208 results about "Large model" patented technology

A method of dynamic model predictive control is presented in which both steady state optimization and dynamic moves calculation of the manipulated variables are determined together as part of one optimization solution (211). The method combines steady state optimization and dynamic move calculation such that the steady state optimal targets are determined consistent with the dynamic moves and the resulting dynamic response of the process so that the controlled variables do not violate their low / high limits in both steady state and dynamically. The method utilizes what is described as high limit dynamic violation variables and low limit dynamic violation variables corresponding to each of the controlled variables. The method offers a unique capability of mixing a part dynamic optimization with a part steady state optimization as it relates to application of a large model predictive control under varying situations, in particular at or near constraints violations that can be done dynamically in real time for improved and robust control performance without having to change the controller tuning.

The application aims to provide a data processing method and device based on a neural network. The method comprises steps that precision conversion for single precision floating point type data of the neural network is carried out; neural network calculation for the low precision floating point type data after precision conversion is carried out. Compared with the prior art, the method and the device are advantaged in that solving a large storage content problem in a low precision mode is facilitated, the memory space occupied by a model can be saved, a larger model can operate through utilizing same hardware configuration, and the memory or hardware space occupied by a data set can be saved; when the model is deployed on a cluster, network bandwidth required during synchronization can be effectively reduced, communication cost can be effectively reduced, and integral performance is improved.

Embodiments disclosed herein provide systems and methods for preparing geometry for 3D printing. In one embodiment, a 3D printing preparation application receives 3D geometry and repairs non-manifold edges and non-manifold vertices, producing a topological manifold geometry. The 3D printing preparation application then welds coincident edges without coincident faces and fills holes in the geometry. The 3D printing preparation application may further perform resolution-aware thickening of the geometry by estimating distances to a medial axis based on distances to distance field shocks, and advecting the distance field using a velocity field. A similar approach may be used to perform resolution-aware separation enforcement. Alternatively, one component may be globally thickened and subtracted from another for separation enforcement. The 3D printing preparation application may also split large models and add connectors for connecting the split pieces after printing. In addition, the 3D printing preparation application may generate a 3D print preview.

A new feedback linearization approach to advanced control of single-unit and multi-unit HVAC systems is described. In accordance with the approach of the invention, this new nonlinear control includes a model-based feedback linearization part to compensate for the nonlinearity in the system dynamics. Therefore, the evaporating temperature and superheat values can be controlled by linear PI control designs to achieve desired system performance and reliability. The main advantages of the new nonlinear control approach include (1) better performance even with large model errors, (2) being able to adapt to indoor unit turn on / off operation, (3) much smaller PI control gains compared to that of current feedback PI controls, (4) much easier design procedures since there is no need for tuning the PI control gains over wide range operation.

The invention discloses an electromagnetic scattering simulation method based on parallel moment method and physical optics mixing to mainly solve the problem that when electrically large models are processed in the prior art, the simulation efficiency of electromagnetic scattering and the precision are low. The electromagnetic scattering simulation method comprises the steps that a rough surface power spectrum and rough surface parameters are input, and a rough surface is simulated based on the Monte Carlo method; geometric modeling is carried out on a radar target through 3dsMax software, and the subdivided radar target is output; a combined model is formed by combining the rough surface and the radar target, the rough surface is divided into a physical optics area, and the radar target is divided into a parallel moment method area; induced currents on the rough surface and induced currents on the surface of the radar target are obtained; a far scattered field of the combined model is calculated based on the induced currents on the surface of the combined model, a radar cross section of the combined model is obtained, and simulation between the rough surface and the combined electromagnetic scattering of the radar target is achieved. According to the electromagnetic scattering simulation method, the memory consumption is little, the simulation efficiency is high, and the electromagnetic scattering simulation method can be used in the study on the rough surface and the radar target combined electromagnetic scattering character in the field of national defense and the civil using.

A scale-Space feature extraction technique is based on recursive decomposition of polyhedral surfaces into surface patches. The experimental results show that this technique can be used to perform matching based on local model structure. Scale-space techniques can be parameterized to generate decompositions that correspond to manufacturing, assembly or surface features relevant to mechanical design. One application of these techniques is to support matching and content-based retrieval of solid models. Scale-space technique can extract features that are invariant with respect to the global structure of the model as well as small perturbations that 3D laser scanning may introduce. A new distance function defined on triangles instead of points is introduced. This technique offers a new way to control the feature decomposition process, which results in extraction of features that are more meaningful from an engineering viewpoint. The technique is computationally practical for use in indexing large models.

