24 results about "Modal synthesis" patented technology

The invention relates to the technical field of artificial intelligence, in particular to a multi-modal data retrieval, generation and synthesis method and system based on artificial intelligence driving, and the method comprises the steps of multi-modal feature extraction, cross-modal alignment, feature fusion, multi-modal retrieval and sorting and multi-modal generation. Different deep learning models are adopted to carry out feature extraction on multi-modal data and convert the multi-modal data into vectors, through cross-modal alignment, different modal feature vectors are mapped to the same vector space, through feature fusion, correlation weights are calculated through an attention mechanism, weighted summation is carried out on fusion features, and through multi-modal retrieval and sorting, multi-modal data are obtained. According to the method, the similarity between a query vector and candidate data is calculated and sorted, and finally, through multi-modal generation, retrieval knowledge is used as external knowledge to be fused into a generation model, and a multi-modal synthesis answer is generated, so that information in multi-modal data can be effectively integrated and utilized.

The application discloses a kind of vibration characteristics analysis methods of water turbine guide vane shaft-top cover rubbing coupling system, comprising the following steps: S1, using finite element method in combination with fixed interface modal synthesis method, the reduced dynamic model of guide vane shaft-top cover rubbing coupling system is established;S2, based on reduced dynamic model analysis mode truncation number, the influence law of friction coefficient and water flow surface load on the prestressed modal characteristics of guide vane shaft-top cover rubbing coupling system first 3 orders;S3, when reduced dynamic model is subjected to simple harmonic excitation, the influence law of near load end contact state, near-far load end contact state and friction coefficient on the vibration response and frequency spectrum characteristics of guide vane shaft-top cover rubbing coupling system;S4, based on the above steps, whether mixed-flow water turbine is stably and safely operated is evaluated.The application is analyzed by vibration characteristics of mixed-flow water turbine guide vane shaft-top cover rubbing coupling system, to monitor and diagnose the structural health state and failure of mixed-flow water turbine.

The invention relates to the technical field of aerospace, in particular to a method for determining the modal and vibration characteristics of a helicopter composite material tail shaft, which comprises the following steps: establishing an equivalent analysis model of the helicopter composite material tail shaft by referring to the structural characteristics of a helicopter; dividing the equivalent analysis model into cylindrical shell substructures with different thicknesses based on a regional decomposition theory, and obtaining a vibration analysis model of the cylindrical shell substructures; introducing an improved fixed interface modal synthesis method to further construct a spectral Chebyshev reduced-order model of the cylindrical shell substructure; performing equivalent treatment on the interfacial force between the adjacent cylindrical shell substructures according to an artificial spring theory; and assembling and analyzing the cylindrical shell substructure to obtain an overall reduced-order model of the helicopter composite material tail shaft, and solving to obtain the modal and vibration characteristics of the helicopter composite material tail shaft. The order reduction model established by the method does not depend on grid quality any more, and meanwhile, the method has remarkable value for improving the effective load and fuel economy of the helicopter.

The invention provides a helicopter tail rotor and tail beam coupling modeling method based on a time-varying mixed interface modal synthesis method, and the method comprises the steps: firstly considering the rotation influence of a blade, and building a blade structure kinetic equation under a physical coordinate system; a blade structure kinetic equation under the modal coordinate system is obtained through modal coordinate transformation, tail beam modal information is obtained, and a tail beam structure kinetic equation under the modal coordinate system is established. Thirdly, a time-varying effect brought by rotation of the blade is considered, a quadratic coordinate transformation formula of modal speed and modal acceleration is obtained through conversion according to the displacement boundary condition between the blade and a tail beam position structure, and quadratic coordinate transformation is conducted on a blade structure kinetic equation under a modal coordinate system based on the quadratic coordinate transformation formula; and finally, assembling the blades and the tail beam according to a number-to-number seating principle to obtain a tail rotor and tail beam coupling time-varying structure kinetic equation under the modal coordinate system. The method is suitable for the coupling dynamics problem of the uniformly and non-uniformly arranged tail rotor and the tail beam, and the calculation efficiency in the modeling process is remarkably improved.

