基于印刷图像信息挖掘的印刷机故障诊断方法研究

项目名称： 基于印刷图像信息挖掘的印刷机故障诊断方法研究

项目编号： No.51275406

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 机械、仪表工业

项目作者： 张海燕

作者单位： 西安理工大学

项目金额： 66万元

中文摘要： 印刷机是一种高精密且结构复杂的特种机械，其频繁故障常导致巨大的经济损失。在印刷过程中印刷和定位等关键部件会产生同频、倍频振动信号，由于采集的信号难以分解的原因，传统的故障诊断方法不能有效的应用于印刷机故障诊断，因此研究新的高精度的故障诊断方法具有科学意义及应用需求。本项目结合印刷工艺特点和图像识别技术，通过研究印刷图像所携带的信息与印刷机故障之间的映射关系反演印刷机故障。首先研究一种新的图像预处理算法，既能消除在线采集印刷图像时光照对印刷图像色差的影响，又能够增强印刷图像所携带的故障信息；其次研究在L*a*b*色彩空间进行印刷图像纹理特征识别与提取的方法；最后研究印刷机故障特征参数的选取，运用KPCA和SVM结合的方法来精确表征印刷机故障。此项目将为印刷包装行业带来巨大经济效益，同时对同类机械的故障诊断提供理论基础和技术手段。

中文关键词： 故障诊断；图像处理；特征提取；多元统计分析；

英文摘要： The printing is a special machinery of high-precision and complex structure, the frequent fault often leads to huge economic losses.Print machine, which generates the same or doubling frequency signals between sets and key parts, is a kind of same-frequency machine. These signals are difficult to analyze because of interference between them. As a result, the fault diagnosis method based on signal vibration does not work efficiently on print machine. So a new fault diagnosis method is needed both by scientific and technologic field. Combined with specialty of print technic and technology of image recognition, this research deduces print machine failure by exploring relationship between information on printed image and print machine failure. First, a new image pretreatment algorithm based on RETENIX and weighted least squares is proposed. This algorithm can not only eliminate the effects of light to print image acquiring online, but also make information generated by Print machine failure more abundant. Second, algorithm for Print machine register failure character picking-up is carried out. At the same time that method for print image texture character which is on L*a*b* colour space recognization and picking-up is carried out. It will be a feasible method for picking-up character on print machine shadow-overlap

英文关键词： Fault Diagnosis；Image Processing；Feature Extraction；Multivariate；