Traditional and new geotechnical engineering problems such as compressed air energy storage, intercepting water with compressed air, carbon dioxide sequestration and oil and gas underground reserve project are all involving air-water two-phase flow and stress coupling problems. For this engineering reality, based on the weak coupling theory of gas-water two-phase seepage and stress in unsaturated soil, a air-water two-phase percolation-stress coupling calculation program based on coupled TOUGH2 and FLAC3D was developed. The calculation program can simulate real air-water two phase flow, and can investigate the gas-water interaction of seepage process. The calculation program considers the direct interaction between gas-water two-phase seepage and soil skeleton deformation, reflects the process of porosity, permeability, capillary pressure and the change of soil physical and mechanical parameters, and achieve a more perfect gas-water two-phase seepage-stress coupling analysis. Furthermore, by comparing with classical drainage test and model test, it is verified that the program can accurately simulate the gas-water two-phase flow-stress interaction.

A bearing RUL prediction method based on multi-scale features and attention mechanism was proposed aiming at the problem that the previous remaining useful life (RUL) prediction methods were insufficient in mining bearing degradation information and ignored the difference in the contribution of different features, which affected the prediction accuracy. Several time-domain and frequency-domain features of the original bearing vibration signal at multiple scales were calculated as the input feature set. The multi-scale feature set was input into the network, and the attention module was used to adaptively assign the best weights to different features. Then the convolutional neural network (CNN) module was used for deep feature extraction and multi-scale feature fusion. The RUL prediction value was obtained through the feedforward neural network (FNN) module mapping. The proposed method was applied to the public bearing datasets for comparative studies. Results showed the superior prediction performance of the proposed method.

A task allocation method for Internet of vehicles spatial crowdsourcing with privacy protection was proposed under the blockchain architecture in order to solve the problem that centralized spatial crowdsourcing server in the traditional spatial crowdsourcing of Internet of vehicles was untrusted and vulnerable to malicious attacks, which posed a great threat to users’ privacy. A distributed and trusted spatial crowdsourcing system of Internet of vehicles was designed based on the blockchain technology. The multi-key homomorphic encryption algorithm was adopted to distribute tasks, which supported task allocation of location ciphertext data of different vehicle users (keys). Then the possibility of privacy disclosure of vehicle users was reduced. The experimental results show that the proposed method can effectively protect users’ privacy information, reduce the computing overhead of task allocation by 34.3% compared with the existing methods, and improve the efficiency of task allocation.

In order to clarify the characteristics of the retaining system of narrow-long deep foundation pit in soft soil in Nansha Port Area, the behavior of retaining system for foundation pit excavation was studied, according to the measured data of Guangzhou deep and narrow foundation pit with diaphragm wall and internal support as retaining system. Results show that the maximum lateral wall displacement ranges from 0.07%H to 0.38%H, with a mean value of about 0.22%H, where H is the excavation depth. The location of the maximum wall displacements H_δm is between H?6 andH+3, and most are above the excavation surface. The wall deformation mainly occurs in the excavation stage of the second and third layers of soil, accounting for 32.6% and 40.1% of the cumulative deformation respectively. Foundation pit excavation has depth effect, and layered excavation of deep foundation pit is very important for wall deformation control. The main influence depth of wall deformation is about twice the excavation depth of foundation pit, and the spatial effect is significant. The variation characteristics of the vertical reinforcement stress and the lateral displacement of the wall are basically similar. With the excavation depth of the foundation pit, the maximum position moves down gradually, revealing the dynamic adjustment process of the wall deformation and stress. The axial force of the support reaches the maximum value after about two weeks after the support is erected, which shows real-time with the excavation of the foundation pit. The axial force of the multi-layer support structure is adapted dynamically with the excavation and support process of the foundation pit to coordinate the deformation development. When the excavation of the foundation pit is completed, the ultimate stable axial force of reinforced concrete support is about 0.73 times of the design value, the axial force of the first and the second steel support is 0.40 and 0.31 times of the design value, and the steel support design is conservative. On the premise of ensuring the stability of the foundation pit, the optimal design of a support scheme can be considered. Research results have important practical significance for the subsequent safety prediction of similar foundation pits in this area, as well as the guidance of similar engineering design and construction parameter optimization.

