59 results about "Dynamic topic model" patented technology

The invention provides a modeling method for a dynamic model of a modular multi-level convector. The method comprises the following steps of establishing an upper bridge single submodule switch period average model and an upper bridge arm single submodule small signal communication model; establishing an upper bridge arm switch period average model and an upper bridge arm small signal communication model; establishing a lower bridge arm switch period average model and a lower bridge arm small signal communication model; establishing a lower bridge single submodule switch period average model and a lower bridge arm single submodule small signal communication model; and establishing a modular multi-level convector switch period average module and a modular multi-level convector small signal module. The established dynamic model of the modular multi-level convector facilitates analysis on a dynamic property and a frequency response characteristic of the modular multi-level converter, facilitates cascade control strategy design of the device, and achieves description of internal state quantity.

The invention discloses a double-arm robot cooperative impedance control method based on an estimated dynamics model. Firstly, the expected trajectory of the Cartesian space at the tail end of two arms of a robot is calculated according to the expected trajectory of the target object in the Cartesian space, then the actual contact force generated by the tail ends of the two arms and the target object is measured, deviation between the actual contact force and the expected contact force is figured out, and then the expected trajectory is corrected; then the joint angle trajectory of the double-arm robot is figured out; and by means of time delay estimation, expected speed feedback and expected position feedback, an estimated dynamic model of the double-arm robot is obtained, control momentsof joints of the double-arm robot are obtained accordingly, and the double-arm robot is controlled to complete movement. According to the method, the coordination of kinematics and dynamics in the interaction process of the double-arm robot and the external environment can be realized, the control torque can be rapidly generated by estimating the dynamics model, and the method can be applied to the motion control of the cooperative operation of the double-arm robot.

The invention provides a spatial robot prediction control method based on a quantum particle swarm optimization algorithm. Firstly, a lagrangian dynamic model of a spatial robot system is establishedon the basis of an extended mechanical arm method, and a discrete state space model is established by combining the dynamic model with a kinematic model; secondly, a corresponding discrete model prediction controller is designed on the basis of a system extended state space model and a Laguerre polynomial; finally, rolling optimization is conducted on performance indexes under the constraint condition through the quantum particle swarm optimization algorithm, and prediction errors are subjected to feedback correction, so that effective tracking of tail end desired trajectory is achieved. According to the control method, effective tracking of the tail end desired trajectory can be achieved under the given constraint condition, and task space trajectory planning does not need to be carried out in advance; the situation that a global optimum solution cannot be found through a conventional quadratic programming algorithm under multi-constraint conditions can be avoided; energy consumptioncan be optimized while the requirement of tracking precision is met.

The invention discloses a breast cancer prognosis survival rate prediction method based on a dynamic Cox model, and belongs to the technical field of a tumor mortality risk regression model algorithm.The method includes: using a log probability regression model to estimate the patient's positive lymph node ratio LNR level; combining the ratio with other features to fit a Cox regression model of Bayesian method based dynamical estimation model parameters; and predicting a survival rate-time characteristic curve of the patient. The simulation results show that the Cox regression model of the Bayesian method based dynamical estimation model parameters of two logarithmic pseudo-marginal LPML indexes of the combined model is superior to the conventional Cox proportional risk model using LNR index features; and the Cox regression model of the dynamical estimation parameters can further improve the prediction precision of a survival analysis model, the computational complexity is within an acceptable range, and the prediction effect is relatively accurate.

The invention discloses a dynamic model-based unmanned car model prediction controller design method. According to the method, a novel vehicle dynamic model is established; an unmanned car model prediction controller on the basis of the novel vehicle dynamic model; (1) the novel vehicle dynamic model obtained through linearization at the current working point is utilized to predict an output stateof a system in a prediction time domain; (2) the obtained predicted output of the system and given reference output are utilized to construct an optimization problem; and (3) the optimization problemis solved, and a first component of U (k) is taken as an optimal control quantity of the current moment to act on the system. According to the method, vehicle speed is added to serve as the control quantity, and error compensation is introduced to make up accumulative errors caused by model precision, linearization and discretization in the prediction process, so that unmanned cars can have a better trajectory tracing effect.

