17 results about "Full model" patented technology

The invention discloses a welding simulation heat source parameter calibration method and system based on a hybrid agent model, and belongs to the technical field of welding numerical simulation and optimization, and the method comprises the following steps: S1, problem definition and experiment design; s2, analyzing a product structure and constructing a joint model; s3, parameter space sampling; s4, performing simulation calculation; s5, constructing a mixed agent model; s6, proxy model optimization based on a genetic algorithm; s7, parameter verification and model updating; and S8, optimizing and outputting optimal parameters. According to the method, most time-consuming full-order finite element calculation is replaced by the proxy model and the simplified joint model, the parameter calibration process is shortened from several days or weeks to several hours, and the optimization efficiency is improved; a simplified model and a complete model are used for optimization, firstly, the simplified model is used for rapidly obtaining an optimal parameter approximate value, and the complete model is used for obtaining accurate parameters.

Systems and methods are provided for creating a machine learning (ML) model staging repository, where ML models are stored as an open container image (OCI). The OCI comprises layers or portions of the complete ML model, so that the model can be stored separately and as a smaller files. The ML models may be pre-packaged for automated downloads and integration at the customer site. In some examples, the OCI can identify / store the model in a directory structure that defines the model and its profile. The OCI can comprise a combination of layers and profiles that allows the AI platform to optimize the storage and simplify the process of downloading or updating the given user namespace instead of downloading a single large file for the model.

This invention discloses a method and apparatus for accelerating distributed training of large AI models based on model prediction. The method includes: constructing a hierarchical parameter dataset; training a prediction model for predicting model parameters of a target model using an evaluation method based on prediction results; computing nodes updating the model parameters of some layers of the target model and then pushing them to a parameter server; on the parameter server, aggregating the received model parameters and predicting the model parameters of the remaining layers of the target model using the prediction model; concatenating the predicted and aggregated model parameters into a complete model parameter set and pushing it to all computing nodes; and replacing the local model parameters with the received complete model parameter set. The prediction model adopts a "convolution + channel attention mechanism + pooling" structure. This invention can significantly reduce communication volume without introducing a large amount of computational overhead while maintaining the accuracy of the target model.

The application discloses a kind of temperature field simulation calculation method and device based on transformer model, by obtaining and according to transformer parameter, transformer three-dimensional one-grade model is constructed, wherein the transformer three-dimensional one-grade model includes core, single-phase winding and oil tank;Based on the model single-phase current of the transformer three-dimensional one-grade model is calculated, the model single-phase current calculated is input into the transformer three-dimensional one-grade model, and the single-phase loss density of transformer is obtained;The model simulation parameter of transformer is set, according to the model simulation parameter and the single-phase loss density, the transformer three-dimensional one-grade model is iterated and simulated, and the transformer temperature field simulation result is obtained.Compared with prior art, the technical scheme of the present application simplifies the existing transformer model, solves the problem of large amount of calculation, difficult mesh partitioning and long calculation time in the existing three-dimensional full model, and can improve the simulation calculation efficiency of transformer three-dimensional one-grade model.

The invention relates to a model training method and device based on data flow, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps that under the condition that a reading condition is met, training data are read from a preset cache region, and the preset cache region is used for storing real-time data of a data stream; based on the training data, performing parameter fine tuning training on an increment module in the to-be-trained model, and taking the trained to-be-trained model as an increment model; the to-be-trained model is obtained after an increment module is loaded on the main model; and generating a target model according to the main model and each incremental model. Therefore, by introducing the increment module, the traditional fine-tuning training of full-amount model parameters of the main model is changed into efficient parameter fine-tuning of the lightweight increment module, so that the computing resource consumption and the time cost in the model training process are effectively reduced, and the model training efficiency is improved; and the updating timeliness of the model adaptive business data is ensured.