The invention relates to an ultra-large type geotechnical engineering three-dimensional model test system, which comprises an ultra-large type model test device, a hydraulic loading system and a stress displacement test system, wherein the hydraulic loading system and the stress displacement test system are connected with the ultra-large type model test device respectively; the ultra-large type model test device comprises a door-type frame; the lower end of the door-type frame is arranged on a bottom plate device; and a model reverse wall positioned on the lower part of the door-type frame is arranged on the bottom plate device. The ultra-large type geotechnical engineering three-dimensional model test system has the advantages that: 1) the ultra-large type geotechnical engineering three-dimensional model test system has compact integrated structure, high strength and large rigidity and can bear kiloton reverse force; 2) the ultra-large type geotechnical engineering three-dimensional model test system can satisfy the requests of the different height tests; 3) the bottom plate is assembled and provided with a T-shaped slide groove; and the reverse wall can be expanded front and back or left and right or fixedly shrunk so as to meet requirements on different model sizes; 4) the model test device is conveniently mounted and dismounted to finish different plane tests; 5) a stress displacement test system can automatically and highly-precision collect and analyze the test result, which is conveniently operated; and 6) the large three-dimensional model tests for crossed tunnels are accomplished.

The present invention is a method and system for interactive rendering of large polygonal environments on commodity PC hardware. The system allows a user to walk through a large model at interactive frame rates on machines with limited memory. It works by first creating a hierarchical spatial decomposition of the model on disk using a fast and incremental out-of-core preprocessing algorithm. At running time, the system and method uses an approximate from-point visibility algorithm to dynamically determine which parts of the model to retrieve from disk. Multiple threads and a speculative prefetching algorithm are used to improve frame rates.

The invention relates to a novel model box testing device for an electroosmosis consolidation treated soft soil foundation. Electroosmosis consolidation can be simulated in a way of better approaching actual conditions through opening setting, dimension setting and strength setting of the testing device, device installation in a test, and various treatment means in test process control to consolidate the soft soil foundation. The model box can simulate electroosmosis consolidation, preloading, vacuum preloading, forced ramming and other foundation treating technologies to jointly treat the foundation. Besides functions of controlling electrification, being suitable for different electrode materials and controlling properties of a soil sample of the traditional electroosmosis consolidationtesting device, the model box has the advantage that different electrode arrangement forms and the application of an electrode conversion technology are fully considered, and has main characteristicsof economy, simpleness and multiple functions.

The invention relates to 3D (3-dimensional) multifunctional color food printing equipment. The equipment is shown as figure 1, and the working principle of the equipment is basically the same as that of a conventional 3D printer. Besides dynamic three-dimensional accurate positioning in X, Y and Z directions, a printing system comprises three spray nozzles, three dosing pumps provided with servo motors, three containers and a conveying pipe, wherein the spray nozzles adopt a circumferential uniform distribution structure and realize printing of three colors or objects through a control system. The equipment further has the advantages that a touch control screen and templates are arranged in a printer, a larger model base is established in the printer, so that various food prototypes can be printed rapidly and conveniently, or food with irregular animal forms is printed through combined templates, human dimensional imagination and creativity can be particularly stimulated, and green delicious food is made with own hands. The forming precision is higher, the speed is high, the technology is simple, the cost is low, and the equipment can be widely applied to the field of catering industry, the family field and the teaching field.

The Enviromentally Clean Rechargable Power Supply is a small portable totally electric rechargable source of both 12 volt DC 15 amp electricity and up to 700 watts and 6.5 amps of 110 volt AC electricity. Invention can be recharged with a conventional battery charger. Invention is enviromentally clean and can be operated indoors or outdoors. Uses include but but are not limited to emergency power to light homes and apartments, back up power to operate computers in homes and offices, power for lighting boats and RVs, power to operate trolling motors, charge automotive and marine batteries and provide a source of electricity for camp lighting, radios etc. Larger models can provide power to operate small scooters, go karts etc. with electric drive motors.

A method of dynamic model predictive control is presented in which both steady state optimization and dynamic moves calculation of the manipulated variables are determined together as part of one optimization solution (211). The method combines steady state optimization and dynamic move calculation such that the steady state optimal targets are determined consistent with the dynamic moves and the resulting dynamic response of the process so that the controlled variables do not violate their low / high limits in both steady state and dynamically. The method utilizes what is described as high limit dynamic violation variables and low limit dynamic violation variables corresponding to each of the controlled variables. The method offers a unique capability of mixing a part dynamic optimization with a part steady state optimization as it relates to application of a large model predictive control under varying situations, in particular at or near constraints violations that can be done dynamically in real time for improved and robust control performance without having to change the controller tuning.

