51 results about "Enhancement Technologies" patented technology

The invention relates to the technical field of aero-engines, and discloses an automatic design method for aero-engine turbine blades, which analyzes design requirements through a language large model, combines a retrieval enhancement technology and a turbine blade expert knowledge base, performs interdisciplinary coupling verification by utilizing a step-by-step reasoning template, and generates and optimizes part design schemes and parameters. And then calling parametric modeling and an AI simulation agent model to carry out multi-physics field prediction, comparing a result with a requirement, if the result meets the requirement, outputting design, otherwise, automatically iterating and adjusting parameters until the parameters reach the standard or reach the maximum number of iterations. According to the method, end-to-end intelligent closed-loop design from design requirements to design results can be realized, manual intervention is reduced, a design scheme meeting performance requirements can be quickly generated, the design efficiency can be improved, the design cost can be reduced, and the design quality can be improved.

The invention relates to the technical field of data acquisition, in particular to an AOP-based user behavior data acquisition method. The method comprises the following steps: firstly, identifying user behavior related code features by statically analyzing Java source codes of an application program, and constructing a feature template library; during a compiling period, according to a template library identification target method, embedding a point burying instruction by utilizing a byte code enhancement technology, and constructing a monitoring section framework; and during running, dynamically adjusting an execution strategy according to the running state of the application program by virtue of a dynamic rule engine. Meanwhile, an annular buffer area is adopted to asynchronously process user behavior data streams, cache data are managed in a sub-generation mode, and asynchronous processing is triggered when a threshold value is exceeded. The method also constructs an adaptive thread pool model based on a hierarchical thread pool architecture, distributes tasks according to data value weights, monitors the task backlog rate of a high-priority sub-pool, adjusts the sampling frequency or acquisition dimension when the task backlog rate exceeds a threshold value, forms adaptive closed-loop control, and effectively acquires and processes user behavior data.

The invention discloses a patch position correction method and system, relates to the technical field of biomedical engineering and digital health, and adopts a hardware-level time synchronization and motion-artifact coupling modeling technology to effectively solve the problems of time delay drift and motion interference of signal acquisition in a dynamic environment and remarkably improve the basic quality of electrocardiosignals. Secondly, by establishing a body surface reference coordinate system and a pose compensation parameter generation algorithm, the micro-displacement state of the patch can be accurately recognized, and visual position adjustment guidance is generated, so that the patch is always kept at the optimal measurement position, and the phenomenon of signal attenuation caused by poor contact is fundamentally reduced; a signal quality multi-dimensional evaluation system adapts to quality discrimination requirements in different motion states through an intelligent weighted fusion mechanism, and combines an adaptive filtering and deep learning enhancement technology, so that the morphological integrity of a QRS waveform is kept, various interference components are effectively inhibited, and the system can still keep excellent electrocardiosignals in a strenuous motion state.

The invention discloses an AI service intelligent fault tolerance and degradation method, device and equipment based on a Java bytecode enhancement technology, a medium and a product, and relates to the technical field of AI service integration. According to the method, in a class loading period of Java application starting, a loaded class is scanned to identify target methods marked with declarative fault-tolerant annotations, a bytecode operation framework is utilized to enhance the target methods so as to implant a unified fault-tolerant logic interception entry, and then in a running period, when the target methods are called, the target methods are subjected to fault-tolerant annotation processing. And triggering a fault-tolerant interceptor through the interception entrance to analyze the annotation to obtain a fault-tolerant strategy, allocating an independent fault-tolerant component instance to the fault-tolerant interceptor based on the unique identifier of the method, constructing and executing an enhanced call chain containing a fusing check and retry mechanism, and finally, if the execution of the call chain fails, executing the fault-tolerant interceptor. And if yes, a preset degradation method in the fault-tolerant strategy is automatically executed, so that complete non-intrusive decoupling of AI service fault-tolerant management and service logic can be realized, and the reliability of the system is remarkably improved.

