基于机器视觉的冷凝器污垢特性参数在线测量方法研究

This project focuses on studying the online measurement of fouling characteristic parameters in condenser based on machine vision for the purpose of solving the practical problem of poor control effect caused by only relying on fouling heat transfer parameter to determine the control quantity. The main research points are as follows. Firstly, studying the image preprocessing methods to solve the problem of image quality degradation under the complex dynamic environment in condenser tubes; Secondly, studying real-time tracking method of the fouling particle trajectory to realize the online measurement of deposition rate and removal rate; In view of the particles overlap, occlusion and flocculation deformation in the process of tracking, studying edge correction algorithm based on image sequences, particle matching algorithm based on clustering prediction to achieve accurate tracking trajectory; Thirdly, studying feature extraction method of fouling particles and frame difference hierarchical memory method to realize the online recognition of fouling types, space position and component ratio of the various fouling types in the fouling compound body; Finally, studying elastic-plastic model equivalent method for fouling compound body, developing the strength parameter prediction model and studying model calibration, optimization algorithm to achieve accurate measurement of fouling strength parameters. The research findings of the project will provide theory evidence for the efficient fouling control and safe operation of condenser, it has great significance in energy saving, consumption reducing and economic benefits improving in electric power, chemical industry and machinery industry, and has a good prospect of application.

本项目旨在研究基于机器视觉的冷凝器污垢特性参数在线测量方法，解决生产现场只依靠污垢传热参数确定控制量而导致的污垢控制效果差的难题，主要研究内容为：研究冷凝器管内复杂环境下的图像预处理方法，解决成像质量劣化问题；研究污垢颗粒运动轨迹实时追踪方法，实现污垢沉积率、剥蚀率的在线测量，针对追踪过程中存在的颗粒重叠、被遮挡以及絮凝变形现象，研究基于序列图像的边缘修正补偿算法、基于聚类预估的颗粒匹配算法，实现运动轨迹的准确追踪；研究污垢颗粒的特征提取方法、帧间差分层叠记忆方法，确定污垢混合体结构，实现混合体内污垢类型、各类污垢空间位置以及成分比例的在线识别；研究污垢混合体的弹塑性模型等效方法，建立强度参数预估模型并研究模型校正、优化算法，实现污垢强度参数的准确测量。 项目研究成果将为冷凝器的高效污垢控制奠定基础，对促进电力、化工、机械等行业节能降耗、提高经济效益具有重要意义，应用前景广阔。

本项目研究基于机器视觉的冷凝器污垢特性参数在线测量方法，解决生产现场只依靠污垢传热参数确定控制量而导致的污垢控制效果差的难题。针对冷凝器管内热流扰动、水波动散斑、管壁光反射、湍流等复杂环境，从图像去噪、图像增强、图像复原入手，提出系列图像预处理方法，解决成像质量劣化问题；提出污垢颗粒运动轨迹实时追踪方法，实现污垢沉积率、剥蚀率的在线测量，针对追踪过程中存在的颗粒重叠、被遮挡以及絮凝变形现象，研究基于序列图像的边缘修正补偿算法和基于聚类预估的颗粒匹配算法，实现颗粒运动轨迹的准确追踪；提出基于颜色与纹理特征提取的污垢类型在线识别方法和基于帧间差分层叠记忆的污垢成分及其空间位置在线识别方法，确定污垢混合体结构参数；提出污垢混合体的弹塑性模型等效方法和基于贝叶斯变点识别的结垢阶段检测方法，建立强度参数预估模型并研究模型校正、优化算法，实现了污垢强度参数的准确测量。.项目研究成果为冷凝器的高效污垢控制提供了坚实基础，对促进电力、化工、机械等行业节能降耗、提高经济效益具有重要意义，应用前景广阔。