49results about How to "Speed ​​up the calculation process" patented technology

The invention discloses a lithium battery instantaneous heat production power measurement method, which comprises the following steps of: spraying blackbody paint on the surface of a lithium battery sample, and placing the lithium battery sample on a heat insulation bracket of a constant-temperature box; the lithium battery sample is connected with a charging and discharging device, and surface temperature change in the heat production process of the lithium battery sample and surface temperature drop data after charging and discharging are measured through a thermal infrared imager; calculating an equivalent surface heat exchange coefficient of the lithium battery sample according to the surface temperature drop data of the lithium battery sample; and establishing a thermal simulation model according to the thermophysical parameters of the lithium battery, and dynamically optimizing the heating power data according to the goodness of fit between the simulation temperature change and the actually measured temperature change to obtain the instantaneous heat production power data of the lithium battery sample. According to the method, the infrared non-contact temperature measurement technology is used for recording the surface temperature rise data of the lithium battery in the charging and discharging states, the thermal simulation model is established according to the thermophysical parameters, the instantaneous heat production power of the lithium battery is calculated by using the dynamic parameter optimization algorithm, the calculation process is rapid, and the test precision is high.

The invention discloses a carrier roller fault diagnosis method based on multi-modal information fusion, and particularly relates to the field of carrier roller fault diagnosis. The method comprises the following steps: collecting various modal signal data of a carrier roller, and preprocessing original one-dimensional signal data; simplifying the one-dimensional time sequence data of each signal by using classification aggregation approximation; converting each signal into a two-dimensional feature image by using a Markov field and synthesizing the two-dimensional feature image into a three-channel image; a carrier roller fault diagnosis model based on a Swin Transform network architecture is established; the Swin Transform network architecture is trained and verified, if the network achieves the optimal effect, the next step is carried out, otherwise, the previous step is returned, and initial parameters are modified to continue training; and inputting the test set into the optimal Swin Transform network architecture to carry out carrier roller fault diagnosis, and finally obtaining a carrier roller fault classification result. According to the technical scheme, the problems that existing carrier roller fault diagnosis is based on single-mode information, and information is insufficient and inaccurate are solved, and a foundation is laid for follow-up control and maintenance implementation.

The invention discloses a hotel cleaning personnel cleaning compliance detection method and system based on RFID and deep learning, and electronic equipment, an action detection model based on a machine learning model is supplemented and added on the basis of the technology of a detection mechanism based on RFID tag type matching, and the detection result of the model is combined to detect the cleaning compliance of hotel cleaning personnel. And the operation normalization of the cleaning personnel is jointly judged, so that the detection and judgment are more accurate. In addition, a traditional behavior recognition method based on RFID data is complex in model structure, too many in neural network layer number and large in model parameter quantity, the calculation and storage overhead is increased, the model does not have the rapid deployment and response capacity, and in the application scene of the method, efficient operation on equipment with limited calculation resources such as a mobile phone cannot be achieved. The cleaning action recognition module based on the deep learning algorithm recognizes whether cleaning personnel do cleaning actions such as wiping while performing RFID tag type matching detection, so that the detection accuracy and rationality of the system are improved.

The invention provides a cache system, a cache processing method, electronic equipment and a storage medium, the cache system comprises a plurality of processing cores and a configuration module, each processing core comprises a target storage area, the configuration module is used for determining multi-level cache configuration information of a first processing core, and according to the multi-level cache configuration information, the first processing core is configured to process a cache; first cache regions are divided from target storage regions of a plurality of processing cores, the first cache regions correspondingly divided by the plurality of processing cores form a multi-level cache system of the first processing cores, and the processing cores where the first cache regions of different levels are located are related to position information of the first processing cores. The first processing core is any one of the plurality of processing cores; and the first processing core is used for searching and acquiring data indicated in a data access request from the multi-level cache system of the first processing core when the data access request is received. In this way, the multi-level cache system of different processing cores can be flexibly configured, and the performance is improved.

