82558 results about "Neural network nn" patented technology

The invention provides a news scene three-dimensional reconstruction and visualization method based on multi-source remote sensing data, and relates to the technical field of three-dimensional reconstruction, and the method comprises the steps: extracting multi-modal data feature points through employing a spatial position weighted iterative nearest point algorithm and a tensor projection algorithm; after bidirectional attention deep neural network processing and local geometric consistency constraint screening, registration multi-source remote sensing data is generated, data enhancement is carried out through a variational auto-encoder network and a generative adversarial network, the enhanced data is used for depth information generation and texture feature migration, an initial three-dimensional grid model is generated based on a Poisson reconstruction algorithm, and the initial three-dimensional grid model is constructed; and carrying out topological optimization by combining multi-view projection consistency constraint and a grid optimization algorithm to obtain an optimized three-dimensional scene model, setting a space division function, carrying out space division, constructing a grade model and a depth region, generating a scene semantic segmentation map, and updating a visual scene.

The invention relates to a dynamic data pipeline construction method based on artificial intelligence and multi-modal data processing, and belongs to the technical field of data fusion treatment and artificial intelligence. According to the method, a multi-modal deep learning model is adopted, semantic analysis is performed on business documents, user requirements and function specification instructions, and key indexes, data requirements and logic constraints are extracted; constructing a multi-modal knowledge graph, performing unified representation and feature extraction on structured and unstructured data resources, and performing embedded learning by using a graph neural network; based on a deep semantic matching model and a hybrid recommendation system, realizing accurate matching of business requirements and data resources; modeling generation of a data processing flow into a Markov decision process by adopting a reinforcement learning strategy, and automatically generating an optimized data processing workflow; the adaptability of the model to new business requirements is improved by utilizing a meta-learning algorithm; and dynamic scheduling and execution of tasks are realized through a workflow engine. According to the invention, the data processing efficiency can be improved.

The invention relates to the technical field of industrial automation control, and discloses a multi-machine collaborative industrial robot intelligent scheduling system and an application method.The system comprises a data acquisition module for acquiring real-time motion data through a multi-modal sensor; the feature fusion module is used for generating environment dynamic feature data based on graph neural network fusion data; the strategy generation module is used for generating collaborative strategy parameters by utilizing a collaborative reinforcement learning model; the trajectory optimization module is used for constructing a multi-target trajectory optimization model to plan an optimal synchronous trajectory; the motion control module is used for establishing a layered motion control model to realize cooperative motion control; a dynamic resource allocation module can also be included to optimize resource allocation. The application method sequentially executes corresponding steps according to the modules of the system. According to the method, the environment can be accurately sensed, the cooperation strategy can be optimized, the efficient track can be planned, accurate motion control and reasonable resource allocation can be realized, and the cooperative motion control capability and the production efficiency of the industrial robot in the complex environment can be effectively improved.

Provided is a road crack detection method, a medium and a product. The lightweight YOLO-MCS road crack detection model is constructed based on an improved yolov8 network. A method of improving the yolov8 network includes: replacing a feature extraction network backbone with a lightweight convolutional neural network MobileNetV3, and embedding a Coordinate Attention Mechanism (CA) module in the lightweight convolutional neural network MobileNetV3; adding a small target detection layer and a Squeeze and Excitation (SE) module at a neck end; and introducing a loss function of power Intersection over Union (IoU) into a head prediction structure. The road crack detection method overcomes the problem that the existing detection algorithm has low detection precision for small cracks and is difficult to be applied to edge devices with limited computing resources.

The invention provides a home abnormal state signal detection method and system based on multi-modal sensing, and relates to the technical field of detection, and the method comprises the steps: collecting human body motion, acoustics and environment parameters through a millimeter wave radar, an acoustic sensor and an environment sensor, inputting the parameters into a deep fusion network, a spatio-temporal attention mechanism and a multi-scale convolutional neural network are used to extract a spatio-temporal feature sequence, and joint probability features are formed in combination with an acoustic feature spectrum analysis result. And inputting the joint probability features into a time sequence knowledge graph, and obtaining scene adaptive features after processing by a graph attention network and a double-flow auto-encoder network. And finally, the multi-task learning network is combined with the risk propagation neural network and the recursive neural network to carry out anomaly prediction and risk level evaluation, and is combined with historical early warning information to output an early warning strategy. According to the invention, multi-modal data can be effectively fused, the accuracy and reliability of home abnormal state detection are improved, and a more accurate risk assessment and early warning strategy is provided.