The application discloses a kind of based on rough contact interface turbomachinery center pull rod rotor modeling analysis method, belong to turbomachinery high temperature rotor technical field.The present application includes calling pre-trained contact stiffness prediction model, input current operating condition roughness parameter and load parameter, obtain the predicted stiffness of contact interface;The three-dimensional finite element model of rotor is established, the mesh division is carried out to contact interface area, and nonlinear spring damping unit is set between the corresponding nodes of contact interface;The predicted stiffness obtained is given to the nonlinear spring damping unit, and the reduced degrees of freedom are carried out to the finite element model after stiffness assignment using hybrid modal reduction method, and the reduced model is solved for vibration response, the fidelity of the model is ensured, the full model solving degree of freedom is greatly compressed by hybrid reduction method, while not affecting the accuracy of the model, reduce the amount of calculation, facilitate the analysis and solution of model.

The application discloses a kind of based on the construction method and system of long-term future flow section in power grid of multi-version model.The method is based on model multi-version library to query future state model version information;According to future state model version information, corresponding model data and current real-time state model data are extracted in model multi-version library to form complete full model data by model splicing, and real-time state estimation mode data is inserted into full future model;Then, for the incremental equipment in future version model, automatic grid-connection processing is carried out relying on deep topological search technology, and complete initial mode data is given to the future model;Finally, according to multiple future model versions, the future section of multiple time scales is intelligently grouped, future boundary planning data is superimposed, and the construction of long-term future flow section is completed.The application can effectively improve the accuracy and reliability of future flow calculation.

Provided are computer-implemented systems and methods which improve embedding generation model performance with respect to localized quality issues. In particular, the proposed techniques can operate to improve localized embedding model performance without incurring significant re-training costs and / or re-deployment costs associated with full model re-training.

This invention provides a parametric modeling method and system for a full model of a side-type inlet / outlet, relating to the field of hydraulic engineering technology. The method includes: generating a coordinate transformation matrix based on the engineering parameters and number of inlets / outlets; integrating and processing the two-dimensional control points of the inlet / outlet structure's outline with unit normals to obtain a homogeneous unit normal matrix, and determining equidistant offset lines tangent to the pier head arc of the inlet / outlet structure; generating two-dimensional control points for the pier head arc based on the equidistant offset lines; stacking all two-dimensional control points of the inlet / outlet structure into a control point matrix to generate a three-dimensional model of the inlet / outlet structure; combining the engineering parameters of other structures to generate three-dimensional models of other structures, and then combining the coordinate transformation matrix to generate a full model of the side-type inlet / outlet. This invention improves the speed of parametric modeling of a full model of a side-type inlet / outlet reservoir area and enhances the stability of the modeling process.

The invention discloses a wind tunnel full-model test section overall frame structure which comprises an upper frame, a lower frame and stand columns, the stand columns are connected between the upper frame and the lower frame, the upper frame, the lower frame and the stand columns jointly form a front-back through box type frame, and the front ends of the upper frame, the lower frame and the stand columns are connected with inlet flanges. The rear ends of the upper frame, the lower frame and the stand columns are connected with outlet flanges. The wind tunnel test platform has the beneficial effects that a traditional passive support platform is upgraded into an active controllable platform, the pain points of geometric stiffness, difficult transformation, difficult interference control and the like of a static frame are overcome through modular and mechanical-electrical integration design, and cooperative rise of efficiency, precision and flexibility of a wind tunnel test is realized; and physical and control interfaces are reserved for future function extension.

The invention relates to a nozzle throat liner ablation prediction system based on data mining, which belongs to the technical field of aerospace and comprises a data import and loading module, a full model training module, a selective verification module, a prediction module and a result export module. The data importing and loading module is used for importing historical data and inputting the historical data into the model training module; the full model training module processes the data through a plurality of data mining algorithms to generate a prediction model; the selective verification module is used for selecting different types of models for training and verification according to requirements; the prediction model performs ablation prediction through real-time data input; the result export module stores the prediction result; the method has the beneficial effects that the ablation trend of the nozzle throat liner can be predicted in advance, replacement is carried out in time, nozzle faults caused by serious ablation are avoided, assistance is provided for designers to accurately estimate an inner trajectory and provide accurate ablation data, and the safety and reliability of a system are improved.

