305 results about "Dynamic learning" patented technology

The invention relates to the technical field of data governance, and discloses a data governance method based on an intelligent decision engine, which comprises the following steps: carrying out business semantic classification and marking on preliminarily processed real-time streaming data, and constructing a data portrait library; constructing a dynamic topological graph, learning an abnormal propagation rule based on a graph neural network, analyzing an influence range and establishing an influence grading mechanism; performing multi-dimensional quality evaluation on the data, and generating a dynamic data quality score and a grading strategy; constructing a data quality historical problem and reason case library, and generating a quality anomaly root cause judgment and influence quantification report by using a large language model agent; generating a candidate strategy set, and selecting an optimal governance strategy from the candidate strategy set by establishing a multi-objective optimization model; and performing compliance test and conflict identification on the optimal governance strategy by using a large language model agent, and dynamically adjusting the decision weight of a rule engine by using a reinforcement learning algorithm to realize a closed loop of data governance and dynamic learning.

A dynamic learning server-based logistics system includes a node-based logistics receptacle operative to receive a delivery item as part of a logistics transaction and also includes a backend server maintaining a management profile related to operation of the node-based logistics receptacle. The node-based logistics receptacle includes a plurality of monitored receptacle components, a wireless accessory sensor node having a plurality of sensors, and a bridge node operative to retrieve event information from the wireless accessory sensor node. The backend server receives the retrieved event information, compares the retrieved event information with the management profile, identifies a threshold change condition from the comparison, dynamically revises the management profile when the threshold change condition is identified, and transmits an adjustment message to the bridge node. The adjustment message is based upon the revised management profile, and the adjustment message initiates a timing change to operation of the bridge node.

The invention relates to the technical field of defect detection systems, and discloses a titanium alloy surface crack defect detection system based on deep learning. According to the system, a crack feature extraction module is used for collecting a titanium alloy surface image and extracting multi-scale crack features including crack trend distribution features and micro-crack density features; the multi-modal data fusion module is used for receiving the multi-scale crack features and performing space-time alignment on the multi-scale crack features and ultrasonic reflection wave features collected in real time to generate a fusion defect feature matrix; the dynamic learning engine module is used for constructing a crack propagation prediction model according to the historical change trend of the fusion defect feature matrix and outputting a dynamic defect response vector; the defect positioning module is used for mapping the dynamic defect response vector to a titanium alloy surface three-dimensional coordinate space to generate a defect position thermodynamic diagram; and the self-adaptive scanning control module is used for analyzing the defect confidence of each area in the defect position thermodynamic diagram and dynamically adjusting the scanning path and the focal length parameter of the industrial camera.

The invention belongs to the technical field of semantic understanding systems, and particularly relates to a semantic understanding system based on a large language model.The semantic understanding system is characterized in that firstly, a data preprocessing module is used for conducting cleaning, denoising and cross-modal conversion on input multi-modal data such as texts and images, and standardized data is generated; a semantic feature extraction module extracts general semantic features by using a pre-training model, adapts to field requirements through dynamic learning rate fine adjustment, and outputs scenarized semantic vectors; then, a dynamic semantic-knowledge bidirectional fusion module adjusts token weight according to a dynamic semantic weight algorithm, realizes real-time alignment of semantics and a knowledge graph by means of a knowledge entity association strength algorithm, and a knowledge enhancement fusion module further optimizes knowledge weight and dynamically updates association; then, the semantic reasoning module performs multi-round reasoning based on fusion information, and evaluates the result reliability in combination with a confidence coefficient algorithm; and finally, the output and optimization module generates a structured result, and iteratively optimizes parameters of each module according to feedback data to complete a semantic understanding processing flow.

The invention discloses a dynamic learning path planning method and system based on a learner portrait. The method comprises the steps of collecting learning behavior data, learning result data and background attribute data of a target scholar based on a plurality of data sources, and obtaining knowledge point mastery degree vector representation, learning style preference and cognitive ability level to construct a multi-dimensional dynamic portrait of the target scholar; generating an initial personalized learning path of the target scholar based on a preset learning target and the multi-dimensional dynamic portrait of the target scholar; learning process data of a target scholar for related learning resources is monitored in real time, a learning effect evaluation report is generated according to the learning process data, a multi-dimensional dynamic portrait is adjusted to trigger a path replanning mechanism to generate a brand new personalized learning path, and the brand new personalized learning path is pushed. The learning path and the current state of the learner are kept optimally matched all the time, and personalized static planning is upgraded to dynamic syndrome guidance.