The invention discloses a message processing method and device, equipment and a medium. The method is applied to message middleware and comprises the following steps: acquiring a service request of an application system participant; calling a secret key management interface provided by a hardware encryption and decryption module based on an application programming API interface, and generating a secret key and a secret key handle of an application system participant; broadcasting a public key and a public key handle in the secret key handle through a public key handle broadcasting interface; and calling an encryption and decryption interface provided by a hardware encryption and decryption module through an API interface according to the secret key, and processing the service request to obtain a target message. A standard calling interface is provided for an external application system in a uniform interface mode through native national cryptographic algorithm supporting capacity of message middleware, so that encryption and decryption processing of service requests of participants of the application system is achieved, a target message is obtained, the process of supporting application layer data encryption and decryption is achieved based on local hardware equipment, and the method is simple and convenient to implement. And the efficiency of algorithm calculation and data processing is improved.

The invention discloses a state space model accelerator, a Mama model calculation method and a Mama model calculation system. The state space model accelerator comprises a reconfigurable calculation unit used for linear calculation and element-by-element calculation; the reconfigurable computing unit comprises a reduction tree and a processing unit array; the processing unit array comprises a plurality of reconfigurable processing units; during linear calculation, each reconfigurable processing unit calculates input data; the reduction tree is in a reduction mode, and reduction tree slices in the reduction tree carry out reduction calculation on calculation results of the multiple corresponding reconfigurable processing units; when element-by-element calculation is carried out, each reconfigurable processing unit carries out element-by-element calculation on input data and outputs calculation results in parallel; the reduction tree is in a non-reduction mode. According to the Mama accelerator, a reduction-replaceable PE array architecture is adopted, and element-by-element operation and linear operation can be efficiently supported at the same time.

The invention provides a navigation method and device for a metro vehicle inspection robot in a multi-track scene and a readable storage medium, and aims to solve many problems of navigation of the inspection robot in the multi-track scene in the prior art. According to the method, firstly, a lifting platform template is created on a navigation map, parameters are configured, then a TOPO graph is created in combination with information such as the current position of a robot and the position of a target point, then an improved Dijkstra algorithm is used for generating a global navigation path with the shortest time consumption, then the global path is segmented to obtain sub-target points and related attributes, and finally, the target points are obtained. And finally, when the robot runs, the state of passing through the lifting platform is monitored and predicted in real time, and the robot is controlled to move. Through the steps, the lifting platform can be rapidly deployed, the task configuration process can be simplified, the optimal path can be autonomously planned online, the state of the lifting platform can be accurately controlled, the working efficiency and the operation flexibility of the robot are improved, and the inspection quality and the safe operation of the metro vehicle are effectively guaranteed.

The invention provides an intelligent community management robot system based on an artificial intelligence technology, and belongs to the technical field of intelligent community robots, and the system comprises a community data automatic sensing module which is used for automatically sensing and collecting community data, and obtaining multi-modal data; the multi-modal data integrated management module is used for performing centralized management and processing on the acquired multi-modal data to obtain single information flow data; the sustainable artificial intelligence calculation module is used for carrying out intelligent calculation and analysis on the single information flow data, identifying community service requirements, optimizing a service process and outputting an instruction of the community service requirements; the community governance service function module is used for providing intelligent services; and the community robot task execution module is used for providing a human-computer interaction interface and executing intelligent service according to the instruction required by the community service. The intelligent community management robot system solves the problems that the service function of an existing intelligent community management robot system is difficult to achieve, the resident participation degree is low and the intelligent degree is low.

The invention provides a low-code system for data analysis and data cleaning based on Internet of Vehicles data. The low-code system comprises a multi-protocol adapter, a user-defined analysis rule writing interface, a dragging type rule configuration interface, a data processing module, a forwarding mechanism unit, a monitoring and log tracking unit and a message labeling and fingerprint generation unit. Wherein the multi-protocol adapter is used for supporting log data acquisition of various network protocols; the user-defined analysis rule writing interface allows a user to write specific analysis logic according to a data format and a service requirement; meanwhile, for compatibility of various data codes and formats, rule templates and examples are provided to guide a user to compile analysis rules, and the user is helped to quickly position and correct problems in the analysis rules by integrating corresponding error detection and debugging tools; and a user carries out graphical operation combination input, output and analysis on rule nodes through the drag-and-drop rule configuration interface.

