38 results about "Scaled correlation" patented technology

The invention provides a soft measurement method and device for industrial multi-rate acquisition and a medium, and the method comprises the steps: collecting historical data in an industrial process, and dividing the historical data into a plurality of pieces of sampling rate data; feature extraction is carried out on different sampling rate data, and dimension regularization is carried out to obtain each scale feature; according to each scale feature and the query source, obtaining each bidirectional cross attention matrix, and performing fusion to obtain a bidirectional cross attention feature; dynamically calibrating the bidirectional cross attention features based on historical features to obtain a soft measurement model; and inputting to-be-predicted data in the industrial process into the soft measurement model to obtain a quality variable prediction result in the industrial process. According to the method, the device and the medium, the problem of low industrial process quality variable prediction accuracy caused by insufficient feature extraction, insufficient cross-scale association mining and limited utilization of historical target trend information when a soft measurement method of an existing industrial system processes multi-sampling-rate data can be solved.

The invention provides a dynamic sparse observation-oriented deep neural process ocean data assimilation method, and relates to the field of ocean data processing, and the method specifically comprises the following steps: constructing a training data set; simulating actually observed non-uniform and uncertain characteristics through Gaussian nuclear diffusion; building an ocean assimilation network oriented to sparse dynamic observation, outputting an analysis field and estimating uncertainty; and performing end-to-end training on the ocean assimilation network model by taking the reanalysis true value field as a supervision signal, and optimizing network parameters by combining a minimum error term and a structure constraint term. And after training is completed, inputting the background field in the test stage and sparse observation into the ocean assimilation network model for reasoning to obtain an ocean state reconstruction field conforming to the actual physical quantity scale. According to the technical scheme, the problem that in the prior art, calculation feasibility, cross-scale correlation modeling and credible uncertainty output cannot be considered under the real conditions of sparse observation and dynamic change of spatial-temporal distribution is solved.

The invention relates to a fermentation parameter intelligent decision-making method and device based on deep learning, equipment and a medium. The method comprises the following steps: collecting time series data of two modes in a microbial fermentation process and performing multi-dimensional feature analysis to obtain a morphological feature vector and a metabolic feature matrix; a fusion feature matrix is obtained through time dimension matching, and a dynamic correlation intensity curve is generated through nonlinear coupling modeling; metabolic fluctuation is marked abnormally, a cross-dimension anomaly recognition and multi-mode collaborative prediction model is constructed, data are fused and then input into the prediction model, and a regulation and control strategy is generated through long and short-term memory network optimization derivation. By adopting the method, multi-modal data fusion and cross-scale correlation analysis can be realized, the fermentation abnormity identification accuracy and parameter regulation and control scientificity are improved, and the product yield, purity and batch stability control capability are enhanced.

The invention relates to the technical field of pathological section image splicing, and discloses a pathological section association-based pathological image splicing method and device. The method comprises the following steps: acquiring and preprocessing a serialized pathological sub-slice image; calculating a texture flow direction field representing microstructure arrangement of the tissue, and constructing a multi-scale feature descriptor; on the basis of the multi-scale features, the adjacency relation between the sub-slices is quantitatively evaluated by fusing gradient differences, texture distribution KL divergence and a multi-scale correlation degree function of frequency domain mutual information; constructing a global spliced graph model by taking the degree of association as an edge weight, establishing an energy function containing data fidelity and texture smoothness constraints, solving an optimal splicing parameter through iterative optimization, and controlling an iterative process according to convergence criteria of comprehensive splicing stability, residual matching potential and progress; and finally generating a seamless panoramic pathological image. According to the method, biological structure characteristics are utilized, robustness and matching accuracy under complex conditions are improved, and continuity and consistency of splicing results are ensured.

