71 results about "Model implementation" patented technology

A decision support system for supply chain management is disclosed. In one embodiment, an organizational structure of an enterprise value chain is mock-constructed as a framework model and solutions are logically distributed through the organization in accordance with the model. Product management, demand management and inventory management are performed on an exception basis and these processes are implemented incrementally and organizationally such that enterprise activities may be tracked and monitored, by exception, at multiple levels of granularity. In a general aspect, the invention enables collaborative ordering, forecasting, inventory and replenishment management by implementing such systems through an enterprise organizational model.

Described herein are systems and methods for creating and using Convolutional Recurrent Neural Networks (CRNNs) for small-footprint keyword spotting (KWS) systems. Inspired by the large-scale state-of-the-art speech recognition systems, in embodiments, the strengths of convolutional layers to utilize the structure in the data in time and frequency domains are combined with recurrent layers to utilize context for the entire processed frame. The effect of architecture parameters were examined to determine preferred model embodiments given the performance versus model size tradeoff. Various training strategies are provided to improve performance. In embodiments, using only ˜230 k parameters and yielding acceptably low latency, a CRNN model embodiment demonstrated high accuracy and robust performance in a wide range of environments.

The invention relates to an abnormal power consumption detection method based on a neural network. The method diagnoses and analyzes an operation condition of equipment based on an established abnormal power consumption detection model, judges whether metering equipment is in a normal operation condition and realizes an aid decision-making function. The method concretely comprises the following steps of (1) data acquisition: data are mainly from electric energy metering data, operation condition data and event recording data in an electric energy meter and an acquisition terminal; (2) data cleaning: the used data can enter the model after being subjected to data cleaning and screening; (3) data classification: after data cleaning completes, the data are calibrated, one column of numbers for representing data classification is added at the end of the data for classification, and the data subjected to data calibration are integrated into training data; (4) a modeling process: an algorithm model is constructed in a manner of supervised learning; (5) model implementation; and (6) result analysis: the final accuracy rate of abnormal power consumption found by the model maintains at a high level.

A method of modeling operational and / or logical aspects of a system in a system-of-systems environment. Modeling components of generic structure are used to obtain a logical model of the system. The logical model and the modeling components of generic structure are used to obtain related models targeted toward the aspects. The related models are implemented to determine effects of the aspects on the system and / or system-of-systems.

The invention discloses a method for conducting power system fault diagnosis by combining information theory with expert system, belonging to the technical field of power system safety treatment. Themethod includes the following steps: firstly establishing a fault diagnosis information transfer model of an actual channel, determining all possible pieces of equipment with fault through forward reasoning of the expert system according to a power cut area where the power system fails, then respectively assuming suspicious equipment, and reasoning out the corresponding action state of switching and protection reversely to obtain all possible information sources in actual communication. Finally, a control center obtains a group of data sequences consisting of the action state of switching andprotection, establishes the fault diagnosis information transfer model based on the actual channel and implements examination, repair and removal of faults according to the fault diagnosis informationtransfer model.

The invention relates to a multiple tunnel concurrent model implementation method based on virtual network card technology, which comprises the following steps: when a tunnel is built, a virtual address is obtained by applying the processing course of the procedure, the use case of the virtual network card is carried out, file description words communicating with the virtual network card are established; after the virtual network card receives the data transmitted by a protocol stack, transmits the data packet to the corresponding file description words according to the destination address of a data packet, thus processing the data packet by the right processing course; the processing course of every tunnel is divided into tunnel establishment, data transmitting and tunnel dismantling; when the tunnel is established, a client sends tunnel establishing requests, a new course of the application procedure fork of a tunnel gateway processes the requests; after the tunnel is successfully established, the tunnel gateway is responsible for transmitting the communication data between the client end and an application server; when the client end cuts tcp connection or over time, the tunnel connection is dismantled; the tunnel gateway recovers the virtual address distributed by a recovering address pool, and the virtual example of the virtual network card is canceled.

