基于信度规则库的惯性平台健康状态参数在线估计

 实时准确的健康状态预测是规划惯性平台系统及时、经济的维修策略的关键技术。由于平台系统的健康状态是不能够直接观测的，假设平台系统的特征参数监测数据是可以获取到的，而且平台系统的健康状态与特征量是相关的。基于信度规则库(BRB)，以平台系统的状态监测特征参数作为BRB系统的输入，以平台的健康状态作为输出结果，组建了惯性平台健康状态预测系统。为了克服现有BRB参数优化方法的不足，实现实时状态预测，基于期望最大化(EM)算法，研究了健康状态预测系统的参数在线估计算法。该算法在获取系统新的输入输出信息后，就对参数进行更新。利用本文提出的方法对惯性平台系统的健康状态实时预测问题进行了实验分析，实验结果表明：该方法可以有效地实现惯性平台系统健康状态预测模型参数实时估计；与参数离线优化方法相比，该方法不仅在预测精度上，而且在运行时间上都具有明显的优势；在工程实际中有良好的应用潜力。

Real-time Parameters Estimation of Inertial Platform's Health Condition Based on Belief Rule Base

A real-time and accurate health condition prediction for an inertial platform is essential for cost-effective and timely maintenance planning and scheduling. Due to the fact that the true health condition of the inertial platform cannot be observed directly, it is assumed that the observations of characteristic parameters are available from monitoring, and the characteristic parameters correlate with health condition of the inertial platform. In this article, a health condition prediction system for the inertial platform is established based on belief rule base (BRB), where the characteristic parameters of the inertial platform are used as the inputs of BRB system and the health condition of platform as the output consequence. To overcome the drawbacks of current parameter optimization algorithms for BRB and satisfy real-time prediction, a parameter estimation algorithm is investigated for online updating BRB prediction system based on the expectation maximization (EM) algorithm. When the new input-output information of system operation is available, the model parameter can be updated online. Real-time health condition prediction for the inertial platform system is validated using the established model and the algorithm under investigation. The experimental results show that the proposed method can implement online parameter estimation of health condition prediction for the inertial platform effectively. In addition, compared with offline parameter optimization method, the proposed method can generate better results in terms of prediction accuracy and operating time, and thus has great potential in engineering practice.