Dynamic crude oil fouling prediction in industrial preheaters using optimized ANN based moving window technique

The objective of this paper is to develop and validate a reliable, efficient and robust artificial neural network (ANN) model for online monitoring and prediction of crude oil fouling behavior for industrial shell and tube heat exchangers. To explore the complex dynamics of fouling, a new modeling strategy based on moving-window neural network approach is proposed. The essential character of this modeling approach is online updating of the ANN model whenever a new data block is available, so that it can effectively capture the slowly changing of process dynamics. The results of these models have been compared with appropriate sets of experimental data. The mean relative errors (MRE) of training and prediction subsets were about 6.61% and 8.06%, respectively. Since the data extraction in the refinery was performed every 2 h, the modeling approach led to an MRE of about 8% for fouling rate prediction of the next 50 h.     

Fatigue life prediction of thermal barrier coatings using a simplified crack growth model

Models that can predict the life of thermal barrier coatings (TBCs) during thermal cycling fatigue (TCF) tests are highly desirable. The present work focuses on developing and validating a simplified model based on the relation between the energy release rate and the TCF cycles to failure. The model accounts for stresses due to thermal mismatch, influence of sintering, and the growth of TGO (alumina and other non-protective oxides). The experimental investigation of TBCs included; 1) TCF tests at maximum temperatures of 1050 °C, 1100 °C, 1150 °C and a minimum temperature of 100 °C with 1 h and 5 h (1100 °C) hold times. 2) Isothermal oxidation tests at 900, 1000 and 1100 °C for times up to 8000 h. The model was calibrated and validated with the experimental results. It has been shown that the model is able to predict the TCF life and effect of hold time with good accuracy.     

Development of a shrinkage performance specifications and prediction model analysis for supplemental cementitious material concrete mixtures

The unrestrained shrinkage along with restrained cracking tendency of concrete mixtures typically used by the Virginia Department of Transportation (VDOT) were assessed to establish an appropriate limit on drying shrinkage for use in a performance specification. Five existing shrinkage prediction models were assessed to determine the accuracy and precision of each model as it pertains to the VDOT mixtures used in this study. The five models assessed were the ACI 209 Code Model, Bazant B3 Model, CEB90 Code Model, Gardner/Lockman Model, and the Sakata Model. The Gardner/Lockman Model performed best for the supplemental cementitious material (SCM) mixtures.Based on a comparison of the unrestrained drying shrinkage and restrained cracking tendency, it was determined that the potential for cracking could be minimized by limiting the unrestrained shrinkage of the concrete mixtures. The recommended percentage length change specification limits for the supplemental cementitious material mixtures are 0.0400 at 28 days and 0.0500 at 90 days.     

Rubber fatigue life prediction using a random forest method and nonlinear cumulative fatigue damage model

In this study, a random forest machine-learning method is introduced on the basis of the analysis of measured constant amplitude stress fatigue data. This method aims to predict rubber fatigue life under constant amplitude stress. Strain mean value, strain amplitude, and strain ratio are used as independent variables, and the prediction model of rubber fatigue life under constant amplitude stress is established. A nonlinear cumulative fatigue damage model is proposed to calculate rubber fatigue life under the variable amplitude stress. Results show that the random forest method has high precision and generalization capability for rubber fatigue life prediction under constant amplitude stress and the nonlinear cumulative fatigue damage model could be employed to calculate the fatigue life of rubber under variable amplitude stress with enough accuracy according to the constant amplitude stress fatigue life data. This research can provide a reference for rubber fatigue life prediction. © 2019 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2020 , 137, 48519.     

Failure prediction of adhesively bonded CFRP-Aluminum alloy joints using cohesive zone model with consideration of temperature effect

To predict the failure of adhesively bonded CFRP (Carbon Fiber Reinforced Plastics)-aluminum alloy joints applied to High Speed EMU (Electric Multiple Units) more accurately with consideration of temperature influence, a combined experimental-numerical approach is developed in this study. Bulk specimens and adhesive joints, including thick-adherend shear joints(TSJ), scarf joints(SJ) with scarf angle 30°(SJ30°), 45°(SJ45°), and 60°(SJ60°), as well as butt joints(BJ), were manufactured and tested at 23°C (room temperature, RT), 80°C (high temperature, HT) and ?40°C (low temperature, LT). Quadratic stress criteria built at different temperatures were introduced in the cohesive zone mode (CZM) to conduct a simulation analysis. Test results suggest that the effects of HT on mechanical properties of adhesive are more obvious than the effects of LT. It is also found that TSJ show the greatest improvements in failure strengths at LT due to the occurrence of cohesive failure, while SJ and BJ tend to develop fiber tears due to the presence of normal stress. Stress distributions of adhesive layer are found to be symmetrical except for the normal stress of SJ. This simulation analysis shows that the prediction accuracy is related to quadratic stress criteria applied, and that the relative errors of prediction results are less than 7.5% for engineering applications.     

