应用混合种群遗传神经网络的精梳毛纺工艺参数反演模型

针对传统精梳毛纺工艺参数反演模型收敛性和稳定性不理想、反演精度低等问题，以及标准遗传算法（ SGA）应用于复杂优化问题时存在早熟收敛等缺点，以BP神经网络为基础，提出一种混合种群遗传人工神经网络（ MPG?ANN）反演模型，首先以混合种群遗传算法优化BP神经网络的权值与阈值建立预测模型，在此基础上根据毛纱CV值建立混合种群遗传算法反演模型，用来反演精梳毛纺生产过程工艺参数。以纺纱车间大量现场工艺检测数据为对象进行反演验证，结果表明：MPG?ANN 模型反演精度达97％，相比于标准遗传算法人工神经网络（ SGA?ANN）模型提高4％，同时反演结果波动幅度相比于SGA?ANN模型降低了6?28％。该方法可为精梳毛纺生产过程质量控制提供有效的理论指导，对纺织企业新产品工艺开发设计的快速决策具有很好的借鉴作用。     

基于ANN和LSSVR的造纸废水处理过程软测量建模

针对造纸废水处理系统的时变性、非线性和复杂性等特点,将人工神经网络(ANN)和最小二乘支持向量回归(LSSVR)分别用于造纸废水处理过程中的软测量建模,实现造纸废水处理过程中出水化学需氧量和出水悬浮固形物浓度的预测。ANN采用误差反向传播算法建模,LSSVR通过粒子群优化算法进行模型参数优化。结果表明,与ANN模型预测结果相比,LSSVR模型预测结果的均方根误差降低了50%以上,相关系数提高了近10%,表明LSSVR模型在造纸废水处理过程中的预测精度高于ANN模型。     

基于CBR和ANN的精纺纱智能化加工研究与实践

基于事例的推理(CBR)和人工神经网络(ANN)技术,建立了精毛纺纱线质量智能预报加工模型.通过CBR,根据所加工产品的主要特征,能够从历史数据库中快速提取出与设计特征参数最为相似的案例,相似性高的案例排在检索案例前面.利用ANN预报模型对所调整方案的效果进行评价,一方面可以通过调整相应的加工工艺参数及产品质量指标,预测所加工纱线的质量指标,并与实际要求比较,确定最终的加工参数;另一方面,根据所要加工纱线的质量要求,反演所需毛条的相关质量指标,以便企业能够用较低的原料成本生产同样质量的纱线.     

基于高光谱成像的咸鸭蛋腌制品质快速检测

咸鸭蛋是一种中国传统腌制食品，水分和脂肪含量是评价咸鸭蛋腌制品质的重要指标。本文旨在采用高光谱成像系统结合化学计量学探究不同腌制阶段咸鸭蛋的水分和脂质含量变化及分布状态。实验采用高光谱成像技术获取432～961 nm波长范围内的咸鸭蛋剖面反射光谱。为了减小光谱信号的噪声，通过Savitzky-Golay平滑(SG)、高斯滤波(Gauss)和标准正态变换(SNV)三种方法对原始光谱进行预处理，使用竞争自适应加权采样算法(CARS)挑选最优波长，然后进一步采用偏最小二乘回归算法(PLSR)和人工神经网络(ANN)对水分和脂质含量进行定量预测。结果表明，ANN模型能够更好地对水分及脂质含量进行预测，对蛋白水分、蛋黄水分和蛋黄脂质的预测决定系数分别为0.9306,0.9552和0.8896。最后，为了更加直观、全面地对咸鸭蛋腌制过程中的水分和脂质含量进行评价，根据ANN模型绘制出咸鸭蛋剖面的水分及脂质含量分布图。可视化分布图成功显示了咸鸭蛋在不同腌制时间水分和脂质的空间分布。本研究为咸鸭蛋生产及下游食品加工企业提供了一种基于光谱技术的快速检测方法。     

纺织品市场销售额的综合预测模式

本文讨论了一种关于纺织品市场销售额预测的综合模式，其基本思想 是用多种模型和电种算法同时处理预测问题，并寻求模型怀算的最佳匹配。     

人工神经网络在新鲜奶酪中的应用现状及发展前景

综述新鲜奶酪中马苏里拉奶酪和哈罗米奶酪的研究进展以及人工神经网络（artificial neural network，ANN）在新鲜奶酪领域中的应用现状，并分析当前应用ANN 方法的优势及存在的问题。除此之外，ANN 在新鲜奶酪领域中的未来发展前景是基于深度算法，实现对生产过程的多目标预测和多工艺参数优化，为提高新鲜奶酪的工艺优化效率和优化质量提供理论依据。     