The invention discloses a ring main unit monitoring method and system based on the Internet of Things, and the method comprises the steps: building a multi-source data collection network of a ring main unit, extracting the multi-dimensional time-frequency domain features of a signal through an edge computing device, generating a multi-source feature data set, and constructing a dynamic digital twinborn model, real-time mapping of the running state of the ring main unit is achieved through a physical-virtual synchronization algorithm, and twin data volumes are output; according to the node influence coefficient of the twin data volume, analyzing a topological structure of a ring network system by adopting a graph neural network, identifying a key fault propagation path and generating a priority monitoring node list; based on the priority monitoring node list and the twin data volume, dynamically generating an adaptive monitoring strategy through a reinforcement learning algorithm, executing the adaptive monitoring strategy and collecting feedback data in real time, and completing closed-loop iteration of the monitoring strategy through an online learning algorithm; through closed-loop control of multi-source perception-digital twinning-intelligent decision, high efficiency, intelligence and reliability of ring main unit monitoring are realized.

The invention provides a building dynamic structure health monitoring method. The method comprises the steps that original data of a multi-mode sensor is acquired and preprocessed; constructing topology sensing network representation reflecting the spatial distribution and structure connection relation of the sensors; extracting multiple spatio-temporal scale features; local-to-global structure state mapping is realized by using a recurrence plot evolution predictor; integrating structural dynamics knowledge into the neural network model to perform physical constraint correction; and constructing a multi-dimensional health index set and outputting damage probability evaluation. According to the invention, accurate mapping from the local dense monitoring area to the global structure state under the non-uniformly distributed sensor network is realized, and the accuracy and reliability of structure health monitoring are improved.

The invention relates to the technical field of twinborn model simulation, and discloses a twinborn model simulation method and system for hot working of large forgings. The method comprises the following steps: collecting and preprocessing process parameters, quality data and environment information in a multi-source manner, and obtaining hot working characteristic data; performing correlation analysis to construct a process knowledge graph; calculating the distribution of a temperature field, a stress field and an organization field by using a self-sensing multi-field coupling var value neural network; comparing and analyzing to obtain deviation data and correction parameters; adjusting a network parameter optimization prediction result; and executing multi-objective optimization calculation, and generating a whole-process technological parameter and a control instruction. According to the method, full-process multi-physics field coupling calculation from smelting, casting, forging and pressing to heat treatment can be achieved, model parameters are dynamically adjusted according to real-time production data, the optimal process scheme is generated, and therefore the manufacturing quality and efficiency of large forgings are improved.

The invention relates to a multi-modal dynamic optimization educational resource recommendation system and method, and the system comprises the following modules: a multi-modal data collection module integrates video behaviors, answer tracks, physiological signals and other data through edge calculation, and constructs a learning feature map; the student portrait module adopts an LSTM-Attention network in combination with a graph neural network to dynamically model knowledge mastery and learning styles; the resource matching engine realizes multi-objective optimization of knowledge gain, cognitive load and interest matching based on reinforcement learning and knowledge graph analysis; the tag adaptive module dynamically adjusts resource weights through causal inference and comparative learning, the personalized recommendation module generates a dynamic learning path and pushes adaptive resources based on student portraits and real-time behavior data, and the learning progress tracking module monitors a learning state in real time and feeds back the learning state to the resource matching engine to optimize a recommendation strategy in a closed loop mode. The technical defects that resource recommendation of a traditional education platform is rigid and personalized adaptation is lacked are overcome.

The invention relates to the technical field of model analysis, in particular to a time-space fusion neural network line topology analysis method for a power distribution network. The method comprises the following steps: obtaining original line topology data corresponding to a power distribution network, and carrying out structured disassembly and preprocessing to construct a space-time double graph structure; constructing a bidirectional dynamic feature interaction mechanism based on the space-time double graph structure, performing multi-scale topological feature extraction, and generating a space-time separated feature vector set; performing deep coupling fusion on the feature vector set subjected to time-space separation to generate corresponding unified topological feature representation containing abnormal topology; and constructing a dynamic topology state prediction model based on the unified topology feature representation to optimize a space-time joint loss function and output a corresponding real-time topology connection relationship and an equipment state change trend, and meanwhile, performing dynamic topology reconstruction to generate a current-moment reliable topological graph corresponding to potential branch disconnection and temporary tripping. The topology analysis accuracy of the power distribution network can be improved.

