Data-driven models often neglect the underlying physical principles, limiting generalization capabilities in water distribution systems (WDSs). This study presents a novel spatio-temporal graph physics-informed neural network (ST-GPINN) for water quality prediction in WDSs, integrating hydraulic simulations, physics-informed neural networks (PINNs), and graph neural networks (GNNs) to capture dynamics and graph-based network connectivity while approximating partial differential equations (PDEs). ST-GPINN discretizes WDSs using virtual nodes to enhance spatial granularity, employs an Encoder-Processor-Decoder architecture for predictions. Validated on Network A (a small-scale network with 9 junctions and 11 pipes) and Network B (a real large-scale WDS with 920 junctions and 1032 pipes), ST-GPINN outperforms others, achieving a MAE of 0.0073 mg/L, RMSE of 0.0121 mg/L, and R2 of 88.91% in Network A, and a MAE of 0.008 mg/L, RMSE of 0.0098 mg/L, and R² of 98.91% in Network B. Its scalability and accuracy highlight ST-GPINN’s potential for water quality predictions.
This research is motivated by the practical pre-steelmaking stage in large iron and steel companies, which have steady and heavy demands for the steelmaking production process. Our problem studied the pre-steelmaking stage, which consists of two steps that are needed in each convertor before the steelmaking process. During each step, a necessary transportation must be operated by a crane. In contrast to the classical two-machine flowshop problem during which both machines are fixed, these transporting operations are performed by two mounted, removeable cranes. Our problem is scheduling two-crane operations for the sake of minimizing the last convertors’ completion time (makespan); that is, the last finish time among the total operation of the two cranes is minimized. This study was concerned with resolving the interference between two cranes by determining the sequence of loading operations and how each crane avoids the other in order to let it complete its next operation first. A mixed integer linear programming (MILP) model was developed to represent the problem, and we further present the computational complexity of the problem. The result implies that our problem is very difficult to solve, and it is computationally challenging to solve the model. A special case is provided, which can be optimally solved in polynomial time. Furthermore, an evolutionary algorithm cuckoo search (CS) algorithm was attempted to obtain near-optimal solutions for medium- and large-scale problems. Finally, the efficiency and effectiveness of our methods were validated by numerical results in both simulated instances as well as real data from a practical production process.
A novel non-equilibrium molecular dynamics (MD) model was presented to investigate the transport mechanisms and desalination performance of thin-film composite membranes (TFC). Firstly, polyamide (PA) and polysulfone (PSF), or PSF/UiO-66 models, were constructed as active and support layers. Subsequently, feed solution and draw solutions with 0.5, 1.0, and 1.5 M salt concentrations were generated. Thirdly, these aqueous and membranes were combined to build forward osmosis (FO) models. The FO process was simulated to anatomise the membrane fouling, water flux, and salt rejection, which JAVA and MD programs were combined to calculate. The results indicated that the PA-PSF/UiO-66 membrane exhibited a stable water transport and salt rejection performance compared to the pristine PA and PA-PSF membranes. Finally, experiments and simulations were compared, and the same trends were found. Therefore, this simulation work provided significant insight in guiding the design and synthesis of the FO membrane.
Conventional vehicle character recognition methods often treat detection and recognition as separate processes, resulting in limited feature interaction and potential error propagation. To address this issue, this paper proposes a structure-aware self-supervised Masked Autoencoder (CharSAM-MAE) framework, combined with an independent region extraction preprocessing stage. A YOLOv8n detector is employed solely to crop the region of interest (ROI) from full-frame vehicle images using 50 single bounding-box annotated samples. After cropping, the detector is discarded, and subsequent self-supervised pre-training and recognition are fully executed using MAE without any involvement of YOLO model parameters or labeled data. CharSAM-MAE incorporates a structure-aware masking strategy and a region-weighted reconstruction loss during pre-training to improve both local structural representation and global feature modeling. During fine-tuning, a multi-head attention-enhanced CTC decoder (A-CTC) is applied to mitigate issues such as sparse characters, adhesion, and long-sequence instability. The framework is trained on 13,544 ROI images, with only 5% of labeled data used for supervised fine-tuning. Experimental results demonstrate that the proposed method achieves 99.25% character accuracy, 88.6% sequence accuracy, and 0.85% character error rate, outperforming the PaddleOCR v5 baseline (98.92%, 85.2%, and 1.15%, respectively). These results verify the effectiveness of structure-aware self-supervised learning and highlight the applicability of the proposed method for industrial character recognition with minimal annotation requirements.