基于边界涡量动力学理论的离心泵叶轮水力优化研究

发布时间：2019-06-19 16:24

【摘要】：离心泵在国民经济生产和生活中有广泛的应用前景,但其叶轮水力设计理论和方法陈旧,仍存在诸多问题,进行离心泵叶轮的优化设计研究意义重大。本文立足于江苏省研究生培养创新工程项目“基于边界涡量流的离心泵叶轮内流诊断和水力优化研究”,运用边界涡量动力学理论对离心泵进行内流诊断分析,研究内流参数边界涡量流(BVF)与离心泵水力性能参数之间的内在联系和规律,以内流参数为目标:一方面研究BP网络和径向基函数网络(RBF)两种不同人工神经网络(ANN)对离心泵叶轮内流参数预测可靠性的影响;另一方面研究遗传算法(GA)和粒子群算法(PSO)两种智能优化算法对离心泵叶轮优化设计结果的影响,确定应用于离心泵叶轮水力性能的多目标遗传算法的优化策略。本文的主要研究内容及结论如下:1.BVF分布和离心泵内流场及外特性的关系。首先以一台流道式离心泵为研究对象,采用数值模拟方法获得了模型泵叶轮内部流场的局部流动细节,重点分析了叶片压力面和吸力面附近的不良流动状况;结合边界涡量动力学理论,分析了叶片压力面和吸力面上的BVF、摩擦力线以及涡线分布规律,揭示了BVF分布与叶片表面流动分离、漩涡产生与耗散以及水力性能之间的内在联系。研究表明:叶轮内表面的BVF峰值和均值越低,BVF分布均匀指数越高,叶轮内部流动状况越好,流动分离得到抑制,流体对叶轮的做功效果越好,叶轮的扬程和效率更高。2.人工神经网络在离心泵内流参数预测中的应用研究。基于MATLAB平台的二次开发功能,探究隐含层数、径向基函数的扩展速度分别对BP神经网络和RBF神经网络性能预测精度的影响;然后以预测值和CFD计算值的误差范数作为评判两种神经网络预测性能优劣的标准,选取适用于离心泵叶轮内流参数预测的最优人工神经网络。研究表明:当隐含层取18时,BP神经网络预测误差最小,神经网络结构最优;当径向基函数扩展速度Spread取0.3时,RBF网络预测误差最小,神经网络结构最优。将两者预测结果对比分析后发现:RBF神经网络的预测误差更小,程序运行时间更短,运行稳定性更高。3.基于智能优化算法的离心泵叶轮水力性能优化方法研究。研究并建立了以内流场参数BVF为优化目标的离心泵叶轮水力优化问题的数学模型,确定了对应的约束条件和优化变量;确定了优化目标及优化变量的取值范围和编、解码方案;对比了GA和PSO在离心泵叶轮水力优化问题中的适用性,对算法进行改进获得了适用于离心泵叶轮水力优化的最优算法,并制定了高效可靠的寻优策略。研究表明:以BVF为优化目标,应用GA和PSO对离心泵叶轮进行单目标优化,效果良好,而且GA的优化结果叶轮1,无论从内流场分布还是外特性参数,均优于基于PSO得到的结果叶轮2,但以BVF峰值为目标,无法保证叶轮表面BVF均匀性指数最优;以BVF峰值和均匀指数为目标,应用多目标遗传算法对离心泵叶轮进行求解,优化后的叶轮3的内部流动得到改善,叶轮表面BVF均匀指数也得到提高,且优于叶轮2的相关参数,叶轮的扬程和效率都相应提升。结合RBF神经网络的多目标遗传算法,其全局寻优能力强,程序运行时间短,最优结果的精度高。
[Abstract]:The centrifugal pump has a wide application prospect in the production and life of the national economy, but the hydraulic design theory and method of the impeller are old, there are still many problems, and the optimization design of the centrifugal pump impeller is of great significance. In this paper, based on the "Study on the internal flow diagnosis and hydraulic optimization of centrifugal pump impeller based on boundary vortex flow" of Jiangsu Post-graduate Training Innovation Project, the internal relationship and the law of the internal flow parameter boundary vorticity flow (BVF) and the hydraulic performance parameters of the centrifugal pump are studied by using the boundary vorticity dynamics theory to analyze the internal flow of the centrifugal pump. The internal flow parameters are the target: on the one hand, the influence of two different artificial neural networks (ANN) of BP network and radial basis function network (ANN) on the prediction reliability of the internal flow parameters of the impeller of the centrifugal pump is studied. On the other hand, the influence of two intelligent optimization algorithms of genetic algorithm (GA) and particle swarm optimization (PSO) on the optimization design of the impeller of centrifugal pump is studied, and the optimization strategy of the multi-objective genetic algorithm applied to the hydraulic performance of the impeller of the centrifugal pump is determined. The main contents and conclusions of this paper are as follows:1. The relationship between the BVF distribution and the flow field and the external characteristics in the centrifugal pump. In this paper, a flow channel type centrifugal pump is used as the research object, and the numerical simulation method is adopted to obtain the local flow detail of the flow field inside the model pump impeller, and the adverse flow conditions near the pressure surface of the blade and the suction surface are emphatically analyzed, and the boundary vortex quantity dynamics theory is combined. The relationship between the distribution of BVF and the flow separation of the blade surface, the generation and dissipation of the vortex and the hydraulic performance of the BVF, the friction line and the