基于双阈值AdaBoost算法的4-CBA含量软测量建模

针对PX氧化过程中4-CBA含量无法在线测量的问题,提出了一种基于双阈值更新样本权重的AdaBoost算法,该算法以BP神经网络作为弱学习器,采用轮盘赌方法根据样本权重在训练样本集中选择部分样本训练弱学习器,采用上一轮弱学习器的训练相对误差绝对值来更新所有训练样本的权重,在此基础上,用双阈值对样本误差范围进行划分,然后用不同的权重因子与原来的样本权值相乘实现样本权值的二次更新。该过程降低了含有大误差的样本的权值,增加了较大误差的样本的权值,从而减小了在下一轮训练过程中选到异常样本的概率。分别采用5种不同的方法并用实测的工业数据建立了4-CBA含量软测量模型,仿真结果表明用提出的改进AdaBoost算法建立的4-CBA含量软测量模型,其预测误差小于其他方法建立的模型误差。     

基于双阈值AdaBoost算法的4-CBA含量软测量建模

针对PX氧化过程中4-CBA含量无法在线测量的问题,提出了一种基于双阈值更新样本权重的AdaBoost算法,该算法以BP神经网络作为弱学习器,采用轮盘赌方法根据样本权重在训练样本集中选择部分样本训练弱学习器,采用上一轮弱学习器的训练相对误差绝对值来更新所有训练样本的权重,在此基础上,用双阈值对样本误差范围进行划分,然后用不同的权重因子与原来的样本权值相乘实现样本权值的二次更新。该过程降低了含有大误差的样本的权值,增加了较大误差的样本的权值,从而减小了在下一轮训练过程中选到异常样本的概率。分别采用5种不同的方法并用实测的工业数据建立了4-CBA含量软测量模型,仿真结果表明用提出的改进AdaBoost算法建立的4-CBA含量软测量模型,其预测误差小于其他方法建立的模型误差。     

浆态反应器中苯氯化反应的宏观动力学

采用自行设计的三相鼓泡浆态实验反应器 ,在半连续操作条件下 ,通过改变催化剂粒度 ,消除了内扩散的影响 ;测定了不同催化剂浓度、通氯量以及苯回流量条件下的苯氯化反应的宏观动力学数据。建立了宏观动力学模型 ,并由实验数据估计出模型参数 ,模型计算值和实验值吻合较好     

德惠断陷沙河子组泥岩封盖能力综合评价

通过对德惠断陷沙河子组泥岩盖层的单井数据统计分析,得出沙河子组泥岩宏观发育特征;对德惠断陷沙河子组泥岩取样,进行岩心数据分析,得出沙河子组微观封闭的相关数据进行分析。对沙河子组泥岩盖层进行参数划分以及对各指标赋予不同的权值的方法,对德惠断陷沙河子组泥岩盖层进行评价。     

关于有偏估计提高测量可靠性的探讨

可靠性是评价化工过程测量数据优劣的重要指标。理论与实践中普遍采用的观测数据平均值或加权平均作为待测参数测量值方法的本质是最小二乘无偏估计,而无偏测量数据可靠性与方差一一对应,根据高斯-马尔科夫定理可知无偏测量方差有下界,所以无偏测量数据不一定具有高可靠性。提出采用有偏估计改善测量数据的可靠性,首先分析未知参数测量与参数估计过程之间的等价性;其次给出同时采用方差和偏差定量表示有偏测量数据可靠性的方法;最后研究偏参数对有偏测量数据可靠性的影响,并采用数值法求解了偏参数的最优值。仿真结果表明偏参数合理取值范围内有偏测量的可靠性总是优于无偏测量的可靠性。     

基于AHP和模糊数学方法的气井产能影响因素评价

导致苏里格气田气井低产低效的原因复杂,影响因素较多,如何科学有效地评价其影响程度成为该技术领域的难题.基于储层物性、生产动态和排采工艺现场数据,首次将AHP和模糊数学方法应用到气井产能影响因素评价中.通过建立问题的递阶层次结构、计算单一准则下元素的相对权值和层次总排序系数,得到了区块各因素对气井产能的影响权重值.基于现场大量生产数据,确定了评价参数和隶属度的取值.首次提出了一套低产低效气井产能影响因素评价模型.结合案例分析认为,影响气井产能的主要因素为储层物性,气井排采效果和产量递减规律对产量影响效果较小.该方法为判断气井低产低效的原因提供了理论依据.     

