3298 results about "Network layer" patented technology

In example embodiments, an industrial technology stack for an industrial environment includes a set of computational resources and a set of layers executed by the set of computational resources, the set of layers including a governance layer, an enterprise layer, an offering layer, a transaction layer, an operations layer, a network layer, a data layer, and a resource layer. In example embodiments, the industrial technology stack may include one or more artificial intelligence models for implementing one or more components of one or more layers of the set of layers.

An artificial intelligence driven system of systems may include a layered architecture for providing transaction support to various types of enterprises. A governance layer implements automated governance and policy enforcement through specialized governance modules utilizing generative AI technology. An enterprise layer supports enterprise functions by integrating management and control platforms with digital infrastructure. An offering layer creates and manages system offerings via content generation, personalization, and smart product modules. A transactions layer enables automated transaction orchestration through API integration, execution, and fulfillment modules. An operations layer manages AI systems through generation, training, verification and orchestration modules. A network layer provides adaptive networking capabilities through routing, protocol selection and communication modules. A data layer processes fused data from multiple sources using machine learning and AI systems. A resource layer manages computing, storage, and other resources through specialized resource modules.

A small-size vehicle detection deep learning model based on feature fusion of multi-scale modules is provided, which solves the problem of small-size vehicle image detection. The model includes a Backbone network, a Neck layer and a Head network, wherein a C2f_DCNv3 module based on the combination of deformable convolution v3 (DCNv3) and a cross stage feature fusion (C2f) module and an SPPF_LSKA module based on the combination of a spatial pyramid pooling fast (SPPF) layer and a large separable kernel attention (LSKA) module are introduced into the Backbone network; a C2f_SCConv module based on the combination of spatial and channel reconstruction convolution (SCConv) and a C2f module is introduced into the Neck layer; and a multi-scale kernel detection (MSK_Detect) module is introduced into the Head network.

The invention discloses a network threat multi-modal detection method based on a large model, and belongs to the technical field of network security, and the method comprises the steps: collecting three types of heterogeneous data of NetFlow flow of a network layer, a system call chain sequence of a host layer and a protocol load of an application layer, and carrying out the desensitization processing and feature coding to generate a unified tensor format; the method comprises the following steps: through network security threat intelligence and MITRE ATTamp; performing supervision fine tuning on the large model by using a CK attack chain sample, and constructing a network threat identification special model; cross-device behavior characteristics are extracted through a model self-attention mechanism, and a dynamic behavior map is constructed; and finally, comprehensively evaluating an attack mode matching degree, a node vulnerability mean value, historical alarm association and an attack path risk by adopting a weighted fusion algorithm, and triggering a high-confidence alarm when a comprehensive score exceeds 0.8. According to the method, through multi-modal data fusion and dynamic graph analysis, the detection precision and response efficiency of the complex attack chain are improved.

The invention discloses a mechanical equipment state monitoring method and system based on multiple sensors, and the method comprises the five core steps: multi-modal data collection and preprocessing, dynamic feature fusion, adaptive threshold diagnosis, digital twin fault tracing and predictive maintenance decision. All-domain coverage of equipment is realized through a three-layer sensor network architecture, the problems of data synchronization and interference resistance are solved by utilizing a temperature and vibration integrated sensor, deep fusion and anomaly detection of multi-source data are realized in combination with an attention mechanism, a Gaussian mixture model, a three-dimensional convolutional neural network and the like, and finally a precise maintenance strategy is generated through digital twinning and reinforcement learning. The multi-sensor-based mechanical equipment state monitoring system comprises a sensor network layer, an edge computing layer, a cloud platform layer and a man-machine interaction layer, supports federated learning to protect data privacy, improves real-time diagnosis capability through edge-cloud collaboration, and enhances a reality interface to realize intelligent operation and maintenance interaction.

The invention discloses a supply chain full-process traceability system based on digital twinning and block chains, and belongs to the technical field of digital twinning and block chains, and the system comprises a data collection layer which obtains original data from each link in real time, and carries out the preprocessing of the original data through an edge computing device; the twinborn model building layer is used for building digital twinborn models corresponding to all links of the supply chain respectively and generating link-level optimization parameters through simulation analysis; the block chain network layer is used for verifying and storing shared data and key traceability information among the digital twin models; the collaborative optimization module is used for dynamically adjusting a cross-link collaborative strategy based on a multi-agent reinforcement learning algorithm; and the data traceability layer responds to an external traceability query request and generates a full-link visual traceability map. By means of the collaborative algorithm based on multi-agent reinforcement learning, automation and intellectualization of information interaction among all link models are guaranteed, and personalized and multi-terminal traceability service can be provided for users.

