A thorough analysis and cross-comparison of recent relevant works was provided, outlining a closed-loop process for EEG data analysis based on deep learning. EEG data were introduced, and the application of deep learning in three key stages: preprocessing, feature extraction, and model generalization was unfolded. The research ideas and solutions provided by deep learning algorithms in the respective stages were delineated, including the challenges and issues encountered at each stage. The main contributions and limitations of different algorithms were comprehensively summarized. The challenges faced and future directions of deep learning technology in handling EEG data at each stage were discussed.

A new multimodal sentiment analysis model (MTSA) was proposed on the basis of cross-modal Transformer, aiming at the difficult retention of the modal feature heterogeneity for single-modal feature extraction and feature redundancy for cross-modal feature fusion. Long short-term memory (LSTM) and multi-task learning framework were used to extract single-modal contextual semantic information, the noise was removed and the modal feature heterogeneity was preserved by adding up auxiliary modal task losses. Multi-tasking gating mechanism was used to adjust cross-modal feature fusion. Text, audio and visual modal features were fused in a stacked cross-modal Transformer structure to improve fusion depth and avoid feature redundancy. MTSA was evaluated in the MOSEI and SIMS data sets, results show that compared with other advanced models, MTSA has better overall performance, the accuracy of binary classification reached 83.51% and 84.18% respectively.

A power source contribution coefficient-based optimized method was proposed to increase the reasonability and efficiency of the power source operation mode selection in relay setting calculation. The concept of power source contribution coefficient and corresponding calculation method based on the node impedance matrix were presented. A system operation mode selection method based on the power source contribution coefficients of the protected line and infeed branches was developed. The effectiveness of the proposed method was verified through the case study of New England 39-Bus test system. Results show that by utilizing the proposed infeed coefficient calculation method, the specific operation modes of different power sources corresponding to the maximal and minimal infeed coefficient values can be determined precisely. This process does not need permutation or combination, which improves the efficiency and accuracy of the automatic relay setting calculations.

A new composite bent cap consisting of a shell made of steel plate and ultra-high-performance concrete (UHPC) and cast-in-place core normal concrete (NC) was proposed in order to realize the assembly and rapid construction of ultra-large-scale bent cap for urban viaducts or highway reconstruction and expansion projects. Parametric analysis of different UHPC and steel plate thickness was conducted in order to analyze the influence of the thickness of UHPC and steel mold plate on its stress performance. Results showed that the stiffness of the shell was affected by the thickness of UHPC and steel plate and their ratio together under the action of self-weight. The thicker the UHPC and steel plate are, the better the stress performance of the shell is, but the economy will be reduced when tensioning prestress and casting concrete. It is recommended to use UHPC thickness of 70 mm and steel plate thickness of 6 mm. A piece of 1∶2.5 scaled-down model was designed and static loading test was conducted in order to verify the feasibility and safety of this scheme. Results show that the new UHPC-NC composite bent cap has good force performance and high safety reserve, which can provide reference for the assembly construction of bent cap.

A spatial-temporal multi-graph convolution traffic flow prediction model by integrating static and dynamic knowledge graphs was proposed, as current traffic flow prediction methods focus on the spatial-temporal correlation of traffic information and fail to fully take into account the influence of external factors on traffic. An urban traffic knowledge graph and four road network topological graphs with distinct semantics were systematically constructed, drawing upon the road traffic information and the external factors. The urban traffic knowledge graph was inputted into the relational evolution graph convolutional neural network to realize the knowledge embedding. The traffic flow matrix and the knowledge embedding were integrated using the knowledge fusion module. The four road network topology graphs and the traffic flow matrix with fused knowledge were fed into the spatial-temporal multi-graph convolution module to extract spatiotemporal features, and the traffic flow prediction value was outputted through the fully connected layer. The model performance was evaluated on a Hangzhou traffic data set. Compared with the advanced baseline, the performance of the proposed model improved by 5.76%-10.71%. Robustness experiment results show that the proposed model has a strong ability to resist interference.

A multi-behavior aware service recommendation method based on hypergraph graph convolutional neural network (MBSRHGNN) was proposed to resolve the problem of insufficient high-order service feature extraction in existing service recommendation methods. A multi-hypergraph was constructed according to user-service interaction types and service mashups. A dual-channel hypergraph convolutional network was designed based on the spectral decomposition theory with functional and structural properties of multi-hypergraph. Chebyshev polynomial was used to approximate hypergraph convolution kernel to reduce computational complexity. Self-attention mechanism and multi-behavior recommendation methods were combined to measure the importance difference between multi-behavior interactions during the hypergraph convolution process. A hypergraph pooling method named HG-DiffPool was proposed to reduce the feature dimensionality. The probability distribution for recommending different services was learned by integrating service embedding vector and hypergraph signals. Real service data was obtained by the crawler and used to construct datasets with different sparsity for experiments. Experimental results showed that the MBSRHGNN method could adapt to recommendation scenario with highly sparse data, and was superior to the existing baseline methods in accuracy and relevance.