The invention provides a rail train operation state prediction method based on a deep neural network structure model, and the method comprises the steps: obtaining the real-time data of a rail train operation state, and carrying out the preprocessing of the data; establishing a deep neural network structure model according to the preprocessed data; dividing the preprocessed data into a training data set and a verification data set, and training and verifying the deep neural network structure model through the training data set and the verification data set; And carrying out online prediction on the rail transit train through the verified deep neural network structure model. According to the method, a dynamic model is established on the basis of comprehensively considering the difference factors between vehicle individuals and the operation environment, and powerful support is provided for operation of a train in a complex environment.

The invention provides an intelligent contract forming method. The method is characterized by comprising the following steps of converting a non-formalized intelligent contract description document into a description document in a formalized specification; according to the description document in the formalized specification of an intelligent contract, creating dynamic models of the intelligent contract; achieving conversion among different dynamic models of the intelligent contract; verifying the correctness of the dynamic models of the intelligent contract; after the correctness verificationis successful, automatically generating intelligent contract codes according to the dynamic models of the intelligent contract; conducting consistency detection on the description document in the formalized specification of the intelligent contract and the generated intelligent contract codes.

The invention discloses a hybrid modeling method for a feeding system of a numerical control machine tool. The hybrid modeling method comprises a dynamic basic model and a neural network deviation model based on big data, wherein the dynamics basic model is obtained through dynamics theory analysis and parameter identification; the neural network deviation model is obtained by analyzing and training an instruction sequence, simulation prediction data of the dynamic basic model and actual response data; and the instruction sequence is input into a system mixing model, and the actual response sequence is predicted to obtain a mixed prediction sequence. Compared with a pure dynamic model, the technical scheme of the invention has the advantages that the simulation of a highly nonlinear process (such as a reverse process) is more accurate, and compared with a pure neural network model, the generalization ability under different processing technologies is stronger. Through a hybrid modelingmode, accurate simulation of a complex dynamic feeding system is realized.

A system and method that facilitate modeling unobserved speech dynamics based upon a hidden dynamic speech model in the form of segmental switching state space model that employs model parameters including those describing the unobserved speech dynamics and those describing the relationship between the unobserved speech dynamic vector and the observed acoustic feature vector is provided. The model parameters are modified based, at least in part, upon, a variational learning technique. In accordance with an aspect of the present invention, novel and powerful variational expectation maximization (EM) algorithm(s) for the segmental switching state space models used in speech applications, which are capable of capturing key internal (or hidden) dynamics of natural speech production, are provided. For example, modification of model parameters can be based upon an approximate mixture of Gaussian (MOG) posterior and / or based upon an approximate hidden Markov model (HMM) posterior using a variational technique.

The invention discloses a refined assessment model of regional wind energy resources, which comprises the following steps: S1, using anemometer tower observation data, SCADA data of wind turbines and terrain parameters respectively belonging in small-scale wind field ranges to establish separate multi-reference-point wind energy resource assessment models of small-scale regions, wherein the small-scale wind field ranges are 1-20 km; S2, using weather station observation data included in regional ranges of 20-200km and the existing models in the step S1 to establish correlation models, obtaining corresponding weight coefficients, and establishing dynamic models; S3, using numerical weather forecast data to establish mesoscale models of regional ranges of 200-500km; S4, using a combination of the mesoscale models, a wind farm power prediction system, a GIS geographic information model and the existing models in the step S1 and the step S2 to establish a refined self-adaptive model; and S5, using a combination of the refined self-adaptive model and a power load dispatching system to realize refined regional wind energy assessment and visual dynamic dispatching management.