The invention discloses an AI-based computer basic course personalized learning path recommendation method and system, and belongs to the technical field of AI-based data processing. According to the system, behavior data, cognitive data and course interaction data of a learner are acquired through a multi-dimensional data acquisition module, a computer basic course knowledge point association network is established in combination with a dynamic knowledge graph construction module, and a personalized learning path is generated by using an improved deep reinforcement learning algorithm. And the path is dynamically adjusted through the real-time feedback module. The core of the method is that a learner portrait is fused with space-time correlation features of a knowledge graph, a cognitive evaluation model is updated in real time through a Bayesian network, the problems that in a traditional recommendation method, paths are solidified, and the dynamic learning state of an individual is ignored are solved, more accurate personalized learning guidance is achieved, and the learning efficiency and effect of a computer basic course are improved.

The invention discloses a digital power grid network security situation self-adaptive evaluation and defense strategy generation method, system and device and a medium, and belongs to the technical field of digital power grids, the digital power grid network security situation self-adaptive evaluation and defense strategy generation method comprises a causal knowledge graph construction module which is configured to construct and maintain a causal situation graph of a digital power grid; the situation assessment module is configured to perform principle-level anomaly detection by injecting a topology security potential field into the causal situation map and monitoring a potential energy track of a data stream, and dynamically learn and update an unrecognized operation instruction sequence in the causal situation map by adopting an immune learning mechanism, calculating a comprehensive security situation vector based on the updated causal situation map; and the strategy generation module is configured to verify the credibility of the comprehensive security situation vector based on the consensus information of the distributed equipment, and generate a defense strategy sequence with the optimization of the comprehensive security situation vector as a target under the multi-target constraint of considering the service influence and the resource cost.

The invention discloses an intelligent editing style learning and shot arrangement method and device, and the method comprises the steps: firstly carrying out the shot segmentation of an original material video, obtaining a plurality of independent video clips, carrying out the multi-dimensional analysis of each video clip, generating a structured label text containing image semantic features and shooting grammar features, and carrying out the segmentation of the shot; semantic segmentation is carried out on a manuscript to be edited to obtain a plurality of text paragraphs, each text paragraph is matched with at least one edition example, finally prompt words are constructed and input into the generative language model to carry out shot arrangement, and the generative language model outputs a shot arrangement sequence of the current paragraph. The problem of error accumulation caused by the fact that a traditional intelligent editing method depends on an editing script intermediate link is solved, the defect that in the prior art, it is difficult to dynamically learn a specific editing style is overcome, collaborative optimization of semantic matching and editing grammar is achieved, and shot selection accuracy, arrangement fluency and style adaptability are remarkably improved.

The invention relates to the technical field of artificial intelligence and data processing, and discloses a multi-mode identity relation inference system based on a graph neural network. The system comprises a multi-modal feature extraction module, a cross-modal alignment module, a graph structure construction module, a dynamic relation reasoning module and a decision output module. According to the method, the cross-modal alignment module is introduced to project the image features and the text features to a unified public semantic space, so that the nonlinear distribution difference of heterogeneous modals in an embedding space is effectively eliminated, and cross-modal alignment errors are avoided from the source; by integrating the attention mechanism of modal perception in the graph neural network, the system can dynamically learn the semantic association strength between the nodes in different modals, adaptively adjust the weight distribution in the neighborhood information aggregation process, and significantly improve the accuracy of node characterization.

The invention relates to a bearing fault diagnosis method based on variational mode decomposition and time sequence partitioning cross attention fusion, which comprises the following steps: acquiring an original vibration acceleration signal of a rolling bearing, and constructing a standardized original data set; segmenting the standardized original data into a plurality of data blocks, and generating a time domain embedding feature; based on the time domain embedded features, extracting high-order global time domain features by using a multi-head self-attention mechanism, residual connection and a feedforward neural network; decomposing the standardized original data into a plurality of intrinsic mode functions, and extracting frequency domain distribution features through a convolutional neural network; taking the frequency domain distribution characteristics as query vectors, retrieving and matching related fault context information in the global time sequence characteristics, realizing weighted fusion of time-frequency modes, inputting fused fault representation vectors into a classifier, and calculating a result of a bearing health state; and constructing a loss function containing label smoothing and a dynamic learning rate scheduling strategy, and carrying out iterative optimization on model parameters until the model converges.

