Comparison of hydrogenated shale oils with standard jet fuels

The characteristics of jet fuels obtained from typical U.S. shale oils (Geokinetics, Occidental, Paraho and Tosco II) were compared with standard petroleum jet fuels in order to study the possibility of using these shale oils as a substitute. The shale oil fractions distilling below 343°C were catalytically hydroprocessed at low, medium and high severities and fractionated to the jet fuel range (121–300°C). The hydroprocessed products and jet fuels were compared for composition and physical properties. High severity hydroprocessing of shale oils decreased the nitrogen, sulfur, olefin and aromatic content while increasing the hydrogen content. The nitrogen content in shale oil jet fuels was considerably higher even after the high severity treatment. The aromatic content, except in Paraho shale oil, was relatively higher and the hydrogen content was slightly lower. Sulfur and olefin contents were lower at all severities. The physical properties and heat of combustion, except the high freezing point of shale oil jet fuels, were comparable to those of standard petroleum jet fuels.     

A cost-effective approach to generate accurate correlations via analyzing a minimum number of data points; Case study on the convection heat transfer problem

It is very common in the heat transfer area to analyze and design heat equipment using the past available heat correlations. Basically, demanding higher-accuracy correlations enforces the heat laboratories to test and collect larger banks of laboratory data. However, this conversely affects the laboratory cost. Therefore, it becomes challenging to create new approaches that let the correlation developers use smaller experimental datasets and provide correlations with sufficient accuracies. To surmount this challenge, the present work develops a new approach that benefits from the computational fluid dynamics method as a reliable and cheap tool and adequately enriches the original, insufficient dataset. Then, suitable enhanced correlations are developed using the new enriched experimental-numerical-based dataset. In parallel, the artificial neural network (ANN) is used to enrich the original insufficient dataset separately. Using this experimental-ANN-based dataset, it provides a totally ANN-based correlation. It is shown that the results of enhanced correlations are as accurate as those of the ANN-based correlation. However, the point is that the use of the present approach is about 100 times faster than using the ANN. The typical forced convection heat transfer through a pipe is examined here to show the capabilities of the current approach.     

人工神经网络在含氧有机化合物定量结构-保留关系中的应用

采用人工神经网络（ANN）对一系列含氧有机化合物的气相色谱保留指数建立定量结构-保留关系（QSRR）的模型,和多元线性回归（MLR）预测结果进行比较。在OV-1固定相上,相关系数R分别为0.989 1,0.991 1;在SE-54上,相关系数R分别为0.989 2,0.991 7。结果表明：MLR建立的模型优于ANN,具有良好的拟合度和预测精度。     

应用基于粗集的模糊神经网络进行软测量建模的研究

提出将软测量建模与数据挖掘方法相结合的思想。针对模糊神经网络输入维数高，且对应的神经网络是权值不完全连接的网络，结构简单、训练速度快。将该方法用于催化裂化装置的轻柴油凝点的估计，取得良好的效果。     

人工神经网络用于氟化酚的定量构效关系研究

采用误差反传前向人工神经网络（ANN）建立了16种氟化酚的结构与其对梨形四膜虫的毒性之间的定量结构-活性关系（QSAR）模型。以16种氟化酚的量子化学和理化参数作为输入,对梨形四膜虫的急性毒性作为输出,采用内外双重验证的办法分析和检验所得模型的稳定性和外部预测能力,所构建网络模型的相关系数为0.999 8、交叉检验相关系数为0.981 8、标准偏差为0.01、残差绝对值≤0.04,应用于外部预测集,外部预测集相关系数为0.993 6;而多元线性回归（MLR）法模型的相关系数为0.980 2、标准偏差为0.119、残差绝对值≤0.28,外部预测集相关系数为0.980 3。结果表明,ANN模型获得了比MLR模型更好的拟合效果。     

Prediction of the colorimetric parameters and mass loss of heat‐treated bamboo: Comparison of multiple linear regression and artificial neural network method

