基于证据推理的动态系统安全性在线评估方法

综合考虑动态系统历史记录、当前状态以及未来退化趋势信息来对其安全性进行在线评估是极其重要的.本文提出了一种基于证据推理（Evidential reasoning，ER）的安全性在线评估方法.该方法先融合多个安全性指标获得各个时刻的安全性状态，而后融合系统"历史"、"当前"、"未来"时刻的安全性状态，评估得到系统的综合安全性水平.首先，建立了基于三阶Volterra滤波器的在线预测模型，预测指标未来信息；然后，建立了指标最优自适应权重求取模型，计算并更新指标实时权重；最后，提出了基于证据推理方法的融合框架，对"历史"、"当前"、"未来"时刻的信息进行融合，得到系统当前时刻的综合安全性评估结果.通过对某惯性平台系统的安全性评估实例验证了所提方法的有效性.

Online Safety Assessment Method Based on Evidential Reasoning for Dynamic Systems

It is of great significance to online assess the safety of a dynamic system by taking into account historical records, current state, and degradation trend. This paper proposes a new safety assessment method based on the evidential reasoning (ER) approach. To obtain the integrated safety level, multiple safety indicators are fused at first and the "history", "current" and "future" safety states are then integrated. Firstly, a forecasting model based on a third-order Volterra filter is proposed to online predict the safety indicators' information. Secondly, an optimal adaptive fusion weighting model is developed to calculate and automatically update the weighing coefficient. Finally, a safety assessment aggregation scheme based on the ER approach is presented to fuse the "history", "current" and "future" safety information synthetically to obtain a comprehensive safety assessment result of the dynamic system. A practical example of the inertial platform is studied to validate the effectiveness of the proposed ER-based safety assessment method.