67results about How to "Approximation effect is good" patented technology

The invention discloses a nonlinear self-adaption control method of near-space hypersonic vehicle (NHV), which belongs to a flight control method in the technical field of aerospace. The control method mainly comprises three control law parts: a nominal nonlinear generalized predictive control law (NGPC), a B-spline recursive functional linkage network (BRFLN) self-adaption control law, and a robust control law of gain self-adaption adjustment. The invention integrates the simplicity of the NGPC method and the effectiveness of dynamic uncertainty of the BRFLN learning, targets the immeasurable dynamic uncertainty and fast interference of an attitude system in the flight of the NHV, brings favorable learning effect, and realizes the nonlinear accurate control to the attitude angle.

The invention discloses a remote sensing image classification method based on an attention mechanism deep Contourlet network, and the method comprises the steps: building a remote sensing image library, and obtaining a training sample set and a test sample set; then, setting a Contourlet decomposition module, building a convolutional neural network model, grouping convolution layers in the model in pairs to form a convolution module, using an attention mechanism, and performing data enhancement on the merged feature map through a channel attention module; carrying out iterative training; performing global contrast normalization processing on the remote sensing images to be classified to obtain the average intensity of the whole remote sensing images, and then performing normalization to obtain the remote sensing images to be classified after normalization processing; and inputting the normalized unknown remote sensing image into the trained convolutional neural network model, and classifying the unknown remote sensing image to obtain a network output classification result. According to the method, a Contourlet decomposition method and a deep convolutional network method are combined, a channel attention mechanism is introduced, and the advantages of deep learning and Contourlet transformation can be brought into play at the same time.

The invention discloses an intelligent simulation technology, and particularly relates to an improved three-dimensional model voxelization-based inner sphere construction method. An inner sphere construction mode and a hierarchical structure clustering method are improved by steps of surface voxelization of a model, interior voxelization of the model, basic construction of inner spheres, clustering of the inner spheres for the construction of a hierarchical sphere structure and the like. The method has the advantages that the radiuses of the inner spheres are determined during initialization without subsequent processing, so that the method is easy to operate, and the number of the inner spheres can be reduced; and a significant clustering effect can be finally achieved. The method has broad prospect when being applied to the detection of collision between models.

The invention relates to a photovoltaic power generation forecasting method based on a GRNN. According to the method, firstly factors such as solar terms, weather, and sunlight are considered to establish a photovoltaic curve pattern, then a photovoltaic power generation forecasting model based on the GRNN is provided, and a solution algorithm design is conducted. The method has the following advantages that weight does not need to be amended through error reverse calculation in the training process of the GRNN, and the transfer function can be adjusted by just changing a smoothing parameter Sigma, so that the training time is reduced, and the network learning speed is increased; a GRNN forecasting model is high in nonlinear mapping ability and good in approximation performance, and has good robustness, and thus is suitable for processing unstable data; the photovoltaic power generation forecasting technology of the GRNN obviously improves the forecasting accuracy; the forecasting result can provides decision information for grid photovoltaic scheduling, and has great significance in guaranteeing safe operation of the grid.

The invention discloses a method for predicting a user participation behavior of a hot topic by an improved RBF (Radical Basis Function) neural network, and belongs to the field of computer network information technology analysis. Firstly, the situation that a neural network can take a good fitting effect for a complex nonlinear relationship among user behaviors is considered, and further a user participation behavior prediction model is built by adopting the RBF neural network; secondly, a mapping relationship between a user attribute and the participation behavior has uncertainty, and a cloud theory is introduced for optimizing an activation function of a hidden layer in an RBF; and finally, topic popularity is subjected to exponential function model-based parameter fitting by utilizing time discretization and time slicing methods for a characteristic that the participation behavior of a user is changed with time, so that a topic popularity change trend is obtained.

