45results about How to "Improve data consistency" patented technology

The invention belongs to the technical field of artificial intelligence, and particularly relates to a GUI-Agent trajectory data generation method and system based on multi-agent collaborative reasoning. The method comprises the following steps: S1, capturing and analyzing a page state and a structure, and constructing an initial graph; s2, analyzing semantics of interactive elements based on VLM and generating an initial action instruction set; s3, executing the instruction, recording state change, automatically deducing a target intention, forming an initial track, and storing the initial track in a track pool; s4, for the current track, predicting and executing the next action according to the target and the page state of the current track, and updating the track; s5, unexplored elements are recognized, new actions and intentions are generated in combination with the VLM, and a new track is created and added into a track pool; s6, evaluating whether the action result achieves a target or not, and deciding to terminate, adjust or continue to explore according to the target; and S7, circularly executing the steps S4 to S6, performing parallel iteration processing on the trajectory pool until a termination condition is met, and outputting a trajectory data set.

The invention discloses a corpus standardized template construction and file automatic generation method and system, and the system comprises a template construction module, a file generation module, a verification module, a user interface module, a storage and management module, an extension and optimization module, and a quality evaluation module. The standardized management of the corpus is realized, the problem that the structure and format of the corpus lacks unified standards in the traditional method is solved, and the compatibility and sharing of the corpus are improved. Meanwhile, data cleaning, format conversion and multi-modal data integration technologies are combined, corpus files meeting template requirements are automatically generated, the data processing flow is simplified, and the risk of data inconsistency or loss is reduced.

The present disclosure relates to a method and device for identifying tectonic coal based on borehole gamma data. The method comprises: acquiring gamma curve data and mechanical drilling speed data collected in real time during the process of logging while drilling; preprocessing the gamma curve data and the mechanical drilling speed data to obtain depth sequence data; performing segmentation processing on the depth sequence data to obtain a plurality of data sample units; for each data sample unit, extracting a plurality of sample features for representing the response mode of tectonic coal from the data sample unit, fusing the plurality of sample features to obtain a multi-parameter fusion feature vector; and inputting the multi-parameter fusion feature vector into a trained machine learning classification model to obtain a classification result output by the machine learning classification model. The present scheme improves the accuracy and efficiency of tectonic coal identification.

This invention provides a method and system for semantic novelty search of scientific and technological literature based on a large-scale pre-trained model, belonging to the field of information retrieval technology. The method includes: acquiring the target user's query input text and retrieval constraints; performing preprocessing and standardization to obtain preprocessed query input text; performing multi-granularity semantic encoding fusion to determine the semantic vector of the preprocessed query input text; performing domain-adaptive fine-tuning to obtain an optimized BERT model; constructing a feature classifier to generate feature semantic representations to obtain feature semantic representation results; performing comprehensive novelty assessment to obtain a comprehensive novelty rating; and performing automated difference analysis to obtain difference analysis results. This invention solves the technical problem in existing retrieval schemes based on pre-trained models that use single-granularity semantic representations, which cannot fully capture the complex semantic structure of professional terms, resulting in inaccurate novelty search results.

The application discloses a business management and control system for the tobacco industry, and relates to the technical field of information management of the tobacco industry.The system comprises the following: a distributed event coordination bus adopts a multi-level priority queue of a publish-subscribe mode and a differentiated retry and time sequence order preserving mechanism to deliver events; a data acquisition and standardization module switches data processing modes according to network states, completes multi-source business data conflict arbitration and standardization conversion; a data state processing module checks events and updates business states based on a unified metadata model; a streaming monitoring and decision module identifies abnormalities through sliding window aggregation and dynamic threshold identification, generates abnormal instructions and business alarm events; a metadata synchronization module synchronizes metadata changes through hot loading; and a unified operation and maintenance response module executes automatic response arrangement based on events.The application realizes accurate, real-time and active management and control of the whole life cycle of the business of the tobacco industry, and improves the robustness, business efficiency, compliance level and risk response capability of the system.

