Study On Geometry/Flow Parameters Coupling Data Mining And Optimization Of High-Speed Centrifugal Compressors

Modern high-performance centrifugal compressor seeks for the design requirements of high flow rate,high efficiency,high pressure ratio and wide stable operating range.Due to the complexity of the compressor geometry,its internal flow field presents complex flow characteristics such as inverse pressure gradient flow,boundary layer flow,secondary vortex flow,separation flow,tip leakage flow,shock wave effect,etc.,which are the bottlenecks that restrict the multi-objective aerodynamic optimization of transonic centrifugal compressor.Traditional optimization methods are difficult to achieve an all-round improvement of flow characteristics and aerodynamic performance since the physical flow mechanism in the compressor is not sufficiently excavated.In order to solve the problem of multi-objective aerodynamic optimization for high-speed centrifugal compressor,this thesis combines the theory-based data mining,surrogate-model-based data mining and the adaptive sampling hybrid optimization algorithm,and then proposes a multi-objective aerodynamic optimization method for the compressor over the whole operating condition based on geometry/flow parameters coupling data mining.Firstly,the theory-based geometry/flow parameters coupling data mining method is proposed.The universal coupling relationship between the geometry and the flow parameters of centrifugal compressor is derived and established.The key geometry,flow and aerodynamic parameters of the compressor are expressed uniformly as a series of functions of the same dimensionless parameters.Some dimensionless diagrams are drawn according to the universal relationship,revealing the general non-linear coupling and restriction laws between compressor geometry,flow and aerodynamic parameters,and theoretically exploring the feasible region and optimization limits of compressor design space.In order to study the aerodynamic characteristics of off-design conditions,a mathematical model that couples off-design conditions and design condition flow parameters is further established.Therefore,the flow characteristics and aerodynamic performance of the off-design conditions can be explored rapidly according to the parameters under the design condition.This model also theoretically proves that the coupling law of flow parameters under different operating conditions has geometric invariance property.The above models collectively constitute the coupled theoretical model of compressor geometry,flow and aerodynamic parameters under all operating conditions,which provides theoretical support for data mining of compressor internal flow from the physical mechanism level.Secondly,the surrogate-model-based geometry/flow parameters coupling data mining method is established.A new adaptive sampling algorithm is proposed,and an adaptive sampling surrogate model is developed by combining the adaptive sampling with the Kriging surrogate model.The mathematical cases have proven the highefficiency feature of the adaptive sampling surrogate model.Based on surrogate model,the compressor internal flow data mining method is established,in which the data mining algorithms include self-organizing map and analysis of variance.Compared with theory-based data mining,surrogate-model-based data mining can extract the coupling relationship between the local geometry parameters and the global flow as well as aerodynamic parameters.Then,a multi-objective aerodynamic optimization method based on compressor geometry/flow parameters coupling data mining is developed.A gradient mutation hybrid optimization algorithm is proposed by coupling the gradient mutation with the multi-objective genetic algorithm.Some high-dimensional and high-nonlinear mathematical cases are used to verify the global performance and the search efficiency of the optimization algorithm.The sensitivity analysis of key parameters of the algorithm is carried out,showing that it also has good robustness.By coupling the adaptive sampling surrogate model with the gradient mutation hybrid optimization algorithm,an adaptive sampling hybrid optimization algorithm for multi-objective optimization of compressors is deveploed.By combining the theory-based and the surrogate-model-based geometry/flow parameters coupling data mining methods with the adaptive sampling hybrid optimization algorithm,a multi-objective aerodynamic optimization method based on compressor internal flow data mining is established.Finally,the whole speed line multi-objective aerodynamic optimization based on internal flow data mining is implemented for the transonic turbocharger compressor and the Krain impeller.By carrying out theory-based data mining,the coupling and restriction laws between compressor geometry,flow and aerodynamic parameters are obtained.The design variables that play a key role in the flow characteristics and aerodynamic performance are identified.The objective functions that reflect more comprehensive improvements of the flow characteristics and aerodynamic performance are extracted.As a result,the initial design space is obtained,and the physical optimization problem is abstracted into a mathematical optimization model.By conducting surrogate-model-based data mining,the nonlinear coupling and restriction relationships between the design variables and the objective functions are clarified,and the quantitative contributions of the design variables to the objective functions are obtained.Based on this,the design space is refined and the optimization problem is simplified.The refined design space is then optimized using the adaptive sampling hybrid optimization algorithm,and the Pareto optimal solution set is obtained.From the qualitative and quantitative points of view,it shows that the optimization method proposed in this study is superior to the current dominant algorithms in terms of optimization efficiency,global performance,and robustness,among them the optimization efficiency can be boosted by 2-10 times.Meanwhile,the coupling variation laws of geometry,flow and aerodynamic parameters is clarified.Also,the fluid dynamic mechanism behind the improvement of aerodynamic performance is revealed.The multi-objective aerodynamic optimization method proposed in this thesis for the compressor over the whole operating condition based on geometry/flow parameters coupling data mining,provides an efficient new method for the optimization of modern high-performance compressors,which may help the development and localization of design technology for advanced aero-engine compressor,turbocharger compressor,and energy-power industrial compressor.