The invention relates to the technical field of risk control of the Internet financial industry, in particular to a credit evaluation method for optimizing a generalized regression neural network based on a grey wolf algorithm. The method comprises six steps, and compared with common BP and RBF neural networks, the method has the advantages that GRNN selected by the method is strong in nonlinear mapping capability, good in approximation performance and suitable for processing unstable data. The method has the advantages of being good in generalization ability, high in fitting ability, high intraining speed, convenient in parameter adjustment and the like, and compared with common optimization algorithms such as genetic algorithms and particle swarms, the grey wolf algorithm is few in parameter and simple in programming, and has the advantages of being high in convergence speed, high in global optimization ability, potential in parallelism, easy to implement and the like. The grey wolfalgorithm is adopted to optimize the GRNN network model, the prediction precision and stability are high, the defects that the GRNN prediction result is unstable and is very likely to fall into the local minimum value are effectively avoided, and rapid and accurate online real-time prediction of the credit score of the application user is achieved.

The invention discloses an intelligent granary environment safety monitoring system based on a field bus. The system is composed of a granary environment parameter collection platform based on a CAN field bus, and a granary environment safety evaluation subsystem. The system realizes the intelligent detection of granary environment parameters and the intelligent evaluation of granary environment safety. Many problems still existing in the granary environment caused by the unreasonably designed and poor traditional granary environment multi-parameter detection equipment, the incomplete detection system and the like are solved. Based on the nonlinearity and large lag of granary environment parameter changes and the large area and complex structure of the granary environment, the defects of inaccuracy, low reliability and the like of a granary environment monitoring system are overcome, accurate detection and reliable classification of granary environment parameters are realized, and therefore, the accuracy and robustness of granary environment parameter detection are greatly improved.

The invention discloses a primary direction neural network system with a four-layer feedforward structure. The primary direction neural network system comprises an input layer, a first hidden layer, a second hidden layer and an output layer, wherein the input layer comprises D neurons, the first hidden layer comprises K groups of neurons, each group of the neurons comprises three neurons, the second hidden layer comprises K neurons, the output layer is a neuron, and both the D and the K are natural numbers; the input layer is used for receiving the D dimensional vector, and each neuron correspondingly receives one component in the D dimensional vector; the first hidden layer is used for mapping the D dimensional vector received from the input layer to the neurons in the second hidden layer, and each group of the neurons in the first hidden layer corresponds to one neuron in the second hidden layer; the second hidden layer is used for mapping the 3K dimensional vector received from the first hidden layer to the neurons in the output layer; and the output layer performs biased w0 linear weighting to the result of the second hidden layer and then outputs the result. The invention overcomes the extremely small difficulty of being trapped in local and the sensitivity to the noise.

The invention relates to an optimized-neural-network gesture-recognition human-computer interaction method based on GL. Gestures of an operator are collected through a Data Glove14 Ultra data glove are collected, then the collected data is processed, an optimized neural network based on the GL algorithm is subjected to network learning training, finally, the trained algorithm is applied to gesture recognition based on the data glove, and teleoperation of the robot is achieved.

The invention discloses a greenhouse environment multi-parameter intelligent monitoring system based on the Internet of Things. The system is composed of a watermelon greenhouse environment parameteracquisition platform based on a ZigBee network and a watermelon greenhouse environment microclimate factor evaluation subsystem. The system realizes intelligent detection of watermelon greenhouse environment temperature and evaluation of microclimate environment factors. The problems that an existing watermelon greenhouse environment monitoring system does not accurately detect watermelon greenhouse environment parameters and evaluate environment factors according to the characteristics of nonlinearity, large lag, complex greenhouse environment parameter changes and the like of greenhouse environment parameter changes, and therefore the accuracy of predicting and evaluating the watermelon greenhouse parameters is improved are effectively solved.

The invention discloses a micro-gyroscope double-feedback regression neural network sliding-mode control method. On the basis of a conventional single-layer regression neural network, the method adds an output regression item, and carries out simulation on an MEMS micro-gyroscope model. A designed sliding-mode neural network controller can guarantee that a system tracking error is converged to be zero. Moreover, the method employs a double-feedback regression neural network and a common regression neural network, and the results of a neural sliding mode controller of the common regression neural network indicate that the double-feedback regression neural network is better in approaching effect and is more stable in internal state. The neural network designed in the invention can set a central vector and the initial value of a base width randomly, and the central vector and the base width can be stabilized to be optimal values along with a designed adaptive algorithm according to different inputs.

