Prediction of grindability with multivariable regression and neural network in Chinese coal

Grindability index of coal is usually determined by Hardgrove Grindability Index (HGI). The correlation between the proximate analysis of Chinese coal and HGI was studied. It was found from statistical analysis that, the higher the moisture and the volatile matter content in coal, the less the HGI will be. On the contrary, the higher the ash and the fixed carbon content in coal, the higher the HGI will be. But the correlation between proximate analysis and HGI in coals is nonlinear. The prediction equation of HGI reported in literature, which is based on proximate analysis of coal and linear regression method, is not correct for coals in China. In this paper, the generalized regression neural network (GRNN) method was used to predict the HGI. A higher precision in the prediction result was obtained through such new method. By this method, the HGI can be estimated indirectly from the proximate analysis of coal when the HGI measurement equipment is not available.     

Studies of relationship between petrography and elemental analysis with grindability for Kentucky coals

The effects of macerals, ash, elemental analysis and moisture of wide range of Kentucky coal samples from calorific value of 23.65-34.68 MJ/kg (10,170-14,910 (BTU/lb)) on Hardgrove Grindability Index (HGI) have been investigated by multivariable regression method. Two sets of input: (a) macerals, ash and moisture (b) macerals, elemental analysis and moisture, were used for the estimation of HGI. The least square mathematical method shows that increase of the TiO₂ and Al₂O₃ contents in coal can decrease HGI. The higher Fe₂O₃ content in coal can result in higher HGI. With the increase of micrinite and exinite contents in coal, the HGI has been decreased and higher vitrinite content in coal results in higher HGI. The multivariable studies have shown that input set of macerals, elemental analysis and moisture in non-linear condition can be achieved an acceptable correlation, R = 90.38%, versus R = 87.34% for the input set of macerals, ash and moisture. It is predicted that elemental analysis of coal can be a better representative of mineral matters for the prediction of HGI than ash.     

Estimation of free-swelling index based on coal analysis using multivariable regression and artificial neural network

The effects of proximate, ultimate and elemental analysis for a wide range of American coal samples on Free-swelling Index (FSI) have been investigated by multivariable regression and artificial neural network methods (ANN). The stepwise least square mathematical method shows that variables of ultimate analysis are better predictors than those from proximate analysis. The non linear multivariable regression, correlation coefficients (R²) from ultimate analysis inputs was 0.71, and for proximate analysis input variables was 0.49. With the same input sets, feed-forward artificial neural network (FANN) procedures improved accuracy of predicted FSI with R² = 0.89, and 0.94 for proximate and ultimate analyses, respectively. The ANN based prediction method, as a first report, shows FSI is a predictable variable, and ANN can be further employed as a reliable and accurate method in the free-swelling index prediction.     

Prediction of macerals contents of Indian coals from proximate and ultimate analyses using artificial neural networks

Coal, a prime source of energy needs in-depth study of its various parameters, such as proximate analysis, ultimate analysis, and its biological constituents (macerals). These properties manage the rank and calorific value of various coal varieties. Determination of the macerals in coal requires sophisticated microscopic instrumentation and expertise, unlike the other two properties mentioned above. In the present paper, an attempt has been made to predict the concentration of macerals of Indian coals using artificial neural network (ANN) by incorporating the proximate and ultimate analysis of coal. To investigate the appropriateness of this approach, the predictions by ANN are also compared with conventional multi-variate regression analysis (MVRA). For the prediction of macerals concentration, data sets have been taken from different coalfields of India for training and testing of the network. Network is trained by 149 datasets with 700 epochs, and tested and validated by 18 datasets. It was found that coefficient of determination between measured and predicted macerals by ANN was quite higher as well as mean absolute percentage error was very marginal as compared to MVRA prediction.     

煤种成浆性的人工神经网络预测模型研究

为更好地预测煤的成浆性,以大量煤种成浆浓度试验数据为基础,建立了3个输出因子的神经网络成浆浓度预测模型,模型采用L-M算法,对输入数据进行数据预处理,最后对比分析了神经网络预测模型与回归分析模型的预测结果。结果表明,以A_d、哈氏可磨性指数HGI和氧含量O为输入因子的模型预测结果平均绝对误差为0.63%,以M_(ad)、HGI和O为输入因子的模型预测结果平均绝对误差为0.60%,以M_(ad)、HGI和氧碳比O/C为输入因子的模型预测结果平均绝对误差为0.40%,3种组合的模型结果均小于回归分析模型的平均绝对误差1.15%。因此神经网络模型比回归分析模型有更好的预测能力,其中以M_(ad)、HGI和O/C为输入因子的神经网络模型预测结果最好。     

