中文摘要：

利用燃气涡轮发动机数值仿真软件（GSP）建立涡轴发动机性能仿真模型,采用退化因子方法得出部件性能退化后发动机测量参数的变化,并以此分析部件性能退化对发动机性能的影响。针对发动机单个部件性能对整机性能的影响权值难以定量的问题,提出采用随机赋权值的极限学习机（ELM）算法诊断发动机部件性能退化。仿真结果表明,运用ELM算法进行涡轴发动机部件性能退化诊断的平均精度可达97.5%,速度也明显快于BP等传统神经网络。

英文摘要：

The performance simulation model of a turbo-shaft engine was established by gas turbine simulation program （GSP）, and then the change of measuring parameters could be obtained by the use of degradation factor methods. It is difficult to quantitate problems by the effect weights of a single engine component performance on the performance of the whole engine. Thus the algorithm of extreme learning machine （ELM） which weights are randomly assigned to diagnose engine parts performance degradation was proposed. According to the analysis results, the mean diagnosis accuracy of turbo-shaft engine performance deterioration by extreme learning machine algorithm has reached 97.5%, and it is faster than Back Propagation neural network and other traditional methods.