The invention relates to the technical field of materials and data processing, and discloses a method for constructing a damping layer material loss factor and temperature correlation model, which effectively overcomes the defect of insufficient adaptability of a traditional damping material model in a wide temperature range required by a thermal power plant by fusing a molecular chain segment dynamics mechanism and a macroscopic constitutive behavior. The method has the technical advantages that on the basis of the physical basis of a molecular motion energy barrier theory constraint model and in combination with a corrected temperature-frequency equivalent conversion mechanism, a cross-scale correlation framework with clear physical significance is constructed; through collaborative optimization of microscopic activation energy and macroscopic viscoelastic parameters, the reliability of thermal power plant all-working-condition temperature domain loss factor prediction is remarkably improved; an embedded real-time calculation framework is adopted, dynamic evaluation and compensation control of the material damping performance in an engineering scene are achieved, and more accurate technical support is provided for vibration reduction structure design, vibration fault diagnosis and prevention and service life prolonging of key equipment in the fields of thermal power plants, heavy machinery and the like.

The invention discloses a multi-modal geologic feature fusion method and a multi-modal geologic feature fusion system based on a discrete wavelet transform (CLIP)-Stable Diffusion model. The method comprises the following steps: reconstructing multi-modal geological data through differential wavelet transformation, extracting a low-frequency trend from earthquake and deposition data by adopting a Daubechies wavelet, capturing high-frequency details from deposition numerical simulation and logging data by adopting a Coiflets wavelet, and vectorizing a geological text through an orthogonal basis matrix; a CLIP-Stable Diffusion fusion model is constructed, a text encoder is utilized to analyze geological semantic features, an image encoder is utilized to extract spatial features, and a U-net diffusion generator realizes cross-modal alignment under the guidance of text conditions through a cross attention mechanism; and adopting a loss threshold termination mechanism constrained by a geological law in diffusion training, and finally outputting fusion data through wavelet inverse transformation. According to the method, cross-scale correlation deficiency and semantic segmentation limitation of a traditional method are broken through, and reservoir modeling precision and exploration efficiency are remarkably improved.

The invention discloses a PM2.5 complex time sequence prediction method based on a double-path fusion architecture, and belongs to the technical field of PM2.5 complex time sequence prediction methods.According to the method, a double-path fusion structure comprising a local feature extraction path and a global time sequence modeling path is constructed, and combined modeling is carried out on PM2.5 time sequence data of multiple cities, the local path captures short-term fluctuation and high-frequency disturbance characteristics by using a convolution structure, and the global path models long-term trend and multi-scale correlation by using a neural network based on an attention mechanism, so that fine prediction of a complex non-stationary sequence is realized; according to the method, a reversible normalization mechanism is introduced to dynamically adjust input distribution, self-adaptive fusion of local and global results is realized in combination with a double-prediction-head weighted fusion strategy, so that the influence of abnormal disturbance on the model is effectively inhibited, meanwhile, the overall calculation complexity is reduced through a modular design structure, and the calculation efficiency is improved. And the deployability and the real-time performance of the method in a multi-scene air quality monitoring system are enhanced.

The invention discloses a multi-scale time sequence prediction system and method for adaptive hierarchical frequency, which realizes standardized processes of data preprocessing, dynamic hierarchical sampling, frequency attention modeling, cross-scale attention fusion and prediction output through modular architecture design, is clear in interface between modules, is easy to deploy and expand, and is high in practicability. The problem of non-uniform system architecture in the prior art is solved; in the dynamic stratified sampling step, through region division and dynamic sampling rate distribution, adaptive matching of a down-sampling rate and a local frequency characteristic of a non-stationary sequence is realized; in the frequency attention modeling step, through frequency domain conversion and attention weight distribution, the sensitivity of the system to periodicity and frequency characteristics is enhanced, and key frequency components can be focused according to prediction task requirements; according to the cross-scale attention fusion step, through feature dimension unification and cross-scale correlation calculation, the dependency among the scale features is effectively modeled, redundant information is filtered, and the fusion efficiency and the prediction precision are improved.