用神经网络对石蜡无催化剂氧化反应行为的研究

在自制的石蜡氧化装置上对石蜡进行了无催化剂氧化 ,并利用人工神经网络将氧化蜡的酸值和酯值与石蜡氧化条件 (反应温度、空气流量和反应时间 )进行关联。建立了石蜡无催化剂氧化的神经网络模型 ,并用该模型对石蜡的无催化剂氧化规律进行了预测。结果表明 ,该模型不但具有较高的计算精度 ,而且具有满意的预测能力     

Constrained optimization of combustion in a simulated coal-fired boiler using artificial neural network model and information analysis

Combustion in a boiler is too complex to be analytically described with mathematical models. To meet the needs of operation optimization, on-site experiments guided by the statistical optimization methods are often necessary to achieve the optimum operating conditions. This study proposes a new constrained optimization procedure using artificial neural networks as models for target processes. Information analysis based on random search, fuzzy c-mean clustering, and minimization of information free energy is performed iteratively in the procedure to suggest the location of future experiments, which can greatly reduce the number of experiments needed. The effectiveness of the proposed procedure in searching optima is demonstrated by three case studies: (1) a bench-mark problem, namely minimization of the modified Himmelblau function under a circle constraint; (2) both minimization of NO_x and CO emissions and maximization of thermal efficiency for a simulated combustion process of a boiler; (3) maximization of thermal efficiency within NO_x and CO emission limits for the same combustion process. The simulated combustion process is based on a commercial software package CHEMKIN, where 78 chemical species and 467 chemical reactions related to the combustion mechanism are incorporated and a plug-flow model and a load-correlated temperature distribution for the combustion tunnel of a boiler are used.     

Modeling, analysis and optimization of aircyclones using artificial neural network, response surface methodology and CFD simulation approaches

The pressure drop is an important performance parameter to evaluate and design cyclone separators. In order to accurately predict the complex non linear relationships between pressure drop and geometrical dimensions, a radial basis neural network (RBFNN) is developed and employed to model the pressure drop for cyclone separators. The neural network has been trained and tested by experimental data available in literature. The result demonstrates that artificial neural networks can offer an alternative and powerful approach to model the cyclone pressure drop. Four mathematical models (Muschelknautz method “MM”, Stairmand, Ramachandran and Shepherd & Lapple) have been tested against the experimental values. The residual error (the difference between the experimental value and the model value) of the MM model is the lowest. The analysis indicates the significant effect of the vortex finder diameter D_x and the vortex finder length S, the inlet width b and the total height H_t. The response surface methodology has been used to fit a second order polynomial to the RBFNN. The second order polynomial has been used to get a new optimized cyclone for minimum pressure drop using the Nelder-Mead optimization technique. A comparison between the new design and the standard Stairmand design has been performed using CFD simulation. CFD results show that the new cyclone design is very close to the Stairmand high efficiency design in the geometrical parameter ratio, and superior for low pressure drop at nearly the same cut-off diameter. The new cyclone design results in nearly 75% of the pressure drop obtained by the old Stairmand design at the same volume flow rate.     

Treatment of oily wastewater using low cost ceramic membrane: Comparative assessment of pore blocking and artificial neural network models

This work addresses the performance and modeling of the separation of oil-in-water (o/w) emulsions using low cost ceramic membrane that was prepared from inorganic precursors such as kaolin, quartz, feldspar, sodium carbonate, boric acid and sodium metasilicate. Synthetic o/w emulsions constituting 125 and 250 mg/L oil concentrations were subjected to microfiltration (MF) using this membrane in batch mode of operation with varying trans-membrane pressure differentials (ΔP) ranging from 68.95 to 275.8 kPa. The membrane exhibited 98.8% oil rejection efficiency and 5.36 × 10⁻⁶ m³/m² s permeate flux after 60 min of experimental run at 68.95 kPa trans-membrane pressure and 250 mg/L initial oil concentration. These experimental investigations confirmed the applicability of the prepared membrane in the treatment of o/w emulsions to yield permeate streams that can meet stricter environmental legislations (<10 mg/L). Subsequently, the experimental flux data has been subjected to modeling study using both conventional pore blocking models as well as back propagation-based multi-layer feed forward artificial neural network (ANN) model. Amongst several pore blocking models, the cake filtration model has been evaluated to be the best to represent the fouling phenomena. ANN has been found to perform better than the cake filtration model for the permeate flux prediction with marginally lower error values.     