In this study, the colorimetric parameters (L*, a*, b*) and mass loss of heat‐treated bamboo were investigated, and the obtained results were modeled by using two methods: multiple linear regression (MLR) and artificial neural network (ANN). First, bamboo samples were exposed to heat treatment at different temperatures (110°C, 140°C, 170°C, and 200°C) and durations (15, 30, 45, 60, 75, 90, and 115 minutes) in a laboratory oven. Then, the colorimetric parameters (L*, a*, b*) and mass loss of each sample were measured after each period of heat treatment. All data were modeled by using two methods separately for each parameter and the performances of these proposed methods were compared. It was found that color change and mass loss increased with increasing temperature and duration of heat treatment. Mean absolute percentage error (MAPE) values of all obtained MLR ranged from 0.64% to 10.63%, while the all MAPE values of ANN were found to be lower than 1.5%. Based on these results, it can be said that MLR and ANN could be used to evaluate the changes on the selected properties of heat‐treated bamboo samples. On the other hand, it should be emphasized that the ANN gave more accurate results than the MLR method because of its learning capability.     

Gas Analysis by In Situ Combustion in Heavy‐Oil Recovery Process: Experimental and Modeling Studies

Enormous efforts have been made to facilitate produced‐gas analyses by in situ combustion implication in heavy‐oil recovery processes. Robust intelligence‐based approaches such as artificial neural network (ANN) and hybrid methods were accomplished to monitor CO₂/O₂/CO. Implemented optimization approaches like particle swarm optimization (PSO) and hybrid approach focused on pinpointing accurate interconnection weights through the proposed ANN model. Solutions acquired from the developed approaches were compared with the pertinent experimental in situ combustion data samples. Implication of hybrid genetic algorithm and PSO in gas analysis estimation can lead to more reliable in situ combustion quality predictions, simulation design, and further plans of heavy‐oil recovery methods.     

Determination of hydrogen content, gross heat of combustion, and net heat of combustion of diesel fuel using FTIR spectroscopy and multivariate calibration

The precise determination of the heat of combustion is of great importance for trading automotive diesel. The net heat of combustion (NHC) of fuel is related to the hydrogen elemental composition of fuel as obtained by elemental analysis. Heat of combustion expressed as gross heat of combustion (GHC) and net heat of combustion (NHC) have been predicted from data obtained by proximate analysis (density, ash, water and sulphur content) (ASTM D4868). GHC was obtained using bomb calorimetry (ASTM D240). The results of ASTM D4868 and ASTM D240 were found in good agreement. GHC and NHC fall within the relatively narrow range 45.24-46.08 and 41.91-43.27 MJ/kg, respectively. GHCs of tested diesel samples are, on average, about 7% greater than NHCs. The present paper also present a simple analytical method for determination of hydrogen content, GHC, and NHC of automotive diesel fuel using FTIR spectroscopy and partial-least squares calibration (PLS-1). PLS-1 had a high prediction power for prediction of hydrogen from FTIR spectra of diesel samples. The spectral ranges used in calibration were 400-670 and 2846-2970 cm⁻¹. On the other hand, classical least squares calibration (CLS) was found invalid for determination of hydrogen content in diesel. The results obtained by the proposed analytical method were almost to those obtained by ASTM D4868 and ASTM D240. PLS-1 method, offers a simple and reliable analytical method for quantification of hydrogen content in diesel samples without running expensive analysis like those carried out using carbon, hydrogen, and nitrogen (CHN) instruments.     

MODELING AND OPTIMIZATION OF MEMBRANE CHEMICAL CLEANING BY ARTIFICIAL NEURAL NETWORK,FUZZY LOGIC,AND GENETIC ALGORITHM

Robust artificial neural network (ANN) and fuzzy logic (FL) models were derived for chemical cleaning of microfiltration membranes fouled by milk under a wide range of operating conditions. The accuracies of the models were compared with multiple linear regressions (MLR). The developed models are useful tools for predicting the performance of chemical cleaning. The effects of different operating conditions on cleaning performance were elucidated using the ANN developed model. Moreover, optimum cleaning condition was determined by genetic algorithm and ANN model. The current research demonstrated that fuzzy logic and an artificial neural network can quantitatively capture cumulative effects of a range of operating conditions on flux recovery and resistance removal during a cleaning process.     

