基于实时图像处理的镁熔液含氢量自动检测方法与实验研究

Field measuring and calculation analysis of hydrogen content in magnesium melt are very important for scientific purpose and practical application value to control the quality of magnesium melt,microstructure and mechanical performance of magnesium alloys. Present measurement methods such as vacuum extraction or hydrogen separative method can not satisfy the need of field measuring because of longer measuring time or lower precision. This project presents an automatic measurement method of hydrogen content in magnesium melt and verify its effect by a large number of experiments based on real-time images processing technology.At first,an embedded system of images processing would be constructed based on detection thoery of the first bubble method,which was used to study real-time images processing and identification algorithms correspoinding to random precipitation of the first bubble and seek the principles of many algorithm' optimizational assembly and scientific calculation,furthermore,the automatic sorting methods would be established between the bubble and scum; Then, the foundational theory would be proposed for the real-time images processing applied for this detection method; And then, a large number of experiments would be carried through based on the above theoretical study by designing and manufacturing prototype instrument. At last, comparing to experimental data and calibrating results, the theory of real-time images processing would be perfected, and the measure raise of precision of hydrogen content would be explored furtherly and the effect law of the characteristic parameters on the precision would be valued. The sudy of the project will be helpful to seek the behavior motive of hydrogen in magnesium melt and supply the theoretical support for the raise of melting level and the rate of finished castings.

镁熔液含氢量的现场检测及计算分析，对镁合金液态质量控制，组织结构和力学性能等具有重要的科学意义和实际应用价值。现有的检测方法如真空萃取法或氢分离法，因测试时间较长，或测量精确度不高无法满足现场检测的需要。本项目提出了一种基于实时图像处理的镁熔液含氢量自动检测方法并通过大量实验来验证该方法的效果。在第一气泡法检测原理基础上，针对第一气泡析出的随机性问题，研究第一气泡随机析出的实时图像处理与识别算法，探索多种算法优化组合和科学计算的原理，并建立气泡与浮渣图像自动分类方法，提出实时图像处理应用于该检测方法的基础理论。在以上理论研究基础上，设计制作含氢量检测样机并进行大量实验，比较实验数据与标定结果，完善实时图像处理理论，并进一步探索提高含氢量检测精确度的理论方法以及评价各特征参数对精确度的影响规律。本项目的研究有助于探寻镁熔液中氢的行为规律，为提高镁合金的熔炼水平与铸件的成品率提供理论支持。

镁合金铸件中显微气孔的存在严重影响其力学性能，H2析出起了主要作用。现有的检测方法无法满足现场检测的需要。本项目围绕基于实时图像处理的镁熔液含氢量自动检测方法和实验展开了研究。探讨了溶解态氢形核与长大机制及工艺参数的影响，研究了第一气泡随机析出的实时图像处理与识别算法，并建立了气泡与浮渣图像辨识方法，构建了实时图像处理平台并与PLC控制器进行通信，实现了基于实时图像处理的镁熔液含氢量自动检测方法。. 理论推导出了镁熔液中氢气泡析出动力学模型并分析了多个因素对气泡析出的动态变化影响。在满足镁熔液-氢气泡相体系条件下，气泡初始形核高度及镁熔液温度对氢气泡析出变化影响较小，气泡初始半径对其析出影响较为显著。相同时刻下，氢气泡初始半径与气泡上浮速度、上浮高度都成正相关；且在一定变化范围内，随着气泡初始半径增大，气泡到达最大上浮速度的时间滞后性更大。. 研究了第一气泡随机析出对图像检测方法的影响，通过采用不同的图像复杂度因子分别分析不同位置气泡析出时对应的全域背景复杂度和其邻域背景复杂度，探究了镁熔液第一气泡随机析出图像检测方法的普遍适用性，即背景复杂度高的镁熔液第一气泡图像检测方法适用于其他背景复杂低的条件下的第一气泡检测识别。构建了一种基于析出规律差异的气泡与熔渣简单实时辨识方法，获得较好的辨识效果。. 构建了实时图像处理平台，采用C语言进行编程，通过CCS3.3控制算法程序的调试、优化、移植等，并烧写到VPM642板块，结果表明基于DM642的视觉检测平台可以用于镁熔液第一气泡的有效检测。实现了实时图像处理平台和PLC控制器两种通信方式，即I/O通信，基于Modbus协议的通信，还扩展了DM642与运动控制器的通信。经过大量实验研究，证实了基于实时图像处理的镁熔液含氢量自动检测方法的可行性和有效性。. 本项目的研究有助于探寻镁熔液中氢的行为规律，能为提高镁合金的熔炼水平与铸件的成品率提供理论支持。