The invention discloses a new material performance prediction method and system based on artificial intelligence, and belongs to the technical field of new material performance prediction. The system comprises a memory, a processor and a computer program, and the processor executes the program to realize the performance prediction method. The method comprises the following steps: obtaining target new material component data and preparation process parameters, and constructing a material feature matrix containing atomic bonding and microstructure features by means of a multi-modal feature extraction network; correcting process parameters through a process stability evaluation model, generating an optimization decision vector, dividing a matrix region, and monitoring phase transition temperature and grain boundary energy data of a key region; inputting a co-evolution network to obtain a performance influence factor sequence, and carrying out weighted fusion to generate an optimized feature tensor; and the cross-scale correlation model matches historical data and outputs a prediction interval, and the customer interaction unit displays and adjusts the matching dimension according to the demand and outputs again.

The application discloses a method for extracting middle layer top temperature multi-period characteristics and analyzing influencing factors, and belongs to the technical field of atmospheric detection and meteorological data analysis. The method first extracts time series data of the middle layer top temperature based on the 90km height and the lowest temperature point standard, then completes time domain trend and mutation analysis through seasonal departure and Mann-Kendall test, adopts discrete wavelet decomposition to obtain four main oscillation periods of 3 years, 7 years, 11 years and 22 years, then uses continuous wavelet to obtain the length of the fine-calibrated period, and finally performs cross wavelet analysis on the quasi-biennial oscillation, the El Nino effect and the solar activity index to obtain the final correlation and time lag results. The application realizes joint time-frequency domain analysis of discrete wavelet transform, continuous wavelet transform and cross wavelet transform, effectively improves the cycle recognition accuracy, quantifies the multi-scale correlation and time sequence hysteresis of each factor, and provides a new research idea for high-altitude atmospheric climate influencing factor analysis.

The invention discloses an intelligent concrete slump detection method based on multi-modal information fusion, and belongs to the technical field of concrete quality detection. The stirrer video data and the main shaft current signal are synchronously acquired. For video data, video image sequence apparent features are extracted and fused with dense optical flow field features, and a video fusion feature sequence is obtained. Specifically, the method comprises the following steps: extracting inter-frame dense fluid motion features by adopting a Farneback algorithm; and synchronously extracting image apparent visual features by using a deep network. The two are cooperated to realize cross-scale correlation of the material surface morphology and the pixel-level fluid flow state, and a video fusion feature sequence with both microscopic dynamic and global appearance is formed. And for current data, features are synchronously extracted through a deep network, a video and current parallel processing structure is integrally formed, a video fusion feature sequence and current features are subjected to feature superposition and cross attention module fusion, a multi-modal prediction model is constructed, and the slump detection accuracy is improved.

The invention discloses a mechanical production well scaling state intelligent identification method and system. The method comprises the steps that abnormal values and noise information in a mechanical production well data set are filtered out; obtaining a reliable scaling label; influence factors related to the scaling height of the mechanical production well are screened out from the original data, and scaling related characteristics are determined; a neural network model based on CNN-LSTM is constructed, an SE attention mechanism is introduced to construct an SE-LSTM algorithm, and an Adam optimizer is adopted to optimize model parameters; according to the separable variables and the scaling correlation characteristics, inputting the separable variables and the scaling correlation characteristics into a CNN-LSTM-based neural network model for training, using multi-classification cross entropy as a loss function of training, and outputting a scaling type; and using the trained CNN-LSTM-based neural network model to predict scaling data of an unknown mechanical production well. By adopting the technical scheme, the scaling characteristics of the mechanical production well can be accurately identified.

