China Railway Group (China)

Prompt and accurate detection of rolling-element-bearing defects is essential for the reliable operation of rotating equipment. This study presents an enhanced diagnostic strategy that couples Fourier–Bessel empirical wavelet transform (FBEWT) with an enhanced incremental relevance vector machine (EIRVM). FBEWT’s core novelty lies in using amplitude-modulation information to preset its frequency boundaries, yielding stable results even when sampling rate or observation window changes. Meanwhile, the upgraded incremental RVM handles the final classification of bearing health states. Test data show that FBEWT outperforms conventional empirical wavelet decomposition, and EIRVM surpasses the standard RVM classifier. Overall, the integrated FBEWT–EIRVM framework demonstrates strong capability for diagnosing rolling-element-bearing faults.

To address the problem of temperature cracking caused by the concentrated release of hydration heat in mass concrete, this study developed a low-heat composite cementitious material (CWCM) by partially replacing conventional mineral admixtures with construction spoil. A multi-factor synergistic optimization design based on response surface methodology (RSM) was conducted. The water–binder ratio, spoil replacement ratio, curing temperature, and ball-milling time were selected as influencing factors, while the 28-day flexural strength, 28-day compressive strength, and 72 h cumulative hydration heat were used as response variables. A four-factor, three-level Box–Behnken model was established. The results show that the regression model exhibits good fitting performance, and the prediction errors between the predicted and experimental values of all response variables are within a reasonable range. Under the optimized mixture proportion (15% spoil replacement), the system achieves a 28-day compressive strength of 61.03 MPa, while the 72 h cumulative hydration heat is reduced by approximately 15%, meeting the requirements for low-heat cement. Microstructural analyses using XRD, SEM, and TG/DTG indicate that a decrease in the Ca/Si ratio and an increase in the Al/Si ratio promote the formation of a denser C-(A)-S-H gel structure, enhancing the pozzolanic reaction. This mechanism plays a key role in achieving the synergistic regulation of strength enhancement and hydration heat reduction. Compared with conventional fly ash or slag systems, this study innovatively utilizes construction spoil as a partial substitute for traditional mineral admixtures. While maintaining satisfactory mechanical performance, the proposed system effectively reduces hydration heat release, providing a new pathway for temperature control design in mass concrete engineering and high-value resource utilization of construction waste.

Accurate in-motion inspection of railway wheels requires reliable estimation of wheel diameter, flange geometry, and localized tread spalling under partial views, noise, and surface wear. Many existing systems either infer wheel condition indirectly, measure only a subset of parameters, or suffer from ambiguous alignment on nearly axis-symmetric wheel surfaces and biased diameter fitting when defect points are included. These issues limit practical wayside deployment where observations are sparse and partially occluded. This paper presents a stereo-vision sensing system and an integrated 3D measurement pipeline for high-precision wheel condition assessment. The proposed approach is novel, to our knowledge, in combining keypoint-constrained congruent-set registration to resolve near-symmetry ambiguity with rim-corner plane-circle fitting to suppress wear/defect-induced bias in rolling-radius estimation. Coarse alignment is obtained via constrained congruent-set sampling and refined by least-squares pose optimization to fuse sparse observations into a consistent wheel model. Wheel diameter is estimated from rim-corner points, while flange height and thickness are extracted using projection-based interval sampling and NURBS profile fitting. Spalling regions are segmented by boundary extraction and clustering, and spalling length and depth are quantified through cylindrical projection. Experiments validated against gauge-based ground truth showed mean absolute errors of 0.08mm (rolling radius), 0.05mm (flange height), 0.06mm (flange thickness), 0.02mm (spalling length), and 0.09mm (spalling depth), and reduced rolling-radius error by 38.5% compared with a conventional baseline. The results demonstrate a practical, submillimeter solution for in-motion wheel condition monitoring.

To achieve accurate prediction of tunnel boring machine (TBM) penetration rate (PR) under complex geological conditions, this study proposes an interpretable machine learning (ML) framework optimized using Bayesian optimization (BO). A dataset comprising 411 samples collected from two TBM projects was established, incorporating key geological parameters (uniaxial compressive strength, elastic modulus, and Poisson’s ratio) and operational parameters (thrust force and rotation speed). Seven ML models, including decision tree, Gaussian process regression, k-nearest neighbors, random forest (RF), support vector regression, extreme gradient boosting, and light gradient boosting machine, were systematically optimized using BO under five-fold cross-validation. The results demonstrate that ensemble learning models exhibit superior predictive performance. Among them, the RF model achieved the highest accuracy on the testing set with an R 2 of 0.9582, followed by the light gradient boosting machine with an R 2 of 0.9565. SHapley Additive exPlanations (SHAP) analysis indicates that uniaxial compressive strength is the dominant factor controlling PR, while thrust force act as the primary controllable parameters. The proposed framework effectively integrates high predictive accuracy with strong interpretability, providing a reliable methodological basis for intelligent TBM performance prediction and construction parameter optimization.