The invention discloses a crop planting scheme generation method and device based on a large language model, equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the natural language understanding of to-be-processed agricultural information, obtaining language understanding information, and converting the language understanding information into an information vector; constructing an agricultural knowledge graph based on the to-be-processed agricultural information; fusing the information vector and the agricultural knowledge graph to obtain a fusion result, and determining a retrieval result which corresponds to the agricultural knowledge graph and comprises a target entity and entity relationship path in the agricultural knowledge graph by utilizing a knowledge retrieval enhancement technology and based on the fusion result; determining a target context based on the retrieval result and an initial context corresponding to the large language model, and generating a crop planting scheme for the target context by using the large language model to obtain a target crop planting scheme; the target crop planting scheme comprises the explanation information corresponding to the planting suggestions, so that the efficiency of generating the crop planting scheme is improved.

The invention belongs to the technical field of RAG memory enhancement, and particularly relates to a self-adaptive RAG memory enhancement method, which comprises the following steps of: 1, determining an application scene and task requirements; 2, a backbone LLM is initialized and configured; step 3, attribute mining and identification; step 4, attribute labeling granularity selection and labeling; step 5, evaluating and sorting attribute priorities; step 6, constructing memory instance embedding representation; step 7, implementing a retrieval strategy based on attributes and embedding; step 8, integrating and optimizing retrieval results; and step 9, performing experimental evaluation and continuous improvement, namely determining the specific application scene of the self-adaptive RAG memory enhancement technology, retrieving appropriate text content for recommendation according to the theme and requirements input by the user, and determining the specific requirements of the task according to different scenes. According to the method, memory information is enriched by deeply mining and marking key attributes, and meanwhile, two efficient retrieval means are adopted to ensure quick positioning of related information.

The application belongs to the field of chemical mass and heat transfer enhancement technology, and discloses a macro-quantity structured micro-scale vortex packing structure, method and application. The packing structure is a packing with a multi-scale microstructure on the surface. The microstructure causes boundary layer separation and vortex shedding when fluid flows through, generates stable micro-scale vortexes at the microstructure, and forms a macro-distributed vortex group in the packing. The vortex can cut the boundary layer, stretch and fold the fluid micro-element, and promote the periodic update of the interface, thereby improving the gas-liquid mass and heat transfer efficiency. The application significantly reduces the pressure drop. Under the same operating conditions, the pressure drop of the packing structure of the application is reduced by more than 50% compared with the conventional packing, and the pressure drop increment caused by the vortex structure is less than 15%, realizing the synergistic optimization of mass transfer enhancement and low pressure drop.

The invention relates to a layout generation method and device, equipment, a storage medium and a program product. The layout generation method comprises the following steps: acquiring a target graph and target layout precision; determining a numerical value effective standard according to the target layout precision; and based on the numerical value effective standard and the geometric information of the target graph, generating a periodic arrangement graph of the target graph, and obtaining a target layout. According to the method and the device, the periodic arrangement graph with the target layout precision can be generated, so that advanced resolution enhancement technologies such as reverse photoetching technology correction and the like can be conveniently implemented on the full-chip layout.

This invention discloses a method, apparatus, device, and medium for implementing an intelligent quality valve in the requirement submission process, relating to the field of requirement process quality review in fintech. The method includes: acquiring a requirement description and uploading it to a created large-scale intelligent platform; expanding the document based on comparison with existing similar requirements and performing standardization processing to fill in missing elements, generating a requirement document; acquiring voice data from requirement review meetings and uploading it to the large-scale intelligent platform; analyzing the voice data to generate a requirement review meeting summary and submitting the summary to a DevOps platform; creating a requirement review model, which uses retrieval enhancement technology to acquire private domain data in real time to review the requirement document currently awaiting review. This application enables automated review operations, significantly reducing the burden of manual review and improving the accuracy and efficiency of the review process.

The invention relates to the technical field of PCB (Printed Circuit Board) detection, in particular to a method and a system for detecting the edge deviation of the appearance of a PCB based on a welding component. The polarization characteristic pattern is generated by acquiring the multi-spectral original image and the polarization state parameter of the to-be-detected area of the PCB, the reflection exposure abnormal area and the reflection exposure normal area can be effectively positioned, the fusion weight is automatically adjusted according to the spectral response characteristic of the signal area through the self-adaptive spectrum fusion enhancement technology, and the detection accuracy is improved. The problem that in a traditional method, image enhancement is not fine enough is solved, the adaptive fusion enhancement technology enables the multispectral image to better conform to the actual situation during enhancement, more detail information is reserved, key features can be accurately extracted by generating a semantic segmentation map and combining three-dimensional standard coordinate information, and the image enhancement accuracy is improved. The detection precision and reliability of the welding component and the appearance edge of the PCB are effectively improved, the deviation can be accurately positioned and quantified, and the quality control level in the production process of the PCB is improved.

