25 results about "Model synthesis" patented technology

The various embodiments herein provide a system and method for networked digital twins with autonomous collaborative decision-making. The system comprises a Digital Twin Engine for real-time data acquisition, model synthesis, and simulation, an AI Module for advanced data analysis, an autonomous collaborative decision-making module for optimized decision-making, a communication layer for secure data exchange, and supporting modules for coordination, storage, security, and user interaction. The method for generating and deploying digital twins comprises data collection, transmission, preprocessing, model synthesis, simulation, validation, and deployment. The method for networking and collaboration comprises AI-based data processing, complex event processing, autonomous decision-making, task distribution, decision communication, real-time monitoring, and continuous improvement. This system enhances operational efficiency, scalability, and security, reducing the need for human intervention and providing a comprehensive management solution for complex systems.

The invention discloses a fine-grained argument structured automatic evaluation method and a fine-grained argument structured automatic evaluation system. The method comprises the following steps of: firstly, marking a data set by using a large model of a marking device, marking argument components of an argument statement, a structural relationship of the argument components and scores of all dimensions, and forming a training data set after checking by an artificial expert; then, a large language model is finely adjusted based on the data set, and a discussion component recognition model is constructed; performing feature representation on the hierarchical argument structure by using a graph neural network, synthesizing the original argument and the representation by using a large model, performing multi-dimensional scoring by means of a hybrid expert architecture, and constructing a composition scoring model; and finally, enabling the composition scoring model to generate a multi-dimensional comment by using the cue word text, and feeding back and outputting the multi-dimensional comment together with the obtained discussion structure and the score. According to the invention, the automatic evaluation effect and interpretability of the arguments can be effectively improved, and a two-way learning and teaching assistance platform is provided for students and teachers.

The invention discloses a construction method of a synthetic rock mass digital model containing a three-dimensional master control fracture, and belongs to the technical field of rock mass numerical simulation. The method comprises the following steps: collecting a complete rock mass and a crack-containing rock sample on site, and preparing a coal rock sample; reconstructing a rock core through CT (Computed Tomography) scanning and Avizo software, and segmenting a mineral matrix and fractures; based on a Fisher statistical distribution method, performing standardization processing on the orientation characteristic parameters of the three-dimensional fracture of the rock; carrying out grouping and dominance analysis on the fractures; the method comprises the following steps: carrying out coal sample fracture analogue simulation by adopting a DFN in software 3DEC, and generating a DFN-Voronoi block model; a DFN-Voronoi SRM model is manufactured; carrying out a compression experiment on the model to monitor stress and strain; and carrying out DFN-Voronoi SRM model parameter sensitivity analysis based on a single-factor experimental method, and proposing a parameter calibration method. According to the method, the mechanical property and the dominant fracture group of the rock sample are considered, and a parameter acquisition and calibration method of the model is completely given in the aspect of model calibration; fracture network construction and model synthesis are optimized, and DFN-Voronoi SRM model parameter sensitivity analysis and calibration are summarized.

Methods, systems, and apparatus, including medium-encoded computer program products, for computer aided design of physical structures using three-dimensional model synthesis processes. A method includes: obtaining a build angle, a manufacturing direction, a design space, and one or more design criteria for use in a shape synthesis process, the build angle and the manufacturing direction being for an additive manufacturing process; performing the shape synthesis process including applying a build-angle filter to a boundary-based computer-data representation of an intermediate version of the shape of the modeled object during multiple iterations, including removing a portion of material from the intermediate version of the shape in accordance with the build angle and the manufacturing direction to make the intermediate version of the shape self-supporting in a vicinity of the portion of material; and providing the shape of the modeled object for use in manufacturing using the additive manufacturing process.

The invention discloses a three-dimensional model synthesis method based on multiple photos. The method comprises the following steps: S1, respectively exporting three-dimensional models generated by the photos as a three-dimensional model A and a three-dimensional model B; s2, performing one-to-one mapping on the feature points on the three-dimensional model A and the corresponding feature points on the three-dimensional model B, and performing connection; s3, cutting and deleting the overlapped part of the three-dimensional model A and the three-dimensional model B, and exporting a three-dimensional model C and a three-dimensional model D; s4, splicing the three-dimensional model A and the three-dimensional model B to form a complete three-dimensional model E, and redistributing uv for the three-dimensional model E; and S5, baking the color maps of the three-dimensional model C and the three-dimensional model D onto the three-dimensional model E to form a three-dimensional model. According to the method, the problems of low model integrity and poor color information accuracy when a special object is processed by an existing photo modeling technology are solved, and the quality and the reality sense of the three-dimensional model are greatly improved.

