【作者】 倪永成；

【导师】 杨建国；

【作者基本信息】 东华大学 ， 机械电子工程， 2007， 硕士

【摘要】 过程是一组将输入转换为输出的相互关联或者相互作用的活动。加工过程强调制造过程中每个工站的对产品质量的影响。质量预测强调对产品质量的提前控制，避免不良产品的出现。使用标准的方法将工件的质量特征参数值测量出来，再将质量特征值与造成此参数变化的因素联系起来，建立关系模型。通过对过程参数的测量，再根据其关系模型即可实现对质量参数的预测。质量控制属于质量管理的范畴，其目的为了使产品的固有特征能够达到规定的要求。质量控制贯穿产品形成的全过程，强调对产品质量加工过程的质量进行控制，突出通过控制过程质量的方法来控制最终的产品实物的质量。可以预见，面向产品质量加工过程，研究基于过程的质量控制方法必将成为质量控制技术研究领域的重点。本文针对棉纺纺纱工艺的特点，研究了在复杂工艺条件下的质量预测与控制技术。首先，针对棉纺工艺多工序，复杂工艺的情况，我们采用分段工艺质量预测模型对纺纱工艺的每道工序进行建模。建模的过程充分考虑每道工艺的输入，当前工序的工艺与设备参数，原棉质量参数的影响和每道工序的输出。我们使用ANN对采集的棉纺历史数据进行训练，得到适应企业实际生产、工艺和质量条件情况下的数学模型，利用该模型可以实现对当前纱线质量的有效预测。其次，为了有效的利用企业长期积累的历史数据，提取有价值的潜规则，对企业生产决策提供依据，我们使用数据挖掘领域的粗糙集理论，对某棉纺厂的原始数据进行规则提取。相对传统的人工智能方法，粗糙集理论对于不准确、不完整和冗余数据的离散化数据的规则提取具有一定的优势。最后，我们研究的基于SPC控制图原理的模式分类。我们以X-bar图作为基础来对原始数据进行预处理，采用支持向量机分类算法作为我们模式分类的工具，在算法的实现过程中我们使用网格搜索法对支持向量机参数进行了优化。经实验表明，支持向量机分类机是很好的模式分类算法，可以达到较高的预测精度。针对以上的研究领域，我们开发了棉纺质量预测与控制平台，该平台可以帮助管理人员顺利的使用上述工具进行建模、规则提取和模式分类。更多还原

【Abstract】 Process is a group of interactive association or function which transforms input to output. Machining process emphasize the action of each step to product in the manufacture course. Quality forecast emphasize the ahead control to the product’s quality, to avoid the defective product. With the standard method we can measure the quality character values of a workpiece, then we associate this quality character values with the correlative factors to build the association model. Via the measuring of the process parameters and the association model we can implement the forecast of the quality parameters. Quality control belongs to the field of quality management, in order to guarantee the product’s inter characters to meet the stated requirement. Quality control runs through each step of the production, emphasize the control of the product’s quality machining process, stress the method of control the process to ensure the final product’s quality. We can predict that the research about the quality control method based on the product machining process will be the hotspot.In this paper we research the quality forecast and control technology in the complex technics, especially in the cotton’s spinning process.Firstly, aiming at the multi working procedure and complex technics of the cotton’s spinning process, we use the subsection technics forecast model to build each model to each step of spinning process. During this course, we think about the effect of each step’s input, output, process and equipment’s parameters and raw cotton’s quality parameters. We utilize the ANN tool to build the model based on the history datum of one cotton spin company, which can get a good effort of forecasting the yarn’s quality for this model depends on the datum of this company.Secondly, we use data mining algorithm: rough set theory to extract the potential and valued rules based on the raw datum, which can help the company’s decision. Comparing with the other AI method, Rough Set has a advantage in extracting rules from those uncertain、redundant、abridged datum.Finally, we research the pattern classify based on the SPC control diagram. We scalar the raw datum according to the X-bar control diagram, then use the SVM classification algorithm to implement the pattern classification and optimize the SVM parameters via the grid-search method. According to the experiment result, we conclude the SVM is an effective pattern classify algorithm, can reach a high forecast accuracy.We also implement the platform of cotton’s spinning quality forecast and control system based on the above research field, which can help the management employee to use those tools to build model, extract rules and pattern classify.更多还原