CN-121998618-A - Indoor intelligent exhibition digital twin management system based on big data driving

CN121998618ACN 121998618 ACN121998618 ACN 121998618ACN-121998618-A

Abstract

The invention discloses a big data driven indoor intelligent display digital twin management system, which comprises a behavior acquisition module, a behavior analysis module, a prevention maintenance module, a control module and a control module, wherein the behavior acquisition module is used for deploying multi-source heterogeneous sensing equipment to acquire movement tracks and interaction behavior original data of observers, the behavior event log data of each display is obtained through data cleaning fusion and space association analysis, the behavior analysis module is used for converting the behavior event log data into analog load time sequence data by using a display digital model machine and training a physical information neural network according to the analog load time sequence data to infer display hidden physical state vectors, and the prevention maintenance module is used for carrying out feature discretization and mutual information filtering on the display hidden physical state vectors and constructing fusion information gain and mutual information filtering And the optimized decision tree model of the coefficient double-splitting criterion acquires a preventive maintenance work order, and digital twin management of the indoor exhibition closed loop is realized.

Inventors

HE ZHEFENG
YANG AIQUN
HE YINQUAN

Assignees

太原泰森智能科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260123

Claims (8)

1. Indoor intelligent exhibition utensil digital twin management system based on big data drive, its characterized in that includes: The behavior acquisition module is used for acquiring the movement track and the interaction behavior original data of the observer by deploying the multi-source heterogeneous sensing equipment, and acquiring behavior event log data of each display through data cleaning fusion and space association analysis; The behavior analysis module is used for converting the behavior event log data into analog load time sequence data by using a digital prototype of the display device, training a physical information neural network according to the analog load time sequence data and reasoning out a hidden physical state vector of the display device; the preventive maintenance module performs characteristic discretization and mutual information filtering on the display implicit physical state vector and constructs fusion information gain and fusion information gain Obtaining a preventive maintenance work order by using an optimized decision tree model of a coefficient double-splitting criterion; And the digital management module is used for driving the digital twin body to perform visual early warning and work order dispatching based on the preventive maintenance work order, so that the digital twin management of the indoor exhibition tool closed loop is realized.
2. The big data driven based digital twin management system for an indoor intelligent display as defined in claim 1, wherein the process of obtaining the behavioral event log data is: obtaining movement track original data of a viewer With the original data of interaction behavior Abnormal points in the moving track data are removed through a spatial neighborhood filtering rule, time alignment is carried out on the interactive behavior data, and multi-source data in the same time window are fused to obtain a fused track point set ; Track point set based on fusion Construction of display space The subdivision model divides the display space into subareas corresponding to each display, and the track points of each display are corresponding to each observer Calculating the space distance between the locus point and each display device to determine the display device to which the locus point belongs Is defined by the sub-region of (2): track data and corresponding kth display Associating; At the same time, combine interactive behavior In a display device Generating behavior event log data corresponding to each display 。
3. The big data driving-based indoor intelligent display digital twin management system according to claim 2, wherein the conversion process of the analog load time sequence data is as follows: Constructing a display dynamics equation: ; Wherein, the In order to develop a quality matrix for the display, In order to provide a damping matrix, In the form of a matrix of stiffness, In the form of a generalized coordinate vector, Is an external excitation vector; logging behavior event generated by behavior acquisition module Inputting into a dynamic equation of the exhibition tool according to the behavior type Duration and duration of time Determining a corresponding external stimulus And solving a dynamic equation of the display to obtain the simulated load time sequence data of the display 。
4. The big data driven indoor intelligent display digital twin management system according to claim 3, wherein the training of the physical information neural network according to the simulated load time sequence data comprises the following steps: the physical information neural network adopts a double-branch architecture of a coding branch and a decoding branch, and the coding branch adopts a one-dimensional convolution layer and a gating circulation unit The decoding branches map coding features to a display physical state space through a full-connection layer, and are embedded into a display dynamics equation to serve as hard constraint; To simulate load time sequence data Monitor data corresponding to physical state of the display For training samples, the display physical state monitoring data Including component wear data, fatigue accumulation data, and structural stress data; Carrying out time domain processing on the load data, normalizing the monitor data of the physical state of the display, and simulating the load time sequence data according to the time domain processed load time sequence data And normalized monitor data of physical state of the display Constructing a loss function of the physical information neural network, and verifying to obtain the trained modulo physical information neural network 。
5. The big data driven indoor intelligent display digital twin management system according to claim 4, wherein the process of reasoning out the display hidden physical state vector is: current analog load time sequence data Real-time input trained physical information neural network Extracting time domain features of the current load through the coding branch, mapping the time domain features to a physical state space through the decoding branch, and outputting a display hidden physical state vector 。
6. The big data driven indoor intelligent display digital twin management system according to claim 5, wherein the process of performing feature discretization and mutual information filtering on the display recessive physical state vector is: to display implicit physical state vectors To input the feature set, integrate the history maintenance full data, extract the maintenance type, the maintenance time and the maintenance effect as the maintenance label, and form the training data set with the label ; Dividing continuous implicit physical parameters into 5 sections through equal frequency discretization, converting numerical characteristics into category characteristics, calculating mutual information values of the characteristics and maintenance decision labels, and filtering redundant characteristics with the mutual information values lower than 0.1.
7. The big data driven indoor intelligent exhibition digital twin management system according to claim 6, wherein the process of obtaining the optimized decision tree model is: based on the traditional decision tree, splitting is carried out on top-level nodes of the first 3 layers of the traditional decision tree by adopting an information gain criterion, and switching is carried out on bottom-level nodes of the 4 th layer and below of the decision tree Splitting the coefficient criterion; Implicit physical state vector from a display by feature discretization and mutual information filtering Candidate splitting characteristics during splitting of decision tree are screened out Based on candidate split features Calculating a dynamic splitting criterion score automatically selects an optimal criterion, expressed as: ; Wherein, the For the current candidate split feature Training data set at current node The score of the split on the basis of the score, For the gain of the information it is, Is that Coefficients; Setting a split stop threshold, all candidate split features Stopping splitting when the splitting scores of the two are less than the splitting stop threshold value, and adopting the minimum description length And (3) calculating the comprehensive value pruning redundant branches of the coding length and the data fitting error of the decision tree by using the pruning strategy, verifying the optimized decision tree by using a time sequence cross verification method, and obtaining an optimized decision tree model after the verification is completed.
8. The big data driven indoor intelligent display digital twin management system according to claim 7, wherein the process of visual early warning and work order dispatch based on preventive maintenance work order driven digital twin body is: self-adaptive preventive maintenance work order After generation, the structured data display in the work order is first identified Type of maintenance Maintenance priority Maintenance execution time Synchronizing to digital twins of corresponding spreaders through data interfaces, the digital twins being based on the spreader identification And (3) binding the physical display device and the virtual twin body, and triggering visual early warning and worksheet dispatching flow.

