Content uploaded by Akbar Irandoukht
Author content
All content in this area was uploaded by Akbar Irandoukht on May 31, 2018
Content may be subject to copyright.
A preview of the PDF is not available
Modeling and preparation of activated carbon for methane storage I. Modeling of activated carbon characteristics with neural networks and response surface method
Abstract and Figures
Numerous methods have been proposed previously to describe the characterization of porous materials; however, no well-developed theory is still available. Three different modeling methods were employed in this study to explore the relationship between the characterization parameters of activated carbon (AC) and its methane uptake. The first and the second methods were based on the Radial Basis Function (R.B.F) neural networks. At the first R.B.F. modeling, the neural networks algorithm was designed using the Gaussian function. The collected data for modeling were divided into two parts; (i) the data used for training the network and (ii) the data used for testing the predicted network. At the second R.B.F. modeling, the MATLAB toolboxes for designing the R.B.F. neural networks were applied. The response surface method was employed as another model, using different functions in proportion to the way that the parameters affect the methane uptake. Concerning the error minimization in the estimation of the response and some statistical methods, the suitable model was selected. The results revealed that all these models were suitable for modeling the relation between the characterization parameters of the activated carbon and the methane uptake. However, the best response was provided by the neural networks modeling with the MATLAB toolboxes, demonstrating the smaller difference.
Figures - uploaded by Akbar Irandoukht
Author content
All figure content in this area was uploaded by Akbar Irandoukht
Content may be subject to copyright.
... The experimental data, including surface area, micropore volume and packing density as inputs and uptake as target, were gathered from the literature and included three inputs and one target. The data and their references are listed in Table 1 [4]. Data are randomly divided into two groups. ...
... Train and test collected data from literature[4] ...
Optimization and characterization of nanoadsorbents have already been investigated in numerous publications. Since surface area, micropore volume and packing density play an important role in the adsorption process; as a result, it is of considerable importance in chemical processes like PSA. In this study, a CSA-LSSVM method has been used to describe the characterization of nanoadsorbents to explore the relationship between its characterization parameters and CH 4 uptake. For this purpose, experimental data was gathered from the literature for modeling and was divided into two parts, namely train data and test data. Significance of the proposed model was determined using statistical methods. The revealed results of this study indicate that intelligent modeling algorithm was suitable for modeling the relation between the characterization parameters of the nanoadsorbents and the CH 4 uptake.
... Response surface methodology (RSM) is a technique which provides optimized solutions to a variety of industrial problems. 36 It provides efficient and cost-effective solutions for assessing experimental variables. 37 The RSM analysis proposes optimized data sets with the best system performance and saves both time and money. ...
For evaluating the significance of renewable alternative fuels for optimized engine performance and lower emissions, methanol has been extensively utilized as a blend with gasoline in spark-ignition engines. However, rare attempts have been rendered to examine the consequence of methanol-gasoline fuel blends (M6, M12, and M18) on lubricant oil operating for a longer period in engines. The highest and least decrease of 9.62% and 6.68% in kinematic viscosity (KV) was observed for M0 and M18, respectively. However, the flash point (FP) of degraded lubricant oil for M6, M12, and M18 was 3%, 5%, and 7% higher than that of M0, respectively. Total acid number (TAN) and ash content of degraded lubricant oil for M18 were the highest among M0, M6, and M12. An inclusive optimization of engine performance, emissions, and lubricant oil properties has been made for various methanol-gasoline fuel blends at distinct operating conditions by employing the response surface methodology (RSM) technique. RSM-based optimization portrayed the composite desirability value of 0.73 for 2137.13 watt brake power (BP), 6.08 N-m torque, 0.37 kg/kwh brake-specific fuel consumption, 22.10% brake thermal efficiency, 4.02% carbon monoxide emission, 7.15% carbon dioxide emission, 134.12 ppm hydrocarbon emission, 517.02 ppm nitrogen oxides emission, 12.44 cst KV, 203.77°C FP, 2.23 mg/g KOH TAN, and 2.65%wt ash content as responses for fuel blend M8 at 3400 rpm and higher loading condition. RSM predicted results demonstrated significant compliance with empirical findings, with absolute percentage error (APE) below 5% for each response. However, the highest APE of 4.68% was obtained for FP owing to inefficient desirability as a consequence of manual testing. The least APE of 1.57% was obtained for torque because of the highest desirability. Overall, the RSM predicted results of the designed models are effective and viable. RSM technique was found to be effective for the optimization of the broader engine characteristics spectrum.
