132results about How to "Balance performance" patented technology

The invention provides a valve leakage diagnosis method and system based on an improved Hereum algorithm optimized SVM (Support Vector Machine), and relates to the technical field of valve leakage diagnosis. Decomposing the collected original signal to obtain a limited number of IMF components; calculating a fuzzy entropy, and constructing a feature vector; taking the feature vectors as data samples, and dividing the data samples into a training set, a verification set and a test set; using the training set to train the SVM model, and according to the classification error rate of the verification set on the trained SVM model, performing iterative optimization on the parameters of the SVM model based on an improved He-Ma algorithm to obtain an optimized SVM model; inputting the test set into the optimized SVM model to obtain a valve leakage diagnosis model; and acquiring a new acoustic emission signal, and inputting the new acoustic emission signal into the valve leakage diagnosis model to obtain a diagnosis result. The method not only can improve the classification accuracy and generalization ability of the diagnosis model, but also can enhance the stability of the model under different working conditions and environmental noise.

The invention relates to the technical field of sea surface target multi-mode fusion recognition, in particular to a multi-mode fusion sea surface target recognition method of a high-dynamic unmanned platform. According to the method, a multi-modal fusion recognition framework based on a multi-kernel learning method is provided, and a small-sample multi-modal fusion target recognition algorithm for a sea-surface target is designed on the basis of the multi-modal fusion recognition framework. Target detection and feature extraction are respectively performed according to data characteristics of different sources of visible light images, SAR images and active and passive radar one-dimensional range profile signals, multi-mode heterogeneous features are mapped by using a multi-scale kernel method, weighted fusion of a kernel matrix is performed, global training optimization is performed, and finally a classification result of each sample can be output. The algorithm not only considers the performance of a data-driven target recognition algorithm, but also considers platform hardware computing power limitation and small sample conditions, so that the balance among the performance, computing power and samples can be realized.

The invention discloses a data storage system of a digital country based on cloud computing, which comprises a client, a data transmission module and a cloud computing center, relates to the technical field of data storage, stores hot data on a high-performance hot data storage layer, ensures that frequently accessed data can be quickly responded, improves the system performance, and at the same time, stores the hot data on a high-performance hot data storage layer. The cold data is stored in a low-performance cold data storage layer, the storage cost is reduced, effective balance of performance and cost is achieved, the data is automatically migrated among different storage layers along with change of data access frequency, so that the performance and the cost are balanced, the frequently accessed data is migrated into a cache, the access speed of the data is increased, and the data access efficiency is improved. And establishing a relationship between a time interval and a data type of historical data, predicting data which are possibly accessed in a given time interval, adding the data into the cache, dynamically adjusting the cache, and ensuring the validity of the cache.

The invention discloses a crowdsourcing acquisition map data desensitization method and system based on deep learning. According to the method, a'high-precision map desensitization rule and target sensitivity level table 'is constructed for grading and classifying a to-be-desensitized target, and a desensitization method is specified and comprises a shape type, a privacy type and a character type. For shape type and privacy type sensitive targets in a high-precision map, the positions of the sensitive targets are detected by using a sensitive target detection algorithm based on Faster RCNN, and different desensitization methods are used according to the sensitive levels of the sensitive targets. For character sensitive targets in a high-precision map, all character areas in the map are detected by using a natural scene text detection model I3CL, the text areas are recognized by using a text recognition model SVTR to obtain text contents of the character areas, sensitive characters in the character areas are matched through a sensitive word matching module, and the sensitive characters in the text areas are matched through a sensitive word matching module; and a sensitive area removing model based on LaMa is used for the sensitive character area, and sensitive characters are removed from the graph.