The present application belongs to the field of computer vision, and provides an RGB-D saliency object detection method based on depth quality weighting, comprising the following steps: 1) obtaining an RGB-D dataset for training and testing the task, and defining the algorithm target of the present application; 2) constructing an RGB encoder for extracting RGB image features and a depth (Depth) image feature encoder; 3) constructing a cross-modal weighted fusion module, and guiding the weighted fusion of the extracted RGB image features and Depth image features through a depth quality evaluation mechanism guided by a weighting formula; 4) constructing a bidirectional scale correlation convolution mechanism for multi-scale feature extraction and fusion, so as to enhance the advanced semantic information of multi-modal features; 5) establishing a decoder to generate a saliency map P est ; 6) calculating the loss of the predicted saliency map P est and the manually labeled saliency object segmentation map P GT ; 7) testing the test dataset to generate a saliency map P est , and performing performance evaluation using evaluation indexes. The present application can effectively integrate complementary information from different modal images, and improve the accuracy of saliency object prediction in complex scenes.

The invention relates to the technical field of fault diagnosis, in particular to a big data-based display fault prediction system, which comprises a gray-scale synchronization module, a fluctuation identification module, a stage judgment module, a trajectory comparison module and a trend early warning module, analyzes gray-scale time sequence data output by a gray-scale optical acquisition instrument based on a display device, and determines whether a fault occurs or not by checking data integrity. And comparing the gray scale content of each frame with a driving chip synchronizing signal. According to the invention, by checking the synchronism and sequence integrity of the collected data, accurate arrangement of information on a multi-time sequence level, continuity analysis based on a change track and feature node extraction are realized, the difference expression of a key moment and a turning point of a focusing stage in a display process is clarified, and the brightness and a gray scale correlation trend are subjected to multi-dimensional comparison, so that the accuracy of the display process is improved. The dynamic state of the risk area is comprehensively assessed in combination with the trend continuity between the nodes, the capability of capturing display abnormal symptoms is improved, early sensing and timely early warning are realized, and the stability and risk prevention capability of fault prediction of the display device are effectively enhanced.

The invention relates to a lightweight stereo matching method based on weight sharing and channel attention. The left image and the right image pass through a weight sharing feature extractor to obtain a feature map, and an SE channel attention mechanism is inserted behind a residual block of the feature extractor for feature calibration; when a multi-scale correlation body is constructed, sparse indexing is carried out on parallax dimensions by adopting Hash coding based on space coordinates, and voxel sampling and trilinear interpolation with constant time complexity are realized; and obtaining a final disparity map through a variable-resolution iterative updating strategy. According to the method, the quantity of model parameters is effectively reduced, EPE and D1 indexes on a Middlebury data set are superior to those of an existing RAFT-Stereo method, and the method is suitable for real-time scenes such as automatic driving and robot navigation.

The application discloses a rock and outcrop cross-scale correlation method and system based on multi-modal data, belongs to the technical field of three-dimensional data platforms and geological learning, and comprises the following steps: S1, geological outcrop data acquisition and geological outcrop modeling are carried out, a three-dimensional geological outcrop model is obtained, multi-modal data are obtained by simultaneously collecting basic rock sample data, rock sample analysis and test data and rock sample explanation information; S2, the multi-modal data are managed by using a database and a file management mode, and the rock sample and the outcrop are correlated to obtain correlation data. Through the management of the multi-modal data, the deficiency of single expression mode is solved, and the fusion and integration of the multi-modal data are realized. Through the correlation mechanism of the rock and the outcrop, the deficiency of the existing system in the data correlation is solved.

This invention discloses a fault location method for power cables compatible with both high and low resistance, comprising: connecting to a test terminal to acquire voltage and current responses, determining the initial impedance state, and setting scanning excitation parameters; applying an exponentially increasing scanning excitation signal, calculating the complex impedance gradient characteristics, and determining the threshold energy level range; expanding a fractal excitation sequence within the threshold range, collecting propagation response data, and constructing a response matrix; normalizing the propagation response matrix data, calculating the cross-scale correlation strength, and forming a consistency curve; constructing a propagation consistency vector field, performing density clustering and topology analysis, and determining the fault stability region; and calculating the fault location based on the propagation time at the center of the stability region and the propagation velocity. This invention achieves stable identification and accurate location of faults in both high-resistance and low-resistance power cables through exponentially increasing scanning excitation and fractal energy spectrum multi-scale propagation analysis.

