基于径向基神经网络的叶轮轴面投影图优化
发布时间:2018-08-11 17:53
【摘要】:为了提高余热排出泵的效率,采用拉丁超立方试验设计方法对叶轮轴面投影图上的前盖板圆弧半径、后盖板圆弧半径、前盖板倾角和后盖板倾角4个几何变量进行35组叶轮方案设计,应用ANSYS CFX 14.5软件对余热排出泵进行定常数值模拟,得到设计工况下的效率,应用径向基神经网络建立效率与轴面投影图的4个几何变量之间的近似模型,最后采用遗传算法对近似模型进行极值寻优,获得最优的轴面投影图几何参数组合。研究结果表明:对比原始泵的数值模拟性能曲线和试验外特性能曲线,两者吻合较好;径向基神经网络能较好地预测泵设计点效率;优化的轴面投影图使得余热排出泵的水力效率提高了6.18个百分点,改善了叶轮内流场特性。因此,叶轮轴面投影图的优化设计方法是可行的。
[Abstract]:In order to improve the efficiency of waste heat pump, a Latin hypercube experimental design method was used to design 35 impellers with four geometric variables, i.e. the arc radius of the front cover plate, the arc radius of the rear cover plate, the inclination angle of the front cover plate and the inclination angle of the rear cover plate. The constant value of the waste heat pump was simulated by ANSYS CFX 14.5 software. An approximate model between the efficiency and the four geometric variables of the axial projection graph is established by using the radial basis function neural network. Finally, the optimal geometric parameters of the axial projection graph are obtained by optimizing the approximate model with the genetic algorithm. The radial basis function neural network (RBF-NN) can predict the design point efficiency of the pump, and the optimized axial projection diagram can improve the hydraulic efficiency of the waste heat pump by 6.18 percentage points and improve the flow field characteristics of the impeller. Therefore, the optimization design method of the axial projection diagram of the impeller is feasible.
【作者单位】: 江苏大学国家水泵及系统工程技术研究中心;宜兴优纳特机械有限公司;
【基金】:“十二五”国家科技支撑计划资助项目(2011BAF14B04) 国家自然科学基金资助项目(51349004) 江苏省自然科学基金青年基金资助项目(BK20140554) 中国博士后科学基金面上资助项目(2014M560402) 江苏省博士后科研资助项目(1401069B) 江苏省普通高校研究生科研创新计划资助项目(KYLX_1042) 江苏省高校优势学科建设工程资助项目(PAPD)
【分类号】:TM623
[Abstract]:In order to improve the efficiency of waste heat pump, a Latin hypercube experimental design method was used to design 35 impellers with four geometric variables, i.e. the arc radius of the front cover plate, the arc radius of the rear cover plate, the inclination angle of the front cover plate and the inclination angle of the rear cover plate. The constant value of the waste heat pump was simulated by ANSYS CFX 14.5 software. An approximate model between the efficiency and the four geometric variables of the axial projection graph is established by using the radial basis function neural network. Finally, the optimal geometric parameters of the axial projection graph are obtained by optimizing the approximate model with the genetic algorithm. The radial basis function neural network (RBF-NN) can predict the design point efficiency of the pump, and the optimized axial projection diagram can improve the hydraulic efficiency of the waste heat pump by 6.18 percentage points and improve the flow field characteristics of the impeller. Therefore, the optimization design method of the axial projection diagram of the impeller is feasible.
【作者单位】: 江苏大学国家水泵及系统工程技术研究中心;宜兴优纳特机械有限公司;
【基金】:“十二五”国家科技支撑计划资助项目(2011BAF14B04) 国家自然科学基金资助项目(51349004) 江苏省自然科学基金青年基金资助项目(BK20140554) 中国博士后科学基金面上资助项目(2014M560402) 江苏省博士后科研资助项目(1401069B) 江苏省普通高校研究生科研创新计划资助项目(KYLX_1042) 江苏省高校优势学科建设工程资助项目(PAPD)
【分类号】:TM623
【参考文献】
相关期刊论文 前10条
1 杨军虎;张云周;孟瑞锋;王s,
本文编号:2177776
本文链接:https://www.wllwen.com/kejilunwen/dianlilw/2177776.html