MODELING AND SIMULATION OF SMART AUTOMOTIVE COOLING SYSTEM. Karpagam College of Engineering, Coimbatore.

Transcription

1 Volume 118 No , ISSN: (on-line version) url: ijpam.eu MODELING AND SIMULATION OF SMART AUTOMOTIVE COOLING SYSTEM 1 R.Arul Murugan, 2 R.Mahesh, 3 C.Krishnaraj, 4 T.Soundharya 1,2,4 Assistant professor, 3 Professor, 1,2,3,4 Department of Mechanical Engineering, Karpagam College of Engineering, Coimbatore. Abstract: Cooling of Internal combustion engines are generally done by engine coolant fluid flowing through the engine block and transfers of heat to the atmosphere takes place in radiator. The heat from the fluid is transmitted to the atmospheric air by the radiator, thereby chills the fluid and the chilled fluid decreases the temperature of the engine. In a conservative cooling system, water pumps are used. Such a system is fully driven by engine output adding an additional load to the engine. The integration of artificial neural network in automotive cooling system can improve coolant temperature regulation and servomotor power burning ups and the system can react for unpredictable atmospheric changes. Artificial Neural network based cooling systems are put up to characterize the dynamical conduct of superior automotive cooling systems. In this system, cooling process is accomplished by the usage of electro-mechanical components. In the proposed system water pump and radiator fan were integrated with computer controller and also variable position smart valves were employed to enhance the efficiency of the system and thereby reducing the consumption of fuel, freeloading losses and the emission too. In sophisticated automotive cooling systems, the electrical and hydraulic actuator substitutes the fan in radiator and water pump driven through mechanical means. The engine speed, coolant temperature and atmospheric conditions are closely monitored by sensors and its output is given to the microcontroller. The adaptive cooling system can be modified to air cooled systems. The variable speed cooling fan can plays an imperative role in this system. Keywords: fuel consumption, effective cooling, minimized cooling losses, automotive cooling system. 1. Introduction Superior automotive cooling systems preserve considerable improvement on the response of diesel engine by healthier improvement on its temperature regulation, reducing the consumption of fuel and decrease freeloading losses beside with tailpipe discharge [1-3]. Integrated computer-controlled circulation pump, cooling fan, and variable point elegant valve into the engine cooling system enhance the system efficiency when compared with the conservative cooling system [4, 5]. A drop of 5% fuel consumption by engine and a drop of 10% and 20% in HC and CO through exhaust pipe emissions can be attained through replacing mechanical cooling system components with electrical cooling system components. In superior cooling systems, the circulation pump and cooling fan are substitute with actuators driven through electrical means [6] or by actuation through hydraulic [7] for a lone loop cooling system [8] or for a cooling system with multi-loop [9]. The foremost loop of cooling must make certain that the excess heating of engine block to be avoided which leads to boiling of coolant. Likewise, an additional heat exchanger used to cool the transmission oil generally situated within the radiator. Multilayer feed forward technique system in Artificial Neural Network system is utilized to guess the dynamic performance of the cooling system. The main purpose is to efficiently perform the cooling procedure. 2. Literature Review Research finds its height on the improvement of the conventional cooling system and its components. No additional investigation is able to done on these systems. For a better cooling performance and diminished freeloading losses beside with tailpipe discharge can be attained by electro-mechanical means which could be controlled through artificial neural networks rather than depending on mechanical systems. Through Neural Networks, a nonlinear architecture can be developed because they illustrate better ability in modeling nonlinear controllers. Modeling of automotive cooling system based on ANN, with back propagation learning method would helps us to calculate specific fuel consumption and temperature of exhaust gas for a diesel engine with variable attributes of the cooling system. 3. Automotive Cooling System Three main components of automotive cooling system are Radiator, Cooling fan and feed water pump. The 1671

2 flow of coolant is as shown in the Figure 3.1. The coolant with absorbed heat from the engine is made to pass inside the radiator. Heat exchange to the atmosphere occurs in the radiator. The stream of coolant executed by water pump. The heart of the cooling system is the water pump. The Main criteria for an engine cooling system must maintain the engine surrounded by working temperature. The preferred temperature maintains the engine in its utmost efficiency. The variations in the temperature would affect the drivability, and emissions. Geels et al. [3] reported that about 65% of the engines heat is removed with exhaust gas, conducted by metal parts etc., only 35% is removed by cooling system. 4. Artificial Neural Network no sole prescribed definition for what an artificial neural network is. Commonly, a group of arithmetic replica might be identifying as "neural". A group of adaptive weights, i.e. arithmetical factor are tuned by a skilled algorithm, and are proficient of resembling nonlinear functions of their inputs. The adaptive weights are theoretically link powers among neurons, which are triggered at some point in training and prediction. Neural networks identical to biological neural networks in achieving functions jointly and in analogous by the units, to a certain extent there being a clear explanation of subtasks to which a mixture of units are allocated. The term "neural network" frequently refers to models engaged in data, artificial intelligence and cognitive psychology. The features of Artificial Neural Network, 1. Group of dispensation units 2. Formation state of each unit 3. Connection among each unit 4. Propagation law 5. External input 6. Method of information gathering 7. An environment in which the system can operate. 5. Methodology Figure 1. Artificial Neural Network Artificial Neural network, a simple processing unit can be communicated by distribution of signal through huge amount of biased connections. They inspire human brain. It consists of a processing units and connections (weights). The knowledge to ANN system is stimulated with precise values stored in weights makes the network adaptive in nature to trained, remember and generate connection among data. In general, Artificial Neural Networks includes three layers (a) input layer, (b) hidden layer(s), and (c) output layer. Every layer consists of numerous required computational secret nodes or neurons. The returns of utilizing ANNs are ease, speed, and capacity to educate history data to offer the needed forecasts. ANN has been employed in extensive spectrum of applications such as regression, recognition, bunching, optimization and prediction. The idea for the neural system arrived from assessment of middle nervous systems. In artificial neural network, plain fake nodes, called "neurons","neurodes", "processing elements" or "units", are associated collectively to shape a network which imitates a biological neural network. At present there is Thermostats are used to measure the temperature of the coolant at various points. The input voltage is passed to an ADC and sent to an ECU which depending on the program generates an output signal. The output digital signal is passed to a DAC and the voltage is given to the servo motor mounted over the valves. The operation of the valves controls the flow of coolant. The controller performs the actuation of cooling fans. Fans are provided to aid the cooling. Feedback from fan is also provided to the controller. A redundancy system is incorporated by a failsafe ECU to give a warning to the driver in case of failure of any of the main systems and also to control the vehicle till the main system is checked and rectified. All the collected datas for different environmental conditions were used for training the ANN. The observations developed from neural network toolbox can be described as net = perceptron; net = configure(net,[0;0],0); inputweights = net.inputweights{1,1} inputweights =delays: 0 initfcn: 'initzero' learn: true learnfcn: 'learnp' In Multilayer Perception Artificial Neural Networks, after sixty nine epochs the training was stopped. The amount of validation test and the quantity of the gradient error were utilized to weigh up the act of ANNs. While teaching, the gradient error diminishes till the guidance execution circumstances were met. The numbers of confirmation test point out the quantity 1672

