2219铝合金厚板搅拌摩擦焊缺陷在位辨识与精准定量数据集

为提高铝合金搅拌摩擦焊焊缝缺陷超声检测分辨力，结合信号处理和机器学习技术，提出铝合金缺陷亚波长级超声检测技术，突破瑞利衍射极限，将超声检测分辨力由波长级提升至亚波长级。具体包括基于自回归谱外推方法、稀疏解卷积-合成孔径聚焦技术、基于模型解释策略优选特征的机器学习方法和基于反向传播神经网络解耦混叠信号的提高超声全聚焦成像分辨力的方法，数据集主要记录了原始超声数据、基于经验选择模型参数的谱外推超声数据、基于自回归方法确定模型参数的谱外推超声数据、常规SAFT缺陷超声信号数据、Sparse-SAFT缺陷超声信号数据、基于压缩感知的碳钢缺陷超声检测频谱数据、机器学习模型预测的缺陷信号到达时间、机器学习模型处理后的高分辨力超声信号、基于高维特征空间预测的缺陷信号到达时间、基于高贡献子集预测的缺陷信号到达时间等观测值。超声检测实验均严格按照相控阵超声检测国家标准GB/T 32563-2016进行，由实验室专业人员进行试验测试和数据采集，保证科学数据质量。To improve the ultrasonic testing resolution of defects in aluminum alloy friction stir welding (FSW) seams, a sub-wavelength scale ultrasonic testing technology for aluminum alloy defects is proposed by combining signal processing and machine learning technologies, which breaks through the Rayleigh diffraction limit and upgrades the ultrasonic testing resolution from the wavelength scale to the sub-wavelength scale. Specifically, the proposed technology covers four core technical approaches: autoregressive spectral extrapolation method, sparse deconvolution-synthetic aperture focusing technique (Sparse-SAFT), machine learning method with feature optimization via model interpretation strategies, and the method for enhancing the resolution of ultrasonic total focusing method (TFM) imaging by decoupling aliased signals using backpropagation neural network (BPNN). The dataset mainly records the following observations: original ultrasonic data, spectral extrapolation ultrasonic data with model parameters empirically selected, spectral extrapolation ultrasonic data with model parameters determined via autoregressive method, conventional SAFT defect ultrasonic signal data, Sparse-SAFT defect ultrasonic signal data, compressed sensing-based ultrasonic testing spectral data for carbon steel defects, defect signal arrival times predicted by machine learning models, high-resolution ultrasonic signals processed by machine learning models, defect signal arrival times predicted based on high-dimensional feature space, and defect signal arrival times predicted based on high-contribution feature subsets. All ultrasonic testing experiments were carried out strictly in accordance with the Chinese national standard GB/T 32563-2016 for phased array ultrasonic testing, and were performed and data acquired by professional laboratory personnel to guarantee the quality of the scientific dataset.

搜集汇总

背景与挑战

背景概述

该数据集专注于2219铝合金厚板搅拌摩擦焊缺陷的超声检测，通过结合信号处理和机器学习技术，将检测分辨力提升至亚波长级，实现了缺陷的在位辨识与精准定量。数据集包含原始超声数据、基于多种先进方法处理的高分辨力信号数据等观测值，实验严格按照国家标准进行，确保科学数据质量。