基于量子过程神经网络的旋转机械健康状态预测研究

发布时间：2018-04-21 20:46

本文选题：旋转机械 + 量子计算　；参考：《哈尔滨理工大学》2017年硕士论文

【摘要】：工业领域中旋转机械的应用十分广泛,随着旋转机械向大型化和集成化方向的快速发展,其设备的结构变得越来越复杂,其运行状态更加受到人们的关注。通过对旋转机械设备健康状态趋势进行预测可以及时发现故障,提高旋转机械设备的安全性、利用率与可靠性。因此,对旋转机械健康状态进行深入研究具有非常重要的现实意义。本文针对旋转机械复杂的工况环境及非线性的时变特性的问题,对振动数据降噪方法、全局预测模型和集成预测模型进行了研究。提出一种基于相关系数的时频矩阵SVD降噪方法,消除了强背景噪声对奇异值选取的影响,实现信号重构阶次的自适应性选择。将该方法应用到滚动轴承健康状态信号中,降噪结果证明了该方法的有效性。针对目前在旋转机械领域网络预测模型存在的一些不适应性,提出一种基于受控H门的量子过程神经网络模型,实现了模型信息的并行处理及网络参数的多维调整,并避免了过程神经网络在函数拟合和基函数展开的过程中存在误差的问题。滚动轴承性能预测实验表明,基于受控H门的量子过程神经网络具有良好的预测精度。针对单一全局模型复杂难以优化的问题,提出了基于GA-Ada Boost.RT的集成预测模型。分别以过程神经网络、离散输入过程神经网络作为弱学习机构建集成预测模型。通过对滚动轴承的健康状态预测实验结果表明,由结构简单弱学习机构建的集成预测模型具有比全局模型更好的预测效果。基于上述研究成果,结合VC++和Matlab软件的优点,开发了旋转机械健康状态监测系统。通过对旋转机械健康状态预测的研究,可缩短维修时间,减少维修费用,提高设备的利用率、可维修性和可靠性,避免灾难性事故的发生,对拓展及丰富旋转机械健康状态预测方法具有重要的理论意义和实用价值。
[Abstract]:Rotating machinery is widely used in industry. With the rapid development of rotating machinery to large-scale and integrated direction, the structure of its equipment becomes more and more complex, and its running state is paid more attention by people. By predicting the health state trend of rotating machinery equipment, the fault can be found in time, and the safety, utilization ratio and reliability of rotating machinery equipment can be improved. Therefore, it is very important to study the health state of rotating machinery. In this paper, the noise reduction methods of vibration data, the global prediction model and the integrated prediction model are studied in view of the complex operating conditions and nonlinear time-varying characteristics of rotating machinery. A time-frequency matrix SVD denoising method based on correlation coefficient is proposed, which eliminates the influence of strong background noise on singular value selection and realizes the adaptive selection of signal reconstruction order. The method is applied to the health state signal of rolling bearing, and the result of noise reduction proves the effectiveness of the method. A quantum process neural network model based on controlled H gate is proposed to solve the inadaptability of the network prediction model in the field of rotating machinery at present. The parallel processing of model information and the multi-dimensional adjustment of network parameters are realized. The problem of error in the process of function fitting and basis function expansion is avoided. The experiment of rolling bearing performance prediction shows that the quantum process neural network based on controlled H gate has good prediction accuracy. An integrated prediction model based on GA-Ada Boost.RT is proposed to solve the problem of complex and difficult optimization of a single global model. Process neural networks and discrete input process neural networks are used as weak learning machines to construct integrated prediction models. The experimental results on the health state prediction of rolling bearings show that the integrated prediction model constructed by simple weak learning machine has better prediction effect than the global model. Based on the above research results and the advantages of VC and Matlab software, a rotating machinery health monitoring system is developed. Through the research on the health condition prediction of rotating machinery, the maintenance time can be shortened, the maintenance cost can be reduced, the utilization ratio, maintainability and reliability of the equipment can be improved, and the catastrophic accident can be avoided. It has important theoretical significance and practical value to expand and enrich the health state prediction method of rotating machinery.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17;TP183

【相似文献】