107 results about "Radial basis function kernel" patented technology

The invention provides a combined wind power prediction method suitable for a distributed wind power plant. The method comprises the following steps: step 1, acquiring data and pre-processing; step 2, utilizing a training sample set and a prediction sample set which are normalized to build a wind speed prediction model based on a radial basis function neural network and predict the wind speed and variation trend of distribution fans at the next moment; step 3, building a distributed wind power plant area CFD (computational fluid dynamics) model and externally deducing the prediction wind speed of each fan in the plant area according to factors such as the terrain, coarseness and wake current influence of a distributed wind field; step 4, acquiring the power data of an SCADA (supervisory control and data acquisition) system fan of the distributed wind field; and step 5, adopting correlation coefficients. The invention firstly provides a double-layer combined neural network to respectively predict the wind speed and power. Models are respectively built through adopting appropriate efficient neural network types, and improved particle swarm optimization with ideas of 'improvement', 'variation' and 'elimination' is additionally added to optimize the neural network, so that the speed and precision of modeling can be effectively improved, and the decoupling between wind speed and power is realized.

The invention discloses a method for recognizing a road traffic sign for an unmanned vehicle, comprising the following steps of: (1) changing the RGB (Red, Green and Blue) pixel value of an image to strengthen a traffic sign feature color region, and cutting the image by using a threshold; (2) carrying out edge detection and connection on a gray level image to reconstruct an interested region; (3) extracting a labeled graph of the interested region as a shape feature of the interested region, classifying the shape of the region by using a nearest neighbor classification method, and removing a non-traffic sign region; and (4) graying and normalizing the image of the interested region of the traffic sign, carrying out dual-tree complex wavelet transform on the image to form a feature vector of the image, reducing the dimension of the feature vector by using a two-dimension independent component analysis method, and sending the feature vector into a support vector machine of a radial basis function to judge the type of the traffic sign of the interested region. By using the method, various types of traffic signs in a running environment of the unmanned vehicle can be stably and efficiently detected and recognized.

The invention relates to an intelligent vehicle lane change path planning method based on a polynomial and radial basis function (RBF) neural network. The intelligent vehicle lane change path planning method comprises the following steps that: the state information of obstacles and lane change vehicles in lanes are detected and determined according to a vehicle-mounted sensor, and the state information comprises positions, speed, acceleration and shapes; the lane change vehicles and the obstacles are geometrically covered, and in addition, a lane change path model using the time as the independent variable is built; boundary conditions of the lane change vehicles are obtained by the dynamic RBF neural network; the lane change path parameter is subjected to traversing in a certain range according to a certain step length, and the calculation of a polynomial method is combined to obtain the lane change path set under the specific boundary conditions; index functions for evaluating the merits of the lane change patch performance are defined, the optimal path in generated lance change paths is screened according to the index functions and is applied to the practical lane change process of vehicles; and whether the RBF neural network is updated or not is determined according to the merits of the boundary conditions of the generated lane change paths. The neural network has good self-adaption capability, so that the problem that the RBF neural network structure is oversize or undersize is solved.

The invention relates to a cutter abrasion online monitoring method based on wavelet packet analysis and a radial basis function (RBF) neural network. The method comprises the steps that shear force coefficients and cutting edge force coefficients of tangential force and radial force in different cutter abrasion states are calibrated by means of an instantaneous cutting force coefficient recognition method; and by analyzing the correlation between cutting force coefficients and cutter abrasion, the coefficients are taken as cutter abrasion characteristic parameters and input into a RBF neutralnetwork model after being subjected to normalization processing. An input layer of a RBF neutral network monitoring model training process comprises cutting force characteristics, cutting vibration characteristics, the shear force coefficients and the cutting edge force coefficients after being subjected to normalization processing; and an output layer comprises the cutter rear cutter surface abrasion capacity after being subjected to normalization processing; a hidden layer comprises neurons obtained through radial basis function iterative optimization; and it is verified that the RBF neuralnetwork monitoring model has the advantages of high response speed and high recognition precision through cutter abrasion monitoring experiments.

