基于模型预测控制技术的切换系统的故障检测方法研究

Noting that the shortcoming of the existing methods of fault detection (FD), where the strict requirement of precision of modeling or quality of collecting data must be met, a novel FD method based on the model predictive control is proposed, including the establishment of switched predictive model in low order latent subspace, multi-mode fault prediction, compensation and feedback correction and the construction of the local optimal residual and dynamical threshold. In addition, the following key problems will be solved: the construction of the invariant set depending on the multi-mode, the design of the finite-time fault prediction reference trajectory and the disturbance attenuation in the process of feedback correction. The proposed results combine with the advantages of model predictive control scheme: First, the predictive function is only considered, which degrades the requirement of model. Based on the function, not only the current but also future fault information is sufficiently used. Further, the future information compensates the current one, which upgrades the quality of key data. Finally, the simple feedback correction will further improve the effect of FD. The unmanned helicopter flight control systems are considered to illustrate the effectiveness of the proposed method. The study of the project will provide some novel ideas for the FD problem of complex switched systems, which will be of great value to theory and applications.

现有故障检测方法对系统的建模精度或采集数据的品质具有极高的要求，本项目针对这一问题，以切换系统为研究对象，模型预测控制技术为解决途径，提出新的故障检测方法，主要包括低阶隐空间内的切换预测模型建立，多模态下的故障信息预测，补偿和反馈校正，局部最优残差及动态阈值的滚动构造。重点解决一些关键科学问题：多模态下预测不变集的构造，有限时间预测参考航迹的设计，反馈校正过程中的噪声抑制等。提出的方法汲取了模型预测控制技术的优点：注重模型的预测功能而非具体的形式，以此降低对建模精度的要求；通过预测功能实现系统“当前”和“未来”故障信息的有效结合，用“未来”补偿“现在”，提升采集数据信息的品质；简单易行的反馈校正将进一步提升系统的故障检测效果。最后，以多姿态下的无人直升机飞行控制系统为应用对象，实现对上述方法的验证。该问题的研究将为复杂切换系统故障检测问题的解决提供新的思路，具有重要的理论和应用价值。

为了解决故障检测方法对系统的建模精度或采集数据的品质提升问题，以切换系统为研究对象，模型预测控制技术为解决途径，提出新的故障检测方法，主要包括低阶隐空间内的切换预测模型建立，多模态下的故障信息预测，补偿和反馈校正，局部最优残差及动态阈值的滚动构造。重点解决一些关键科学问题：多模态下预测不变集的构造，有限时间预测参考航迹的设计，反馈校正过程中的噪声抑制等。提出的方法汲取了模型预测控制技术的优点：注重模型的预测功能而非具体的形式，以此降低对建模精度的要求；通过预测功能实现系统“当前”和“未来”故障信息的有效结合，用“未来”补偿“现在”，提升采集数据信息的品质；简单易行的反馈校正将进一步提升系统的故障检测效果。该问题的研究将为复杂切换系统故障检测问题的解决提供新的思路，具有重要的理论和应用价值。