The invention discloses a characterization method of material fatigue, creep, and fatigue-creep interaction service life, and belongs to the field of service life prediction of aero-engine critical materials. The characterization method is used for solving service life characterization and prediction problems of materials under low cycle fatigue, creep, and fatigue-creep interaction conditions. According to the principles, a power function form service life prediction method non-linear behavior characterization capacity is established via fatigue-creep interaction, low cycle fatigue, and creeptests of materials at different holding time, and obtaining of effective holding time and normalization dimensionless service life via normalization calculation method. The characterization method iscapable of realizing accurate characterization of service life of materials under low cycle fatigue and fatigue-creep interaction conditions, and especially, consideration and accurate prediction ofcreep service life can be realized. The advantages of the characterization method are that consideration of both physical mechanisms and model implementation convenience is realized, and material lowcycle fatigue, creep, and fatigue-creep interaction service life characterization and prediction problems are solved effectively.

The present application provides a method and system for outlier detection, anomalous behavior detection, missing data imputation and prediction of consumption in energy data for one or more energy sensors by using a unified model. The application discloses a data collection module for collect a time series data to be used as training data, a model training module for training the unified model using the collected time series data to enable computation of a plurality of parameters, and a model implementation module for implementing, by the trained unified model, the plurality of parameters on a new data of energy consumption wherein the plurality of parameters are used perform at least one from a group of outlier detection, anomaly detection, missing data imputation and prediction of consumption in energy data.

A process modelling technique uses a single statistical model, such as a PLS, PRC, MLR, etc. model, developed from historical data for a typical process and uses this model to perform quality prediction or fault detection for various different process states of a process. Training data sets of various states of the process are stored and the training data divided into time slices. Mean and / or standard deviation values are determined for both the time slice parameters and variables and the training data. A set of deviations from the mean are determined for the time slice data and the model generated based on the set of deviations. The modeling technique determines means (and possibly standard deviations) of process parameters for each of a set of product grades, throughputs, etc., preferably compares on-line process parameter measurements to these means and use these comparisons in a single process model to perform quality prediction or fault detection across the various states of the process. Because only the means and standard deviations of the process parameters of the process model are updated, a single process model can be used to perform quality prediction or fault detection while the process is operating in any of the defined process stages or states. Moreover, the sensitivity (robustness) of the process model may be manually or automatically adjust each process parameter to tune or adapt the model over time. An alternative aspect is a method of displaying process alert information using a user interface having multiple screens.

The invention provides a bearing fault detecting and locating method and a detecting and locating model implementation system and method. Data preprocessing is performed on the no-tag classification data of a rolling bearing and then the data are inputted to a trained feature learning and detection model so that the fast detecting and locating problem of the rolling bearing under multiple fault modes can be solved, and statistics of the probability of each type of classification result is performed through voting by the minimization loss function; and the certain fault feature of the most votes is determined as the currently estimated fault mode and the fault part is located. The whole feature learning process does not require any manual feature extraction process, the original data act asthe input of the feature learning algorithm, the unsupervised feature learning process is used in the learning process, and the extracted bearing fault features can be efficiently self-expressed through deep data expansion and projection so that the problem of acquisition difficulty of the tag data can be solved, and the method has the characteristic of high detecting and locating accuracy.

The present invention provides a 3D printing method, comprising the following steps: (11) obtaining a target accuracy grade and a target printing model, and establishing a mapping relationship among an accuracy grade, a printing model, and printing parameters corresponding to the accuracy grade and the printing model; (12) determining, according to the mapping relationship, target print parameters corresponding to the target accuracy grade and the target printing model; (13) performing slicing processing on the target printing model according to the target printing parameters to obtain a print file; and (14) printing the target printing model according to the print file. According to the method, a problem that a user operation in a current 3D printing process is relatively complicated can be solved. The present invention further provides a 3D printing system.

The invention discloses a fault detection method based on a dimensional variable type independent component analysis model. The method comprises the following steps: in an offline modeling phase, according to differences of values of various column vector elements of a separation matrix in a traditional independent component analysis (ICA) model, various variables are firstly and correspondingly endowed with different weights to reflect differences of dimensions; afterwards, because each column vector in the separation matrix represent a difference of the corresponding measurement variable in a projection direction, the dimensions have a plurality of variable forms, and a plurality of ICA fault detection models can be established correspondingly; and, when online monitoring in performed, the multiple ICA models are called to calculated corresponding monitoring statistics, and a final probability type monitoring index is obtained by utilizing Bayesian reasoning in order to provide convenience for fault decision-making. Compared with a traditional method, the method is advantageous in that modeling considers uncoordinated importance of the measurement variables, and fault detection is implemented by utilizing the multiple modes at the same time. According to the method, description for a normal data characteristic is comprehensive, and an excellent fault detection effect is obtained through utilization.