... In addition, the error signal was also backpropagated until the minimum of signal could reach based on many training iterations (so-called epochs). The LAs or backpropagation training commonly used were train-bfg, -lm, -rp, -scg, -gdx, -cgf, and -gda [49]. After the training stage was completed, the testing and validation phase was important because it offered a thorough awareness of the effectiveness of ANN prediction models which were technologically advanced. ...
Biodiesel has been emerging as a potential and promising biofuel for the strategy of reducing toxic emissions and improving engine performance. Computational methods aiming to offer numerical solutions were inevitable as a study methodology which was sometimes considered the only practical method. Artificial neural networks (ANN) were data-processing systems, which were used to tackle many issues in engineering and science, especially in some fields where there was a failure of the conventional modeling approaches. Thus, it was believed that the best choice was the development of a novel approach like the ANN model to anticipate engine performance and exhaust emissions with high accuracy. In this review paper, the structure and applicability of the ANN model were comprehensively evaluated. More importantly, the use of ANN with trained, tested, and validated data was introduced to determine the performance and emission characteristics of a diesel engine fueled with biodiesel-based fuel. In general, the ANN model could supply a relatively high determination coefficient as compared between predicted results and experimental data, showing that the ANN model could have a good ability to predict the engine behaviors with an accuracy higher than 95%.
... This paper is aiming to quantitatively examine the optimum conditions of the reforming process by HYSYS simulation and to validate optimized HYSYS results by a statistical approach called response surface methodology. RSM is a promising tool for optimization studies because this method accounts for the interactions between process variables simultaneously (Bezerra et al. 2008;Körbahti and Rauf 2008;Liu et al. 2015;Montgomery 2004;Namvar-Asl et al. 2008;Pradhan et al. 2012). Traditionally, optimization problems have been carried out by observing the effect of one variable at a time on an experimental response by changing that variable only and keeping others at a constant level. ...
The present work optimize the reforming process for syngas production in regasified liquefied natural gas (R-LNG) based ammonia plant using central composite design in response surface methodology and then to perform a comparison study with the optimization results obtained from the Aspen HYSYS simulation and experimental plant data. The optimization study has been conducted based on the effect of six independent parameters including feed input temperature, feed input pressure, wall temperature of the reformer tube, steam to carbon ratio, hydrogen to carbon ratio and R-LNG feed flow rate on two responses, hydrogen production and amount of unreacted methane in the output stream. Finally, the optimum values of the process variables have been recorded, which results in maximum hydrogen concentration and minimum methane content in the output stream. The average error was about 6.7% for RSM model, and it was approximately 3.8% for HYSYS model, which implies that the discussed models have an acceptable validity.
Abbreviations: ANOVA: Analysis of Variance; CCD: Central Composite Design; CHR: Carbon to Hydrogen Ratio; FACT: Fertilisers And Chemicals Travancore Limited; GHR: Gas Heated Reformers; HYSYS: Hyprotech System Metric Ton Per Day; OFAT: One-Factor-at-A-Time; PBD: Plackett–Burman Design; PFD: Process Flow Diagram; PR: Peng-Robinson; R-LNG Regasified Liquefied Natural Gas; RSM: Response Surface Methodology
... Worthwhile of mentioning, application of intelligent approaches for the prediction of some properties of other nanomaterial such as activated carbon (AC) has been under great of attention. These intelligent methods have been performed from the estimation of optimal conditions for AC synthesis for prediction of various gases and liquid uptakes in porous materials [19][20][21][22]. ...