The invention provides a submarine cable defect detection method based on time-space texture cooperative enhancement, and aims to solve the problems of insufficient defect feature extraction in a low signal-to-noise ratio environment, high false detection rate caused by dynamic background interference and lack of interpretable basis for detection results in the existing submarine cable detection technology. In order to solve the problems, a space-time texture collaborative enhancement mechanism is introduced, and defect detection in a complex environment is realized through multi-modal feature fusion. Specifically, in order to solve the problems of dynamic background interference and weak target features in submarine cable detection, the invention provides a space-time difference enhancement technology, designs a composite texture analysis architecture, extracts linear structure features and local statistical distribution features of a cable, and constructs a dual verification mechanism of defect characterization; a dynamic feature weight distribution strategy is further provided, and the fusion proportion of the spatial-temporal features and the texture features is adjusted in a self-adaptive mode according to the environment complexity. Meanwhile, the method provides reliable technical support for underwater operation and maintenance decisions.

The application relates to the technical field of aero-engines and discloses an automatic design method for a turbine blade of an aero-engine, which comprises the following steps: analyzing design requirements through a language large model, combining retrieval enhancement technology with a turbine blade expert knowledge base, utilizing step-by-step reasoning templates to perform cross-disciplinary coupling verification, generating and optimizing a component design scheme and parameters, subsequently calling a parameterized modeling and an AI simulation agent model to perform multi-physical field prediction, comparing the result with requirements, outputting a design if the requirements are met, and otherwise automatically iteratively adjusting parameters until the requirements are met or the maximum number of iterations is reached. The application can realize end-to-end intelligent closed-loop design from design requirements to design results, reduce manual intervention, quickly generate a design scheme meeting performance requirements, improve design efficiency and reduce design cost, and improve design quality.

This invention discloses a reinforcement learning-based underwater image enhancement training optimization method, belonging to the field of underwater image enhancement technology. Addressing the problems existing in current technologies, this method treats low-frequency color correction and high-frequency texture reconstruction in underwater image enhancement as two enhancement objectives with different optimization focuses. It utilizes a reinforcement learning mechanism to dynamically adjust the training strategies for both, solving the problem that existing fixed training strategies are difficult to adapt to changes in training states. A frequency-domain perceptual state representation is constructed by combining low-pass filtering and residual decomposition to explicitly describe the relative optimization state between low-frequency color restoration and high-frequency texture restoration during the current training process, providing a decision-making basis for the policy controller. An adaptive training optimization mechanism based on frequency-domain state feedback is built, automatically completing policy scheduling according to the actual restoration state during training, reducing the reliance on manually designed training strategies and improving the adaptability, generalization, and scalability of the method in complex underwater degradation scenarios.

This invention provides a turnout design method and system, relating to the field of key characteristic enhancement technology for urban rail transit turnouts. The method includes: obtaining the target train speed, existing turnout dimensions, turnout frog angle, and the range of values ​​for the turnout guide curve radius; calculating the design target speed based on the target speed; back-calculating the lateral speed based on the existing turnout dimensions and iteratively lengthening and adjusting to obtain a first parameter; optimizing the radius based on the range of values ​​for the turnout guide curve radius to obtain the optimal guide curve radius; determining the half-section and switch rail impact angle based on the optimal guide curve radius, the turnout front length, and the turnout rear length to obtain the optimal parameter combination; and verifying the optimal parameter combination to obtain the final turnout alignment design scheme. The method and system described in this invention improve the lateral throughput speed of turnouts, meet the turnout speed requirements of signaling professionals, and enhance the turnout performance of existing turnouts.