The invention relates to an autonomous generation agent system and method for a multi-mode creation task. The system comprises a master control planning unit, an extensible tool library, a scheduling execution and evaluation unit, a conversational progressive combination generator and a multi-modal synthesis unit. Generating a directed acyclic graph by disassembling the constituent elements; identifying sub-tasks without preposition dependence in the directed acyclic graph through a topological sorting algorithm and adding the sub-tasks into a queue to be executed; matching and calling a tool based on the ReAct framework, receiving a tool execution result and generating a structured evaluation report; and the user interaction window can be circulated until the user accepts the core content, and a final multi-modal file is synthesized. By disassembling a complex task, executing automatic process management and planning the task into a directed acyclic graph, the total consumed time can be shortened; the conversational progressive combination generator allows a user to guide, check and finely adjust the generation process in a conversational mode at key nodes, and the quality and stability of generated content are improved.

The present application belongs to the technical field of image fusion, and particularly relates to a weakly supervised medical image fusion method for modality synthesis and enhancement. In view of the problems of human body injury and high cost caused by multiple imaging of the patient for multi-modal image fusion, long time and high difficulty of image fusion requiring registration of a large number of paired data sets, and unclear texture and fine structure of the fused image, the present application provides a weakly supervised medical image fusion method for modality synthesis and enhancement based on CycleGAN and OctopusNet, which is composed of three parts of modality synthesis, image enhancement and image fusion.

The present invention provides a method, medium and equipment for predicting time-varying dynamic parameters in the milling process of a thin-walled workpiece, comprising: establishing a finite element model of the thin-walled workpiece and dividing the grid, dividing the substructure into a finished product area substructure and a material to be removed area substructure, and sorting the substructures according to internal nodes and interface nodes. An improved fixed interface modal synthesis method is used to establish a degree of freedom reduction model for the finished product area substructure. Considering the material removal effect, the degree of freedom reduction model for the material to be removed area substructure is updated in real time. The two substructure reduction models are coupled and connected according to the equilibrium condition of the substructure interface, and the model calculation is performed. Finally, the above steps are iterated according to the tool feed position to update the time-varying dynamic parameters in the milling process of the thin-walled workpiece. The present invention takes into account the influence of the material removal effect of the thin-walled workpiece in actual processing, reduces the calculation time while ensuring its prediction accuracy, and improves the calculation efficiency.

The invention discloses an overseam belt presser foot gap noise analysis method which comprises the following steps: step 1, establishing a rigid body dynamic model of a presser foot mechanism and a tooth feeding structure, extracting a contact force value and a change rule in an impact process between parts, and determining a noise contribution part; 2, the finite element mesh of the noise contribution part is flexible through a modal synthesis method and replaced into a rigid body dynamic model of a presser foot mechanism and a tooth feeding structure, a rigid-flexible coupling model is built, displacement, speed response and surface vibration radiation noise ERP of the noise contribution part in the movement process are obtained, and the ERP is calculated; and vibration response area distribution of the noise contribution part is obtained.

An embodiment of the present application provides a method, system and readable storage medium for cross-modal synthesis of medical images, the method including determining a pair of medical images corresponding to the same part, the medical image pair including a matching MRI image and a CT image for training; performing cross-modal synthesis of the MRI image based on a constructed initial generative adversarial network model to obtain a corresponding CT image, the initial generative adversarial network model including a joint attention residual processing module for jointly extracting channel attention features and global attention features from the image to determine the corresponding joint attention features; performing model training based on the matching medical image pairs, and during the training process, performing downsampling and upsampling operations on the input MRI image in combination with the joint attention residual processing module, and obtaining a target generative adversarial network model at the end of the training; inputting the acquired MRI image to be processed into the target generative adversarial network model to obtain a corresponding CT image synthesis result.

The invention discloses a medical image cross-modal synthesis method, a terminal device and a storage medium, and the method comprises the steps: building a bridge type diffusion generation path with a source mode and a target mode as end points, and improving the cross-modal alignment and structure fidelity; putting forward an internal and external joint condition injection strategy, generating preliminary estimation by using external conditions, refining based on self-conditions, and realizing reuse of intermediate results and stable detail recovery; and by constraining the consistency of output prediction of adjacent time steps, the time sequence coherence and robustness of the model in the generation process are enhanced.

