【作者】 马海涛；

【导师】 闫纪红；

【作者基本信息】 哈尔滨工业大学 ， 机械电子工程， 2009， 硕士

【摘要】 随着制造企业对先进制造技术及装备的需求日益增强,设备的可靠性、运行效率、以及无故障时间对于实现安全运行、提高企业的经济效益和市场竞争力至关重要。采用先进的智能维护管理模式,利用智能预诊技术和网络通讯手段,充分发挥信息化技术在设备维护管理中的作用,进而达到对重大设备和关键部件的运行状况进行实时性能评估及剩余寿命的动态预测,对实现设备的全生命周期智能管理,以及近零故障运行具有重要意义。因此,本文立足于智能预诊的实施及网络化实现,对构建远程智能预诊工具进行研究。首先,本文从实施智能预诊的角度出发,提出智能预诊统一的四层框架结构,完整地描述了各子模块的功能,并特别指出智能预诊数据格式规范化对拓宽智能预诊适用范围的重要性。在此基础上,分析智能预诊网络化实现的可行性,研究服务器资源及网络承载能力有限给智能预诊网络化实现带来的问题,并提出传输特征数据和应用快速收敛算法的解决方案。其次,本文采用支持向量回归机算法解决了智能预诊网络化实现快速响应的问题,分析学习参数对回归模型精度的影响,分别采用遍历寻优、遗传算法寻优、多核结合的方法逐步改善模型的整体性能,保证了模型回归精度,同时提高了模型的泛化能力。将该方法应用于汽轮机转子实际寿命损耗预测,取得了良好的效果。第三,对智能预诊网络化实现的关键技术进行研究,分析比较动态网页开发技术、数据库及接口技术、Matlab与Web接口技术等关键技术,提出离线模式与在线B/S模式相结合的远程预诊模式,并确定使用“Apache + MySQL + PHP”软件组合搭建预诊服务平台。最后,本文以汽轮机转子为应用对象,支持向量回归机为智能预诊方法,结合智能预诊框架,在Windows系统环境下构建基于Web的远程智能预诊工具。其实施过程包括:服务平台搭建、系统程序及显示结构设计和主要功能模块开发等。通过将工具移植到Linux系统下,提高工具的执行效率和安全性。该工具的成功实现,为基于Web的智能预诊提供了参考范例。更多还原

【Abstract】 As the requirement of advanced manufacturing technology and equipments increases, it is an essential issue to keep the reliability and improve the efficiency of the equipments in enterprises. It is significant for the equipments whole life intelligent managing and near-zero-downtime to implement real time evaluation and rest life dynamic prediction of the crucial equipments and their fatal parts, by adopting intelligent maintenance management, using intelligent prognostics and network instruments, and enhancing the role of the information-based technology in maintenance management. To meet this situation, a tool of remote intelligent prognostics is developed aiming at the implementation of intelligent prognostics and its web realization.Firstly, a uniform four layers framework is proposed in the perspective of implementing intelligent prognostics, and the function of modules in each layer is described clearly. Especially, it is crucial to standardize the data format for increasing the application area of intelligent prognostics. The feasibility of web based intelligent prognostics is analyzed, and the transmission of feature data and the application of intelligent algorithm with high convergence speed is researched to resolve the limitation of server resource and network load for implementing web based intelligent prognostics.Secondly, support vector regression machine (SVR) is employed to meet the fast response request of web based intelligent prognostics. The influence of SVR learning parameters to the precision of regression model is researched, and the regression precision and generalization capability is improved gradually by using the gird-on search, GA and mixed kernels, respectively. The method is applied in prediction of life loss of rotors by selecting the feature of turbine rotor actual running data and the results show the effectiveness.Thirdly, the key techniques of implementing web based intelligent prognostics, such as the development technology of active pages, database and its interface, interface technology of Matlab and Web, etc. A remote prognostics mode is presented by combining the outline with online B/S mode, and Apache, MySQL and PHP integrated software is used to build the prognostics service platform.Finally, a web based tool of intelligent prognostics is built for the rotor by using SVR as the method under the framework of intelligent prognostics on the platform of Windows System. The process of implementation is as follows: setting up the service platform, designing of the system program and display structure, and developing main functional modules. And it is successfully transplanted to Linux System to improve the execution efficiency and security after tested. The achievement of this tool provides guidance to implementation of the web based intelligent prognostics.更多还原