摘要
多级涡轮三维气动优化设计由于计算量大、计算时间长、变量样本空间过于庞大,在实践中往往设计周期长,且难以有效实现。随着计算机硬件和计算软件的发展,计算能力已经大为改善,多种设计方法亦实现了有效融合。大力开展多级涡轮三维气动优化设计研究,将传统设计方法与现代自动优化设计方法相结合是解决前述困难,实现多级涡轮优化设计的一个有效途径。文中分析了将准三维设计和多级局部优化联合实现多级涡轮三维设计的可行性,给出了一个多级涡轮气动优化设计流程。准三维设计主要是S2流面正问题计算,通过准三维设计进行初步设计,初步提高性能,确定总体参数,为下一步的优化设计打下基础。然后采用多级局部优化设计,多级局部优化过程使用Numeca/design 3D软件,优化联合采用人工神经网络和遗传算法,通过提高局部性能来提高总体性能。流场计算采用全三维粘性流N-S方程求解,并以一个3级涡轮和一个4级涡轮为例,说明此方法的可行性。
Due to the massive computation load and time as well as an excessively huge variable-sample database space specific to the three-dimensional aerodynamic optimization design of a multi-stage turbine,a long design cycle often results,which is difficult to cope with effectively in practice.With the development of computer software and hardware the computation ability of computers has seen a dramatic improvement.As a result,an effective integration of varied design methods has been implemented.A vigorous development of the three-dimensional aerodynamic optimization-design study of a multi-stage turbine,which combines a traditional design method with that of a modern automatic optimization design,represents an effective approach for overcoming the above-mentioned difficulties and realizing an optimization design of the turbine in question.The feasibility for combining a quasi-three-dimensional design with the multi-stage local optimization to realize a three-dimensional design of the turbine was analyzed with the aerodynamic optimization design process of the turbine being given.The quasi-three-dimensional design mainly involves a direct problem computation of stream surface S2.Based on the design in question,a preliminary design was performed for improving performance and determining the overall parameters,thus setting the stage for a further optimization design.Then,by employing a multi-stage local optimized design and process,Numecadesign 3D software was used.By an optimized joint employment of an artificial neural network and a genetic algorithm,the general performance can be enhanced by way of an increase in localized performance.The flow field thus involved was calculated by seeking a solution for the full three-dimensional viscous flow N-S equation.Moreover,the authors have verified the feasibility of the method under discussion with a three-stage turbine and a four-stage one serving as examples.
出处
《热能动力工程》
EI
CAS
CSCD
北大核心
2008年第1期11-15,共5页
Journal of Engineering for Thermal Energy and Power
关键词
叶轮机械
多级涡轮
气动优化设计
准三维设计
设计流程
遗传算法
人工神经网络
turbo-machinery,multi-stage turbine,aerodynamic optimization design,quasi three-dimensional design,design flow path,genetic algorithm,artificial neural network