The application discloses a 3D modeling repair method based on a point cloud deep learning incomplete model, and relates to the technical field of 3D modeling repair.The specific steps of the method are as follows: firstly, pre-processing and normalization are performed on cross-source heterogeneous incomplete point clouds collected by multiple sensors to unify the input format;then, Bayesian confidence probability modeling and topology propagation are performed to optimize the confidence;then, the high-uncertainty area is actively calibrated;finally, the calibrated confidence is used as the weight to fuse features and complete the incomplete point cloud by using a cross-source Bayesian weighted fusion repair algorithm to generate a complete 3D model.The application uniformly formats the point cloud data of multiple sensors by pre-processing and normalization, accurately calculates and optimizes the confidence, and reduces the noise error;then, the calibrated confidence is used as the weighted fusion basis to accurately fuse geometric features and supplement missing data, effectively restores the complete form of the incomplete model, improves the accuracy and integrity, and adapts to various 3D modeling repair requirements.

The invention discloses a simulation method and related equipment for a vacuum infusion process of a wind power blade pultrusion beam, and the method comprises the steps: building a three-dimensional solid model based on an actual layering scheme of the wind power blade pultrusion beam, and building a simulation model of the vacuum infusion process according to the three-dimensional solid model; giving material attributes and boundary conditions to the simulation model to obtain a complete model; integrating the complete model into a resin flow control equation for simulation, and predicting dynamic flow parameters of the resin; the dynamic flow parameters comprise a flow front form, a pressure field and filling time. According to the embodiment of the invention, the accuracy of predicting the race flow effect can be improved. The method can be widely applied to the technical field of process simulation.

The invention relates to the technical field of robot control, and provides a control method, device, equipment, medium and program for an intelligent robot with a body, and the method comprises the steps: receiving a task language instruction and environment visual observation data collected by the robot in real time; obtaining a corresponding original soft prompt vector according to the task language instruction; inputting the environment visual observation data and the task language instruction into a preset finite-state machine in the intelligent robot system with the body, executing task logic state verification, and outputting a prompt update judgment signal based on a verification result; when the prompt updating judgment signal indicates that updating is needed, a correction soft prompt vector is generated based on the environment visual observation data; and fusing the corrected soft prompt vector and the original soft prompt vector to generate an updated prompt. The method is used for improving the control adaptability and reliability of the intelligent robot with the body in a dynamic environment through a lightweight dynamic prompt mechanism, precise action generation and long-range task logic coherence are achieved, and full-model fine adjustment is avoided.

A computer-implemented method in a model-based reinforcement learning (RL) system with logic states includes switching an agent between a first mode and a second mode, the first mode being a probabilistic planning mode and the second mode being an information gathering mode. In response to the agent being in the probabilistic planning mode, the agent computes a predictive state representation, given a history of observations and actions taken, and the agent scores action candidates based on planning with the predictive state representation so that actions with resolved plans with confidence to achieve a goal state are scored higher than actions without resolved plans. In response to the agent being in the information gathering mode, the agent scores action candidates based on a Q function of a value of expected information to be gathered from a given pair of state and action.

The application provides a cableway bridge parameterized modeling method based on geometric form finding and space topology mapping, relates to the technical field of bridge engineering digital design and modeling, and the technology comprises the following steps: constructing a calculation domain, analyzing form finding, constructing an assembly reference, component topology mapping and full model generation. The application establishes a full parameterized driving dynamic model: all geometric features of the model are driven by input design parameters and topology relations. Through preset parameter dependence relations and update sequences, the system can automatically and orderly update the whole three-dimensional model when the parameters are changed, realizes "one modification, everywhere update", significantly improves the design iteration and scheme comparison and selection efficiency, and solves the core problem of parameter and model disconnection.

The application discloses a device overhaul simulation control method based on a dynamic loading model, and relates to the technical field of virtual simulation; the method aims at the problem of large device full model rendering lag, and proposes a resource scheduling strategy based on a mixed weight, comprehensively considers the visual centrality of components in a view frustum and the semantic association strength in the current overhaul task, constructs a loading priority index model, realizes high-precision priority loading of key operation components and on-demand degradation of irrelevant backgrounds, and combines an asynchronous multithreading architecture and a memory dynamic release mechanism based on future task prediction, so that the smoothness of core interaction is ensured, the video memory occupation is greatly reduced, and the hierarchical light rendering logic is matched, thereby realizing high-fidelity and low-delay simulation of a complex industrial device overhaul process under limited hardware resources.