化学还原法处理含铬废水

铬是人体必需的一种元素,但是六价铬是环境及水质中的一种主要污染物。本实验用化学还原法处理含铬废水,研究了以硫酸亚铁为还原剂,改变p H值和还原剂投料量,对废水中六价铬去除率的影响。在硫酸亚铁理论投料量下,改变待测废水pH值;在最佳pH值下,改变还原剂硫酸亚铁的投料量。实验数据对比发现,pH值为中性时,硫酸亚铁的投料量为理论投料量的1. 12倍时,对废水中六价铬的去除率最高。     

熵权法在重大危险源应急救援评估中的应用

针对重大危险源发生事故时的特点和影响应急救援能力的因素,建立应急救援能力评价指标体系和层次结构模型,采用层次分析法构建不同层次及各类因素中全部指标的判断矩阵,通过矩阵运算和一致性检验,得到各级指标的权重,并用熵权法对得出的权重进行修正。将2种方法所得出的结果进行对比,结果表明:在熵权法修正指标权重之后,火灾监测仪表的配备、维护及定期检测情况,预警设备的灵敏度和准确度,重大危险源泄漏报警装置的灵敏度和准确度,监测信息识别、诊断、评价4个指标的权值大幅度增大。同时,救援组织能力这个指标的权值在修订以后则大幅度的减小。相比之下熵权法修订后的结果在一定程度上增强了评估结果的客观性。     

填充剂变化对橡胶配方的影响

在橡胶技术中采用毛细管式流变仪已经超过45年。毛细管式流变仪是研究提高剪切速率对胶料粘度影响的实用方法。在静态剪切试验中,当用毛细管式流变仪测定时,通过提高一定的剪切速率可降低胶料的粘度。众所周知,不同原料聚合物能赋予不同程度的剪切稀化特性。本研究着重研究不同类型的橡胶填充剂、不同填充剂表面积(与粒度成反比关系)和不同填充剂填充量对剪切稀化曲线的影响。     

基于O_3/UV法在线COD检测的气体溶解量估计方法

针对臭氧协同紫外方法 (O_3/UV)检测化学需氧量(COD)时存在溶解性气体影响测量精度的问题,提出了一种COD检测过程中气体溶解量的估计方法,用于对COD检测模型的补偿。采集不同浓度的COD标准水样在消解过程中的测量数据和实验分析数据,基于PLS-LSSVMs建立溶解氧量和溶解二氧化碳量的估计模型,将模型的输出作为COD检测模型的补偿项。实验结果表明,基于PLS-LSSVMs建立的模型比PLS或者LSSVMs单独建立的模型估计精度高。采用溶解气体量估计模型进行补偿后的O_3/UV法检测COD与国标法测量结果相对误差均小于5%。对提高O_3/UV法检测COD精度具有重要意义。     

Fractionation of a polydisperse polymer using an antisolvent. Application of continuous thermodynamics

Continuous thermodynamics is used to calculate liquid-liquid equilibria in a ternary system containing polydisperse polymer, one good solvent and one poor solvent (antisolvent). The polymer has a semiinfinite molecular-weight distribution which can be described by a gamma distribution. From Flory-Huggins theory, the logarithm of the distribution ratio of the polymer is a linear function of the molecular weight. Because the distribution of polymer between the two liquid phases is very sensitive to the choice of binary Flory parameters. The Flory parameters must be fitted to the ternary data. Results are given for the system benzene, ethanol, and polydisperse polystyrene.     

喷雾场多参数层析算法的仿真效验与实验测量

根据层析技术原理,对喷雾雾滴粒径和浓度的空间分布开展了研究,并编制了喷雾粒径和浓度层析计算程序.对给定粒径和浓度空间分布的喷雾场,根据脉动法测量原理在计算机上仿真了雾滴粒径和浓度的线投影值,根据仿真的投影值重建了雾滴粒径和浓度的空间分布,重建结果与给定的分布结果完全吻合.在喷雾实验台上用光脉动法装置测量了喷雾雾滴粒径和浓度的线投影值,并用编制的程序对喷雾雾滴粒径和浓度的空间分布进行了层析计算.     

The influence of molecular weight on the melt rheology of polypropylene

Melt viscosity and melt elasticity data were obtained over a broad range of temperatures and shear rates on a series of four polypropylenes of different molecular weight but approximately the same molecular weight distribution. The superposition technique was used with both temperature and molecular weight to shift flow curves for all four materials at three temperatures each along the shear rate axis to generate a master flow curve at a given temperature and molecular weight. For polypropylenes of this type, and molecular weight distribution shift, factors which can be used to extend the useful range of experimentally obtained flow data were determined. The dependency of apparent viscosity on weight average molecular weight at shear stresses as high as 10⁶ dynes/cm² is shown. The dependency of melt elasticity on molecular weight and temperature is discussed.     

Construction and testing of a specific viscosity detector for size exclusion chromatography

This article describes our experiences incorporating a viscosity detector directly within a conventional differential refractive index (DRI) detector. This results in a system in which the collection of the necessary data for universal calibration in size exclusion chromatography (SEC) can be readily attained. The system described is constructed from available materials, and yields output which can be used directly in the calculation of the intrinsic viscosity of the eluant from a chromatographic column. Values for various polymer molecular weight distribution parameters, as well as reasonable estimates of the Mark—Houwink constants, can be obtained. Additionally, a measurement related to long-chain branching in the polymer is available using this method.     