3 of successive iterations where the justification on prediction fails to decrease the gradient slip further. forward network by training it with different parameters like engine speed, coolant flow rate, prevailing environmental condition etc and the output is plotted below for both the response. Figure 4. Investigated temperature in radiator and its feedback for a step input Figure 2. Dissimilarity among gradient error and validation checks for MLP Network A comparison made between coolant temperature in conventional systems and the proposed system for idling condition of the engine. 6. Results and Discussion The engine temperature calculated through experimentally by using various thermostats at different point and the temperature calculated through feed forward network by training it with different engine parameters like engine speed, coolant flow rate, types of lubrication oil used and its heat carrying capacities, prevailing environmental condition etc and the output is plotted below for both the response. Figure 5. Coolant temperature Vs time 7. Conclusion Figure 3. Experimental and estimated engine temperature responses for a step input The radiator temperature calculated through experimentally by using various thermostats at different point and the temperature calculated through feed The results show experimental data and the data received through the MLP type artificial neural networks was an optimized one. This proposed system makes use of experimental data in the development process of the ANN modeling to predict the actual performance of dynamic model in simulation. Statistical methods were employed to reveal the effectiveness of the proposed model. The incorporation of Artificial Neural Network in the proposed system maintains the coolant temperature within the limit. The cooling fan operation also monitored through this system and the fan speed also increased according to the temperature of engine and coolant in the radiator. This system allows the cooling 1673

4 fan to run in variable speed at temperature above 90 and stops below it improve the efficiency of the cooling system and thereby reducing the servomotor power utilization or an engine power. The proposed system enhances the efficiency of the cooling system and thereby reducing the consumption of fuel, free loading losses and the emission too. The proper combustion of fuel at the prescribed temperature improves the diesel engine efficiency and reduces toxic emissions. This system can be a better replica for an air conditioning system. References [1] Torregrosa A, Broatch A, Olmeda P. and Romero, (2008) Assessment of the influence of different cooling system configurations on engine warm-up, emissions, and fuel consumptionǁ International Journal ofautomotive Technology, vol. 9, no. 4, pp [9] Çay Y, Korkmaz I, Çiçek A, and Kara F,(2013) "Prediction of engine performance and exhaust emissions for gasoline and methanol using artificial neural network," Energy, vol. 50, no. 1, pp , [10] Oğuza H, Sarıtasb I, and Baydanc H, (2010) "Prediction of diesel engine performance using biofuels with artificial neural network," Expert Systems with Applications, vol. 37, no. 9, pp [11] Gregor P. J. Schmitz, Chris Aldrich, and S.Francois (1999) ANN-DT:An Algorithm for Extraction of Decision Trees from Artificial Neural Networksǁ. IEEE transactions on Neural Networks. [2] Melzer F, Hesse U, and Schmitt M, (1999) "Thermomangemnt." SAE Paper Brace C, Burnham H, Slipper R, Wijetunge N, Vaughan, Wright and Light D,(2001) ǁIntegrated cooling systems for passenger vehiclesǁ SAE Paper [3] Allen D, and Lasechi M, (2001) "Thermal management evolution and controlled coolant flow," SAE Paper [4] Choukroun and M. Chanfreau (2001) Automatic control of electric actuators for an optimized engine cooling thermal managementǁ SAE Paper [5] Mitchell T, Salah M, Wagner J, and Dawson D,(2009) "Automotive thermostat valve configurations enhanced warm-up performance," ASME Journal of Dynamic Systems, Measurements, and Control, vol. 131, no. 4, pp to , 200 [6] Salah M.H, Mitchell, Wagner and Dawson D,(2010), A smart multiple-loop automotive cooling system model, control, and experimental studyǁ IEEE/ASME Transactions on Mechatronics, vol. 15, [7] Deng, Jiamei, Stobart and Richard Maass (2010) The applications of artificial neural networks to enginesǁ Industrial and Control Engineering Applications, Vol 1. [8] J. Eberth, J. Wagner, B. Afshar and Foster (2004) Modelling and validation of automotive smartǁ thermal management system Architectureǁ SAE paper

5 1675

6 1676