The application discloses a kind of multi-modal fatigue driving detection methods based on thermal imaging, comprising: using thermal imaging equipment to obtain driver face image data, using YOLO v8 model to carry out face and key point detection, extract the region of interest at forehead and nose, and calculate the mean of image pixel gray level to obtain heartbeat and respiratory signal;Signal processing is carried out to the physiological signal obtained, and the physiological characteristics related to fatigue are extracted, and fatigue and non-fatigue data set is constructed to train fatigue detection model;The thermal imaging image of fatigue and non-fatigue is extracted by face key point, and a two-classification model is trained to detect fatigue;The decision fusion of multi-model is carried out to face fatigue feature model and physiological characteristic model, and the fatigue state of driver is judged using multi-modal synthesis.The application improves the accuracy and robustness of fatigue driving detection model by the decision fusion of multi-model, can adapt to more complex light environment and individual difference, effectively improves the precision of fatigue driving recognition.

This application discloses a method and related apparatus for determining an image synthesis model. The embodiments of this application can be applied to various scenarios such as cloud technology, artificial intelligence, and medical imaging. In this method, the image synthesis model learns the modal differences between a first modality and a second modality through training, reducing the impact of these differences on the image synthesis model and improving feature extraction accuracy. This enhances the transferability of effective features in cross-modal synthesis, enabling the image synthesis model to accurately convert the image features of the target modality image based on the output features of the source modality image, synthesizing an image that more closely resembles the real target modality, and improving image acquisition efficiency in multimodal scenarios.

The invention provides a multi-flexible-body coupling dynamics modeling method and device for a rotary non-inertial system gear system, and the method comprises the steps: dividing the gear system into a shaft system substructure and a box body substructure, and respectively building the kinematic equations of the shaft system substructure and the box body substructure; a generalized coordinate kinetic equation of the box body is obtained, the generalized coordinate kinetic equation comprises fixed-axis rotation coordinates and flexible node coordinates, and the non-inertial system effect of random body rotation is included; a shaft system generalized coordinate kinetic equation is obtained, discretization is carried out, and a composite motion effect is included; performing order reduction on the box body and the shaft system through a state space fixed interface modal synthesis method, retaining interface nodes and mapping internal nodes to a low-order modal space; bearing coupling is achieved through multi-point constraint, and shafting bearing force is equivalently distributed to flexible joints of a box body; establishing a gear meshing coupling model, and loading a gear backlash, a meshing error and time variable stiffness; and all subsystem equations are coupled to form an integral reduced-order model, and a dimension-increasing fine integration algorithm is adopted for solving. The invention provides an efficient and accurate tool for the design of a gear transmission system.

The invention relates to a method for acquiring dynamic characteristics of a stacked satellite assembly, which belongs to the field of spacecraft design and kinetic analysis, and comprises the following steps: firstly, establishing a single-layer satellite kinetic model, then acquiring actually measured modal parameters of a single-layer satellite through a ground modal test, and correcting the single-layer satellite kinetic model by using a modal confidence criterion; establishing an interface model, and obtaining an interface equivalent dynamic model through theoretical calculation and a simulation method; simulating an axial pretightening force, a shear load and a bending moment condition under an actual launching working condition, and correcting the interface equivalent dynamic model; based on a modal synthesis method, obtaining an N-layer stacked satellite combination dynamic model; acquiring actual modal parameters and dynamic response data of the stacked satellite assembly, and performing reverse correction according to the actual modal parameters and the dynamic response data; and assembling the combination dynamical model and the rocket dynamical model, carrying out satellite-rocket integrated joint simulation, and optimizing the N-layer stacked satellite combination dynamical model according to a simulation result.

