基于Python机器学习的可视化麻纱质量预测系统

发布时间：2018-02-12 10:37

本文关键词： Python 机器学习 GUI MEA-LM-BP神经网络小波神经网络　出处：《东华大学》2017年硕士论文　论文类型：学位论文

【摘要】：纺织加工过程中,原料的纤维指标与成纱质量指标之间存在一定的联系,我们可以借助神经网络算法来进行成纱质量预测。但是目前,我国的麻纱纺织行业企业,智能化程度仍然不够,例如,对数据的处理仍采用人工处理,虽然已有很多的研究表明神经网络在预测成纱质量方面有很高的精确度,但是普通纺织企业并不能将其运用到实际的生产过程中去。事实上,目前并没有以普通纺织行业的普通工人为受众的成纱质量预测系统的开发,导致这些预测模型没能实现用户化。(1)本文在前人对成纱质量预测算法的研究基础上,通过对苎麻纤维指标:束纤维平均断裂强(CN/Tex),精干麻平均支数(Nm),精干麻硬条率,精梳条麻粒,精梳条并丝,精梳条重量不匀,精梳条重量偏差,精梳条回潮率进行训练,对成纱不匀,成纱条干,成纱麻粒三个性能指标进行预测。首先采用MEA-BP预测模型进行建模分析,在此基础上,采用LM算法优化改进MEA算法,从而提高了模型的预测精度,加快了收敛速度。提出了基于MEA-LM-BP的神经网络算法对成纱质量进行预测。首次引入小波神经网络算法对成纱质量进行预测。通过建模分析得出小波神经网络在预测成纱质量方面有较高的预测精度。(2)本文借助Python平台设计并实现了基于模块化设计的成纱质量预测系统,设计了预测模块,数据处理模块以及可视化模块,并且提供了便捷的GUI界面,用户在使用时只需要以点击按钮的方式就可以实现使用模型完成成纱质量预测的操作。(3)运用web技术单独设计了一个简单的在线麻纱数据管理页面,方便数据的在线管理,方便相关人员可以随时随地进行数据的增、删,改等操作,最重要的意义在于此web页面可以直接与麻纱预测系统进行对接,甚至可以实现对实时数据的操作,这样的设计更符合大数据时代的潮流。
[Abstract]:In the process of textile processing, there is a certain relationship between fiber index of raw material and yarn quality index. We can use neural network algorithm to predict yarn quality. The degree of intelligence is still insufficient. For example, data processing is still done manually, although many studies have shown that neural networks have a high degree of accuracy in predicting yarn quality. But ordinary textile enterprises cannot apply it to the actual production process. In fact, at present, there is no development of a yarn quality prediction system aimed at ordinary workers in the ordinary textile industry. In this paper, based on the previous research on the prediction algorithm of yarn quality, the index of Ramie fiber, such as the average breaking strength of Ramie fiber, the average number of branches of dried Ramie, the percentage of hard sliver, the combed yarn, the average number of Ramie fiber, the average number of Ramie fiber, the quality of Ramie fiber and the quality of Ramie fiber, were studied in this paper. Combed sliver silk, combed sliver weight unevenness, combed sliver weight deviation, combed strip moisture regain are trained to predict three performance indexes: yarn unevenness, yarn drying and yarn granulation. Firstly, MEA-BP prediction model is used to model and analyze. On this basis, the LM algorithm is used to optimize and improve the MEA algorithm, thus the prediction accuracy of the model is improved. A neural network algorithm based on MEA-LM-BP is proposed to predict yarn quality. Wavelet neural network algorithm is introduced to predict yarn quality for the first time. Through modeling and analysis, it is concluded that wavelet neural network can predict yarn quality. This paper designs and realizes the yarn quality prediction system based on modular design with the help of Python platform. The prediction module, the data processing module and the visualization module are designed, and the convenient GUI interface is provided. Users can use the model to complete yarn quality prediction by clicking the button.) using web technology to design a simple online linen yarn data management page to facilitate the online management of data. The most important meaning of this web page is that it can connect directly with the yarn prediction system, and even realize the operation of real-time data. This design is more in line with the trend of the big data era.
【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS103.62;TP181

【参考文献】