The present application relates to the technical field of batteries, and discloses a thermal runaway risk prediction method and apparatus, a device, and a storage medium. The method comprises: processing, by at least two neural network layers in a target prediction model, thermal runaway risk parameters layer by layer, wherein the target prediction model is obtained by pre-training on the basis of state vectors of a plurality of time steps, so that the memory capability of the model for past state sequences can be enhanced, and thus the model can better learn the dynamic characteristics of an energy storage battery system, and captures a complex temporal association relationship among multiple variables, thereby improving the accuracy of thermal runaway risk prediction and reducing the safety risk of the energy storage battery system.

A system and method for neural network weight compression with intrusion detection capabilities that optimizes model storage and transmission while providing security. The system analyzes weight characteristics to identify statistical properties within different neural network layers, generates optimized encoding schemes based on the analysis, and creates reference distributions for security verification. The compression process employs a multi-resolution approach that produces a progressive representation with base and enhancement layers, enabling flexible deployment across diverse computing environments. Security markers and statistical fingerprints can be embedded throughout the encoded representation, allowing for detection of unauthorized modifications during transmission or deployment. The system monitors encoded weight streams, measures distribution divergence against reference baselines, and generates alerts when statistical anomalies indicate potential tampering. This approach achieves superior compression ratios while maintaining model performance and providing robust protection against increasingly sophisticated attacks targeting neural network weights.

The invention relates to the technical field of bridge detection, in particular to a bridge defect automatic identification and inspection method based on deep learning, which comprises the following steps: S1, multi-modal data acquisition: acquiring multi-modal data of the surface and internal structure of a bridge; s2, multi-modal data preprocessing: generating a space-time aligned enhanced data set; s3, feature fusion and extraction: generating a fusion feature spectrum; s4, adaptively optimizing the characteristic spectrum: adjusting the connection weight between network layers according to the real-time environment parameters; s5, defect positioning and type identification: positioning space coordinates of defects and synchronously outputting probability distribution of defect types; and S6, defect analysis and report generation: generating an inspection report including defect evolution trend analysis based on the result of defect positioning and type identification. According to the invention, the bridge inspection efficiency is improved, the labor cost is reduced, and the intelligent upgrading of the bridge management field is further promoted.

The invention discloses a medical image super-resolution reconstruction method based on multi-level attention guidance, and the method comprises the following steps: S10, constructing a deep learning network model based on a generative adversarial network architecture, which comprises a generator and a discriminator; the generator is based on an improved U-Net architecture, a hierarchical attention module and a dual-path feature processing module are configured in an encoder and a decoder of the generator, the hierarchical attention module adopts different attention strategies according to network levels to consider structure and texture, and the dual-path feature processing module separates and processes low-frequency and high-frequency information; the generator further comprises a multi-level feature fusion module for integrating the multi-scale features of the decoder, and an attention guide up-sampling module for final enhancement and dimension raising. The discriminator adopts a spectrum normalization U-Net architecture and uses multi-scale features for matching; s20, training the network model by adopting a composite loss function comprising pixels, adversarial, perception and total variation loss; and S30, inputting the low-resolution image into the trained model, and outputting a high-resolution image. According to the method, through deep fusion of multi-level attention and multi-scale feature processing, the image restoration quality can be remarkably improved, the texture detail definition can be enhanced, the anatomical structure accuracy can be ensured, and the noise robustness can be improved.

The invention relates to the technical field of part detection, and discloses a part defect automatic detection method based on machine vision. The method comprises the following steps: firstly, acquiring three-dimensional geometric parameters of a target part, and matching a historical defect sample set in a visual sample library according to the three-dimensional geometric parameters; performing defect type clustering division on the set to obtain a plurality of defect type subsets; processing the subsets one by one to execute multispectral feature extraction, and obtaining a reference detection area and a defect diffusion range parameter corresponding to each defect category; utilizing defect diffusion range parameters to configure the scanning step length of the multi-stage detection network layer, and generating a plurality of scale defect feature maps; and finally, performing cross-level association fusion on the feature maps to generate a fusion defect feature map, and outputting the fusion defect feature map as a final detection result. According to the method, three-dimensional geometric features and historical data are combined, and the comprehensiveness and accuracy of part defect detection are improved through multispectral extraction, adaptive scanning and feature fusion.

