology, 2016. – No 8 (3). – pp. 14902-14910.
  6. N. Deng, C.-S. Jiang. Fault diagnosis technology based on the fusion of neural network and fuzzy Logic // 2012 International Conference on Systems and Informatics, 2012. – pp. 419-422.
  7. Z.T. Yao, H.X. Pan. Engine fault diagnosis based on improved BP neural network with conjugate gradient // Applied Mechanics and Materials, 2014. – №536-537. – pp. 296-299.
  8. Zubkov E.V., Galiullin L.A. Hybrid neural network for the adjustment of fuzzy systems when simulating tests of internal combustion engines // Russian Engineering Research, 2011. – №31 (5). – pp. 439-443.
  9. R.A. Valiyev, L.A. Galiullin, A.N. Iliukhin. Design of the modern domain specific programming languages // International Journal of Soft Computing, 2015. – №10 (5). – pp. 340-343.
  10. R.A. Valiev, A.Kh. Khairullin, V.G. Shibakov. Automated Design Systems for Manufacturing Processes // Russian Engineering Research, 2015. – №35 (9). – pp. 662–665.

653 total views, no views today

Download PDF File

About the author: admin