A ship object detection algorithm was proposed based on a multi-head self-attention (MHSA) mechanism and YOLO network (MHSA-YOLO), aiming at the characteristics of complex backgrounds, large differences in scale between classes and many small objects in inland rivers and ports. In the feature extraction process, a parallel self-attention residual module (PARM) based on MHSA was designed to weaken the interference of complex background information and strengthen the feature information of the ship objects. In the feature fusion process, a simplified two-way feature pyramid was developed so as to strengthen the feature fusion and representation ability. Experimental results on the Seaships dataset showed that the MHSA-YOLO method had a better learning ability, achieved 97.59% mean average precision in the aspect of object detection and was more effective compared with the state-of-the-art object detection methods. Experimental results based on a self-made dataset showed that MHSA-YOLO had strong generalization.

Two typical cases including the stiff system of ordinary differential equations ROBER problem and the steady-state mixture fraction equation in jet flame were selected in order to efficiently embed the complex physicochemical information of turbulent combustion into physics-informed neural networks (PINNs). The potential of PINNs in solving combustion chemical differential equations was explored. Results show that the PINNs model can correctly capture the evolution of the zero-dimensional stiff reaction system. PINNs solution accorded well with the conventional numerical solution for steady jet flame. The selection of residual points was particularly important for solving complex differential equations in the field of combustion and chemistry, which should be considered based on the specific configuration in detail.

Composite materials with different proportions were fabricated in order to characterize the modification effect of polydimethylsiloxane (PDMS) in metakaolin-based geopolymers and improve their waterproof and corrosion resistance. Contact angle measurement, thermogravimetric and strength tests were performed to characterize the wettability, water retention and mechanical strength of the material. Mercury intrusion porosimetry (MIP), scanning electron microscopy (SEM), backscattering (BSE), and energy dispersive (EDS) tests were performed to analyze the pore distribution, microstructure, and chemical composition. Results show that PDMS can hydrophobically modify the geopolymer gel extensively and uniformly, and the contact angle is about 130° when the mass ratio of PDMS to MK (m_PDMS/m_MK) is 0.025. Adding the silane coupling agent and drying treatment can improve hydrophobicity. A proper mass ratio of PDMS (0.01≤m_PDMS/m_MK≤0.025) can improve the strength and toughness of the geopolymer, because the gel structure of the composite is denser. PDMS enhances the binding force of geopolymers to moisture at low temperature, which is significant for reducing drying shrinkage.

A fast algorithm for all zero block detection was proposed in order to reduce the computational complexity. A fixed threshold was derived based on the hard decision quantization formula in order to detect genuine all zero blocks. Pseudo all zero blocks were further detected by the adaptive threshold related to the transform block size and quantization parameter (QP). The decision was made based on the fully connected neural network (FCNN) for the remaining blocks by extracting eight features that were closely related to the result of quantization. The experimental results showed that the proposed fast algorithm achieved up to 7.382% and 7.237% coding complexity saving under Low Delay B and Random Access configurations with only 0.458% and 0.575% performance loss on average, respectively.

A hyperspectral anomaly detection algorithm based on fractional Fourier transform (FrFT) and total variation regularization constraint was proposed aiming at the challenge of insufficient utilization of spatial information and contamination of background dictionary in low-rank and sparse representation based hyperspectral anomaly detection algorithms. The high-dimensional image data was mapped to multiple subspaces through the clustering algorithm. A pure background dictionary was obtained by constructing the FrFT-RX operator in order to enhance the discrimination between anomalies and background. A total variation regularization constraint was introduced into the low-rank and sparse representation model in order to describe the spatial features of background and anomalies in the intermediate domain after FrFT transformation. The optimal solution of the model was obtained by using the alternating direction method of multipliers. The anomaly detection results were obtained. The experimental results on three real hyperspectral datasets demonstrate that the proposed algorithm has a higher detection rate and a lower false alarm rate compared with the other five anomaly detection algorithms.

