09-11-2016, 11:02 AM
ESTIMATION OF OUTPUT PARAMETERS FOR 4-STROKE C I ENGINE USING ARTIFICIAL NEURAL NETWORK WITH PONGAMIA AND BIOGAS AS A FUEL
1467396208-ELSEWEARPAPER.DIC.docx (Size: 356.33 KB / Downloads: 5)
ABSTRACT
This paper describes the technical feasibility of using Pongamia (Honge) oil and Biogas under Dual-fuel mode. This technology can be applied in rural area for electricity generation in developing countries. The use of Honge oil and Biogas is considered as sustainable energy supply, when both are produced locally. The experiment is carried out to study the performance of diesel engine (CI Engine) under dual-fuel mode, which is carried out on 5KW diesel generator set. The esterified honge oil (EHO) blends with diesel and bottled biogas was used for experimentation, and the gas is directly added to inlet air by modifying the induction manifold. The experiment is carried out for varying lamp loads. The engine shows considerably high thermal efficiency for EHO and biogas combination. The mechanical efficiency was improved than diesel biogas operation. One more point noticed that introduction of biogas drastically reduces EHO blends and diesel consumption. The part of paper also describes application of Artificial Neural Network (ANN) to estimate Thermal efficiency and BSFC of the engine, from comparison and observation it clears that ANN estimates close to experimental value when 90% of data is at training set. Estimated Thermal efficiency and BSFC using Artificial Neural Network correlates well with measured value. The mass flow rate, speed temperature, fuel consumption rate and time are used as a input parameters.
Keywords. Bio-Gas, C I engine, esterified honge oil (EHO), Artificial Neural Network (ANN).
1 Introduction
The most peoples of developing countries don’t have accessing modern energy sources for their daily energy needs. They are mainly depending on traditional fuels like coal, wood and petroleum fuels. In world the major part of all energy consumed from fossil sources. These fossil sources are limited and will be exhausted by the near future. Thus many countries looking for alternative sources such as, hydro, biomass, wind, solar etc. the nonconventional fuels have ability to solve many of current problems such as, air pollution, global warming and sustainability issues[1].
Biogas and Vegetable oil is one of the renewable, non depleting sources of energy. The content of Vegetable oil close to diesel fuel. Recently Vegetable oils have become more attractive because of their advantage is more over petroleum fuels. The direct use of Vegetable oil in diesel engine leads to problem because of high viscosity and low volatility; doesn’t burn completely and forms deposits in fuel injector and combustion chamber. This problem can be eliminated through various techniques; among those most important one is subjecting oil to transesterification with ethanol gives low viscous ethyl esters known as biodiesel (BD). The BD can be prepared by using several types of Vegetable oils such as honge, soybean, rapeseed; sunflower and palm are most studied [2]. Pongamia is one of the nitrogen fixing tree which grows in dry area, produces seeds containing 30-40% of oil. This is medium sized ever green tree which can be planted banks of rivers or sea costs [3].
India is the largest cattle breeding country; there is abundance of raw material for producing biogas and also municipal sewage can be used for this purpose. One of the alternate technologies ‘Sulabh’ propagates the biogas plant that utilizes human excreta as its raw input. From the last 35 years, it has setup more than thousands of such plants throughout India. Biogas obliquely replaces fossil fuels and the methane contained in it is a natural and the methane is collected and utilized commonly for electricity as well as for combined heat and power (CHP) production. Recent life cycle assessment studies have demonstrated that biogas derived methane is one of the most energy efficient and environmentally sustainable vehicle fuels. The country occupies second position in the world for biogas utilization and fifth in wind energy and photovoltaic production. Renewable energy contributes to about 7-9% of the total power generating capacity in the country [4].
2. ARTIFICIAL NEURAL NETWORKS
The structure of the brain is found to be a highly developed mechanism, which is capable of performing immensely impressive tasks. The things that the computer is capable of doing, the brain manages exceptionally well, and the idea behind neural imputing is that by modeling the major features of the brain and its operation, the computers that exhibit many of the useful properties of the brain can be produced. Though the structure of the brain is complex, it can be viewed as a highly interconnected network of relatively simple processing elements, there is a need for the model that can capture important features of real neural systems in order that it will be similar behavior. However, the model must deliberately ignore many small effects, if it is to be simple enough to implement and understand. The aim of a model is to reduce simplified version of a system, which retains the same general behavior, so that stem can be more easily understood.