Optimum design of water vapor separation process (dehumidification) using adsorption process mostly depends on the selection of appropriate porous materials or adsorbents with the highest equilibrium storage capacity of the vapor. Equilibrium capacity is generally evaluated through cost-demanding experiments via direct measurement of the vapor isotherm. Reliable prediction of the vapor adsorption capacity in porous materials provides a robust tool to a quick screening of porous materials appropriating for dehumidification process. In this article, adsorption capacity of water vapor in metal–organic framework (MOF) materials is predicted using two robust “artificial neural network (ANN)” and “Adaptive network-based fuzzy inference system (ANFIS)” methods. The three parameters of the surface area, pore volume and pore diameters are selected as input and the water vapor adsorption capacities of MOFs were computed as the output of the models. Comparison of the obtained results and real experimental data implied the superiority of the ANFIS and ANN models to predict the water vapor adsorption capacity into MOFs with a mean squared error (MSE) of 0.005 and 0.002, respectively. This clearly indicates a great potential for the application of both ANN and ANFIS methods to rapid screen MOFs suitable for water vapor adsorption.
In the realm of internal combustion engines, there is a growing utilization of alternative renewable fuels as substitutes for traditional diesel and gasoline. This surge in demand is driven by the imperative to diminish fuel consumption and adhere to stringent regulations concerning engine emissions. Sole reliance on experimental analysis is inadequate to effectively address combustion, performance, and emission issues in engines. Consequently, the integration of engine modelling, grounded in machine learning methodologies and statistical data through the response surface method (RSM), has become increasingly significant for enhanced analytical outcomes. This study aims to explore the contemporary applications of RSM in assessing the feasibility of a wide range of renewable alternative fuels for internal combustion engines. Initially, the study outlines the fundamental principles and procedural steps of RSM, offering readers an introduction to this multifaceted statistical technique. Subsequently, the study delves into a comprehensive examination of the recent applications of alternative renewable fuels, focusing on their impact on combustion, performance, and emissions in the domain of internal combustion engines. Furthermore, the study sheds light on the advantages and limitations of employing RSM, and discusses the potential of combining RSM with other modelling techniques to optimise results. The overarching objective is to provide a thorough insight into the role and efficacy of RSM in the evaluation of renewable alternative fuels, thereby contributing to the ongoing discourse in the field of internal combustion engines.
The use of NG in the transportation sector is becoming an appealing option to diesel and gasoline fuels, presenting higher benefits. ANG technology offers a secure, cost-effective, energy-efficient strategy for the storage of NG in porous sorbents at reasonable gas densities. The major goal for its extensive utilization is the requirement of effective storage materials under practicable conditions. Recently, there has been increased attention in utilizing bio-wastes for the preparation of microporous carbons. In this contribution, our growing knowledge on the use of biobased materials and the processing strategies in an effort to predictively produce effective porous carbons appropriate for ANG technology have been reviewed. By careful literature selection, different precursors with different alternative processes to convert low-cost bio-wastes into porous carbons and achievements in methane storage are presented. To gain deeper insight into the technology, the correlation between the structural and chemical properties of materials and the factors affecting the storage performance are highlighted. The utilization of bio-wastes for the development of microporous carbons with facile methods emerged to be encouraging, which would be significant in larger scale applications. Bio-waste processing for ANG storage is valued over many other techniques, and the products are able to store substantial levels of methane. This review could help improve researchers’ evaluation of the methods as a guideline for ANG. Further studies for achieving an accomplished interconnection between the structural characteristics and the methane storage capacities with different bio-wastes and optimization strategies would be beneficial.Graphical abstract
In this study, the effects of JP-8 military aviation fuel and biodiesel sunflower oil on the performance and emissions of the internal combustion engine were investigated by creating a mixture of fuel in different concentrations. Experimental and statistical methods were used in the study process. All of the experiments were carried out on an experimental setup with a single-cylinder compression-ignition engine at 7.5, 12.5 and 17.5 Nm engine load conditions. A total of 5 different fuels were used, including pure JP-8 fuel, 3 different blended fuels and pure biodiesel fuel. The design of the Response Surface Method, which is used as a statistical method, was carried out with Central Composite Design in the Design Exper 11 program environment. After the created design, 13 different experiments were carried out. With the tests performed, the values of IMEP, BSFC, COVimep, ignition delay, equivalence ratio, exhaust gas temperature, CO, and NOx emission response parameters were determined. Quadratic models were created by performing variance analysis for the response parameters obtained. After the optimization, it was concluded that the engine load should be 14.49 Nm and the JP-8 fuel biodiesel ratio should be 2.47%. In case the internal combustion engine is operated with optimum input parameters, IMEP value is 3.71 bar, BSFC value is 193.46 g/kWh, COVimep value is 5.99 bar, CO value is 724 ppm and NOx value is 168.62 ppm.
