Indoor bioaerosols present a significant risk to human health, contributing to the transmission of respiratory and infectious diseases. Ultraviolet (UV) radiation has emerged as a widely used approach for controlling indoor bioaerosols. Simulation-based design provides a powerful approach for predicting the disinfection effectiveness of UV radiation. A critical prerequisite for simulations is the accurate calculation of the UV irradiance distribution. However, the widely used view factor method falls short in accounting for the occlusion of UV radiation by irregular objects, which can lead to an overestimation of the effectiveness of UV disinfection. To address this limitation, this study proposed an improved view factor method that incorporated UV shading effects. The proposed method was first validated in a full-scale chamber equipped with a heated manikin and a table. The calculated UV irradiance distribution and bioaerosol concentrations were compared against experimental measurements. Results showed that our method successfully predicted the UV shading effects caused by irregular objects and the variations in bioaerosol concentrations. The method was subsequently applied to a shared office case to demonstrate the importance of accounting for shading effects in the calculations. Results revealed that neglecting the shading effect overestimated the UV disinfection efficiency by a factor of 1.3-2.4 in calculation. By incorporating shading effects into calculations, the method delivered more reliable UV dose predictions, enabling safe and accurate design of indoor UV systems while mitigating occupant exposure risks of bioaerosols.

Subgraph matching, a cornerstone of relational pattern detection in domains ranging from biochemical systems to social network analysis, faces significant computational challenges due to the dramatically growing search space. Existing methods address this problem within a filtering-ordering-enumeration framework, in which the enumeration stage recursively matches the query graph against the candidate subgraphs of the data graph. However, the lack of awareness of subgraph structural patterns leads to a costly brute-force enumeration, thereby critically motivating the need for intelligent navigation in subgraph matching. To address this challenge, we propose Neural Graph Navigation (NeuGN), a neuro-heuristic framework that transforms brute-force enumeration into neural-guided search by integrating neural navigation mechanisms into the core enumeration process. By preserving heuristic-based completeness guarantees while incorporating neural intelligence, NeuGN significantly reduces the First Match Steps by up to 98.2% compared to state-of-the-art methods across six real-world datasets.

Since urban residents heavily rely on rail transit on a daily basis, controlling airborne infectious disease transmission in railway trains is critical for public health. In recent years, 222 nm far-ultraviolet C (UVC) lamps have emerged as a promising solution for disinfection of bioaerosols in occupied environments with enhanced safety. However, there is a lack of experimental data in railway train compartments, especially regarding the effectiveness of far-UVC against bioaerosols that are suspended in the air and that settle on handrails. Therefore, this study experimentally and numerically evaluated the effectiveness of far-UVC disinfection of bioaerosols suspended in air and deposited on surfaces using a simplified, scaled railway carriage model. The experimental data show that, under a high ventilation rate of 107 air changes per hour (ACH), the far-UVC lamp still provided supplementary reduction of viable airborne bioaerosols, achieving a disinfection efficiency of 23

Far-UVC radiation can be an efficient and secure approach for bioaerosol sterilization in occupied high-risk settings like healthcare facilities. In dental clinics, aerosols produced by dental procedures are characterized by the existence of co-generated particles such as grind tooth fragments, restorative materials, and dental calculus. In this study, the influence of particles on far-UVC disinfection efficiency was investigated. First, the susceptibility of aerosolized bacteria and viruses to far-UVC radiation was experimentally measured under three conditions, i.e., without particles, with co-generated particles, and with ambient particles. The findings indicated that co-generated particles, but not ambient particles, shielded bioaerosols from far-UVC inactivation. The far-UVC susceptibility of aerosolized E. coli and MS2 with co-generated particles was only 49 % and 40 % of that without particles, respectively. Then, the impact of co-generated particles on far-UVC disinfection performance and occupational exposure to viable bioaerosols in a dental clinic was investigated by numerical method. It was found that, due to the shielding effect of co-generated particles on bioaerosols against far-UVC inactivation, the dentist’s exposure to viable bioaerosols increased by 41 %. Thus, it is crucial to consider the shielding effect of co-generated particles during the far-UVC system design to ensure reliable bioaerosol inactivation.