The invention discloses a dynamic GM (1, 1) model based on a cubic spline interpolation. The dynamic GM (1, 1) model is applied for a thought of a piecewise polynomial interpolation and an overall trend of dynamic prediction time series data. The dynamic GM (1, 1) model based on the cubic spline interpolation theoretically analyzes a background value of a GM (1, 1) model. Firstly, a through of a cubic spline interpolation is put forward. Then, a method of combination of a piecewise linear interpolation function and a cubic spline interpolation formula to construct a new class grey prediction model. A constructing process of the background value can be improved so as to overcome the shortcomings of an existing grey improving model and provide a new means to improve prediction accuracy. Lastly, the model can be utilized to implement prediction. The dynamic GM (1, 1) model based on the cubic spline interpolation is scientific in concept, simple in calculation, small in workload and high in prediction accuracy. The dynamic GM (1, 1) model based on the cubic spline interpolation has a better use value and a wide application prospect in a prediction technique field.

The invention discloses a river flow monitoring data quality control method based on a chaotic neural network. The method comprises the following steps: a, sorting historical flow data in a chronological order to obtain time series data; b, normalizing the time series data; c, calculating the optimal time lag [tau] and the optimal embedding dimension m, and converting the one-dimensional time series data into multi-dimensional spatial sample data; d, proportionally dividing the multi-dimensional spatial sample data into multi-dimensional spatial training sample data and multi-dimensional spatial test sample data; e, using the multi-dimensional spatial training sample data to train and construct a GMDH neural network, using the multi-dimensional spatial test sample data to test the GMDH neural network so as to obtain a GMDH neural network model; f, detecting the abnormal value of river flow monitoring data; g, inspecting the completeness of the data within 24 hours of a day; and h, storing the revised river flow monitoring data is stored in a database. The method is able to improve the quality of river flow monitoring data.

The invention discloses a global sliding mode control method for an active power filter based on a regression neural network of a double hidden layer, characterized in that the method comprises the following steps: 1) establishing a mathematical model of the active power filter; 2) establishing a global sliding mode controller for the active power filter based on the regression neural network of the double hidden layer, designing a control rule which is used as a control input for the active power filter; and 3) designing an adaptive rule, and verifying the stability of the global sliding modecontroller of the active power filter based on the regression neural network of the double hidden layer based on the Lyapunov function theory. The global sliding mode control method for the active power filter based on the regression neural network of the double hidden layer has advantages of: improving the approximation accuracy and generalization ability of the network, reducing the number of network parameters and weights, speeding up the network training speed; being able to store more information; having a better approximation effect; and being able to improve the compensation current tracking accuracy and system robustness of the active power filter system in the presence of parameter perturbation and external interference.

The invention discloses a metacognitive fuzzy neural network-based active power filter sliding mode control method. The method includes the following steps that: S1, an active power filter dynamic equation is established; and S2, a control law is designed to be an equation described in the descriptions of the invention, and f<^> is an estimation value obtained by using a metacognitive fuzzy neuralnetwork f(x). According to the method of the present invention, the metacognitive method is adopted to adjust the structure of the fuzzy neural network online; the structure of the fuzzy neural network is dynamically adjusted according to a tracking error design rule addition, parameter update and rule deletion algorithm; and therefore, the compensation current tracking performance and system robustness of an active power filter system under parameter perturbation and external interference can be improved.