A construction method and system for lifting adjustable balance beams in prefabricated staircases are disclosed. This method employs machine vision-based artificial intelligence detection technology to extract central data from key frames in the lifting monitoring video of the adjustable balance beam. This focuses on the distribution of the beam's center of gravity, and further extracts the multi-scale correlation characteristics of the relative temporal changes in this center of gravity data. This information is then used to detect the lifting stability of the adjustable balance beam. This allows for accurate detection and judgment of the lifting stability, enabling the generation of an instability warning when instability is detected, thus ensuring construction safety.

The invention discloses a quality prediction method and system based on bottle blank visual features, and belongs to the technical field of quality monitoring, and the method comprises the steps: obtaining a multi-view forming image of a to-be-predicted bottle blank, carrying out the real-time timing sequence tracking through a timing sequence node, constructing a timing sequence image set in combination with a historical bottle blank, and obtaining a visual feature vector, so as to construct an image resume; generating a fusion feature vector based on the image resume; constructing a quality state vector of the to-be-predicted bottle blank, performing defect evolution prediction by using a time sequence prediction model, generating a predicted quality state vector, generating a trajectory unit through a decoding hierarchical architecture, calculating and correcting an initial failure probability, thereby generating a defect evolution trajectory, and marking a risk early warning point; a process optimization suggestion is generated by dividing granularity nodes, constructing weighted edges, generating a cross-scale association graph, constructing an association library in combination with a mapping model, constructing a directional causal association graph and carrying out causal association mining, and updating and feedback optimization are carried out on an image resume.

The application discloses a method for constructing a three-dimensional geological body, which comprises the following steps: establishing a three-dimensional seismic attribute body about a work area to be studied; based on the three-dimensional seismic attribute body, combining conventional logging data and imaging logging data, respectively dividing rock types under geological scale, logging scale and imaging scale; according to the rock type division results under different scales, combining core data, establishing a multi-scale correlation model for sequentially performing scale coarsening and attribute prediction from core, imaging, logging to geology; and updating the original three-dimensional geological body according to the multi-scale correlation model. The application can apply fine scale data to the construction of a macroscopic geological body.

The present application relates to pathological section image splicing technical field, and disclose pathological image splicing method and device based on pathological section association.The present application includes: obtaining and preprocessing serialization pathological sub-section image;Calculate the texture flow field representing the arrangement of microstructure of tissue, and construct multi-scale feature descriptor;Based on multi-scale feature, through the multi-scale correlation function of fusion gradient difference, texture distribution KL divergence and frequency domain mutual information, the adjacent relationship between sub-sections is quantitatively evaluated;With the correlation degree as the edge weight, the global splicing graph model is constructed, the energy function containing data fidelity and texture smoothing constraint is established, the optimal splicing parameter is solved by iterative optimization, and the iterative process is controlled according to the convergence criterion of comprehensive splicing stability, residual matching potential and progress;Finally, seamless panoramic pathological image is generated.The present application uses biological structure characteristics, improves the robustness and matching accuracy under complex conditions, and ensures the continuity and consistency of splicing results.

The invention discloses a strong shear turbulence dynamic grid adaptive simulation method coupled with a k-epsilon series model. The method mainly comprises the following steps of: 1, judging whether a shielding function is applied or not; step 2, calculating a Vman dynamic coefficient; step 3, identifying the self-adaptive length scale of the shear layer; step 4, constructing a scale-related adjustment function based on the turbulence energy spectrum integral; 5, reconstructing the turbulence viscosity of the k-epsilon series model by using an adjustment function; and 6, performing turbulence simulation by using the reconstructed turbulence viscosity. According to the method, the size of a local grid is identified by constructing a Vman dynamic coefficient, the length scale of the local grid is jointly determined in combination with the length scale of the shear layer adaptive sub-grid, and then the scale correlation function is constructed through the turbulence energy spectrum integral to reconstruct the turbulence viscosity, so that grid adaptive simulation is realized; an effective numerical simulation method is provided for solving low-consumption and high-precision simulation of strong shear turbulence in a complex engineering flow problem.

