基于粒子群最大似然估计的焊缝早期隐性损伤磁记忆精确定位模型

发布时间：2018-05-16 21:08

本文选题：磁记忆 + 最大似然估计　；参考：《工程科学学报》2017年10期

【摘要】：针对由于焊接残余应力、磁场噪声等干扰,造成磁记忆检测在焊缝早期隐性损伤位置定量评价上的困难,提出基于粒子群算法优化的最大似然估计磁记忆梯度定量模型.通过对预制未焊透缺陷的Q235焊接试件进行焊缝疲劳拉伸实验,同步对比扫描电镜和X射线检测结果,发现磁记忆信号梯度对早期隐性损伤位置反应比较敏感,并获得了梯度随着与隐性损伤的距离增大而减小的衰减变化规律,构建隐性损伤位置参数与磁记忆梯度的非线性函数,考虑磁场噪声对隐性损伤定位结果的影响,引入最大似然估计建立目标函数,进一步考虑目标函数的非线性容易陷入局部极值而非全局极值的问题,采用具有全局搜索能力的粒子群算法对目标函数进行优化,建立基于粒子群最大似然估计的焊缝隐性损伤位置磁记忆定量模型,验证结果表明定位误差仅为3.48%,为实际工程中利用磁记忆技术及时发现早期隐性损伤并精确定位提供了新的思路.
[Abstract]:In view of the interference of welding residual stress and magnetic field noise, the difficulty in quantitative evaluation of the location of early recessive damage in the weld is caused by magnetic memory detection. The maximum likelihood estimation of magnetic memory gradient quantitative model based on particle swarm optimization is proposed. The fatigue tensile test of Q235 welded specimens with prefabricated defects is carried out. Compared with the results of scanning electron microscopy and X ray detection, it is found that the magnetic memory signal gradient is sensitive to the position of the early recessive damage, and the attenuation change law decreases with the increase of the distance between the recessive damage and the recessive damage. The nonlinear function of the recessive damage position parameter and the magnetic memory ladder is constructed, and the recessive loss of the magnetic field noise is considered. The maximum likelihood estimation is introduced to establish the target function, and the target function is further considered as the target function is easy to fall into the local extremum rather than the global extremum. The target function is optimized by the particle swarm optimization with the global search ability, and the hidden damage location of the weld is established based on the maximum likelihood estimation of the particle swarm. The magnetic memory quantitative model shows that the positioning error is only 3.48%, which provides a new idea for the timely discovery of early recessive damage and accurate positioning by magnetic memory technology in practical engineering.
【作者单位】：东北石油大学机械科学与工程学院;哈尔滨工业大学航天学院;
【基金】：国家自然科学基金资助项目(11272084,11072056,11472076) 中国石油科技创新基金资助项目(2015D-5006-0602) 中国石油和化学工业联合会科技指导计划资助项目(2017-01-05) 东北石油大学研究生创新科研项目资助(YJSCX2016-024NEPU)
【分类号】：TG441.7

【相似文献】