An architecture of deep learning in QSPR modeling for the prediction of critical properties using molecular signatures

Deep learning rapidly promotes many fields with successful stories in natural language processing. An architecture of deep neural network (DNN) combining tree-structured long short-term memory (Tree-LSTM) network and back-propagation neural network (BPNN) is developed for predicting physical properties. Inspired by the natural language processing in artificial intelligence, we first developed a strategy for data preparation including encoding molecules with canonical molecular signatures and vectorizing bond-substrings by an embedding algorithm. Then, the dynamic neural network named Tree-LSTM is employed to depict molecular tree data-structures while the BPNN is used to correlate properties. To evaluate the performance of proposed DNN, the critical properties of nearly 1,800 compounds are employed for training and testing the DNN models. As compared with classical group contribution methods, it can be demonstrated that the learned DNN models are able to provide more accurate prediction and cover more diverse molecular structures without considering frequencies of substructures.     

Multiple order model migration and optimal model selection for online glucose prediction in Type 1 diabetes

To address the problem of insufficient available modeling data for glucose prediction, as well as modeling burden, a model migration method was developed in a previous work to quickly transfer an old model to a new subject using a simple parameter adjustment. However, this method, which is referred to as first‐order model migration (FOMM), only considers a single order for each exogenous input, and may not produce an optimal model structure for accurate prediction. To overcome this problem, a multiple order model migration (MOMM) algorithm is proposed in this study. For different numbers of modeling samples, including glucose and two exogenous inputs (meal and insulin), the optimal modeling method may be different, and therefore, must be properly determined for each modeling scenario. First, the optimal model order is determined for each input and a multiple order prediction model is used. Then, a MOMM algorithm is developed based on particle swarm optimization (PSO) to simultaneously revise multiple parameters. The multiple order parameters of each input in the old model are quickly customized so that the revised model can be used for new subjects with desirable prediction accuracy. In particular, the influence of the number of modeling samples is analysed to check the applicability of different methods; this analysis determines the appropriate selection guidelines for the optimal model in response to different data sizes. The proposed method was evaluated using thirty in silico subjects and clinical data from seven individuals with Type 1 diabetes mellitus (DM). Overall, the MOMM algorithm presented superior results when the time period for collecting the samples was larger than 10 h (50 samples). In particular, the size of the modeling samples was separated into three different regions by evaluating the glucose prediction performance and the comparison between different algorithms for both in silico and clinical subjects. In Region I, the FOMM method achieves the best performance. In Region II, the MOMM method should be used and the prediction accuracy is superior in Region II in general. With enough samples (Region III), the subject‐dependent model (SM) algorithm can be chosen. The MOMM algorithm is demonstrated to be able to transfer models for new subjects with improved model structure. © 2017 American Institute of Chemical Engineers AIChE J, 64: 822–834, 2018     

速生桉木硫酸盐法制浆过程中甲醇生成量的预测模型

以速生桉木硫酸盐法制浆为研究对象,考察了制浆过程中有效碱浓度、H因子(时间和温度)对甲醇生成量以及残余有效碱浓度变化的影响,研究发现制浆过程残余有效碱浓度与蒸煮H因子的负指数呈线性关系,结合这一关系从动力学角度建立了引入碱浓动态变化的制浆过程中预测甲醇生成量的数学模型。结果表明该模型能很好地预测速生桉木硫酸盐法蒸煮过程中各工艺条件下的甲醇生成量(R² 0.990)。模型的预测效果表明,温度对甲醇生成量影响显著,而初始有效碱浓度对甲醇生成量影响较小。实际生产中,可根据该模型对各工艺条件对甲醇生成量影响程度的预测效果适当改变工艺条件,以控制制浆过程中单位木素脱除量的甲醇生成量。     

Deducing ink thickness variations by a spectral prediction model

Most existing techniques for regulating the ink flow in offset presses rely on density measurements carried out on specially printed patches. In the present contribution, we develop a methodology to deduce ink thickness variations from spectral measurements of multichromatic halftone patches located within the printed page. For this purpose, we extend the Clapper‐Yule spectral reflectance prediction model by expressing the transmittance of the colorants composed of superposed inks as a function of the ink transmittances and of fitted ink layer thicknesses. We associate to each ink an ink thickness variation factor. At print time, this ink thickness variation factor can be fitted to minimize a difference metric between predicted reflection spectrum and measured reflection spectrum. The ink thickness variations deduced from multichromatic halftones allow to clearly distinguish between normal ink volume, reduced ink volume, or increased ink volume. This information can then be used for performing control operations on the printing press. © 2009 Wiley Periodicals, Inc. Col Res Appl, 34, 432–442, 2009     

