51 results about "Global modeling" patented technology

A method and system for designing a probe card from data provided by prospective customers via the Internet is provided. Design specifications are entered into the system by prospective customers and compiled into a database. The collective feasibility of each set of design specifications is determined by an automated computer system and communicated to the prospective customer. If feasible, additional software enables prospective customers to create verification packages according to their respective design specifications. These verification packages further consist of drawing files visually describing the final design and verification files confirming wafer bonding pad data. Verification packages are reviewed and forwarded to an applications engineer after customer approval. An interactive simulation of probe card performance is also provided. Data on probe card performance is incorporated into an overall modeling exercise, which includes not only the probe card, but data on the device(s) under test and wafer, as well as data on automated test equipment.

The invention discloses an unsupervised crowd abnormity monitoring and positioning method based on recurrent neural network modeling. The utilizes the combination of the time sequence characteristic of a monitoring video and the long time dependence of a recurrent neural network, individually models each grid after dividing a video scenario into grids, performs unsupervised learning on a crowd dynamic sequence in a normal case by selectively using optical flow statistical characteristics, trains a model by using a Hessian-Free Optimization method, loads abnormal data into the model, and monitors and positions a crowd scenario at a t+1 moment by measuring the distance between a histogram of optical flow at the t+1 moment and a histogram of optical flow at the t moment. The unsupervised crowd abnormity monitoring and positioning method achieves global modeling in the time and space, well keeps a relation among the crowd dynamic sequence characteristics at various moments, simplifies the complexity of the model, reduces RFF training difficulty, and has good abnormal event monitoring and positioning instantaneity and accuracy.

The invention discloses a method for optimizing a dynamic cloud workflow by using a genetic algorithm. The algorithm aims at minimizing periodic average expense of the cloud workflow in iteration execution, and limiting the total execution time that the workflow is executed for on period in a dynamic environment each time within a maximum execution deadline defined by a user. As the execution mode of the workflow is dynamic and variable in the cloud realization environment, the method carries out overall modeling on all flow topology results which may occur, then the characteristics of dynamic time-variation of the cloud workflow are comprehensively considered, the average performance of the execution of the cloud workflow in the dynamic environment is optimized by using the genetic algorithm, and the execution efficiency of the cloud workflow is improved.

The invention discloses a human body detection and attitude estimation combined deep network learning method. The method comprises: combining a CNN model with overall and local information to carry out detection, the model efficiently extracting underlying features through a shared convolution layer, and then the features respectively passing through two branches connected in parallel to carry outhuman body detection and attitude estimation. According to the invention, a Fusion model with a hidden tree structure reasoning algorithm is used to fuse results of human body detection and attitudeestimation, so that a robust and reliable human body detection box is obtained. According to the method, through an NMS algorithm (poseNMS), the obtained information of the human body parts is utilized, and all the individuals which are shielded with one another are effectively reserved. According to the method, a tree structure model is used for embedding information of each part into a detectedbounding box, and a convolutional network is used for realizing an inference algorithm. The method integrates the advantages of overall modeling and local modeling, has a very good detection effect oncrowded and shielded people and pedestrians with uncommon posture behaviors, and can be better integrated into practical application.

The invention discloses a self-adaptive topology control method for a wireless sensor network based on a cellular automaton and an intelligent agent. A multi-level cellular automaton method is adopted on overall modeling and the topology of the heterogeneous wireless sensor network is expressed by the change of a simple discrete state; the concept of the intelligent agent is introduced to abstract partial key nodes into the intelligent agents, so that the ability of the network for dealing with complex tasks independently is improved. According to the invention, relevant knowledge of complex systems and artificial intelligence and the wireless sensor network are combined together and a self-adaptive topology method is designed; and on the premise that the network coverage and the degree of communication are guaranteed, the energy consumption is saved and the service life of the network is prolonged.

