摘要：The ultra-low specific speed centrifugal blower is widely used in energy industries due to its features such as low flow rate,high pressure and low manufacturing cost. However,the width-to-diameter ratio of the above blower becomes relatively small to satisfy the needed operation condition and its performances are considerably degraded as a result of relatively high leakage,disc friction and passage friction loss consequently. The purpose of this paper is to improve its performance through the optimization design of the blade’s profile properly. Based on artificial neural networks (ANN) and hierarchical fair competition genetic algorithms with dynamic niche (HFCDN-GAs),the optimization design approach is established. By conjoining Bezier parameterization and FINE/TURBO solver,the optimized blade is designed by adjusting the profile gradually. An industrial ultra-low specific speed centrifugal blower with parallel hub and shroud has been selected as a reference case for optimization design. The performance investigations of the centrifugal blowers with different types of blades are conducted. The conclusions of the performance improvement of the optimized blade provide positive evidences in the application of the optimization design of the above blower blade.

摘要：运用计算流体力学(Computational fluid dynamics,CFD)技术数值模拟带径向进气室的某石化离心压缩机级内部的流动情况,结果表明进气室内部的流场较为复杂,存在较大的旋涡结构,流动损失较大,并造成在叶轮进口处存在气流畸变,因此进气室的结构形状对整机的性能影响显著。为改善进气室内部的流动情况和提高级的性能,发展一种基于CFD技术的数值优化方法,并对该进气室的几何形状进行了优化设计。优化后的整机外特性整体得到了改善,在保证质量流量和总压比不低于初始值的约束条件下,设计工况点的整机等熵效率提高了1.26%。对比分析优化前后整机内部的速度、熵和总压等参数分布,发现优化模型内部的流动结构更为合理,流动损失更小。优化结果也验证了该优化设计方法的有效性。

摘要：在离心泵叶轮的基本外尺寸(叶轮内外半径、叶轮进出口宽度、叶片进出口安放角及叶片数)和设计转速相同的情况下,采用三次曲线对叶片进行造型。设计叶片包角分别为59°、75°和91°的三副离心叶轮C1、C2、C3,其中叶轮C1与传统的单圆弧叶型非常接近。三副叶轮的同台试验结果显示,叶轮C2的最高效率比叶轮C1、C3的效率高1.28%、1.43%。采用数值模拟得到设计工况下三副叶轮内的相对速度场和各流道内的载荷分布,C1叶轮内有明显的回流区,且各流道内的载荷存在较大差异。研究表明:在叶轮外尺寸相同的情况下对叶片造型设计时,存在最佳的叶片包角,单圆弧叶型不是最佳叶型。

摘要：采用一种接近美国NIST标准物性的实际物性模块,研究了某化工行业离心压缩机高压级的流动压缩性能,重点关注了模型级的进口参数如压力、温度、流动角、体积流量对性能的影响,并与理想物性下计算得出的性能作了相应比较,指出了临界点附近物性参数的敏感性,为CO2高压压缩机设计与性能分析提供了依据。

摘要：精准识别建成环境能够为城市空间结构与城市交通互动关系研究提供科学依据,为合理配置城市空间资源提供数据基础。研究了一种基于语义识别的深度学习模型,在融合兴趣点与兴趣面基本分析单元的基础上,提取兴趣面内兴趣点的高维语义特征向量,通过聚类分析识别得到功能用地类型。结果表明该深度学习模型识别精度的准确率达到80%,模型可以为建成环境识别提供新的方法思路。