CN-122021352-A - Wind tunnel control driving parameter intelligent generation method based on XGBoost machine learning method

CN122021352ACN 122021352 ACN122021352 ACN 122021352ACN-122021352-A

Abstract

The application belongs to the technical field of aerospace vehicle ground simulation tests and machine learning, and discloses an intelligent generation method of wind tunnel control driving parameters based on a XGBoost machine learning method, which comprises the steps of data screening and processing, screening input/output variables participating in modeling from wind tunnel historical test operation data, extracting the control driving parameters after a flow field is stable, and forming a data set; the method comprises the steps of feature analysis, input/output variable determination of a model according to the control relation between the position of a wind tunnel flow field adjusting mechanism and flow field pressure parameters and by combining a correlation analysis map, model construction, dividing a data set into a training set and a testing set, respectively establishing each driving parameter prediction model by adopting an XGBoost algorithm, iteratively developing parameter tuning, training, testing and verifying, parameter generation and verification, inputting current test conditions based on the trained XGBoost parameter prediction model, and intelligently generating and controlling driving parameter combination.

Inventors

JIA PENG
YUAN JIALING
MA YUANYE
YU WENSHAN
XU XIN
LIU GANG
LIU LONGBING
Su beichen
CAO XIN
ZHANG DONG
HUANG CHEN

Assignees

中国空气动力研究与发展中心高速空气动力研究所

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (9)

1. The wind tunnel control driving parameter intelligent generation method based on XGBoost machine learning method is characterized by comprising the following steps: s1, data screening and processing, namely screening input/output variables participating in modeling from wind tunnel historical test operation data, and extracting control driving parameters after flow field stabilization to form a data set; S2, characteristic analysis, namely determining input/output variables of a model according to the control relation between the position of a wind tunnel flow field regulating mechanism and flow field pressure parameters and combining a correlation analysis map; S3, model construction, which comprises the steps of dividing the data set obtained in the step S1 into a training set and a testing set, respectively establishing each driving parameter prediction model by adopting XGBoost algorithm, and iteratively carrying out parameter tuning, training, testing and verification; And S4, parameter generation and verification, namely inputting current test conditions based on a trained XGBoost parameter prediction model, and intelligently generating and controlling driving parameter combinations.
2. The method for intelligently generating wind tunnel control driving parameters according to claim 1, wherein the data screened in the step S1 comprises: time sequence data, main leading pressure Auxiliary pressure Main adjustment displacement Parking shift Displacement of gate finger Displacement of standing stream Displacement of main row Angle of attack alpha, total pressure Static pressure Reference Mach number Pressure of air source , Nominal mach number Atmospheric pressure Test section Type and model alpha = 0 ° of incoming flow blockage Data, incoming flow blocking degree Is the sum of the blockage degree of the model, the supporting mechanism and the corresponding accessory mechanism.
3. The method for intelligently generating wind tunnel control driving parameters according to claim 2, wherein in step S1, the process of forming the data set includes: At total pressure And reference Mach number After stabilization, extracting alpha from time sequence data, starting operation time, and adjusting the position of each flow field regulating mechanism and the value of flow field pressure parameter, and adjusting the flow field pressure parameter and the flow blockage degree Atmospheric pressure Nominal mach number A dataset is formed.
4. The intelligent wind tunnel control driving parameter generating method according to claim 3, wherein the control relation between the position of the wind tunnel flow field adjusting mechanism and the flow field pressure parameter is: Main pressure of leading By main adjustment of displacement Closed loop regulation of total pressure Displaced by main row Closed loop control and take over pressure Influence of auxiliary pressure By dwell displacement Closed loop control of static pressure By displacement of gate fingers Standing-flow displacement Individually or jointly controlled and subject to pilot pressure Influence; At the same time, the incoming flow blocking degree For static pressure Influence and total pressure And static pressure Determining a reference Mach number Is of a size of (a) and (b).
5. The method for intelligently generating wind tunnel control driving parameters according to claim 1, wherein the step S3 comprises the steps of establishing XGBoost output regression tasks, respectively establishing a prediction model for each control parameter to be intelligently generated, and respectively performing super-parameter tuning.
6. The method for intelligently generating wind tunnel control driving parameters according to claim 5, wherein in step S3, a bayesian optimization method is used for super-parameter tuning, root mean square error RMSE of a data set is used as an optimization criterion, super-parameter tuning is performed on each independent XGBoost model, and the models are trained on a training set by using optimal parameter combinations respectively.
7. The method for intelligently generating wind tunnel control driving parameters according to claim 6, wherein step S3 further comprises the step of using a test set to evaluate the performance of each model, and the performance evaluation process adopts the training set and the RMSE of the test set as evaluation indexes.
8. The intelligent generation method of wind tunnel control driving parameters according to claim 4, wherein in step S4, the known amount in the current test condition is total pressure Nominal mach number Pressure of air source Atmospheric pressure Degree of incoming flow blockage Test section ; The control parameter combination to be intelligently generated is that the main leading pressure Auxiliary pressure Main adjustment displacement Parking shift Displacement of gate finger Displacement of standing stream Displacement of main row 。
9. The method for intelligently generating wind tunnel control driving parameters according to claim 8, wherein the step S4 further comprises parameter verification, and based on the absolute error delta, the intelligently generated driving parameters, the manually given driving parameters and the closed loop steady state values in the time sequence data are compared and verified, so that feasibility of the model intelligent parameters is verified.