The invention relates to a method for determining oil driving efficiency and sweep coefficients of all layers and positions of a heterogeneous reservoir stratum. The method includes the following steps of firstly, establishing an oil driving efficiency and sweep coefficient mathematical model; secondly, manufacturing a heterogeneous reservoir stratum large model, a core calibration parallel sample and an electrode core calibration parallel sample; thirdly, measuring parameters of the core calibration parallel sample and parameters of the electrode core calibration parallel sample; fourthly, solving the oil driving efficiency and the sweep coefficients of the different layers and the different positions of the heterogeneous reservoir stratum large model and the total oil driving efficiency and the total sweep coefficient of all the layers. When the oil driving efficiency calculated according to the method is low, the oil driving efficiency can be improved by replacing chemical agents or increasing the concentration of the chemical agents. When the sweep coefficients are low, the sweep coefficients can be increased by increasing the generated pressure difference, or the sweep volume can be increased by drilling an infill well. The method provides visual and accurate foundation data for oilmen, and provides reference data for regulating the exploitation scheme of an oil field.

The invention provides a model compression method based on pruning sequence active learning. an end-to-end pruning framework based on sequential active learning is provided; The method has the advantages that the importance of all layers of the network can be actively learned, the pruning priority is generated, a reasonable pruning decision is made, the problem that an existing simple sequential pruning method is unreasonable is solved, pruning is preferentially carried out on the network layer with the minimum influence, pruning is carried out step by step from simplification to difficulty, and the model precision loss in the pruning process is minimized; And meanwhile, the final loss of the model is taken as a guide, and the importance of the convolution kernel is evaluated in a multi-angle, efficient, flexible and rapid manner, so that the compression correctness and effectiveness of the whole-process model are ensured, and technical support is provided for subsequent transplantation of a large model to portable equipment. Experimental results show that the model compression method based on pruning sequence active learning provided by the invention is leading under the conditions of multiple data sets and multiple model structures, can greatly compress the model volume under the condition of ensuring the model precision, and has a very strong practical application prospect.

The invention discloses a neural network model training method. The method comprises a step of obtaining a first image feature and a second image feature, wherein the first image feature is an image feature outputted by a picture A through a trained model, and the second image feature is an image feature outputted by the image A through a model to be trained, a step of obtaining a classification probability which is a classification probability outputted by the second image feature through a classification layer of the model to be trained, and a step of updating the parameters of the model tobe trained until the second image feature and the first image feature are fitted according to the first image feature, the second image feature and the classification probability. The invention also discloses an electronic device and a storage medium. The invention provides the neural network model training method, the electronic device and the storage medium, the parameters of the model to be trained are updated according to the first image feature, the second image feature and the classification probability, a small model with the same precision of a trained large model is obtained through training, and the calculation speed and accuracy are ensured.

The invention discloses a design method of a scale model of an extra-high voltage converter transformer. According to the principle of similitude, the structural characteristics of the extra-high voltage converter transformer and the application occasions of the scale model are taken into account, typical physical quantities needing controlling are set according to the research goal, corresponding resemblance constants are set, the resemblance constants are related through a control equation, and finally a total scale formula of the converter transformer is obtained. The scale model is designed according to the total scale formula of the converter transformer, the consistency of the scale model and an original model is verified by simulating and measuring the electromagnetic field distribution of the scale model, and finally the electromagnetic field distribution characteristics of an original large model are predicted through the scale model. By means of the design method of the scale model, the designed scale model can effectively reflect the electric characteristics of the original model, can be used for carrying out experiments and research, lowers experiment cost, shortens the experiment period, and has good economical benefits for researching large equipment like the extra-high voltage converter transformer.

A system and method for computing unemployment insurance premiums that directly supplements government unemployment benefits. To facilitate risk calculations, the work force is classified by industry, occupation, income level, geographic location and many other factors. The insurance premiums are calculated as part of a larger model of private unemployment insurance for high-income workers. If the worker's income exceeds a certain threshold, the insurance would supplement the replacements the government already provides. Additionally this process includes the methods in which the claimant's eligibility is assessed, and how the program is administered (including the claims process). Claim verification relies on piggybacking on the state's determination of a claimant's eligibility for unemployment compensation payments, and the premium computation process takes into account a state's history with respect to granting and denying such claims. The process can also be extrapolated to a non pure supplemental policy.