The invention discloses a storage chip energy consumption test method and device based on multi-objective optimization. The method comprises the following steps: S1, initializing a test environment; s2, setting a multi-objective optimization model and a fuzzy logic system; s3, constructing an initial test data set; s4, calculating the membership degree of each test target; s5, generating a group of initial optimization parameter combinations; s6, dynamically adjusting the test parameters on the basis of the initial optimization parameter combination; s7, generating an optimized test parameter combination; s8, repeating the steps S6 and S7 until a preset optimization convergence condition is reached, and generating a final test parameter combination; s9, obtaining final test result data; and S10, analyzing the final test result data, comprehensively evaluating the performance of each test target through a fuzzy logic system, and generating a multi-target optimization test report of the storage chip. According to the method, the problems of multi-target test balance, test process optimization and balance between efficiency and accuracy in the prior art are effectively solved.

The invention provides a defect classification method and system for a semiconductor chip and a readable storage medium. The training method comprises the following steps: acquiring training set data, wherein the training set data comprises different types of defect images of the semiconductor chip; inputting the training set data into an initial defect classification model based on a deep convolutional neural network so as to train the initial defect classification model; in the training process, an attention mechanism module is introduced to the tail of at least part of convolution blocks of the initial defect classification model, and residual connection is introduced to at least part of the convolution blocks; and a target defect classification model is obtained after multiple rounds of iterative training. The method effectively balances model performance and computing resource requirements, and ensures that higher efficiency is kept while the classification precision is improved.

The invention relates to the technical field of battery energy storage system evaluation, and relates to the technical field of battery energy storage, in particular to a method for balancing the service life and performance of a fuel cell, which comprises the following steps of: 1, acquiring multi-source data of the fuel cell, and applying a series of preset pulse current load sequences to the fuel cell, measuring response parameters output by the fuel cell in real time; 2, Fourier transform processing is carried out on the response parameters, and frequency domain signals of the response parameters are obtained; the performance and service life balance model is constructed based on the fusion feature vector containing the dynamic performance, the service life characteristic and the operation state of the fuel cell, the load distribution weight can be dynamically adjusted, the performance and the service life of the single fuel cell are kept balanced, and therefore the overall service life of the fuel cell pack is prolonged, and the reliability of the fuel cell pack is improved. And the influence on the performance of the whole system due to overload or too fast loss of the single fuel cell is avoided.

The invention discloses a hydrogen fuel cell hybrid power system grid-connected control method, which comprises the following steps of: analyzing and establishing a plurality of models of a system, comprehensively considering a plurality of cost weights on the basis of a hydrogen consumption model of the system, and constructing a comprehensive cost function; establishing a system operation state model; the speed, the acceleration and the system operation state serve as input of a self-adaptive Elman neural network together, a state transition matrix real-time correction factor of the system in the state is obtained, and a basic state transition matrix is adjusted through the real-time correction factor to obtain a real-time state transition matrix after the influence of the current driving condition on system state transition is considered; and according to the real-time state transition matrix, predicting the system demand power within a period of time, solving the comprehensive cost to enable the cost to be minimum, and completing the power distribution between the multiple stacks of fuel cells and storage batteries. According to the method, the fuel economy of the system is ensured, the internal balance of the multi-stack fuel cell system is improved, the service life of the system is prolonged, and the real-time prediction capability of the system is improved.

The invention discloses an RFID antenna power control method and system considering partition identification, electronic equipment and a storage medium. The control method comprises the steps that S1, an RFID antenna transmits an electromagnetic signal to an RFID tag according to a preset transmitting power level; s2, the RFID tag carried by each electric energy meter responds to the electromagnetic signal and receives a reference message fed back by the corresponding electric energy meter; s3, the RFID reader-writer receives and records the RFID data transmission rate, the response signal strength, the corresponding electric energy meter reference information and the transmitting power level of each RFID tag under different transmitting power levels; s4, calculating a characteristic value according to the received data, and calculating a confidence coefficient according to the fed-back quantity of the electric energy meters; and S5, determining the optimal transmitting power level of the RFID antenna by combining the feature values and the confidence coefficients under different transmitting power levels. The problems that an existing RFID system is difficult to achieve accurate fine-grained identification in a complex storage environment, and the complexity of data management is increased due to the signal reflection and interference problems are solved.