Description

Wind tunnel control driving parameter intelligent generation method based on XGBoost machine learning method Technical Field The application belongs to the technical field of aerospace vehicle ground simulation tests and machine learning, and particularly relates to an intelligent generation method for wind tunnel control driving parameters based on XGBoost machine learning method. Background The temporary impact transonic wind tunnel has high fidelity geometric simulation and high Reynolds number test capabilities, and plays an important role in aerodynamic force simulation test of the aerospace craft in China. As a temporary flushing transonic wind tunnel, the wind tunnel has the remarkable characteristics of multiple controlled objects, strong coupling of flow field parameters and rapid transient/steady switching. In order to effectively shorten the test time and improve the accuracy of flow field control, the control driving parameters are required to be accurately set in each test, and the initial preset values of the positions of flow field regulating mechanisms such as a main pressure regulating valve (hereinafter referred to as main regulation), a main exhaust valve (hereinafter referred to as main exhaust), a resident chamber flow valve/pressure regulating valve (hereinafter referred to as resident flow/resident regulation), a grid finger mechanism and the like are included, and the pressure parameters of a main ejector (hereinafter referred to as main guide) and a resident chamber ejector (hereinafter referred to as auxiliary guide) are included. The initial preset value of the position of the flow field regulating mechanism is accurately set, so that the transient pressurizing process of the wind tunnel is quickly cut into the steady-state regulating stage of the flow field, the steady-state regulating process of the flow field is greatly shortened, meanwhile, the accurate flow field pressure parameter setting ensures that the closed-loop control of the flow field is in the optimal regulating working condition, and a foundation is laid for the high-precision control of the flow field. Under normal conditions, the setting of the control driving parameters is finally determined by post personnel according to priori knowledge and manual inquiry historical driving number data, and the parameter setting method based on experience often has the following problems that 1, the driving of each working condition needs longer time for preparing the parameters for the first time, 2, the experience difference of the post personnel can cause larger control parameter deviation, and 3, the manual inquiry historical driving number data can be limited to the coverage range of the existing experience case, so that the change requirement of the new working condition can not be fully met. The control of the driving parameters can directly influence the length of the blowing time and even the success or failure of the blowing, and has important practical significance for establishing test flow field conditions and shortening the blowing time more quickly and saving energy for wind tunnels with huge energy consumption. Disclosure of Invention The application aims to overcome the problems in the prior art, and discloses an intelligent generation method of wind tunnel control driving parameters based on XGBoost machine learning method, which is used for automatically giving accurate control driving parameters. The aim of the application is achieved by the following technical scheme: A wind tunnel control driving parameter intelligent generation method based on XGBoost machine learning method comprises the following steps: s1, data screening and processing, namely screening input/output variables participating in modeling from wind tunnel historical test operation data, and extracting control driving parameters after flow field stabilization to form a data set; S2, characteristic analysis, namely determining input/output variables of a model according to the control relation between the position of a wind tunnel flow field regulating mechanism and flow field pressure parameters and combining a correlation analysis map; S3, model construction, which comprises the steps of dividing the data set obtained in the step S1 into a training set and a testing set, respectively establishing each driving parameter prediction model by adopting XGBoost algorithm, and iteratively carrying out parameter tuning, training, testing and verification; And S4, parameter generation and verification, namely inputting current test conditions based on a trained XGBoost parameter prediction model, and intelligently generating and controlling driving parameter combinations. According to a preferred embodiment, the data screened in step S1 comprises: time sequence data, main leading pressure Auxiliary pressureMain adjustment displacementParking shiftDisplacement of gate fingerDisplacement of standing streamDisplacement of main rowAngle of attack alph