The topology of neural network refers to its framework as well as its interconnection scheme. The framework is often specified by the number of layers and the number of nodes per layer. The types of layers include Input layer, Hidden layer and Output layer. A connection between nodes in different layers is called an 'interlayer connection'. The term 'connectivity' refers to how nodes are connected. Full connectivity means that every nodes in one layer is connected to every node in its adjacent layer. Yusuf cay approched ANN to model gasoline engine for prediction of engine performance ,the predicted values are with in acceptable limits[5].there are numerous published works in various journals on sucessful of ANN approch in automotive field for example, Kiani et al. Estimated six engine out put parameter for gasoline engine using back propogation ANN[6] with R values equal to 0.98, 0.96, 0.90, 0.71, 0.99 and 0.96 for carbon monoxide carbon dioxide (CO2), unburned hydrocarbon (HC), nitrogen oxides (NOx) , torque and brake power, respectively. Likewise many models have been developed to successfully model engine output response. Accurate predictions are usually achieved by utilizing back-propagation neural networks with Levenberg–Marquardt (LM) as the learning algorithm [7–13]
However the ANN architecture various from case to case such as suitable number of neuron selection for hidden layer, the transfer functions for hidden layer and output layer. For example kamyar et al., successfully investigated influence of operational parameter on CRDI engine using back-propagation network with bayesaian training algorithm [14]. A slightly different approach was done by Togun and Baysec where individual network architecture was built for each output to improve the accuracy of the ANN predictions [8]. Brahma et al., on the other hand, successfully incorporated physical model and data partitioning based on the physical processes method to increase the accuracy of the ANN prediction of in-cylinder pressure trace[13]. Harun Mohamed Ismail et al., modeled a light duty diesel engine powered with biodiesel blends to predict engine output response using back-propagation feed forward ANN was used with combination of tansig/purelin transfer functions, TRAINLM training algorithm and an neuron value of 10 are the optimum configuration to predict the correlations between four engine control parameters and nine engine-out responses [15]. Sayin and Ertunc used ANN for the modeling of a gasoline engine to predict BSFC, BTE, exhaust gas temperature and exhaust emissions. They observed that mean relative errors for the whole of training data and the test data were within 2–7% [16].
3 Experimental Setup
Experiment was conducted in energy conversion laboratory at Sambhram research center, Bangalore. The engine was a 5HP, water cooled 4-stroke engine. It was coupled with alternator through universal joint. Fig.1 shows the experimental set-up.
The engine was started by cranking manually and run for a certain period with no load condition allows to stabilize, the required engine speed was achieved by adjusting governor mechanism. Initially run the engine for diesel-biogas combination that is recorded as a standard set. The biogas flow rate was fixed by adjusting the knob in gas flow meter and recorded time taken to consume fixed cc of fuel for different loading. The experiment were conducted for constant engine speed of 1500RPM, the speed of the engine was monitored regulaloy using digital speedometer of contact type. Temperature is monitored using thermocouples and also air consumption rate was measured using U-Tube monometer. The above procedures were repeated for different trials.
Construction of ANN model
ANN is an approach inspired by brain structure and tries to simulate the brain processing capabilities. For this experimental study feed-forward back propagation network was designed. This network has six input layers, ten hidden layers and two output layers. The input data sets and corresponding output data sets are require to train to test the network. To develop ANN model the available experimental data set has been divided into two sets, one set for training the network, remaining was used to verify the capabilities of network. Haykin has presented mathematical model for testing and training ANN[17]. The weights are adjusted by training the network, using input and output data sets which is obtained by experiment. The weights are adjusted to minimize the error between output of network to actual value. Once the training is completed estimation of new set of data may be done using already trained network. The input parameter are load, gas flow rate, percentage of biodiesel and the out put parameters are BSFC, thermal efficiency. The neural network toolbox of MATLAB07 was used to develop the network and tangent sigmoid transfer function was used in the hidden layers. The training was done using Levenberg-Marquardt method. The performance index of TrainLM algorithm is the mean squared error (MSE) [18] and it is formulated as given below.
(1)
Where, yi is the predicted value of the i th pattern, yk is the target value of the i th pattern and N is the number of pattern.
Result and discussion
The performance of diesel engine was studied using Pongamia and biogas with varying load. The behavior of the engine for various operating condition are discussed below.
4.1 Specific Fuel Consumption (BSFC)
Specific fuel consumed by the engine per hour was computed. The experimental observation clears that Specific fuel consumption descries with increase in load which is normal behavior of diesel engine from the fig.1, it shows that SFC for pongamia and biogas combination is slightly high compared to diesel, biogas mixture. This is because of low heating value of Pongamia(H) and biogas combination.
Brake Thermal Efficiency (BTE)
Generally brake thermal efficiency increase with increase in load. From fig.3 it clearly shows that the maximum BTE were observed for diesel and biogas combination at a load of 800w to 1000w,brake thermal efficiency for pongamia and biogas considerably higher as compared with diesel. The maximum thermal efficiency at full load is may be due to higher heat release rate [5]. Fig.3 indicates the predicated value of thermal efficiency is close to measured value but a slight deviation was observed at predicted value this is because of less data used for trailing network at load of six hundred watts.
Conclusion
The study concludes that practical feasibility of biogas application with pongamia for compression ignition engine. Dual- fuel mode is recommended for CI engine with biogas, operation. The experimental study clears that a drastic reduction of diesel and pongamia consumption was noticed when biogas is supplied through intake manifold with air and also ,an improved thermal efficiency and reduced total fuel consumption (TFC). At 100% pongamia+ biogas combination gives a mechanical efficiency of 52 %. The rise in exhaust temperature was recorded with biogas and also rough engine noise was noticed when supplement fuel ( biogas) proportion was gradually increased the ANN model was developed to predict engine performance such has BSFC and BTE.The predicted values are with in acceptable limit when more experimental data was used for train the network