This invention discloses a method for predicting the bearing stratum of cast-in-place piles based on deep fusion of multimodal data. By integrating the original borehole data and multimodal data of cast-in-place piles, and leveraging deep learning to mine information, it reduces additional drilling and lowers costs. Simultaneously, through data preprocessing and rapid calculation using a deep learning model, the results are obtained simply by inputting the pile location coordinates after model training, significantly shortening the cycle and improving project progress. Furthermore, by integrating multimodal data to construct a comprehensive feature vector, the model deeply fuses and analyzes the data, learning location correlation patterns to achieve bearing stratum prediction for large-area sites, resulting in more comprehensive and continuous results, overcoming the shortcomings of traditional methods. In addition, the multimodal feature vector encompasses multiple types of information, and deep learning techniques such as cross-modal attention fusion modules consider intermodal relationships. The model comprehensively mines deep feature patterns from various factors, enabling more accurate prediction of key information such as bearing stratum type, burial depth, and thickness.

The application discloses a kind of based on virtual flat adit's slope structural plane model construction method, comprising: generating and each reality flat adit corresponding considering structural plane spatial variability reality flat adit model;Comprehensively consider the geometric characteristics of reality flat adit model and the certainty structural plane feature, generate several considering structural plane spatial variability while considering structural plane along elevation continuity random change virtual flat adit model.The application according to the actual flat adit data of site to the blank area that has not carried out flat adit exploration restores its structural plane feature as far as possible, establishes the structural plane network model in conformity with geological authenticity, to carry out subsequent slope stability evaluation work.The application realizes that the structural plane network constructed in model meets the exposure condition of reality flat adit, while considering the spatial variability of structural plane distribution characteristics in different parts of slope, solves the problem that structural plane model deviates from reality in rock slope numerical calculation and influences the slope stability evaluation.

This invention discloses a comprehensive evaluation method for large-scale models based on multi-dimensional index fusion. The comprehensive evaluation method includes: Step S1: constructing a hierarchical evaluation system; Step S2: constructing the evaluation set and managing data; Step S3: constructing an automated evaluation process; Step S4: performing result analysis and feedback. A systematic and standardized evaluation method for large-scale models is established to quantify the comprehensive capabilities of models in general, financial, and specific financial scenarios.

The various embodiments herein provide a system and method for networked digital twins with autonomous collaborative decision-making. The system comprises a Digital Twin Engine for real-time data acquisition, model synthesis, and simulation, an Al Module for advanced data analysis, an autonomous collaborative decision-making module for optimized decision-making, a communication layer for secure data exchange, and supporting modules for coordination, storage, security, and user interaction. The method for generating and deploying digital twins comprises data collection, transmission, preprocessing, model synthesis, simulation, validation, and deployment. The method for networking and collaboration comprises Al-based data processing, complex event processing, autonomous decision-making, task distribution, decision communication, real-time monitoring, and continuous improvement. This system enhances operational efficiency, scalability, and security, reducing the need for human intervention and providing a comprehensive management solution for complex systems.

This invention belongs to the field of building construction and digital twin technology, and relates to an intelligent matching method and system based on artificial intelligence. This invention collects construction environment data in real time and identifies risk markers, calculates component deformation using a time-series prediction model, and corrects the virtual components generated by the BIM model for environmental compensation. It employs 3D point cloud registration technology for geometric matching analysis to determine hole alignment errors and interface gaps. Based on a dynamic decision tree model, it dynamically adjusts the installation sequence by comprehensively considering the degree of geometric mismatch, environmental risks, and construction logic. Furthermore, it incrementally updates the model based on actual installation deviations. This invention effectively solves the technical problem of component deformation and installation mismatch caused by environmental factors such as temperature and humidity during construction, achieving dynamic adaptation and precision control of the construction process, improving the matching success rate and construction efficiency of component assembly, and continuously optimizing the system's adaptability and reliability through continuous learning.