Exposure to aerosolized bacteria poses a significant threat to human health as they can cause various respiratory infections. In dental clinics, the bacteria-laden aerosols generated from dental procedures may disperse to the dentist's breathing zone via air and cause cross-infection. A unique feature for aerosols generated during dental procedures is that they contain solid particles. In this study, the impact of co-generated solid particles on the activity of aerosolized bacteria and its influence on exposure assessment were investigated. This study first compared the recovery of aerosolized bacteria with and without co-generated solid particles. The results showed that with co-generated solid particles, the number of recovered bacteria increased by 142 %. Then, the effects of three influencing factors were analyzed, including solid particle size, particle concentration in suspension, and relative humidity of air. Finally, the influence of co-generated solid particles on the exposure assessment in a dental clinic was numerically evaluated. The results showed that, if the impact of co-generated solid particles was not considered, the dentist's exposure would be under-estimated by 58.6 %. This study revealed the phenomenon of co-generated solid particles acting as protectants to enhance the bacterial survival in aerosols, providing valuable insights for accurate exposure and risk assessment.

Temporal Graph Transformers (TGTs), which incorporates Transformer into the temporal graph learning models, are powerful tools for analyzing and predicting temporal graph data. However, most existing TGT models focus on one-hop interactions due to sequence correlation and computational complexity caused by neighborhood explosion. This limited focus on local subgraph structures restricts the representational power of current TGTs. Additionally, the introduction of higher-order structures exacerbates efficiency issues in TGTs, with the time-consuming feature processing stage often neglected, leading to low training efficiency. To address these challenges, we propose QSFormer (Quick Sense Temporal Graph TransFormer), a solution designed to enhance local sensation ability and accelerate training efficiency in TGTs. QSFormer includes a sense augmentation strategy that incorporates high-order neighbor-hoods with position-differentiated encoding and extends common neighbor. Furthermore, QSFormer implements a quick training framework for TGTs to accelerate feature processing and model convergence, including padded parallel sampling and adaptive mini-batch generation. Extensive experiments demonstrate that QSFormer consistently outperforms existing baselines, including TGTs such as DyGFormer and HOT. Notably, QSFormer surpasses these TGTs in training speed by over four and seven times, respectively. Our code is publicly available at https://github.com/Stephanie0002/QSFormer.

Controlling indoor air contaminants is crucial for human health. Among the various control methods, source control stands out as the most direct and effective approach. To control an unknown source, locating the air contaminant source is essential. The rapid development of robotics and computing technologies opens up opportunities for advanced methods of locating air contaminant sources in indoor environments. This study developed an advanced algorithm for locating an air contaminant source by means of a sensing robot with a pre-calculated air contaminant distribution database. The algorithm was based on a single and fixed source with a constant emission rate. The air contaminant distribution database was established by computational fluid dynamics (CFD) simulations. The feasibility of the proposed method was assessed by computer experiments in four typical indoor environments. When the sensing robot was used with a pre-calculated air contaminant distribution database, the advanced algorithm was able to identify the source location within one grid cell (< 0.72 m) from the real location in 93% to 97% of the cases for the four environments. Furthermore, on average, the advanced algorithm only took up to 1.72 more steps than the shortest route to locate the air contaminant source for the four typical indoor environments.

Memory-based temporal graph neural network (MTGNN) models are effective for predicting temporal graphs by using node memory and message-passing modules to capture temporal and structural information, respectively. However, distributed training for large graphs presents challenges such as accuracy loss and decreased efficiency due to remote features and memory transmission. Despite improvements in MTGNN system optimizations, issues like dynamic load imbalances, communication overhead, and memory staleness persist. To tackle these challenges, we introduce MemShare, a distributed MTGNN system. MemShare introduces a novel shared node memory paradigm that utilizes a small subset of shared nodes across machines and GPUs to reduce distributed communication for memory management. It incorporates techniques like shared nodes-centric graph partitioning, shared nodes-aware boundary decay sampling, and shared nodes-targeted synchronous smoothing aggregation. Experiments show that MemShare outperforms existing distributed MTGNN systems in accuracy and training efficiency.