A scale-Space feature extraction technique is based on recursive decomposition of polyhedral surfaces into surface patches. The experimental results show that this technique can be used to perform matching based on local model structure. Scale-space techniques can be parameterized to generate decompositions that correspond to manufacturing, assembly or surface features relevant to mechanical design. One application of these techniques is to support matching and content-based retrieval of solid models. Scale-space technique can extract features that are invariant with respect to the global structure of the model as well as small perturbations that 3D laser scanning may introduce. A new distance function defined on triangles instead of points is introduced. This technique offers a new way to control the feature decomposition process, which results in extraction of features that are more meaningful from an engineering viewpoint. The technique is computationally practical for use in indexing large models.

The invention relates to a rapid parking space detection method based on deep learning, belongs to the technical field of driving, and is used for solving the problems of poor parking space detectionenvironment adaptability and large model calculation amount, and the method comprises the steps: an offline step: offline collecting image data including parking spaces, and establishing a training and verification data set; training, evaluating and optimizing a neural network model, wherein the neural network model is used for performing semantic segmentation on parking space sidelines in the image data; an online steps: collecting image data containing parking spaces online, carrying out semantic segmentation on parking space side boundary lines through a trained neural network model to obtain parking space side boundary masks, carrying out fitting, clustering and combination on the obtained side boundary masks, and obtaining a geometrical shape composed of the side boundaries; and according to a set shape discrimination condition, screening the geometrical shape to determine the parking space. The method is high in environmental adaptability; the size of the adopted model is very small, the calculation amount is low, and the requirement for calculation resources is low; and the system is low in cost and has large-scale application potential.

The invention discloses a lightweight vehicle detection model construction method, system and device. The method comprises the following steps of: adding a lightweight channel attention module at thetail end of a bidirectional feature pyramid module of a target detection algorithm EfficientDet model with an EfficientNet backbone network, and learning a relationship of feature maps among differentchannels through the channel attention module by one-dimensional convolution with a certain step length; carrying out positive and negative sample selection by adopting an adaptive training sample selection algorithm, and training the EfficientDet model; training a high-precision model by using a large model in an original data set by adopting a distillation learning method; and then compressingthe trained large model to construct a smaller model, and guiding the small model to carry out distillation learning of an intermediate feature layer, distillation learning of a target classificationresult and result distillation learning of target frame regression prediction by using the large model as a teacher network to obtain a final lightweight vehicle detection model.

The invention discloses a large model test device simulating deep vertical shaft freezing. The large model test device comprises a model test barrel, a slurry loading system, a hydraulic system, an excavation simulation system, a refrigerating system, a monitoring system and a data collection and control system, wherein the excavation simulation system is located in a soil sample, the soil sample is located in the model test barrel, the slurry loading system, the hydraulic system and the data collection and control system are all located outside the model test barrel, one part of the monitoring end of the monitoring system is buried into the soil sample, the other part of the monitoring end of the monitoring system is located in the excavation simulation system, the output end of the monitoring system is connected with the input end of the data collection and control system, the hydraulic system provides hydraulic driving force for the excavation simulation system, a slurry output opening of the slurry loading system is communicated with cavity fluid between the inner wall of the model test barrel and the soil sample, and the refrigerating system provides circulation low temperature liquid for soil sample freezing. By means of the large model test device simulating deep vertical shaft freezing, when a test is carried out, formation of a freezing wall in the deep vertical shaft freezing process and displacement change conditions of well side during shaft excavation can be really reflected.

The invention discloses a direct making method of a large model fracture core maintaining original oil-water saturation. The method includes: (1) determining the volume V, porosity phi, permeability K, oil saturation So, water saturation Sw and the like of a fracture core; (2) preparing simulated oil according to a stratum oil-water viscosity ratio, and determining the used oil mass mo=Vo*rhoo; (3) under the circumstance of no consideration of oil saturation, acquiring mass of the used water, cement and quartz sand, i.e. a, b and c respectively; (4) while establishing oil saturation, acquiring the mass of water mw=a-Vo*rhow; (5) mixing oil, water and an emulsifier evenly to prepare a oil-in-water emulsion; (6) adding cement and quartz sand into the emulsion and stirring the substances evenly to obtain a cement slurry; (7) when a cement sample is in a semi-solidified state, cutting the cement sample according to an experiment required fracture direction with steel wire; and (8) solidifying the cement sample to the end. The method provided by the invention is low in cost and convenient to operate, can control the distribution of fracture in the core, and realizes simultaneous establishment of the model and saturation.