The invention discloses a ship field-oriented intelligent meta-model construction and dynamic adaptation method and system. The method comprises the following steps: constructing a ship field system meta-model library comprising concept, interaction and behavior meta-models; performing semantic understanding and intention recognition on natural language requirements of a user by using a large language model of an integrated domain dictionary, and extracting structured modeling elements; word vector cosine similarity and semantic distance calculation based on domain ontology are fused to generate a candidate set; a multi-objective decision model is adopted, semantic fitting degree, coupling degree and multiplexing frequency are comprehensively considered to optimize and screen the candidate set, and an optimal meta-model adaptation sequence is output; instantiating the meta-model and constructing an incidence relation; performing consistency verification including static grammar, dynamic semantics and cross-view logic on the instance set; and performing version persistent storage on the verified model. According to the method, automation and intelligence of ship domain model construction are realized, and the modeling efficiency and the model accuracy and reliability are remarkably improved.

Methods, systems, and apparatus, including medium-encoded computer program products, for computer aided design of physical structures using three-dimensional model synthesis processes. A method includes: obtaining a draft angle, an ejection direction, and a 3D shape for a casting or molding process; modifying data values in discrete elements (from which the 3D shape is determinable) to add material to the modelled object, including, for each of the discrete elements, changing a current value of the discrete element based on a comparison of the current value with an other value and a constant value, the other value being determined from one or more values found in one or more other discrete elements in the data structure located away from the discrete element along the ejection direction, and the constant value being determined for the draft angle; and providing the data structure with the modified data values for further processing.

The invention relates to the technical field of artificial intelligence and energy power, in particular to a power generation scene-oriented large model performance evaluation method and related equipment, and the method comprises the steps: constructing an evaluation index system including task performance, power generation equipment adaptation and power generation data characteristics; distributing initial weights for the indexes by combining expert knowledge in the power generation field, and generating an initial weight vector; evaluating the candidate large model based on a pre-selected test set, obtaining test data and constructing an initial performance matrix; and adjusting the initial weight vector through an optimization algorithm to obtain an optimized weight vector, and carrying out weighted calculation on the initial performance matrix to obtain a comprehensive adaptation degree score of each model so as to realize performance evaluation of the large model of the power generation scene. According to the method, both the index specialty and the evaluation scientificity are considered, and a reliable basis can be provided for model selection and optimization of a power generation scene large model.

The invention discloses an interpretable method and device for a model prediction process based on a causal graph, and the method comprises the steps: constructing a global causal graph used for describing the correlation between features through employing the high-dimensional structural data of a target scene; nodes in the global causal graph are variables, and directed edges are causal relationships among the variables; clustering and pruning the global causal graph to obtain a plurality of pruned sub-graphs; performing classification prediction on each sub-graph by using a preset graph neural network model to obtain a GNN classification sub-model corresponding to each sub-graph; synthesizing the GNN classification sub-models corresponding to each sub-graph into an integrated model to obtain a prediction model; extracting a common part of the plurality of sub-graphs to obtain a common sub-graph; and according to the common sub-graph, carrying out interpretability on the prediction model to obtain an interpretability report. Therefore, by adopting the embodiment of the invention, the strict requirement of the special scene on the interpretability of the model can be met, and the trust mechanism of the special scene on the model decision can be effectively established.

The invention provides a complex multivariate combination model comprehensive element intelligent extraction and recombination publishing method, and relates to the field of system comprehensive modeling simulation. The complex multivariate combination model comprehensive element intelligent extraction and recombination publishing method is realized through the following technical steps: step 1, preparing and preprocessing multi-source model data, analyzing a *. Mo file of a target system through a Modelica compiler, and generating an AST syntax tree containing component declarations and equation sets; and importing STEP format geometric data of the CAD model, and carrying out grid repair and feature edge extraction by using Paraview. According to the method, a lightweight digital fingerprint algorithm oriented to multi-source heterogeneous data is developed, while encryption strength is kept, feature extraction time of a gigabit-level CAD model is shortened to be within 30 seconds, an interpretable intelligent confirmation framework is constructed, a model element feature map is constructed, a technical barrier is broken through, a model element association rule base based on a semantic network is developed, and a multi-source heterogeneous data encryption algorithm is developed. And automatic binding and intelligent extraction of the geometric parameters and the multi-physical field equation are realized.