The invention discloses a radial basis function (RBF) neural network parameter self-optimizing-based multi-step prediction method for water quality. The method comprises the following steps of: first storing the data of each monitoring station into a database of a local server by using the remote transmission of an online water quality monitoring instrument; then performing normalization processing on a water quality sample sequence, calculating an autocorrelation coefficient to determine an input variable of an RBF neural network, and converting sample data into a standard dynamic sequence data format trained and predicted by the RBF neutral network; next searching for and determining an optimal value of a spreading coefficient spread of the RBF neural network by utilizing a differential evolution algorithm and taking a relative standard error as a target function to obtain an optimal prediction model; and finally sampling water quality data in real time, performing multi-step prediction by using the obtained optimal prediction model and adopting a single-point iteration method, and evaluating a water quality prediction result to realize an early warning function. The water quality can be intelligently warned.

The invention provides a method for controlling a flexible structure and a self-adaptive changing structure by a radial basis function (RBF) neural network, belonging to the field of aviation. The method aims at solving the problem that the existing method can not preferably solve the conflict between the shake of a solar sailboard and the high-precision control target of an attitude control system. The method comprises the following steps: an E1 input forming module is used for converting an inputted expected satellite attitude angle theta d into a response uE1, and outputting the response uE1 to a nominal system and a flexible spacecraft; the nominal system is used for outputting expected satellite attitude information xm (t), and the flexible spacecraft is used for outputting practical satellite attitude information x (t) to obtain an error e (t) by comparing the xm (t) with the x (t); a sliding film face control module is used for obtaining a proper sliding film face s according to the error e (t), and transmitting the s to the RBF neural network and a self-adaptive locoregional control module; the self-adaptive locoregional control module is used for outputting a self-adaptive locoregional control u* to the RBF neural network; and the RBF neural network is used for obtaining and adjusting a locoregional control un and an adding result between the un and the uE1 according to the s and the u* to control the satellite attitude of the flexible spacecraft to achieve an expected value.

The invention discloses a multisource disturbance actuator saturation integrated spacecraft relative attitude control method. The method includes: starting from spacecraft relative attitude kinematicand kinetic equations to perform characteristic analysis and classification of multisource disturbances in a system; then, estimating disturbances with different characteristics by taking advantages of estimation performances of a disturbance observer, a radial basis function neural network and an extended state observer; next, designing an anti-saturation compensator to avoid damages caused by saturation upper limit exceeding of a spacecraft execution mechanism; finally, designing a composite controller according to a backstepping method on the basis of multisource disturbance observation values and anti-saturation compensation items. Therefore, adverse effects of multisource disturbances on the system are avoided, system robustness is improved, accurate spacecraft relative attitude control in a saturation range is guaranteed by the execution mechanism, and smoothness in space operation task completion is guaranteed.

The invention discloses an urban water disaster risk prediction method based on an RBF (radial basis function) neural network-cloud model. The method includes (1) determining evaluation factors, levels and the indicator range under corresponding levels; (2) determining an expectation Ex and an entropy En of the cloud model; (3) determining the weight of each evaluation factor according to measured values of the evaluation factors and the indicator range of each level; (4) training the RBF neural network, finishing model establishment for the RBF neural network, inputting the measured values of the evaluation factors of the cloud model to the trained RBF neural network to perform simulated prediction, and obtaining a prediction value of each evaluation factor; and (5) substituting the prediction value of each evaluation factor to the integrated cloud model to allow the integrated cloud model to calculate corresponding certainty degree of the prediction value of each evaluation factor belonging to each risk level and multiply the corresponding weight to obtain integrated risk level distribution. The urban water disaster risk prediction method is visualized and reliable and strong in operability, and accuracy of prediction is improved.

The invention discloses a self-adaptive teleoperation control method for a neural network based on a radial basis function. The self-adaptive teleoperation control method comprises three steps of respectively establishing dynamics models for a master manipulator end and a slave manipulator end in a teleoperation system, designing a slave manipulator end controller, and finally designing a master manipulator end controller. According to the self-adaptive teleoperation control method, the stability and relatively good operating performance in the teleoperation process can be guaranteed. When the slave manipulator end for teleoperation, namely a slave end manipulator grabs a target object, uncertainty of kinematics and dynamics parameters of the system occurs. The controller is designed for the slave manipulator end by use of a self-adaptive controller for an RBF neural network, so that the advantages of a self-adaptive control method can be played, the self-learning ability for and the adaptivity to the uncertainty in the teleoperation system are realized, and further the parameter uncertainty and the influence of unknown interference on the teleoperation system are overcome.