An enterprise system/process modeling (ESPO) system and method that permits efficient representation of enterprise systems/processes (ESP) and calculation of global solutions to models associated with these ESPs is disclosed. The system/method incorporates an enterprise system/process modeling definition subsystem (ESPD) in which an interconnected processes, resources, constraints, and objectives may be defined to describe a global enterprise system/process model (ESPO). This ESPO describes the enterprise to be modeled and the various boundaries within which enterprise modeling is to take place. This ESPO is translated and transformed into an intermediate indexed format by an enterprise global modeling subsystem (EGMS) for use in generating the coefficients of an equation matrix. The equation matrix is solved for a solution state space conforming to desired enterprise objectives and the results are presented for review by an enterprise model reporting subsystem (FSPR).

The invention provides a method and apparatus for analyzing, regulating and controlling the active power of a wind farm based on wind farm region real-time wind speed. The method and apparatus are used for a control system provided with a data acquisition system, a wind farm modeling system, a wind farm data storage system, a wind turbine active power forecasting system, a wind farm active power distribution system and a control effect feedback system. According to the method and apparatus adopted, overall modeling is carried out based on a regional geography model, a wind turbine model, a wind turbine wake flow model and a regional identical wind belt model; and a fuzzy control theory, a neural network algorithm and an adaptive adjustment algorithm are adopted to uninterruptedly predict real-time wind speed of wind turbines at a next time point. The method and apparatus of the invention can accurately and effectively predict theoretical active power upper limit of each wind turbine in a short time through monitoring wind farm regional real-time wind speed; and by means of the theoretical active power upper limit, wind farm active power control efficiency and accuracy can be effectively improved.

The invention discloses an incremental Kriging-based sequential optimization test design method. The method comprises the following steps of performing initial test design; performing initial modeling; performing model verification; performing optimization sampling; judging an update criterion of a Kriging model by adopting a six sigma update criterion; and performing incremental Kriging modeling or DACE modeling. The whole algorithm is divided into two parts according to the design of the algorithm: the first part is building of incremental Kriging, namely, quick building of the Kriging model is realized by performing transform, Joris factor decomposition and maximum likelihood estimation operations on a correlation matrix, a target function matrix and a regressive function matrix after addition of sampling points; and the second part is a sequential optimization test design stage, namely, the optimization sampling of new design points is finished through maximized variance optimization and the six sigma update criterion, and the operation is repeated, so that global modeling efficiency of the Kriging model is greatly improved under the condition of meeting model precision.

The invention discloses an elastic operation control method for an AC/DC hybrid distribution network. By establishing a distribution network elastic evaluation index, an energy storage model, a wind power and photovoltaic output model, and constructing a AC/DC distribution network timing sequence transfer model, overall modeling is performed. Since the model is a nonlinear mixed integer programming model and is difficult and inefficient to solve by an artificial intelligence algorithm or enumeration method. Therefore, the invention provides a greedy search algorithm for solving and optimizingthe model. When a fault occurs in the AC/DC hybrid distribution network, the provided model takes account of the transfer and control capabilities of the energy storage, transfers the loads in the faulted distribution network area in cooperation with distributed power sources such as wind power and photovoltaic power in the distribution area and the standby power supply capacity of an adjacent tieline, calculates the elastic index according to a transfer result, optimizes the output of each energy storage period by an improved particle swarm optimization algorithm to obtain an optimal energystorage output mode, and finally achieves the maximum elastic operation of the distribution network in the fault area.

The invention discloses a fermentation process fault monitoring method based on a just-in-time learning local model and relates to the field of data driving fault diagnosis. Global modeling can not overcome the problems of multiple stages of production process, seasonal effects and material quality effects during actual production, and a large number of false alarms will be produced during time-varying process. A local modeling method based on a just-in-time learning strategy is put forward to solve the problem of model mismatching during actual fault monitoring, and faults are monitored through a local partial least square model. Information entropy is introduced into the just-in-time learning strategy, and similar sample points are automatically selected for modeling. Due to the fact that the local model can represent the current system state, stage identification is not needed, the calculated amount is reduced, and problems which are brought to monitoring by time-variant characteristics in the fermentation process are overcome. The false alarm rate is effectively reduced, the fault monitoring accuracy rate is increased, and safety and economy of production are guaranteed.