Data science has been rapidly developed and implemented in diverse scientific and technological fields over the past decade, to accelerate new knowledge generation and develop high‐impact applications. Recently, different data science tools and techniques have been widely used, such as optimizations, regressions, and classifications of data (tabular, spectral, or visual). In this review, data science tools and techniques are discussed for their adoption for the synthesis, characterization, and applications of carbon‐based materials. Materials synthesis in conjunction with data science has resulted in optimal growth conditions with desired properties and processing techniques. Regarding characterization, molecular structures can be reconstructed using microscopy images, and a particular property can be predicted based on the other properties of the desired carbon material. Moreover, for the applications of carbon materials, data science has enabled prediction of the water treatment efficiency, classification of electronic signals, prediction of the biological activity, and virus classification. It is clear that by combining data science and carbon‐related materials, it is now possible to accelerate theory‐experimental research in the quest for novel materials and their emerging applications.
Adsorbed Natural Gas (ANG) technology is a safe and low-cost approach for natural gas storage. Improving the volumetric adsorption capacity of the adsorbents in the ANG tank can enhance the fuel density and make this technology cost-effective compared to the other available CH4 storage approaches. For this purpose, the present research focuses on maximizing CH4 uptake on low-cost and available anthracite-based carbon materials via experimental and analytical investigation. The effect of preparation variables of chemical agent (KOH) impregnation ratio to the anthracite (2.6-4.3 g/g), activation temperature (665-834 °C) and retention time (39-140 min) on the specifications of the coal-based activated carbons (ACs) and their CH4 adsorption capacity was examined experimentally. The results were analyzed through empirical models including Response Surface Methodology (RSM), our in-house developed models, namely Regularization Networks (RN) and Adaptive Neuro-Fuzzy Interface Systems (ANFIS). The statistical assessment revealed that all three established models could effectively predict the methane adsorption capacity of the carbon samples based on their preparation conditions. The superior performance of our in-house RN is dedicated to its robust theoretical backbone in the Regularization theory. Finally, the carbon sample prepared under the optimized preparation condition, based on RSM and genetic algorithm (GA), showed the highest CH4 uptake of 175 cm3(STP)/cm3. Based on the authors’ knowledge, the volumetric CH4 capacity of the optimized AC is one of the highest values reported in the literature among different classes of the adsorbent.
Consider the ridge estimate (λ) for β in the model unknown, (λ) = (XX + nλI) Xy. We study the method of generalized cross-validation (GCV) for choosing a good value for λ from the data. The estimate is the minimizer of V(λ) given bywhere A(λ) = X(XX + nλI) X . This estimate is a rotation-invariant version of Allen's PRESS, or ordinary cross-validation. This estimate behaves like a risk improvement estimator, but does not require an estimate of σ, so can be used when n − p is small, or even if p ≥ 2 n in certain cases. The GCV method can also be used in subset selection and singular value truncation methods for regression, and even to choose from among mixtures of these methods.