人工神经网络用于三苯基丙烯腈衍生物的定量结构-活性关系模型

采用人工神经网络(ANN)BP算法探讨了24个三苯基丙烯睛衍生物的lg1/C(C为半致死浓度)与X位羟基指示数I、分子表面积SA和B环上原子净电荷之和QB之间的关系,以20个样本为训练集建立了定量结构-活性关系(QSAR)模型,其相关系数和标准偏差分别为R=0.9969和SD=0.0164,其余4个样本为测试集,得到R=0.9913和SD=0.1533;用多元线性回归(MLR)方法建立的QSAR模型R=0.9360,SD=0.3779。结果表明,ANN方法具有良好的预测能力,比MLR方法更精密。     

Thermal analysis of alternative diesel fuels from vegetable oils

The relatively poor cold-flow properties of monoalkyl esters of vegetable oils and animal fats (biodiesel) present a major obstacle to their development as alternative fuels and extenders for combustion in direct injection compressionignition (diesel) engines. In this work, differential scanning calorimetry (DSC) heating and cooling curves of methyl soyate (SME), methyl tallowate (TME), SME/TME admixtures, and winterized SME were analyzed. Completion of melt, crystallization onset (Onset), and other temperatures corresponding to melting and freezing peaks were correlated to predict cloud point (CP), pour point (PP), cold filter plugging point (CFPP), and low-temperature flow test (LTFT) data. Effects of treating methyl esters with cold-flow improvers were examined. Low-temperature flow properties of biodiesel may be accurately inferred from subambient DSC analyses of high-melting or freezing (β-form) peaks. The temperature of maximal heat flow for freezing peaks gave the best accuracy for predicting CP, PP, and CFPP, while freezing point gave the best accuracy for predicting LTFT. Onset also gave good correlations with respect to predicting PP, CFPP, and LTFT. Cooling scan parameters were more reliable than heating scan parameters. Presented at the 88th American Oil Chemists’ Society’s Annual Meeting & Expo, Seattle, Washington, May 11–14, 1997.     

SYSTEMATIC DERIVATIONS OF MODEL PREDICTIVE CONTROL BASED ON ARTIFICIAL NEURAL NETWORK

This article presents systematic derivations of setting up a nonlinear model predictive control based on the artifical neural network. Unlike most research in the past, the control law is mathematically developed in detail so that the performance of the ANN-based controller can be improved. In this paper, a three-layer feedforward neural network with hyperbolic tangent functions in the hidden layer and with a linear function in the output layer is used. The two-stage scheme including pseudo Gauss-Newton and least squares is proposed for training ANN. This training method is better than the traditional algorithm in terms of training speed. The Levenberg-Marquardt approximation is also utilized for the minimum of the predictive control criterion. Two typical chemical processes are simulated and the ANN model predictive control applications can reach fairly good results.     

Artificial and Hybrid Fuzzy Linear Neural Network‐Based Estimation of Seed Oil Content of Safflower

Inexpensive and rapid methods for measurement of seed oil content by near infrared reflectance spectroscopy (NIRS) are useful for developing new oil seed cultivars. Adopting default multiple linear regression (MLR), the predictions of safflower oil content were made by 20–140 samples using a Perten Inframatic 8620 NIR spectrometer. Although the obtained interpolation results of MLR had desired accuracy, the extrapolation was extremely poor. The extrapolation determination coefficient (R²) and standard error (SE) of cross validation for MLR models were 0.63–0.78 and 3.71–4.44, respectively. In order to overcome the accuracy limitation of linear MLR models, a common suggestion is to use a nonlinear artificial neural network (ANN); however, it needs a large number of data to yield significant accurate results. We developed a novel robust hybrid fuzzy linear neural (HFLN) network to capture simultaneously linear and nonlinear patterns of data with a limited number of safflower samples. Empirical extrapolation results showed that the HFLN had higher R² (=0.85) and lower SE (=1.83) compared to those obtained by MLR and ANN models. It is concluded that hybrid methodologies could be used to construct efficient and appropriate models for estimation of seed oil content set up on NIR system.     