Graph Neural Networks have emerged as an effective machine learning tool for multi-disciplinary tasks such as pharmaceutical molecule classification and chemical reaction prediction, because they can model non-euclidean relationships between different entities. Particle crushing, as a significant field of civil engineering, describes the breakage of granular materials caused by the breakage of particle fragment bonds under the modeling of numerical simulations, which motivates us to characterize the mechanical behaviors of particle crushing through the connectivity of particle fragments with Graph Neural Networks (GNNs). However, there lacks an open-source large-scale particle crushing dataset for research due to the expensive costs of laboratory tests or numerical simulations. Therefore, we firstly generate a dataset with 45,000 numerical simulations and 900 particle types to facilitate the research progress of machine learning for particle crushing. Secondly, we devise a hybrid framework based on GNNs to predict particle crushing strength in a particle fragment view with the advances of state of the art GNNs. Finally, we compare our hybrid framework against traditional machine learning methods and the plain MLP to verify its effectiveness. The usefulness of different features is further discussed through the gradient attribution explanation w.r.t the predictions. Our data and code are released at https://github.com/doujiang-zheng/GNN-For-Particle-Crushing.

Mini-batch Graph Transformer (MGT), as an emerging graph learning model, has demonstrated significant advantages in semi-supervised node prediction tasks with improved computational efficiency and enhanced model robustness. However, existing methods for processing local information either rely on sampling or simple aggregation, which respectively result in the loss and squashing of critical neighbor information. Moreover, the limited number of nodes in each mini-batch restricts the model’s capacity to capture the global characteristic of the graph. In this paper, we propose LGMformer, a novel MGT model that employs a two-stage augmented interaction strategy, transitioning from local to global perspectives, to address the aforementioned bottlenecks. The local interaction augmentation (LIA) presents a neighbor-target interaction Transformer (NTIformer) to acquire an insightful understanding of the co-interaction patterns between neighbors and the target node, resulting in a locally effective token list that serves as input for the MGT. In contrast, global interaction augmentation (GIA) adopts a cross-attention mechanism to incorporate entire graph prototypes into the target node representation, thereby compensating for the global graph information to ensure a more comprehensive perception. To this end, LGMformer achieves the enhancement of node representations under the MGT paradigm. Experimental results related to node classification on the ten benchmark datasets demonstrate the effectiveness of the proposed method. Our code is available at https://github.com/l-wd/LGMformer.

Traffic trajectories of various vehicles, bicycles and pedestrians can help understand the traffic dynamics in a fine-grained manner like traffic flow, traffic congestion and ride -hailing demand. The comprehensive usage of traffic trajectory data has not been fully investigated due to the prevalent privacy concerns and commercial limitations. The traffic trajectory simulation task has emerged to generate high-fidelity trajectories in demand for downstream tasks to fill the gap between the scarce trajectory data and the widespread applications. Previous state-of-the-art methods build the spatiotemporal dependencies of trajectories with Graph Neural Networks (GNNs) under generative adversarial training, yielding better yet unstable trajectory quality. We observe that the unsatisfied synthetic trajectories are caused by the insufficient spatiotemporal modeling of road networks and trajectory semantics. In this paper, we propose a novel SpatioTEmporal GAted (STEGA) framework equipped with semantic -aware graph learning for traffic trajectory simulation to enable the explicit modeling of spatiotemporal dependencies throughout the learning pipeline. On the one hand, STEGA employs a graph encoder with the semantics of road networks for the spatial points of a trajectory, together with a time encoder for the time points. On the other hand, STEGA devises two spatiotemporal gates with the semantic graphs for the predictions of the future trajectory. Boosted by the semantic -aware graph learning, the proposed STEGA outperforms the state-of-the-art counterparts consistently at both macro- and micro -level metrics on two datasets. Elaborate ablation studies and downstream tasks of the synthetic trajectories further demonstrate the superiority of STEGA. Our code is available at https://github.com/Star607/STEGA.