The application relates to the technical field of cognitive function assessment, in particular to a cognitive assessment method and system based on eye movement test. The cognitive assessment method based on eye movement test comprises the following steps: a cognitive task covering multiple cognitive subdomains is measured on a subject in a preset order, eye movement data corresponding to each cognitive task is obtained by taking a preset interest area of each cognitive task as a spatial constraint; quality control and preprocessing are performed on the eye movement data, multi-dimensional eye movement features are extracted and standardized; each cognitive task score and overall cognitive ability quantitative score are calculated to obtain standardized individual cognitive parameters; a normal population cognitive parameter distribution is taken as a benchmark to construct a multi-cognitive subdomain parameter distribution model, a subject cognitive state is judged by combining a machine learning classification model, and a mild cognitive impairment assessment result is obtained. The cognitive assessment method based on eye movement test can comprehensively depict the cognitive state of an individual and realize multi-dimensional joint assessment.

The utility model belongs to the field of torque measurement, and particularly relates to a ball screw friction torque measuring device which comprises a base assembly, a fixing plate and a measuring assembly. A sliding groove is formed in the top of a base of the base assembly, and four mounting blocks distributed in a rectangular shape are arranged on the outer surface. A motor and an electric push rod are installed on the fixing plate. The motor transmits power through a coupler. The measuring assembly comprises a connecting block, a first clamping plate, a second clamping plate and a long bolt, the first clamping plate and the second clamping plate are embedded into the first torque sensor and the second torque sensor respectively, the bottom of the connecting block is in sliding fit with the sliding groove, and the output end of the electric push rod is fixedly connected with the connecting block. The electric push rod drives the measuring assembly to move on the sliding groove, clamping and position adjusting of the ball screws of different specifications are achieved, the double torque sensors are used for synchronously collecting friction torque data, and the measuring efficiency and reliability of the friction torque of the ball screws can be effectively improved.

The invention relates to the technical field of soil testing, and discloses a lateral pressure adjustable expansive soil expansion rate measuring device, which comprises an experiment box, a sample cylinder and a measuring module, symmetrical through holes are formed in the side wall of the sample cylinder, a lateral pressure loading system is correspondingly arranged on the outer side of each through hole, and each lateral pressure loading system comprises a rigid piston head capable of extending into the cylinder, a piston rod, a closed medium pressure cavity and a miniature constant pressure pump; the pump is connected with the pressure cavity through a pipeline, and the piston head is driven to apply adjustable lateral pressure to the sample by accurately feeding and discharging a medium; the measurement module comprises a vertical deformation and lateral pressure measurement system and integrates free expansion and lateral confinement expansion measurement functions; the rigid piston and the water medium are adopted for constant-pressure loading, the structure is simplified, operation is convenient and fast, rapid response and high-stability control over lateral pressure are achieved, the membrane embedding effect is thoroughly eliminated, a multi-stage constraint continuous test can be completed through one-time sample loading, and the test efficiency and data consistency are remarkably improved.

The present application relates to the technical field of industrial unmanned aerial vehicle application and intelligent scheduling, and particularly discloses an unmanned aerial vehicle highway inspection task intelligent planning management system, which comprises a task input module, an intelligent planning engine, a task execution management module, a data hub and a feedback optimization module; the task input module receives and checks and analyzes inspection task requirements and highway basic information; the intelligent planning engine decomposes task priorities based on highway multi-dimensional attributes, dynamically allocates unmanned aerial vehicle resources and generates an optimal path meeting multiple constraints; the task execution management module issues adaptive instructions, monitors unmanned aerial vehicle states and processes inspection data; the data hub uniformly stores whole-process data; and the feedback optimization module optimizes a decision model based on historical data. The present application improves inspection planning accuracy and resource utilization, guarantees data quality, realizes self-evolution and ensures efficient and reliable highway inspection.

The present application relates to a kind of additive manufacturing process crack real-time monitoring method, device, equipment and medium, comprising: S1.in the additive manufacturing process, the laser radiation signal for forming piece is collected in real time and conversion, and the converted electric radiation characteristic signal is handled, and the radiation signal thermogram of forming piece is obtained;S2.to the forming piece simultaneously carries out CT scanning and obtains image data, and after processing, crack characteristic distribution chart is obtained;S3.radiation signal thermogram and crack characteristic distribution chart correspond with the layer number of forming piece, and the radiation signal crack characteristic data set corresponding to layer by layer point by point is obtained;S4.data set is used to train machine learning or deep learning model, and crack determination model is obtained;S5.crack determination model is used for online monitoring to carry out crack identification.The method of the present application can form the radiation signal crack database corresponding to layer by layer point by point in three-dimensional space, and the data consistency and migratability are improved.