The invention provides an image self-adaptive grid generation variational method, and relates to image approximation and piecewise polynomial approximation. The method comprises: S1, inputting an image, and setting related parameters; S2, generating original triangular mesh partition; S3, according to an energy function and related gradient and Hessian matrix information, calculating a new positions of vertexes of triangular meshes, and moving vertexes to the new positions; S4, updating connection relations of the triangular meshes whose vertexes are moved; S5, executing the step S3 to S4 for a plurality of times in a circulating manner, until iterations reach J, and triangular grids whose partition structure is very close to an original image being generated in an image region; and outputting optimal triangular grids and a corresponding approximating polynomial set. Using a piecewise polynomial fitting method combining with a Newton iteration optimization method, triangulation is distributed along image characteristic lines in a self-adaptive manner. Polynomial approximation is used to obtain good visual and numerical effect of original image approximation on triangulation. The method can be used for image approximation and image vectorization.

The invention discloses a subspace clustering method and device for potential low-rank representation, and the method comprises the steps: obtaining data, and carrying out the preprocessing of the data, and obtaining a feature matrix; potential low-rank representation subspace clustering of an unobserved data sample is considered, a Schatten-p norm is used as a regular term to replace a rank function, and a problem that an NP is difficult to solve is converted into a solvable problem; introducing an lp norm constraint error term to construct a potential low-rank representation subspace clustering optimization objective function; then solving the optimization objective function to obtain a low-rank representation matrix; calculating an affinity matrix based on the low-rank representation matrix; and calculating and segmenting the affinity matrix by using a spectral clustering algorithm to realize potential low-rank representation subspace clustering of the data. According to the method,the problems that low-rank representation samples are insufficient and rank functions are difficult to solve are solved, the robustness of potential low-rank representation subspace clustering is enhanced, and the performance of potential low-rank representation subspace clustering is improved.

The invention discloses an ultra wideband interference suppression technique of minimum bit error rate criterion, which comprises the following steps of: 1, establishing a nonlinear equalizer model; 2, establishing a target function by using minimum bit error rate as criterion; and 3, adjusting the equalizer parameter by adopting a sliding window random gradient algorithm. The bit error rate of asystem is lower, the interference suppression capability of the system is stronger, the nonlinear function has better approximation performance, and the nonlinear equalizer has more excellent interference suppression capability; the system is simple to control, and the nonlinear equalizer can be switched between the minimum bit error rate criterion and the minimum mean square error criterion; thetechnique is simple to implement, the parameter can be adaptively adjusted, and the parameter of the equalizer can be adaptively adjusted on line by using the sliding window random gradient algorithm; and the technique has wide application range, is used for narrowband interference suppression of an ultra wideband communication system, and is also used in the fields of interference suppression, mode identification and the like of other communication systems.

The invention discloses a method for learning a non-linear face movement manifold based on a statistical shape theory. A method for pre-processing face shape based on the statistical shape theory comprises the following steps of: (1) demeaning, normalizing and pluralizing the shape of each frame in a face movement sequence; (2) removing redundant information in complex representation; and (3) by combining Riemannian geometry tangent space mapping, projecting the face movement sequence of the complex representation into a tangent space of the movement manifold to form a face movement locus. Byusing a Gaussian process latent variable model, the method for learning the face movement manifold comprises the following steps of: (1) calculating a mean value and a covariance function of a Gaussian process, and determining a probability density function of the constructed Gaussian process; and (2) solving a latent variable by using a scaled conjugate gradient method to obtain a dimension reduction result which corresponds to the face movement locus. In the method, the dimension of face movement data is reduced by using a true manifold distance and using a good dimension reduction method, so that the structure of the face movement manifold is more accurately described.

The invention discloses an AUV recovery docking dynamic positioning control method based on model predictive control. For dynamic positioning control in an AUV recovery docking process, state estimation is completed through an adaptive unscented Kalman filter algorithm, a state estimation result is compared with an expected state to complete feedback, linear processing is carried out on a linear part in a complex nonlinear system of the AUV, meanwhile, the approximation capacity of an RBF neural network to a nonlinear function is utilized, a compensation amount is generated through an RBF neural network compensator with a self-adjusting function according to the nonlinear part of the AUV, and closed-loop control over a dynamic positioning system is achieved through a model predictive controller. According to the method, a neural network and model predictive control are combined to solve the influence caused by an inaccurate model in the AUV dynamic positioning process, and the dynamicpositioning control precision and stability are improved.