The invention provides a denoising method for a magnetic resonance spectrum based on Transform, and relates to the technical field of signal processing and artificial intelligence, and the method comprises the steps: firstly obtaining a noisy free induction decay signal (FID) and a corresponding label signal, constructing a training data set, converting the signal into a frequency domain signal through fast Fourier transform, and carrying out the amplitude normalization; a real part and an imaginary part of a frequency domain signal are respectively projected to a high-dimensional feature space, and are respectively sent to a multi-head self-attention mechanism-based multi-layer Transform encoder through two channels, so as to capture the global correlation of a long sequence. And the coded output is separated and decoded, a real part and an imaginary part are recovered respectively, and a denoised complex frequency domain signal is obtained through combination. In the training process, a complex mean square error is used as a loss function, and dynamic learning rate scheduling and an early stop mechanism are combined to improve convergence stability and model training efficiency.

The invention relates to the technical field of medical information processing, and discloses an acute pulmonary embolism auxiliary diagnosis decision-making system based on a large model, and the system comprises a data collection and preprocessing module which is used for receiving multi-source clinical data and preprocessing the multi-source clinical data; the large model reasoning module is used for inputting the preprocessed multi-source clinical data into a large model fused with a multi-modal causal attention and identification adversarial learning mechanism, and outputting a reasoning result; the dynamic learning module is used for accessing desensitized case data of a hospital electronic medical record system, continuously learning new cases through federal learning, and optimizing the recognition capability of a large model pair; the diagnosis suggestion generation module is used for generating structured diagnosis suggestions according to the reasoning result output by the large model; the result output and interaction module is used for displaying diagnosis suggestions to doctors and updating reasoning results in real time by a large model when the doctors supplement data; the diagnosis efficiency is improved.

The invention provides a multi-agent cooperation-oriented process planning and execution system and method, and the system comprises an autonomous process planning engine which dynamically generates and optimizes a process path based on a reinforcement learning model according to a real-time business rule; the multi-Agent collaborative execution layer receives a flow path, performs task decomposition and adaptive load balancing to intelligently distribute sub-tasks to a plurality of Agents for parallel execution, and adopts a consensus protocol to guarantee the consistency of key data; and the fault self-healing guarantee system continuously monitors the running state of the system, identifies a source through an intelligent positioning module when detecting an abnormality, and automatically generates and executes a recovery strategy in combination with a preset template and a dynamic learning mechanism. By constructing an intelligent closed loop of planning, execution and self-healing, dynamic self-adaption, efficient cooperation and high-reliability operation of the business process are achieved, complex business changes and system anomalies can be coped with without manual intervention, and the intelligent level and the business continuity guarantee capability of an automatic system are remarkably improved.

The invention relates to the technical field of heart failure auxiliary diagnosis and treatment, in particular to a heart failure auxiliary diagnosis and treatment knowledge distillation method and system.The knowledge distillation method comprises the steps that an age-crossing fusion data set is built by combining ancient medical case text data related to traditional Chinese medicine heart failure; performing field fine tuning on the pre-training model to generate a basic teacher model; performing semantic alignment, and outputting a cross-time contrast database; constructing a curative effect scoring model by using the time sequence model, and outputting a quantitative association database; using a graph neural network to output an intelligent heart failure knowledge graph, and generating a comprehensive teacher model; performing knowledge distillation on the comprehensive teacher model to generate a clinical student model; the knowledge distillation system is applied to the knowledge distillation method, has a dynamic learning ability, can fuse ancient and modern trans-time knowledge and combine traditional Chinese medicine ancient book wisdom to carry out heart failure auxiliary diagnosis and treatment, and can give consideration to model performance and deployment efficiency through knowledge distillation.

The invention discloses a smart home energy efficiency optimization method and system based on multi-source perception learning. The method comprises the steps of multi-source data acquisition and preprocessing, multi-source perception feature extraction, user behavior and environment dynamic learning, demand prediction, scene recognition, multi-target collaborative optimization and control strategy generation and strategy execution and feedback learning. The invention relates to the technical field of smart home, in particular to a smart home energy efficiency optimization method and system based on multi-source perception learning, and according to the scheme, multi-source perception data are collected, a multi-target optimization model is constructed, smart home scheduling is driven by using real-time electricity price, and user satisfaction is improved; an improved self-adaptive optimization algorithm is constructed, parameters of a multi-objective optimization model are optimized, dynamic electricity price and user behavior changes are responded in a mode of searching an optimal solution, energy efficiency is optimized, and interference on living habits of users is reduced to the maximum extent.