The invention discloses a large model comprehensive evaluation system based on artificial intelligence. The system comprises a data set construction module, a preliminary evaluation model construction module, a depth evaluation model construction module and a comprehensive evaluation module. The invention relates to the technical field of large model evaluation, in particular to a large model comprehensive evaluation system based on artificial intelligence, which improves the accuracy and stability of the evaluation process by introducing a self-adaptive vector set center adjustment mechanism, dynamically optimizing the number of sets, fusing a similarity measurement method and a self-adaptive termination criterion; a multi-dimensional evaluation mechanism of feature importance, distribution rationality and structural stability is introduced, data vectors are comprehensively scored through a dynamic weighted fusion strategy, the comprehensive evaluation ability of the data vector quality is improved, excellent data vectors are effectively identified through a deep screening mechanism, and the accuracy of the quality of the data vectors is improved. The accuracy and hierarchy of the evaluation result are optimized, and finally the comprehensive quality evaluation precision and reliability of the large model are improved.

The application discloses a kind of video image data cleaning methods and systems for urban rail transit engineering, method includes: calculating video adjacent frame structure similarity index SSIM, video block is divided according to threshold value and index set is constructed;Select no less than one multi-modal large language model, score is obtained by feature pre-processing, prompt word construction, model input calculation, the grouping voting mechanism is used to judge feature, obtain feature matrix and retain background image;Definition feature block, construct matrix to obtain sampling set, calculate sampling interval according to training sample requirement, realize redundancy removal by interval sampling or supplementary synthesis data;The amount of use of model synthesis data is used to classify and augment, adapt by discriminator and autoencoder, fusion adaptation data and original data complete augmentation, achieve video image data automatic cleaning.Reduce data storage and management cost, so that data resources are more reasonably configured.

The invention discloses a virtual adit-based slope structural plane model construction method. The method comprises the steps of generating a real adit model which corresponds to each real adit and considers spatial variability of a structural plane at the same time; geometric features and deterministic structural plane features of a real footrill model are comprehensively considered, and a plurality of virtual footrill models which consider structural plane space variability and consider continuous random change of the structural plane along the elevation simultaneously are generated. According to field actual footrill data, structural plane characteristics of a blank area without footrill exploration are truly restored as much as possible, and a structural plane network model conforming to geological authenticity is established for subsequent slope stability evaluation work. According to the method, the structural plane network constructed in the model meets the actual adit revealing conditions, meanwhile, the spatial variability of the distribution characteristics of the structural planes of different parts of the slope is considered, and the problem that in rock slope numerical calculation, the structural plane model deviates from reality and influences slope stability evaluation is solved.

The invention discloses a deep learning-based method and related equipment for synthesizing a CT image from a multi-part MR image. The method comprises the following steps: acquiring training data; inputting the first MR image, the first CT image and the clinical text information into a preset network model for training to obtain a pre-training image synthesis model of the plurality of parts after the plurality of training cycles are finished; based on the quantitative proportion in the training data applied by the synthesis of each part, superposing the pre-training weights of the plurality of pre-training image synthesis models corresponding to the training period according to the proportion to serve as an initialization weight initialization preset network model, and then performing re-training to obtain a target image synthesis model; and inputting the real MR image to be synthesized and the real clinical text information into the target image synthesis model to obtain a target CT image. The method can process MR image synthesis of multiple parts, widens the application range of the model, introduces personalized feature information to enrich the input of the model and improve the model synthesis effect, and can be widely applied to the technical field of image processing.

The invention discloses a dynamic optimization layout system and method for a desertification construction camp in combination with digital twinning, and relates to the field of dynamic optimization layout of camps, and the method comprises the steps: constructing a camp candidate region evaluation model, and screening out first five construction camp candidate regions with model comprehensive evaluation values arranged from large to small; scanning the construction camp candidate area to generate a candidate area live-action three-dimensional model; constructing a digital twin three-dimensional model of the camp; carrying out analogue simulation on the five candidate areas, and determining a unique construction camp candidate area through a full-connection neural network; and adopting a mean square error as a deviation index, and dynamically adjusting preset camp condition label data of the digital twin three-dimensional model of the camp. The method has the advantages that the optimal construction camp candidate area is effectively screened for the to-be-constructed environment of the planned large new energy base, the loss of manpower and material resources caused by blind site selection is avoided, and the method for dynamically optimizing the camp layout after the camp is built is provided.