The invention relates to a point-level supervision video clip positioning method and system based on corresponding learning and application, and belongs to the technical field of target video clip positioning, and the method comprises the steps: firstly extracting features through a pre-training model, obtaining an enhanced visual representation and an enhanced text representation, processing the enhanced visual representation and the enhanced text representation, and obtaining a pre-training model; obtaining a multi-granularity text representation and a multi-granularity visual representation; finally, the multi-granularity text representation is subjected to maximum pooling, and representation integrating semantic information of the whole sentence is obtained; in the training stage, Gaussian distribution is combined with key frames, and different video clips are measured; in the reasoning stage, the best matching key point of the video and the query is identified firstly, and then the point is expanded to obtain a predicted video clip most similar to the query; and a model loss optimization model formed by three parts of global alignment loss, frame-fragment corresponding loss and fragment-frame corresponding loss is utilized. According to the method, the granularity and scale problems existing in video clip positioning under point supervision are solved, and the model performance is effectively improved.

The invention belongs to the technical field of engines, and discloses a silencing ring gap control method, a supercharger and a vehicle. The noise reduction ring gap control method comprises the steps that a noise reduction ring gap calibration value is obtained based on the current engine working condition; the silencing ring is adjusted, so that the silencing ring gap is equal to the silencing ring gap calibration value; the calibration method of the noise reduction ring gap calibration value comprises the steps that supercharger noise corresponding to different noise reduction ring gaps is tested under the working condition of a to-be-tested engine; and extracting the width value of the silencing ring gap when the noise is minimum under the working condition of the to-be-tested engine, calibrating the width value as the silencing ring gap calibration value under the working condition of the to-be-tested engine, and obtaining the corresponding relationship between the working condition of the engine and the silencing ring gap. According to the noise reduction ring gap control method, accurate control over the noise reduction ring gap can be achieved, the problem of supercharger noise caused by improvement of the engine performance is solved, and effective balance between the engine performance and noise control can be achieved.

The invention provides a large 3D scene texture dynamic loading and unloading method, medium and system, and belongs to the technical field of 3D scene loading.The large 3D scene texture dynamic loading and unloading method, medium and system comprise the steps that three-dimensional scene parameters are collected, texture frequency parameters are decomposed to obtain basic and fluctuation components, a texture loading model is constructed, and a coefficient matrix is generated; packaging the texture file into a fixed-size block, building a double-layer resource optimization model based on a coefficient matrix, setting global and local optimization targets, and calculating a comprehensive contribution coefficient; an initial resource allocation scheme is obtained through an evolutionary algorithm, and a final scheduling scheme is obtained through dynamic programming; according to the texture blocks, loading the texture blocks, determining a recovery strategy in combination with life cycle, grouping association and state transition prediction, and executing resource recovery; the process comprises texture analysis, model construction, resource optimization, scheduling and recovery, and efficient management of three-dimensional scene textures is realized; the resource scheduling decision-making mechanism can be optimized, and the user experience is improved.

The invention discloses a tundish cover castable, and relates to the technical field of tundish refractory materials. The tundish cover castable is prepared from the following raw material components in percentage by mass: 7.8 to 9.6 percent of alumina of 8 to 5mm, 2.6 to 4.3 percent of alumina of 5 to 3mm, 2.6 to 4.3 percent of alumina of 1 to 0mm, 17.4 to 20.8 percent of alumina of 200 meshes, 9.6 to 13 percent of mullite of 5 to 3mm, 22.6 to 26 percent of mullite of 3 to 1mm, 9.6 to 13 percent of mullite of 1 to 0mm, 7 to 8.7 percent of mullite of 200 meshes, 2.6 to 3.5 percent of composite binding agent, 0.8 to 1.7 percent of aluminum silicate, 0.05 to 0.07 percent of SiC nanowire, 0.2 to 0.3 percent of aluminum titanate and the balance of water. The invention discloses a tundish cover castable which comprises the following components in percentage by weight: 0.4-0.7% of alumina micro powder, 0.08-0.11% of sodium tripolyphosphate, 2.6-3.5% of silica fume, 0.87% of steel fiber and 1.3-1.7% of high-alumina cement. Compared with a traditional process, the tundish cover castable has the advantages that the overall performance of the tundish cover castable is remarkably improved, the breaking strength and compression strength of the castable are effectively enhanced, the service life is remarkably prolonged, and the production cost is reduced. And the production cost is greatly saved, and the economical efficiency and the operation rate of continuous casting operation are improved.