This invention is focused on a new method of numerical modeling of the Direct Metal Additive Manufacturing process with the layer-by-layer building of the metal products. This method separates the global macro-scale modeling and the local micro-scale modeling, with a database to link in between. The database containing the micro-scale modeling results can be established well before the global scale product simulation is conducted. As a result, this invention only uses the global modeling and database to simulate the additive manufacturing of the whole product, without using the time-consuming micro-scale modeling simultaneously. This new method enables very rapid simulation of the additive buildup process of products and prediction of product qualities, which only takes minutes to complete the simulation instead of weeks needed by the conventional methods of simulation known to those skilled in the art.

The invention provides a video small target detection method based on a depth convolution neural network,. Based on the characteristics of multi-layer nonlinear structure of deep convolution neural network, the video object features are modeled as a whole. Combined with the fast and accurate detection of common video objects by Faster RCNN and the effective path added by ResNet in the feature extraction layer of depth network, With the simplification of the network structure, a depth network structure ERF-Net (Efficient Residual Faster rcnn) based video target detection method is proposed, which can detect video objects quickly and small objects at the same time.. The invention has the advantages that the targets with different distances and different scales in the video are accurately andquickly detected, the effect and efficiency of small target detection are improved, and a good foundation is provided for subsequent target tracking, target recognition, and the like.

The invention relates to a method for quickly analyzing bonding strength of a side window of a high-speed train. A global modeling and local modeling combination method is adopted for a train body, a bonder, glass and a window frame during finite element modeling; the train body is modeled by adopting a large-sized unit; and the bonder, the glass and the window frame are modeled by adopting small-sized units. The method comprises the steps of firstly performing strength analysis on the train body to obtain a deformation displacement of the train body; secondly performing transfer from the deformation displacement of the train body to a displacement of the bonder by adopting a triangle area interpolation method; and finally loading a displacement obtained by interpolation to finite element models of the bonder, the glass and the window frame, considering actual working conditions of the side window, and performing submitting and calculation so as to quickly analyze the bonding strength of the bonder and lay a foundation for strength check and lightweight design of a bonding structure in the later period. The invention proposes a method for quickly calculating the bonding strength of the side window of the high-speed train. According to the method, repeated modeling and workload can be reduced; and in addition, the number of the finite element models can be reduced and the calculation efficiency can be improved.

A system analysis device 100 includes a whole model generation unit 1021 which generates a whole model which is obtained by modeling elements or whole of a system and which includes a plurality of partial models, a core model generation unit 1023 which extracts, from a plurality of whole models generated on the basis of the same event, the partial models whose prediction precision satisfies a predetermined criteria, and generates a core model by integrating the extracted partial models, and a threshold setting unit 1024 which calculates a distance between the core model and the plurality of whole models using a predetermined definition, and outputs the distance as a threshold value for error determination using the core model.

The invention discloses a method for converting a three-dimensional geologic model into a numerical calculation model. The method comprises the following steps of 1, performing region cutting on the geologic model according to a calculation range; step 2, extracting control data from the cutting model to perform grid reconstruction and obtain a cutting area surface model, wherein the step comprises grid reconstruction of a stratum interface and grid reconstruction of the surrounding surface of the stratum; 3, automatically subdividing finite element grids for the surface model of the cutting area, and generating topological data and geometric data of the finite element grids; and 4, importing the grid data into a numerical analysis system to complete modeling. According to the method, a three-dimensional geological model is converted into a numerical calculation model meeting the requirements of finite element grids in a specific mode; a complex model grid can be accurately generated through conversion, the overall modeling speed is high, the automation degree is high, the excavation process can be conveniently simulated, the method can be used for different numerical analysis systems, and the efficiency of preprocessing work for solving problems through finite elements is greatly improved.