The application discloses an interactive method for quickly adding a recurring schedule in a waterfall month view. The method involves the following steps: providing a schedule template selector, performing schedule drop, conflict detection and prompt, and differential submission and server verification in the waterfall month view. The application provides a schedule template selector in the user interface, displays a plurality of predefined schedule templates, selects at least one schedule template from them, applies the schedule template to the corresponding date to generate a schedule instance and performs a preview in response to the user's operation on the date grid, performs conflict detection and prompt during the preview, and submits the data to the server in response to the user's submission operation. The application improves the accuracy and efficiency of the interactive operation process of quickly adding a recurring schedule in a waterfall month view, and solves the problems of incoherence, low efficiency and high error rate in the prior art.

A dynamic statistics and result analysis system for watershed river and lake shoreline development and utilization comprises a remote sensing data acquisition module, a data preprocessing and classification module, a database construction module, a change trend analysis module, a river and lake directory and association retrieval module and an interaction management module. The data preprocessing and classification module is used for data preprocessing and statistical classification, the database construction unit is used for constructing a system database, the variation trend analysis module is used for variation trend analysis and statistical report generation, and the river and lake directory and association retrieval module is used for directory maintenance and result association. According to the dynamic statistics and result analysis system for watershed river and lake shoreline development and utilization, a remote sensing data statistical classification algorithm based on a neural network is proposed to classify remote sensing data; and a river-lake shoreline development and utilization change trend analysis algorithm based on deep learning is proposed to analyze the change trend.

The invention discloses an unstructured data report filling method and system based on a large language model, and belongs to the technical field of artificial intelligence, and the method comprises the steps: preprocessing unstructured data, and generating preprocessed unstructured text data; based on the preprocessed unstructured text data, constructing and executing a semantic structured extraction chain; field mapping and report template binding are carried out based on the constructed semantic structured extraction chain, and a field mapping rule is established; based on a field mapping rule, automatically generating and filling a report, and automatically generating the report according to the extracted structured data; and interpretable proofreading and manual auditing are carried out on the automatically generated report. According to the method, implicit information can be accurately extracted, and semantic inference and association are supported; automatic report filling is achieved, and the manual arrangement workload is remarkably reduced; and the output is explainable, and each field can be positioned to the original text, so that the auditing credibility is improved.

The invention provides a document synchronization method and device, a storage medium and electronic equipment. The method comprises the steps of obtaining first document content of a to-be-synchronized target local document, and obtaining a document content version of a target cloud document corresponding to the target local document in current cloud storage; second document content is determined based on the document content version, and a first document change operation set is determined according to the difference between the first document content and the second document content; generating a first document synchronization instruction according to the first document change operation set; and sending the first document synchronization instruction to the target server, so that the target server updates the content of the target cloud document based on the first document synchronization instruction. Only the change data is transmitted to the server, redundant data and meaningless time consumption generated by full-amount synchronization are reduced, the change operation is re-determined based on the document content indicated by the cloud document version, data loss or coverage errors of the cloud document are avoided, and data consistency and integrity in the data synchronization process are improved.

The invention relates to the field of artificial intelligence and pattern recognition, and discloses an archive category automatic judgment and description template matching method based on pattern recognition. The method comprises the following steps: acquiring original data and metadata of an archive; performing structured analysis and semantic feature extraction to generate a multi-dimensional feature vector; the archive category is judged through the ensemble classification model; searching a description template with the highest matching degree in the dynamic template library, and automatically generating and registering a new template if there is no high matching template; and finally, automatic recording is completed. Through a dynamic template mechanism and online incremental learning, full-automatic, self-adaptive and sustainable evolution of archive description is realized, and description efficiency, consistency and system robustness are improved.

The present application relates to the technical field of robots, and in particular to a motion generation method, system, device and storage medium for a humanoid robot, comprising: physical simulation rendering generates a visual video containing visual representation of joint motion state of the robot; analyzing the visual video to obtain frame images, natural language instructions and motion sequences, decoding output trajectory data; and outputting control instructions through a hierarchical architecture of high-level motion planning and bottom-level dynamics tracking to drive the robot to perform actions. The present application implicitly encodes the joint state of the humanoid robot in the visual features, improves the inference efficiency and generalization ability of the model, and the hierarchical decoupling architecture of the large model high-level trajectory planning and the reinforcement learning bottom-level dynamics tracking controller pays a small performance cost, which can compress the inference time delay to within 50ms and the parameter quantity to below 2 billion, and balances the high generalization of the stateless VLA and the small model, few-step inference, edge low latency, and low power requirement.

