27 results about "Madaline" patented technology

The application discloses a kind of based on multi-modal perception and adaptive model's cold-rolled emulsion concentration intelligent closed-loop control method and system.Method includes: through heterogeneous sensor array multi-modal perception, data is fused using neural network based on attention mechanism, obtain comprehensive state information;Adaptive model predictive controller and the meta-learning optimizer based on deep reinforcement learning are used to generate control instructions by two-layer intelligent decision engine constituted;Combining digital twin model carries out forward-looking simulation and optimization;Finally drive high-precision actuator to complete closed-loop control.System includes multi-modal perception module, intelligent decision and control module, digital twin and optimization module, high-precision execution module and data communication network.The application solves the problems of low sensor reliability, rigid control model and lack of prediction ability in the prior art, and realizes high-precision, adaptive and intelligent closed-loop control of emulsion concentration.

The application discloses a kind of based on low-rank self-adaption's federal remote sensing big model training method and system, comprising: using respective remote sensing data, pre-training big model and local low-rank self-adaption model is carried out individualized heterogeneous data knowledge learning;Using the local low-rank self-adaption model parameter initialized global shared low-rank self-adaption model parameter after training, based on pre-training big model and global shared low-rank self-adaption model is carried out collaborative federal training, learns global remote sensing task field knowledge;By heterogeneous rank alignment fusion mode, the alignment of local heterogeneous data knowledge and global remote sensing task knowledge is realized after, output is composed of pre-training big model and the low-rank self-adaption model parameter of alignment after local by federal remote sensing big model, such as greatly reduce the parameter quantity and communication overhead of federal training upload, guarantee the efficient training of model, also realize local heterogeneous data and global remote sensing task self-adaption alignment, guarantee the versatility and robustness of model.

This invention discloses an adaptive model reuse method based on adapter fine-tuning. Taking object detection in computer vision as an example, firstly, a model pre-trained on a large-scale image dataset is obtained as the original model, and candidate adapters are set for it according to the structure of the pre-trained model. Then, the Taylor expansion method is used to calculate the global gain of adding adapters at different positions in the network for images with localized bounding boxes and class labels in the target domain, thereby achieving adaptive adapter configuration. In the fine-tuning stage, the global gain of adapters at different positions for labeled image samples is recalculated. The original model parameters are frozen, and the adapter parameters are updated using gradient backpropagation with a learning rate weighted by the global gain softmax. Finally, the fine-tuned model is reused in downstream tasks to locate objects in images and provide class predictions. This invention achieves adaptive adapter configuration based on global gain, realizing more accurate object localization and class prediction in object detection tasks.

This invention discloses a predictive control method and system for energy conservation in urban rail transit electromechanical systems based on an adaptive model, belonging to the field of energy conservation control technology for urban rail transit. Existing control methods for ventilation and air conditioning systems in subway stations struggle to balance environmental comfort and operational energy consumption. Online simulation is time-consuming, online strategy trial-and-error carries significant risks, and real-time adaptive adjustment is difficult. This method deploys an offline-trained neural network proxy model and a control strategy model on edge computing devices in subway stations. The neural network proxy model outputs predicted temperature field distribution data and equipment energy consumption data for multiple areas of the station based on outdoor environmental parameters, passenger flow parameters, and equipment operating parameters. The control strategy model uses this proxy model as an interactive environment to construct a reward function based on thermal comfort and equipment energy consumption indicators for reinforcement learning pre-training. This invention is primarily used for energy-saving optimization control of ventilation and air conditioning systems in subway stations.

The application relates to the technical field of solid state disks, and discloses a fault diagnosis method and system for a solid state disk master control chip. The method comprises the following steps: collecting operation logs, delay distribution and multi-source information of the master control chip, generating a data set through time series analysis and extracting dynamic change characteristics; determining an association mode of a health state based on the characteristics and a clustering algorithm; when the mode shows that delay fluctuation exceeds a threshold value, adjusting a standard by fusing the multi-source information to identify potential abnormal fluctuation; integrating abnormal fluctuation and operation logs by using a decision tree algorithm to obtain an early warning index, constructing an adaptive model to update a diagnosis strategy; if the update involves a high load environment, filtering normal fluctuation through an anomaly detection mechanism to realize fault positioning; classifying and integrating fault-related multi-source information, evaluating overall system performance risks and generating a health state report. The application effectively improves the accuracy of fault early warning and the adaptability of diagnosis strategies of the master control chip of the solid state disk.

