基于图像分割的蒸发器二次侧管板清洗喷嘴自适应定位

Adaptive Positioning of Tubeplate Cleaning Nozzle on Secondary Side ofEvaporator Based on Image Segmentation

针对在核电蒸发器二次侧传热管管板清洗过程中清洗喷嘴定位困难的问题,提出了一种基于图像分割的清洗喷嘴自适应定位方法。清洗喷嘴与摄像头以固定距离偏置安装,通过设计一个轻量级的全卷积神经网络语义分割模型对摄像头采集的传热管图像进行实时分割,得到完整的传热管轮廓;然后对轮廓进行分析,获得传热管中心像素位置和位置偏差;最后完成位置偏差补偿,实现清洗喷嘴的自适应定位。经过与多种网络模型的类比实验可知,该网络模型计算效率高以及分割性能优异,可快速准确实现清洗喷嘴在蒸发器内的自适应定位。

Aiming at the difficulty of locating the cleaning nozzle in the cleaning process of the secondary side heat transfer tube plate of nuclear evaporator,an adaptive locating method of the cleaning nozzle based on image segmentation was proposed.The cleaning nozzle and camera are mounted with fixed distance bias.A lightweight full convolutional neural network semantic segmentation model was designed to segment the heat transfer tube image from the camera in real time,and the complete heat transfer tube contour was obtained.Then the contour is analyzed for the pipe center pixel location and position deviation,the final location deviation compensation is completed to realize adaptive localization of cleaning nozzle.The simulation experiments with many network models show that this network model has high computational efficiency and excellent segmentation performance,and can quickly and accurately realize the adaptive positioning of the cleaning nozzle in the evaporator.