Development of Emission models by Conventional Method and Artificial Neural Network for Direct injection VCR Engine using Eco-diesel (Plastic oil) #1 D. S. Barbole, #2 P. A. Deshmukh, #3 Abhay A. Pawar #1 PG Student, Rajarshi Shahu College of Engineering, Tathawade,SP Pune University, Mumbai, India #2 Professor, Rajarshi Shahu College of Engineering, Tathawade,SP Pune University, Mumbai, India ABSTRACT Artificial neural network( ANN) predicts the performance and emissions of diesel engine with great accuracy. The present study uses conventional method and artificial neural network to form emission model when Ecodiesel is used as fuel in variable compression ratio (VCR) diesel engine. Based on experimental data an ANN model is developed to predict CO2,, HC and CO with load, torque, Brake power, Indicated power, Indicated mean effective pressure, Air flow, Fuel flow, Specific fuel flow rate, Air fuel ratio as input parameter for the network. The study was carried out with 70% of total experimental data selected for training the neural network, remaining 15% data has been used for testing the performance of the trained network and15% data for validation purpose. The developed ANN model was capable of predicting the performance and emissions of the experimental engine with excellent agreement as observed from correlation coefficient value which is 0.97912 for overall training, validation and testing. This emission model is step towards formulating universal emission model. Keywords Eco-diesel (Plastic oil), Artificial neural network, Emission model, VCR engine, Eco-diesel Blends. I. INTRODUCTION Petroleum resources are finite and most of energy needs of the world are provided by fossil fuels which are depleting at high rate. In recent years, the consumption of petroleum products in India has been increased significantly. If concerned to India the need to search alternative fuels urged to meet the demand for transportation, agricultural sector is very high. Other than depleting resources of petroleum, another important aspect of their use is the great rise of emissions from IC engines like CO, HC,, CO 2 etc. by automobiles and various industries which have huge effect on human life and vegetation. Plastics materials have proved their reputation and have gained popularity as they are lighter in weight, does not rust, lower in cost, reusable and conserve natural resources. Now a day waste plastics have created a very serious environmental challenge because of their huge quantities and sorting of plastic from other waste is problem. Plastics are produced from petroleum derivatives that contain additives such as antioxidants, colorants, and other stabilizers and are composed primarily of hydrocarbons. Pyrolysis is the most attractive technique of chemical feedstock recycling of the polymeric materials by using heat in the absence of oxygen. Compression ignition engines have proved to be the best option in heavy duty and light duty applications like transportation, power generation and light utility vehicles, but due to rapid depletion of conventional fossil fuels, their rising prices, ever increasing environmental issues and power dominance are the major concerns. The present study deals with performance and emission analysis of blends of waste plastic oil obtained by catalytic pyrolysis of waste polyethylene with diesel in a CI engine with varying loads and also with varying compression ratio and what are the changes in emission.[1] Method of production of Eco - diesel Pyrolysis is a thermal degradation process in the deficiency of oxygen, performed to obtain eco-diesel by using silica alumina as a catalyst. For ease of handling the process different sizes and shapes of waste plastics are collected, cleaned and crushed with shredder. The reactor chamber is fed with fine crushed plastic particles. The chamber is heated and maintained at a temperature range of 320 C 800 C by using copper coil placed around for 3 4 hours duration. Waste plastic gets vaporized at this high temperature and passes through the condenser devices. Latent heat transfer occurs by condensing the waste plastic vapour because of the cold water present inside the condenser. Condensed waste plastic vapour is then stored in the oil collector in the form of plastic oil. Following output products were collected from the pyrolysis treatment the Waste Plastic Oil (Eco- diesel) 75% to 90%, Gas 5% to 20% and Residual 5% to 10%. [2] This process has great efficiency as exhaust gases are burned inside pyrolysis chamber to heat plastic at required temperature. Experimental setup 2015, IERJ All Rights Reserved Page 1
