基于RBF神经网络的挤出吹塑中型坯尺寸的预测

在利用BP神经网络预测挤出吹塑中型坯尺寸工作的基础上,采用径向基神经网络(RBF)来预测挤出吹塑中型坯尺寸,并与BP神经网络的预测结果进行了比较。结果表明,虽然RBF与BP神经网络均能较好地预测挤出吹塑中型坯尺寸,RBF网络的训练时间比BP少很多,只是BP的0.7%。     

挤出吹塑中型坯成型的有限元模拟——型坯尺寸预测

建立了塑料挤出吹塑中平直型机头内非牛顿粘弹性熔体的流动模型，并利用POLYFLOW有限元软件进行求解，预测了不同型坯长度和不同流量时的型坯尺寸分布。     

尤里卡沥青软化点人工智能软测量技术探索

人工神经元网络(ANN)应用于软测量是人工智能方法在石化过程中成功应用的热点,ANN不需要系统模型就能映射复杂非线性关系,特别适合炼油生产过程的建模与预测工作,研究了用ANN对尤里卡沥青软化点进行软测量的方法,和现用的吴羽公司kθ法相比,ANN方法测量精度高,具有学习能力和联想记忆能力,健壮性好,它与生产装置的DCS硬件相结合能够达到优化生产控制的目的。     

人工神经网络在材料开发中的应用研究进展

介绍了人工神经网络(ANN)的发展历程、模型特性与分类,以及反向传播(BP)神经网络模型及其改进算法,重点论述了ANN在高分子聚合反应过程和质量控制、成型加工工艺设计与条件优化、材料使用与服役性能预测方面的应用进展,以及在辅助性能表征与分析等方面的应用研究状况,并指出了ANN在未来新材料开发中应用的发展方向和亟待解决的问题。     

重质油黏度的定量结构-性质关系研究

研究了重质油黏度的定量结构-性质关系。将量子化学参数和拓扑指数相结合作为结构描述符,分别用多元线性回归(MLR)和人工神经网络(ANN)建立了结构描述符和黏度之间的校正模型。用留一交叉验证法,验证、评价所建立的MLR和ANN模型的预测能力。对于MLR模型,验证的均方根相对误差为7.77,对于ANN模型,验证的均方根相对误差为7.21,说明建立的MLR和ANN模型都可用于预测重质油的黏度,但ANN模型优于MLR模型。     

基于电性拓扑状态指数的烷烃密度定量结构-性质关系研究

研究了烷烃密度的定量结构-性质关系。以电性拓扑状态指数为结构描述符,分别用多元线性回归(MLR)和人工神经网络(ANN)建立了结构描述符和密度之间的校正模型。用留一交叉验证和外部测试集验证评价所建立MLR和ANN模型的预测能力。对于MLR模型,这两种验证的均方根相对误差分别为3.37和1.92。对于ANN模型,这两种验证的均方根相对误差为1.06和1.34。这说明建立的MLR和ANN模型都可用于预测烷烃的密度,但ANN模型优于MLR模型。     

人工神经网络在注塑制品质量优化中的应用进展

大部分复杂的塑料制品均采用注射成型生产。注塑制品质量的预测和制品质量的优化是注射成型过程中的重要步骤。人工神经网络(ANN)作为人工智能最常用的方法已经被应用到注射成型中,但是,仍存在训练成本较高、构建模型复杂等缺陷。ANN预测模型可以拟合注塑过程,并且,优化注塑制品质量。以工艺参数和过程参数作为输入数据的ANN预测模型不仅可以预测注塑制品质量,还可以结合智能优化算法优化注塑工艺参数。并且,对减少ANN预测模型训练成本的方法进行了综述。最后,总结了ANN预测模型在注塑制品优化中的进展和发展方向。     

基于LM算法的棒束通道内空泡份额预测模型研究

应用Anklam T M和Byong-Jo Y的实验数据成功构建了基于LM算法优化的人工神经网络(ANN),用训练成功的ANN对棒束通道内的空泡份额进行预测,并得出了新的空泡份额预测关系式,其预测的均方根误差为7.80%。将ANN的预测结果与Cunningham J P and Yeh H C模型、Kamei A模型、Paranjape S模型的预测结果进行对比,结果表明:ANN的预测结果优于Cunningham J P and Yeh H C模型、Paranjape S模型的预测结果,与Kamei A模型的预测结果相近。通过输入变量对输出变量影响的敏感性分析,发现测点轴向距离与当量直径之比Z/DH、质量流密度G、加热棒束的热流密度q对棒束内空泡份额有很大的影响。     

嵌入岭回归的BP算法及其在软测量中的应用

针对三层神经网络(ANN)最佳隐节点个数难以确定和随着隐节点个数增加ANN模型易出现过拟合等缺点,提出了嵌入岭回归(RR)的误差反传算法(BP).BP-RR根据样本规模自适应确定隐节点个数,并通过BP算法充分提取样本数据信息.然后,针对隐含层输出可能存在的复共线性,采用RR以预测性能为指标,通过进化算法确定最佳岭参数,进而重新确定隐含层与输出层之间最佳的权值和阈值,克服ANN过拟合,建立具有良好预测性能的模型.将BP-RR应用于建立石脑油干点软测量,结果显示,BP-RR模型具有良好的预测性能.与ANN相比,BP-RR模型鲁棒性强,预测精度高.     