The invention provides an equipment fault early warning method based on a neural network, a medium and equipment, and the method comprises the steps: collecting equipment state information at regular time, recognizing abnormal equipment, obtaining operation parameters and structure parameters of the abnormal equipment and a layout schematic diagram of a system where the abnormal equipment is located, constructing a dynamic coupling topology link, and extracting associated parameters; verifying and generating a first error weight in combination with a physical kinematics equation, generating a second error weight by matching historical fault data, and comprehensively generating fault state information; a detection mode is switched based on risk level evaluation, a system dynamic coupling topology link, associated parameters, abnormal equipment information and fault state information are input into a neural network model, fault link information is output, and finally first early warning information and second early warning information are generated according to a dynamic detection mode to realize two-stage early warning. According to the method, the fault conduction relation between equipment is modeled through the graph neural network, and fault path prediction and systematic risk early warning are realized.

The invention discloses an electric energy metering box fault prediction method and system based on big data analysis, relates to the technical field of smart power grids, and solves the problems of progressive aging missing detection and instantaneous interference misjudgment caused by dependence on single parameter threshold alarm and fault positioning misalignment caused by multi-source data isolated analysis in the prior art. According to the scheme, electrical, environment and equipment state parameters are collected in real time through a multi-dimensional sensing network; the error drift of the mutual inductor is dynamically predicted based on an LSTM-Kalman filtering model, and core breakdown early warning is realized in combination with wavelet transform; outputting a corrected resistance value and a fault mark by using a BP neural network; predicting the life of the piezoresistor by adopting a gradient boosting decision tree and fusing lightning overvoltage characteristics; the transient interference is suppressed through the combination of a Transform self-attention mechanism and dynamic time warping; according to the method, the aging detection precision and the complex environment adaptability are remarkably improved, the misjudgment rate is reduced, and the multi-fault associated positioning and active defense capability is realized.

The invention provides a pre-control and monitoring method for a whole process of actual grouting engineering based on digital geological model, comprising: extracting discontinuous fracture surfaces in a geologic body structure, to build a underground geologic body structure model; build a multi-source geologic body attribute model, optimizing and solving grouting simulations of a disaster high-risk region by using multiphase flow calculation method, and controlling grouting equipment by using optimized grouting parameters to carry out an actual grouting engineering; obtaining a multi-factor diffusion range of variable coefficients by adjusting physical values of the parameters in the grouting simulation and attribute data structure model by using a control variable method, learning and capturing complex mapping relationship between different parameters by using a neural network, optimizing the grouting parameters in the actual engineering in real-time, so as to realize the pre-controlling, monitoring and optimization of the whole process of the actual grouting engineering.

The invention discloses a virtual power plant peak regulation optimization scheduling method and system, electronic equipment and a medium, and the method comprises the steps: obtaining multi-source data of distributed energy, energy storage equipment and a load side, and carrying out the cleaning and normalization processing; utilizing an LSTM neural network and an XGBoost model to carry out short-term and medium-and-long-term load prediction respectively; based on the prediction result and the resource state, generating a scheduling optimization strategy through a deep Q network (DQN) algorithm; according to an optimization strategy, distributed energy power generation, energy storage charging and discharging and load side response are dynamically managed, and multi-resource cooperative control is achieved; and through a feedback optimization module, DQN model parameters and strategy updating are automatically adjusted according to monitoring data, and the scheduling effect is optimized. The system comprises a data acquisition module, a data preprocessing module, a load prediction module, a scheduling decision module, a resource control module, and a monitoring and feedback optimization module. The method has the beneficial effects that the load prediction precision is improved, the resource scheduling is optimized, the operation cost is reduced, and the method is suitable for scheduling management of a multi-source energy system.

Disclosed are a complex device fault diagnosis method and system based on multi-dimensional features, and the method comprises the following steps of: acquiring fault data of a target complex device, and carrying out input coding mapping and position coding; utilizing a multi-head attention mechanism and multi-head dilated convolution series connection, and acquiring fault features covering global and local deep information in combination with a feedforward neural network; fusing the extracted fault features to acquire a multi-dimensional feature vector; and acquiring a fault type probability score matrix by a classifier, and outputting a fault diagnosis result of the target complex device. The global modeling advantage of multi-head self-attention and the local multi-scale feature perception advantage of multi-head dilated convolution are fused, and a device fault diagnosis model in which a Transform model and the multi-head dilated convolution are mutually fused is established.