The invention discloses a graph neural network accelerated training method for a GPU, a storage medium and an electronic device, and the method comprises the steps: obtaining an initial graph, segmenting the initial graph into N sub-graphs, and enabling the initial graph to be a static graph corresponding to a moment in a dynamic graph sample; calculating loads of the sub-graphs are estimated, the N sub-graphs are distributed to the M GPUs based on the calculating loads of the sub-graphs, training of the graph neural network is carried out, and at least two sub-graphs are distributed to each GPU; incremental data is obtained, the subgraph is updated based on the incremental data, and the incremental data is used for representing the change of a static graph in the dynamic graph sample scanned at the current moment relative to the static graph currently processed by the GPU; and distributing the N sub-graphs containing the incremental data to M GPUs and carrying out graph neural network training. According to the technology, the calculation and resource utilization rate is improved, the training time consumption is reduced, and the training efficiency of the graph neural network during training is improved.

The invention discloses an instruction execution method of a digital signal processor and the digital signal processor. The instruction execution method comprises the following steps: acquiring a target instruction and a peripheral clock enable signal of a corresponding peripheral unit; determining an instruction type of the target instruction through an instruction structure of the target instruction and a preset instruction setting condition; determining a signal state of a peripheral clock enable signal; executing a priority processing strategy when the instruction type is a priority instruction and the signal state is an enabling state, the priority processing strategy comprising: sending a partner core starting signal to a priority processing unit, and sending a main core starting signal to a main processing unit to start the priority processing unit to load and process a target instruction, and the main processing unit sequentially processes other instructions in the instruction set through the multi-stage assembly line at the same time. According to the instruction execution method provided by the embodiment of the invention, the instruction priority can be improved, and the operation and processing of special instructions can be accelerated, so that the response speed is improved, and the processing capacity of the digital signal processor is improved.

The invention discloses a left ventricle segmentation method based on an optimized UNet network and an improved CBAM attention mechanism, and belongs to the technical field of image processing. The method specifically comprises the following steps: constructing a lightweight attention mechanism SCBAM, combining the lightweight attention mechanism SCBAM with a convolutional VGG network, cutting a UNet network to L3, and carrying out lightweight design to reduce calculation and storage overhead. By combining an SCBAM attention mechanism, a VGG network and jump connection, an iterative up-down sampling block Nestedunet is constructed, the accuracy of cardiac function evaluation is improved, and a lighter and more efficient solution is provided for automatic auxiliary diagnosis in practical application.

The present disclosure provides a sampling circuit, a control method and a hardware accelerator applied to a lattice cryptography algorithm, and relates to the technical field of post-quantum cryptography. The sampling circuit comprises a padding module configured to pad a seed according to a hash padding rule; a hash operation module configured to perform a hash operation on the padded seed to generate a pseudo-random sequence; a cache module configured to store the pseudo-random sequence generated by the hash operation module; and a sampling module configured to sample the pseudo-random sequence stored in the cache module to obtain a sampling result. The present disclosure provides the cache module, and in the sampling process, the cache module provides data for sampling to the sampling module without relying on the hash operation module to provide data. Therefore, the sampling module and the hash operation module in the present disclosure are independent of each other and can work simultaneously, thereby improving the operation efficiency of the sampling circuit and effectively accelerating the calculation process of the lattice cryptography algorithm.

The present invention discloses an underwater 3D visualization intelligent survey system, which relates to the technical field of industrial solid waste treatment material management, and includes an ultrasonic sensor and a sonar sensor detection module: a water pressure sensor, an ultrasonic sensor and a sonar sensor are integrated on an underwater survey device to obtain sensor data of underwater detection in real time, and then the sensor data collected in real time is transmitted to a cloud data interaction processing module through a communication protocol; the present invention realizes an intelligent function integrating ultrasonic sensors, sonar sensors, automatic orientation technology, automatic water surface keeping technology, GPS satellite positioning technology, information denoising technology, and on-site and cloud data instant interaction technology by setting ultrasonic sensors and sonar sensor detection modules, a sensor automatic orientation module, an automatic water surface keeping module, a cloud data interaction processing module, a GPS satellite positioning module, and a data fusion analysis and calculation module.