基因表达式编程在含氧有机化合物定量结构-保留关系中的应用

基因表达式编程(GEP)是一种新颖的遗传算法,是一种高度有效、稳定的随机搜索方法。采用GEP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,并与人工神经网络(ANN)预测结果进行比较。GEP和ANN在OV-1固定相上,相关系数R分别为0.9908,0.9892;在SE-54上,相关系数R分别为0.9956,0.9891。结果表明:GEP建立的模型优于ANN,具有良好的拟合度和预测精度。     

New Strategies for Predicting Parison Dimensions in Extrusion Blow Molding

In this work, two new strategies were proposed for predicting the parison thickness and diameter distributions in extrusion blow molding. The first one was a finite-element-based numerical simulation for the parison extruded from a varying die gap. The comparison of simulated and experimental parison thickness distributions indicates that the new method has certain accuracy in predicting the parison thickness from a varying die gap. The second one was an artificial neural network (ANN) approach, the characteristics of which are in sufficient patterns that can be obtained without doing too many experiments. The diameter and thickness swells of the parisons extruded under different flow rates were obtained by a well-designed experiment. The obtained data were then used to train and test the ANN model. The dimension of one location on the parison can provide one pattern to train the ANN model. Trained and tested ANN model can be used to predict the dimensions at any location on the parison within a given range. The proposed two strategies can help search the processing conditions to obtain optimal parison thickness distributions.     

A modeling approach to the effect of resin characteristics on parison formation in extrusion blow molding

The most critical stage in the extrusion blow‐molding process is the parison formation, as the dimensions of the blow‐molded part are directly related to the parison dimensions. The swelling due to stress relaxation and sagging due to gravity are strongly influenced by the resin characteristics, die geometry, and operating conditions. These factors significantly affect the parison dimensions. This could lead to a considerable amount of time and cost through trial and error experiments to get the desired parison dimensions based upon variations in the resin characteristics, die geometry, and operating conditions. The availability of a modeling technique ensures a more accurate prediction of the entire blow‐molding process, as the proper prediction of the parison formation is the input for the remaining process phases. This study considers both the simulated and the experimental effects of various high‐density polyethylene resin grades on parison dimensions. The resins were tested using three different sets of die geometries and operating conditions. The target parison length was achieved by adjusting the extrusion time for a preset die gap opening. The finite element software BlowParison® was used to predict the parison formation, taking into account the swell and sag. Good agreements were found between the predicted parison dimensions and the experimental data. POLYM. ENG. SCI., 2009. Published by Society of Plastics Engineers     

模拟技术在塑料中空成型中的研究进展

介绍了模拟技术在吹塑型坯成型中的研究和发展状况，对国内外学者针对型坯成型过程运用的数值分析方法和理论依据进行了论述，并着重介绍了神经网络方法。指出型坯成型过程是吹塑过程的核心，受聚合物材料性能、熔融温度、成型加工条件等因素的综合影响，有待进一步研究。     

A comprehensive experimental study and numerical modeling of parison formation in extrusion blow molding

Parison dimensions in extrusion blow molding are affected by two phenomena, swell due to stress relaxation and sag drawdown due to gravity. It is well established that the parison swell and sag are strongly dependent on the die geometry and the operating conditions. The availability of a modeling technique ensures a more accurate prediction of the entire blow molding process, as the proper prediction of the parison formation is the input for the remaining process phases. This study considers both the simulated and the experimental effects of the die geometry, the operating conditions, and the resin properties on the parison dimensions using high density polyethylene. Parison programming with a moving mandrel and the flow rate evolution in intermittent extrusion are also considered. The parison dimensions are measured experimentally by using the pinch-off mold technique on two industrial scale machines. The finite element software BlowParison® developed at IMI is used to predict the parison formation, taking into account the swell, sag, and nonisothermal effects. The comparison between the predicted parison/part dimensions and the corresponding experimental data demonstrates the efficiency of numerical tools in the prediction of the final part thickness and weight distributions. POLYM. ENG. SCI., 47:1–13, 2007. © 2006 Society of Plastics Engineers     

塑料挤出吹塑的机理问题

采用不同的方法对挤出吹塑过程的型坯成型、型坯吹胀与制品冷却三阶段的机理问题进行了研究.采用人工神经网络方法预测了受模口温度和挤出流率影响的型坯成型阶段的膨胀.利用建立起来的神经网络模型预示的膨胀与实验结果很吻合,且可在一定范围内,预示不同工艺条件下型坯的直径膨胀和壁厚膨胀,为型坯的直径和壁厚的在线控制提供了理论依据.基于薄膜近似和neo-Hookean本构关系,建立了描述型坯自由吹胀的数学模型,并通过实验方法获得了型坯吹胀的瞬态图象.     