The invention discloses an edge computing resource allocation optimization method and system based on reinforcement learning, and relates to the technical field of edge computing resource management.The method comprises the steps that computing load data and network state data are collected, a historical load sequence is formed, a state space matrix is constructed, the historical load sequence is mapped into a discrete state value, and the discrete state value is obtained; establishing a deep neural network model; inputting the state space matrix into a deep neural network model, and generating a resource allocation strategy set according to the priority and time delay requirements of the calculation tasks; iteratively training the resource allocation strategy set by adopting a double Q learning algorithm, setting a reward function based on task completion time and a resource utilization rate, and selecting an optimal resource allocation strategy; and adjusting computing resource allocation of the edge nodes according to the optimal resource allocation strategy, and meanwhile, feeding back an execution result to the deep neural network model for online learning updating. According to the method, intelligent optimization of edge computing resource allocation is realized by introducing a deep reinforcement learning method.

The invention relates to the technical field of fault detection, in particular to an industrial fault diagnosis method and system based on intelligent causal correction. The method comprises the following steps: acquiring monitoring data and user task requirements; constructing task structure information based on the obtained user task demand; processing the monitoring data through a three-layer cascaded framework to generate metadata; constructing an initial causal graph structure by utilizing the task structure information and the metadata; performing deep optimization on the initial causal graph structure through a graph neural network to obtain a causal composite relation graph; fault diagnosis and information retrieval are carried out by utilizing the causal composite relation graph and combining a knowledge base; and generating a structured diagnostic report. Through an innovative structure integrating the LLM and the graph neural network, the system can dynamically optimize a causal relationship model, automatically learn and correct fault association, enhance the interpretability of a causal graph by using semantic reasoning of the LLM and a graph attention mechanism, and significantly improve the robustness and accuracy of diagnosis.

The invention discloses a power equipment fault cross-domain collaborative analysis system and method, and relates to the technical field of power grid dispatching, and the method comprises the steps: obtaining preprocessed multi-source heterogeneous data of power equipment, constructing a cross-domain knowledge graph based on the topological relation of the preprocessed data and historical fault data, and marking a fault propagation path. And a graph neural network is adopted to carry out embedded representation. Designing a space-time multi-branch network, respectively extracting space, time sequence and modal interaction features by using the space-time multi-branch network, and performing fusion in a feature fusion layer to obtain fusion features and branch weights; according to the method, mapping knowledge domain embedded representation is combined, a collaborative reasoning model is constructed by utilizing a Bayesian network, reasoning decision is performed on fusion features, finally, a cross-domain collaborative analysis result of the power equipment fault is obtained, and fusion and efficient reasoning of multi-source heterogeneous data are realized through combination of the mapping knowledge domain and a space-time multi-branch network. And the accuracy and efficiency of fault diagnosis are improved.

The invention provides an industrial park multi-target collaborative optimization scheduling method and system based on artificial intelligence, relates to the technical field of intelligent scheduling of industrial parks, and aims to establish a digital twinborn model and a dynamic topological graph, perform resource flow relationship modeling by using a graph neural network, and perform resource demand analysis in combination with a multi-model prediction framework. A multi-agent reinforcement learning algorithm is adopted to generate a scheduling strategy, migration from a global strategy to a local decision module is realized through a strategy distillation technology, robustness of the scheduling strategy is simulated and verified in a digital twin model, the strategy is adjusted according to a simulation result, and finally efficient scheduling of park resources is realized. The method effectively improves the intelligentization and collaboration level of the operation of the industrial park, reduces the energy consumption and carbon emission, enhances the ability of the park to cope with complex scenes, and is suitable for the efficient management of the modern industrial park.

The invention discloses a traffic supervision system applied to an intelligent street lamp and an intelligent supervision method thereof, relates to the technical field of intelligent traffic, and solves the problems that an existing intelligent street lamp system lacks a physical-digital mapping relation, edge computing resource allocation is low in efficiency and cloud computing delay is high. According to the scheme, on the basis of multi-sensor data fusion, space-time reference unification is carried out by adopting an atomic clock and a GNSS, and a dynamic causal graph is constructed through a graph neural network, so that abnormal event detection is optimized; an improved Jaccard space-time similarity algorithm is adopted to optimize calculation task allocation, an edge calculation cluster is constructed based on 5G-V2X, and high-risk region identification and traffic flow prediction are carried out; a LiFi or 5G-UWB communication medium is adaptively selected through a multi-modal fusion reinforcement learning algorithm, and efficient early warning information synchronization is realized; according to the method, the multi-source data fusion value and the early warning precision are remarkably improved, the computing power resource utilization rate is optimized, and the instruction real-time performance and the system self-adaptive capability in a complex environment are enhanced.