The invention discloses a metal defect identification method based on YOLOv8, and the method comprises the steps: carrying out the semi-automatic labeling of a collected sample data set through a LabelMe semi-automatic labeling tool, and carrying out the detection feedback through an existing YOLOv8 model. Carrying out manual labeling correction on a labeling sample with a relatively poor detection result through a LabelImage labeling tool; replacing a feature fusion part in an original YOLOv8 network with a combination of a weighted splicing operation, a feature enhancement layer and a gating mechanism based on a bidirectional feature pyramid network idea, and performing adaptive fusion on features of different scales; inserting a GCNECA module into the last layer of the backbone network of the original YOLOV8 network; the method comprises the following steps of: inserting a TransformerSE module behind a P3 layer in a neck network layer of an original YOLOV8 network; a MultiScaleLSTMSimAM module is inserted behind a P5 layer in a neck network layer of an original YOLOV8 network, so that the detection effect of large target metal defects is enhanced. According to the method, the metal defect identification capability in a natural scene is enhanced, and higher robustness, accuracy and efficiency are achieved.

An electronic circuit system implementing and executing machine learning inference engines. While ML inference engines are based on (architectures and parameters defined by) configured, trained and tuned machine learning models, our design has the novel ability to support data driven, on-the-fly-reconfigured model runs. Reconfiguration and tuning operations include dynamic computational graph modifications, define-by-run alterations, changes to network depth (number of layers) and width (neurons per layer), and adjustments to weights, biases, plus activation function parameters. Neural networks supported include Feed-Forward, RNN, CNN, and Hopfield architectures, plus Ensemble, Federated, Cooperating, Adversarial, and Swarm collections. Decision Trees and Forests are also supported, as are more esoteric approaches such as ART and KAN. Our invention is capable of running both standalone and cooperatively, the cooperative processing being local and / or remote / cloud based, interfacing with telemetry applications to feed data, and machine learning software to feed new or updated models.

The invention discloses a storage optimization method and device, electronic equipment and a storage medium, and relates to the technical field of data processing.According to the storage optimization method and device, the electronic equipment and the storage medium, the occupied space of caches is compressed by adopting retention and selection strategies, and when the caches are allocated to each network layer, the cache allocation efficiency is improved. According to the method, key value caches from a low layer to a high layer are distributed in a gradually decreasing mode, so that keyword elements are prevented from being omitted in the low layer, unimportant lexical elements are prevented from being stored in the high layer, and then the target attention value is obtained by utilizing the attention focusing degree of the input lexical elements and the cached lexical elements. The keyword elements which have the most influence on the input lexical elements are selected from the cache lexical elements to be cached, then representative representative lexical elements are selected from the remaining lexical elements to be stored, and only the lexical elements which have the influence on the model processing data can be stored. And the technical effect of compressing the occupied space of the KV cache so as to provide the capability of processing more requests is achieved.

The invention discloses a land-sea-air-space holographic sensing and collaborative decision-making system based on multi-mode edge intelligence. The system comprises a sensing layer which is used for carrying out three-dimensional monitoring on a ship driving process, the sea, the sky, a shore base and a space formed by the ship driving process, the sea, the sky, the shore base and the sea, the sky and the shore base; the network layer is used for constructing a global communication basis, preprocessing monitoring data, judging whether the ship has an abnormal condition or not, and realizing data transmission of the ocean, the sky, a shore base and a space formed by the ocean, the sky and the shore base; and the application layer is used for judging whether an abnormal condition exists in the ship driving process or not based on the data of the ocean, the sky, the shore base and the space formed by the ocean, the sky and the shore base transmitted by the network layer, performing collision early warning decision based on the abnormal condition, and performing dynamic channel capacity prediction. And navigation guidance and route tracking in extreme weather are carried out based on environmental risk early warning, underwater obstacle collision is avoided, and early warning information and decision information are fed back to a sensing layer.

