542 results about "Neural network learning" patented technology

The invention relates to the technical field of environment monitoring, and provides a collaborative control method and device for a measurable, displayable, adjustable and controllable environment system, and the method comprises the steps: collecting an original data set of a monitoring region, carrying out the preprocessing, and constructing a multi-dimensional environment data knowledge graph containing time-space correlation attributes; through a neural network learning environment parameter and health risk space-time correlation characteristic, outputting a healthy human settlement environment comprehensive evaluation result based on a fuzzy comprehensive evaluation matrix of adaptive threshold adjustment; inputting the comprehensive evaluation result of the healthy human settlement environment into a Kriging interpolation algorithm, generating an indoor four-dimensional space-time dynamic thermodynamic diagram in combination with environment parameter space-time distribution characteristics, generating a health risk prediction thermal layer based on risk weight superposition, and forming a visual composite atlas; outputting a regulation and control strategy by taking energy consumption, equipment cost and a health risk value as optimization targets; and controlling the household cooperative equipment. According to the invention, predictive control is carried out on the environmental equipment, and the accuracy and effectiveness of environmental regulation and control are improved.

The invention provides a hybrid representation learning-based omni-directional motion control method for a quadruped robot in a three-dimensional environment, aiming at solving the problem that an existing motion control method for the quadruped robot is limited by a narrow forward view angle and external perception noise of the robot, so that the omni-directional motion capability in the three-dimensional environment is difficult to obtain. Comprising the steps of designing various input and output contents of a first-stage end-to-end neural network learning framework, designing a training process including mixed representation learning based on a reinforcement learning framework, constructing a specific neural network model of each module, and constructing a quadruped robot simulation motion control training environment. The method has the advantages that network input and output content beneficial to efficiently learning the relation between actual observation of the robot and the environment truth value is provided, and a mixed representation learning method including supervised learning and comparative learning is introduced, so that the robot can effectively deduce environment information beyond the visual field range of the robot in the training process.

The invention discloses a trajectory prediction method and device based on a time-space diagram neural network learning model, and the method comprises the steps: carrying out the data cleaning and trajectory compression based on to-be-classified original trajectory data, and outputting a compressed grid sequence; based on the grid sequence, static track information classification is executed, and track static attributes are extracted; dynamic space-time image track classification is executed, and track dynamic space feature attributes are extracted; the method comprises the following steps: pre-constructing a directed weighted network based on a driving road section connectivity attribute feature, and completing training of a trajectory prediction model based on a trajectory static attribute, a dynamic space feature attribute and the directed weighted network; and based on the trajectory prediction model, inputting spatial-temporal characteristic data of a trajectory to be predicted, and obtaining an optimal trajectory path prediction result. According to the method, uncertain track geographical background information on the road network can be effectively utilized, and meanwhile, errors caused by map matching are avoided.

An air conditioner load monitoring and anomaly detection method and system. The method comprises: acquiring user power consumption data and meteorological data, the user power consumption data and the meteorological data comprising the total power of an electricity meter, the power of an air conditioner, and the outdoor temperature, preprocessing the data, and computing electrical characteristics of the user power consumption data; on the basis of the electrical characteristics, learning by using a two-tower neural network so as to obtain operating rules of the air conditioner, and extracting an air conditioner operation curve; on the basis of the air conditioner operation curve, computing electrical characteristics and energy consumption characteristics of the air conditioner, and performing cluster analysis by using a K-means clustering algorithm; and, on the basis of clustering results, comparing the clustering results with clustering results of collected electrical data of the air conditioner under normal and abnormal operating conditions to determine whether the air conditioner is in an abnormal state, so as to implement detection of anomalies in the operation of the air conditioner, thereby improving the precision of load decomposition of the air conditioner, reducing application costs, providing a decision-making reference for equipment maintenance personnel, and prolonging the service life of the equipment.

The invention provides a 6061 aluminum alloy rheological behavior constitutive relation prediction method based on physical information machine learning, and the method comprises the steps: carrying out the interception of a yield point of experimental stress-strain curve data, retaining a plastic segment, and combining with a corresponding experimental condition, thereby obtaining a data set for the training of a physical information machine learning model; performing ANN model training on the data set to obtain a data driving loss item between an ANN model prediction stress value and an experimental stress value; secondly, a physical constitutive model is embedded into a loss function of an ANN model, a physical law serves as a constraint condition to be introduced into neural network learning, and an additional physical driving loss item is formed by calculating the difference between neural network prediction and physical results described by the constitutive model; and establishing a loss function by combining the physical driving loss item and the data driving loss item according to different weights, training a physical information machine learning model according to the loss function, and outputting a final prediction stress-strain curve.