Fig. 3.1 VCR Engine test setup 1 cylinder, 4 stroke, Diesel engine.[3] An experimental setup made by using Kirloskar engine with Type 1 cylinder, 4 stroke Diesel, water cooled, power 3.5 kw at 1500 rpm, stroke 110 mm, bore 87.5 mm. 661 cc, CR 17.5, Modified to VCR engine CR range 12 to 18 is used here for testing of eco-diesel which is attached with dynamometer of eddy current type, water cooled, with loading unit. This setup has fuel tank Capacity 15 lit with glass fuel metering column. Calorimeter used here for measuring is of pipe in pipe type. To acquire data generated by this setup data acquisition device NI USB-6210, 16-bit, 250kS/s is used. Other devices used are as follow 1. Temperature sensor Type RTD, PT100 and Thermocouple, Type K 2. Load indicator Digital, Range 0-50 Kg, Supply 230VAC 3. Rotameter Engine cooling 40-400 LPH; Calorimeter 25-250 LPH 4. Crank angle sensor has Resolution of 1 Deg at Speed of 5500 RPM with TDC pulse. 5. Pump Type: Monoblocks 6. For engine cylinder pressure measurement piezo sensor of Range 5000 PSI, with low noise cable is used. 7. Load sensor Load cell, type strain gauge, range 0-50 Kg 8. Fuel flow transmitter DP transmitter, Range 0-500 mm WC 9. Overall dimensions W 2000 x D 2500 x H 1500 mm Software used is EnginesoftLV for Engine performance analysis. Test engine coupled with electrical dynamometer to apply load on the engine. Electrical Dynamometer consists of electrical powerbank which applies loads in range of 0 to 50 kg loads on an engine and it is controlled with the aid of ammeter and voltmeter. Engine is connected with the computer to record and analyze the output data. The combustion parameters such as cylinder pressure, instant heat release rate and ignition delay are evaluated. AVLDiGas444 exhaust gas analyzer is used to measure engine emissions such as, unburnt hydrocarbon (HC), carbon monoxide(co) and Carbon dioxide(co2). Smoke opacity of the exhaust gas is measured with the use of AVL437C smoke meter.[3] Fuel properties Table No: 4.1 Eco-diesel [Plastic oil] properties Sr. No. Test description Unit RESULTS TEST METHOD 1 Ash Content % by Wt. 0.01 IS 1448(P4) 2013 2 Density at 15 C kg/m3 842.5 IS 1448 (P16) 2013 3 Flash Point (Abel) C < -6.0 IS 1448(P20) 2008 4 Gross Calorific Value Cal/g 10470 IS 1448(P6) 2013 5 Kinematic Viscosity at 40 C CSt 1.815 IS 1448(P25) 2013 6 Pour Point C 0 IS 1448 (P10,Sec2) 2013 7 Total Sulphur % 0.176 ASTMD 4294 2010 8 Water Content % by Vol 0.45 IS 1448 (P40) 2011 [Chem-tech laboratories Test Report No: R151200052.001 from BVG India.](4) 2015, IERJ All Rights Reserved Page 2
Method Along with pure diesel four different blends are prepared to test VCR diesel engine. No any extra additives are added inside eco-diesel. Fine filtered eco- diesel is taken for testing purpose. Firstly engine is made to run on diesel for 15 min to remove out carbon particle present inside cylinder block. 1. 100% pure diesel. 2. 5% blend of eco-diesel with pure diesel. 3. 10% blend of eco-diesel with pure diesel. 4. 15% blend of eco-diesel with pure diesel. 5. 20% blend of eco-diesel with pure diesel. Here test are taken at constant speed of 1600 rpm with varying compression ratio from 16 to 18 in step of one. At the same time different loads are applied on the engine using dynamometer. The range selected here is 0 to 9 in a step of 3. Result and calculations Result can be discussed in two parts by two methods as follows. Conventional analysis By using regression analysis we can derive relation between various variables. Here is derived equation for NOx emission by using dimensional analysis method. Load = [M¹L 0 T 0 ] Indicated Power= IP = [M¹L 2T 3] BSFC = [M L ²T²] NOx = [M¹L ³T ] Air flow kg/hr = [M¹L T ¹] N=No. of parameters =5 M= No. of fundamental dimensions = 3 Parameters in terms = M + 1 =3+1 = 4 Let Equating power of M, L, T we get For M: 0 = a + b +1.......... (1) For L: 0 = 0 + 2b 2c 3...... (2) For T: 0 = 0 2b + 2c......... (3) Solving we get a = 2, b = 3, c =. Now Equating power of M, L, T we get For M: 0 = a + b +1........ (1) For L: 0 = 0 + 2b 2c....... (2) For T: 0 =0 3b + 2c 1..... (3) Solving we get a = 0, b = 1, c = 1. Now according to theorem 2015, IERJ All Rights Reserved Page 3
Rearranging and removing proportionality sign, This is dimensional analysis for NO. Now by using MINITAB software for performing regression analysis we can get conventional equation. After finding value of K, A, B we can write NOx emission equation as follow, ANN analysis Artificial neural network(ann) predict the performance and emissions of diesel engine with great accuracy. To get the best prediction by the network, several structures were evaluated and trained using the experimental data. By generalizing the Widrow-Hoff learning rule to multiple-layer networks and nonlinear differentiable transfer functions Back-propagation is a network created. Input vectors and the corresponding target vectors are used to train a network until it can approximate a function.