The invention discloses a territorial space planning surveying and mapping three-dimensional analogue simulation system, device and terminal, and the system comprises a data collection and processing module, a three-dimensional modeling module, a planning scheme importing and editing module, an analogue simulation and analysis module, a result output and sharing module and an intelligent decision support module which have corresponding functions and performance indexes. The method has the advantages that data acquisition precision, model precision, simulation accuracy and the like are achieved, various innovative functions and technologies such as model dynamic updating, multi-person collaborative editing, scene simulation and the like are achieved, and scientificity and efficiency of territorial space planning are improved; the precision and efficiency of territorial space planning are improved, accurate data acquisition, three-dimensional fine modeling, comprehensive simulation analysis, intelligent decision support and diversified result output are realized through a multi-module cooperation and innovation technology, scientific and reasonable planning is promoted, and industry development is promoted.

The invention discloses a vehicle energy management method. The method comprises the following steps: acquiring a preset vehicle operation condition; constructing a state variable equation and a control variable equation under a preset vehicle operation condition, and determining an algebraic constraint equation based on the state variable equation and the control variable equation; constructing a fuel cell attenuation equation based on the vehicle operation load and the vehicle start-stop times; constructing an ampere-hour circulation equation of the power battery; solving the algebraic constraint equation on the basis of a nonlinear programming method by taking the fuel cell attenuation amount and the power cell ampere-hour circulation amount as target quantities and aiming at prolonging the service life of the fuel cell and the service life of the power cell in a preset vehicle working condition to obtain an optimal state variable and an optimal control variable; and performing energy management on the fuel cell hybrid vehicle based on the optimal state variable and the optimal control variable.

The invention discloses a server board card, a performance adjusting method of the server board card and a chip, and belongs to the field of servers. The server board card comprises a management processor and a computing chip, and the management processor is used for managing elements in the server board card; the management processor is used for acquiring historical operation parameters of the elements, and the historical operation parameters comprise parameters used for reflecting the operation performance of the elements; the calculation chip is used for predicting a current operation parameter of the element according to the historical operation parameter of the element, and the current operation parameter is used for representing an operation parameter corresponding to a current moment in an operation parameter change trend corresponding to the historical operation parameter; and the management processor is also used for adjusting the performance of the element according to the current operation parameter of the element. According to the method and the device, the limitation of adjusting the performance of the board card based on the preset performance level is broken, the performance adjustment with finer granularity is realized, the power supply use and performance of the server board card are balanced, and the response capability of the server board card for different performance requirements is enhanced.

The invention discloses an optical fiber sensing system and method for on-line monitoring in human knee arthroplasty. The optical fiber sensing system comprises an optical fiber grating pressure sensing probe 1, a sensing optical cable 2, a signal acquisition and communication module 3, an optical fiber grating demodulation module 4 and a computer display module 5, the pressure sensing probe is connected to the fiber bragg grating demodulation and processing module through a transmission optical cable for signal decoupling, result data is transmitted to the computer for processing through the signal acquisition and communication module, the influence of temperature fluctuation on the measurement precision is eliminated through a difference algorithm, and the display screen displays the pressure at different positions in the operation process in real time. And chart analysis and surgical operation parameter AI optimization can be carried out on the stored data. Due to the application of the system technology, inherent defects of knee joint balance caused by subjective judgment of surgeons are greatly reduced, reference can be provided for the doctors to evaluate soft tissue tension in total knee arthroplasty, and then the success rate of the operation is increased.