The invention provides a wind power prediction method and system, and relates to the field of wind power prediction, and the method comprises the steps: collecting weather forecast data and actually measured wind power data at all historical moments, and constructing a historical data set; using a BP neural network to learn the relationship between each meteorological factor in the historical data set and the wind power to obtain a power prediction model, and performing preliminary prediction on the wind power; based on the historical data set, establishing an error inference model by using a Seq2Seq-AM algorithm, and performing prediction error inference on the preliminary prediction result; and establishing an error correction discrimination model by using an auto-encoder algorithm, and discriminating whether to use the prediction error to correct the preliminary prediction result so as to obtain the final wind power of the distributed wind power plant. Based on the Seq2Seq-AM and the auto-encoder, the prediction error is deduced, and whether the prediction error is used for correction or not is judged, so that accurate prediction of the wind power of the target station is realized.

The invention discloses a personalized commodity recommendation optimization method based on deep learning. The method comprises the following steps: S1, constructing a data set; s2, constructing a multi-dimensional knowledge graph, and embedding the multi-dimensional knowledge graph into a low-dimensional space by using a knowledge graph embedding algorithm; s3, constructing a user interest track graph by using the time sequence graph model, and generating a user interest vector based on capturing long-term and short-term interest changes; s4, obtaining an optimized user interest vector by using an improved moth fire suppression algorithm; s5, based on the optimized user interest vector, combining a graph neural network to learn relationships between users and commodities, between users and between commodities; and S6, generating a personalized commodity recommendation list according to a graph neural network learning result and real-time behavior feedback of the user. According to the method, a multi-dimensional knowledge graph, time sequence graph modeling, a moth fire suppression algorithm, a graph neural network and the like are fused, and optimization of personalized commodity recommendation is realized.

The invention discloses an intelligent operation and maintenance distributed cloud native application computing method, and relates to the technical field of cloud computing, and the method comprises the steps: collecting heterogeneous data of an application layer, a middleware layer, a container layer and a service governance layer, constructing a dynamic multilayer dependency graph, and explicitly modeling a cross-layer dependency relationship and implicit dependency caused by resource competition; based on the spatio-temporal features of the spatio-temporal diagram neural network learning map, predicting a cascade fault risk and outputting a high-risk sub-graph; candidate operation and maintenance operations are generated through reinforcement learning, indirect negative influences of the operations on non-target components are evaluated in combination with a causal inference model, and risk filtering operations are verified through a multi-dimensional safety threshold value; and after the safety operation is executed, continuously optimizing the model and the atlas through a feedback closed loop. According to the method, deep insight, accurate risk prediction and self-adaptive decision making of side effect minimization can be carried out on a complex dependency relationship, and the system stability and the operation and maintenance efficiency are remarkably improved.

The invention provides a winding insulation residual life prediction method and system, and relates to the technical field of winding insulation residual life prediction. According to the method, firstly, a data set is divided through segmentation threshold values, and on the basis of an Arrhenius equation, a segmentation mathematical model is constructed to describe the degradation process of an insulating material; then, a boundary penalty function of mathematical model parameters is introduced, a neural network model is constructed, the mathematical model and a neural network are jointly optimized, and meanwhile, a weighted loss function is designed in combination with statistics of degradation data; and finally, performing progressive training on the neural network by using each sub-data set, and generating a prediction result through superposition of a theoretical aging component of the mechanism model and a residual compensation component learned by the neural network. According to the method, segmented dynamic modeling and neural network learning are combined to realize accurate prediction of the non-linear time-varying trajectory of the insulation state characteristic quantity, and the characterization capability of the model on the complex degradation process of the insulation material is remarkably improved.