碳氢燃料JP-10高温液态黏度测量和推算模型构建方法研究

基于流体动力学层流哈根-泊肃叶（Hagen-Poiseuille）定律,利用双毛细管法,对高密度空天动力燃料JP-10液态黏度进行实验测量,测温范围326.6~671.2 K,测量压力2.0、 3.0、 4.0 MPa,扩展相对不确定度2.88%~4.96%（置信因子k=2）。通过纯物质环己烷动力黏度的测量,对实验系统进行了标定,实验结果与NIST数据库平均相对偏差在1.22%以内,最大相对偏差绝对值为2.04%,实验结果与推荐黏度值在2.0 MPa时平均相对偏差为1.25%,4.0 MPa时平均相对偏差为1.61%,最大相对偏差绝对值为3.50%,验证了实验系统的可靠性。选取临界压力状态的黏度值作为参考状态值,通过引用Yaws液相有机化合物的黏度经验公式,结合SRK状态方程对绝对速率理论黏度模型进行了改进,耦合实验数据,建立了一种适用于碳氢燃料的高温高压液相黏度的推算模型。采取共轭梯度法和遗传算法对模型参数进行拟合,计算结果与实验结果的平均相对偏差值在2.00%以内,最大相对偏差绝对值小于4.50%,验证了模型的精确性。     

两段法固定床甲醇制芳烃产物预测多元非线性回归模型

针对甲醇制芳烃机理建模假设多、产物预测过程复杂、计算成本高等问题,提出了一种多元非线性回归分析的数据驱动建模方法。在两段法固定床反应系统中研究了反应压力、甲醇体积空速、一段中心温度、二段中心温度、装置运行时间、累计甲醇进料量及其交互作用对芳烃产物的影响。利用最小二乘法进行参数估计,建立四个六元二次非线性产物分布回归模型。测试结果表明,测试样本的总体决定性指标{\rho }^{\mathrm{2}}为0.9576,均方误差MSE为0.0037,相比于传统的动力学模型,具有产物预测精确度高、计算量小以及泛化性强的特点;空速和反应压力的交互作用对产物中芳烃选择性影响显著,在空速增大的情况下,产物中最大芳烃占比对应的压力峰值也随之增高。     

A prediction model for the critical gas velocity and pressure gradient of continuous liquid-carrying in inclined pipes

A generic model for the prediction of critical gas velocity and pressure gradient in slightly inclined pipes (β ≤ 6°) is presented in this article. The gas–liquid configuration was determined based on the minimum energy principle and consideration of wettability and surface tension. A visualization experiment was conducted to obtain the critical gas velocity and the critical pressure gradient of a gas–liquid flow through the 40 and 60-mm pipe diameter. The theoretical study shows that the configuration is close to a convex interface shape at the critical conditions, which is in accord with the experimental phenomenon. Experimental study shows interfacial waves are the main cause of increased interfacial friction factor and a linear functional relationship between the inclination angle and the flow correction factor f(β). The results demonstrate that the new model is capable of providing satisfactory prediction results for the critical gas velocity, pressure drop, and liquid holdup.     

配煤对煤灰熔点的影响及灰熔点预测模型研究

针对皖北刘桥二矿煤（A）属于高灰熔点煤,无法满足Shell气化炉液态排渣的需要。考察了采用配煤技术降低煤A的灰熔点的效果,结果表明,配煤可以显著的降低煤A的高灰熔融性。使其能够满足Shell气化炉液态排渣工艺的要求。并采用最小二乘法对灰熔点与煤灰灰成分之间建立并回归了预测模型,预测模型方程表明,若能增加配煤煤灰中MgO的含量可显著降低煤灰熔点,增加配煤煤灰中CaO的含量可使煤灰熔点降低,在煤灰中SiO2和Al2O3总含量一定的条件下,高硅低铝的配煤煤灰可进一步降低煤灰熔点。同时该模型能较好地预测三种原煤配煤的灰熔点。     

某油田储层测井解释模型的建立

为提高油、气、水在储层内部的分布和流动特性等参数的计算精度,在对某油田储层测井参数解释模型研究中,进行基础性数据整理,建立了储量计算所需参数的孔隙度和渗透率解释与预测模型。     