The invention discloses a multi-window rendering optimization method for containerized swan application, and the method comprises the steps: building a window adaptation layer at a host side to achieve the instruction conversion from a container side to the host side, analyzing a rendering mode at the container side when the swan application Abilitity is started, generating a structural body containing window full-amount attributes, and transmitting the structural body to the host side; the adaptation layer creates a corresponding native window to establish mapping between a swan gap window and a host window, a Mesa compatible buffer pool is initialized for the window according to a rendering mode, when an application initializes a frame buffer creation request, a host side allocates an idle buffer from the corresponding buffer pool, a container side establishes mapping between a virtual frame buffer object and a host side buffer, and the virtual frame buffer object is mapped to the host side buffer; when the swan application submits display, the cross-container sharing handle of the rendering buffer area is exported from the container side and sends buffer area data to the host side, the host side imports the sharing handle and then binds the GPU buffer area to the surface of a target window to submit a refreshing request, and the synthesizer synthesizes application rendering content and original window decoration and sends the application rendering content and the original window decoration to the display equipment to complete screen display.

The invention provides an inverter type selection method of a photovoltaic power generation system and a related product. The inverter type selection method of the photovoltaic power generation system comprises the following steps: acquiring a photovoltaic design scheme of the photovoltaic power generation system; determining electrical parameters of three groups of different types of inverters according to the electrical parameters, wherein the three groups of different types of inverters comprise a string type inverter, a centralized inverter and a micro inverter; determining the mechanical performance requirements of the three groups of different types of inverters according to the installation mode; selecting matched alternative equipment models of three different types of inverters from an inverter model selection library according to the electrical parameters and the mechanical performance requirements; generating three groups of alternative inverter type selection schemes according to the selected alternative equipment types of the three different types of inverters; and selecting a target inverter from the three groups of alternative inverter type selection schemes. The scheme provided by the invention is wide in application scene, and suitable inverter schemes can be provided for various photovoltaic design schemes.

The invention discloses a frequency adjustment method and device, electronic equipment and a readable storage medium, and relates to the technical field of processors, the frequency adjustment method and device are applied to the electronic equipment, the electronic equipment comprises multiple processors, and the method comprises the steps that a target processor set, used for executing task loads, of the electronic equipment and load information of the task loads are determined; obtaining first information of each first processor, wherein the first information comprises at least one of frequency upper and lower limit information, operation frequency when each first processor executes the task load and occupied information; based on load information and the first information, determining a processor needing frequency adjustment as a second processor through an adjustment model, and determining a target adjustment strategy of frequency adjustment of the second processor; and adjusting the frequency of the second processor based on the target adjustment strategy. The power consumption is reduced as much as possible on the premise of ensuring that the electronic equipment is not blocked in operation, and the relative balance between performance and power consumption is achieved.

The invention relates to the technical field of battery packs, and discloses a bottom protection structure for a battery pack and the battery pack, the bottom protection structure for the battery pack comprises a liquid cooling plate, a bottom protection plate and a buffer layer; the bottom of the liquid cooling plate comprises a non-flow channel area and a protruding flow channel area; the bottom protection plate is arranged at the bottom of the liquid cooling plate, and the bottom protection plate and the liquid cooling plate are arranged at intervals; the buffer layer is arranged between the liquid cooling plate and the bottom protection plate, and the bottom of the buffer layer abuts against the bottom protection plate; in the flow channel area, the top of the buffer layer abuts against the liquid cooling plate; according to the present invention, the impact force can be effectively absorbed and dispersed when the different regions of the battery pack are impacted so as to ensure that the battery cells in the battery pack are not extruded when the battery pack is subjected to the external force effect, such that the battery pack can be prevented from being damaged, and the battery pack is prevented from being damaged. And the protection level of the battery cells in the battery pack is improved, and the overall safety of the battery pack is ensured.