The present invention provides a facial recognition system using deep learning methodologies while integrating privacy-preserving capabilities. This system employs convolutional neural networks (CNNs) to extract and classify facial features, ensuring high accuracy in recognition tasks. Moreover, the system addresses privacy concerns by incorporating techniques such as facial feature encryption and anonymization, thereby enhancing user privacy and data security. This invention is applicable across various domains, including security, surveillance, access control, and personalized services, where facial recognition is utilized while preserving individual privacy.

The invention provides a multi-modal enterprise credit risk assessment method based on a knowledge graph, which integrates data such as enterprise relationships, industry policies and supply chain information by constructing an enterprise financial knowledge graph, processes entity static attributes and associated information by using a multi-modal embedding technology, captures the associated information in combination with a heterogeneous graph neural network, and evaluates the credit risk of an enterprise. And the dynamic space-time attention mechanism mines time and space features of the time series data, identifies a core risk conduction path based on an attention weight, and finally fuses graph-level features, dynamic space-time features and business rules to output a structured evaluation result. According to the method, multi-modal data is effectively integrated, the problem of incidence relation modeling deficiency is solved, deep fusion of enterprise multi-source data and accurate extraction of risk features are realized, and the accuracy and interpretability of enterprise credit risk assessment can be effectively improved.

The invention discloses an industrial robot real-time adaptive control method and system based on digital twinning, and relates to the technical field of industrial robots. The digital twin engine module runs a high-fidelity dynamics simulation model and an environment interaction model, performs real-time state estimation, abnormal working condition recognition and twin parameter dynamic updating, is seamlessly integrated with the control execution module, and provides decision support with high robustness and high adaptability for an industrial scene; the adaptive control module performs online rolling optimization on a control strategy based on a deep reinforcement learning algorithm, generates joint space trajectory correction, tail end precision compensation and dynamic load adaptability optimal instructions, and realizes parameter adaptive setting through fuzzy logic or a neural network; and the fault diagnosis module performs multi-scale time sequence analysis by using an LSTM and convolutional neural network fusion model, detects position offset, moment sudden change or temperature overrun and other abnormalities, and triggers emergency shutdown, sound-light alarm and an adaptive recovery strategy.

The invention relates to a water quality change trend rapid prediction method based on multi-source data fusion and physical constraint, and the method specifically comprises the following steps: 1, synchronously collecting spectral information, DO, COD, temperature, pH and other data at a key monitoring station, constructing a hydrodynamic water quality coupling equation, simulating the spatial-temporal dynamic distribution of water quality parameters, and calculating the water quality change trend; 2, outputting a water quality sensitive area through a hydrodynamic force-water quality model, screening sensor layout point positions in combination with information entropy evaluation and spatial clustering, realizing low-cost water quality sensor network deployment through a multi-objective optimization algorithm, and calculating a water body global water quality distribution diagram by adopting a spatial interpolation method, 3, synchronously collecting spectral information according to key monitoring sites, analyzing main pollution sources, and adopting a principal component analysis and attention mechanism neural network; 4, based on real-time optical characteristic value-DO data, in combination with a spatial topology network, a water quality gradient and a cross-regional covariance, capturing water quality parameter spatial correlation among different sites, and determining the water quality parameter spatial correlation among different sites; an optical characteristic value-DO-COD dynamic prediction model is constructed; a COD predicted value is corrected by combining pollution traceability and spectral characteristics, and multi-source data is assimilated by adopting ensemble Kalman filtering, so that the model precision is improved.

The invention relates to the technical field of information, in particular to a multi-task data analysis method and device and a storage medium, and the method comprises the steps: constructing a multi-dimensional task state feature vector as the input of a task loss prediction model by collecting a task execution progress, a resource use condition and an abnormal log index in a task operation process; constructing a task dependency graph and a resource conflict graph according to the dependency relationship between the tasks and the resource occupation condition, learning the relevance between the tasks and a resource competition mode through a graph neural network algorithm, and obtaining an influence factor and a propagation path of task loss; the performance of a task loss prediction model is continuously tracked, model parameters are dynamically updated through an incremental learning mechanism, the method adapts to changes of a multi-task environment, prediction accuracy and real-time performance are improved, and a reliable decision basis is provided for task scheduling and resource optimization.