The invention relates to the technical field of mutual inductor tests, and provides a mutual inductor test abnormal data automatic filtering analysis method and system, and the method comprises the steps: obtaining magnetic field data, electric field data and thermal field data, respectively carrying out the feature extraction and weighted fusion, and obtaining a multi-physical field fusion feature vector; constructing an electromagnetic induction chain type propagation graph, calculating the weight of an edge in the graph, constructing a magnetic flux conservation constraint graph convolutional neural network, learning propagation and evolution modes of transformer test abnormity, and obtaining a coupling abnormity feature vector; the coupling anomaly feature vectors are classified, normal data, single physical field abnormal data and multi-physical field coupling abnormal data are recognized, and corresponding data are judged as abnormal data and filtered; and carrying out physical mechanism analysis on the filtered abnormal data to obtain an abnormal detection report. According to the invention, high-precision automatic identification, filtering analysis and physical mechanism traceability diagnosis of transformer test abnormal data are realized.

An image classification model, e.g., a neural network model, may be trained on a set of training medical imaging exams each including a training report and a training medical image. A model generation module or device may, for each of the training medical imaging exams: use finding item criteria to reorganize text of the training report into a list of finding items, each associated with text extracted from the training report text, use natural language processing to analyze the resultant text associated with each finding item to determine an associated classification of each finding item, store, in a training dataset, the training medical imaging exam, the associated finding items, the matching text, and the classifications resulting from the analysis of the matching text. The trained neural network learns to output, for other medical imaging exams, one or more finding items and an associated classification data.

The invention discloses an energy-saving ground source heat pump control method and system, and the method comprises the steps: building a ground source heat pump system dynamic model which comprises an indoor heat balance model, a ground source heat pump unit model and a soil heat exchange model, obtaining a topological network through coupling, designing a distributed control and fault-tolerant operation strategy according to the topological network structure, and carrying out the control of the ground source heat pump system. Each node is locally controlled according to the position and adjacent node information, a state space model is constructed, an optimal control action sequence is solved through quadratic programming to balance energy consumption and an indoor and outdoor temperature deviation objective function, the optimal operation state is obtained, and the mapping relation between the optimal operation state and environment parameters is learned through a neural network; a ground source circulation flow control algorithm is designed, and circulation pump flow distribution and unit working parameter automatic adjustment are achieved. According to the method, the control precision and stability of the ground source heat pump system can be effectively improved, reliable operation of the ground source heat pump system under complex working conditions is ensured, the energy utilization efficiency is remarkably improved, energy conservation and consumption reduction are achieved, and meanwhile good interpretability is achieved.

The invention discloses an intelligent identification method for a rock tension-shear rupture mechanism based on acoustic emission RA-AF parameters, and belongs to the technical field of intelligent identification. The method comprises the following steps: carrying out indirect tension test and variable-angle shear test on rock, acquiring RA-AF data of the rock under tension and shear loads by adopting an acoustic emission technology, and respectively constructing RA-AF standard databases corresponding to tension and shear fracture mechanisms based on a DBSCAN clustering algorithm; wherein 4 / 5 is used as training set data, and 1 / 5 is used as test set data. A probabilistic neural network is used for learning RA-AF data feature information of a pulling and shearing fracture mechanism, an intelligent classification model of the pulling and shearing fracture mechanism is established, the accuracy rate of the model is detected, and the model which reaches the detection standard through adjustment is applied to quantitative recognition of the fracture mechanism of the same rock under different static loads. According to the method, the data processing workload of workers can be reduced, and the classification efficiency is improved.

The invention belongs to a robot control method, and provides a human-guided robot-environment interaction self-adaptive control method and a related device for solving the technical problems that an existing human-guided robot-environment interaction control method is insufficient in motion stability, low in human operator safety and low in control precision. And estimating an optimal impedance parameter through a state space model, and learning a human-guided reference trajectory through a neural network in combination with the optimal impedance parameter. Then, in combination with a robot Jacobian matrix and a human-guided reference trajectory, a reference joint speed and an estimated acceleration of a joint space are obtained through closed-loop inverse kinematics, then network parameters of a robot dynamic model are obtained through approximation of an uncertain dynamic model, and finally, the robot dynamic model is obtained through combination with the reference joint speed and the network parameters of the joint space. And the mechanical arm control torque is obtained through the offset width fuzzy neural network. Accurate and compliant control under uncertain motion and dynamics conditions is realized.