An NFT-based method for assetization of physical assets on blockchain was proposed aiming at the problem that the application of non-fungible token (NFT) was currently limited to virtual digital assets, such as digital collectibles, cryptographic artworks and game props, which was hard to cover widely existing physical assets. Smart contracts and automatic safe deposit boxes were combined to achieve strong bindings between off-chain physical assets and on-chain virtual assets based on the ability of NFT to confirm the rights of digital assets. A digital asset must be minted or circulated on blockchain (in virtual world) after its physical entity is locked off blockchain (in physical world). The corresponding NFT must be destroyed on blockchain before a physical asset is circulated off blockchain. The method avoids cross-domain double-spending attacks between the virtual world and the physical world, and offers a secure and trustworthy approach to transfer value and confirm ownership of physical assets on blockchain, which provides a feasible technical way for future circulation and trading of physical assets in the metaverse.

A missing data repair method based on spatio-temporal fusion graph convolutional network was proposed in order to solve the problem of insufficient traffic flow characteristics mining by existing spatio-temporal correlation repair method. Two types of functions were used to respectively calculate the temporal autocorrelation coefficient and spatial correlation coefficient of traffic flow data by analyzing the spatio-temporal characteristics of traffic flow. The deployment position of the traffic detector was used as a node to form a geometric topology graph, and a spatio-temporal fusion matrix was constructed by linear fusion rules, which replaced the adjacency matrix of the graph convolution input layer to capture the fine-grained spatio-temporal relationship of the traffic flow. The lightweight one-dimensional convolution layer was used to learn the temporal characteristics of multi-channel time series vectors in order to speed up the convergence speed of the model. The graph convolutional layer was used to learn the spatial characteristics of traffic flow data. A spatio-temporal fusion graph convolution network repair model was constructed. The experimental results show that the repair accuracy and model convergence speed of the method in multi-detector scenarios were improved compared with other repair methods, which can effectively repair the missing traffic flow data.

Five expansion pattern types and eighteen structure types were proposed, as well as the "four elements" of the expansion system based on the investigated cases of underground space expansion projects in order to evaluate the disturbance effect of urban underground space network extension construction. The total evaluation objective composed of four sub-objectives, including extensional structure, existing structure, stratum and surrounding environment, and six stress and displacement evaluation indexes were determined based on the analysis of the expansion disturbance conduction path and the internal working mechanism of its elements. The analytic hierarchy process (AHP) was used to construct the judgment matrix. The evaluation index weight was determined by expert scoring. Then the control value grading method was applied to determine the membership degree of the evaluation index, thereby constructing a five-level urban underground space network expansion disturbance classification evaluation standard. The application of the above evaluation method was conducted. The order of expansion disturbance degree of different expansion methods is expansion by small>vertical expansion>horizontal expansion >connected connection, and the order of expansion disturbance degree of different stratum is soft soil >loess >alluvial-diluvial strata>rock. The expansion disturbance degree will decrease significantly with the increase of the proximity distance for horizontal expansion.

A method based on attention mechanism and multi-scale information fusion was proposed to resovle the problem of low accuracy of the traditional single shot multibox detector (SSD) algorithm in detecting small targets. The algorithm was applied to the vehicle detection task. The feature maps of the target detection branch were fused with 5 branches and 2 branches respectively, combining the advantages of the shallow feature map and the deep feature map. The attention mechanism module was added between the basic network layers to make the model pay attention to the channels containing more information. Experimental results showed that the mean average precision of the self-built vehicle data set reached 90.2%, which was 10.0% higher than the traditional SSD algorithm. The detection accuracy of small objects was improved by 17.9%. The mAP on the PASCAL VOC 2012 dataset was 83.1%, which was 6.4% higher than the current mainstream YOLOv5 algorithm. The detection speed of proposed algorithm on the GTX1 660 Ti PC reached 25 frame/s, which satisfied the demand of real-time performance.