The invention discloses a network element model implementation method supporting complex multi-network construction. The method first describes the networking mode and directionality of the network, then defines a node capability attribute, and finally establishes a link relation and the time-varying function of the link. The method abstracts the core characteristics of the network node, and on this basis, fully considers the network hierarchy, a node motion attribute, an application characteristic, the link direction and the link time variability. The network is modeled by G (V, E, H). In the description of the link E, a time-varying relation function related to the motion characteristics of both end nodes vi and vj is introduced, and the direction and weight information W can be added so as to express a networking strategy. By using the nodes suitable for routing computation as the same reference plane, the network is divided into multiple hierarchies, and the hierarchical structure of the network is described by using the network hierarchy attribute H. The method is especially suitable for the formalized description of a satellite network, a wireless network, and a wired network under the form of dynamic networking and hybrid networking.

The invention discloses a machine-learning-based daily access model implementation method and system. The method includes steps of A: setting a time range of traffic self learning; B: setting a networking terminal range of traffic self learning and a to-be-accessed service system list; C: collecting and analyzing traffic; D: forming a traffic analyzing result; E: creating an abnormity access rule of a traffic model; and F: generating the traffic model according to the analyzing result and the abnormity access rule and monitoring network access constantly through the traffic model. According to the invention, machine learning on actual traffic condition of an enterprise can be realized. Through a period of time of self learning, a daily access rule (which is a rule of access to an enterprise service system by a network terminal) meeting the practical condition of the enterprise can be obtained. A security manager only needs to make fine adjustment on a practical access rule according to a practical access control requirement of the enterprise, so that abnormal access precision of the enterprise can be improved. Besides, by using the daily access model provided by the invention, optimization or detection of safety equipment strategies can be performed.

The invention discloses a quality-related fault detection method based on two variable blocks. According to the method, a genetic algorithm is combined with a neighbor component analysis algorithm, the input variables are divided into the two variable blocks which are related and not related to quality. Then, a partial least squares (PLS) model between the quality-related variable block and the output is established for carrying out the quality-related fault detection, and the quality non-related variable block is combined with the PLS model input residual error to carry out the quality non-related fault detection. Compared with a traditional moving method, according to the method disclosed by the invention, the quality-related and quality non-related measurement variables are distinguished optimally by combining the genetic algorithm with the NCA. In addition, according to the method disclosed by the invention, the PLS model input residual error of the quality-related variable is combined with the quality non-related measurement variable to carry out the quality non-related fault detection, and all the quality non-related component information is comprehensively utilized. Therefore, the method provided by the invention can give more accurate quality-related fault detection results.

The invention provides a simulation model interface adaptation development system and a working method thereof, and the working method comprises the following steps: creating model definition file content or editing existing model definition file content, and generating a model user code file; wherein the model user code file comprises a user model definition file and a user model implementation file; after a model user code file is generated, establishing a mapping relationship between an existing model interface data structure and a one-dimensional expanded interface variable required in a model standard interface; and generating interface code files and model description files corresponding to various models according to the mapping relationship. The method is provided with a simple andunified function interface, conversion and generation of the FMU and S-Function models are supported at the same time through one-time coding, and the generation efficiency of the models and the useconvenience of users are improved.

The invention relates to a three-dimensional grid splitting method of blind watermark. The process of embedding the watermark comprises the steps of 1) processing an original model using an NPCA algorithm, and separating all peaks into multiple aggregations; 2) splitting the grid, mapping all peak aggregations to be front view, selecting proper pixel from the front view to obtain an additional information sampling view; and 3) embedding the watermark information; the process of extracting the watermark comprises the steps of a) processing an attacked model using the NPCA algorithm; b) extracting the additional information sampling view of the attacked model; c) subtracting the additional information sampling view of the original model by the additional information sampling view of the attacked model, aligning the grid and primarily estimating the spit grid; and d) extracting the watermark. The invention only carries a little information, can synchronously split the attacked model, extracts the watermark without the original model, and can predict the watermark grid splitting boundary with less auxiliary data.