The invention discloses a multi-layer flame-retardant fabric with low-temperature cryogenic protection and flashover prevention functions and a preparation method of the multi-layer flame-retardant fabric. The multi-layer flame-retardant fabric comprises a thermal protection outer layer, a heat buffer layer, a flame-retardant thermal insulation layer and a skin-friendly comfortable layer; the thermal protection outer layer is a three-dimensional warp-knitted spacer fabric subjected to hydrophobic treatment and comprises a thermal protection outer layer surface layer and a thermal protection outer layer inner layer, and spacer yarns are alternately connected back and forth between the thermal protection outer layer surface layer and the thermal protection outer layer inner layer in the single vertical direction to form the three-dimensional warp-knitted spacer fabric; the spacing distance between every two adjacent spacer yarns is 0 to 10 mm; the heat buffer layer is a coating formed on the inner layer of the thermal protection outer layer; the flame-retardant thermal insulation layer is a thermal insulation felt made of a flame-retardant material The skin-friendly comfortable layer is formed by flame-retardant cotton fabric, polyimide fabric or aramid fiber / flame-retardant viscose blended fabric. The combined fabric provided by the invention has the advantages of low-temperature cryogenic protection, flame retardance, high-heat protection performance and high-temperature resistance, and realizes the balance of structure, function and performance.

The invention discloses a rare earth aluminum alloy cable material and a preparation method and application thereof, and relates to the technical field of aluminum alloy material preparation. The rare earth aluminum alloy cable material is prepared from the following raw material components in percentage by mass: 0.3 to 0.6 percent of Ce, 0.3 to 0.6 percent of Sc, 0.05 to 0.15 percent of Y, 0.10 to 0.20 percent of Fe, 0.05 to 0.15 percent of Si and the balance of Al. The preparation process comprises the steps of melting, refining, casting, homogenizing treatment, hot rolling, drawing and aging treatment. The rare earth elements Ce, Sc and Y are added, a specific heat treatment process and a processing process are combined, and the three rare earth elements Ce, Sc and Y have a synergistic effect, so that the performance and the electric conductivity of the cable material are balanced, and the cable material has good electric conductivity and elongation on the premise of ensuring certain tensile strength, and can meet the requirements of the cable industry.

The invention discloses a game scene management method and system, and the method comprises the steps: 1, carrying out layered loading, and dividing a map into a plurality of small regions; dynamically loading and unloading the area according to the position of the player; 2, activating as required, wherein objects in the scene are kept in a non-activated state when not needed; 3, optimizing the data flow: optimizing the data flow, and ensuring that the loading and unloading sequence of the scene data is reasonable; 4, scene dependency management: managing the dependency relationship between scenes to ensure that the dependent resources are loaded or are loaded when one scene is loaded; 5, resource optimization: optimizing the model, the texture and the animation; step 6, asynchronous loading: loading resources by using an asynchronous or background thread; and 7, illumination and shadow management: using a combination of pre-calculated illumination and real-time illumination to limit the number of dynamic light sources, and arranging the positions of the light sources. According to the invention, the game scene can be better integrally managed.

The invention relates to the technical field of polymer composite material preparation, in particular to a super-tough core-shell stereocomplex polylactic acid-based composite material and a preparation method thereof. The super-tough core-shell stereocomplex polylactic acid-based composite material is prepared from the following raw materials in parts by mass: 70 to 100 parts of polylactic acid and 0 to 30 parts of the modified elastomer in claim 4, and the dosage of the modified elastomer is not 0; the modified elastomer is prepared by melting and blending stereocomplex polylactic acid powder and a functional elastomer copolymer; the stereocomplex polylactic acid powder is prepared by reactive melt blending of PLLA and PDLA in an equal mass ratio or an equal molar ratio. The prepared super-tough core-shell stereocomplex polylactic acid-based composite material has excellent tensile strength and impact resistance, can be applied to the field of engineering plastics such as automobile parts, electronic and electric appliance shells, household products, catering tableware and the like, and has remarkable environmental protection benefits and industrial popularization value.

