融合多源信息与知识模型的印刷机故障逆向定位及耦合机理研究

The printing machine is a type of equipment which is characterized with high-speed operation and complex structure. The frequent failure during operation often leads to a waste of numerous printed materials. It is difficult to apply the well-established failure diagnosis method to accurately diagnose the fault and analyze the coupling mechanism of printing failure, due to the strong coupling between the multiple actuators of press groups and multi-media materials system. In this project, we propose a new failure diagnosis method by integrating multi-source information and knowledge model to achieve the reverse positioning of the failure and analyze the coupling mechanism to address the above limitations. Firstly, the printing press fault feature database of rich information will be established with the construction of a multi-sensor system to monitor the working state of printing machines, as well as collecting multi-source information, combining with implicit failure information form the printing screen. Secondly, multivariate statistical approaches will be utilized to help characterize the main element feature of printing press, establishing the mapping relationship and the affecting weights between the failures of machine. As a result, the reverse fault source trace will be completed, which will bring about the analysis of the coupling mechanism of printing press fault. Finally, with the help of the artificial knowledge model, the intelligent decision-making algorithm of the printing press will be established, such that an accurate diagnosis of printing machine fault will be achieved.This project will provide a new method for printing press fault diagnosis and condition monitoring, as well as a helpful probe in fault diagnosis for large complex process equipment, which will enrich the theoretical system of the mechanical failure diagnosis.

印刷机是一种结构复杂、高速度运转的高精密设备，其运行过程中频繁的故障常导致大量印刷物料的浪费。由于印刷机中多组执行机构与多介质物料系统之间存在强耦合作用，已有的故障诊断方法难以有效应用于印刷机故障诊断和耦合机理解析。针对上述问题，本项目提出融合多源信息与知识模型的印刷机故障诊断方法，以实现故障逆向定位，阐明耦合机理：首先，构建多传感器系统监测印刷机运行状态，采集多源信息，结合印刷画面中隐含的印刷机故障信息，建立信息丰富的印刷机故障特征集；其次，通过多元统计方法构建表征印刷机状态的主元特征，确立故障现象之间的映射关系及影响权重，完成故障源的逆向追溯，从而实现故障耦合机理解释；最后，融合人工知识模型，建立印刷机故障诊断的智能决策算法，完成印刷机故障的精确诊断。本项目的研究将为印刷机故障诊断和状态监测提供一种全新的方法，同时为大型复杂过程装备故障诊断提供有益的探索，将丰富机械故障诊断的理论体系。

印刷机是印刷机是集机、光、电、液、气、控制、化学、网络等技术于一体，并且结构复杂的精密机电产品，在知识传承、产品包装、有价证券、印刷电子等领域具有广泛的应用，对推动社会文明的进步和人类社会信息的交流发挥了重要的作用。由于印刷机中多组执行机构与多介质物料之间存在强耦合作用，对其进行故障诊断是故障诊断方法应用于特种设备挑战性难题。.本项目探索性地对印刷机的故障诊断展开研究，从印刷机的产品中提取故障标识的图像信息，挖掘图像信息中设备运行状态信息特征，构建了表征印刷单元状态的多元特征集以及图像与故障模式间的映射关系，实现了利用印刷画面特征进行印刷机故障诊断；构建了印刷机多源信息的测试系统，获取了印刷机中关键传动部件轴承、齿轮等关键零件的故障信号，提出了基于自适应的的多源信息特征提取方法，结合模式识别方法对印刷机的轴承与齿轮故障进行了分类，完成了未知故障的诊断；在此基础上，引入声音信号，振动信号构建了多源信息特征集，提出了基于流形学习的多源特征的融合方法，结合支持向量机分类模型化LE方法，实现了印刷机轴承及齿轮故障的分类与诊断；构建了基于知识规则和图像信息的印刷机故障诊断分析系统，利用印刷机产品中蕴含大量机构状态信息以及积累的大量人工知识经验等信息源。研究了基于灰度共生矩阵的印刷画面特征提取方法，基于主元分析的冗余信息去除方法以及基于支持向量机的决策网络的构建方法，开发了基于画面的印刷机故障诊断分析系统，通过实验验证了诊断的可靠性。.迄今为止，项目组在国内外学术期刊及会议上共发表论文9篇，培养2名博士研究生和5名硕士研究生，申报发明专利3项，目前已授权2项。