The application discloses a kind of LLM privacy inference safety evaluation method and system based on multimodal synthesis, belong to model safety analysis field.Based on the preset virtual privacy attribute, using the way of multi-agent cooperation, the multimodal synthesis data of simulated social scene is automatically generated, and the multimodal synthesis data includes user portrait instance, theme list, event list, post text, picture prompt word and associated image;The theme list, post text and associated image in the multimodal synthesis data are input into the large language model to be evaluated, the inference of the virtual privacy attribute is carried out, the inference result is obtained, and the original virtual privacy attribute corresponding to the user portrait instance is compared with the inference result, to generate quantitative privacy risk score.The application can automatically generate simulated multimodal social content, quantify model privacy inference risk, establish unified benchmark horizontal comparison model privacy protection level, and construct end-to-end evaluation closed-loop framework.

The application discloses a kind of shipboard aircraft dynamics fast calculation method.Method includes following process: establishing shipboard aircraft catapult take-off multi-flexible body system dynamics model, establishes shipboard aircraft catapult take-off multi-flexible body system topology, utilize finite element method to establish main tire refinement finite element model, then tire flexible body dynamics characteristic parameter is exported and deduces the dynamics equation of tire flexible element, based on multi-body system transfer matrix method (MSTMM) and finite element modal synthesis method deduce shipboard aircraft catapult take-off multi-flexible body system total transfer equation and total transfer matrix, solve the inherent frequency of shipboard aircraft multi-flexible body system considering inflatable flexible tire and the dynamic response of catapult take-off;The method proposed in the application overcomes the defects of high matrix order and large amount of calculation of traditional methods when dealing with complex coupling system of shipboard aircraft containing inflatable flexible tire, has the advantages of fast calculation speed, high degree of programming and strong universality, provides an efficient calculation method for the dynamic analysis of complex shipboard aircraft multi-flexible body system.

The invention provides a modal parameter prediction method for hybrid modeling of a simulation model and a test model, which adopts a free interface modal synthesis method based on residual dynamic flexibility, forms a comprehensive equation of a system by utilizing section-level simulation analysis or dynamic test data, and solves modal parameters of the system. The method is suitable for a complex system with a plurality of substructures and a plurality of butt joint surfaces, and can provide data support for rocket structure design, flight attitude control system design and load and environment analysis.

The invention provides a marginal plate damping device optimization method based on Monte Carlo simulation and Latin hypercube design, and the method comprises the steps: selecting a certain order of vibration mode of a blade mode to carry out the structure optimization design of a damper based on the vibration mode of the blade mode, and determining the harmonic response frequency range of the blade; processing a blade group substructure by a dual-mode synthesis method, and combining an overall structure matrix through a substructure reduction matrix; obtaining a parameter matrix and a two-dimensional contact surface friction model, and constructing a dynamic equation set with dry friction nonlinearity; a dynamic equation set with dry friction nonlinearity is obtained, and an amplitude-frequency characteristic curve is solved through a harmonic balance method and an arc length continuation method; monte Carlo simulation and Latin hypercube design are carried out, a multi-parameter Monte Carlo simulation analysis program is written, and parameter optimization analysis with the optimal damping effect of the damping device as the target is completed; and obtaining an optimal parameter group to further optimize the original damping structure. The model is processed through a substructure method, and the vibration response of the blade group with dry friction damping under simple harmonic excitation can be predicted.

The invention discloses an LLM privacy reasoning security assessment method and system based on multi-modal synthesis, and belongs to the field of model security analysis. Based on a preset virtual privacy attribute, multi-modal synthetic data for simulating the social scene is automatically generated in a multi-agent cooperation mode, and the multi-modal synthetic data comprises a user portrait instance, a theme list, an event list, a post text, a graph speaking prompt word and an associated image; and inputting the theme list, the post text and the associated image in the multi-modal synthetic data into a to-be-evaluated large language model, executing multi-modal virtual privacy attribute reasoning to obtain a reasoning result, and generating a quantitative privacy risk score by comparing the reasoning result with an original virtual privacy attribute corresponding to the user portrait instance. According to the method, the simulated multi-modal social content can be automatically generated, the privacy reasoning risk of the model is quantified, the privacy protection level of a unified reference transverse comparison model is established, and an end-to-end evaluation closed-loop framework is constructed.