A system and method for performing Automatic Target Recognition by combining the outputs of several classifiers. In one embodiment, feature vectors are extracted from radar images and fed to three classifiers. The classifiers include a Gaussian mixture model neural network, a radial basis function neural network, and a vector quantization classifier. The class designations generated by the classifiers are combined in a weighted voting system, i.e., the mode of the weighted classification decisions is selected as the overall class designation of the target. A confidence metric may be formed from the extent to which the class designations of the several classifiers are the same. This system is also designed to handle unknown target types and subsequent re-integration at a later time, effectively, artificially and automatically increasing the training database size.

The invention discloses an intelligent transformer fault diagnostic method based on an RBF (radial basis function) neural network. The intelligent method includes (1), acquiring three ratios of five gases C2H2 / C2H4, CH4 / H2, C2H4 / C2H6 as training sample data by the utilizing IEC (international electrotechnical commission) three ratio method; (2) performing fuzzily processing on the three ratios by utilizing a membership function; (3), coding fault types; (4), training the RBF neural network according to the training sample data until the RBF neural network meets precision requirements; (5), inputting to-be-diagnosed samples after being fuzzily processed; and (6), outputting diagnosed results. The intelligent transformer fault diagnostic method has good reasoning ability and high diagnosed precision, overcomes the defects of the IEC three ratio method, and can precisely display all transformer fault problems.

An inverted pendulum self-adaptive iterative learning inversion control method is characterized in that: aiming at an inverted pendulum system containing unknown input saturation, a self-adaptive iterative learning inversion controller is designed through utilization of a neural network and an inversion control method in combination with self-adaptive iterative learning control; the construction of an integral lyapunov function solves the control problem caused by derivation of a unknown gain function; based on the median theorem, a hyperbolic tangent function is adopted to approximate an input saturation term; then, a radial basis function neural network is adopted to approximate and compensate uncertain unknown items of a system, and two combined self-adaptive laws are adopted to updatethe weight of the neural network and the bound of estimation errors. Under the condition that the system has input saturation, the invention provides the control method which can compensate the unknown uncertainty of the system, solve the control problem caused by derivation of the unknown gain function and realize two-norm convergence of a tracking error of the system to be near zero within limited iteration times.

The invention relates to an adaptive RBF (radial basis function) neural network control technique for three-phase parallel active filters, belonging to an active power filter control technique. The invention provides an adaptive RBF neural network control method for three-phase parallel active power filters, which is used for controlling a compensation current output by a three-phase parallel active power filter through a controller, thereby eliminating harmonic waves and improving the power supply quality of a power grid. According to an adaptive control rule provided by the invention, the boundedness of weights is ensured, and the stability of the controller is proved by using a Lyapunov stability theory; and simulation results show that the control method effectively reduces the distortion factor of harmonic waves and is good in dynamic response, and when parameters change, the controller has good robustness and adaptability.

The invention discloses a vehicle speed tracking method of a radial basis function neural network based on particle swarm optimization. The invention constructs an automobile dynamic model through anengine model, a transmission system model, a vehicle model and a brake model. The parameters of radial basis function neural network model are calculated by gradient descent method, and the PID controller adjusts the parameters adaptively by radial basis function neural network model. Parameters of particle swarm optimization are obtained by off-line optimization of particle swarm optimization algorithm. The PSO parameters are initialized and assigned to the radial basis function neural network PID controller. The initial throttle opening or the initial brake pedal position is obtained by theinitialized radial basis function neural network PID controller and input to the vehicle dynamics model to calculate the actual tracking speed. The actual tracking speed and the output of PID controller are inputted into the neural network, and the parameters of RBF neural network and PID controller are adjusted according to the feedback error of the speed. The invention realizes safe and stable tracking target speed.

