自适应遗传神经网络算法在推力估计器设计中的应用

为了在全包线内能够准确方便估计出航空发动机推力,提出了一种自适应遗传神经网络算法:将遗传算法和神经网络技术相结合充分发挥遗传算法和神经网络各自的全局收敛性和局部搜索快速性的优点,其中通过自适应概率遗传操作及局部寻优算子直接优化出神经网络拓扑结构及权值(包括阈值),克服了神经网络隐层节点需凭经验尝试的缺点和神经网络对初始权值(包括阈值)敏感的缺点,再应用神经网络对上述优化的权值(包括阈值)进行"精调",最后设计出全包线推力估计器.经验证,此推力估计器具有较高估计精度和良好泛化能力.

Application of adaptive genetic neural network algorithm in design of thrust estimator

An adaptive genetic neural network algorithm(AGNNA) was proposed for estimating aeroengine thrust accurately and easily in a full flight envelope.It is based on the combination of genetic algorithm and neural network technology which could make full use of the global convergence property and rapid local searching capability separately.The topology structure and weights(including threshold) of neural network could be directly optimized through adaptive probability genetic operator and local optimizing operator,thereby overcoming the disadvantage of hidden layer nodes of neural network requiring empirical effort and the sensitivity of neural network to its initial weights(including threshold).Then,neural network was applied to "fine-tune" the above-specified weights(including threshold).Finally,the thrust estimator for the full flight envelope was designed with the algorithm.The experimental results show that the thrust estimator has good estimation accuracy and generalization ability.