The invention provides a vehicle fault diagnosis method and device based on multi-source data fusion and a storage medium. The method comprises the following steps: collecting real-time state data and real-time alarm data of a target part, and obtaining external environment data and map navigation data; preprocessing the data, fusing the preprocessed data to obtain a multi-source data set target feature, inputting the target feature extracted from the multi-source data set into an adaptive neural fuzzy inference system, mapping the target feature to a plurality of fuzzy sets, and fuzzifying by using a member function; calculating the rule weight of each target feature to reflect the fuzzy intensity; and training an inference system by adopting a neural network learning algorithm, adjusting a member function and a rule weight, and converting fuzzy output into a specific numerical value or a fault level by utilizing an output member function in a defuzzification process so as to judge the health state of the part. According to the invention, high-interpretability fault diagnosis output is provided, the accuracy is improved, and fault diagnosis can be carried out in a wide vehicle operation environment.

The invention discloses a molecular structure characterization-based drug-drug interaction prediction method, which comprises the following steps of: acquiring drug molecule data and drug-drug interaction data, performing standardized preprocessing on the data, and constructing a drug molecule map according to the drug molecule data; inputting the drug molecule map into a model, extracting drug molecule multi-scale features through a multi-scale map convolutional network, and fusing the drug molecule multi-scale features through a dynamic attention mechanism to obtain drug molecule features; based on the drug molecular characteristics, through a full-connection neural network, learning the relationship between the drug molecular characteristics and the drug-drug interaction, and obtaining the prediction probability of the drug-drug interaction; and training the model by adopting a joint loss function according to the drug-drug interaction data and the prediction probability, and applying the trained model to drug-drug interaction prediction. And through the multi-scale image convolutional network, dynamic attention fusion and joint optimization, the prediction accuracy is significantly improved.

The invention discloses a refrigerator defrosting control method based on BP (Back Propagation) neural network and fuzzy control coupling, which is characterized by comprising the following steps: realizing deep perception of dynamic characteristics of a refrigerating system through data-driven modeling: learning actual data signals of a compressor and an evaporator by the BP neural network, and indirectly estimating recessive parameters such as refrigerant charge and frost layer form distribution; and the fuzzy controller dynamically adjusts a defrosting strategy according to a neural network prediction result, and collaborative optimization of refrigeration cycle parameters and the frosting process is achieved. Environment parameters are learned online through a neural network, and the fuzzy rule base is optimized in real time; the neural network directly maps bottom layer physical characteristics such as a compressor operation mode and evaporator structure parameters, and outputs a dynamic compensation signal; and the model is continuously iterated by accumulating historical data to adapt to regional / seasonal differences.

The invention discloses a high-voltage cable outer metal sheath grounding defect on-line monitoring method and monitoring system, which can work under the condition that a cable is not powered off, realizes on-line monitoring, and realizes on-line monitoring through the cooperation of a current sensor, a voltage sensor, a single-chip microcomputer and an upper computer. And the equipment can quickly detect when a fault occurs and is convenient for maintenance workers to timely eliminate obstacles, so that the risk of major electric power accidents is reduced. The algorithm is similar to an improved neural network algorithm and is used for diagnosing the defects of the outer metal sheath of the cable, the weight is obtained through training learning, and finally the defects of the outer metal sheath of the high-voltage cable are determined. The protection content covers each component of the technical scheme, and comprises a monitoring method, a device structure, a sensor technology, a data processing algorithm, a neural network learning algorithm and the like.

An image classification model, e.g., a neural network model, may be trained on a set of training medical imaging exams each including a training report and a training medical image. A model generation module or device may, for each of the training medical imaging exams: use finding item criteria to reorganize text of the training report into a list of finding items, each associated with text extracted from the training report text, use natural language processing to analyze the resultant text associated with each finding item to determine an associated classification of each finding item, store, in a training dataset, the training medical imaging exam, the associated finding items, the matching text, and the classifications resulting from the analysis of the matching text. The trained neural network learns to output, for other medical imaging exams, one or more finding items and an associated classification data.

The invention is suitable for the technical field of automatic driving vehicle control, and provides an automatic driving track rapid tracking control method based on online neural network learning, which comprises the following steps: firstly, designing an adaptive model prediction controller, then making a data set and training a fuzzy radial basis function neural network; then, online learning of the fuzzy radial basis function neural network is carried out; and finally, accelerated optimization of the training process of the fuzzy radial basis function neural network is carried out. The automatic driving track rapid tracking controller based on online neural network learning provided by the invention has the remarkable advantages that the neural network model is trained by using control data of MPC (model predictive control), and a nonlinear relation is fitted to approximate the MPC, so that the calculation efficiency is effectively improved, and the rapidity of the controller is improved; the controller can collect real-time data, update a data set and learn new neural network weight parameters online in the process of tracking a reference trajectory, and shows better adaptability and precision compared with a traditional model prediction controller.