The application discloses a kind of based on multi-sensor fusion's aero-engine test bed performance parameter synchronous acquisition and analysis system;System is through the hierarchical linkage of sensor layer, data acquisition layer, data processing layer and analysis decision layer, integrates multiple types of high-precision sensors, adopts the dual time alignment mechanism of hardware synchronous trigger and software compensation combination, fuses the multi-modal algorithm of Kalman filter, SVM classifier, deep neural network (DNN) and D-S evidence reasoning, with double-layer inverse Broyden iteration and online learning adaptive model, realize the accurate synchronous acquisition of engine temperature, pressure, vibration and other key parameters, real-time analysis and fault early warning.System is compatible with multiple aviation buses, with redundant transmission and anti-interference design, visual interface supports multidimensional data linkage and playback, can adapt to the engine test demand under different conditions, provides reliable technical support for aero-engine research and development, maintenance and performance optimization.

A cross-condition fault diagnosis method based on visibility enhanced graph data and structure information aware graph convolutional neural network is provided to solve the problem of low efficiency of domain adversarial training. The method first uses an enhanced visibility weighted graph to map the sampling points with visibility in the vibration sequence to graph nodes and establish topological connections through the horizontal visibility rule, so that the graph sample can reflect the time sequence visibility and its weight characteristics of the signal. Then, a domain adversarial adaptive model based on structure information aware graph convolutional neural network is constructed. In the model, the topological connection and edge weight information generated by the visibility rule are used to weight and aggregate the node features and enhance the secondary features. Combined with source domain supervised classification and domain adversarial alignment, preliminary migration is realized. The domain adaptive training module driven by the combined loss weights and fuses the source domain classification loss, domain adversarial loss, target domain entropy minimization loss and high confidence pseudo label supervision loss, and optimizes them in stages, so as to reduce the prediction uncertainty of the target domain and enhance the class separability.

The application discloses an alloy process parameter optimization method and system based on transfer learning, belongs to the technical field of industrial intelligent control, and is used for solving the technical problems that current alloy process parameter optimization methods cannot adapt to new specification cold start, black box constraint processing capacity is limited, and single round batch output differentiated candidate process parameter points cannot be supported. The method comprises the following steps: training a historical prior model through a historical data set; completing the first round optimization of the current parameter optimization task according to the historical prior model; training a task adaptive model according to a measured performance data set obtained after the first round optimization; performing knowledge fusion on the historical prior model and the task adaptive model according to a self-defined historical confidence weight function; constructing a collection function with different search preferences in a process parameter search space and performing parameter optimization to obtain multiple groups of candidate process parameters and acquire measured performance data, retraining the task adaptive model until a preset convergence condition is met, and outputting optimal process parameters.

This invention relates to a model training method for unsupervised graph domain adaptation tasks, belonging to the field of graph domain adaptation technology, and solves the problem of negative transfer affecting model performance in existing technologies. The method includes: acquiring labeled source domain graph data and unlabeled target domain graph data; pruning the source domain graph structure based on topological deviations; constructing an adaptive model, which includes a graph encoder, a domain discriminator, and a classifier; the graph encoder is used to encode features into the input graph data; the domain discriminator is used to determine whether the data comes from the source domain or the target domain based on the encoded features; the classifier is used to predict labels based on the encoded features; and the adaptive model is trained on the pruned source domain graph data and target domain graph data to obtain a classification model corresponding to the target domain. This significantly improves the classification accuracy of the model in the target domain.

The application discloses a wireless communication network intelligent interference management system based on deep learning and belongs to the technical field of communication network management; by processing and fusing multipath, mobility and environmental factors, the problem of inaccurate channel state information caused by the traditional model ignoring non-stationary interference can be solved; the constructed sequence simultaneously captures time trend and spatial cooperation, which can provide high-quality input for subsequent adaptive models; by obtaining cross-scene general parameters through meta-learning pre-training, realizing millisecond-level adaptive dynamic channels through online fine-tuning, and finally outputting accurate interference prediction, the adaptability of the model to complex channels and the real-time performance of interference estimation can be significantly improved, which provides key support for the anti-disturbance ability of trajectory control; through the cooperation of the above steps, the cross-scene generalization foundation is laid through the pre-training stage, the new scene is quickly adapted through the online fine-tuning stage, and the key compensation information is output in the interference prediction stage, forming a closed-loop logic of pre-training, fine-tuning and prediction.