The application discloses a kind of ship navigation intelligent obstacle avoidance methods based on multi-modal perception, it is related to ship intelligent navigation technical field, including the following steps: acquisition multi-modal perception data and standardization, generate ship navigation perception dataset;Arrange perception node, generate ship multi-node perception layout;Establish ship circumferential perception coordinates and generate ship circumferential layered covering relationship;Divide far distance early warning area, near distance identification area and berthing monitoring area, generate ship zoned perception state;Process far distance radar data and near distance millimeter wave radar data, generate radar target perception result;Process visual image data, generate visual target perception result and ship look-around panoramic image;Match radar target perception result and visual target perception result, combined with ship body attitude change is directed to compensation and collision risk judgment, generate ship navigation intelligent obstacle avoidance result.The application has the advantages of wide perception range, target fusion accuracy and obstacle avoidance judgment stability.

The application relates to the technical field of electric power inspection, and discloses a method and system for power grid inspection and identification of an unmanned aerial vehicle, which comprises the following steps: acquiring channel path data, tower position data and inspection task configuration data; generating a channel inspection main flight path and a tower fine inspection sub-flight path; establishing a channel inspection mode parameter set and a tower fine inspection mode parameter set, and setting a mode switching trigger area; controlling the unmanned aerial vehicle to execute the channel inspection mode along the channel inspection main flight path; after completing tower inspection, executing key component collection integrity judgment, and executing supplementary shooting control when there are uncovered key components or unqualified images; then switching back to the channel inspection mode to continue flying; and identifying and analyzing the channel inspection data and the tower fine inspection data, and performing unified spatial marking and task association output. The application realizes integrated execution of channel inspection and tower fine inspection, reduces repeated flight, and improves inspection data integrity.

The invention discloses a double-front-axle steering point position sensitivity analysis and joint optimization method and system, and relates to the technical field of steering point position optimization. Comprising the following steps: S1, acquiring data of each corner, and preprocessing the data; s2, performing frequency domain analysis on the turning angle of the front axle steering drop arm, judging the non-constant-speed disturbance intensity of a universal joint, and analyzing and removing a periodic disturbance component in the turning angle of a wheel to obtain a turning angle trend of the wheel; s3, an ideal steering relation is established, the deviation between the wheel steering angle trend and the ideal steering relation is evaluated, and the steering geometric error degree is recognized; carrying out single-axis step length adjustment on the coordinates of the key structure points, and analyzing and screening out optimized candidate positions; and S4, constructing a joint optimization target, and performing iterative updating in combination with the non-constant-speed disturbance intensity of the universal joint to obtain optimized key structure point coordinates. The problem that structure point position adjustment is misjudged due to the fact that universal joint non-constant-speed disturbance and steering geometric deviation are difficult to distinguish in the point position optimization process of the double-front-axle steering system is solved.

The invention discloses an auxiliary device for mutual inductor wiring, and belongs to the technical field of mutual inductor wiring. The auxiliary device comprises a base, a positioning rod, a lifting rod, a fixing assembly, an L-shaped pressing plate, a circular pressing plate, a crimping assembly and a cleaning belt. According to the invention, the L-shaped pressing plate is used for crimping a secondary side wiring copper plate, and the circular pressing plate is used for crimping a primary side wiring bolt; the wiring quality and safety are greatly improved; a gear and rack transmission and trapezoidal block slope jacking structure is adopted, the full-automatic process of ascending, rotating and pressing of the lifting rod is achieved, dust, oxide layers and other impurities on the upper surface of a wiring copper plate and the lower surface of an L-shaped pressing plate can be cleaned through automatic circulating rotation of a cleaning belt, and the verification precision of a mutual inductor is effectively improved; the problem that in a traditional wiring mode, due to poor wiring contact, verification errors are generated by the mutual inductor is solved.

The present application relates to the technical field of medical devices, and more particularly to an intraocular pressure measurement calibration device and calibration method, the intraocular pressure measurement calibration device comprising a lifting focusing mechanism, an overall rotating mechanism and an angle adjusting mechanism; wherein one end of the lifting focusing mechanism is connected with a shooting unit, and the other end is connected with the angle adjusting mechanism; the angle adjusting mechanism is engaged with the overall rotating mechanism. Through the cooperation of the mechanisms, the present application realizes the precise calibration of focal length, direction and angle, greatly reduces the operation difficulty, and can adapt to the clinical application scenarios of implantable intraocular pressure monitoring systems; through the step-by-step calibration of image definition optimization, stripe posture correction and incident light axis alignment, the present application effectively solves the problems of poor interference stripe quality and difficulty in demodulation in the prior art, and significantly improves the accuracy and data consistency of intraocular pressure measurement; the device structure is compact and has strong compatibility, and can be used with existing implantable intraocular pressure sensors and image processing systems, and has good clinical promotion value.