The invention discloses an uncertain optimization method of gun projectile initial disturbance, which establishes a dynamical model of gun-projectile integration, screens test points in the value range of uncertain factors, substitutes the finite element analysis into the dynamical model of gun-projectile integration, and establishes a BP neural network agent model of gun-projectile integration dynamical model, carried out the Monte Carlo simulation on the BP neural network surrogate model of the integrated dynamics model of projectiles and artillery and the mean and determines the variance ofthe initial disturbance of projectiles; constructs the probability constraint function of interior ballistic performance index, and constructs the uncertain objective function by the mean and standard deviation of initial disturbance, and establishes the stochastic uncertain optimization model of interior ballistic performance. The stochastic uncertain optimization model of launching performancein bore is solved by using multi-objective genetic algorithm, and the optimal distribution interval of design variables is obtained. The invention realizes the uncertain optimization of the initial disturbance of the artillery projectile, and can obtain the parameter interval of the artillery projectile conforming to the target performance.

The invention provides a static meteorological satellite channel modeling method, which comprises the following steps of: according to the spatial distribution situation of a satellite-to-ground link, successively analyzing the influence of free-space loss, ionospheric scintillation, atmospheric absorption, multipath and shadow on a channel link; according to the weather situation, establishing a static meteorological satellite Rice channel model and a Suzuki channel model; and according to a Markov chain, linking the Rice channel model and the Suzuki channel model in one dynamic model to reflect the influence on the transmission characteristic of the static meteorological satellite channel by the weather state change. The method disclosed by the invention has an important value on the signal modulation, the coding mode and the power control technology of static meteorological satellite communication.

When the placement of the elements (mass points, links having inertia, etc.) of a model expressing a robot 1 is determined according to a first geometric restrictive condition from an instantaneous desired motion of the robot 1 that has been created using a dynamic model, this placement is defined as a first placement, and the placement determined according to a second geometric restrictive condition from a corrected instantaneous desired motion that has been obtained by correcting the instantaneous desired motion is defined as a second placement. The corrected instantaneous desired motion is determined such that the moment component calculated from the difference between the first and the second placements approximates a predetermined value. The instantaneous desired motion is created using a dynamic model of the robot. Thus, the motion of the instantaneous desired gait of the robot that has been created using the dynamic model is properly corrected without using a dynamic model, leading to higher dynamic accuracy of an instantaneous desired gait including the corrected motion.

The invention discloses a regional land use control policy simulation and effect analysis method, and relates to the field of land use and cover change simulation. By integrating the system dynamics model, the cellular automata model and GIS, the method realizes the real-time data exchange between the three at the grid level, realizes the effective interaction between graphs and numbers, and improves the efficiency of land use change. The accuracy of the simulation can expand the scope of application of the model, and provide better auxiliary tools for the regulation and control of land policies, which can be effectively used in the simulation and effect analysis of land use regulation policies in various regions, dimensions, and scales.

A novel electronic system provides fast three-dimensional model generation, social content sharing of dynamic three-dimensional models, and monetization of the dynamic three-dimensional models created by casual consumers. In one embodiment, a casual consumer utilizes a dedicated real-time 3D model reconstruction studio with multiple camera angles, and then rapidly create dynamic 3D models with novel computational methods performed in scalable graphics processing units. In another embodiment, uncalibrated multiple sources of video recording of a targeted object are provided by a plurality of commonly-available consumer video recording devices (e.g. a smart phone, a camcorder, a digital camera, etc.) located at different angles, after which the uncalibrated multiple sources of video recording are transmitted to a novel cloud computing system for real-time temporal, spatial, and photometrical calibration and 3D model reconstruction. The dynamic 3D models can be uploaded, listed, and shared among content creators and viewers in an electronic sharing platform.

