基于多时间尺度模型的变工况大型旋转机械健康管理研究

项目名称： 基于多时间尺度模型的变工况大型旋转机械健康管理研究

项目编号： No.51475455

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 机械、仪表工业

项目作者： 李伟

作者单位： 中国矿业大学

项目金额： 80万元

中文摘要： 大型旋转机械系统复杂、工况多变、信号频率多样，其突发故障与性能退化现象在不同时间尺度上呈现。首先，提出随机窗口分析方法，解决高频振动信号与低频缓变信号特征的统一表征问题，在多时间尺度上描述系统多源输入输出信号，并运用拟合、辨识及统计等技术，建立小时间尺度稳态拟合模型、瞬态状态空间模型和大时间尺度运行统计模型，表述复杂系统在多时间尺度上的输入输出耦合关系；其次，拓展拟合模型左零空间解耦方法和状态空间模型观测器解耦方法，解决工况变化与系统输出耦合的问题，在小时间尺度上生成仅反映故障信息的残差，实现突发故障快速诊断；然后，基于大时间尺度一致性假设，探讨系统退化程度与运行统计模型参数的关系，解决系统在复杂工况下长期运行时的寿命预测问题，并利用在线规划技术，研究成本优化的维护方法；最终，形成基于多时间尺度模型的变工况大型旋转机械健康管理方法，为复杂机电系统的可靠性、安全性和可维护性研究提供思路。

中文关键词： 多时间尺度；故障诊断；寿命预测；智能维护；旋转机械

英文摘要： Large-scale rotating machinery is with complicate operating conditions and complex sensor signals. This project employs dynamics theory, fault diagnosis and prognostic theory to develop a multi-time-scale health management method for rotating machinery under varying operating conditions. Firstly, with the proposed statistical window analysis method, the signals of high frequency and very low frequency are described through a unified way. Secondly, the small time scale input-output fitting model, the small time scale input-output state-space model and the large time scale statistical model are created by applying fitting methods, subspace identification method and statistical method. Thirdly, faults are diagnosed based on the left null space decoupling method for fitting model and observer decoupling method for state-space model, where the effects of varying operating conditions are eliminated and the performance of fault diagnosis is also evaluated. Fourthly, the residual life time percent is taken as the health indicator, and the health state is predicted based on the large time scale statistical model with neural network techniques. The maintenance cost of the whole machine per unit time is minimized by applying a preventive mainte nance strategy. The results will improve the safety, reliability and arability of rotating machinery, and reduce the maintenance cost.

英文关键词： Muti-time-scale;fault diagnosis;life prediction;intelligent maintenance;rotating machinery