Modelling

The air-cooling system for injection compressors in underground gas storage (UGS) facilities is critical to operational safety. Its control must balance outlet temperature safety with fan energy savings. However, traditional empirical methods struggle to balance these dual objectives, especially for parallel multi-unit operations where quantitative scheduling guidelines are lacking. To address this, we take the eight parallel air-cooling units at the Chongqing Xiangguosi UGS facility as the research object. We propose an integrated framework that covers multi-condition outlet temperature prediction, heat transfer performance evaluation under a unified baseline, and hierarchical fan speed control. First, an adaptive sample selection mechanism is established using joint evaluation criteria that include gradient uncertainty and diversity metrics. By selecting samples with high information density, a deep neural network (DNN) model is trained to predict the outlet temperature of the air-cooling system. Second, a two-dimensional operating-condition zoning system is constructed based on ambient temperature and natural gas volumetric flow rate. This prediction model is used to fill data gaps under various conditions, enabling a performance ranking of all units under the same baseline. This ranking provides a quantitative basis for optimal unit selection at the system level. Furthermore, for individual units, multi-condition fan speed sweep simulations are conducted using HTRI Xchanger Suite. These simulations establish the mapping between thermal margin and fan speed. Consequently, a hierarchical control strategy is developed that coordinates system-level scheduling with equipment-level speed regulation. This strategy covers both normal energy-saving scenarios and over-temperature emergency scenarios, while clearly defining engineering safety constraints for low-speed operation. The application results show that the proposed method significantly reduces fan energy consumption while ensuring process safety. It offers a valuable technical reference for intelligent and energy-efficient operation of similar multi-unit heat exchange systems.

Porous asphalt mixtures (PAC) with high connectivity porosity can significantly reduce surface runoff and play a more critical role in sustainable paving. This study investigated the permeability performance of PAC-13 porous asphalt mixtures with three target porosities and four design-density levels using computational fluid dynamics (CFD) and the discrete element method (DEM). Solid-volume-fraction mapping, pressure and velocity responses, and permeability coefficients were used to characterize the seepage behavior, while equivalent void diameter and fractal dimension were introduced to describe mesoscopic void features. The correlation between mesoscopic void structure and permeability was analyzed. Results showed that the internal void distribution of PAC-13 exhibited clear spatial heterogeneity, and the overall porosity alone could not fully characterize the pore structure. Increasing the target porosity from 20% to 22% reduced the average peak pressure by approximately 23.5%. The average permeability coefficient increases with the porosities from 18%, 20%, and 22%, with the highest value of 0.00393 m/s obtained at 22% target porosity and a design-density level of 0.90. Correlation analysis showed that the permeability coefficient was positively correlated with target porosity, equivalent void diameter, and fractal dimension. A regression model incorporating these macro–mesoscopic parameters achieved an (R²) of 0.91. These findings indicate that PAC-13 drainage performance should be evaluated by combining overall porosity with mesoscopic void-structure parameters. The findings rely exclusively on computer simulation, and further experimental tests should be conducted.

Electrocardiogram (ECG) signals exhibit distribution shifts across recording devices, populations, sampling rates, and lead configurations, impairing generalization and long-term feature extraction. To address these issues, this paper proposes a multihead self-attention and adaptive multiscale dilated convolution model (MHA-MSD). First, a multihead encoder projects ECG features into a high-dimensional space to encode both temporal dynamics and inter-lead correlations. A subsequent self-attention mechanism then enhances cross-dataset generalization. Next, an adaptive multiscale dilated convolution module constructs a multibranch network with different dilation rates and uses a gating network to dynamically adjust branch weights according to signal length and feature distribution. Finally, a spatial attention module fuses multibranch features and refines features during dimensionality reduction to enhance global feature learning. On CPSC2018, the model classifies nine categories (Normal, AF, I-AVB, LBBB, RBBB, PAC, PVC, STD, and STE), achieving 93.6% accuracy and a 92.9% mean F1 score in the best-performing fold among the 10 cross-validation iterations. On the MIT-BIH-to-INCART transfer, the model classifies normal (N), supraventricular ectopic (S), and ventricular ectopic (V) beats and achieves 97.38% accuracy and 97.41% mean F1 score, while class-discriminative attention maps align with clinically relevant waveform segments. Cross-dataset classification of N, LBBB, and RBBB among CPSC2018, PTB-XL, and CODE-15% yields respective F1 scores of 0.87, 0.96, and 0.86 for CPSC2018-to-PTB-XL transfer.