The invention discloses a short-term load prediction method and device for a power utilization region. The method comprises the steps of collecting historical load index data of a to-be-predicted power utilization region; respectively inputting the historical load index data into corresponding classification load prediction models; performing the operation of the classification load prediction model, obtaining unit short-term load prediction results corresponding to different user categories, wherein the classification load prediction model is a neural network model trained by taking a morletmother wavelet function as a transfer function based on load index principal components of the same user category, and the load index principal components are obtained through a principal component analysis mode according to historical load index data of the same user category; and summing the short-term load prediction results of the units to obtain a short-term load prediction result of the to-be-predicted power utilization region. The technical problem of low prediction precision of an existing short-term load prediction mode of the power utilization region is solved.

The invention discloses a near-field dynamics modeling method for composite materials in a thermal conduction problem. The method comprises the following steps that (1) an entity model is established;(2) the entity model is divided, layered area positions of all materials in the entity model are determined, and corresponding material attributes are given to all areas; (3) the entity model is divided into unit grids; (4) initial temperature values of all units of the entity model are set; (5) based on a near-field dynamics constitutive relation of mass diffusion, model parameters are integrated, temperature change calculation methods of all the units in the whole entity model are determined, and operation parameters of all the units in the entity model are calculated; (6) calculation is submitted. The modeling method is high in precision and good in grid self-adaptivity, all the parameters in the model are pre-processed, the complexity degree during operation can be greatly lowered, and an algorithm is greatly optimized. Moreover, the modeling method and the operation process are wide in application range and suitable for solution of multiple problems.

The invention relates to a Backstepping-based lower limb exoskeleton control method, which comprises the following steps of: 1, establishing a lower limb exoskeleton dynamical model, and converting the lower limb exoskeleton dynamical model into a control system state equation; 2, designing a Backstepping controller, and carrying out the control of the Backstepping controller; 3, improving an RBF disturbance observer, including designing the RBF disturbance observer and an RBF neural network adaptive law; and 4, performing control implementation, and enabling the lower limb exoskeleton to move according to the expected track. According to the method, for external random disturbance, the approximation characteristic of an RBF neural network is utilized, a disturbance observer is designed, and the external random disturbance is approximated; and aiming at the network approximation error existing in the RBF neural network, the disturbance observer is further improved, and the auxiliary variable is introduced to compensate the network approximation error, so that the external random disturbance is compensated, the approximation of the external random disturbance is closer to a real value, and the approximation error of the RBF neural network is reduced.

The invention discloses an active power filter inversion control method based on a metacognitive fuzzy neural network, which includes the following procedures: S1, establishing the dynamic equation ofthe active power filter, S2, designing the control law(shown in the description), where f(>hat)(x) is the estimate obtained by using metacognitive fuzzy neural network to approximate the unknown f (x) of the system. A metacognitive method is introduce to adjust that fuzzy neural network structure on line, the fuzzy neural network structure can be adjusted dynamically according to the increase oftracking error design rules, parameter updating and rule deletion algorithm, which can improve the compensation current tracking performance and system robustness in the presence of parameter perturbation and external disturbance.

The invention discloses a robot controller design method, device and equipment and a readable storage medium. According to the method, a kinetic model of the robot is established, a target control model of the robot is determined, an event triggering mechanism of the robot is designed according to the kinetic model, and model description is performed on the target control model, so that a function approximation model of the target control model is determined; and on the basis of the event triggering mechanism, a control law and an adaptive law of the robot is constructed by using the function approximation model, a closed-loop control model of the robot is determined, and a target controller of the robot is established by using the control law, the adaptive law and the closed-loop control model; the nonlinear problem of an event triggering mechanism is solved, and the stability and feasibility of the control system of the fan detection robot operation arm based on event triggering are improved.

