Reinforced concrete slabs are widely used in building structures due to their economic, durable, and aesthetic advantages. The determination of their ultimate strength often hinges on punching shear strength. Presently, methods such as closed hoops, steel bending, and fiber reinforcement are employed to enhance punching shear strength, with fiber reinforcement gaining popularity due to its ease of implementation and efficacy in improving concrete durability. This study introduces a novel approach employing six machine learning algorithms rooted in decision trees and decision tree-based ensemble learning to predict punching shear strength in steel fiber-reinforced concrete slabs. To overcome experimental data limitations, a data augmentation approach based on the Gaussian mixture model is employed. The validation of the data augmentation is conducted through “synthetic training—real testing” and “real training—real testing”. Additionally, the best machine learning model is analyzed for explainability using Shapley Additive exPlanation (SHAP). Results demonstrate that the proposed data augmentation method effectively captures the original data distribution, enhancing the robustness and accuracy of the machine learning model. Moreover, SHAP provides better insights into the features influencing punching shear strength. Thus, the proposed data enhancement model offers a reliable approach for modeling small experimental datasets in structural engineering.

Metal Magnetic Memory (MMM) exhibits the advantage of not requiring embedded sensors or external excitation, making it suitable for inspecting ferromagnetic components in engineering structures. This study introduced MMM into stress detection of steel strands. Graded tensile tests were conducted on the steel strands to investigate the correlation between Self-Magnetic Flux Leakage (SMFL) signals and stress levels. Different spatial detection positions with varying Lift-Off Values (LOV) and Rotation Angle Values (RAV) were set to examine the distribution of spatial SMFL field under load. Furthermore, a magnetic characteristic parameter A_N was proposed to assess the stress level of the steel strands. The results indicate that the rate of change in the middle region of the SMFL curve was lower than that at the beginning and the end. Additionally, with increased applied load, the SMFL curve exhibited systematic variations, and the dispersion of the normal component curve gradually decreased. By utilizing the magnetic characteristic parameter A_N, the stress in the steel strands can be calculated, with the parameters determined based on LOV and RAV. This achievement expanded the nondestructive testing methods for steel strands and holds significant research value.

Reliable short-term wind speed prediction is one of the core technologies in the strong wind warning system for railway applications, which is of great significance for ensuring the safety of high-speed train operations and ancillary railway facilities. To improve forecasting accuracy, decomposition-based methods have attracted extensive attention due to their superior ability to address complex data characteristics (e.g., nonstationarity and nonlinearity). Currently, there are two pre-processing schemes for decomposition-based methods, i.e., one-time decomposition and real-time decomposition. In order to apply them better, this paper first expounds the difference between them, based on a combination of DWT (discrete wavelet transform) and CKDE (conditional kernel density estimation). The results show that although the one-time decomposition-based method has an unexceptionable accuracy, it only can provide offline prediction and thus may not be practical. The real-time decomposition-based method possesses stronger practicability and is able to provide online prediction, but it has limited accuracy. Then, an improved ensemble strategy is developed by optimizing the selection of appropriate decomposed components to conduct the prediction on the basis of real-time decomposition. This improved ensemble strategy provides an effective guidance for this selective combination, including taking historical information into consideration in the data. Finally, numerical examples and practicality analysis using two groups of measured wind speed data demonstrate that the proposed method is effective in providing high-precision online wind speed prediction. For example, compared with CKDE, the average degrees of improvement achieved by the proposed method in terms of MAE, RMSE, and MRPE, are 16.25%, 17.66%, and 16.93, respectively, while those compared with the traditional real-time decomposition method are 17.11%, 18.54%, and 16.84, respectively.