The invention relates to the technical field of structural engineering, and particularly discloses a multi-scale integrated design optimization method and system, and the method comprises the steps: defining a design target, and collecting first performance data; performing multi-scale hierarchical division on the first performance data; carrying out first scale hierarchy attribute modeling on the multi-scale hierarchy, and carrying out second scale hierarchy modeling by adopting an analysis method; topological structure form optimization is carried out based on a design target, and parallel processing is carried out through multi-target optimization. According to the method provided by the invention, the performance of the whole system is improved through cross-scale optimization, and design deviation caused by inconsistency of optimization targets at all levels is avoided. The lightweight level of the design is remarkably improved, and meanwhile the durability and stability of a mechanical system are guaranteed. The method greatly accelerates the simulation and calculation process, effectively shortens the design period, supports the rapid evaluation and comparison of a plurality of optimization schemes, and improves the design efficiency and optimization precision.

The present invention discloses an instruction execution method of a digital signal processor and a digital signal processor, the instruction execution method includes: obtaining a target instruction and a peripheral clock enable signal of a corresponding peripheral unit; determining the instruction type of the target instruction through the instruction structure of the target instruction and a preset instruction setting condition; determining the signal state of the peripheral clock enable signal; executing a priority processing strategy when the instruction type is a priority instruction and the signal state is an enable state, the priority processing strategy includes: sending a companion core start signal to a priority processing unit, and sending a main core start signal to a main processing unit to start the priority processing unit to load and process the target instruction, and the main processing unit simultaneously processes the remaining instructions in the instruction set through a multi-stage pipeline sequence. The instruction execution method of the embodiment of the present invention can improve the instruction priority, accelerate the operation and processing of proprietary instructions, thereby improving the response speed and the processing capacity of the digital signal processor.

The utility model discloses an embedded control system based on an FPGA, the system comprises an FPGA processor, a power supply module used for supplying power to the FPGA processor, a Flash module, a USB module, an Ethernet module, an expansion module, a JTAG module and a man-machine interaction module, the expansion module is used for being connected with external expansion equipment, the JTAG module is used for internal testing of the FPGA processor, and the man-machine interaction module is used for being connected with the FPGA processor. The man-machine interaction module is connected with the FPGA processor through a corresponding protocol data line, the USB module is used for data transmission and power supply, the operation and control capability of the system is accelerated, the development cycle of a user is shortened, the expansibility is improved, and the power consumption is reduced.

This invention discloses a network-based data weaving integrated machine system architecture, belonging to the field of data management and integration technology. It includes a hierarchically deployed basic node layer, composed of multiple physical or virtual device nodes, each node comprising a computing unit, storage unit, caching unit, network communication unit, and query execution engine; a node control layer, configured to uniformly manage all nodes, including a node registration module, a node discovery module, a task scheduling module, and a routing control module; a data weaving capability layer, configured to provide data management functions, including an active metadata management module, a federated query module, a semantic layer module, and an access control module; a network collaboration layer, configured to realize multi-node collaboration, including a cross-node query collaboration module, a metadata synchronization module, and an access control mapping module; and an intelligent collaboration layer, configured to provide intelligent computing interfaces. This invention employs the above system architecture to construct a scalable, networkable, and collaborative network-based data infrastructure.

The present invention discloses a method for calculating the grounding resistance of a horizontally laid rectangular grounding body, belongs to the technical field of calculating high soil resistivity, and solves the problem that the resistivity of the grounding resistance cannot be calculated in high soil areas. The grounding device of the present invention comprises a rectangular panel, and the rectangular panel is placed and laid at a depth of 0.7m from the horizontal ground; the electric potential between the rectangular panel and any point parallel to the ground can be accurately calculated through a calculation formula; the present invention calculates the grounding resistance by transitioning from the rectangular panel to the ground soil current density, and can specifically calculate the grounding resistance value between the rectangular panel and the ground in the high soil area, thereby realizing the accurate calculation of the grounding resistance of the rectangular panel. The present invention has the advantage of solving the calculation method of the resistivity in the high soil area.

The invention provides a bonding potential energy calculation method and device, a terminal, a storage medium and a product, and the method comprises the steps: constructing a simulation calculation model, and dividing the simulation calculation model into a plurality of simulation calculation sub-models with independent calculation capability; obtaining atomic information, and generating an atomic list; determining a calculation atom according to a preset allocation rule, and allocating a bonding potential energy calculation task to the calculation atom; determining the affiliation of a simulation calculation sub-model of the calculation atom, and storing the bonding potential energy calculation task in the simulation calculation sub-model; and performing parallel calculation according to a preset calculation formula. According to the bonding potential energy calculation method and device, the terminal, the storage medium and the product provided by the invention, the calculation task of the bonding potential energy can be reasonably allocated to each simulation calculation sub-model, and the simulation calculation sub-models do not have a data interdependence relationship, so that the acceleration of the calculation process can be realized by adopting a parallel method; therefore, the engineering development efficiency is greatly improved.

