2015, (6): 89-95.
doi: 10.11729/syltlx20140074
Abstract:
To improve the flow field quality of the 2 .4m transonic wind tunnel in CARDC,it is necessary to build an efficient flow field control model for controller design verification plat-form of the wind tunnel.The data driven modeling method becomes an important alternative because accurate aerodynamic models are difficult to establish while a lot of experimental data have been accumulated in the long running of the 2.4m wind tunnel.As to the hardware,a flow field control simulation system is established based on the memory reflective technology to get the measurement data in situ.The data driven modeling method is adopted based on the system identification theory,which regards the system as a “black box”and establishes the relationship between the inputs and the outputs by algorithms using the data obtained from the actual meas-urements.Non-linear Auto-Regressive Moving Average Model with Exogenous Inputs (NAR-MAX)model is selected as the wind tunnel’s data model.Mutual information algorithm,curve fitting algorithm and false nearest neighbors algorithm are respectively used to identify sampling interval,time delay and order.The effects of least square linear regression,BP neural network and supporting vector machine (LS SVM)on model fitting are compared,and LS SVM is cho-sen.To improve the precision of simulation,the running process of wind tunnel can be divided into three stages:charging stage,start-up stage and adjusting stage according to respective run-ning features and then each stage is modeled separatedly.Due to the MIMO features,large time delay and high order characteristics of the wind tunnel,a data compression method based on in-formation entropy is adopted which scales down the model.Finally,the model for the whole run-ning process is realized by fusing the sub-models of the three stages.The total pressure at stag-nation and the static pressure at test section are calculated via the model and the Mach number can be obtained by the relationship between the pressure and Mach number.The simulation re-sults show that within the scope of the test conditions,the total pressure prediction precision of the model reaches 0.1%,and the Mach number prediction precision generally reaches 0.001.The goal has been attained.In conclusion,an intermittent transonic wind tunnel flow field control model is systematically established,which will play a significant role in the future design of the controller based on the modern control theory.