测量仪表过失误差的单结点识别方法

提出了一种新的过失误差识别方法 ,即结合仪表的可靠度、精度等级等信息 ,利用单结点的约束残差进行过失误差识别 .给出仿真实例 ,通过与其他几种常用的误差识别方法对比对这一方法进行了评价 .     

Gel permeation chromatography: Automatic data acquisition and reduction in real-time with a process computer

A real-time computer system is described which minimizes the total analysis time in gel permeation chromatography by automatically acquiring the data, performing the necessary calculations and displaying the molecular weight distribution and molecular weight averages as soon as the polymer peak has eluted. The calculation method described here is valid for linear homopolymers only, and does not include corrections for instrumental peak broadening. The experimental results show good reproducibility and precision. On-line changes of the constants and other parameters may be implemented by means of an auxiliary program which runs in parallel with the data analysis program. A system such as that described here is suitable for utilization as a measurement technique in process control applications. The programs are written in CORAL 66 for a Ferranti Argus 700 E computer.     

Analysis of FBRM measurements by means of a 3D optical model

Continuous process monitoring is desirable for many particulate processes such as the crystallization of active pharmaceutical ingredients. Only an in-line measurement technique can achieve such a continuous monitoring.A popular in-line measurement technique, which can be applied without dilution, is the focussed beam reflectance measurement (FBRM). However, FBRM is at this point mainly used for qualitative measurements. The measured chord length distribution is different from a particle size distribution. For a quantitative measurement a sound understanding of the measurement principle is necessary. In this paper, an optical model of the FBRM probe and a three-dimensional simulation of the measurement are presented.A three-dimensional particle field is generated with a Monte-Carlo approach. The back scattered light intensity is calculated as a function of the position of the laser beam with respect to this particle field. A vector of scattering intensities is obtained for a given laser path. This vector is processed with the simulated electronics of the Lasentec FBRM system. The output of this processing step is a chord length distribution which can be compared to the output of the Lasentec FBRM system.Simulation studies with mono-disperse polystyrene particles of different sizes and concentrations are conducted and compared to measurements of a Lasentec D600L FBRM probe. With the presented model yet unexplained massive over-estimation of small particles and concentration-dependent changes in the chord length distribution can be described.     

A relationship between steady-state shear melt viscosity and molecular weight distribution in polystyrene

A model that relates to the molecular weight distribution (MWD) of high-density polyethylene to the steady-state shear melt viscosity has been applied to polystyrene melts. Relations are developed for predicting the rheological flow curve from the molecular weight distribution. Relationships are also developed to predict the MWD from the flow curve, although practical limitations to this procedure are given. From a consideration of predictions of the model and experimental data, it is concluded that the transition for a given molecular species from Newtonian to non-Newtonian flow is sharp. Additionally, the calculated empirical parameter that partitions the MWD into molecules that act in a Newtonian fashion and those that do not is shown to be equivalent to the largest molecular weight homolog that can still undergo Newtonian flow at a given shear rate for monodisperse fractions. The temperature dependence of the relaxation times is found to be somewhat higher than that predicted by the Rouse theory. An activation energy of 30 kcal/mole for η0 was used to fit the experimental viscosity data adequately at 190° and 225°C. The terminal relaxation spectrum for a narrow-MWD polystyrene standard is calculated and found to agree well for long relaxation times with that reported in the literature.     

Determination of the molecular weight distribution of cellulose on calibrated gel columns

CMC samples obtained by gel filtration on a preparative scale were used to calibrate an agarose column. The relation between molecular weight and the peak elution volume for monodisperse samples were calculated as well as the relation between band broadening and the peak elution volume. In the calculations of the molecular weight distribution curves of the studied CMC samples, consideration was given to the variable band broadening by proper transformations of the experimental data, thus making it possible to obtain numerical solutions free from oscillations. The method admits a rapid determination of the molecular weight distribution of cellulose and related materials.     

Measurement of the droplet size distribution of fog using multi-wavelength lidar: A feasibility study

The extinction and backscatter coefficients of an aerosol are determined in part by the aerosol size distribution. If specified at more than one wavelength these optical data can be used to infer the aerosol size distribution if the relationship between the optical data and the size distribution is well defined. In the case of fog the index of refraction is known and the fog droplets are spherical so that the relationship between the optical data and the droplet distribution is given by Mie theory. The inversion procedure of Twomey (1977) is adapted to derive the size distribution of an example fog using the extinction and backscatter coefficients at wavelengths between 0.53 μm and 10.6 μm. Such measurements could be made using a multi-wavelength lidar. The accuracy of the solution as a function of intrinsic inversion error and measurement error is discussed.