The invention belongs to the technical field of digital twinning communication of drainage basin infrastructure, and discloses a digital twinning drainage basin infrastructure control system for communication engineering, when a communication detector monitors signal attenuation, the sampling interval of a hydrological sensor can be automatically shortened, the frequency of an infrastructure collector is improved, and the communication efficiency is improved. Microcosmic signal fluctuation is associated with macroscopic hydrology and mesoscopic infrastructure data in real time; meanwhile, data of different scales are subjected to layered denoising, the cross-scale correlation degree is calibrated through wavelet transform and an attention mechanism in combination with a Pearson coefficient, the space error of the fused data is controlled within + / -1.5 m in cooperation with a space deviation compensation equation, and deep correlation fusion of the multi-scale data is achieved; the output calibration data can effectively reduce communication control errors, and the communication control problem caused by inaccurate data is avoided; a cross-domain physical rule base is established based on calibration data, the cross-domain physical rule base comprises dam stress-strain and base station tilt-link attenuation relations, and water conservancy-communication multi-physics field coupling equation quantitative correlation is further established.

The application provides a deep neural process ocean data assimilation method for dynamic sparse observation, relates to the field of ocean data processing, and specifically comprises the following steps: constructing a training data set; simulating the uneven and uncertain characteristics of actual observation through a Gaussian kernel diffusion; building an ocean assimilation network for sparse dynamic observation, outputting an analysis field and uncertainty estimation; using a reanalysis true value field as a supervision signal to perform end-to-end training on the ocean assimilation network model, and optimizing network parameters by jointly minimizing an error term and a structure constraint term. After training is completed, the background field and sparse observation in the test stage are input into the ocean assimilation network model for reasoning, so that an ocean state reconstruction field conforming to the actual physical quantity scale is obtained. The technical scheme of the application overcomes the problem in the prior art that the calculation feasibility, cross-scale correlation modeling and reliable uncertainty output cannot be considered under the realistic conditions that the observation is sparse and the spatiotemporal distribution dynamically changes.

The invention provides a potential emotion intelligent analysis method based on head micro-vibration monitoring, and the method comprises the steps: reconstructing a discrete vibration point source into a mutual traction elastic curved surface network in a three-dimensional space through non-uniform tension mapping; forming a mechanical conduction topological graph; detecting a micro vortex core in the topological graph; the method comprises the following steps: by constructing phase space projection of a vortex community, tracking gravity mutation and dynamic evolution of a rejection boundary caused by emotion conflict or conversion between adjacent vortexes; the method mainly captures two critical events of internal core energy level transition and external circulation tearing recombination of vortex fingerprints, and deduces a phase change trajectory spectrum of emotional conflicts according to a time-space sequence and an energy cascade scale of the two critical events. According to the method, multi-scale correlation modeling of the emotional state from micromechanics characteristics to macroscopic evolution laws is realized; wherein the elastic curved surface network provides a space substrate, vortex core detection ensures characteristic specificity, and phase space projection completes dynamic deduction.

The application discloses an intelligent data analysis method and system based on big data, relates to the technical field of data analysis, and comprises the following steps: sequentially unifying multi-source heterogeneous business data to obtain a unified sequential business data set; extracting features according to the unified sequential business data set to obtain an evolution feature set; compressing the feature space according to the evolution feature set to obtain key analysis features; performing multi-scale correlation analysis according to the key analysis features to obtain a business correlation structure; quantifying the correlation degree according to the business correlation structure to obtain a correlation weight result; and comprehensively reasoning according to the correlation weight result to obtain an intelligent analysis result. The application converts the key analysis features from independent individuals into a mutually correlated overall structure, enables the correlation relationship to simultaneously have short-term and long-term business explanation capabilities, enhances the time semantics of the correlation analysis result, and improves the stability and explainability of the correlation analysis result.

