作　　者： (魏芳）; (梁茂宗）; (马晶玮）; (周登极）; (张会生）;

机构地区： 中国航发商用航空发动机有限责任公司,上海201108 上海商用飞机发动机工程技术研究中心,上海201108

出　　处： 《燃气涡轮试验与研究》 2017年第4期11-15,22,共6页

摘　　要： 针对多方法计算所得发动机部件性能降级无法直接应用Dempster-Shafer(D-S)证据理论融合的问题,提出一种实现降级参数量化融合的方法。以一段时间内计算结果作为样本,融合性能模型输出结果和神经网络输出结果,通过划分区间统计样本落入区间的频率构造基本概率赋值(BPA),从而实现性能降级的量化融合。对比了独立区间划分法和嵌套区间划分法两种BPA构造方法的融合结果,得出了嵌套区间划分法构造BPA具有更适合量化融合的主要结论。 Aiming at the problem that performance degradation calculated from multiple methods cannot be directly fused by Dempster-Shafer(D-S) evidence theory,a method was proposed.The results calculated over a period of time were taken as sample,and the outputs of the performance model and the ANN model were fused.Through building basic probability assignment(BPA) by calculating the frequency in divided intervals,the quantitative fusion of degradation was realized.After compared the fusion results of two BPA building methods(independent interval method and nested interval method),the main conclusion was drawn that nested interval method was preferable for quantitative fusion.

关 键 词： 航空发动机 证据理论 性能降级 基本概率赋值 嵌套区间划分 量化融合 信息融合