The invention relates to the technical field of data processing, in particular to a personalized learning path planning method and system. Comprising the following steps: acquiring a behavior data sequence arranged according to a time sequence, and extracting a repetitive mode in a learning behavior and a difference point deviating from a standard learning process to obtain a behavior difference index; dynamically adjusting an association structure between nodes in the knowledge graph, identifying key knowledge nodes, determining a missing concept set, and inserting missing concepts into a current learning path to generate a supplementary path draft; by simulating a plurality of alternative learning paths and calculating coherence scores of the alternative learning paths, screening and sorting to obtain a plurality of personalized learning path options; and integrating the behavior data newly generated by the learner through a feedback module, and outputting a final dynamic learning path. According to the method, the problems of insufficient dynamic adaptability and difficulty in effectively utilizing the behavior data of the learner in the existing learning path planning are solved, and dynamic and personalized learning path planning based on the behavior difference of the learner is realized.

The invention belongs to the technical field of physical exploration, and particularly discloses a deep geophysical prospecting weak abnormal signal enhancement method and system, and the method comprises the steps: generating a geostructured prior guide field, carrying out the targeted decomposition of an original geophysical prospecting signal, and synchronously estimating an uncertainty region, thereby achieving the real-time enhancement of a weak abnormal signal while maintaining the fidelity of background information; focusing the enhanced resources in the information fuzzy region; the decomposition signals are input into a dynamic learning model, and random noise and structured interference generated by multiplicity are effectively suppressed through antagonistic learning; the enhanced sub-band signals are fused into reconstructed signals with high geological interpretability by using a deep reconstruction network through a self-attention mechanism and compound loss function optimization; a closed-loop feedback mechanism is established, newly disclosed high-confidence anomaly features are fed back and integrated to a prior field, an iteration guide field is formed to drive multi-round iteration, step-by-step mining and locking of weak anomalies are achieved, and finally a high-reliability result is stably converged.

The invention relates to the technical field of knowledge tracking, and discloses a graph enhanced double-memory collaborative knowledge tracking model based on an ACT-R cognitive architecture. Comprising a static knowledge structure coding module based on hypergraph projection, a batch-level dynamic learning track construction and coding module, a cross-graph gating fusion mechanism, a sequence modeling module and an expert hybrid prediction module. According to the method, long-term stable structured semantic association between concepts in declarative memory is modeled through a static knowledge structure diagram, a dynamic learning trajectory diagram based on batch reconstruction is designed to accurately capture an evolution rule of a behavior sequence in programmed memory, and on the basis, a cross-diagram gating fusion mechanism and a hybrid expert mechanism are introduced, so that the evolution rule of the behavior sequence in the programmed memory is accurately captured. And self-adaptive fusion and multi-path decision of double-graph features are realized.

The invention provides a dynamic adaptation method and system for a multi-mode power supply compatible charging pile system, and the method comprises the steps: building a multi-protocol physical layer compatible connection between a charging pile and a power supply module through a multi-communication interface layer, and receiving a data frame inputted by the charging pile; analyzing a baud rate, a check bit and an instruction set of the data frame, and identifying a charging pile protocol type and a power module protocol type; querying a matching mapping relationship between the charging pile protocol type and the power supply module protocol type in the dynamic adaptation matrix; if the matching mapping relation exists, data field conversion is executed, and a target protocol frame is generated; and if the matching mapping relation does not exist, starting a machine learning process, creating a temporary mapping rule, and effectively solving the problems of poor compatibility and high manual configuration cost of the charging pile and the power supply module in a traditional protocol mode through the modes of protocol analysis, intelligent matching and dynamic learning.

The invention discloses an autism detection method based on multi-mode collaborative embedding, and belongs to the technical field of medical image analysis and artificial intelligence. The method comprises the following steps: firstly, obtaining resting state functional magnetic resonance imaging data and non-imaging data of a subject; a Markov transition field is utilized to encode the time sequence into an image so as to retain dynamic features, and feature extraction is carried out through an efficient multi-scale attention module; then realizing effective fusion and semantic alignment of multi-modal information by adopting a three-level fusion architecture and a joint loss function; and then adaptively constructing a graph structure based on the fusion features, dynamically learning a node relationship by using a graph attention network, and completing a classification decision. According to the method, the defects of a traditional method in the aspects of dynamic feature modeling, multi-modal fusion and heterogeneous graph structure processing are effectively overcome, the autism detection accuracy and robustness are remarkably improved, and a reliable tool is provided for clinical intelligent diagnosis.