The invention relates to the technical field of remote sensing image change detection, in particular to a remote sensing image unsupervised change detection method and system for joint learning of time domain difference and consistency, and the method comprises the steps: constructing a multi-temporal remote sensing image sample through data enhancement transformation; constructing a training loss function based on time domain contrast learning and space contrast learning of consistency constraints, and training the image change detection model based on the training loss function and by using the multi-temporal remote sensing image sample; and inputting the multi-temporal remote sensing image of the target area into the trained image change detection model, and obtaining change information in the multi-temporal remote sensing image of the target area by using the image change detection model. According to the method, difference and consistency among multi-temporal observation are captured in combination with comparative learning, pixel-level semantic representation is obtained in combination with a visual basic model, sparse and compact change mapping learned by a neural network is promoted by utilizing grid sparseness loss, and unsupervised change detection of a high-resolution remote sensing image can be accurately and efficiently realized.

The invention relates to an unmanned aerial vehicle track prediction method based on multi-feature LSTM, and belongs to the technical field of data processing, and the method comprises the following steps: 1, constructing a six-degree-of-freedom global motion state matrix of a target unmanned aerial vehicle; 2, performing joint estimation to obtain a three-dimensional wind field vector acting on the target unmanned aerial vehicle, and extracting the real-time speed direction of the target unmanned aerial vehicle; 3, fusing the historical state sequence, the three-dimensional wind field vector and the speed direction unit vector, and constructing a multi-dimensional time sequence feature vector; 4, the LSTM neural network learns a nonlinear maneuvering mode of the target unmanned aerial vehicle under the influence of the wind field through a door control mechanism of a forgetting door, an input door and an output door; and step 5, outputting the position probability distribution of the target unmanned aerial vehicle at a plurality of time points in the future through the LSTM neural network. The method has the advantages that the space-time cone representing the future trajectory is formed, and trajectory prediction of the target unmanned aerial vehicle is completed.

The invention provides a capacitor voltage transformer error online detection method and system, and the method comprises the steps: collecting a secondary output signal of a three-phase capacitor voltage transformer, and carrying out the data preprocessing; based on the collected data, constructing a deep neural network learning model based on a stacked auto-encoder, and extracting deep features of the data through multilayer nonlinear transformation to obtain reconstructed data; taking the original data and the reconstructed data as input data, respectively carrying out co-integration analysis, and calculating a co-integration matrix and an equalization error; combining the reconstruction data with the equilibrium error, constructing a composite statistical magnitude and calculating a statistical magnitude threshold value; and collecting online measurement data of the capacitor voltage transformer, calculating monitoring statistics and carrying out abnormity diagnosis. The technical problems that disturbance components and error information in measurement data are difficult to distinguish, non-linear and non-stationary components in measurement errors of the capacitor voltage transformer are difficult to analyze, and the online detection accuracy and real-time performance of the measurement errors are low are solved.

The invention relates to the field of intelligent control of marine navigation, and discloses a formation dynamic interception path planning method and device based on a differential game theory, and the method comprises the steps: constructing a differential game model comprising a formation and intercepted ships; designing revenue functions for the two parties; obtaining a Nash equilibrium strategy by solving a Hamiltonian-Jacobian-Bellman equation of the game, and solving an individual optimal control instruction of each formation member; and fusing state information among formation members through a consistency protocol, adjusting individual instructions to meet anti-collision and formation cooperative constraints, and generating a final control instruction. According to the method, historical adversarial data is learned by using the neural network, and the weight of the revenue function is adaptively and dynamically adjusted to cope with different tactical scenes; and the whole decision-making process is set under a model prediction control framework for rolling optimization, so that the real-time performance, the collaboration and the intelligent level of interception path planning are greatly improved, and the interception efficiency is effectively improved.

