摘要
针对缺陷定量辨识的非线性、不适定性提出了采用混沌-LM(Levenberg-Marquardt)混合解法结合单面红外检测技术对缺陷进行定量辨识的新算法,并提出了相对敏感系数分析缺陷可检测性的新概念。通过相对敏感系数分析发现,缺陷尺寸、位置的不同描述参数的可检测性并不相同,而且采用检测面达到最大温差时刻点的瞬态检测比稳态检测更具优越性。文中采用数值实验对瞬态和稳态检测均进行了检验,实验结果与相对敏感系数分析结论一致。此外,数值实验还证明了混沌-LM混合解法的有效性,表明了辨识结果的稳定性和全局最优性。
In order to overcome the difficulty of nonlinearity and ill-posedness of the defect quantitative identification with the single-side infrared inspection,a new algorithm（Chaos-LM hybrid algorithm）,which combined the chaos algorithm with the Levenberg-Marquardt method,was introduced in this paper.And a new concept of relative sensitivity coefficient was also brought forward to analyze the defect inspectability.Through analyzing the relative sensitivity coefficient,it was found that the inspectability of different parameters describing the size and position of the defect was different,and the inspection carried out at the time with maximum surface temperature difference was more desirable than that in steady heat transfer state.Both of the transient and steady inspections were tested by a series of numerical experiments,and the experimental results are consistent with the analysis of relative sensitivity coefficient. Besides,the numerical experiments have certified the effectiveness of the Chaos-LM hybrid algorithm, and show that the identification results are stable and globally optimal.
出处
《红外与激光工程》
EI
CSCD
北大核心
2013年第2期317-323,共7页
Infrared and Laser Engineering
基金
国家自然科学基金(50906099)