The invention discloses a multi-parent genetic algorithm air source heat pump multi-objective optimization control method based on a radial basis function neural network. The method comprises the following steps that 1, input and output variables are input into a system according to user requirements; 2, creating, training and testing a radial basis function neural network; 3, performing multi-objective optimization on an air source heat pump by using a multi-parent genetic algorithm based on the trained radial basis function neural network; and 4, obtaining the parameter value of the input variable of the optimal solution according to the Pareto solution through the above steps, and transmitting the obtained input variable value to the system to adjust the control quantity of the heat pump. The multi-objective optimization of the COP heating capacity Qh or the carbon dioxide release amount m and the heating capacity Qh of the system can be rapidly realized while the precision is high.

The invention discloses a transformer fault diagnosis method based on a radial basis function neural network. According to the method, the content of characteristic gas in insulating oil can be used as input for the radial basis function neural network, transformer faults are output accurately, and accordingly accuracy in transformer fault diagnosis is improved greatly and safe and reliable transformer operation is ensured.

A method and apparatus are disclosed for recommending items of interest by fusing a plurality of recommendation scores from individual recommendation tools using one or more Radial Basis Function neural networks. The Radial Basis Function neural networks include N inputs and at least one output, interconnected by a plurality of hidden units in a hidden layer. A unique neural network can be used for each user, or a neural network can be shared by a plurality of users, such as a set of users having similar characteristics. A neural network training process initially trains each Radial Basis Function neural network using data from a training data set. A neural network cross-validation process selects the Radial Basis Function neural network that performs best on the cross-validation data set. A neural network program recommendation process uses the selected neural network(s) to recommend items of interest to a user.

The invention discloses a multivariate quality process out-of-control signal diagnostic method based on a support vector machine and genetic optimization. The multivariate quality process out-of-control signal diagnostic method based on the support vector machine and the genetic optimization is characterized in that first, the types of signals likely to lead to abnormality of a multivariate process are determined according to the mean value dimensions of the multivariate process, namely the structure of a classifier model is determined; second, radial basis function parameters and penalty factors of the support vector machine are optimized with the genetic algorithm; third, the optimal support vector machine classifier model is obtained through the acquired optimal parameters, and the multivariate process out-of-control signals are diagnosed on the basis of the optimal support vector machine classifier model. The parameters of the SVM are selected dynamically through global searching ability of the genetic algorithm, and thus automatic optimization selection of the parameters of the SVM classifier is achieved, and quality diagnosis effects of the multivariate process are also promoted. The multivariate quality process out-of-control signal diagnostic method based on the support vector machine and the genetic optimization integrates the GA global searching ability and the classifying ability of the SVM, and meanwhile avoids complex calculation, simplifies the network structure of the classifier and promotes generalization ability and identification efficiency of the classifier.

The invention discloses a diversified image marking and retrieving method based on a radial basis function neural network (RBFNN). The diversified image marking and retrieving method comprises the steps of (1) constructing and learning an RBFNN model capable of covering an image sub-concept; (2) inputting data preprocessed by a retrieval database into the RBFNN model constructed in the step (1), carrying out the diversification marking on images in an image library, and meanwhile marking the images in the image library with labels of concepts and sub-concepts; (3) carrying out the diversification retrieval on the marked image library according to retrieval key words and the marked results of the step (2): firstly searching the images marked with the retrieval key words, and sequencing the images according to the similarity of the concepts, and then bringing the images belonging to the different sub-concepts in the front according to the similarity of the concepts; and (4) outputting the retrieval results. The diversified image marking and retrieving method has the advantages that the image retrieving precision is improved, and meanwhile, the diversity of the image retrieval results is greatly enhanced, the retrieval time is saved, and robustness and practical applicability are high.

The invention provides an RBF (radial basis function) neural network-based indoor visual environment control system and method. The system includes a data acquisition module, a data processing and control module and an output driving module; the data acquisition module is used for acquiring indoor and outdoor illuminance values; the data processing and control module is used for obtaining indoor illumination control parameters according to the illuminance values and based on an RBF (radial basis function) neural network algorithm and outputting the indoor illumination control parameters; and the output driving module is used for controlling indoor shutter blinds to rotate and / or an indoor lighting lamp to be turned on according to the indoor visual control parameters so as to realize an indoor visual environment comfortable control effect. According to the RBF (radial basis function) neural network-based indoor visual environment control system and method of the invention, the relatively reasonable indoor visual environment neural network control system is constructed, the control precision and universality of the indoor visual comfortable environment control system can be improved, defects such as instability and limitations of a traditional control system and method can be eliminated; and illumination power resources can be saved to the greatest extent with indoor visual comfort ensured.