This invention relates to the field of dynamic characteristic analysis technology for complex integrated energy systems, and particularly to a method for extracting key features of an electro-gas coupled system based on modal synthesis. The method includes: acquiring the dynamic nonlinear models of components in the electro-gas coupled system; obtaining a unified nonlinear equation set in the electro-gas coupled system based on the dynamic nonlinear models of the components, and then obtaining a unified linear equation set through linearization, thereby obtaining the system state matrix; obtaining the system's full-order eigenvalues ​​using the eigenvalue decomposition method based on the obtained system state matrix, analyzing the system's full-order modes, and then using modal synthesis to analyze the full-order eigenvalues ​​to obtain the system's key dominant modes; analyzing the dynamic process of the system based on the obtained key dominant modes, and performing simulation. This invention reduces the difficulty of dynamic simulation by reducing the system's high-order dynamics to low-order dynamics, effectively solving the problem of inaccurate time scale division in previous methods.

This invention discloses a cross-modal synthesis method, terminal device, and storage medium for medical images. It establishes a bridge-type diffusion generation path with the source modality and target modality as endpoints to improve cross-modal alignment and structural fidelity. It proposes an internal and external joint condition injection strategy, which uses external conditions to generate preliminary estimates and refines them based on self-conditions to achieve reuse of intermediate results and stable detail recovery. By constraining the consistency of output predictions at adjacent time steps, it enhances the temporal coherence and robustness of the model in the generation process.

The invention discloses a CBCT image cross-modal synthesis method based on a mixed state space model. The CBCT image cross-modal synthesis method comprises the steps of data preprocessing and 2.5 D input sequence construction; constructing a focus slice correction branch Branch A, and removing artifacts in the slices and among the slices; constructing a volume scene context modeling branch Branch B, and extracting three-dimensional space consistency features; a mixed global-local Mamba decoder Branch C is constructed, and a high-fidelity sCT image is generated; and training and optimizing the model. According to the method, the physical fidelity of the synthesized CT image and the continuity of an anatomical structure are remarkably improved, the computational efficiency bottleneck of high-resolution medical image processing is broken through, the clinical-level real-time reasoning speed is realized while the 3D-level modeling capability is ensured, an ambiguity elimination mechanism based on multi-scale context comparison is constructed, and the real-time reasoning speed is improved. And the clinical safety and reliability of the generated image are obviously improved.

The invention relates to the technical field of vehicle modeling, in particular to a modeling and simulation optimization method for dynamic transfer characteristics of a tire and related equipment. The method comprises the following steps: establishing a finite element model after a tire and a rim are assembled into a wheel assembly, and performing inflation pressure static analysis on the finite element model to obtain a prestress configuration in an inflation state; performing linearization processing on the prestress configuration, and extracting modal characteristics of the wheel assembly in a preset frequency range; based on the modal characteristics, a modal synthesis method and a transfer function analysis method are adopted, and the dynamic transfer characteristics from the tire grounding point to the wheel center are calculated; wherein the dynamic transfer characteristic is defined as a transfer function from the input force at all grounding action points to the output acceleration at the wheel center within a preset frequency range; and loading the model representing the optimized dynamic transmission characteristics into a whole vehicle noise simulation analysis model, analyzing and optimizing road noise response in the vehicle, and realizing matching of the tire dynamic transmission characteristics and the whole vehicle.

The invention relates to the technical field of artificial intelligence machine learning, and discloses a domain adaptation method from a multi-modal synthesis image to a real image. The method is used for solving the problems that in a traditional method, semantic drift and negative migration are generated by superposition of multi-source domain differences, and a label-free real domain structure keeps measurement constraint and dynamic gating of task consistency. The method comprises the following steps: firstly, establishing a domain adaptation session, a frozen data aperture, a threshold aperture, a resource budget and a version number; performing multi-modal quality, synchronization and geometric consistency verification on the synthetic domain and real domain samples, implementing risk gating admission and initializing modal weights; correspondingly establishing a structure fingerprint baseline according to the synthetic domain label anchor point and the cross-modal geometry, and calibrating a structure threshold boundary; performing style migration and structure checking on the synthetic domain sample, and re-generating unqualified callback intensity; and confronting alignment training and carrying out version sealing, rechecking and rollback to realize reproducible cross-domain adaptation.