通过优化汽轮机叶片的安装顺序,来减少安装后的残余不平衡量。对此提出一种阈值式迭代局部搜索(threshold iterative local search,TILS)算法,该算法在迭代局部搜索(iterative local search,ILS)算法基础上,采用阈值限定扰动与随机扰动...通过优化汽轮机叶片的安装顺序,来减少安装后的残余不平衡量。对此提出一种阈值式迭代局部搜索(threshold iterative local search,TILS)算法,该算法在迭代局部搜索(iterative local search,ILS)算法基础上,采用阈值限定扰动与随机扰动相结合的方法来跳出局部最优解,减少了平均到达局部最优解所需的迭代步数。实验证明,该方法可以在短时间内找到一个近似最优叶片排序组合,相对于ILS算法,搜索效率提高了20%以上。计算得到的合成质径积的近似最优解,相对于现有分组排序法、遗传算法、云自适应遗传算法(CAGA)等方法,分别减小到其最优解的0.33%~31%,且计算时间也大幅度减小。展开更多
Venipuncture robots have superior perception and stability to humans and are expected to replace manual venipuncture.However, their use is greatly restricted because they cannot make decisions regarding the puncture s...Venipuncture robots have superior perception and stability to humans and are expected to replace manual venipuncture.However, their use is greatly restricted because they cannot make decisions regarding the puncture sites. Thus, this study presents a multi-information fusion method for determining puncture sites for venipuncture robots to improve their autonomy in the case of limited resources. Here, numerous images have been gathered and processed to establish an image dataset of human forearms for training the U-Net with the soft attention mechanism(SAU-Net) for vein segmentation. Then, the veins are segmented from the images, feature information is extracted based on near-infrared vision, and a multiobjective optimization model for puncture site decision is provided by considering the depth, diameter, curvature, and length of the vein to determine the optimal puncture site. Experiments demonstrate that the method achieves a segmentation accuracy of 91.2% and a vein extraction rate of 86.7%while achieving the Pareto solution set(average time: 1.458 s) and optimal results for each vessel. Finally, a near-infrared camera is applied to the venipuncture robot to segment veins and determine puncture sites in real time, with the results transmitted back to the robot for an attitude adjustment. Consequently, this method can enhance the autonomy of venipuncture robots if implemented dramatically.展开更多
文摘通过优化汽轮机叶片的安装顺序,来减少安装后的残余不平衡量。对此提出一种阈值式迭代局部搜索(threshold iterative local search,TILS)算法,该算法在迭代局部搜索(iterative local search,ILS)算法基础上,采用阈值限定扰动与随机扰动相结合的方法来跳出局部最优解,减少了平均到达局部最优解所需的迭代步数。实验证明,该方法可以在短时间内找到一个近似最优叶片排序组合,相对于ILS算法,搜索效率提高了20%以上。计算得到的合成质径积的近似最优解,相对于现有分组排序法、遗传算法、云自适应遗传算法(CAGA)等方法,分别减小到其最优解的0.33%~31%,且计算时间也大幅度减小。
基金supported by the National Natural Science Foundation of China (Grant No. U1813209)Self-Planned Task of State Key Laboratory of Robotics and System (Harbin Institute of Technology)(Grant No.SKLRS202112B)。
文摘Venipuncture robots have superior perception and stability to humans and are expected to replace manual venipuncture.However, their use is greatly restricted because they cannot make decisions regarding the puncture sites. Thus, this study presents a multi-information fusion method for determining puncture sites for venipuncture robots to improve their autonomy in the case of limited resources. Here, numerous images have been gathered and processed to establish an image dataset of human forearms for training the U-Net with the soft attention mechanism(SAU-Net) for vein segmentation. Then, the veins are segmented from the images, feature information is extracted based on near-infrared vision, and a multiobjective optimization model for puncture site decision is provided by considering the depth, diameter, curvature, and length of the vein to determine the optimal puncture site. Experiments demonstrate that the method achieves a segmentation accuracy of 91.2% and a vein extraction rate of 86.7%while achieving the Pareto solution set(average time: 1.458 s) and optimal results for each vessel. Finally, a near-infrared camera is applied to the venipuncture robot to segment veins and determine puncture sites in real time, with the results transmitted back to the robot for an attitude adjustment. Consequently, this method can enhance the autonomy of venipuncture robots if implemented dramatically.
