混凝土缺陷智能化快速检测与定量识别技术研究

针对目前混凝土缺陷无损检测技术的不足及相关模型实验研究十分缺乏的现状,结合实际工程中混凝土结构常见的质量缺陷,制作了一系列含有不同类型、性质缺陷及无缺陷的混凝土模型试件,开展了基于先进的信号处理技术和人工智能技术的混凝土缺陷无损检测的模型实验研究。针对冲击回波测试信号非稳态的复杂特性,应用小波变换技术有效地提取了缺陷信号的特征值;并应用极限学习机(ELM)作为分类模型,由此在理论分析和模型试验的基础上,建立了基于小波分析和极限学习机的混凝土缺陷智能化快速检测与定量分类识别系统。结果表明:该系统具有较好的分类识别性能,初步实现了对混凝土缺陷类型、性质和范围的智能化快速定量识别与评价,极大地提高了混凝土缺陷检测与评估的速度及精度。

On nondestructive testing and intelligent quick identificationof concrete defects

Aiming at the current situation of deficiency of nondestructive testing technology of concrete defects and extreme lack of related model test study at home and abroad, in this study we take representative RC plate structure in the civil and hydraulic engineering as the research object and make a series of concrete model specimens. Thus experimental study on concrete defects nondestructive testing is performed based on advanced signal processing and artificial intelligence technologies. In view of the fact that impact echo testing signal has a complex non-stationary characteristic, we adopt wavelet transform based feature extraction for extracting a number of features out of impact echo waveforms and use extreme learning machine (ELM) as classification models. Thus our proposed concrete defect detection and intelligent quick identification model system is established on the basis of theoretical analysis and model test. The system, built on advanced signal processing and artificial intelligence technologies, can achieve better defect detection, defect diagnosis, defect sizing location, and can perform both qualitative and quantitative identification and evaluation quickly. The verification results show that the system is more applicable, efficient and effective in nondestructive testing of concrete defects.