Thermal conductivity prediction of foods by Neural Network and Fuzzy (ANFIS) modeling techniques

A neuro-fuzzy modeling technique was used to predict the effective of thermal conductivity of various fruits and vegetables. A total of 676 data point was used to develop the neuro-fuzzy model considering the inputs as the fraction of water content, temperature and apparent porosity of food materials. The complexity of the data set which incorporates wide ranges of temperature (including those below freezing points) made it difficult for the data to be predicted by normal analytical and conventional models. However the adaptive neuro-fuzzy model (ANFIS) was able to predict conductivity values which closely matched the experimental values by providing lowest mean square error compared to multivariable regression and conventional artificial neural network (ANN) models. This method also alleviates the problem of determining the hidden structure of the neural network layer by trial and error.     

Neural network (ANN) approach to biodiesel analysis: Analysis of biodiesel density, kinematic viscosity, methanol and water contents using near infrared (NIR) spectroscopy

The use of ethanol and biodiesel, which are alternative fuels or biofuels, has increased in the last few years. Modern official standards list 25 parameters that must be determined to certify biodiesel quality, and these analyses are expensive and time-consuming. Near infrared (NIR/NIRS) spectroscopy (4000-12,820 cm⁻¹) is a cheap and fast alternative to analyse biodiesel quality, when compared with infrared, Raman, or NMR methods, and quality control can be done in realtime (on-line).We compared the performance of linear and non-linear calibration techniques - namely, multiple linear regression (MLR), principal component regression (PCR), partial least squares regression (PLS), polynomial and Spline-PLS versions, and artificial neural networks (ANN) - for prediction of biodiesel properties from near infrared spectra. The model was created for four important biodiesel properties: density (at 15 °C), kinematic viscosity (at 40 °C), water content, and methanol content. We also investigated the influence of different pre-processing methods (Savitzky-Golay derivatives, orthogonal signal correction) on the model prediction capability. The lowest root mean squared errors of prediction (RMSEP) of ANN for density, viscosity, water percentage, and methanol content were 0.42 kg m⁻³, 0.068 mm² s⁻¹, 45 ppm, and 51 ppm, respectively. The artificial neural network (ANN) approach was superior to the linear (MLR, PCR, PLS) and “quasi”-non-linear (Poly-PLS, Spline-PLS) calibration methods.     

基于神经网络和遗传算法的螺旋折流板换热器性能预测

将人工神经网络（ANN）应用于非连续螺旋折流板换热器的壳程换热和流阻分析。中试试验研究了具有3个螺旋角和2种管型的换热器。作为人工神经网络最常用的一种类型,将多层感知器神经网络（MLP）应用于本研究,使用一定的实验数据进行网络训练及预测。应用遗传算法（GA）对MLP的初始权值和阈值进行优化,预测结果精确。通过比较不同网络结构的预测误差来选择最适宜的网络结构为9-7-5-2。和关联结果比较可知MLP-GA网络对于换热器性能预测更加适合。此外,当使用MLP-GA方法在训练数据范围以外对壳程换热系数和压降进行预测时,网络预测结果和实验结果吻合程度也较高。因此,MLP-GA混合算法能够用来预测螺旋折流板管壳式换热器的传热和水力学性能。     

费托合成油和石油基油加工产品对比分析

费托合成及煤化工的发展能够有效缓解我国对石油资源的依赖。本文针对低温煤基合成油与传统石油基石脑油、煤油、柴油、润滑油和蜡产品性能进行对比。文章指出与传统石油基产品对比,费托合成油产品具有蜡含量高、无硫、无氮、少芳烃的特性,满足清洁油品的环保要求,同时在高档润滑油基础油和高熔点石蜡等高附加值产品生产方面更具竞争优势。但费托合成油加工的油品普遍存在凝点、冰点、密度等相关关键指标不合格问题,不同油品的生产,均需要通过异构、精制、裂化、重整等技术改善油品的低温流动性,并通过切割、掺炼等工艺改进才能生产符合标准的油品。文章提出结合我国清洁燃料消费及能源结构调整的变化,费托合成工艺和煤制油应发挥高端产品优势,延伸加工产业链,实现粗放型加工向产业链高端迈进。