The present invention relates to the cross - technical field of tunnel engineering and geotechnical mechanics testing, and discloses a triaxial compression grouting test device and method with a variable bearing contact area, which includes a cube - shaped specimen; the loading system includes a reaction frame and a plurality of hydraulic rods rotatably connected to the reaction frame through driving members. The output ends of the hydraulic rods are circumferentially connected with a plurality of pressurizing mechanisms, and the output ends of the hydraulic rods are also detachably connected with a plurality of groups of bearing blocks arranged circumferentially. The front surfaces of the plurality of bearing blocks in each group are arranged in an inclined plane in the counter - clockwise direction, and the outer edges of the plurality of bearing blocks in each group gradually shrink in an arc shape. The plurality of pressurizing mechanisms are respectively used to abut against the front surfaces of the plurality of groups of bearing blocks, and the back surfaces of the plurality of bearing blocks are fitted with corresponding multi - layer nested bearing plates, and the bearing plates are used to abut against the end faces of the specimen. The present invention can more realistically restore the stress state of tunnel surrounding rocks, can make the bearing area adaptively adjust as the volume of the specimen decreases, and avoid stress concentration.

The invention belongs to the technical field of electromagnetic nondestructive testing, and particularly relates to a decarburized layer thickness detection method and device based on pulsed eddy current and a medium. A probe comprising multiple coils and a self-calibration unit is placed on the surface of a workpiece to apply excitation, and positioning is completed; carrying out amplification and phase calculation on the received signal, analyzing and optimizing a coil excitation pickup sequence through sample piece pre-scanning, collecting a response signal based on the optimized sequence, and extracting time-frequency domain composite features; performing multi-frequency scanning on the workpiece by using the optimized sequence, extracting features of a scanning position, inputting the features into a pre-trained deep learning model, and outputting thickness information; and judging whether a standard is met according to the thickness information. According to the method, by optimizing the coil excitation sequence and combining with the deep learning model, the detection precision is improved, and nondestructive evaluation of the thickness of the decarburized layer can be effectively achieved.

This application provides a priority-based database transaction scheduling method and system under the Tianmai operating system, relating to the field of database technology. The method first divides the transactions to be scheduled into hard real-time transactions, soft real-time transactions, and fixed real-time transactions. Then, it determines the initial priority based on the transaction deadline, and subsequently determines the transaction data validity period based on transaction characteristics and the validity period of the accessed data object. The system determines the final deadline based on the smaller value between the transaction data validity period and the transaction deadline, and dynamically adjusts the current priority of the transaction based on this deadline. Subsequently, it determines the transactions to be executed based on the transaction type and current priority, and performs preemption control according to the transaction type order and current priority during lock resource contention. This scheme can improve the real-time performance, transaction correctness, and concurrency control capabilities of high-priority transactions in airborne embedded databases.

The application belongs to the technical field of power secondary equipment modeling, and discloses a power secondary equipment visual modeling and configuration system and method based on external electrical signal labeling, equipment and medium, to solve the problems of large manual workload, poor configuration consistency, missing function model and non-uniform engineering model in traditional processes. The system comprises: a geometric modeling module for establishing a geometric model; a PLP construction module for organizing physical terminals Pin, ports Port and logic layers L to form physical logic ports PLP; an external electrical signal import module for receiving external signals and labeling to Port; a semantic binding module for binding electrical signal semantics to PLP; a Function organization module for combining multiple PLPs to form a function object Function; and a model export module for exporting a structured file. The application realizes the standardization and semanticization of equipment modeling, and improves the design efficiency and data consistency.