A dynamic 3D reconstruction technology for construction sites was proposed to ensure safety and efficiency in the construction process. A Binocular camera was deployed to scan the reconstruction site in 3D to obtain the model base and target activity trajectory. The YOLOv8 model was enhanced with an attentional scale sequence fusion (ASF) module to form the YOLOv8-ASF framework, which improved the accuracy and performance of the model, to solve the pain points such as target occlusion and target loss. The improved semi-global block matching (SGBM) algorithm was fused, and the YOLOv8-ASF-SGBM algorithm was integrated with the YOLOv8-ASF to achieve near-real-time target recognition and localization based on 2D images. The obtained depth information was used to 3D project the behavior trajectories of dynamic elements into the substrate, to realize the near-real-time and full-view monitoring of the real construction site. Experimental results show that the proposed technology reproduces the movement trajectory of construction dynamic elements in high-precision three-dimensional, and the relative error with the real motion trajectory of dynamic elements is less than 5%, which can realize high-precision full-view three-dimensional monitoring based on two-dimensional image and video information, and has good application scenarios and engineering value.

Existing research on the retrieval task of similar cases ignores the legal logic that the model should imply, and cannot adapt to the requirements of case similarity criteria in practical applications. Few datasets in Chinese for case retrieval tasks are difficult to meet the research needs. A similar case retrieval model was proposed based on legal logic and strong interpretability, and a case event logic graph was constructed based on predicate verbs. The statutory knowledge corresponding to various crimes was integrated into the proposed model, and the extracted elements were input to a neural network-based scorer to realize the task of case retrieval accurately and efficiently. A Confusing-LeCaRD dataset was built for the case retrieval task with a confusing group of charges as the main retrieval causes. Experiments show that the normalized discounted cumulative gain of the proposed model on the LeCaRD dataset and Confusing-LeCaRD dataset was 90.95% and 94.64%, and the model was superior to TF-IDF, BM25 and BERT-PLI in all indicators.

The development and research status of YOLO algorithm in traffic object detection were systematically summarized from the perspective of the three core elements of 'people-vehicle-road' in order to comprehensively analyze the important role of YOLO (You Only Look Once) algorithm in improving traffic safety and efficiency. The commonly used evaluation indexes of YOLO algorithm were outlined, and the practical significance of these indexes in traffic scenarios was elaborately expounded. An overview of the core architecture of YOLO algorithm was provided, its development process was traced, and the optimization and improvement measures in each version iteration were analyzed. The research status and application scenarios of YOLO algorithm for traffic object detection were sorted out and discussed from the perspective of the three traffic objects 'people-vehicle-road'. The limitations and challenges of YOLO algorithm in traffic object detection were analyzed, and corresponding improvement methods were proposed. Future research focuses were anticipated, providing a research reference for the intelligent development of road traffic.

A multi-distortion type underwater image enhancement algorithm based on improved CycleGAN was proposed, aiming at the difficulties of underwater image blurring, low contrast and image distortion recognition caused by various factors such as scattering, absorption and color deviation. Firstly, in order to improve the image enhancement effect, Auto-Encoder+Skip-connection network structure was used in the generator of CycleGAN, and global color correction structure was added for global enhancement in terms of pixel as well as color, so as to better capture the color information in underwater images. Secondly, a multidimensional perceptual discriminator was designed to learn the global and local features of the image. This discriminator payed more attention to the local details of the image, effectively targeted scattering and color noise, perceived the image from a multidimensional space, and had a stronger ability to extract the features, thereby enhancing the accuracy of image discrimination. Finally, the experimental results on EUVP, UIEB and U45 datasets showed that the proposed method achieved better results, compared with other algorithms. In processing multi-distortion types of underwater images, the algorithm’s SSIM indicator was higher than that of the second place by an average of 1.57%, the PSNR indicator was higher by 1.836%, the UIQM indicator was higher by 1.324%, and the UCIQE indicator was higher by 1.086%. The proposed method performed well in processing color and noise details.

