基于RBF神经网络的燃气轮机特性计算  

Gas Turbine Performance Calculation based on RBF Neural Networks

作　　者：贾小权[1] 张仁兴[1] 贺星[1] 房友龙[1] 

机构地区：[1]海军工程大学船舶与动力学院,武汉430033

出　　处：《燃气轮机技术》2010年第4期49-53,共5页Gas Turbine Technology

摘　　要：燃气轮机各部件的特性可通过实验获得,但耗费昂贵;且基于知识产权的保护,厂家一般也不会提供完整的燃气轮机特性数据,这对燃气轮机模型建立的精度会造成较大影响。本文根据生产厂家提供的燃气轮机的部分性能曲线,利用人工神经网络的高度非线性映射、自学习和泛化等功能,采用RBF神经网络训练出所需的特性曲线。结果表明:RBF神经网络无论是在训练时间还是训练精度上,均取得了较好的效果。所得的燃气轮机部件全工况特性为燃气轮机仿真建模和性能分析奠定基础。The components performance of gas turbine can be obtained by experiment,but it costs too much.Considering protection of the intellectual property rights,the manufactories don＇t provide the integrated curve of gas turbine.It will influence on precision of the model of gas turbine.Based on the segment curves of gas turbine which provided by the manufactory,some special data of gas turbine components performance were obtained by utilizing the radial basis function（RBF） neural networks which with the capability of multi-nonlinear,self-educated and generalization function in this paper.The results show that the RBF neural network obtain better effect of the train time and the precision of the training goal.The obtained integrated performance of the gas turbine is the groundwork for establishing the gas turbine simulation model and performance analysis.

关 键 词：燃气轮机 特性 计算 径向基神经网络 

分 类 号：TK471[动力工程及工程热物理—动力机械及工程]