At present, the path planning methods based on short-term traffic flow prediction mostly use historical and real-time traffic flow data, and the timeliness of the prediction needs to be improved. In response to this problem, a path planning method based on travel plan data (RPTP) was proposed. The method can proactively obtain the future traffic demand of travelers and provide more reasonable routes for vehicles. Based on the idea of “travel plan”, the overall framework of optimal path planning was designed on the basis of travel plan data. An estimation algorithm was constructed to calculate the road network density in multiple future periods using the route data of the travel plan. The multi-period road network density was integrated using the spatial superposition method, based on which the heuristic function of the D*Lite algorithm was improved. Several simulation experiments were carried out using the SUMO simulation platform, and the simulation effects produced by the RPTP method were compared with that of static path planning method (SPP) and rolling path planning method (RPP) in the same condition. Experimental results show that the RPTP method can improve the traffic efficiency of the road network and alleviate the road traffic congestion, which effectively verifies the superiority of the RPTP method.

A fire detection algorithm based on improved GhostNet-FCOS was proposed in view of the low detection accuracy and high complexity of existing fire detection algorithms. The algorithm was based on FCOS with reduced channel dimensions, and GhostNet was selected as the feature extraction network to implement a lightweight fire detection algorithm. Dynamic convolution was introduced to optimize the basic modules of the backbone without increasing width and depth, resulting in improved feature extraction ability for variable flames. A spatial attention module was introduced into the backbone network in order to optimize the expression of network spatial features. The definition of positive and negative samples and the regression loss function were improved to optimize the network’s attention to different areas in the ground truth box during the training process. The experimental results in self-built fire dataset and public dataset show that the algorithm has advantages in detection accuracy and model complexity. The detection accuracy of the algorithm in the self-built fire dataset was 90.9%, the amount of parameter was 4.58×10⁶, and the floating point operation was 31.45×10⁹.

A two-stage hot spot detection method of aerial infrared image was proposed to realize component level positioning and fine classification diagnosis of hot spot defects in infrared image, aiming at the problems of high cost, low efficiency and low accuracy of traditional inspection technology of photovoltaic power station. This method combined the traditional image processing technology with the deep learning method to further improve the accuracy and efficiency of defect diagnosis. Specifically, firstly, based on the difference between the gray values of the front and back scenes of aerial infrared images, a component segmentation method based on edge detection was proposed to extract the contour of photovoltaic components to achieve component level positioning. This method achieved the effective detection rate of photovoltaic components up to 99.3% with relatively small hardware requirements. Secondly, considering the differences in the causes, hazards and corresponding treatment methods of hot spots, an infrared defect classification model based on EfficientNet was proposed to finely classify the hot spots, so as to provide more accurate decision support for the operation and maintenance personnel of the power station. The model obtained hot spot classification accuracy of 97.0% when it occupied 20.17 MB. Through experimental comparison and analysis, it is demonstrated that the proposed method has greatly improved the efficiency and accuracy of defect diagnosis.