The invention discloses a software monitoring analysis method based on a hierarchical adaptive large model agent, which comprises the following steps of: 1, inputting index data and performing user query on the index data, performing preliminary analysis based on visual perception on the index data, and performing state initialization on a dynamic agent state for recording whole-course information, the input index data is an original time sequence; 2, iterative hierarchical planning and self-adaptive execution: in an iterative loop, selecting an advanced analysis assembly line by a top planner based on the current agent state, dynamically selecting or pruning specific tools and steps in the assembly line according to the context by a bottom executor, executing analysis, and updating the agent state; and step 3, based on the final intelligent agent state after the iteration loop is terminated, synthesizing all information by the multi-mode large model, and generating a final natural language answer to the initial user query.

The invention discloses a visual inspection quality assessment method based on multi-detection model results and large model analysis, which comprises the following steps: inputting image data to be assessed into a plurality of basic target detection models, and obtaining detection results independently generated by each model; performing spatial alignment and structured integration on bounding boxes, categories and confidence information output by multiple models to form unified result representation; if the to-be-evaluated data contains the original annotation information, comprehensively analyzing the difference and consistency between the multi-model detection result and the original annotation by a large model; and if the to-be-evaluated data does not provide the original annotation, generating an annotation quality evaluation result, potential annotation difficulty region analysis and recommended annotation suggestions under an unsupervised condition by the large model based on consistency distribution of a multi-model detection result, a confidence coefficient mode and semantic context reasoning. According to the method, the accuracy, consistency and automation level of labeling work are remarkably improved.

The application discloses a kind of desertification construction camp dynamic optimization layout system and method combined with digital twinning, it is related to camp dynamic optimization layout field, comprising: constructing camp candidate area evaluation model, screening out the construction camp candidate area of five before model comprehensive evaluation value from big to small arrangement;Candidate area real scene three-dimensional model is generated to the construction camp candidate area scanning;Camp digital twinning three-dimensional model is constructed;Five candidate areas are simulated, and the only construction camp candidate area is determined by full connection neural network;Mean square error is used as deviation index, and the preset camp condition label data of camp digital twinning three-dimensional model is dynamically adjusted.The application has the advantages that: the optimal construction camp candidate area is effectively selected for the environment to be constructed in the planned large new energy base, the loss of manpower and material resources caused by blind site selection is avoided, and a method for dynamically optimizing the layout of the camp after the camp is built is provided.

The invention discloses a large language model dynamic valuation method and system based on parallel game, and the method comprises the steps: firstly constructing a parallel game evaluation environment, and setting game scene parameters; and loading the to-be-evaluated large language model as an intelligent agent, and distributing roles and initial contexts. And then driving the intelligent agent to perform multi-round interactive gaming according to rules, monitoring violation in real time and recording data. And multi-dimensional evaluation results such as winning rate, influence, collaborative contribution degree and compliance are calculated based on the data. And finally, performing weighted fusion on the multi-dimensional result according to the evaluation index weight to generate a comprehensive capability estimated value of each model, thereby providing a more comprehensive and accurate basis for model evaluation.

The invention discloses a cast-in-place pile bearing stratum prediction method based on multi-modal data depth fusion, and the method comprises the steps: integrating the original drilling of a cast-in-place pile and multi-modal data, and mining information through deep learning, thereby reducing the extra drilling, and reducing the cost; meanwhile, through data preprocessing and deep learning model rapid calculation, after model training is completed, a result is obtained by inputting pile position coordinates, the period is greatly shortened, and the project progress is improved; besides, multi-modal data are integrated to construct comprehensive feature vectors, model deep fusion analysis data and learning position association rules, large-area site bearing layer prediction is achieved, the result is more comprehensive and continuous, and the defects of a traditional method are overcome. Besides, a multi-modal feature vector covers multiple types of information, a cross-modal attention fusion module and other deep learning technologies consider inter-modal connection, the model integrates various factors to mine a deep feature mode, and key information such as the type, the burial depth and the thickness of a bearing stratum can be predicted more accurately.