The invention provides a star sensor based star tracking method. The method comprises steps as follows: firstly, angular velocity is estimated and filtered; then, star vectors are predicted and transformed into star points; finally, the star points are extracted from a wave gate with a prediction star point as the center. The method does not depend on attitude information, only the star vector of the star point of each frame is required to be modeled as a whole, and a star point predicted value is obtained on the basis of the star vector and a parameter of a star sensor. The method adopts a simple model and is small in computational amount and easy to implement; different from an existing star tracking method used for calculating angular velocity and recursion based on quaternion differential and adopting a guide star catalogue, the method has the advantages that the angular velocity can be directly computed on the basis of star vector differential for star vector recursion, the star tracking is further realized, and the method can be used as an effective complement to existing star tracking methods.

The invention provides a multi-color halftone equipment split modeling method and system oriented to a copy object. The method includes the steps that multi-color halftone equipment is split into multiple four-color models containing black ink in an enumeration mode; a Nigboer primary color step wedge prepared in the ink quantity limited modeling process is used for acquiring color gamut of the four-color models through a convex hull algorithm; a color gamut judgment method is used for acquiring color gamut coverage rates of different four-color model combinations on a to-be-copied image, based on the principle of maximization of the color gamut coverage rates, the four-color model combinations are determined; aiming at the probable condition that multiple sets of combinations meet condition requirements at the same time, a center aberration minimization method is used for determining a final optimal four-color model combination, and equipment split modeling is completed. Aberration samples are prepared in the ink quantity limitation process, under the condition without an additional sample preparation step, the overall modeling process of the multi-color halftone equipment is simplified, and implementation is convenient.

The invention discloses a flexible power distribution network real-time simulation information modeling method, which comprises the following steps of carrying out integral modeling by taking a real-time simulation platform as an IED, taking each simulator as logic equipment, decomposing specific functions of a flexible power distribution network system according to a four-remote function throughan object-oriented method, and establishing corresponding logic nodes; converting the data in a UDP format of the simulation platform into data in an IEC 61850 standard format through re-encapsulationand object mapping of the data; realizing model and service mapping from ACSI to MMS through an IEC 61850 server; starting an MMS service according to the corresponding mapping, and after receiving the data transmitted by the simulator and finding a corresponding IEC61850 mapping node, modifying the data into a corresponding data object attribute, so that the IEC61850 client or the IEC61850 equipment can successfully read and write the related analog quantity and digital quantity information of the power grid model in the simulator, and the functions of remote measurement, remote signaling, remote control and remote regulation of partial data in the flexible power distribution network model are realized.

The present invention relates to a multi-control method for management of the response time of a data-centric real-time application program. The present invention integrally models a first system for controlling the response time of CPU operations and a second system for controlling the response time of accessing a storage medium using a MIMO structure and simultaneously controls the response time of the CPU operation and the response time of accessing the storage medium through the configuration by the integrated modeling. According to exemplary embodiments of the present invention, it is possible to more efficiently control the response time than an existing feedback control method.

The invention relates to the field of computers, discloses an information statistics method based on target identification. Processing is carried out through logic optimization, the motion track of each tracking target can be recorded in detail; the method comprises the following steps: obtaining a plane structure, colors and the like of a plurality of cameras, realizing target tracking with certain reliability through global modeling of view fields of the plurality of cameras in combination with a target detection technology, and if a tracking target is lost or causes error tracking due to partial or complete shielding, correcting and recovering tracking through target information; and if the target information cannot be obtained temporarily, tracking can be carried out by analyzing the target motion track and the color characteristics.

The invention discloses an urban automatic optimization modeling system based on a tilt camera. The system comprises a tilt image collection module, a tilt image processing module, a tilt model generation module, a modeling recognition module and a modeling optimization module, the tilt image collection module collects omnibearing ground object data of front, rear, left, right and vertical anglesthrough a tilt camera, and position information during photographing is recorded through GPS equipment. And the modeling of the whole city can be rapidly and accurately formed. And the range of the model needing to be optimized can be reduced, and the next step of targeted optimization is facilitated. According to different types of models and reasons for generating problems, targeted optimizationis carried out, data are obtained from existing resources, the models with the problems are optimized one by one, leakage checking and deficiency filling are carried out on overall city modeling, andthe integrity and accuracy of the overall city modeling are further improved.