The local pressure in the pipe network did not meet the minimum service water pressure specified by the pipe network due to closing the boundary valve during the implementation of district metered area (DMA) zoning. In order to solve the problem, the optimal partition boundary pipes (BPs) were obtained by taking the number of BPs and its average flow, pipe diameter and length after the spectral clustering algorithm as the objective function, through the function gamultiobj in MATLAB. A series of different minimum service water pressures were set as constraints. The average water age of the nodes and the zoning cost after the partition were taken as the objective function, and the Pareto optimal solution was obtained by gamultiobj optimization calculation. The optimal layout scheme of the equipment on the BPs was determined according to the solution. The simulated annealing algorithm was used to find the best pipe replacement scheme to make the water pressure of the pipe network meet the requirements. Taking the Modena pipe network with only 0.09 m depressurization space as an example, the water quality of users with large volume flow rate and end users after zoning was improved on the basis of the successful completion of zoning. The proposed method can realize the DMA zoning of the pipe network under the minimum depressurization space, and the normal operation of the pipe network can still be ensured after the partition.

In order to explore the applicability of microbial induced calcium carbonate precipitation (MICP) to cement calcareous sand in seawater environment and the wet-dry cycle resistance of MICP cemented bodies, the calcareous sands were cemented in seawater and freshwater environment respectively, and the wet-dry cycles were carried out in seawater environment. The element and mineral composition of the cemented bodies were analyzed based on the energy dispersive spectroscopy(EDS), X-ray diffraction(XRD). In addition, the relationships between mechanical properties, mass loss and wet-dry cycle were established through unconfined compressive strength test, weighing; and the weakening mechanism of the wet-dry cycle on samples were analyzed by scanning electron microscope (SEM). Results show that, in seawater environment, the cementation effect of MICP on calcareous sand is better than that in freshwater environment. The resistance to wet-dry cycle of calcareous sand cemented in seawater environment is larger than that cemented in freshwater environment. After 21 wet-dry cycles, the strength of cemented calcareous sand in seawater and freshwater environment decreased to 30% and 7.53% of the original samples, respectively. The wet-dry cycle reduces the particle surface roughness and intergranular cementation strength, which is manifested in the reduction of strength and stiffness of cemented calcareous sand in macro-characteristics.

A neural network based segmentation model MS²Net was proposed to automatically and accurately extract regions of interest from medical images. In order to better extract context information, a network architecture combining convolution and Transformer was proposed, which solved the problem that traditional convolution operations lacked the ability to acquire long-range dependencies. In the Transformer-based context extraction module, multi-head self-attention was used to obtain the similarity relationship between pixels. Based on the similarity relationship, the features of each pixel were fused, so that the network had a global view, while the relative positional encoding enabled Transformer to retain the structural information of an input feature map. Aiming at making the network adapt to different sizes of regions of interest, the multi-scale features of decoders were used by MS²Net and a multi-scale attention mechanism was proposed. The group channel attention and the group spatial attention were applied to a multi-scale feature map in turns, so that the reasonable multi-scale semantic information was selected adaptively by the network. MS²Net had achieved better intersection-over-union than advanced methods such as U-Net, CE-Net, DeepLab v3+, UTNet on both ISBI 2017 and CVC-ColonDB datasets, which reflected its excellent generalization ability.

The segmentation model construction and training based on single source data may lead to insufficient segmentation accuracy due to the defects of various imaging methods in medical images. Aiming at this problem, a medical image segmentation method based on multi-source information fusion was proposed. The FFDM and DBT data sources in the breast tumour microcalcification cluster lesion were used as examples to verify the effectiveness of the proposed method. The Yolov4 region candidate network was used to screen the suspicious regions of the FFDM data. DBT image was preprocessed by using the suspicious region information. The preprocessed DBT image was used as the input of the improved U-Net model to achieve lesion segmentation. Finally, through the fusion strategy of fault segmentation results based on sequential similarity discrimination, the multi-slice results in DBT were combined to complete the final lesion segmentation. True positive rate of 98.52%, false positive rate of 10.45% and accuracy of 94.07% were obtained from the FFDM and DBT data of 20 patients by using this method. Results show that the medical image segmentation method based on multi-source information fusion can effectively utilize the advantages of multi-source data, and achieve the rapid and accurate segmentation of lesions. The method can provide a novel solution for intelligent medical image diagnosis and treatment.