The invention relates to a data processing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a to-be-deployed original neural network model and a corresponding calibration data set; inputting data in the calibration data set into the original neural network model, and obtaining an activation value corresponding to each model parameter in the original neural network; based on the size of the activation value corresponding to each model parameter, performing quantization processing on the model parameters above the initial quantization retention ratio in the original neural network to obtain a quantized neural network model; and based on the performance of the quantized neural network, performing iterative adjustment on the initial quantization retention ratio until a preset requirement is met, and obtaining a target quantization retention ratio. According to the method, the model storage amount or calculation amount can be effectively reduced under the condition that the model performance is ensured.

The invention relates to a memory refreshing method and device, electronic equipment, a chip and a storage medium, and belongs to the technical field of computers. The method comprises the steps that it is determined that a memory service scene is located in a target scene, the target scene is one of at least two service scenes with different service densities, a target refreshing mode of a memory is determined according to the target scene, and the memory is refreshed based on the target refreshing mode. In the application, the refresh mode can be dynamically selected based on the current target scene of the memory, and a fixed refresh mode is no longer adopted to refresh the memory, so that the balance of performance and power consumption can be realized in business scenes with different business densities, and the overall efficiency of the memory is prompted.

This utility model discloses a novel composite heat exchange vehicle-mounted refrigerator, relating to the field of vehicle-mounted refrigeration technology. It includes a housing, a novel composite heat exchange system, and a control system. The novel composite heat exchange system includes: a main refrigeration unit, utilizing a cold source and a cold storage medium within an energy storage body for refrigeration or cold storage; and a secondary refrigeration unit, including a semiconductor refrigeration element. The cold end of the semiconductor refrigeration element is coupled to the system's airflow path or energy storage body to provide supplementary or deep refrigeration, while its hot end dissipates heat through a heat dissipation structure. The heat dissipation structure is immersed in the cold storage medium, utilizing the cold storage medium to absorb heat from the hot end. This utility model, by combining the main refrigeration unit and the secondary refrigeration unit and optimizing the thermal management of the secondary refrigeration unit, achieves stronger refrigeration capacity and can meet lower refrigeration temperature requirements.

The application provides a dynamic causal reinforcement intelligent medical hallucination detection and correction method and system, and relates to the technical field of medical hallucination detection.The method comprises the following steps: acquiring relevant data sources, and constructing a dynamic medical causal graph according to the relevant data sources; determining a to-be-detected diagnosis result, a to-be-detected diagnosis result evidence source and a to-be-detected diagnosis result confidence; acquiring actual detection data; obtaining a first medical hallucination discrimination result; determining whether the to-be-detected text needs to be corrected; in the case that the to-be-detected text needs to be corrected, processing the to-be-detected text according to a medical hallucination correction model to obtain a first training generated text; obtaining a combination loss function; obtaining a trained medical hallucination discrimination model and a trained medical hallucination correction model; and obtaining a final medical hallucination discrimination result and a final generated text.According to the application, the efficiency and accuracy of medical hallucination detection and correction can be improved.

The application discloses a large model super-long context processing method, belongs to the field of artificial intelligence and natural language processing technology, and realizes large model super-long context processing based on visual text compression and sparse attention cooperation. The implementation of the method comprises the following steps: S1, visual representation learning of text; S2, adaptive multi-granularity visual text compression; S3, dynamic sparse attention of compressed sensing; S4, hierarchical feature fusion and text reconstruction; and S5, end-to-end optimization and adaptive strategy. The application realizes the two-dimensional visualization of text and collaborative design with a sparse attention mechanism, solves the problems of high computational complexity, insufficient semantic modeling and insufficient structural information utilization of traditional methods in processing super-long sequences, significantly reduces the computational and memory overhead, realizes accurate modeling of deep semantic and global structure of long documents, and provides efficient core technical support for long text intelligent analysis, multi-round dialogue systems and other applications.