基于数据驱动模拟电路故障预测算法实现与软件开发

发布时间：2018-06-09 00:39

本文选题：自回归模型 + 灰色模型　；参考：《电子科技大学》2015年硕士论文

【摘要】：据统计,电子设备中80%的故障来自模拟电路,因此,模拟电路的故障预测研究成为提高电子设备可靠性的关键。在故障预测领域,基于数据驱动的故障预测方法是一种使用较为广泛的预测方法,主要是利用设备的仿真数据、故障注入数据等,通过数据分析和处理算法进行趋势预测。自回归预测模型(AR)具有建模简单、计算快捷的优点,灰色模型所需建模数据量少、预测精度高,基于上述原因,本文对故障预测软件的总体设计及关键模块实现做了深入研究,具体的研究工作有:1.预测软件总体框架设计。将预测软件划分为故障预测、数据库管理、界面服务、用户权限管理四个功能模块。在VC++6.0平台下,对软件中的数据流向进行设计,并以属性表单和属性页完成功能模块的集成。2.实现了基于数据驱动的故障预测算法。故障预测算法是预测软件的核心,本文根据自回归预测模型、灰色预测模型以及粒子群优化算法(particle swarm optimization,PSO)的建模步骤,完全采用C++标准库函数进行算法编程和接口设计,没有涉及到与MFC相关的函数库,使得本文实现的预测算法的调用不局限于具体的操作系统,增加了预测算法在不同操作系统之间的移植性以及不同环境中的复用性。3.实现了故障预测软件平台的开发。首先,按照第三范式的要求设计了服务器端数据库,在软件里面采用ADO的连接方式访问数据库,实现了对故障信息、预测结果等数据的存取;其次,采用面向对象的编程思想对各个功能模块所使用的方法和类进行设计和封装,整个软件平台的搭建基于MFC对话框的开发方式进行,充分利用了MFC提供的控件以及Visio ActiveX图形化开发控件,达到人机交互友好的目的。4.预测软件的测试验证。以绝缘栅双极晶体管(Insulated Gate Bipolar Transistor,IGBT)实测数据以及Sallen-key模拟电路仿真数据为测试验证实例,通过将实际的电路剩余使用寿命(Remain Useful Lifetime,RUL)和软件预测的RUL作比较,结果表明软件对模拟电路的RUL具有很高的预测精度,可以为电路系统的健康管理提供实时可靠的理论依据。
[Abstract]:According to statistics, 80% of the faults in electronic equipment come from analog circuits. Therefore, the research on fault prediction of analog circuits becomes the key to improve the reliability of electronic devices. In the field of fault prediction, data-driven fault prediction method is a widely used prediction method, which mainly uses the simulation data of equipment, fault injection data and so on, and carries out trend prediction through data analysis and processing algorithm. The autoregressive prediction model (ARM) has the advantages of simple modeling, fast calculation, less modeling data and high prediction accuracy. Based on the above reasons, this paper makes a deep study on the overall design of the fault prediction software and the realization of the key modules. The specific research work is 1: 1. The overall framework design of prediction software. The prediction software is divided into four functional modules: fault prediction, database management, interface service and user authority management. In VC 6.0 platform, the data flow direction in the software is designed, and the function module integration. 2. 2 is completed by attribute form and attribute page. A fault prediction algorithm based on data drive is implemented. Fault prediction algorithm is the core of prediction software. According to the modeling steps of autoregressive prediction model, grey prediction model and particle swarm optimization algorithm, the C standard library function is used for programming and interface design. There is no MFC related function library, which makes the call of the prediction algorithm not limited to the specific operating system, and increases the portability of the prediction algorithm between different operating systems and the reusability of the prediction algorithm in different environments. The development of fault prediction software platform is realized. First of all, according to the requirements of the third normal form, the server database is designed. In the software, ADO connection is used to access the database, which realizes the access to fault information, prediction results and other data. The methods and classes used in each functional module are designed and encapsulated by object-oriented programming thought, and the whole software platform is built based on the development of MFC dialog box. It makes full use of the control provided by MFC and the Visio ActiveX graphic development control to achieve the purpose of friendly man-machine interaction. 4. Test and verify the prediction software. Taking insulated Gate Bipolar Transistor IGBT (Insulated Gate Bipolar Transistor IGBT) and Sallen-key analog circuit simulation data as test examples, the residual service life of actual circuit is compared with Rul predicted by software. The results show that the software has a high prediction accuracy for the RUL of analog circuits and can provide a real time and reliable theoretical basis for the health management of circuit systems.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN710

【共引文献】