This invention discloses a device for enhancing the stability of a power line joint, relating to the field of power line joint enhancement technology. It includes a base, a connecting rod, and a terminal block. The connecting rod, measuring 200mm x 26mm x 10mm, is located on one side of the base. The terminal block is located on the side of the connecting rod away from the base. The device is gently slipped onto the cable, and then the cable core is carefully inserted into the cable tray, ensuring complete alignment for optimal electrical connection. Next, the terminal block is accurately placed above the core, and the M8 nut and M8 locking bolt are tightened to securely hold the core in place, preventing loosening or detachment. The device limits significant cable sway at the joint, reducing stress concentration and preventing breakage, strand breakage, and unraveling due to stress concentration.

This application discloses an automated detection method for pin defects based on 3D imaging and deep learning, including the following steps: (1) performing structured light 3D adaptive anti-overexposure imaging on the pin to obtain effective point cloud at the micron level of the pin tip; (2) performing local repair on the pin tip defect cloud obtained in step (1), calculating the normal vector and curvature features of the repaired point cloud, fusing 3D point cloud features and 2D texture features to obtain multimodal fusion features of the pin tip; (3) inputting the multimodal fusion features obtained in step (2) into a lightweight twin feature matching network based on few-sample deep learning, combining meta-learning transfer enhancement technology to complete the identification and classification of pin tip micro-breakage and micro-bending defects, and outputting defect location and defect degree parameters; (4) based on normal vector constraints and pseudo-defect feature comparison, performing double filtering on the suspected defects detected in step (3), removing pseudo-defects to obtain the real defect detection results.

The invention relates to the technical field of construction of detection models, in particular to a method for constructing a light-weight small target detection model, which comprises the following steps of: constructing an HS decoupling head structure by using a double-branch collaborative mechanism aiming at a detection head part of the light-weight small target detection model, and extracting a deep channel attention mechanism and multi-scale texture features to obtain a light-weight small target detection model; optimizing spatial precision according to a prediction result, and determining a lightweight small target detection model based on a depth lightweight reconstruction mechanism; according to the invention, by adopting the HS decoupling head technology based on the double-branch cooperation mechanism, the technical effect of significantly improving the small target detection precision is realized; by adopting a depth lightweight trunk reconstruction technology based on MobileNetV3, the technical effect of maintaining strong feature extraction capability under extremely low resource consumption is realized; by adopting the neck network depth separable convolution comprehensive replacement and GSConv enhancement technology, the technical effects of greatly reducing the model calculation complexity and keeping the feature fusion capability are realized.

The invention relates to the technical field of cultural science and technology and artificial intelligence crossing, and discloses an exhibit content generation and vision enhancement system based on a multi-modal generative AI. Comprising the following modules: an exhibit knowledge system construction module, a multi-source data fusion and scene demand analysis module, an AI scenarized multi-modal content generation module, an AR / VR immersive visual enhancement module, a scenarized interaction and knowledge transfer verification module, and a knowledge system and content dynamic iteration module. All the modules are linked in sequence to realize scenarized content generation and immersive visual presentation of the exhibit; according to the invention, by constructing a structured exhibit knowledge system, the limitation of fragmentation of traditional exhibit content is broken, and systematic transmission and extension of knowledge are realized; based on the multi-mode generation and immersive visual enhancement technology, abstract exhibit knowledge is converted into visual and perceptible texts, images, 3D scenes and animation content, and a user can understand culture behind an exhibit more deeply in cooperation with a multi-terminal adaptation and diversified interaction mode.

This invention discloses a patch position correction method and system, relating to the fields of biomedical engineering and digital health technology. The invention employs hardware-level time synchronization and motion-artifact coupling modeling technology, effectively solving the problems of time delay drift and motion interference in signal acquisition under dynamic environments, significantly improving the basic quality of ECG signals. Secondly, by establishing a body surface reference coordinate system and a pose compensation parameter generation algorithm, the micro-displacement state of the patch can be accurately identified and intuitive position adjustment guidance can be generated, ensuring the patch remains in the optimal measurement position and fundamentally reducing signal attenuation due to poor contact. The multi-dimensional signal quality evaluation system, through an intelligent weighted fusion mechanism, adapts to the quality discrimination requirements under different motion states. Combined with adaptive filtering and deep learning enhancement technology, it maintains the morphological integrity of the QRS waveform while effectively suppressing various interference components, enabling the system to maintain excellent ECG signals even under strenuous exercise.