The invention discloses an electromagnetic simulation parallel computing method, which relates to the field of electromagnetic simulation computing, and is characterized in that a plurality of processor IPs are in the same network segment, and each processor is internally provided with a plurality of cores; all the processors exchange and share data; an Ansys Element Desktop module and a same computer management account are deployed in each processor, wherein the Ansys Element Desktop module and the same computer management account are consistent in account. And the electromagnetic simulation calculation tasks of the Ansys Remote Solve Manager module are distributed to each processor through the Ansys Remote Solve Manager module, and the electromagnetic simulation calculation tasks of the Ansys Telenics Desktop module are distributed to each processor through the Ansys Telenics Solve Manager module. The plurality of processors can process different parts of an electromagnetic simulation calculation task in parallel, so that the overall calculation process of the electromagnetic simulation calculation task is improved, the calculation time is remarkably shortened, the research progress in the research fields of microwaves, communication, national defense, military affairs, medical treatment and science is improved, and the development cost of enterprises is reduced.

The invention belongs to the technical field of nuclear power alarm monitoring, and particularly relates to a nuclear power alarm monitoring method and device based on event subscription. Basic data related to alarm monitoring is read from a DCS system through a one-way gatekeeper, a local database is established, an alarm measuring point and an alarm event are configured in a local system, and configuration files of the alarm measuring point and the alarm event are downloaded to an alarm service; the alarm events are divided into multi-measuring-point working condition events and single-measuring-point events; duplicate removal is carried out on the single-measuring-point events requested by subscription of different clients, and unified subscription is carried out on a single-measuring-point event set after duplicate removal; and the alarm service of the multi-measuring-point working condition event performs alarm calculation according to a preset alarm algorithm model, triggers an alarm event when an alarm calculation result meets an alarm condition, and actively pushes the corresponding alarm event to a corresponding client according to an alarm subscription table. According to the invention, the technical problem that the nuclear power alarm monitoring function is limited in the prior art is solved, and the health management of the DCS is enhanced.

The invention discloses a multi-modal ship image classification method based on Mangbar Fourier calculation, and the method comprises the steps: inputting a natural light image and an infrared image of the same ship into a multi-modal classification model at the same time, and obtaining a classification result. The multi-modal classification model comprises a sequence generation module and a classification head, and further comprises N convolution Mangbar modules and N-1 Fourier attention modules. And the convolution Mangbar module is used for performing convolution operation and double-sequence Mangbar calculation on the input features to obtain output basic features. And the Fourier attention module is used for carrying out global interaction and fusion on the two groups of input features through Fourier transform and an attention mechanism to obtain output fusion features. According to the method, the Mangbar model and Fourier calculation are introduced, efficient extraction of ship image features is realized, and local detail information and global semantic information of the image are considered, so that ship features are described more comprehensively, noise and background interference are effectively suppressed, and the accuracy of classification and recognition is improved.

The application provides a privacy protection federated learning method based on a blockchain anti-poisoning attack, and belongs to the technical field of federated learning, privacy protection and blockchain technology. The technical problems of poisoning attack and malicious aggregation in the privacy protection federated learning are solved. The technical scheme comprises the following steps: S10, system global initialization; S20, local model training; S30, malicious participant identification; MO and SM cooperate to determine whether the participant has performed the normalization operation; S40, gradient ciphertext aggregation; S50, participant and MO update model. The application has the beneficial effects that the application protects the privacy of the participant model gradient by using the fully homomorphic encryption algorithm CKKS, identifies the malicious gradient in the gradient ciphertext by using the cosine similarity method, and calculates and verifies the aggregation result by using the consensus mechanism of the blockchain.