The invention relates to the field of computer vision and intelligent traffic, and discloses a road defect detection method and model based on lightweight enhanced feature fusion. The method is based on a YOLO11 architecture and comprises a backbone network, a neck network and a head network. A simplified re-parameter batch normalization module (NRepBN) is adopted in a backbone network, and a residual normalization path is introduced in a training stage to improve stability. An enhanced correlation feature fusion module (ECFM) is introduced into a neck network, through grouping feature focusing and multi-layer feature reconstruction, fine crack features are effectively reserved, and multi-scale correlation is enhanced. In the training process, a self-adaptive threshold focusing loss function (ATFL) is combined, focusing parameters are dynamically adjusted according to prediction confidence, and performance reduction caused by class imbalance is relieved. According to the method and the model, the detection precision of small targets such as micro cracks and pits can be improved in road defect image detection, meanwhile, the lightweight characteristic is kept, and the method and the model are suitable for road defect detection in various scenes and real-time deployment on edge equipment.

The invention discloses a large-model-assisted automatic program verification method and system, and the method comprises the steps: executing data scale correlation analysis for a to-be-verified program to judge whether data scale optimization can be carried out or not, executing data scale optimization if the data scale optimization can be carried out, and verifying an optimization result by using a verification tool; carrying out defect analysis on the to-be-verified program to judge whether a defect exists or not, and if the defect exists, generating a counter example according to the corresponding defect to verify the counter example; otherwise, a verification tool is used for verification, and in the execution process of the method, data scale optimization and defect analysis are both assisted by a large model for analysis. According to the method, a large model is used for assisting in data scale optimization and defect analysis, calculation resource consumption in the verification process is reduced, the verification time is shortened, the verification accuracy is improved, and particularly the verification efficiency of a complex program is improved.

The present application relates to the technical field of intelligent prediction, in particular to a red tide occurrence probability intelligent prediction method based on neural network and key factor identification, comprising S1: obtaining historical red tide event data, multi-station marine environment monitoring data, red tide emergency monitoring data and station continuous hydrological and meteorological observation data of a target sea area, and processing to obtain a standardized monitoring data set; identifying a red tide prediction key area, extracting a multi-period monitoring sequence to form a key area time sequence sample set; S2: performing multi-scale correlation analysis, sensitivity analysis and causal correlation identification to obtain a key factor sorting result; extracting a target key factor affecting red tide occurrence, determining a threshold boundary range of each target key factor, and generating a key factor threshold representation set; S3: inputting a probability prediction network and outputting a red tide occurrence probability result; and generating red tide occurrence early warning information of the target sea area within a prediction period. The present application improves the accuracy and stability of red tide prediction.

The invention discloses a cross-scale matrix flow modeling method combined with a graph theory, and relates to the field of composite material forming process simulation. The method comprises the following steps: establishing a high-precision matrix shear constitutive model through a rheological experiment and symbolic regression; constructing a reinforcement pore wall surface rough model based on a fractal theory; a single horizontal circular tube flow formula considering the roughness effect is deduced by combining the two; carrying out binarization and coarsening processing on the enhanced CT image, constructing a graph network model, and quantitatively defining edge resistance; using a Dijkstra algorithm to identify a minimum resistance path in the graph network as a main flow channel; and finally, based on the main flow channel information and a series calculation model, realizing flow prediction from a nanometer rough scale, a pore scale to a macroscopic Darcy scale. According to the method, the problems of inaccurate rheological characterization, neglected roughness effect, difficulty in cross-scale association and low main channel identification efficiency of a traditional method are solved, and the precision and efficiency of matrix flow prediction in the composite material forming process are remarkably improved.