The application discloses a BAS system attack load automatic generation method and system based on a large language model, and belongs to the technical field of network security. In order to solve the problems that attack load generation depends on manual work, has poor compliance and is difficult to break through dynamic defense, the BAS attack instruction or environment fingerprint is received and modeled as a decision process; search enhancement technology is used to fuse safety knowledge to form prompt information; an initial load is generated by a strategy network under the constraint of a predefined syntax rule; a reward model is trained based on target environment feedback to evaluate bypassing capability; and the strategy network is updated by using a reinforcement learning mechanism and knowledge dynamic backfilling is realized. The application can significantly improve the simulation degree and bypass success rate of the BAS system, realize automatic iteration of attack strategies and mining of new vulnerability variants.

The application discloses a dynamic dotting method and device, wherein the method comprises the following steps: acquiring description information of a setting class file, checking the description information of the setting class file, and acquiring a setting method and the setting class file according to a method parameter, the name of the setting method and the name of the setting class file; intercepting and preprocessing the setting method to set the setting method to a state that can be dotted; calling an interface on a front position or a rear position of the setting method, and reporting dotting data to fill the dotting data to the front position or the rear position of the setting method; redefining the setting class file to obtain an enhanced setting class file; and sending the enhanced setting class file to a setting agent end to replace the setting class file with the enhanced setting class file through bytecode enhancement technology to realize dynamic dotting of a program; and the application implants dotting data when the program runs, thereby improving dotting efficiency and dotting accuracy of the program.

The application discloses a shrimp-rice co-cultivation area hydrological simulation method based on deep learning, relates to the technical field of shrimp-rice co-cultivation, and aims to solve the problem of poor hydrological simulation effect.The application can timely discover and cope with abnormal conditions by monitoring hydrological changes, crop growth conditions and weather conditions in real time, ensure the flexibility and adaptability of the scheduling scheme, and the setting of the simulation scene includes scene definition, initial state setting and boundary condition setting, so that the simulation process can flexibly adapt to different geographical, climatic and ecological conditions, the adaptability and accuracy of the simulation are improved, the accuracy and reliability of the model can be further improved through dynamic optimization, the model can more truly reflect the actual situation, and various model enhancement technologies such as differential equation embedding, boundary condition injection and adversarial training can improve the performance of the model from multiple dimensions.

The invention relates to a point cloud thermal prediction method for a large-scale unstructured 2.5 D integrated circuit. The network comprises three core modules: an adaptive multi-path coupling diffusion module (AMD), a wavelet-based fine-grained recovery module (WFR) and a thermal sensing material expert hybrid adapter (TA-MoME). The AMD module adaptively learns interaction of a thermal diffusion path through an attention mechanism based on serialized gating; the WFR module recovers fine-grained thermal gradient features by using a high-frequency wavelet domain enhancement technology; and the TA-MoME adapter dynamically routes the material specific expert to adapt to the heat distribution difference between the heterogeneous materials. Particularly, the framework also introduces a continuous learning mechanism with efficient parameters to ensure that when the model adapts to a new geometric scale, catastrophic forgetting can be prevented, and cross-scale adaptation and mobility can be realized. Experimental results show that in 80K-scale point cloud prediction, compared with an existing FSA-Heat method, the average precision index is remarkably improved (errors are reduced by about 78% or above), compared with commercial COMSOL, acceleration of 147 times is achieved, and the generalization ability for larger-scale nodes (0.4 M) and zero samples without geometrical shapes is shown.

The application discloses an industry-specific customized model generation method based on a large model, which comprises the following key steps: first, collecting structured and semi-structured multi-modal data containing key entities, concepts and relationships in the target industry; then, constructing a knowledge graph through a deep learning method; in the model construction stage, pre-training of a large language model is selected, and feature engineering and personalized adjustment of the model architecture are carried out according to the characteristics of the industry; on the basis of the constructed knowledge graph, the retrieval enhancement technology based on the knowledge graph is adopted to optimize the answer content of the large language model and inject the knowledge of the target industry field; finally, the reinforcement learning algorithm is used to optimize the model output by training the reward model. The application realizes the professional customization of the industry-specific model, significantly improves the question and answer ability of the large language model in the target industry field, and ensures the high accuracy and high relevance of the model when answering the problems related to the target industry.

