用于航空发动机动态辨识的MSMEA-ELM算法

MSMEA-ELM algorithm applied in aero-engine dynamic identification

针对传统思维进化算法搜索半径缺乏目的性,临时子群体补充缺乏方向性以及神经网络训练速度慢、泛化能力不足,传统极端学习机隐含层神经元个数多的缺点,提出一种多群体自适应思维进化算法优化的极端学习机(MSMEA—ELM)算法,通过传感器数据训练该算法用于对航空发动机大范围动态过程进行辨识。以训练均方误差与权值2范数的加权和最小为优化目标,采用多群体自适应思维进化算法优化极端学习机。以某型涡扇发动机为研究对象,采用MSMEA—ELM算法进行航空发动机动态过程辨识,验证了该算法的有效性。

In traditional mind evolutionary algorithm,the searching radius is lack in target and supplement of temporary groups is lack in direction,neural network had slow training speed and weak generation ability,traditional extreme learning machine has many nerve cells in hidden layer. In order to solve these problems,multi-group self-adaptive mind evolutionary algorithm optimized extreme learning machine（ MSMEA—ELM）algorithm is proposed to solve these problems. Aiming at optimizing the adding of weighted mean square error and weighted 2 norm of weights,multi-group self-adaptive mind evolutionary algorithm is introduced to optimize ELM.The demonstration test is conducted on a turbofan engine and MSMEA—ELM algorithm is applied in aero-engine dynamic identification. The result shows that the introduced algorithm is efficient.