摘要
并行推理是提高实时诊断速度和可靠性的有效途径．根据专家经验由假设建立一个候选集，每个候选故障的行为都是由与状态相关的瞬态分布征兆序列描述的．承担着并行推理任务的各个线程，在测量序列中寻找与候选故障具有一致行为的测量集，从而动态地调整候选集确定故障类型．在诊断推理过程中，根据候选集大小、候选概率和期望后验熵来动态地调整各个线程自身的优先权，以期达到最佳的实时诊断效果．
Parallel inference is an effective way to improve the real time diagnostic speed and reliability. A candidate set is set up with hypothesis according to experts experience, and the behaviors for each candidate fault are described by the temporally distributed symptoms relevant to the state. Each thread that takes on the parallel inference task searches for the behaviors of a measurement set. The behaviors are compatible with the candidate fault's in the measurement sequence, to determine the fault type with the update candidate set dynamically. The priority of each thread can be dynamically adjusted by itself according to the size of candidate set, the candidate probability and the expected posterior entropy in the process of diagnostic inference, so as to reach optimum real time diagnostic results.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
1998年第5期54-59,共6页
Journal of Southeast University：Natural Science Edition
关键词
瞬态分布征兆
实时诊断
并行推理
temporally distributed symptoms
real time diagnosis
parallel inference
