论文部分内容阅读
在动态系统基于模型诊断中,状态空间大小与元件个数、时间是双指数关系.K-Best枚举方法每个时刻只考虑K个可能性最大的状态,有效减小了枚举空间,但当系统复杂庞大或诊断周期长时,状态更新仍是一项巨大工程.提出一种结合粒子滤波和不确定图的动态系统诊断方法 PF_LUG,利用粒子在状态空间的分布近似其概率,并用不确定图标签的反向匹配代替传统的正向轨迹枚举.算法有效解决了由时间导致的计算量增长问题,使时间对复杂度的影响由指数运算降为乘积运算.仿真结果表明该算法的运行时间相对诊断周期线性增长,比K-Best枚举有明显优势.
In the dynamic system model-based diagnosis, the state space size and the number of components and time are exponential. The K-Best enumeration method only considers K most probable states at each moment, which effectively reduces enumeration space, When the system is complex or the diagnosis period is long, state updating is still a huge project.A dynamic system diagnosis method PF_LUG based on particle filter and uncertain graph is proposed, which uses the distribution of particles in the state space to approximate the probability and uses uncertain The reverse matching of the icon label replaces the traditional enumeration of forward trajectory.The algorithm effectively solves the problem of time-induced increase of computation volume and reduces the influence of time on complexity from index operation to product operation.The simulation results show that the operation of the algorithm The linear growth of the relative diagnostic cycle time, than the K-Best enumeration has obvious advantages.