This invention relates to the interdisciplinary field of artificial intelligence and geological disaster monitoring and early warning, specifically to a landslide identification method and system based on phased feature adaptive transfer learning, comprising: constructing an initial landslide identification model F1 with an encoder-decoder; transferring general visual pre-trained weights and utilizing a remote sensing image dataset D from the source domain. S Fine-tuning the encoder parameters of model F1 yields a domain-adapted optimized landslide identification model F2; the source domain remote sensing image dataset D... S And the target domain UAV imagery dataset D T The data is input into model F2 and processed based on the covariance alignment mechanism to obtain the domain adaptive model F3; a small amount of labeled target domain small sample data D′ is then used. T The input is fed into model F3, and the specified parameters in model F3 are fine-tuned through a supervised learning mechanism to obtain target domain adaptation model F4; based on target domain adaptation model F4, a landslide recognition task is performed on the input UAV image to be detected.

The application discloses a high-precision solution method for steady-state magnetohydrodynamic equations based on an adaptive PINN model and belongs to the technical field of magnetohydrodynamic equation solution.The application constructs a complete closed-loop system from network construction, residual diagnosis to optimization adjustment. The sharing-branch network structure is used to realize differentiated learning of physical quantities, the adaptive residual mechanism is introduced in the Adam stage to dynamically focus on the high residual area, the L-BFGS stage is combined with the line search condition and the loss function scale transformation to perform fine optimization, the modules work collaboratively, and a systematic high-precision solution scheme is formed. Through the RAR or RBA mechanism, the model can autonomously diagnose the high residual area and dynamically adjust the training points during the training process, and the fitting capability for the boundary layer and the strong nonlinear region is significantly improved. Meanwhile, the Strong Wolfe line search and the loss function scale transformation are combined to adaptively balance the multi-objective loss scale, and the robustness of the step selection and the convergence precision are ensured to be improved.

The application discloses a kind of self-adapting current protection method, system, equipment and medium, belong to relay protection technical field, method includes: analog power output generates typical scene set, extracts fluctuation characteristics, constructs self-adapting model and generates dynamic reference, combines fault characteristic and outputs real-time threshold and executes protection decision.System includes data generation and scene simulation module, characteristic index extraction module, boundary parameter definition module, self-adapting model construction and reference generation module, real-time threshold output module, protection action decision module: the present application is by constructing distributed power output generation model and self-adapting setting calculation model, in steady state, linear follow-up is realized to power fluctuation using transfer function, in transient state, current limit is calculated in combination with fault ride-through mechanism.Realize that protection setting value is in real time self-adapting matching to complex operating condition, avoid the risk of distribution network protection misoperation and refusal caused by new energy output randomness, guarantee the stable operation of system.

The present application relates to a kind of narrow rectangular flow channel safety monitoring method based on physical prior map self-adaption, belong to nuclear power safety monitoring technical field.It includes: obtaining source domain data and target domain data, by the way of transfer learning training a physical prior map self-adaptive model, for the cross-condition local deformation fault diagnosis of narrow rectangular flow channel;The model includes initial feature extractor, graph feature extractor of fusion fluid motion physical prior and classifier.In the model training process, by calculating physical constraint prior loss to learn source domain adjacency matrix and target domain adjacency matrix;Source domain classification loss is calculated to minimize empirical risk;Class-level subdomain distribution alignment loss and manifold constraint loss are calculated to achieve fine-grained cross-domain feature distribution alignment.Multiple iteration training, after each loss converges, model training is completed.The present application can realize the diagnosis to abnormal state under complex multi-condition environment, and carry out effective safety monitoring.

