Geometry-Aware Crack Identification in Rectangular Cantilever Beam Families Using Frequency-Shift and Curvature-Damage Peak Features: Scripts, Data, and Results

This dataset provides the MATLAB scripts, numerical data, trained regression models, intermediate computational outputs, and evaluation results supporting the study entitled “Geometry-Aware Crack Identification in Rectangular Cantilever Beam Families Using Frequency-Shift and Curvature-Damage Peak Features.” The dataset is organized into three compressed archives: 1. Geometry_Aware_Crack_Identification_Scripts.zip This archive contains the MATLAB scripts used to generate the beam-family data, extract modal features, compute curvature-damage indicators, prepare regression data, train the crack position and depth regression models, and perform the in-grid, noise, off-grid, out-of-family, Abaqus, and experimental evaluations. 2. Geometry_Aware_Crack_Identification_Data_and_Results.zip This archive contains the primary numerical data, intermediate processing outputs, trained model files, and results generated by the main computational workflow. It includes the data and outputs required to reproduce the reported analyses. 3. Geometry_Aware_Crack_Identification_Multi_Seed_Evaluation.zip This archive contains supplementary files generated using three different random seeds. These files are provided to assess the stability of the training procedure under different random initializations. The computational workflow uses three vibration modes and was implemented in MATLAB R2024b. Parallel execution requires the Parallel Computing Toolbox. The scripts may require updating user-specific file paths when transferred to another computer. The archives are provided as supporting research materials for reproducibility and transparency. The primary reported results correspond to the main workflow archive, while the multi-seed archive provides supplementary stability analysis.本数据集配套提供了题为《基于频移与曲率损伤峰特征的矩形悬臂梁族几何感知裂纹识别》的研究所需的MATLAB脚本、数值数据、训练好的回归模型、中间计算结果以及评估结果。 本数据集分为三个压缩归档文件： 1. Geometry_Aware_Crack_Identification_Scripts.zip 该归档包含用于生成梁族数据、提取模态特征、计算曲率损伤指标、制备回归数据集、训练裂纹位置与深度回归模型，以及开展网格内、含噪、网格外、梁族外、Abaqus及实验评估的MATLAB脚本。 2. Geometry_Aware_Crack_Identification_Data_and_Results.zip 该归档包含核心数值数据、中间处理结果、训练好的模型文件，以及主计算流程生成的全部结果，涵盖复现报告中分析所需的全部数据与输出。 3. Geometry_Aware_Crack_Identification_Multi_Seed_Evaluation.zip 该归档包含使用三种不同随机种子生成的补充文件，用于评估训练流程在不同随机初始化条件下的稳定性。 本计算流程采用三种振动模态，基于MATLAB R2024b实现，并行运行需依赖并行计算工具箱（Parallel Computing Toolbox）。将脚本迁移至其他设备时，可能需要更新用户自定义的文件路径。 本归档文件作为支撑研究可重复性与透明度的配套科研材料提供。核心报告结果对应主流程归档文件，而多随机种子归档则提供补充性的稳定性分析内容。