A multimodal feature fusion model based on non-smooth non-negative matrix factorization (nsNMF-PCNN-GRU-MSA) was proposed to address the problems of poor generalisation ability, single feature extraction mode and model uninterpretability in the fatigue driving detection methods. This model detected the level of driver fatigue by analyzing electroencephalogram (EEG) signals. A channel weighting module was designed in the shallow layer of the network, and the non-smooth non-negative matrix factorization (nsNMF) algorithm was introduced to compute the contribution of the electrode channels. A multimodal feature fusion module was designed in the middle layer of the network, where the Gramian angular field imaging method was introduced to map the 1D EEG data into a 2D image, and the spatio-temporal features of different modes were fused in parallel with the PCNN-GRU module. The multi-head self-attention (MSA) mechanism was fused in the deep layer of the network to complete the task of fatigue driving state classification. The experimental results showed that the fatigue detection accuracies of the model on the mixed samples of the SEED-VIG and SAD datasets were 93.37% and 90.78%, respectively, and the lowest accuracies for single-subject data were 86.60% and 85.59%, respectively, which were higher than those of the state-of-the-art models. The analysis method of mapping the feature activation values onto the brain topology map not only improves the interpretability of the model, but also provides a new perspective on fatigue driving detection.

A road pavement distress situation forecasting method employing graph convolutional networks was introduced, addressing the prediction problem of road pavement distress generation and deterioration. Firstly, a topological network was established through clustering algorithms, selecting the main influencing factors of the target pavement distress during its evolution. Subsequently, to enhance the expressive capability of the graph neural network for distress information, a graph topology enhancement method was employed, constructing views related to distress information from both static and dynamic aspects. Finally, an enhanced graph neural network (GNN) architecture was applied, by incorporating attention mechanisms in the view dimension to adjust the influence of different views and utilizing Transformer and GRU modules in the temporal dimension to enhance the predictive performance of the model for pavement distress states over extended time sequences. The internal calibration tests of the model, including ablation studies, multi-sample testing, and hyperparameter control group validation, demonstrated the applicability and stability of the proposed model. For the large and sparse pavement disease dataset, the mean absolute error of this model converged within 4.0, which was better than the results of the traditional prediction algorithms in terms of comprehensive performance.

An improved non-dominated sorting genetic algorithm II (NSGA-II) was proposed in order to address the challenges of time-consuming manual airspace sectorization and the difficulty in comparing the quality of different sectorization schemes. A three-dimensional multi-objective optimization model for sectorization was established by using a grid-region-sector hierarchy in order to balance controllers’ workload within sectors and reduce workload differences between sectors. A fitness evaluation operator, a probability-adaptive combination crossover operator and a dynamic mutation operator were incorporated in the NSGA-II algorithm in order to enhance the number of feasible solutions, solution diversity and computational efficiency. A simulation was conducted for the automatic 3D sectorization of Xi'an high-altitude airspace. Results showed that the optimized scheme improved workload balance within sectors by 37% and reduced inter-sector workload by 24% compared with the current sectorization configuration. The proposed improved NSGA-II provided a broader range of options for decision-makers with varying preferences compared with traditional weighted multi-objective optimization algorithms.

Traditional computational intelligence technology was found to lack real-time capabilities and adaptability, and computational intelligence technology based on parallel computing made computational efficiency improve and addressed the issue of compatible processing of multimodal information. From three branches of computational intelligence: neural networks, evolutionary algorithms, and swarm intelligence algorithms, the current states were reviewed on the integration of computational intelligence and big data-parallel computing. Problems present in parallel computing intelligence were summarized, and some thoughts were given to the development direction of related studies.

For the poor real-time detection capability of the current object detection model in the production environment of electronic components, GhostNet was used to replace the backbone network of YOLOv5. And for the existence of small objects and objects with large scale changes on the surface defects of electronic components, a coordinate attention module was added to the YOLOv5 backbone network, which enhanced the sensory field while avoiding the consumption of large computational resources. The coordinate information was embedded into the channel attention to improve the object localization of the model. The feature pyramid networks (FPN) structure in the YOLOv5 feature fusion module was replaced with a weighted bi-directional feature pyramid network structure, to enhance the fusion capability of multi-scale weighted features. Experimental results on the self-made defective electronic component dataset showed that the improved GCB-YOLOv5 model achieved an average accuracy of 93% and an average detection time of 33.2 ms, which improved the average accuracy by 15.0% and the average time by 7 ms compared with the original YOLOv5 model. And the improved model can meet the requirements of both accuracy and speed of electronic component surface defect detection.