The invention relates to a method for realizing power grid hidden fault diagnosis processing based on a graph convolutional network. The method comprises the following steps of performing data acquisition and preprocessing; constructing a power grid topological graph; a GC-LSTM model is constructed; carrying out model training and inference; and fault classification and positioning are carried out. The method and the device for realizing power grid hidden fault diagnosis processing based on the graph convolutional network, the processor and the computer readable storage medium are high in efficiency and accuracy: power grid topological structure information is extracted by utilizing the GCN, and time sequence dynamic data is processed by combining with the LSTM, so that high-precision identification and positioning of hidden faults are realized. The expandability is high, the number of network layers, input characteristics and adjacent matrixes can be flexibly expanded or cut, and the method is suitable for power grids and data sampling frequencies of different scales. The self-adaptive capacity is high, static equipment attributes and real-time dynamic electrical quantity data are fused, and good diagnosis performance can be kept for the conditions of power grid topology change, load change, equipment aging and the like.

The invention discloses a cross-domain collaborative interface dynamic configuration method and device and a computer program product. According to the cross-domain collaborative interface dynamic configuration method, service flow priorities are determined by analyzing service layer data, and an initial interface configuration scheme is generated in combination with network state information. And when the initial scheme does not meet the transmission requirements, optimizing the interface parameters by adopting a dynamic programming algorithm, generating a new configuration scheme, and issuing the new configuration scheme to the target network management system. In the service switching process, network performance is monitored in real time, interface parameters are dynamically optimized and updated to a network management system, and intelligent management of service flow and efficient utilization of network resources are achieved. The method can adapt to different manufacturer devices and network levels, guarantees the service quality of cross-domain services, and improves the operation efficiency and reliability of an electric power communication network.

The invention provides a target recognition model reasoning optimization method and device, and the method comprises the steps: firstly carrying out the structural analysis and sensitivity evaluation of a pre-training model, extracting the structural features of each network layer, activating the distribution features, carrying out the quantitative sensitivity scoring, and constructing a data set reflecting the hierarchical features and fault-tolerant capability; and querying a quantitative configuration knowledge base based on the data set to generate a heterogeneous quantitative strategy. Layered low-bit quantization is executed according to the strategy, and a layered weighted loss function is introduced to carry out quantization perception training, so that precision loss caused by bit width compression is effectively compensated. According to the method, through hierarchical heterogeneous quantification, the model recognition precision is preserved to the maximum extent while high compression ratio and reasoning acceleration are achieved, and particularly, the performance of a high-sensitivity layer is protected. The generated heterogeneous quantitative model remarkably reduces memory occupation and power consumption, is suitable for an edge hardware platform with limited resources, forms a set of complete automatic process from analysis and configuration to training compensation, and has good universality and engineering practical value.

The invention discloses a remote sensing image multi-class target detection method based on heterogeneous attention fusion modeling, solves the problem that distribution of different types of features of a multi-scale target in a network hierarchy is different, and designs a multi-head self-attention and cross attention collaborative fusion architecture. A bottom layer feature enhancement mechanism under high-level semantic guidance is constructed, and multistage feature deep fusion is realized through bidirectional attention interaction and residual connection; the invention provides a target refinement loss function based on SAM and a class activation graph. In order to solve the problem of complex background interference in a remote sensing image, an SAM segmentation mask and class activation diagram combined constraint is introduced in a training stage, a model is guided to learn finer features in a real target area, irrelevant features are eliminated, and therefore more accurate target positioning is achieved, complex background interference is relieved, and the detection effect of the model is improved.

A method is provided for processing data in a neural network system. The method includes receiving input data; processing the input data through a first set of neural network layers configured to perform data processing using MatMul-free techniques to produce intermediate data; further processing the intermediate data through a second set of neural network layers configured to simulate spiking neural network (SNN) functionalities using MatMul-free techniques; and outputting a result based on the processed data from the second set of neural network layers.

The embodiment of the invention provides a charging service real-time monitoring method and device based on a service probe, and the method and device achieve the precise construction of a service event chain through constructing a multi-layer data analysis mechanism and integrating the data of an application layer, a session layer and a network layer. And designing an exception recognition strategy based on multi-dimensional feature fusion, and establishing a neural network classifier to perform exception event analysis in combination with hardware state features and scene features. A real-time monitoring engine is introduced, and dynamic early warning is carried out on business abnormity through hierarchical analysis and correlation analysis. According to the method, the defects of the traditional technology in the aspects of data analysis, feature analysis, real-time monitoring and the like are effectively overcome, and the intelligent level and the early warning effect of charging service monitoring are remarkably improved.