预测造纸废水出水指标的随机森林建模方法

出水化学需氧量(COD)与出水固形物含量(SS)是评价造纸废水处理工艺好坏的重要指标。为了更好地对其进行预测,提出了一种基于随机森林(RF)模型的方法,并以R语言为工具进行回归预测。对比偏最小二乘(PLS)模型、支持向量回归(SVR)模型、人工神经网络(ANN)模型等常规预测模型,发现RF模型具有预测精度高,结果误差小,泛化能力好,调整参数少等优点。在对出水COD进行预测时,RF模型的相关系数r为0.7954,相比于PLS、SVR、ANN分别提高了8.88%、10.73%、14.68%。在对出水SS进行预测时,RF模型的相关系数r为0.8551,相比于PLS、SVR、ANN分别提高了15.43%、24.25%、30.79%。     

弹性纤维的结构、弹性及其纺织品染整(一)

较系统地介绍了各类弹性纤维的结构和类别，分析了弹性机理和特征，重点探讨了氨纶、PTT弹性纤维纺织品的染整要点，包括纺织品的存放、松弛处理、精练、热定形、染色加工要点和注意事项。最后，对PTT／棉纤维纺织品的分散／活性染料一浴法染色作了深入的探讨。     

弹性纤维的结构、弹性及其纺织品染整(二)

较系统地介绍了各类弹性纤维的结构和类别，分析了弹性机理和特征，重点探讨了氨纶、PTT弹性纤维纺织品的染整要点，包括纺织品的存放、松弛处理、精练、热定形、染色加工要点和注意事项。最后，对PTT／棉纤维纺织品的分散／活性染料一浴法染色作了深入的探讨。     

基于PCA降维模型的造纸废水处理过程软测量建模

针对造纸废水处理过程的复杂特性，本课题将主成分分析（PCA）与人工神经网络（ANN）和支持向量回归（SVR）相结合，构建出两种新的软测量模型：主成分分析-人工神经网络（PCA-ANN）和主成分分析-支持向量回归（PCA-SVR）。本课题将这两种软测量模型应用于造纸废水处理过程中出水化学需氧量（COD）和出水悬浮固形物（SS）浓度的预测。计算结果表明，PCA-ANN和PCA-SVR的预测效果均优于偏最小二乘、支持向量回归和人工神经网络3种常规软测量模型，并且PCA-ANN的预测效果最优。对于出水COD浓度预测，PCA-ANN的决定系数（R2）为0.984，均方误差（MSE）为1.892，较ANN分别优化了9.7%和71.5%。对于出水SS浓度预测，PCA-ANN的R2为0.762， MSE为0.228，较ANN分别优化了31.2%和58.7%。     

Woven fabric design and analysis in 3D virtual reality. Part 1: computer aided design and modeling of interlaced structures

A computer model is developed that enables geometric models for woven fabric structures in three dimensions, which would allow engineers to simulate new textile forms in virtual reality and to study the effects of the process parameters on the textile structure. Mathematical models and algorithms for the generation and manipulation of 3D single and multi-layer fabrics are developed.     

Application of data mining technique in predicting worsted spun yarn quality

Today’s industry gives first priority to information technology. Since understanding the structures and relationships dominated of data can help industrial managers to attend in competitive market successfully, a special mechanism must be developed to process data stored in a system. Hence, the focus on widespread use of data mining gains increasing attention. The purpose of this paper is using data-mining technique in textile industry. More than 150,000 data includes testing of raw materials, manufacturing process parameters and yarn quality parameters, during one year in worsted spinning factory were collected. Next, yarn quality was predicted by using data-mining methods containing clustering and artificial neural network (ANN). In order to evaluate the proposed method, the results obtained were compared with conventional methods based on ANN. The results showed that the performance of data-mining technique is more accurate than that of ANN.     

Viscoelastic Behaviour and Microstructural Modifications in Acrylic Fibres and Yams as a Function of Textile Manufacturing Processing Conditions

Tensile tests on acrylic yarns and fibres extracted from various steps of textile processing reveal mechanical and thermal modifications due to the manufacturing operations. To examine this, viscoelastic modelling based on mechanistic models for fitting experimental data of tensile tests were investigated. Crystalline/amorphous distribution was assessed by the differential solubility technique, and the models account for changes in property values due to microstructure modifications.     

Virtual characterization of MDF fiber network

A virtual design method for medium density fiberboards (MDF) is proposed with the aim to optimize the fiber orientation and lay-up of MDF. The new method estimates the stiffness and strength by using microstructure models of the MDF fiber network. The virtual design is used to improve the manufacturing technology of MDF plates with multilayer oriented fiber structure. Experimental investigations of the mechanical behavior of MDF microstructure for various fiber geometries, glue content and distribution are complicated, time consuming and expensive. On the other side, virtual microstructure design allows to develop a new wood fiber based material with less experimental work. Microstructure models help to better understand the non-linear damage mechanical behavior of a wood fiber network depending on fiber geometrical parameters. Such parameters as crack distribution and fiber deformation on micro-scale level are complicated to experimentally measure, but possible to model using computer simulations. The virtual design tool requires less empirical data. The model takes into account information on average wood fiber orientation, fiber diameter, fiber length and mechanical properties of wood fiber cell wall and glue. The numerical method for strength and stiffness analysis of MDF microstructure was calibrated using standard MDF with non-oriented fibers. It turned out that this method gives precise results for MDF with oriented fibers and even with multilayer structure. The proposed virtual microstructure design tool can significantly improve and speed-up the optimization manufacturing technology of MDF and other wood fiber based composites.     

