基于径向基函数神经网络的多级离心压缩机混合模型

大型离心压缩机作为多影响因素和强非线性的复杂系统,其性能的准确预测难以实现.针对这一问题,结合径向基函数(RBF)神经网络,本文建立了多级离心压缩机性能预测的混合模型.首先基于热力学第一定律和压缩机能量损失机理建立了多级离心压缩机性能预测的机理模型.该模型无需任何实验确定的性能曲线,完全由压缩机的几何结构参数预测出压缩机在设计工况和非设计工况下的性能.然后利用RBF神经网络修正机理模型的误差,并通过对RBF神经网络的不断更新,进一步提高了模型的预测精度和适用性.将所建立的混合模型应用于实际的离心压缩机,结果表明该方法具有良好的预测性能.

Hybrid model for multi-stage centrifugal compressor based on radial basis function neural network

The large centrifugal compressor is a complex system with many factors and strong nonlinearities; the performance of which cannot be predicted accurately. To deal with this problem, we propose a hybrid model for predicting the performance of a multistage centrifugal compressor by employing the radial basis function (RBF) neural network. First, according to the structural parameters of the compressor instead of the experimental characteristic, we deduce a theoretical prediction model based on the first law of thermodynamics and the energy loss mechanism. This model is used to predict the design performance and the off-design performance of the compressor. Then, a RBF neural network, which is updated in time, is applied to the theoretical model to form a hybrid model, in which the error of the theoretical model is continuously corrected to raise its accuracy in the process of performance prediction. This hybrid model has been used to predict the performance of practical multistage centrifugal compressors in industrial applications; the results of performance prediction are satisfactory