The invention discloses an underwater target recognition method based on an improved YOLOv8 algorithm, and the method comprises the steps: inputting an underwater image into an improved YOLOv8n model for underwater target detection, and obtaining an output underwater target recognition result. The method has the advantages that the convolution blocks of the P5 layer of the backbone network and the last layer of the neck network adopt DSConv, so that the network complexity is reduced, and the reasoning speed is increased; a fourth C2f module of the backbone network adopts a C2f DiRMB module in which an inverted residual attention mechanism and dual-channel convolution are introduced, so that the capability of capturing key global information of the network is enhanced, training parameters are reduced, and the understanding of a complex scene is improved; and finally, a small target detection head for improving the small target detection capability is additionally arranged in the head network. According to the underwater target identification method, the mAP (at) is 0.5%, the mAP (at) is 0.5-0.95%, and the accuracy and the recall rate are respectively improved by 0.5%, 0.8%, 0.5% and 1.0%.

The invention belongs to the technical field of new-generation information, and particularly relates to a power plant metal supervision entity relationship extraction method based on dual coding, which comprises the following steps: multi-modal data preprocessing: collecting and cleaning data, and carrying out data labeling; constructing a dual coding joint learning model: designing a network layer architecture, and training the dual coding joint learning model; constructing and querying a dynamic knowledge graph: extracting a model to generate a triple, and performing time sequence evolution analysis, causal reasoning interface and dynamic updating; and incremental knowledge updating and dynamic model optimization: an incremental learning and feedback module forms a bidirectional closed loop between the knowledge graph and the relationship extraction model, and continuous evolution of the system is ensured through dynamic knowledge updating and adaptive model optimization. According to the method, through bidirectional feature modeling of dual paths, collaborative representation optimization between entities and relationships is realized while context information is captured, so that the requirements of complicated data types and diversified semantic associations in an engineering scene are met.

The invention belongs to the field of edge computing, and relates to a model deployment method based on pruning compression in an edge device, which comprises the following steps of: constructing a joint constraint condition set by analyzing hardware architecture configuration of a target edge device, and determining an upper limit of a compression ratio of each network layer of a to-be-deployed model in combination with an interlayer dependency relationship of the to-be-deployed model; the method comprises the following steps: performing hybrid pruning operation on a to-be-deployed model, reserving a redundant weight block on a key network layer, compiling and deploying the pruned model, sensing environment operation interference characteristics in a running period after the model is deployed in real time, and judging whether to trigger an online fine tuning mechanism to update parameters of the redundant weight block. And the pruning boundary parameters are iteratively optimized based on the fine tuning performance feedback data, so that the risk of model collapse caused by a static pruning strategy in the prior art is avoided, continuous self-adaption and performance optimization of the model on the resource-constrained edge device are realized by reserving redundant weight blocks, and the deployment flexibility of the edge device model is effectively improved.

The invention discloses a vehicle communication fault source determination method and related equipment, and relates to the technical field of vehicle communication, the method comprises the following steps: obtaining communication operation information of a plurality of processing levels in a vehicle communication system, the plurality of processing levels comprising a hardware layer, a middle layer, a security layer and a network layer; based on a preset log format, recording the communication operation information, and generating an encrypted log; and performing fault root cause analysis on the encrypted log through a preset fault tree model to obtain a target fault source. Through multi-level log acquisition and fault tree analysis, in combination with a structured format and encrypted storage, vehicle communication fault diagnosis with accurate and efficient fault positioning, data traceability, safety and credibility can be realized.

The invention discloses a multi-modal attack identification method fusing BMama and difference to guide trans-attention, which comprises the following steps: simulating a false data injection attack, a denial of service attack, an address resolution protocol spoofing attack and a domain name system spoofing attack, collecting physical layer sensor data and network layer flow data, and preprocessing multi-modal data; bMama is constructed to perform dynamic time modeling on multi-modal data, a graph neural network is combined to adversariate a variational auto-encoder, features of a power grid system topology and a communication topology structure are fused, and robustness of potential representation is enhanced through adversarial training; the method comprises the following steps of: guiding feature complementary fusion by using modal difference through a difference guide iteration cross-attention fusion mechanism, improving the capability of distinguishing complex attacks, finally carrying out attack detection and classification on fused modals, and executing end-to-end optimization according to a weighted combination of loss of each part. The method can effectively detect and classify the multi-modal attack in the smart power grid, and enhances the safety and reliability of a complex system.