A four-wheel steering vehicle stability control method based on time-varying dynamics model is characterized in that the method comprises a reference model, a tire data processor, an MPC controller and a CarSim vehicle model; The reference model is used to determine the desired yaw rate and the center-of-mass sideslip angle of the vehicle. The tire data processor is used for determining the side angle, side force and side force gradient of the tire. The CarSim automobile model is used to output the information of the actual moving state of the automobile, including the longitudinal velocity, yaw angular velocity, center-of-mass deflection angle and road adhesion coefficient. The MPC controller optimizes the rear wheel rotation angle of the vehicle according to the expected yaw rate, the center-of-mass deflection angle and the actual motion state information of the vehicle, and outputs the result to the CarSim vehicle model to control the vehicle to realize the stability control.

The invention discloses a dynamic topic model-based dynamic text cluster device and method. The device comprises a news acquisition module, a news initial feature extracting module, a dynamic feature converting module and a dynamic cluster module, wherein the news acquisition module is used for acquiring news data in the Internet; the news initial feature extracting module is used for extracting the initial feature of the acquired news data; the dynamic feature converting module is used for carrying out dynamic feature conversion on the extracted initial feature; and the dynamic cluster module is used for carrying out dynamic cluster on a news data set based on the converted feature. With the adoption of the device and the method, the problem of effective cluster of the dynamic text data set is solved so as to combine a dynamic topic model with a dynamic cluster algorithm, and the inherent problems of polysemy, multiple words for one meaning, data sparsity, feature dimension disaster and the like of the text feature based on a term are solved.

The invention discloses a dynamic short text cluster searching method. According to the method, short text stream data are used for building a short-term topic model and a long-term historic topic model is synthesized to amend the short-term topic model in a data stream to obtain the probability distribution of topics and feature words, clustering is performed by the conditional probability of thetext and the topics, and dynamic accurate searching of the keywords is formed. The dynamic topic model is built, the keyword searching function changing along the time is realized, the problems of sparsity of the short text data, information loss and the like are solved by a polynomial mixed topic model, and the efficiency and the performance of information searching are improved.

The invention discloses a thermodynamic system dynamic model establishing method which belongs to the technical field of thermodynamic system dynamic characteristic analysis and optimal operation. The method comprises the following steps of: analyzing the causal relation between main parameters of a thermodynamic system to be modeled; determining a plurality of input variables, output variables and intermediate variables; according to the transfer order of each of the variables, determining an input layer, an output layer, an intermediate layer and the number of nodes in each layer; connecting the nodes with a causal relation or a reverse causal transfer relation by using directed line segments with arrows; determining an overall causal transfer topological structure of a system; marking standard data structures on each directed line segment representing the causal relation; and identifying related parameters in each data structure by operating measured data by the system. The model established by the method can effectively describe a causal dynamic transfer relation of each key physical parameters in the thermodynamic system, is suitable for analyzing the relation among substances, energy transfer and balance of the thermodynamic system from a relatively macroscopic perspective, thereby analyzing and optimizing the operating characteristics of the system.

The invention discloses an emergency public opinion evolution analysis method based on sentiment analysis and theme features. The emergency public opinion evolution analysis method mainly comprises the steps: firstly carrying out the word segmentation of a data text through employing a word segmentation method based on character string matching according to the features of Chinese word segmentation; secondly, utilizing synonym linguistic data to extend weights and features of words after word segmentation, and calculating emotion values; then calculating and carrying out time slicing on the data based on the emotion difference value; and finally, taking the data after the time slicing as the input of a dynamic topic model (DTM) so as to obtain the change and the popularity of the topic word of the emergency in a time sequence, predicting the development trend of the emergency, and verifying the effect of sentiment analysis in a topic evolution process by comparing with a topic evolution result only considering the text similarity slicing.

The invention discloses a method for extracting important time slices in a social media short text. The method comprises the following steps: dividing a text in time; extracting a subject term sequence in the social media short text through a dynamic subject model, searching a monotonous interval with popularity ranking change of each subject term, and combining monotonous intervals which have opposite trends and belong to fluctuation or monotonous intervals which have the same trend and smaller change amplitude; taking intersections of the combined monotonous interval sequences of all the subject terms in sequence, calculating the chaos degree of each intersection, and ranking to obtain a plurality of important time slices determined from a subject evolution perspective; performing sentiment analysis on each text after time period division by utilizing a naive Bayesian classifier, and determining an important time slice union set of each sentiment through a sentiment change amplitudeand a threshold value; calculating the confusion degree in the union set, and ranking to obtain a plurality of important time slices determined from the perspective of emotion conversion; and taking an intersection of the important time slices determined from the two angles to obtain the time slice.