The invention relates to the field of data fusion analysis, in particular to a body feeling time sequence data fusion method and system based on a cross-modal attention mechanism, which can dynamically learn the correlation and weight of different modal data under different time steps and fuse to generate a unified body feeling state representation. The method specifically comprises the following steps: acquiring comprehensive body feeling data of a plurality of training actions corresponding to a monitoring target acquired by a heterogeneous sensor combination in real time so as to extract output synchronization deviation characteristics of various types of body feeling data corresponding to various training actions; evaluating a synchronization deviation characterization parameter of the comprehensive body feeling data to divide a synchronization deviation degree category of the comprehensive body feeling data; and adaptively evaluating, fusing and analyzing the comprehensive body feeling data. According to the method, dynamic alignment and information fusion of asynchronous time sequence data from different modes can be realized, and the robustness and accuracy of dynamic perception of human physiological signals are improved.

The disclosure includes a digital twin system. The digital twin allows for real time classification and ranking of data received from a distributed learning knowledge acquirer. The digital twin system ranks the data while the distributed learning knowledge acquirer is performing a drift evaluation. The distributed learning knowledge acquirer uses the ranked data as training data for discriminative AI models. The digital twin system offers more flexibility and precision in ranking the data. The digital twin system is a digital twin providing contextual active dynamic learning to the distributed learning knowledge acquirer's physical system. Digital twin system causes a model driven approach to allow for superior predictive capabilities by being able to examine large state spaces.

The invention relates to the technical field of employee evaluation, in particular to a high-risk industry multi-dimensional dynamic weight employee evaluation method, which comprises the following steps of: obtaining quantitative scores of three-level indexes of employees and normalizing the quantitative scores by constructing a multi-level evaluation index system; inputting the standardized score vector into a Transform-based weight learning model, dynamically learning an index weight through a self-attention mechanism, carrying out step-by-step aggregation through a multi-layer perceptron, and outputting a comprehensive score of the employee on each first-level index; inputting the comprehensive score and the historical background information into a pre-established large language model, and performing fusion analysis according to a structured template to generate a personalized evaluation text; and finally, based on the score, the text and the employee information, automatically synthesizing a visual comprehensive evaluation report containing the radar map, the information bar and the text. According to the invention, objectiveness, individuation and operability of employee evaluation are realized.

The invention provides an atlas updating method and device based on a large model, electronic equipment and a medium, information is extracted from large-scale text data by utilizing a natural language processing model, and the method mainly comprises the steps of data collection and preprocessing, entity extraction, relation extraction, atlas updating, model evaluation and iteration and the like. According to the method, an automatic crawler technology, multi-language text processing, transfer learning, visual data auxiliary entity extraction, relation extraction based on a pre-training language model and other technologies are comprehensively applied, and multiple beneficial effects of graph timeliness, multi-language adaptability, model generalization ability improvement, extraction accuracy improvement and the like are achieved. The atlas updating and continuous optimization process is optimized by means of an incremental updating strategy, a dynamic learning system, user feedback data, a conflict resolution strategy and the like. In conclusion, according to the method, the accuracy, the integrity and the user friendliness of the atlas are remarkably improved, meanwhile, the dependence on the annotation data is reduced, and the overall calculation efficiency and the model generalization ability are improved.

The invention discloses a financial heterogeneous data real-time cleaning and fusion method, and the method comprises the following steps: S1, obtaining multi-source, multi-time-dimension and multi-granularity financial heterogeneous original data, building a data access channel, and executing format standardization processing; and S2, constructing a cross-source, cross-time and cross-granularity financial index consistency constraint space, dynamically learning systematic deviation distribution of each data source in different market states based on the constraint space, and generating an implicit consistency deviation field. According to the financial heterogeneous data real-time cleaning and fusion method, a cross-source, cross-time and cross-granularity financial index consistency constraint space is constructed, systematic deviation distribution of data sources in different market states is dynamically learned based on the space, and an implicit consistency deviation field is generated; the problem of consistency among multi-source heterogeneous data is effectively solved, and the accuracy and reliability of the data are improved.