distribution of the vortex line on the pressure surface and the suction surface of the blade are analyzed. The results show that the lower the BVF peak and the mean value of the inner surface of the impeller, the higher the distribution of BVF, the better the internal flow of the impeller, the better the flow separation, the better the effect of the fluid on the impeller, and the higher the lift and the efficiency of the impeller. The application of artificial neural network in the prediction of internal flow parameters of centrifugal pump. based on the secondary development function of the MATLAB platform, the influence of the hidden layer number and the expansion speed of the radial basis function on the performance prediction precision of the BP neural network and the RBF neural network is explored, and then the error norm of the predicted value and the CFD calculation value is used as a standard for judging the performance of the two neural networks to predict the performance, The optimal artificial neural network for the prediction of the internal flow parameters of the impeller of the centrifugal pump is selected. The results show that when the implicit layer is 18, the prediction error of the BP neural network is the least, the structure of the neural network is optimal, and when the spreading speed of the radial basis function is 0.3, the prediction error of the RBF network is the least, and the structure of the neural network is optimal. It is found that the prediction error of the RBF neural network is smaller, the program running time is shorter, and the running stability is higher. Research on hydraulic performance optimization of centrifugal pump impeller based on intelligent optimization algorithm. The mathematical model of the hydraulic optimization problem of the centrifugal pump impeller with the internal flow field parameter BVF as the optimization target is studied and the corresponding constraint condition and the optimization variable are determined, and the value range and the coding and decoding scheme of the optimization target and the optimization variable are determined. The applicability of GA and PSO in the hydraulic optimization of centrifugal pump impeller is compared, and the algorithm is improved to obtain the optimal algorithm suitable for the hydraulic optimization of the impeller of the centrifugal pump, and an efficient and reliable optimization strategy is developed. The results show that, with BVF as the optimization target, the single-objective optimization of the centrifugal pump impeller with GA and PSO is optimized, the effect is good, and the optimization result of GA is superior to the result impeller 2 based on the particle swarm optimization, whether the internal flow field distribution or the external characteristic parameter, but the BVF peak is the target, in that invention, the uniform index of the BVF of the surface of the impeller can not be guaranteed, the BVF peak value and the uniform index are used as the target, the multi-objective genetic algorithm is applied to solve the centrifugal pump impeller, the internal flow of the optimized impeller 3 is improved, and the uniform index of the BVF on the surface of the impeller is also improved, And the lift and the efficiency of the impeller are correspondingly improved. Combined with the multi-objective genetic algorithm of the RBF neural network, the global optimization capability is strong, the program running time is short, and the optimal result is high in precision.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH311

【参考文献】