There are a large number of airplanes currently being operated, in which the ventilation system needs to be improved to more effectively remove air contaminants. A potential approach is to adjust the supply air directions with the use of simple airflow deflectors. This study proposed a method for optimizing the supply air direction of ventilation in aircraft cabins based on the Re-field synergy index and Bayesian optimization. A validated numerical model was used to calculate the air distribution and air contaminant transport in a single-row single-aisle aircraft cabin to obtain the Re-field synergy values. The Bayesian optimization approach was used to identify the supply air direction which maximizes the Re-field synergy, namely, maximizes the mass transfer effectiveness. Finally, the air contaminant transport in a 7-row single-aisle aircraft cabin with the optimized supply air direction was evaluated to demonstrate the enhancement of ventilation performance. The results show that the proposed method based on the Re-field synergy index and Bayesian optimization can efficiently optimize the supply air direction in order to enhance the air contaminant removal in aircraft cabins. In the 7-row single-aisle aircraft cabin, the optimized supply air direction can reduce the average air contaminant concentration in the breathing zone of the passengers by up to 23 %.

Indoor environmental quality plays a crucial role in determining the overall quality of life. This study aims to develop a design optimization approach for the floor plan of public housing buildings with modular flat design in Hong Kong, with focus on enhancing natural ventilation, reducing noise levels, and improve daylighting conditions. The evaluation of these environmental factors was conducted using deep neural network models and a mathematically based Calculation of Road Traffic Noise model. A general floor plan representation was developed for three- and four-winged structures of public housing buildings. An optimization approach utilizing Bayesian optimization was applied to three studied cases: Hung Shing House, Hung Hei House, and Cheung Tai House. The optimization process resulted in an average 41.5% improvement in average natural ventilation rate. The optimized building shapes effectively served as noise barriers, leading to an average reduction of 20% in average noise levels. The existing window configurations of each unit type under the modular flat design already provided sufficient daylighting, resulting in only a minor improvement from the optimization process.

Fast calculation of person-to-person particle transport is essential for accelerating the evaluation and design of air distribution for reducing the risk of infection. This study developed a Lagrangian-Markov chain model with coarse grids for fast prediction of person-to-person particle transport in complex indoor environments. Detailed procedures and parameter determination approaches were developed. The proposed Lagrangian-Markov chain model was first validated with experimental data in two real-life cases of person-to-person particle transport, one in an aircraft cabin and the other in a COVID-19 isolation ward. The computing speed of the proposed model was then compared with the flux-based Markov chain, Eulerian, and Lagrangian models. The results show that the proposed Lagrangian-Markov chain model can predict person-to-person particle transport reasonably well in real-life cases with complex geometry and airflow fields. In terms of computing speed, the proposed Lagrangian-Markov chain model with coarse grids can be tens to hundreds of times faster than the three existing models for the two evaluation cases. With its fast computing speed, the Lagrangian-Markov chain model can be applied in the fast design of air distribution for real-life complex indoor environments.

In occupied indoor environments, 222 nm far-UVC is a secure and effective approach for controlling the spread of infectious bioaerosols. To enhance the disinfection effectiveness, it is crucial to carefully design the placement of far-UVC lamps, by considering the impact of airflow pattern, bioaerosol distribution, and irradiance exposure. Using CFD alone for designing the locations of multiple lamps presents challenges due to the high computational cost resulting from the high number of potential combinations. Therefore, this study combined CFD with the Bayesian optimization method to improve the computational efficiency, enabling customized design for complex scenarios. The proposed method was first validated with experimental data and the results from the grid search method. Next, the validated method was applied to design the placement of multiple far-UVC lamps in a railway compartment. In the studied railway compartment, just 1 h of optimization led to 1.05-fold-2.87-fold improvement in bioaerosol disinfection efficiency compared to random selection and uniform distribution. This optimization method for customized location design can maximize the utilization of far-UVC lamps, reducing equipment investment and energy consumption while ensuring effective disinfection.