The invention discloses a drug-target interaction prediction method and system, and belongs to the technical field of biological information. The method comprises the following steps: firstly, acquiring molecular structure data of a drug and sequence and structure data of a target spot; then, respectively extracting molecular map structure characteristics and SMILES sequence characteristics of the medicine, and amino acid sequence characteristics and three-dimensional space structure characteristics of a target spot; further, taking drugs and targets as nodes, taking the fused multi-modal features as node features, and combining known interaction and similarity information to construct an initial heterogeneous graph; inputting the heterogeneous graph into a dynamic graph neural network, dynamically learning an inter-node connection weight by using a graph attention mechanism, and iteratively updating node representation through multi-layer message transmission to obtain depth feature representation of drugs and targets; and finally, splicing the depth features, inputting the depth features into a multi-layer perceptron classifier, and predicting the drug-target interaction probability. According to the method, through multi-modal feature fusion and dynamic graph structure learning, the prediction accuracy and robustness are remarkably improved.

The invention discloses a special glass surface microcrack detection system based on AI vision, and particularly relates to the technical field of optical imaging and computer vision, and the system comprises an optical image acquisition module, a primary detection module, a defect feature extraction module, a defect prediction module, a secondary detection module, a final judgment module and a database updating module. According to the method, multi-mode optical imaging and AI dynamic learning are combined, online detection is carried out on a special glass production line, and time sequence change rate analysis is introduced when defects are judged based on a multi-angle optical feature extraction result, so that accurate judgment of defect types is realized, and hard samples are automatically screened and pushed to experts for re-checking; according to the method, the product quality safety can be guaranteed to the maximum extent, meanwhile, production line maintenance and process optimization can be better guided, potential safety hazards are reduced, in addition, the efficiency bottleneck and cost rise caused by excessive dependence on manual reinspection can be avoided, and the long-term operation and maintenance cost is reduced.

The invention discloses an engineering cost real-time evaluation and early warning system based on multi-source data fusion and dynamic learning, and the system comprises a multi-source data collection module which is used for collecting construction process data, historical engineering data, market information data, and design drawing data; the preprocessing module is used for preprocessing and fusing the collected multi-source data to form a multi-dimensional feature vector with a unified timestamp; the hybrid neural network evaluation module is used for inputting the multi-dimensional feature vector into a pre-constructed hybrid neural network model for cost evaluation; the dynamic optimization learning module is used for carrying out online learning and dynamic optimization processing on the hybrid neural network evaluation module based on a transfer learning strategy; the block chain evidence storage module is used for storing key data generated in the evaluation process; and the progress monitoring and early warning module is used for monitoring the construction progress in real time, calculating progress deviation and triggering corresponding early warning. Accurate evaluation, real-time early warning and dynamic optimization of the construction cost are realized.

The invention relates to the technical field of virus traceability, and particularly discloses a coronavirus sample detection method which comprises the following steps: S1, acquiring high-throughput original sequencing data of a target sample; s2, inputting the original sequencing data into a dynamic learning type recognition model, and outputting each virus pedigree and the credibility of each pedigree; s3, performing noise perception variation detection on the original sequencing data; s4, aiming at each virus lineage, determining a variation point; s5, detecting a mixed infection indicator; s5, constructing a Bayesian network to determine the genetic relationship among the variation points, and generating a virus haplotype sequence; s6, calculating a genetic distance, and deducing a propagation path through a maximum likelihood method; and outputting a traceability report. The method solves the problem of frequency conflict when multiple pedigree coexist, is suitable for the situation that multiple pedigree viruses coexist to form mixed infection or co-infection, and avoids misjudgment of attribution of variation sites.

The invention discloses a double-section type drying temperature control system, which relates to the technical field of drying and comprises a drying state sensing module, a state recognition and analysis module, a thermal load prediction modeling module, an inter-section switching judgment module, a thermal inertia compensation control module, a parameter self-adaptive adjustment module and a control logic learning optimization module. The drying state sensing module is used for acquiring temperature, humidity and heat flow data in the material drying process to construct an inter-segment multi-dimensional temperature feedback matrix; and the state recognition and analysis module is used for executing manifold clustering analysis based on the inter-segment multi-dimensional temperature feedback matrix and extracting dry state distribution characteristics. According to the method, accurate identification and intelligent switching control of the drying state are realized through multi-point temperature and humidity sensing, manifold clustering and thermal load modeling; thermal inertia prediction and power decline regulation are introduced, so that temperature overshoot is effectively inhibited; and a closed-loop steady-state control and dynamic learning mechanism is constructed by combining self-adaptive adjustment of a proportional-integral-derivative controller, so that the drying quality and the operation stability are improved.