【作者】 侯彩虹；

【导师】 崔运花；

【作者基本信息】 东华大学 ， 纺织工程， 2004， 硕士

【摘要】 为了实现织物的智能感官评估，本论文采用了模糊技术、神经网络等方法对来自于不同专家感官评估的感官数据以及来自于仪器测量的客观数据进行了分析和处理，并且建立了客观评估与感官评估之间的联系，在此基础上，开发了基于Internet的纺织品智能感官评估系统。 在感官数据处理中，采用了新的以语言两元组模型为基础的方法对织物风格评估中不同专家小组给出的感官数据标准化。在客观数据处理中，为了避免高维的问题，用主因子分析和主成分分析方法从量化的数据和人类知识中选择特征参数并提取模糊规则，依此描述客观数据中的特征参数与感官数据之间的联系。在纺织品智能感官评估系统开发中使用了Java语言和SQL Server数据管理系统语言。这样，如果给出一块新的织物样品的客观数据，系统就可以给出它相应的感官数据和综合等级；如果同时给出织物的感官数据，就可以直接从感官数据得出织物的综合手感等级。 第一章介绍了纺织品感官评估的背景知识及相关知识，并简单介绍了本论文的研究内容。 第二章从理论基础和实际需求出发，对纺织品智能感官评估系统进行了总体设计，并简单地介绍了系统各组成模块的功能以及模块之间的关系。 第三章对感官数据进行了处理。包括采用一种新的以语言两元组模型为基础的方法评估织物的风格，将感官数据融合为统一的形式。并利用所建立的模块对不同小组的专家以及评估用语进行分析。 第四章对客观数据进行了处理。处理模块包括客观数据处理模块和客观数据聚类模块。在前一个模块中，分别采用主因子分析和主成分分析从KES参数中提取特征参数，通过降低数据空间的维数来精简系统。在后一个模块里，以客观数据为基础用分级聚类方法和模糊毛均值聚类方法对织物样品进行风格分类。 第五章采用自适应神经网络的模糊推理系统建立客观与感官数据间的联系。通过这个模块，用户可以在输入相关的客观数据后得到单个指标的感宫评估结果或者织物的综合评估结果，同样，如果输入具体织物样品的感官数据就可得到其相应的物理参数值。 第六章开发了在线智能感官评估系统。其中，用Java语言编写了主体程序，用SQL语言管理数据系统，用Fro址Page制作了网页。 第七章论文结论。 本论文的研究工作着重技术的创新，主要的创新点如下:(l)与以往的经典计算方法不同，本文采用了模糊、神经网络方法处理数据。(2)采用功temet作为感宫评估的服务平台，并用一系列新型的软件编写了相关的程序以实现纺织品感官评估的网络化、智能化。更多还原

【Abstract】 To complete the intelligent sensory evaluation for Textile Products, we use the methods of fuzzy techniques, neural networks, classical factorial analysis and other data analysis to analyze and transact the sensory data which from different panels, as well as the objective data from KES, Moreover we also build the relationship between them. Based on which we establishe the Web-Based Intelligent Sensory Evaluation System for Fabrics.In disposing the objective data, in order to avoid of high dimension problem, we develop some methods for selecting the relevant rules that are used to build the relationship between the selected objective feature parameters and the sensory fabric data. We use Java language and SQL database to develop the sensory evaluation system in Internet environment. So given the objective data of a new fabric sample, the system can provide its sensory data and its total hand grade. In meantime, if the sensory data are also given the total hand grade can be obtained directly from the sensory fabric hand dataIn the first chapter, we firstly introduce several basic conceptions and then present contents of the thesis.In the second chapter, taking into account the theory bases and the practical demands of market, the definitions of sensory evaluation and intelligent sensory evaluation are introduced. Based on it we put forward the whole model of intelligent sensory evaluation system, as well as its sub-models. Then the function of individual model and the relationships among them are explained.In the third chapter, we treat with the sensory date by the model of Sensory Data Aggregated Model (SDAM), in which a new linguistic 2-tuple fuzzy model based approach is proposed for evaluating fabrics, and aggregating the sensory data. Otherwise using the built model to analyze different panels and evaluation results.In the fourth chapter, we deal with the objective date, including the Objective Date Processing Model (ODPM) and the Objective Date Clustering Model (ODCM). In the former model, using PCA we extract the features from KES parameters to reduce the dimension of discussed data space. In the latter one, using grade clustering method and FCM we give the classes of the samples based on physical parameters.In the fifth chapter, we take use of ANFIS to build the relationship between the sensory data and objective data. Through this model, given theobjective data we can get one or all sensory evaluation of criterion, conversely we can also get the sensory data.In the sixth chapter, we develop the WISES, in detail Java language is used to compile the mail program, SQL is used to manage the data system and Frontpage is used to make web page.In the final chapter, we prospect the next study work for making progresses.The work in this thesis puts focus on the innovation in techniques, following is the main innovation: (1) fuzzy logic and neural networks are used to deal with data which differ from old classical methods. (2) the Internet is used as the service platform, we use several kinds of software to write programs concerned.更多还原