This invention discloses a real-time monitoring method for hydropower station gate flow based on big data, belonging to the field of hydropower technology. The method includes the following steps: forming a flow state perception model; constructing a flow anomaly disturbance identification model to calculate the flow disturbance difference; constructing a gate self-feedback correction model based on deviation back mapping; constructing a gate group collaborative optimization mechanism; and constructing an adaptive model detection mechanism. This invention significantly improves the system's response capability to nonlinear disturbances and the security and reliability of monitoring data by innovatively introducing technologies such as a flow anomaly disturbance identification model, a gate self-feedback correction model, a time-series drift tracking model, and a structurally trusted blockchain mechanism. Compared with traditional methods, this invention not only possesses higher flow prediction accuracy and adaptive control capabilities but also constructs a multi-node asynchronous consensus verification mechanism for monitoring data, effectively supporting intelligent scheduling decisions and event tracing applications across departments and time periods.

The application discloses an alloy process parameter optimization method and system based on transfer learning, belongs to the technical field of industrial intelligent control, and is used for solving the technical problems that current alloy process parameter optimization methods cannot adapt to new specification cold start, black box constraint processing capacity is limited, and single round batch output differentiated candidate process parameter points cannot be supported. The method comprises the following steps: training a historical prior model through a historical data set; completing the first round optimization of the current parameter optimization task according to the historical prior model; training a task adaptive model according to a measured performance data set obtained after the first round optimization; performing knowledge fusion on the historical prior model and the task adaptive model according to a self-defined historical confidence weight function; constructing a collection function with different search preferences in a process parameter search space and performing parameter optimization to obtain multiple groups of candidate process parameters and acquire measured performance data, retraining the task adaptive model until a preset convergence condition is met, and outputting optimal process parameters.

This invention discloses a deep learning-based intelligent interference management system for wireless communication networks, belonging to the field of communication network management technology. By processing and fusing multipath, mobility, and environmental factors, it can solve the problem of inaccurate channel state information caused by traditional models ignoring non-stationary interference. The constructed sequence simultaneously captures temporal trends and spatial cooperation, providing high-quality input for subsequent adaptive models. Through meta-learning pre-training to obtain cross-scenario general parameters, online fine-tuning achieves millisecond-level adaptation to dynamic channels, and finally outputs accurate interference prediction, which can significantly improve the model's adaptability to complex channels and the real-time performance of interference estimation, providing key support for the anti-disturbance capability of trajectory control. Through the synergistic cooperation of the above steps, the pre-training stage lays the foundation for cross-scenario generalization, the online fine-tuning stage quickly adapts to new scenarios, and the interference prediction stage outputs key compensation information, forming a closed-loop logic of pre-training, fine-tuning, and prediction.

The application provides a desulfurization control method based on federated learning and adaptive model predictive control, comprising the following steps: dividing distributed control nodes for a wet desulfurization system and constructing a federated learning architecture; each distributed control node trains a local model and uploads model gradient parameters of the model to a federated aggregation server for weighted aggregation to obtain a global coupled model; and an adaptive model predictive controller is constructed based on the global coupled model to output a cooperative control instruction for cooperative control of the wet desulfurization system. The application avoids the privacy leakage risk and data transmission constraints of sensitive data upload by constructing a federated learning architecture, while realizing cooperative optimization control of multiple operating parameters of the desulfurization system, improving the control stability under wide load and coal quality fluctuation conditions, and balancing desulfurization efficiency and cost. Further, by dividing the working condition interval to switch the controller control parameters, adaptive adaptation of the control strategy under different operating conditions is realized.

The application discloses a kind of real-time data question and answer training and chart generation system, including data access and preprocessing module, field self-adaptive training module, real-time question and answer interaction module, intelligent chart generation module and knowledge graph and feedback optimization module;The data access and preprocessing module is used for the access of multi-source heterogeneous real-time data, cleaning, standardization processing and dynamic caching, and outputs structured data and real-time data stream index;The field self-adaptive training module is based on prompt auxiliary self-learning mechanism, and a small amount of labeled field data and a large amount of unlabeled data are used for model fine-tuning, to generate field-specific question and answer model, while constructing field terminology dictionary and updating in real time.The application realizes the integration function of field self-adaptive model training, real-time accurate question and answer, intelligent chart generation and continuous iteration optimization, improves user data query efficiency and data exploration ability.