Provided in the invention is a magnetic fluid motion control system based on parameter uncertainty quantization. The system comprises a controller; a magnetic fluid system model, an objective function, an objective function gradient A or an objective function gradient B are arranged in the controller. According to the invention, the controller design is simplified; the random disturbance and uncertain factor influences in the actual movement of the magnetic fluid can be simulated; and the applicability problem of the control variable parameterization to the model uncertainty problem is solved.The applicability problem of the control variable parameterization to the model uncertainty problem is solved. The total time required by the control algorithm is greatly reduced by being compared with the time required by conventional control algorithms; and thus the possibility of real-time control of the magnetic fluid is increased.

The invention discloses a uranium and zirconium system X-ray fluorescent substrate effect correcting method. The uranium and zirconium system X-ray fluorescent substrate effect correcting method comprises the following steps that a uniform design theory is used for arranging uranium and zirconium system test points; an X-ray fluorescent spectrograph is used for collecting the signal intensity of uranium and zirconium elements; a multiple regression model of the signal intensity and the quality concentration of the uranium and zirconium elements is built; an SPSS technology is used for solving the nonlinear multiple regression equation; the uranium and zirconium system substrate effect correcting method based on the multivariate regression analysis is built. The test points are arranged by using the uniform design theory, so that the arrangement uniformity is realized; the test times can be effectively reduced; the calculation time is saved; the SPSS measure is creatively provided for solving the problem of solving the nonlinear multiple regression model; the multiple regression model is used for building the signal and mass concentration relationship of the uranium and zirconium elements; the complicated substrate effect expression problem is avoided; the difficult problem of uranium and zirconium system substrate effect correction is solved.

The embodiment of the invention discloses an image zooming method. The method comprises the following steps of: determining an image area to which each pixel belongs in red component intensity, green component intensity and blue component intensity of a source image respectively, wherein the image area comprises an edge area and a flat area; if it is determined that the pixel belongs to the flat area, carrying out interpolation calculation on the red component intensity, the green component intensity and the blue component intensity corresponding to the pixel belonging to the flat area according to a binary three-point Lagrange interpolation algorithm to calculate a pixel value of a target image; if it is determined that the pixel belongs to the edge region, carrying out the interpolation calculation on the red component intensity, the green component intensity and the blue component intensity corresponding to the pixel belonging to the edge area according to a gravity Newton binary mixed rational interpolation algorithm to calculate the pixel value of the target image; and generating the target image according to the pixel value. By employing the method and the system, the definition of the target image after zooming processing can be increased, and the quality of the target image after the zooming processing can be obviously improved.

The invention relates to the technical field of wireless communication, in particular to an energy efficiency optimization method for Sink nodes in a sensing cloud network, which comprises the following steps of: S1, constructing a sensing cloud network, forming a sensor area with a honeycomb structure by a plurality of sensor nodes in the bottom layer of the sensing cloud network, and setting oneSink node in each sensor area; S2, constructing a reference algorithm model to obtain a test optimal beam forming vector of the Sink node; S3, constructing a deep neural network model, and training and testing the deep neural network model through the data of the reference algorithm model; and S4, optimization energy efficiency: obtaining an optimal beam forming vector of the Sink node through the trained and tested deep neural network model. According to the invention, the energy efficiency of the Sink node can be improved, the decision time of the sensing cloud network is reduced, and the real-time performance of the system is improved.

The invention discloses a styrene bulk polymerization anti-interference distribution shape control method based on an interference observer. The method comprises the steps of obtaining a parameterized model of a polymer according to an element reaction in a styrene bulk polymerization reaction; introducing a B spline function to approximate a particle size distribution function output by the system, calculating a weight vector at a corresponding moment, and establishing a state space model between the weight vector and control input by adopting a subspace identification method based on an input and output data pair; respectively designing a PI type controller and an interference observer according to the state space model and considering the existence of interference to realize estimation of unknown interference and effective control of an output distribution function; and solving the corresponding controller gain and the observer gain by combining the Lyapunov stability analysis method so as to complete the anti-interference control of the styrene bulk polymerization process. According to the invention, a reliable anti-interference performance can be provided, and the stability of the styrene bulk polymerization process is improved.