The invention discloses an SDN GIS network topology model implementation method, and relates to the SDN field. The method comprises the steps: enabling a client to request a background for the data of topology resources, and converting the requested data into a customized topological tree shaped model object; carrying out the recursive traversal of the customized topological tree shaped model object, drawing each node, adding circular cover objects corresponding to the nodes to a set synchronization model; obtaining a set of all nodes which need to be connected if the nodes are logic domain nodes and are not expanded, and enabling the unexpanded logic domain nodes or network element nodes in the set to be connected with the nodes; and calculating the centers of the circular cover objects through the longitude and latitude of a logic domain on a GIS map. The method displays the network topology on the GIS map, facilitates the management and positioning of the geographic positions of network resources, achieves the effective operation and control of the resources, and improves the efficiency.

The invention discloses a modeling implementation system based on steel crude fuel purchasing valuation. The modeling implementation system comprises a data preparation module and a data processing module, wherein the data preparation module is used for providing steel crude fuel composition definition information, valuation standard and rule information and collection, checking and matching information, including metrical information and checking and testing information, of various data; the data processing module is used for reading corresponding valuation standard information and valuation rule information in the data preparation module according to business-related information set in an information system, matching the valuation standard information and the valuation rule information with the complete metrical information and the checking and testing information in a combined mode, and contrasting actual checking and testing result values according to material information and predetermined valuation standards and valuation rules to obtain a settlement valuation model of crude fuel. The modeling implementation system based on steel crude fuel purchasing valuation can realize accurate and efficient processing of crude fuel settlement valuation for an iron and steel enterprise, can be adapted to rapid configuration after crude fuel settlement management rules change, and can realize crude fuel settlement management and systematic management of a purchasing and supply management system and a peripheral management system.

Certain embodiments provide systems and methods for dietary planning and monitoring according to a clinical element model. A method includes creating a dietary plan for a patient. The dietary plan includes dietary items and relationships between dietary items implemented according to a clinical element model, for example. The method also includes receiving input regarding a dietary item. The input includes an identifier for the dietary item and contents of the dietary item, for example. The method further includes mapping the dietary item to a clinical element model based on the input. The method additionally includes providing the clinical element model for the dietary item in conjunction with the dietary plan.

The invention discloses a fault diagnosis method based on the binary classification Fisher discriminant analysis and aims to improve the applicability and the classification accuracy through variable selection when Fisher discriminant analysis models are used for fault diagnosis. The method comprises the steps that a set of characteristic variables through which the fault types are most distinguished from normal data is selected by means of the genetic algorithm, then a binary classification Fisher discriminant analysis model between the normal data and each type of fault data is built by means of the characteristic variables, and finally, fault classification diagnosis is performed by means of a plurality of binary classification Fisher discriminant analysis models. Because the genetic algorithm is adopted to optimize and select the set of characteristic variables, the disturbing influence of non-characteristic variables can be maximally reduced, and a dimensionality reduction effect can be further achieved, so that the limitation of the limited quantities of reference fault samples to modeling is reduced to a certain extent. Besides, the binary classification discriminant analysis models are adopted, so that each model is targeted on a specific fault type, and accordingly the model classification accuracy can be improved.

The embodiment of the invention discloses a training model implementation method and device and a computer readable storage medium. The method comprises the steps that a training program loads a dynamic library generated by a service algorithm logic layer; And during forward propagation and backward propagation calculation, a forward propagation program and a backward propagation program of the dynamic library is called for training through a forward propagation interface and a backward propagation interface to obtain a training model. The training program and the service algorithm logic layerare decoupled, the training program and the service algorithm logic layer can run independently, the training program can load the service algorithm logic layer to run, and respective development andmaintenance are facilitated; Compilation and operation only involve service algorithm logic layer codes, the compilation time is saved, and the version development efficiency is improved.

A style model implementation that is applied to objects within a user interface to define any attributes of the objects. The model is divided into two parts: a “runtime” which uses a style definition to affect the look and feel of the user interface, and a “design-time” which is the experience presented by WYSIWYG tools used to define the look and feel of the user interface. The design-time may be implemented in terms of the run-time objects. Properties associated with the styles can be set on the objects using the design-time UI and enforced at runtime. Styles may be grouped into themes to provide an easy mechanism to apply changes to many objects.