The invention relates to a carbon emission checking system and method based on a block chain, and the system comprises a basic resource layer which collects enterprise energy consumption data in real time, employs cloud computing resources to provide data storage and computing capability, and employs a Hash algorithm to process the enterprise energy consumption data, and obtains a data abstract; the block chain network layer carries out distributed storage through an alliance chain architecture, carries out consensus evidence storage on the data abstract by adopting a PBFT consensus algorithm, and generates an encrypted data abstract; the intelligent contract layer receives the encrypted data abstract transmitted by the block chain network layer, and executes verification, calculation and supervision rules; and the application service layer calls the processing result of the intelligent contract layer, and respectively provides an enterprise carbon checking terminal service interface, an institution carbon checking terminal service interface and a competent department carbon supervision terminal service interface for the enterprise, a third-party checking institution and the competent department to check the carbon emission. According to the invention, high-efficiency check, high-credibility verification and whole-process traceable supervision of the carbon emission data are realized.

In a traffic scene, a traditional target detection algorithm always faces the problems of strong background interference, difficulty in small target detection and the like, and detection precision and robustness are affected. Therefore, the invention provides an improved YOLOv12n vehicle detection method in which an EMA (Empirical Multi-scale Attention) attention mechanism and an SFA (Space Feature Aggregation) attention mechanism are fused. The invention further provides a method for detecting the YOLOv12n vehicle based on the improved YOLOv12n vehicle based on the attention mechanism of the EMA (Empirical Multi-scale Attention) and the attention mechanism of the SFA (Space Feature Aggregation). The SFA module is deployed in a shallow network, key target area expression is enhanced by aggregating spatial features, and background noise interference is suppressed; the EMA module is embedded into a neck network, and the global information capture and multi-scale sensing capabilities are improved by adopting multi-scale convolution, cross-space modeling and feature grouping mechanisms. According to the method, the real-time performance is kept, meanwhile, the detection precision in a complex scene is remarkably improved, and particularly, higher robustness is shown in the aspects of small target recognition and shielding processing.

The invention relates to the technical field of intelligent dormancy control of Internet of Things equipment, and discloses an intelligent dormancy control system of the Internet of Things equipment based on environment self-sensing, which comprises an environment feature extraction layer module, an environment dynamic model construction module, a self-adaptive dormancy decision engine module, a lightweight state cache module and a power supply cooperative control module. By acquiring the physical layer signal characteristics and the network layer connection state information of the NB-IoT protocol stack in real time, a dynamic model can be constructed based on the environmental signal fluctuation, the stability of the environment where the equipment is located can be accurately evaluated, the dormancy triggering condition can be dynamically adjusted, false wake-up caused by a fixed threshold value is avoided, dependence on a traditional sensor is reduced, and the reliability of the equipment is improved. The energy efficiency and adaptability of the equipment are improved, the anti-interference capability in an interference environment is improved, and the long-term stable operation of the equipment in a complex industrial environment is ensured.

The invention discloses a multi-mode industrial anomaly detection method fusing image masks and mixed experts. According to the method, an image mask technology and a hybrid expert method are fused to guide an industrial multi-modal pre-training model, and based on a multi-modal industrial data pre-training model, the pre-training large model can learn features which are difficult to express by a single mode. In the training stage, part of pixel blocks of a multi-modal input image are randomly masked by using a mask technology, the model can reconstruct visual features aligned with image texts of masked image blocks according to visible image blocks through training, and the model can learn low-level geometric structure features and high-level semantic information of the image from the model at the same time, so that the multi-modal input image is obtained. Therefore, most visual information required by downstream tasks is covered. An encoder part of an industrial large model adopts a hybrid expert structure to replace a forward propagation layer, and a small part of network layers can be dynamically activated during data processing, so that the model training and reasoning speed is increased while the model parameter quantity is kept.