A scale-Space feature extraction technique is based on recursive decomposition of polyhedral surfaces into surface patches. The experimental results show that this technique can be used to perform matching based on local model structure. Scale-space techniques can be parameterized to generate decompositions that correspond to manufacturing, assembly or surface features relevant to mechanical design. One application of these techniques is to support matching and content-based retrieval of solid models. Scale-space technique can extract features that are invariant with respect to the global structure of the model as well as small perturbations that 3D laser scanning may introduce. A new distance function defined on triangles instead of points is introduced. This technique offers a new way to control the feature decomposition process, which results in extraction of features that are more meaningful from an engineering viewpoint. The technique is computationally practical for use in indexing large models.

The invention relates to a spatial structure integral model test cable force self-balanced loading control method which belongs to the technical field of structural engineering. The method is based on an integral self-balanced loading system formed by a test model, a bracing cable, a base plate girder, a bracing cable displacement loading device and a strain gauge force plate and is realized in a loading control system formed by a data input module, a loading simulation analysis module, a displacement loading control module and an internal force fine-adjustment control module according to specific steps. The method is high in adaptability and simple in operation, is suitable for the large-sized model test loading and can accurately realize the predetermined step loading target through the dual control technologies of displacement loading and internal force fine-adjustment.

The invention belongs to the technical field of construction projects and relates to a sand rain process automatic sander. The sand rain process automatic sander comprises a PLC, servo motors, track guide rails, a sand charging barrel, a hose, nozzles, a sanding cylinder, filter plates and a screen, wherein the PLC is connected with three servo motors, and the servo motors are operated to control the track guide rails in the respective direction to move; the sand charging barrel is suspended at a high part, the sand can flow downwards, and the bottom of the sand charging barrel is connected with the top of the nozzles by virtue of the hose; the bottom of the nozzles are screwed with the sanding cylinder, the filter plates of different sizes are arranged among the nozzles and are used for controlling the flow, and the nozzles are fixedly arranged on the track guide rails; and the screen is arranged at the bottom of the sanding cylinder. According to the sand rain process automatic sander disclosed by the invention, the instrument is easy to operate, the three-dimensional motion is automatically controlled, the manual operation errors are reduced, the prepared sample is high in uniformity and repeatability, the relative compactness range is wide, and most of experimental requirements can be met. Moreover, the sand rain process automatic sander is conveniently applied to large model experiments, particularly a centrifugal model experiment.

The invention relates to a method for constructing a microporous structure of a bionic support, which comprises the following steps: performing overall scanning of a natural bone by using Micro-CT technology to extract data of a spongy bone and reconstructing the spongy bone to obtain a model of the microporous structure of the spongy bone; extracting a model of a unit body from the model of the spongy bone; obtaining a model of a three-dimensional microporous structure by mirroring; and finally, performing Boolean intersection operation with the model of the three-dimensional microporous structure and a model of a damaged bone to obtain a model of the microporous structure of the bionic support matched with a damaged part. According to the design method, the amount of processed data is reduced in a reconstruction process. The design method of obtaining a large model by mirroring the unit body solves the problem of local fault in the micro structure. According to design requirements, the model of the microporous structure, which is obtained by mirroring, may have an infinite size. The microporous structure of the obtained model of the bionic support is similar to the structure of a real bone, which is more favorable for cell adhesion, cell climbing and osteogenic replacement. The bone support designed by the method has the same profile as the real bone and thus can be implanted more conveniently.

An apparatus that accurately measures rope, wire, tubing and other materials is herein disclosed. A variety of different-sized models would possess different sized openings. One model with a one-inch opening would be used to measure small wire and cable. Another model with approximately a 3-inch opening would be used for larger electrical cables, rope, tubing and the like. An even large model would be available for measuring ductwork, hoses, and other large diameter material. As the material is pulled through the opening, an optical sensor, similar to that used on an optical computer mouse, accurately measures the material to the fraction of an inch. A mechanical sensor, utilizing a guide wheel is also envisioned. The length is displayed on a digital readout on the perimeter of the invention that is provided with a power button, reset button and the like. It is envisioned that various attachments such as a belt loop, a spool attachment and other items to increase the versatility of the invention would also be provided.