The application relates to a model prediction task sampling meta-reinforcement learning and domain randomization training method, wherein the method comprises the following steps: constructing a risk random function of an online constructed deep generative model, determining a risk function distribution of the deep generative model, and inferring an approximate posterior of the risk function distribution; estimating a function posterior distribution of the deep generative model to construct a target collection function, and performing random sampling in a target task space by using the target collection function to obtain a prediction value of a posterior task adaptive loss and corresponding collection data, and generating a target scenario optimization task batch by using the prediction value and a risk learner to predict a collection score of the collection data; determining a domain randomization and meta-reinforcement learning update rule of a machine learner, so as to perform a decision model update training operation in an adaptive model of a target zero sample or a target small sample on the machine learner. Therefore, the problems that the prior art is difficult to improve the adaptive robustness and improve the calculation efficiency of the model are solved.

The application discloses a model fine-tuning method for privacy protection, comprising the following steps: embedding an adaptive module into a feature extraction network to obtain an adaptive model; pre-training the adaptive model by using first data, and updating adaptive model parameters except the adaptive module in the pre-training process to obtain a pre-trained model; and distributing the pre-trained model to a user end, so that the user end can train the pre-trained model by using second data, and update parameters of the adaptive module in the training process until the pre-trained model converges, thereby obtaining a user end inference model. The user end inference model can obtain personalized user information, and realize risk control in the personalized learning process. Accordingly, the application also discloses a risk control method for privacy protection.

This invention discloses a method and system for harmonic reducer fault diagnosis based on federated prototype domain adaptation, belonging to the field of machine fault diagnosis technology. It addresses the problems of large differences in the distribution of harmonic reducer data among users and the scarcity of labeled data for some operating conditions, leading to low model diagnostic accuracy. It also addresses the lack of semantic hierarchical guidance in existing federated domain adaptation methods and the inability to measure the contribution of each source domain user model. The method comprises: first, constructing a multi-user prototype collaborative interaction mechanism using user model parameters and class prototypes as the core interaction information of the federated learning framework, achieving effective knowledge transfer between source and target domain users. Second, designing a class prototype-guided local model training method to align local data features with class prototypes while protecting privacy. Finally, proposing a difference-aware adaptive model aggregation strategy to improve the diagnostic performance of the global model. This invention is applicable to harmonic reducer fault diagnosis.

The application discloses a kind of marine pipeline anticorrosion optimization methods based on adaptive algorithm, comprising: S1, through multi-source sensing array real-time acquisition marine pipeline original data;S2, the original data collected is preprocessed, generates standardization data set;S3, corrosion prediction model is established by dynamic collaborative perception network, obtains pipeline corrosion rate and corrosion risk thermodynamic diagram, outputs corrosion risk assessment result;S4, according to corrosion risk assessment result, establish anticorrosion digital twin model, design anticorrosion optimization strategy;S5, optimal anticorrosion parameter sequence in anticorrosion digital twin model is solved by using adaptive model predictive control mechanism, generates optimal anticorrosion scheme;S6, execute optimal anticorrosion scheme, and real-time monitoring pipeline response data;S7, based on pipeline response data, online updates dynamic collaborative perception network parameters and anticorrosion digital twin model, forms closed loop optimization.The application improves anticorrosion effect, reduces operating cost, with economy and sustainability.

This disclosure proposes an actuator failure compensation and safety control method, storage medium, and system based on a two-layer evaluation architecture. The bottom layer uses an adaptive model to predict the nominal control output command. Above this, a reinforcement learning model is built to output high-frequency residual compensation commands to eliminate model mismatch. A second evaluation network is independently constructed to assess the future global runaway risk of the vehicle. Lagrange dual optimization is introduced to transform the macro-risk assessment into mathematical constraints to guide the correction of the first strategy network. Continuous flexible degradation is achieved by using a global risk index feedforward to adjust the target cost weight of the predictive control. Discrete quadratic programming command projection is performed by triggering a control barrier function at the limit boundary. This invention overcomes the conservative strategy degradation defects caused by traditional mixed rewards, establishes an absolute safety defense line, and improves the control accuracy and safety robustness of the vehicle under complex nonlinear conditions.