The relation between the inlet and outlet temperature of catalyst and gas emissions by using artificial neural networks

Pages 5
Views 6

Please download to get full document.

View again

of 5
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
The relation between the inlet and outlet temperature of catalyst and gas emissions by using artificial neural networks CHARALAMPOS ARAPATSAKOS, DIMITRIOS CRISTOFORIDIS Department of Production and Management
The relation between the inlet and outlet temperature of catalyst and gas emissions by using artificial neural networks CHARALAMPOS ARAPATSAKOS, DIMITRIOS CRISTOFORIDIS Department of Production and Management Engineering Democritus University of Thrace V. Sofias Street, 671, Xanthi GREECE Abstract: - The present paper examines the relation between Tout and Tin temperature of gas emissions in a Three Way Catalyst (TWC) and CO, HC, NOx emissions, by using artificial neural networks (Multilayer Perceptron). In the present paper the tests are extended using the ECE driving cycle. Key-Words: - Catalytic Converters, Gas emission, Driving Tests 1 Introduction The development of the three-way catalysts must be considered a very important step towards the elimination of automotive emissions from the top position of urban pollution sources. However, to take full advantage of this reduction, the car exhaust and fuel control system must be in good working order [1,2]. In order to ensure periodic catalyst inspection, many countries have passed legislation forcing 12 or 6 months regular exhaust gas tests. They also require the use of an official card file containing the results of all tests performed on the car exhaust system. Although the periodic catalyst performance monitoring by the garage personnel is mandatory and essential, there is also a strong need for a continuous catalyst evaluation by the driver during driving, so that any emission problem can be discovered at the earliest opportunity. In this way the last excuse of the conscienceless driver is eliminated and severe fines can be administered to the polluters. To alert the driver for possible faults in the emission control system, the state of California requires that all cars sold in this state after 1993 must be equipped with an onboard catalyst performance assessment system [3, 4, 5], This system should be able to monitor catalyst deterioration before it results in exhaust emissions greater than.6 gr/mile HC or an increase greater than.4 gr/mile HC [6]. Although the present research effort should continue on the OBD systems, the development of cleaner engines and more efficient catalytic systems, it is the authors opinion that car manufacturers should also investigate the possibilities of catalyst service procedures. Presently, the catalyst is a part that is not serviced during its active life and is readily replaced when the car emissions exceed the state or country limits. The time period, during which the catalyst converter functions effectively the legislated limits, is defined as operating life. The operating life decreases due to a combination of thermal, chemical and mechanical effects. Some of the chemical effects that cause catalyst efficiency reduction are reversible, as HC and CO storage due to temporary λ sensor malfunction or engine misfire, whereas other processes as lead, sulphur, zinc poisoning are considered permanent. Thermal effects as Pt/Ph, Pd/Rh sintering is also permanent [4]. Higher temperatures are normally measured near the exit of the converter, whereas the chemical deactivation of the catalyst surface is greater at the inlet section and gradually progresses to the interior as the catalyst ages. In general, the decrease of the converter efficiency begins at the front part and gradually spreads backwards. Therefore, as the converter ages and the chemically active region retreats to the exit sections, the temperature of the exhaust gases that reach this region becomes progressively lower due to the heat convection losses in the previous inactive region [5]. This effect reduces the overall efficiency; the heat production and the gas temperature rise even further. The leading part of a catalyst under normal condition is the one mostly strained from thermal and chemical factors [6]. The performance index of TWC is the difference between the exhaust gas temperature in TWC minus inlet temperature of gas in TWC. However, the results from the experiment measurements have shown that there is a relation between Tout and Tin temperature of gas emissions in TWC and CO, HC, NOx emissions, when it has ISBN: been used the artificial neural networks (Multilayer Perceptron). 2 Instrumentation and experimental results The experimental measurements were carried out on a SUZUKI VITARA model with a five gear manual transmission. This is a 1598 cm3 four-cylinder engine with a 97 DIN PS output at 56 rpm and 132 Nm at 4 rpm. The engine has single point injection and uses a monolithic type catalyst. The experimental procedure was adjusted according to the requirements of the legislated exhaust measurement tests. The instrumentation included, a power brake with two cylinders (Ward Leonard system controlled by a computer) and sampling system for exhaust gases CO, HC. For the exhaust gas measurements the following instruments were used:i) CO, Signal 2/NDIR,ii) HC, Signal 4/NDIR. The series of tests, which were held, included, the normal and inverse converter mounting with or without a converter. In the case of inverse mounting, short flange adaptors were used. The two cases tested were namely: Hot NEDC (New European Driving Cycle) European cycle with engine warming, Cold NEDC (New European Driving Cycle) + Hot UDC (Urban Driving Cycle) European cycle, with cold engine start and urban part of the European cycle with warm engine. The experimental conditions were almost identical for all of the tests, while, in addition, all the legislated measuring procedures were held for tests of this type. During