Parison Inflation in Extrusion Blow Molding: A Theoretical Analysis for Identifying Critical Process Parameters

Abstract

An analysis for describing parison (cylindrical) inflation behavior in the extrusion blow molding process is presented. A general growth equation is developed starting from the basic conservation principles. Assuming the polymer melt constituting the parison to behave as a purely viscous Generalized Newtonian Fluid, the effect of different process and material parameters on the inflation process is investigated. From the numerical results, it is inferred that the growth behavior for inelastic liquid exhibits a general tendency of approaching exponential (constant stretch rate) growth as elapsed time progresses. Besides, the initial parison dimensions are determined to play a very significant role in governing the inflation process. Moreover, the inertial contribution owing to fluid motion is found to exert an appreciable influence on the growth dynamics, and hence cannot be neglected without introducing severe approximations in the analytical development.     

Variables affecting parison diameter swell and their correlation with rheological parameters

An important factor in the selection of blow molding resins for producing handled bottles is the effective diameter swell of the parison. Ideally, the diameter swell is directly related to the weight swell and would require no separate consideration. In actual practice, the existence of gravity, the finite parison drop time and the anisotropic aspects of the blow molding operation prevent reliable prediction of the parison diameter swell directly from the weight swell. The parison diameter swell is a complex function of the weight swell, the rate of swell and the melt strength. Elements of this function are presented which show the effect of extrusion rate, parison drop time and parison weight. A technique is presented which allows the estimation of local weight and diameter swell ratios. Their direct relationship is confirmed by data obtained on several blow molding resins. The relationship between weight swell and diameter swell is definitely anisotropic. A mathematical model for swell is proposed which incorporates experimentally determined rate constants and swell coefficients. Correlations are given which suggest fundamental relationships between these derived coefficients and basic variables such as resin properties or process conditions. The model's predictive capability is demonstrated by using it to back calculate parison dimensions.     

Modeling parison formation in extrusion blow molding by neural networks

A series of experiments were carried out on the parison formation stage in extrusion blow molding of high‐density polyethylene (HDPE) under different die temperature, extrusion flow rate, and parison length. The drop time of parison when it reached a given length and its swells, including the diameter, thickness, and area swells, were determined by analyzing its video images. Two back‐propagation (BP) artificial neural network models, one for predicting the length evolution of parison with its drop time, the other predicting the swells along the parison, were constructed based on the experimental data. Some modifications to the original BP algorithm were carried out to speed it up. The comparison of the predicted parison swells using the trained BP network models with the experimentally determined ones showed quite a good agreement between the two. The sum of squared error for the predictions is within 0.001. The prediction of the parison diameter and thickness distributions can be made online at any parison length or any parison drop time within a given range using the trained models. The predicted parison swells were analyzed. © 2005 Wiley Periodicals, Inc. J Appl Polym Sci 96: 2230–2239, 2005     

Numerical simulation of suction blow molding process for producing curved ducts

During suction blow molding process, the extruded parison undergoes twisting deformation within the mold cavity, as the air drawing flow around the deforming parison exerts non‐uniform shear stresses on its surface. Such twisting deformation can compromise the specific radial and circumferential variations in parison thickness that are intentionally generated during extrusion. This research is devoted in developing a fluid–structure interaction model for predicting parison deformation during suction blow molding process, with a specific emphasis on the suction stage. A fluid flow model, based on Hele‐Shaw approximations, is formulated to simulate the air drag force exerted on the parison surface. The rheology of the material of the parison is assumed to obey the viscoelastic K‐BKZ model. As the suction process also involves the sliding of the parison within the mold cavity, a modified Coulomb's law of dry friction is used to simulate the frictional contact between parison and mold. The numerical results of this study allowed identifying a clear correlation between the twisting deformation undergone by the parison during the suction stage, also observed experimentally and the design parameters, namely, the air drawing speed, the geometry of the duct mold cavity, and the parison/mold eccentricity. POLYM. ENG. SCI., 59:418–434, 2019. © 2018 Her Majesty the Queen in Right of Canada