The invention provides a data replication method and device, a medium and electronic equipment. The method comprises the steps of obtaining attribute replication requirements of a source object type and a target object type; searching a copy tool class corresponding to the copy operation from the source object type to the target object type in a global cache; under the condition that the copy tool class is not found, dynamically generating the copy tool class by utilizing a byte code enhancement technology, and storing the copy tool class in a cache; and calling a copying method in the generated copying tool class, and executing an attribute copying operation from the source object type to the target object type according to an attribute copying demand. Therefore, the problem that according to an existing scheme, manual hard coding needs to be carried out according to different business data objects, so that the flexibility of data copying is poor is solved.

The application relates to the technical field of program enhancement, and provides a function dynamic enhancement method based on an application program, which comprises the following steps: when a target program runs, whether a preset key event exists is listened to; when the preset key event exists, event trigger information is acquired; according to the event trigger information, a local plug-in is called through a JVM Tool Interface; and according to the local plug-in, original byte code is dynamically adjusted. The application can endow the target application program with the ability of dynamically adding new functions or modifying existing functions without modifying the source code of the target application program, and can realize independent development, independent operation and flexible reuse when applied to the application program of the java language.

The invention discloses an edge intelligent application cooperative processing system and method based on a video enhancement technology. The cooperative processing system comprises a video enhancement module, a task fragmentation module and a dynamic scheduling module. The video enhancement module is used for enhancing the initial video; the task fragmentation module is used for calculating the content complexity of the enhanced video, dividing calculation tasks and constructing a calculation task calculation power matrix; and the dynamic scheduling module is used for generating a task allocation strategy and carrying out computing power resource allocation on each enhanced video according to the task allocation strategy. According to the cooperative processing system and method, computing power resources can be reasonably distributed according to the complexity of the video content, so that the video processing efficiency is effectively improved.

The invention relates to a log desensitization method, system and device capable of dynamically updating rules and a program product. The system comprises a configuration management center and a log desensitization agent. And the configuration management center is used for centrally storing and managing the desensitization rules and actively sending update notifications to the associated agents after the rules are modified. The log desensitization agent is packaged into a Java Agent, and parameter loading is started through the JVM of the target service application. When the application is started, the agent implants desensitization logic into the log output class through a byte code enhancement technology; the method comprises the following steps: acquiring an initial rule from a configuration management center and caching the initial rule in an own memory; and when a rule updating notification is received, dynamically pulling a new rule and updating the memory, so that the rule takes effect in real time. And when the business application runs, the agent carries out desensitization processing on the log output according to the latest rule in the memory.

This invention discloses an audio enhancement system and method for underground broadcasting systems, relating to the field of audio enhancement technology. The system acquires noise signals and original audio signals from a mine roadway, processes them through standardization, encodes and transmits them, and then decodes them to obtain a standardized noise signal. The standardized noise signal undergoes frequency band amplitude shaping and time delay compensation, followed by a phase inversion operation to generate a reverse noise signal. Based on the original audio signal and a roadway sound reflection model, the reverberation component within the roadway is predicted. After phase inversion, a reverse reverberation signal is generated, and a reverberation gating coefficient is introduced for weighted control to obtain the actual reverse reverberation signal. The original audio signal, the reverse noise signal, and the actual reverse reverberation signal are then superimposed to obtain the final audio signal. Utilizing the principle of destructive interference, the system effectively reduces the impact of environmental noise and reverberation.

The invention discloses a real-time video QoE prediction method and system based on wireless network parameters, and the method comprises the steps: connecting an OAI signal through terminal equipment, collecting a QoE parameter index in a video playing process from a video from WebRTC, and obtaining a corresponding wireless network state parameter index as original data; a training set is generated through preprocessing, and the training set is used for training a constructed two-stage QoE perception and prediction model based on a Transform model after data enhancement processing; and in the online stage, inputting real-time data into the pre-trained two-stage QoE perception and prediction model based on the Transform model to carry out QoE perception prediction. The adaptive capacity of the model to different network conditions and video content changes is effectively improved through a robust data enhancement technology, and the generalization and robustness of the model are enhanced. Real-time QoE prediction is completed in a short time interval through two-stage QoE perception and prediction based on a Transform model, and the requirements of a real-time XR video communication system are met.