A new arc additive manufacturing process—droplet+arc additive manufacturing (DAAM) technology was applied to manufacture aluminum alloy samples in order to improve the quality and the efficiency of aluminum alloy. A new droplet generation system (DGS) was applied instead of the conventional wire feeding system, which makes the material addition and arc energy independent of each other. The formed material is 2219 aluminum alloy, and a trace amount of Mg element was added through the DGS. A thin-walled structure was deposited using the DAAM system at a significantly higher deposition rate (160 $ {\mathrm{m}\mathrm{m}}^{3}/\mathrm{s} $) than conventional wire and arc additive manufacturing techniques. The microstructure of the cross section of the thin-walled structure was observed and analyzed. Results showed that the grain morphology of the thin-walled structure was dominated by columnar crystals and exhibited a periodic distribution of inner-layer columnar crystals and inter-layer equiaxed crystals. The average tensile strengths in the horizontal and vertical directions were 455.4 MPa and 417.0 MPa after T6 heat treatment, while the yield strengths were 342.2 MPa and 316.4 MPa, respectively. The comparison results with the previous studies show that the addition of Mg element increases the yield strength of 2219 aluminum alloy, but leads to a corresponding decrease in elongation.

The effects of leg structure design, foot-end design and sensor design on touchdown detection were comprehensively discussed by analyzing the existing legged robot touchdown detection methods. The touchdown method for direct detection of external sensors, the touchdown detection method based on kinematics and dynamics, and the touchdown detection method based on learning were summarized. Touchdown detection methods were summarized in three special scenarios: slippery ground, soft ground, and non-foot-end contact. The application scenarios of touchdown detection technology were analyzed, including the three application scenarios of motion control requirements, navigation applications, and terrain and geological sensing. The development trends were pointed out, which related to the four major touchdown detection methods of hardware improvement and integration, multi-mode touchdown detection, multi-sensor fusion touchdown detection, and intelligent touchdown detection. The specific relationships between various touchdown detection algorithms were summarized, which provided guidance for the development of follow-up technology for touchdown detection and specific applications of touchdown detection.

A lightweight classification model MB-CNN for optical coherence tomography (OCT) images was proposed to accurately and conveniently identify multiple types of fundus lesions. By reducing the number of convolution cores and adjusting the proportion of convolution blocks in each stage, a lightweight backbone network L-Resnet was designed, and the extraction of deep-layer semantic information was enhanced by deepening the network depth. The multi-scale convolution block MultiBlock was designed using depthwise seperable convolution, and the features of the lesion area was mined. Different convolution kernels were used to extract the lesions features of different sizes to improve the recognition ability of the network to the OCT image of the lesion. The feature fusion module FFM was constructed, and the shallow layer information and deep layer information were fused, the texture and semantic information of the pathological features were extracted, and the recognition ability of small target lesions was improved. Experimental result showed that the overall classification accuracy of MB-CNN in the three datasets of UCSD, Duke and NEH was 97.2%, 99.92% and 94.37% respectively, the amount of model parameters were significantly reduced. The proposed model can classify various fundus lesions.

A design strategy through appropriately opening the passing space interface in response to climate was proposed as a solution to high energy consumption and formal convergence within the high-speed railway station buildings. Performance simulation tools based on building information modeling (BIM) were used to build a typical model, and the opening time period for different climate zones were determined according to wind and thermal environment simulation analysis. Results show that it is feasible to open up the passing space interface, meeting the requirement of indoor thermal comfort, in the case of a typical summer calculation day in climate zones except in hot summer and warm winter zone (Guangzhou as an example). Meanwhile, during the particular time periods of a year, interface opening is beneficial to energy savings and emission reduction in station buildings, especially in hot summer and cold winter zone (Shanghai as an example) and cold zone (Beijing as an example). The energy-savings reached up to 44.8% and 32.2%, respectively, as well as carbon reduction rates of 36.1% and 21.3%. Hence, the proposed strategy has significant application potential in the green design schemes of high-speed railway station buildings and can explore ideas for regional expression of spatial forms.

A low-jitter high-speed charge-pump phase-locked loop (CPPLL) suitable for high-speed SerDes serial link was designed using 65 nm CMOS technology. Loop bandwidth and circuit structure of voltage-controlled oscillator (VCO), charge pump (CP), phase frequency detector (PFD) were optimized to reduce jitter caused by voltage ripple and internal noise. CPPLL can achieve a stable clock signal with the smaller jitter offset while meeting the wide frequency range and high speed requirements of SerDes link. Chip area including the entire pads is 0.309 mm². The measurement results show that CPPLL can generate a 10.9-12 GHz clock signal and exhibit a phase noise of ?111.47 dBc/Hz and a reference spur of ?25.14 dBc and a figure-of-merit (FoM) of ?223.5 dB at 10 MHz offset. It takes 600 μs to generate a stable 11.3 GHz clock signal, and its RMS jitter is 973.9 fs when the reference frequency is 706.25 MHz, which is approximately 0.065 UI. The power consumption is 47.3 mW at the supply voltage of 1.2 V. The proposed phase-locked loop (PLL) is suitable for high-speed communication link systems at 20 Gb/s and above.