高温陶瓷过滤管性能退化建模及实时寿命预测

高温陶瓷过滤管由内部孔径较大的支撑体和外部孔径较小的过滤膜双层结构构成,在实际应用中,存在大量粒径较小的粉尘颗粒,会穿过过滤膜沉积到支撑体内部,脉冲反吹无法有效清除. 因支撑体内颗粒沉积及管壁外残余粉尘层不断压缩,使陶瓷过滤管渗透率不断下降,残余压降逐渐增加。本工作基于高温陶瓷过滤管壁内颗粒沉积特性及残余粉尘层压缩不可直接观测的特点,结合贝叶斯估计理论,利用过滤管运行期间采集的残余压降数据,提出一种基于状态空间模型的过滤管性能退化建模方法。该方法能融入最新采集到的残余压降数据,实时对模型参数进行更新,可对陶瓷过滤管的剩余寿命进行实时预测,同时对陶瓷过滤管剩余寿命的失效概率密度分布及陶瓷过滤管的退化状态变化率进行预测。对某高温试验装置及壳牌煤气化装置中的陶瓷过滤管残余压降分析表明,预测剩余寿命准确率随残余压降数据增加而逐渐增加,后期预测准确率高于95%,且陶瓷过滤管退化状态变化率逐渐变小,与陶瓷过滤管残余压降前期增加快后期增加慢的现象一致。     

Dynamic rate-based and equilibrium models for a packed reactive distillation column

Dynamic rate-based and equilibrium models were developed for a packed reactive distillation column for the production of tert-amyl methyl ether (TAME). The two types of models, consisting of differential algebraic equations, were implemented in gPROMS and dynamic simulations were carried out to study the dynamic behaviour of reactive distillation of the TAME system. The dynamic responses predicted by the two types of models are similar in general, with some differences in steady-state values. The influence of manipulated variables, such as distillate flow rate, reflux ratio, and reboiler duty, on the dynamic responses of the controlled variables (reactant conversion and product purity) was studied. The dynamic response of reactant conversion is very nonlinear and unconventional, but the response of product purity is well approximated by a linear first-order differential equation. The CPU time required to complete a dynamic simulation of the rate-based model is at least an order of magnitude higher than that for the equilibrium model. Therefore, the dynamic rate-based model is much more complicated than the equilibrium model, and simplification of the rate-based model is necessary for the implementation of model-based control. A new approach was proposed to simplify the dynamic rate-based model by assuming that the mass transfer coefficients are time invariant. This approach was demonstrated to be superior to the conventional simplification methods. It can reduce the number of equations by up to two-thirds and still accurately predicts the dynamic behaviour.     

Online prediction of subcutaneous glucose concentration for type 1 diabetes using empirical models and frequency‐band separation

Online glucose prediction which can be used to provide important information of future glucose status is a key step to facilitate proactive management before glucose reaches undesirable concentrations. Based on frequency‐band separation (FS) and empirical modeling approaches, this article considers several important aspects of on‐line glucose prediction for subjects with type 1 diabetes mellitus. Three issues are of particular interest: (1) Can a global (or universal) model be developed from glucose data for a single subject and then used to make suitably accurate on‐line glucose predictions for other subjects? (2) Does a new FS approach based on data filtering provide more accurate models than standard modeling methods? (3) Does a new latent variable modeling method result in more accurate models than standard modeling methods? These and related issues are investigated by developing autoregressive models and autoregressive models with exogenous inputs based on clinical data for two groups of subjects. The alternative modeling approaches are evaluated with respect to on‐line short‐term prediction accuracy for prediction horizons of 30 and 60 min, using independent test data. © 2013 American Institute of Chemical Engineers AIChE J 60: 574–584, 2014     

基于经济模型预测控制的金氰化浸出过程动态实时优化

针对金氰化浸出过程时间常数大、不确定性强等问题,提出了一种基于经济模型预测控制(EMPC)的动态实时优化方法。不同于传统的模型预测控制,EMPC将经济指标直接作为滚动优化的目标函数,在每个采样时刻求解滚动窗口内的最优操作序列。和稳态优化方法相比,基于EMPC的方法能保证动态最优性,提高经济收益。此外,金氰化浸出过程受随机噪声、未知参数可变等不确定性影响,提出使用扩展卡尔曼滤波(EKF),通过构造增广系统对状态变量及不确定参数进行在线同步估计,加强EMPC的准确性和可靠性。仿真结果表明,提出的EMPC+EKF策略能有效提高金氰化浸出过程的经济性能。     

Dynamic behaviour of electrochemical reactors for a step change in inlet concentration under potentiostatic control according to the dispersion model

The dynamic modelling of electrochemical reactors under potentiostatic control is performed according to the dispersion model. The response of the system in concentration and current is discussed for a step change in the inlet concentration in terms of the Peclet number. The stirred tank and plug flow behaviours are analyzed as limiting cases of the dispersion model.