Recently, there has been a surge of interest in hypergraph neural networks (HNN) due to their excellent ability to represent non-pair-wise interactions, such as communities in social networks, partnerships in academic cooperation networks, and biochemical interactions in biological networks. Existing works, which classify nodes of hypergraphs, mainly focus on the homogeneous hypergraph. However, the non-uniform and heterogeneous hypergraph is more common in realistic scenarios. Worse yet, homogeneous hypergraph approaches show limited accuracy on the heterogeneous hypergraph. In this paper, we propose a multi-channel hypergraph convolution framework (HHNN) to classify the nodes in the large, non-uniform, and heterogeneous hypergraph. Our framework is composed of two modules: The first decomposition module transforms arbitrary heterogeneous hypergraph to bipartite-form hypergraph (bfHG) so that we can introduce hypergraph structure more efficiently than the other methods. Second, we devise a multi-channel attention hypergraph convolution module (HHCNN), aiming at fusing various information from different kinds of nodes. We conduct experiments on four real-world heterogeneous hypergraph datasets, and the results show that our framework significantly improves the accuracy compared with six state-of-the-art approaches.

In the fifth wave of the COVID-19 epidemic in Hong Kong in early 2022, the large number of infected persons caused a shortage of ambulances and transportation vehicles operated by the government. To solve the problem, taxi drivers were recruited to transport infected persons to hospitals in their taxis. However, many of the drivers were infected after they began to participate in the plan. To tackle this issue, the present study numerically evaluated the effectiveness of several intervention measures in reducing the infection risk for taxi drivers. First, experiments were conducted inside a car to validate the large-eddy simulation (LES)-Lagrangian model for simulation of particle transport in a car. The validated model was then applied to calculate the particle dispersion and deposition in a Hong Kong taxi with intervention measures that included opening windows, installing partitions, and using a far-UVC lamp. The results show that opening the windows can significantly reduce the driver's total exposure by 97.4 %. Installing partitions and using a far-UVC lamp can further reduce the infection risk of driver by 55.9 % and 32.1 %, respectively. The results of this study can be used to support the implementation of effective intervention measures to protect taxi drivers from infection.

Correctly predicting indoor particle transport is crucial for the design of air distribution to reduce the transmission of airborne infectious diseases. In recent years, a model for predicting particle transport based on the Markov chain technique has emerged, with the advantage of fast computing speed. However, the turbulence diffusion modeling approach in this Markov chain model is semi-empirical without justification. Thus, the present study aims to increase the accuracy of Markov chain model in predicting particle transport by modifying the turbulence diffusion. Laboratory experiments on transient particle transport under unsteady periodic ventilation were conducted to validate the proposed model. The results show that the improved Markov chain model can correctly predict transient particle transport in comparison with the experimental data and was more accurate than the existing model in predicting transient particle transport.

In this paper, we strive to develop an interpretable GNNs' inference paradigm, termed MSInterpreter, which can serve as a plug-and-play scheme readily applicable to various GNNs' baselines. Unlike the most existing explanation methods, MSInterpreter provides a Message-passing Selection scheme(MSScheme) to select the critical paths for GNNs' message aggregations, which aims at reaching the self-explaination instead of post-hoc explanations. In detail, the elaborate MSScheme is designed to calculate weight factors of message aggregation paths by considering the vanilla structure and node embedding components, where the structure base aims at weight factors among node-induced substructures; on the other hand, the node embedding base focuses on weight factors via node embeddings obtained by one-layer GNN.Finally, we demonstrate the effectiveness of our approach on graph classification benchmarks.