The invention relates to a method for predicting mesoscopic fuel consumption on the basis of RBFNN (radial basis function neural networks). The method includes determining road energy consumption influence factors; dividing vehicle travel tracks into travel fragments; computing average energy consumption of vehicles in each form fragment; analyzing average energy consumption distribution laws of road sections and computing average energy consumption of the road sections; determining setting of parameters such as the road energy consumption influence factors; utilizing obtained data sets as training sets for neural networks and carrying out model learning; inputting test data sets and acquiring road fuel consumption prediction results by means of computing. The method has the advantages that large quantities of observation samples of input parameters and road energy consumption output parameters in regard to road section types, average speeds of the vehicles and the like can be accumulated under the support of large data volumes of energy consumption track data sets and can be trained, laws of correlations between the road energy consumption influence factors and average energy consumption of the roads can be mastered, accordingly, the energy consumption can be predicted for other road sections, with insufficient quantities of energy consumption track samples, in road networks, energy consumption laws can be extensively popularized, and the method is high in precision in the aspect of monitoring granularity.

The invention provides a front car identification method based on monocular vision. The method includes the steps that (1), an original image is collected from a vehicle-mounted camera, the edge of the image is extracted according to a Canny edge extraction method, influence of noise points is eliminated through morphological filter, projection is carried out in the horizontal direction, and an area of interest of a front car is obtained according to projection characteristics; (2), a shadow area at the car bottom is extracted and judged according to the geometrical shape of the shadow at the car bottom, edge characteristics are overlaid, and a car area is judged; (3), graying, normalization and binary tree complex wavelet transformation are carried out on small color images of candidate car areas of different shapes, and characteristic vectors are obtained; (4), the number of dimensions of the characteristic vectors is decreased through a two-dimension independent component analysis algorithm, the characteristic vectors are fed into a support vector machine based on a radial basis function kernel to be classified, and it is judged that whether the candidate car areas are the car area. Cars on the road ahead are detected accurately, and real-time and reliable road condition information can be supplied for unmanned cars.

The invention belongs to the technical field of information retrieval and database structures, and discloses a surface roughness monitoring model based on data mining and a construction method. The surface roughness model is established based on variance analysis and regression analysis, the incidence relation between cutting force and vibration signals and the surface roughness is determined according to the clustering result, and blindness in the cutting signal selection process is greatly reduced. A multi-sensor technology is applied, force and vibration signals in the cutting process are collected in real time, the cutting signals are decomposed and reconstructed on the basis of singular spectrum analysis, interference generated by noise signals can be effectively reduced, and characteristic quantity extraction is facilitated. Time domain and frequency domain results of cutting force and vibration signals are analyzed, feature extraction is carried out through correlation selection, a surface roughness prediction model is established through a radial basis function neural network, prediction precision and the intelligent level can be greatly improved, and online real-time prediction can be achieved.

The robust controller for nonlinear MIMO systems uses a radial basis function (RBF) neural network to generate optimal control signals abiding by constraints, if any, on the control signal or on the system output. The weights of the neural network are trained in the negative direction of the gradient of output squared error. Nonlinearities in the system, as well as variations in system parameters, are handled by the robust controller. Simulation results are included in the end to assess the performance of the proposed controller.

The invention discloses a sewage TP (Total Phosphorus) soft measurement method during sewage treatment based on a self-organized particle swarm and a radial basis function neural network, and aims to solve the problems that during current sewage treatment, the effluent TP measurement process is complicated, the cost of instruments and equipment is high, and the reliability and the accuracy of the measurement results are low. The effluent TP soft measurement method is calibrated by real-time data, so as to predict the effluent TP during the sewage treatment process, and solve the problem that the effluent TP is difficult to measure; the results indicate that the effluent TP soft measurement method can quickly and accurately predict the concentration of effluent TP, and is favorable for strengthening delicacy management in an urban sewage treatment plant, and promoting the real-time water quality monitoring level.