基于模糊递归神经网络的污泥容积指数预测模型

污泥容积指数（SVI）,一个关键的污泥沉降性能评价指标。针对污水处理过程中污泥膨胀关键水质参数污泥容积指数难以准确在线测量,且实验室取样测量方法时间久、精度低,提出了一种改进型的模糊递归神经网络（HRFNN）用来预测污泥容积指数的变化,通过在网络第三层加入含有内部变量的反馈连接来实现输出信息的反馈。实验结果表明,与其他模糊神经网络相比,该网络的规模小、精度高,处理动态信息的能力明显加强。     

基于BP神经网络的煤与瓦斯突出危险性的预测研究

应用BP神经网络的理论和方法,结合贵州某矿山的实际情况,建立了基于BP神经网络的煤与瓦斯突出危险性预测的数学模型,通过数学软件matlab 7.0对煤层的突出危险程度进行了预测。结果显示,煤与瓦斯突出危险性的预测与实际情况相符,表明采用BP神经网络模型进行预测是可行的,为煤与瓦斯突出的预测提供了一种精度较高的方法。     

基于BP人工神经网络的生化处理水水质预测

神经网络具有强大的非线性映射能力和并行处理能力,近年来在水处理领域中被广泛地应用.利用神经网络算法对某污水处理厂的污水处理系统进行了出水水质预测.结果表明,基于BP神经网络的水质预测模型拟合效果较好,模拟出来的化学需氧量(COD)、pH、固体悬浮物(SS)及生物需氧量(BOD)的数值范围均较接近于实际值,其平均相对标准差分别为6.96%、1.31%、12.09%、15.18%.     

Experimental investigation,modeling and optimization of membrane separation using artificial neural network and multi-objective optimization using genetic algorithm

In this work, treatment of oily wastewaters with commercial polyacrylonitrile (PAN) ultrafiltration (UF) membranes was investigated. In order to do these experiments, the outlet wastewater of the API (American Petroleum Institute) unit of Tehran refinery, is used as the feed. The purpose of this paper was to predict the permeation flux and fouling resistance, by applying artificial neural networks (ANNs), and then to optimize the operating conditions in separation of oil from industrial oily wastewaters, including trans-membrane pressure (TMP), cross-flow velocity (CFV), feed temperature and pH, so that a maximum permeation flux accompanied by a minimum fouling resistance, was acquired by applying genetic algorithm as a powerful soft computing technique. The experimental input data, including TMP, CFV, feed temperature and pH, permeation flux and fouling resistance as outputs, were used to create ANN models. This fact that there is an excellent agreement between the experimental data and the predicted values was shown by the modeling results. Eventually, by multi-objective optimization, using genetic algorithm (GA), an optimization tool was created to predict the optimum operating parameters for desired permeation flux (i.e. maximum flux) and fouling resistance (i.e. minimum fouling) behavior. The accuracy of the model is confirmed by the comparison between the predicted and experimental data.     

Estimation of gross calorific value of coals using artificial neural networks

The gross calorific value (GCV) is an important property defining the energy content and thereby efficiency of fuels, such as coals. There exist a number of correlations for estimating the GCV of a coal sample based upon its proximate and/or ultimate analyses. These correlations are mainly linear in character although there are indications that the relationship between the GCV and a few constituents of the proximate and ultimate analyses could be nonlinear. Accordingly, in this paper a total of seven nonlinear models have been developed using the artificial neural networks (ANN) methodology for the estimation of GCV with a special focus on Indian coals. The comprehensive ANN model developed here uses all the major constituents of the proximate and ultimate analyses as inputs while the remaining six sub-models use different combinations of the constituents of the stated analyses. It has been found that the GCV prediction accuracy of all the models is excellent with the comprehensive model being the most accurate GCV predictor. Also, the performance of the ANN models has been found to be consistently better than that of their linear counterparts. Additionally, a sensitivity analysis of the comprehensive ANN model has been performed to identify the important model inputs, which significantly affect the GCV. The ANN-based modeling approach illustrated in this paper is sufficiently general and thus can be gainfully extended for estimating the GCV of a wide spectrum of solid, liquid and gaseous fuels.     

Proximate analysis based multiple regression models for higher heating value estimation of low rank coals

In this paper, multiple nonlinear regression models for estimation of higher heating value of coals are developed using proximate analysis data obtained generally from the low rank coal samples as-received basis. In this modeling study, three main model structures depended on the number of proximate analysis parameters, which are named the independent variables, such as moisture, ash, volatile matter and fixed carbon, are firstly categorized. Secondly, sub-model structures with different arrangements of the independent variables are considered. Each sub-model structure is analyzed with a number of model equations in order to find the best fitting model using multiple nonlinear regression method. Based on the results of nonlinear regression analysis, the best model for each sub-structure is determined. Among them, the models giving highest correlation for three main structures are selected. Although the selected all three models predicts HHV rather accurately, the model involving four independent variables provides the most accurate estimation of HHV. Additionally, when the chosen model with four independent variables and a literature model are tested with extra proximate analysis data, it is seen that that the developed model in this study can give more accurate prediction of HHV of coals. It can be concluded that the developed model is effective tool for HHV estimation of low rank coals.     