我国纺织业发展趋势分析

本文通过对我国纺织产业的种植、产品质量等宏观环境进行分析,并结合纺织加工、流通及其产业链运行的情况,从规模化种植、纺织纤维多元化和面料功能化、加大服装企业创新能力三个方面进行剖析,达到提高纺织产业竞争力、改善产业运行效率,为纺织业的发展提供依据。     

MODELING AND OPTIMIZATION OF CONSTANT RETORT TEMPERATURE (CRT) THERMAL PROCESSING USING COUPLED NEURAL NETWORKS AND GENETIC ALGORITHMS

Coupled artificial neural network (ANN) models and genetic algorithms (GA) were applied for developing prediction models and for optimization of constant temperature retort (CRT) thermal processing of conduction heating foods. ANN prediction models were developed for process time (Pt), average quality retention (Qv), surface cook value (Fs), equivalent energy consumption (En), final temperature difference (T_g) at can center, and lethality ratio (p, heating/total lethality). The processing conditions as inputs for ANN models were as follows: retort temperature (RT = 110–140C), thermal diffusivity (α= 1.1–2.14*10^?7m²/s), volume of can (V = 1.64–6.55*10^?4m³), ratio of height to diameter of can (R_dh= 0.2–1.8), total desired lethality value (F₀= 5–10 min) at can center and quality kinetic destruction parameters: decimal destruction time (D_q– 150–300 min) and their temperature dependence (z_q= 15–40C). The data for training and testing ANN models were obtained from a finite difference computer simulation program. A second order central composite design was used for constructing the experimental data for training ANN models, while an orthogonal experimental design with 6 factors and 3 levels was used for the generalization of trained ANN models. ANN model linked Genetic Algorithms (GA) were employed for searching for the optimal quality retention and corresponding retort temperature, and for investigating the effects of main processing factors. ANN‐based prediction models successfully described the various outputs of CRT thermal processing (correlation coefficients: R² > 0.98; relative errors: Er ≤ 3%). The coupled ANN‐GA models, verified under several typical processing conditions, could be effectively used for optimization of CRT thermal processing. The main processing conditions and their interactions in the order of their importance with respect to the optimal quality retention and corresponding retort temperature were: V >z_q >R_dh >; and z_q >F_d > R_dh >V, respectively.     

基于蜂窝纸芯模块的西瓜包装设计

为了实现运输包装及产品包装一体化,研究了西瓜的包装结构与造型设计,探索蜂窝纸芯模块应用于相关包装产品领域的数字化设计与绿色制造方法。以虚拟设计和虚拟制造的方法进行新型包装结构产品开发,通过最终产品形态逆向反求蜂窝纸芯模块数据和加工尺寸,以数控CNC加工中心完成蜂窝模块的加工制作。结果表明,此法可实现包装的个性化定制,有效提升产品的品牌附加值。     

小轴类零件CAD/CAPP/CAM信息集成系统模型

通过对计算机辅助设计CAD、计算机辅助工艺设计CAPP、计算机辅助制造CAM系统间信息共享和集成存在问题的分析,建立了以产品数据管理PDM为数据平台的CAD/CAPP/CAM集成系统的体系框架结构,使得零件产品设计信息、工艺信息和加工信息等与零件加工制造过程相关的信息以统一的数据格式存储到PDM系统中,实现了产品数据信息安全、有效的管理。根据轴类零件加工特点的分析,建立了零件的几何特征模型、工艺信息模型和加工仿真模型。通过对小轴类零件的实例分析,利用VB高级程序语言集成接口进行设计,自动生成零件的工艺工序图,达到了设计、管理、制造过程中信息交换和共享的目的,从而大大提高了管理和生产效率。     

Simulating the drawing of spunbonding nonwoven process using an artificial neural network technique

The statistical and artificial neural network (ANN) models are established for predicting the fiber diameter of spunbonded nonwovens from the processing parameters. The ANN of Bayesian frameworks produces smaller prediction errors and thus is determined to be the preferred network. Results show that the ANN model yields more accurate and stable prediction than the statistical model, and a reasonably good ANN model can be established with relatively few data points. Four methods are used to reveal the relative importance of the processing parameters in terms of their effect on the fiber diameter. It is found that the initial polymer temperature plays an most important role in reducing the fiber diameter, while the effect of the initial air temperature is not significant. Using an established ANN model, computer simulations of the effects of the processing parameter on the fiber diameter are carried out. It is found that higher polymer melt index, smaller polymer flow rate, higher initial polymer temperature, higher initial air temperature, and higher initial air velocity can all produce finer fibers. This area of research has great potential in the field of computer-assisted design in spunbonding technology.