The invention relates to a claim settlement customer risk identification method and system. The method comprises a data preparation stage, a model training stage and a model implementation stage. Thedata preparation stage comprises the following steps of obtaining data related to a claim settlement customer risk from medical claim settlement big data, wherein the data comprises personal information and doctor-seeing information of a claim settlement customer, and the doctor-seeing information includes doctor-seeing expense information; and extracting characteristics required for training a GBM model from the data to form a data set. The model training stage comprises the following steps of taking part of the data in the data set as a training set to be input to a trainer of the GBM modelfor performing training; and generating a trained GBM model. The model implementation stage comprises the following steps of collecting current claim settlement data of the claim settlement customer,and extracting the characteristics required by the GBM model; and inputting the characteristics to the GBM model to generate an identification result, wherein the identification result comprises a doctor-seeing expense possibly generated in the future by the claim settlement customer.

The invention discloses a component-based open architecture model implementation method, wherein the open architecture model implementation comprises a definition open accessing connector, an implementation open accessing connector and a definition insertable object; wherein the definition open accessing connector comprises a definition standard accessing pointer; the implementation open accessing connector comprises business accessed an implementation appointment layer; the definition insertable object comprises an accessing connector of definition standard. The method supports loading plug-in units when system software is in motion and is started, and supports accessing in different layers.

The invention discloses a page region weight model implementation method. According to the principle of administrative region division, a geographic information library based on administrative region division and a relational diagram of adjacent geographic positions at the same level are established, a region queried by a user and a related region information weight queue are output dynamically with region information queried by the user and a weight value queue as input, a correcting algorithm is used for correcting the geographic information weight queue which is output dynamically, and then the geographic information weight queue which is corrected is output. A retrieval program conducts retrieval according to the output region weight queue, and therefore the effect of geographic ranking of page output is achieved. The page region weight model implementation method comprises the following steps: the geographic information library based on administrative region division is established, an adjacent relation information library is established, weight retrieval is conducted, weight is corrected, and page retrieval is conducted. The page region weight model implementation method is simple in algorithm and easy to implement, optimizes search results, enhances information localization and individuation, and is high in practicability and usability.

The invention provides an industrial process fault detection method based on data neighborhood feature preservation and aims at solving the main technical problem of comprehensively considering the distance, time and angle neighborhood features of process data in a process of establishing a fault detection model. The method comprises the following steps: defining a corresponding neighborhood set for each sampling data point, wherein the neighborhood set comprises data samples similar to the data point in distance, time and angle; then, constructing a feature value problem to solve a projection transformation vector, and establishing a corresponding fault detection model on the basis; and finally, implementing online fault detection by use of the model. Compared with the traditional method, the fault detection model established by the method minimizes the risk of information loss, and a more reliable and accurate result can be obtained.

The invention relates to a methodology and computer executable instructions configured to implement a prediction system. The invention deals with the use of a configuration file specifying at least the interactions to be completed between components of the prediction system, where this configuration file is transmitted to an implementation site. At the implementation site the configuration file is supplied as an input to at least one autonomous software agent where this agent or agents run the components of the prediction system as specified by the interactions defined within the configuration file. An extension to this method is also disclosed where the prediction system is built or constructed at the implementation site using the configuration file.

The invention relates to the technical field of data management, in particular to a power grid natural-disaster data warehouse model implementation method. A power grid natural-disaster device operation and maintenance system is imperfect and is simple in data storage mode, a traditional data window changes deserved natural-disaster data into information isolated islands. The power grid natural-disaster data warehouse model implementation method comprises the steps that a natural-disaster classification system is established according to a natural-disaster monitoring device; a monitoring device standing book entity model is established; the monitoring device entity model is converted to generate one or more theme example models; according to theme examples, theme type standards are subdivided for the natural-disaster classification system; according to specific items of natural-disaster monitoring, a theme attribute basic-standard library is established; a relation model is established; a big data technology is applied to establish a non-relational database natural-disaster data storage model. The power grid natural-disaster data warehouse model implementation method facilitates information calling and use and meanwhile provides a reliable unified data warehouse for future natural-disaster information data mining and analysis.