the tests the gases (CO, HC) were continuously monitored. The using of the ECE test-driving schedule for NEDC + Hot UDC is illustrated in fig. 1.The changes in temperature differences ( Τ = Tout - Tin) between the catalyst inlet and outlet during circle are shown in the figures 2 and 3. The catalyst efficiency increases, as the temperature difference (Tout - Ti) across the converter increases. Comparing figures 2 and 3 it can be easily observed that the warm-up time for the inverse mounting condition is smaller. This means that the efficiency of the catalyst converter is higher in the inverse mounting condition. The improvement of the light off time during the transitional stages is also an important factor, because in these cases it is clear that an old catalyst converter can t readily respond during a cold engine start. The smallest temperature difference is being observed during UDC and NEDC, indicating low efficiency. ARTIFICIAL NEURAL NETWORKS First attempts at building artificial neural networks (ANN) were motivated by the desire to create models for natural brains (figure 1.1 biological neurons). Much later it was discovered that ANN are a very general statistical framework for modeling posterior probabilities given a set of samples (the input data). Figure 1. Biological neurons The basic building block of a (artificial) neural network (ANN) is the neuron. A neuron (figure 1.2) is a processing unit which have some (usually more than one) inputs and only one output. Figure 2. Neuron. First each input xi is weighted by a factor wi, and E = N wx i i i= 1 ISBN: the whole sum of inputs is calculated. Then an activation function f is applied to the result E. The neuronal output is taken to be y=f(e) (Figure 1.3) and that s the input for the next neuron and so on. Figure 3. A biological neuron a general artificial neuron weights to decrease the output error. Their main advantage is that they are easy to use, and that they can approximate any input/output map. Building the MLP In this experiment used a Multilayer Perceptron (MLP) with one input layer, one hidden layer, and one output layer. The input layer has two processing elements (PE) which are temperature before catalytic converter (T1) and temperature after catalytic converter (T exhaust). The hidden layer has 3 processing elements (PE) and finally output layer has three processing elements (PE) which are CO, HC and NOx. First we randomized our data and also the initial weights in order not to overtrain our neural model. Then we used 7% (15) of the total data to train the MLP network and 3% (6) of the data to test them. In the hidden layer the transfer function f was TanhAxon and applies a bias and tanh (hyperbolic tangent) function to each neuron in the layer. This will squash the range of each neuron in the layer to between -1 and 1. Such nonlinear elements provide a network with the ability to make soft decisions. MULTILAYER PERCEPTRON An Artificial Neural Network called Multilayer Perceptron is a feed forward network (FF) which has one or more layers between input and output layer (figure 1.4) Figure 5. Hyperbolic tangent Figure 4. Multilayer Perceptron with l layers of units The Learning Rule we used was gradient descent with Momentum (,7) for both hidden and output layer with step size 1, for hidden layer and,1 for output layer. The maximum epochs we used to train the MLP network was 3 and the Mean Square Error (MSE) was,8175. The multi-layer perceptron (MLP) is a universal function approximator as proven by the Cybenko theorem. Also the learning process in MLP is iterative and essentially consists in adjusting the ISBN: Figure 6. Gradient- Descent Method 3 Experimental measurements and Experimental results During the experiments, it has been counted: The percent of (%) (CO) Το ppm(parts per million) HC Το ppm(parts per million) NOx CO CO Output Figure 7. Desired Output and Actual Network Output about CO HC HC Output Figure 8. Desired Output and Actual Network Output about HC NOx NOx Output Figure 9. Desired Output and Actual Network Output about NOx O u tp u t CO HC NOx CO Output HC Output NOx Output Figure 1. Desired Output and Actual Network Output about CO, HC, NOx Performance CO HC NOx MSE ,6 755, ,23 NMSE, ,66423, MAE 384, , ,35562 Min Abs Error, ,292523,65327 Max Abs Error 14629, , ,89854 r, , , Table 1. Desired Output and Actual Network Output about CO, HC, NOx 4 Conclusion By taken into consideration all the above, it can be said that there is a relation between Tout and Tin temperature of gas emissions in the Three Way Catalyst (TWC) and CO, HC, NOx emissions too, by using the artificial neural networks (Multilayer Perceptron. ISBN: References: [1]. Funabiki, M., Yamada, T (1985): A Study of Three-Way Conversion Catalyst Thermal Deactivation and Improvement, SAE Paper , pp [2]. Pelters, S., Schwartzenthal, D., Maus, W., Swars, H., Bruck, R. (1993):,Alternative Technologies for Studying Catalyst Behaviour to Meet OBD II Requirements, SAE Paper [3]. Church, L.M., Thoss, E.J., Fizz, L.D., (1991): Operating Temperature Effects on Catalyst Performance and Durability, SAE Paper 91845, pp [4]. Sparis, P.D., Botsaris, P.N., Karkanis, A.,(1995): An Investigation of Three - Way Catalyst Rich Mixture Poisoning, ISATA Paper 95ATS86. [5]. Burkholder, S., P., Cooper, B., J., (1991): Effect of Aging and Testing Conditions on Catalyst Performance, SAE Paper , pp [6]. Wendland, D.W., Matthes, W.R., Visualisation of Automotive Catalytic Converter Internal Flows, SAE Paper , pp , [7]. The ANN Book R. M. Hristev Edition 1 (1998) by R. M. Hristev [8]. An Introduction to Neural Networks, Ben Krose, Patrick van der Smagt, Eighth edition November [9]. Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, Nikola K. Kasabov, A Bradford Book The MIT Press Cambridge. Massachusetts, London, England. [1]. Neural Networks Theory, Alexander I. Galushkin, Springer-Verlag Berlin Heidelberg 27 ISBN:
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!