The invention discloses a microgrid equivalent modeling method based on LSTM neural network. The method comprises specific steps of 1, collecting The disturbance data of the common coupling point of the microgrid during the disturbance period; 2, according to the equivalent modeling requirement of the microgrid, determining the number of input and output nodes of the LSTM neural network, and offline training the LSTM neural network by utilize the disturbance data collected in the step 1; 3, according to the neural network trained offline in the step 2, obtaining a nonlinear equivalent model which can represent the running state of the microgrid. The invention utilizes artificial neural network to have good ability to deal with complex non-linear problems, and at the same time can well capture the dynamic characteristics of the electric power system, and the structure and parameters of the dynamic model are determined by the structure and parameters of the LSTM neural network. Only themeasured values of the common coupling points of the micro-grid are needed, and the specific parameters and topological structure of the micro-grid system are not required to be mastered. Moreover, adefinite model is not required to be established in advance when the micro-grid system is equivalent. Once the model is trained and tested, the dynamic equivalent model based on LSTM neural network can meet the needs of system simulation and analysis.

The invention provides a social network user multi-label classification method based on a dynamic multi-view learning model, and the method comprises the following steps: building multi-view representation of a user for a specific social network data set; based on the user representation, constructing a deep fusion representation model among the multi-view data; using a dynamic routing model, updating parameters, and optimizing multi-view characteristics; and introducing a shared representation model, and constructing an objective function for the features in the step 3; and through model optimization, obtaining an optimal shared representation matrix, and finally, realizing multi-label classification of any user by utilizing the shared matrix. According to the method, multi-label efficient classification of network users is realized. The problems of model learning performance reduction, limited view fusion quantity, incapability of meeting multi-classification task requirements of the model and the like caused by data loss are solved, and the method can be widely applied to scenes such as user accurate analysis, abnormal user detection, user relationship mining and unknown user identification in the network.

The invention discloses a robot dynamics model identification method, and relates to the field of robot dynamics models. The dynamics model identification method comprises three cycles from inside tooutside; the inner-layer cycle is configured to converge a covariance matrix of errors in a dynamics model to a constant value, so that the statistical characteristics of the errors are more accurately calculated; the middle-layer cycle is configured to remove data points not conforming to the model in the data so as to improve the consistency of model parameter estimation; and the outer-layer cycle is configured to update nonlinear friction force model parameters so as to improve the accuracy of model parameter estimation. Compared with a traditional dynamics model identification method, thenovel method is used for analyzing a parameter identification problem from the point of statistics, an iterative weighted least square method is adopted to remove an outlier, a new friction force model is adopted to improve the identification precision, and dynamics model parameters can be identified accurately and robustly in an off-line mode, so that a firm foundation is laid for the design andapplication of a robot.

The invention discloses a software customization method and system; the method comprises the following elements: an application layer used for modeling according to client requirements, wherein one model file is matched with one suite in the modeling process; an execution layer used for automatically loading the corresponding model file when a user carries out business operation, explaining the model file, and converting into a business order; a realization layer used for finishing business operation steps according to the business order when the business order is received; an adaption layer used for providing an man-machine interaction interface according to the client requirements. The system comprises a rule modeling device, a rule translator, a rule file database and an adaptor. The method can shorten a delivery period, can save software coding / testing works taking the biggest time consumption in a developing process, thus realizing software design field delivery, and improving developing efficiency; the method uses dynamic model loading to realize field design delivery, i.e., seeing is obtaining. The model in the system can realize functions so as to reduce or omit conventional assembly language programming.