This application provides a circuit and system for voltage information diagnosis and self-calibration. By adding resistors R1 and R2 in series, the total cell voltage is divided, reducing the sampling point voltage Vsample to the measurement range of the analog-to-digital converter (ADC) of the microcontroller (MCU). The MCU accumulates and calculates the total voltage Vafe by sampling the voltages of each cell from the analog front-end (AFE). Based on the Vafe data from the AFE, the total cell voltage Vmcu is automatically calibrated to obtain the calibrated total cell voltage Vcalib. Simultaneously, the MCU compares the calibrated total cell voltage Vcalib with the Vafe voltage in real time to obtain the deviation value Vdiag. When the deviation value Vdiag exceeds a set threshold, fault diagnosis is triggered, indicating an abnormality in the cell voltage sampling function, and a safety protection mechanism is activated to meet functional safety requirements. This circuit, through reasonable hardware design, voltage protection, cell total voltage calibration, and fault diagnosis mechanism, not only improves the accuracy of system measurements but also achieves self-calibration and fault diagnosis, making it suitable for applications with high stability requirements.

A fluid and solid contact interface simulation method based on flow-solid interface coupling dynamic mesh, the initial mesh of the target geometric model is divided, and the flow field, temperature field and interface evolution driving field are solved according to the initial conditions and boundary conditions, after the theoretical displacement amount of the mesh node in each time step is obtained, based on the interface coupling dynamic mesh (ICDM) method, the local real normal displacement scalar is sequentially executed node state lock protection, equivalent normal projection and parallel synchronous processing; then through the grid node movement, the grid coordinates are updated and the flow field, temperature field and interface evolution driving field are re-solved on the updated grid to obtain a new round of node displacement, and whether the simulation reaches the preset time or the deformation target is judged; finally, the final interface topography, grid node coordinates, grid quality, flow field distribution, temperature field distribution, pressure field distribution and / or heat exchange performance parameters are output. The present application can realize the synchronous iterative update of the deposition behavior and the temperature field and even the flow field, dynamically reflect the influence of deposition on the heat exchange performance, thereby greatly improve the simulation accuracy and physical consistency and be suitable for the subsequent machine learning reverse design of a large number of structure parameters and heat exchange performance samples.

Some embodiments of the present application show a display device and a channel program information display method, the method comprising: in response to an instruction for playing a program of a first channel, if a playing signal source of the first channel is broadcast, respectively acquiring and storing a broadcast electronic program guide and a network electronic program guide of the first channel; in response to the instruction of displaying the program information, if the network electronic program guide of the first channel is incomplete, acquiring a broadcast electronic program guide of the first channel; fusing a network electronic program guide and a broadcast electronic program guide; and displaying the program information of the first channel based on the fused electronic program guide. According to the embodiment of the invention, when a channel program is played by a broadcast signal source, broadcast and network electronic program guides can be stored at the same time. When the program information of the channel needs to be displayed, the network electronic program guide is preferentially acquired, and if the network electronic program guide is incomplete, the broadcast electronic program guide can be fused, so that more complete program information is displayed to a user.

The invention provides an urban waterlogging prevention system optimization method and system based on point cloud data, and relates to the technical field of data processing. The method comprises the following steps: acquiring laser point cloud data of a target town area, and carrying out filtering and feature extraction processing on the laser point cloud data to generate point cloud topographic feature data; extracting structural form data and material attribute data of the waterlogging prevention structure; establishing a semantic mapping relationship between the structure type and the hydraulic structure data in the waterlogging prevention semantic map, and performing data association on the semantic mapping relationship and the point cloud topographic feature data; generating a three-dimensional hydrodynamic model by utilizing hydraulic structure data provided by the waterlogging prevention semantic map; performing water flow calculation on different rainfall scenes by using the three-dimensional hydrodynamic model to generate hydrodynamic simulation data for optimizing the waterlogging prevention system structure; and determining an optimization strategy of the waterlogging prevention structure in the drainage blocked area by combining the waterlogging prevention semantic map. According to the scheme, the modeling precision and optimization reliability of the urban waterlogging prevention system can be improved.

The application provides a Cache write cache device and a control method thereof, and relates to the technical field of artificial intelligence chips. The device comprises: a write cache table for storing a plurality of write cache entries, each write cache entry corresponding to attribute information of a Cache line; a memory for storing data of the Cache line corresponding to each write cache entry; a write cache control unit for processing storage requests and loading requests based on the write cache table, and for writing the data of the Cache line corresponding to the write cache entry from the memory to a data storage unit when a preset write-back condition is met; a loading request queue for buffering loading request information from a previous execution unit or a previous Cache; and an effective address generation module for generating a corresponding effective write-back address according to a write data effective flag signal recorded in the write cache entry when performing a write-back operation. The application effectively reduces the write-back pressure on the lower-level storage, eliminates the structural blockage of the storage backend to the computing frontend, and significantly improves the overall memory access efficiency and computing throughput.