(6) Here for input layer we select load, torque, Brake power, Indicated power, Indicated mean effective pressure, Air flow, Fuel flow, Specific fuel flow rate, Air flow ratio to get out layer as emission of IC engine namely CO, CO2, HC. First by using 70% value network is trained then 15% data is used for testing and 15% data is used for validation purpose. Fig. 6.2.1 Artificial neural network The neural network architecture refers to the arrangement of neurons into layer(s) and the connection patterns between the hidden layers, the choice of activation functions and most pertinently the number of hidden neurons in the hidden layers. In the present study we used a feed-forward neural network model to predict the output parameters. A multilayer feed-forward neural network is consists of one input layer, unknown hidden layer(s) and one output layer. Input layer consist of 9 neurons and output layer consist of 4 neurons. The input layer is interconnected with the hidden layers and similarly hidden layer is connected with the output layer with the help of synaptic weights. During the training phase of the network, synaptic weights are modified on each iteration, in order to learn the underlying patterns existing in the data. The states of the hidden and output layer determine the output of the network.(5) ANN analysis Result: Artificial neural network gives correlation coefficient close to 1. For training of the network correlation coefficient is 0.9988. For validation of the network correlation coefficient is 0.93429. For testing of the network correlation coefficient is 0.97871. Overall correlation coefficient is 0.97912. 2015, IERJ All Rights Reserved Page 4
Fig. 6.3.1 ANN regression plot Fig. 6.3.2 Actual and ANN CO2 emission comparison Fig. 6.3.3 Actual and ANN HC emission comparison 2015, IERJ All Rights Reserved Page 5
Fig. 6.3.4 Actual and ANN emission comparison Fig. 6.3.5 Actual and ANN CO emission comparison II. CONCLUSIONS 1) For emission conventional model is developed. 2) If the relation between input and output parameters is linear in that case dimensional analysis successful. If it is non-linear then ANN is useful. This limitation of dimensional analysis is overcome by Artificial neural network technique. 3) In dimensional analysis number of parameters can be used but lengthy calculations are there. 4) Artificial neural network(ann) predict the performance and emissions of diesel engine with more accuracy. 5) The present study uses conventional method and artificial neural network to develop emission model when Eco-diesel is used as fuel in VCR diesel engine. 6) Artificial neural network gives less error in predicting, HC, emissions than CO emissions from the graph it is clearly visible. 7) Emissions are less at 10% blend of eco-diesel. III. ACKNOWLEDGMENT Author sincerely thanks Rajarshi Shahu College of engineering for offering the setup of IC engine for experimental studies. REFERENCES 1. Sachin Kumar, R. Prakash, S. Murugan, R.K. Singh, ( 2013), Performance and emission analysis of blends of waste plastic oil obtained by catalytic pyrolysis of waste HDPE with diesel in a CI engine, Energy Conversion and Management 74 (2013) 323 331. 2. P. Senthil Kumar, G. Sankaranarayanan,(2016), Investigation on environmental factors of waste plastics into oil and its emulsion to control the emission in DI diesel engine, Ecotoxicology and Environmental Safety. 3. Instruction manual (2008), VCR Engine Test Set Up, Apex technology Pvt. Ltd. page 1-5 and 51-60. Test report no. R151200052.001(2015), Pyrolysis Oil sample analysis, Chem-tech laboratories, Pune. 2015, IERJ All Rights Reserved Page 6
4. Sumit Roy, Rahul Banerjee, Probir Kumar Bose,(2014), Performance and exhaust emissions prediction of a CRDI assisted single cylinder diesel engine coupled with EGR using artificial neural network, Applied Energy 119 (2014) 330 340. 5. M. Kiani Deh Kiani, B. Ghobadian, T. Tavakoli, A.M. Nikbakht, G. Najafi,(2009), Application of artificial neural networks for the prediction of performance and exhaust emissions in SI engine using ethanol- gasoline blends, Energy 2009.08.034,1-5. 6. Rahul Krishnaji Bawane, SV Chanapattana, Abhay A Pawar, Performance Test of CI Engine fueled with Undi Oil Biodiesel under Variation in Blend Proportion, Compression Ratio & Engine Load, International Journal of Science, Engineering and Technology Research (IJSETR), Volume 3, Issue 8, August 2014. 7. C. Srinidhi,S. V. Channapattana, J.A. hole, A.A.Pawar, P.G.Kamble, Investigation On Performance and Emission Characteristics Of C.I. Engine Fuelled With Honne Oil Methyl Ester, International Journal of Engineering Science Invention, Volume 3 Issue, May 2014, PP.59-66. 2015, IERJ All Rights Reserved Page 7