试车数据驱动的涡轴发动机起动过程模型辨识

针对传统的采用解析法建立涡轴发动机起动过程模型复杂的问题,提出了一种基于变步长萤火虫算法优化的有外部输入的非线性自回归网络(CSFA-NARX)的涡轴发动机起动过程模型辨识方法。以涡轴发动机起动过程试车试验数据为数据样本,利用CSFA-NARX网络模型辨识得到涡轴发动机起动过程模型,并采用留一交叉验证方法对辨识模型的性能进行验证。结果表明:得到的辨识模型输出参数,如燃气发生器转速n_g、输出轴转速n_r和涡轮后温度T₄都较好地逼近了试车实测数据,各参数验证样本最大相对误差平均值分别为0.90%、1.51%、和2.01%;在相同训练与验证样本情况下,得到的辨识模型精度优于采用萤火虫算法优化的NARX网络(FA-NARX)、NARX网络和变步长萤火虫算法优化的BP网络(CSFA-BP)模型精度。

Model Identification of Turboshaft Engine's Starting Process Driven by Test Data

Aiming at the complex problem of establishing a turboshaft engine starting process model by traditional analytical methods,a nonlinear autoregressive network model method for turboshaft engine starting process based on the change of step of firefly algorithm optimization with external input was proposed.Taking the test data of the turboshaft engine starting process as the data sample,the model of the turboshaft engine starting process was obtained by using the CSFA-NARX network model identification,and the performance of the identification model was verified by the retention cross-validation method.The results show that the obtained identification model output parameters,such as gas generator speed n_g,output shaft speed n_r,and post-turbine temperature T₄ all well approximate the test data,the n_g,n_r,T₄ average maximum relative error between the training and the verification samples is respectively 0.90%,1.51%,and 2.01%.In the case of the same training and verification samples,the accuracy of identification model is better than that of the NARX network optimized by firefly algorithm(FA-NARX),NARX network and BP network optimized by change of step of firefly algorithm(CSFA-BP).