The invention discloses a knowledge base system construction method and device for multi-source heterogeneous data and a medium, and relates to the technical field of artificial intelligence and natural language processing. The method comprises the following steps: integrating a dynamic graph convolutional network and a hierarchical attention mechanism to construct a multi-modal document analysis engine; performing semantic structure analysis on the original heterogeneous document on the basis of a multi-modal document analysis engine to extract document structure features and content semantic features, and constructing an original document relationship model on the basis of the document structure features and the content semantic features; based on the original document relationship model, performing classification fusion on heterogeneous data in the original heterogeneous document to obtain a to-be-stored heterogeneous data corpus, and processing the to-be-stored heterogeneous data corpus by using a graph neural network to establish a cross-modal semantic association index; and based on the cross-modal semantic association index, performing classified storage on the to-be-stored heterogeneous data corpora by utilizing a preset heterogeneous database so as to complete knowledge base system construction of the multi-source heterogeneous data.

According to the PCB production line defect detection system and method based on image processing, a plurality of high-resolution cameras, an infrared thermal imaging camera and a 3D laser scanner are adopted, a dynamically-controlled annular LED light source, polarized light illumination and a structured light projection technology are combined, a foreground and a background are separated through a self-adaptive threshold Otsu method, and the defect detection accuracy is improved. A convolutional neural network denoising model and a frequency domain filtering technology are introduced, welding spot and line features are extracted from a preprocessed image, a coding scheme is designed, multi-modal features are fused, spatial-temporal feature analysis is introduced, fine classification and positioning are performed in combination with SVM preliminary screening and a YOLOv5 deep learning model, an online learning mechanism is introduced, model parameters are dynamically optimized, and a multi-modal image is obtained. An intelligent early warning mechanism is established, automatic grading and alarm sending are carried out, a statistical report is automatically generated, production process parameters are recorded, root cause analysis is carried out in combination with big data analysis, and potential defects are early warned in advance.

The invention provides a decision model construction method based on a big data environment. The method belongs to the economic data decision field. The method comprises the following steps: firstly, acquiring multi-source heterogeneous data including sensor time sequence data, an expert rule base and equipment causal priori knowledge, and constructing a standardized training data set; then, extracting an explicit causal relationship from the historical data by utilizing a causal discovery algorithm, and generating an interpretable knowledge graph in combination with priori knowledge of an equipment manual; thirdly, constructing a neural symbol joint model; neural network parameters and rule confidence coefficients are synchronously adjusted, and collaborative learning of data driving and knowledge driving is realized. And then, carrying out multi-dimensional logic verification on the trained model, including rule conflict detection, anti-factual reasoning and decision path traceability verification. The method can effectively improve the accuracy and interpretability of the decision model, and is suitable for processing decision tasks in a big data environment.

The invention provides a chemical industrial park major hazard source safety management method and system, and relates to the technical field of safety management, and the method comprises the steps: constructing a chemical industrial park hazard source knowledge graph, learning node features through employing a graph neural network, recognizing the potential correlation between hazard sources in combination with a pre-trained deep learning model, and constructing a hazard source propagation path. And then, calculating the risk probability of each propagation path, and generating a hidden danger risk assessment result. And finally, constructing a hazard source causal chain based on a knowledge graph inference engine, and generating a hidden danger traceability analysis report and a risk prediction report in combination with the risk probability. According to the invention, through the knowledge graph and the deep learning technology, intelligent safety management of the chemical industry park hazard source is realized, potential hazard association can be effectively identified, risks can be predicted, decision making can be assisted, and the safety level of the chemical industry park can be improved.

The invention discloses a cross-modal information fusion and retrieval method and system driven by semantic comprehension, and the method comprises the steps: obtaining text, image and audio original data, and extracting an initial feature set of each modal through a deep neural network; dynamically distributing each modal weight coefficient based on an attention mechanism, and performing weighted fusion on the initial feature set to obtain cross-modal fusion feature representation; through a cross-modal semantic association analysis model, high-dimensional semantic association features are extracted from the fusion feature representation, and semantic enhancement feature vectors are generated; constructing a cross-modal semantic graph network based on the vector, complementing missing modal features, and generating an optimized multi-modal feature set; and inputting the optimized feature set and the query sample into a contrast learning model, calculating a semantic similarity score, and generating a cross-modal retrieval result sorting list according to the score.