The problem of the approximation of nonlinear mapping, (especially
continuous mappings) is considered. Regularization theory and a
theoretical framework for approximation (based on regularization
techniques) that leads to a class of three-layer networks called
regularization networks are discussed. Regularization networks are
mathematically related to the radial basis functions, mainly used for
strict interpolation tasks. Learning as approximation and learning as
hypersurface reconstruction are discussed. Two extensions of the
regularization approach are presented, along with the approach's
corrections to splines, regularization, Bayes formulation, and
clustering. The theory of regularization networks is generalized to a
formulation that includes task-dependent clustering and dimensionality
reduction. Applications of regularization networks are discussed
This book introduces readers to the Artificial Neural Network (ANN) and Hybrid Neural (HN) models: two effective tools, which can be exploited to design and control industrial processes. Different topics including modeling, simulation and process design are covered. More efficient analyses and descriptions of real case studies, ranging from membrane technology to the obtaining of second-generation biofuels are also provided. One of the major advantages of the described techniques is represented by the possibility of obtaining accurate predictions of complex systems, whose behaviors might be difficult to describe by conventional first-principle models. One of the major impacts of the present book is to show the true interactions and interconnectivities among different topics belonging to chemical, bio-chemical engineering, energy, bio-processes and bio-technique research fields. Some of the main goals are here are to provide a deep and detailed knowledge about the main features of both ANN and HN models, and to iterate possible topologies to integrate in these ANN and mechanistic models; to cover a wide spectrum of different problems as well as innovative and unconventional modeling techniques; to show how various kinds of advanced models can be exploited either to predict the behavior or to optimize the performance of real processes.
Activated carbons prepared by KOH activation of an anthracite were studied for methane storage applications. The effect of the different variables of the activation process (KOH/anthracite ratio, pyrolysis temperature and time) on methane storage and methane delivery was analyzed. Methane delivery was obtained in two different ways: calculated from the isotherm and measuring the volume of methane delivered from a carbon-filled vessel (5 cm3). Both methods give similar values. In addition to the well-known effect of the micropore volume and packing density, special attention was paid to the effect that the micropore size distribution has in methane storage performance. It was shown that this parameter is also a key parameter in the application of activated carbons for methane storage applications. Activated carbons prepared from a cheap raw material and using a single stage activation process have reached very high values of methane storage (155 V/V) and delivery (145 V/V).
This paper presents an overview of the results of our research group in methane storage, in which the behaviour of different carbon materials in methane storage has been studied. These materials include physically activated carbon fibres (ACFs), chemically activated carbons (ACs) and activated carbon monoliths (ACMs), all of them prepared in our laboratories. These results have been compared with those corresponding to commercial ACFs, commercial activated carbon cloths and felts, and a commercial activated carbon.An in depth analysis (different raw materials, activating agent and preparation variables) has been done in order to obtain the carbon material with the best methane adsorption capacity by unit volume of adsorbent. The important effect of the micropore volume, micropore size distribution (MPSD) and packing density of the carbon materials in the methane adsorption capacity and delivery has been analysed. After this study, activated carbons with volumetric methane uptake as high as 166 v/v and delivery of 145 v/v have been prepared. In addition, ACM with methane uptake of 140 v/v and a delivery of 126 v/v has also been obtained.Moreover, the results corresponding to preliminary in situ small angle neutron scattering (SANS) study of CD4 adsorption under pressure in different porous carbons and a zeolite are also included. These experiments have established SANS as a viable technique to investigate high-pressure methane adsorption. CD4 adsorption at supercritical conditions produces changes in the SANS curves. The changes observed are in agreement with theoretical speculations that the density of the adsorbed phase depends upon the pore size.
Learning an input-output mapping from a set of examples, of the type that many neural networks have been constructed to perform,
can be regarded as synthesizing an approximation of a multidimensional function (that is, solving the problem of hypersurface
reconstruction). From this point of view, this form of learning is closely related to classical approximation techniques,
such as generalized splines and regularization theory. A theory is reported that shows the equivalence between regularization
and a class of three-layer networks called regularization networks or hyper basis functions. These networks are not only equivalent
to generalized splines but are also closely related to the classical radial basis functions used for interpolation tasks and
to several pattern recognition and neural network algorithms. They also have an interesting interpretation in terms of prototypes
that are synthesized and optimally combined during the learning stage.