The invention discloses a magnetic anisotropy energy research method, device and equipment, a medium and a program product, and relates to the technical field of magnetic materials, and the method comprises the steps: obtaining a to-be-researched target magnetic material, and configuring the initial magnetic state of the target magnetic material; performing density functional theory calculation of primary spin-orbit coupling on the target magnetic material in the initial magnetic state to obtain an electronic structure physical quantity of the target magnetic material; and according to the physical quantity of the electronic structure in combination with a Green function method, performing one-time calculation to obtain a magnetic anisotropy energy result pointed by each magnetic moment of the target magnetic material, and performing magnetic anisotropy energy analysis based on the magnetic anisotropy energy result pointed by each magnetic moment of the target magnetic material. According to the method, the complex calculation process of the density functional theory of adding spin-orbit coupling for multiple times can be avoided, the problem that the calculation process is difficult to converge is avoided, and the efficiency of calculating and analyzing the magnetic anisotropy energy is improved.

The application provides an energy efficiency service deployment and delivery method and system for an edge computing network, and relates to the technical field of edge computing.The method comprises the following steps: extracting multi-dimensional service features, encoding and fusing the extracted multi-dimensional service features to generate a global context feature representation; combining system load dynamic prediction results to adaptively adjust an upper-layer decision time scale based on the global context feature representation; executing service deployment decisions and base station sleep / activation switching decisions under the adjusted upper-layer time scale; executing service delivery decisions and resource allocation decisions based on real-time service requests under a fine-grained lower-layer time scale; and coordinating the decision-making process through a double-time-scale hierarchical learning framework.The method can effectively reduce long-term network overhead, effectively respond to dynamic service requests, improve the processing capacity for time coupling relationships between deployment and delivery cycles, and reduce system costs.

The invention belongs to the technical field of traffic plan recommendation, and discloses a traffic plan recommendation method based on artificial intelligence, which adopts a double-task soft sharing neural network model to perform traffic plan recommendation and mobile phone model recommendation, gives consideration to directional traffic features and text features, and improves the traffic plan recommendation efficiency. According to the method, a full connection layer shared by two weights is used for carrying out re-representation and feature extraction on traffic features, a cosine similarity merging layer is used for analyzing the matching degree between directional traffic and user requirements, and in addition, a soft sharing mechanism is used for repeatedly utilizing the features used for traffic package recommendation tasks to carry out mobile phone model recommendation. By means of the technical means, the problems that an existing traffic plan recommendation model does not conduct targeted feature analysis on traffic plan features and data values are not fully developed are solved, and soft-sharing, high-precision and double-task traffic plan recommendation is achieved.

The present invention discloses a method and system for calculating fine land use types, and relates to the field of remote sensing computing technology. The method comprises: obtaining an original high-resolution image of a target area; extracting the original high-resolution image to obtain fine land-level patches and land use types; inputting the fine land-level patches into a classification recognition model, determining the land use type corresponding to each of the fine land-level patches, and adjusting the land use type according to the land use type confidence; the classification recognition model is constructed using a hierarchical perceptron and a multi-granularity decision maker. The present invention can enhance the discrimination and calculation of land use types of land plots.

The invention discloses a federated learning method based on a model deviation and variance distribution model, and relates to the field of machine learning. According to the method, the variance of the model uploaded by the client in each round is calculated on the server, the offset parameter of the model of each client relative to the global model is learned, the personalized information of each client is extracted, and a personalized initial model is provided for local learning of the client in each round. Through personalized model learning, the learned model has better performance on local data, the problem that existing personalized federated learning depends on extra overhead on the client is solved, extra data does not need to be transmitted between the client and the server, and privacy of client data is better protected. In addition, by means of hierarchical calculation of the model, calculation and storage expenses in the learning process are reduced, meanwhile, the performance of the personalized model is further improved, and the prediction accuracy of the data is improved.

The application provides a switch frequency determination method and device for an LCC resonant converter, the method comprising: obtaining an input voltage and an output voltage of the LCC resonant converter; determining a starting mode in which the LCC resonant converter is currently located by using the output voltage; determining a calculation program corresponding to the starting mode by using the starting mode in which the LCC resonant converter is currently located, wherein the calculation program is an operation logic program designed for the starting mode by using a field programmable logic gate array; inputting the input voltage and the output voltage into the field programmable logic gate array, and calculating the switch frequency of the LCC resonant converter by using the calculation program, so that the switch frequency determination method improves the adjustment rate of the switch frequency of the LCC resonant converter, the rising speed of the output voltage, and further improves the dynamic response speed of an existing power supply.