The invention relates to the technical field of electric vehicle charging pile planning, in particular to an intelligent charging pile layout optimization method based on a multi-scale space-time diagram neural network. Comprising the following steps: S1, constructing a heterogeneous dynamic graph which represents a potential charging pile position node set, epsilon t represents a time-varying edge set, At represents a time-varying adjacent matrix, and Xt represents a node feature matrix; s2, calculating a self-adaptive adjacency matrix with a specific relationship, and calculating a multi-type dynamic relationship between capture nodes of the self-adaptive adjacency matrix based on the constructed heterogeneous dynamic graph model; s3, updating a structure bias matrix, and capturing dynamic change characteristics of the network; s4, calculating distance measurement between nodes, and fusing geographic information and semantic information; s5, through graph neural network learning, based on the constructed heterogeneous dynamic graph and the calculated adaptive adjacency matrix, executing a graph neural network learning process, and extracting node space-time representation; and S6, predicting a future charging demand based on node representation obtained by graph neural network learning, and generating based on the future charging demand.

According to an aspect of the present disclosure, a method executed by an information processing apparatus in order to cause a neural network to learn a Tth task corresponding to a Tth learning set is provided. The method includes: for each of a plurality of units, determining an importance degree of the unit in the Tth task; for each of a plurality of layers, determining dissimilar tasks from among a first task to a (T−1)th task, the dissimilar tasks being not similar to the Tth task in terms of behaviors in the layer; and in learning that uses the Tth learning set, suppressing updating of weight parameters of the plurality of units included in the plurality of layers in accordance with importance degrees in the dissimilar tasks determined for each of the plurality of layers.

The invention discloses an anti-parasitic drug target affinity prediction method based on multi-modal feature fusion. The method comprises the following steps: 1, acquiring an anti-parasitic drug related data set; 2, constructing a multi-scale convolutional neural network, and obtaining sequence features of drugs and proteins; 3, constructing a graph convolutional neural network, and obtaining structural features of the medicine; 4, constructing a graph sampling and aggregation network, and obtaining structural features of the protein; 5, constructing a self-adaptive gating network, fusing two modal characteristics of the drug and the protein, and generating drug-protein joint expression; 6, carrying out feature alignment on the two modal features of the drug and the protein by utilizing comparative learning; and 6, updating model parameters according to an output result of the prediction layer. According to the method, multiple neural networks are adopted to learn sequence features and structural features of drugs and proteins, and information of two modal features is balanced and integrated by utilizing comparative learning, so that drug-protein interaction expression is enhanced, and the accuracy of predicting the affinity of anti-parasitic drugs and proteins is improved.

The invention discloses an electro-hydraulic system adaptive neural network preset time control method, which integrates a neural network real-time learning technology, innovatively introduces a preset time performance function, and designs an adaptive neural network preset time performance controller. For the electro-hydraulic system position tracking control problem, it can be ensured that system output transient performance and steady-state performance converge to a specified performance index range within a preset time, safe and reliable operation of the system is ensured, the unknown dynamic state of the system can be learned in real time through a neural network, high-precision motion control performance is achieved, and the control precision is improved. And the problem of differential explosion in traditional backstepping control of the electro-hydraulic system can be avoided, and the influence of measurement noise on the control precision is reduced.

The invention belongs to the field of track planning and intelligent control of a reusable vehicle, and relates to an intelligent track planning method for a cruise section of the reusable vehicle. The method comprises the following steps: constructing a three-degree-of-freedom centroid motion dynamics model of a cruise section vehicle, generating an offline optimal trajectory sample library covering wide working conditions based on a Gaussian pseudo-spectral method, designing a mapping relation between deep neural network learning task parameters and trajectory features, and outputting a high-quality trajectory planning initial value online. The distribution of collocation points is dynamically adjusted in combination with a self-adaptive collocation point method, a nonlinear programming problem is iteratively solved with a high-quality initial value as a starting point, and efficient and accurate optimization of the trajectory is achieved. Simulation results show that the method significantly reduces the sensitivity of online optimization to initial guess, still has fast response ability and high robustness under complex constraint and uncertainty working conditions, effectively improves the autonomous trajectory planning ability of the cruise section of the reusable vehicle, and ensures flight reliability and control precision.

The invention relates to a turbine blade design method based on a neural network, and the method comprises the steps: carrying out the modeling of turbine blade structure parameters, and obtaining a parameterized model; based on the parameterized model, performance evaluation calculation is conducted on the turbine blade structure, and performance evaluation results are obtained; selecting a neural network, and learning the turbine blade structure parameters and the performance evaluation result through the neural network to obtain a mapping relation; and designing the turbine blade to be designed based on the mapping relation. According to the method, the mapping relation between the turbine blade structure parameters and the performance evaluation result is learned through the neural network, and the design efficiency and accuracy of the turbine blade can be improved.