The invention relates to a magnetic flux leakage testing defect reconstruction method based on an improved artificial bee colony algorithm. According to the method, a radial basis function neural network is used as a forward model, and an error square sum of a magnetic flux leakage signal predicted by the forward model and an actually measured magnetic flux leakage signal is used as a target function to improve the artificial bee colony algorithm; a current individual optimal solution and a global optimal solution are introduced to accelerate algorithm convergence speed; the improved artificial bee colony algorithm is used as an iterative algorithm to solve a reconstruction problem, and the finally obtained global optimal solution is a reconstructed defect outline. The magnetic flux leakage testing defect reconstruction method based on the improved artificial bee colony algorithm improves speed and precision of magnetic flux leakage testing defect reconstruction.

The invention discloses a method for optimizing a disaggregated model by adopting genetic algorithm. The method comprises the following steps of: 1, acquiring a training sample, wherein the acquiring process includes signal acquisition, characteristic extraction and sample acquisition; 2, selecting kernel function: radial basic function is selected as the kernel function of a disaggregated model needing to be established, and the disaggregated model is a support vector machine model; and 3, determining penalty parameter C and kernel parameter gamma: a genetic algorithm is adopted to optimize the penalty parameter C of the disaggregated model needing to be established and the kernel parameter gamma of the selected radial basic function and the optimization process includes population initialization, calculation on the fitness value of each individual in the initialized population, selection operation, interlace operation and variation operation, calculation on the fitness value of each individual in the offspring, selection operation and judgment on whether the termination condition is met. The method is reasonable in design, simple and convenient in operation, convenient to realize and good in use effect and high in practical value; the classification precision of the obtained disaggregated model is high, the training speed is high and the number of support vectors is less.

The invention provides an indoor wireless positioning fingerprint generating method based on artificial neural networks and belongs to the technical field of wireless communication. The invention particularly relates to an indoor wireless positioning fingerprint generating method based on artificial neural networks. The method uses data collected at a plurality of survey points selected in a positioning area to generate a radial basis function (RBF) network to achieve mapping of target point coordinates in the positioning area to the received signal strength indicator (RSSI) of an access point (AP), and fingerprint data are calculated on the basis of the mapping. Without many survey data, the method uses the RBF network to generate other fingerprint data to improve the forming efficiency of a fingerprint database.

The invention discloses a soft measuring system based on kernel principal component analysis and radial basis function neural network, and the system can be used to measure parameters hard to measure in a generating set or in the complex industrial process. An intelligent instrument for measuring auxiliary variables, a DCS database for storing data and the soft measuring system are included, wherein all measurable variables of the generating set are measured by the onsite intelligent instrument and stored in the DCS database, the DCS database stores all data of the set, and the soft measuring system comprises a PC used for modeling, a server for predicating a soft measurement model and a device for displaying data. The invention also discloses a soft measuring method based on kernel principal component analysis and radial basis function neural network. The system and method in the invention have high precision, generalization capability and performance, are suitable for modeling in the complex industrial process, are general and universal, can solve the problems in soft measurement of operation parameters in complex environments including high temperature, high voltage, corrosion and electromagnetic interference, and improve the system safety and reliability.

The invention provides a gravity field density inversion method based on a quasi-radial basis function neural network. The method comprises a step of establishing a gravity observation system, a stepof establishing a gridded model, a step of establishing a gravity forward kernel function matrix, a step of establishing a radial basis function neural network, a step of training the neural network,and a step of outputting an inversion result. According to the method, a model space is compressed by using a radial basis function, and the dimensionality reduction of inversion parameters is achieved under the premise of ensuring complex model representation ability. A pseudo-neural network structure is proposed, the training of a sample tag is not needed, the difficulty of establishing a training data set is avoided, and a gravity field density inversion algorithm is achieved based on the pseudo-neural network structure. The vertical resolution and reliability of an inversion result are improved, the method has strong anti-noise ability, and the application field of a gravity inversion method is extended.