Prediction of water retention capacity of hydrolysed electrospun polyacrylonitrile fibers using statistical model and artificial neural network

Box Behnken design of experiment was used to study the effect of process variables such as alkali concentration, temperature and time on water retention capacity of the alkaline hydrolysed electrospun fibres. The hydrolysis of electrospun polyacrylonitrile fibres was carried out using sodium hydroxide with different processing conditions like concentration of alkali, temperature and time. With the increase in the concentration of alkali, time and temperature, the water retention capacity of membrane was found to increase in the membranes. Water retention capacities of the membranes were modeled and predicted using empirical as well as artificial neural network (ANN model). The fiber diameter and mean flow pore diameter of electrospun polyacrylonitrile fibers and hydrolyzed fibers shown in SEM images were 310 ± 50, 275 ± 75 nm, 0.9258 and 1.12 microns, respectively. The present study indicated that the nanofibrous membranes have potential for the water absorbing applications. © 2009 Wiley Periodicals, Inc. J Appl Polym Sci, 2009     

A novel adaptive sampling based methodology for feasible region identification of compute intensive models using artificial neural network

Identification of feasible region of operations in multivariate processes is a problem of interest in several fields. This is particularly challenging when the process model is black-box in nature and/or is computationally expensive, as analytical solutions are not available and the number of possible model evaluations is limited. An efficient methodology is required to identify samples where the model is evaluated for developing a computationally efficient surrogate model. In this work, an artificial neural network based surrogate model is proposed which is integrated with a statistical-based approach (Jack-knifing) to estimate the variance of the surrogate model prediction. This allows implementation of an adaptive sampling approach where new samples are identified close to the feasible region boundary or in regions of high prediction uncertainty. The proposed approach performs better than a previously published kriging based method for different dimensionality case studies.     

人工神经网络-基团键贡献耦合模型预测煤液化油的偏心因子

为探索预测煤直接液化油窄馏分的偏心因子的新方法,建立了基于人工神经网络-基团键贡献耦合模型(ANN-GBC),以煤直接液化油包含的45个基团键和常压沸点(T_b)共46个参数作为该模型的输入参数,研究了煤直接液化油15个窄馏分的偏心因子与分子结构之间的相关性。结果表明,通过计算20个模型化合物的偏心因子,表明ANN-GBC模型具有较好的模拟推算功能,计算值与理论值平均相对误差均在2.5%以下。偏心因子ω随蒸馏切割馏分温度的升高而增大,ANN-GBC模型预测值普遍高于Watanasiri、NEDOL关联式的计算值。380℃馏分ω小于1,相对偏差较小;380℃馏分ω偏差较大;针对420℃馏分,因仅能定性定量分析其中20%物质,不同物质的含量差异导致个别结果的跳跃,ω偏差较大。     

基于遗传BP神经网络预测硫在高含硫气体中溶解度

为更精确地关联预测硫在高含硫气体中的溶解度,提出将遗传算法(GA)和LM-反向传播神经网络(LM-BP ANN)相结合的预测模型。设计了该模型的计算过程,讨论了模型参数的设置。以温度、压力和气体组分作为BP神经网络预测模型的输入变量,利用GA优化了BP神经网络的初始权值和阈值,采用遗传算法优化后的BP神经网络计算了元素硫在高含硫气体中的溶解度。结果表明,该模型训练结果与实测值之间的平均相对误差为5.90%,测试结果与实测值的平均相对误差为5.54%;该方法较BP神经网络模型具有预测精度高、收敛速度快的优点;该模型具有较好的模拟及内推、外推功能。     

Artificial neural network approaches on composition-property relationships of jet fuels based on GC-MS

The relationships of composition-properties of 80 jet fuels concerning chemical compositions and several specification properties including density, flashpoint, freezing point, aniline point and net heat of combustion were studied. The chemical compositions of the jet fuels were determined by GC-MS, and grouped into eight classes of hydrocarbon compounds, including n-paraffins, isoparaffins, monocyclopraffins, dicyclopraffins, alkylbenzens, naphthalenes, tetralins, hydroaromatics. Several quantitative composition-property relationships were developed with three artificial neural network (ANN) approaches, including single-layer feedforward neural network (SLFNN), multiple layer feedforward neural network (MLFNN) and general regressed neural network (GRNN). It was found that SLFNNs are adequate to predict density, freezing point and net heat of combustion, while MLFNNs produce better results as far as the flash point and aniline point prediction are concerned. Comparisons with the multiple linear regression (MLR) correlations reported and the standard ASTM methods showed that ANN approaches of composition-property relationships are significant improvement on MLR correlations, and are comparable to the standard ASTM methods.