Description

Indoor intelligent exhibition digital twin management system based on big data driving Technical Field The invention relates to the technical field of new generation information, in particular to an indoor intelligent exhibition digital twin management system based on big data driving. Background Indoor exhibition tool management technology has developed from early manual inspection and single-point sensor monitoring to centralized monitoring based on the Internet of things. The more advanced method introduces a digital twin concept, and a plurality of physical sensors (such as stress, displacement and vibration sensors) are deployed on the display to directly collect state data, so that a virtual model of the display is constructed, state visualization and alarm based on a fixed threshold are realized, and a certain support is provided for preventive maintenance. The existing indoor intelligent display digital twin management system still has significant limitations, firstly, physical sensors are densely deployed on a display body to cause high system deployment cost and complex transformation and possibly influence the appearance and user experience of the display, secondly, the digital twin body mainly realizes monitoring but not prediction, state sensing is passive and lagged, the non-directly measurable degradation processes such as structural fatigue accumulation, component hidden abrasion and the like cannot be effectively observed, finally, maintenance decision is usually based on a simple static threshold value, dynamic and accurate maintenance strategies cannot be generated according to display actual load history and future use plan, and real intelligentization and resource optimization are difficult to realize, so that the indoor intelligent display digital twin management system based on big data driving is provided. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides the following technical scheme for achieving the purposes: Indoor intelligent exhibition utensil digital twin management system based on big data drive includes: The behavior acquisition module is used for acquiring the movement track and the interaction behavior original data of the observer by deploying the multi-source heterogeneous sensing equipment, and acquiring behavior event log data of each display through data cleaning fusion and space association analysis; The behavior analysis module is used for converting the behavior event log data into analog load time sequence data by using a digital prototype of the display device, training a physical information neural network according to the analog load time sequence data and reasoning out a hidden physical state vector of the display device; the preventive maintenance module performs characteristic discretization and mutual information filtering on the display implicit physical state vector and constructs fusion information gain and fusion information gain Obtaining a preventive maintenance work order by using an optimized decision tree model of a coefficient double-splitting criterion; And the digital management module is used for driving the digital twin body to perform visual early warning and work order dispatching based on the preventive maintenance work order, so that the digital twin management of the indoor exhibition tool closed loop is realized. The process of obtaining the behavior event log data comprises the following steps: obtaining movement track original data of a viewer With the original data of interaction behaviorAbnormal points in the moving track data are removed through a spatial neighborhood filtering rule, time alignment is carried out on the interactive behavior data, and multi-source data in the same time window are fused to obtain a fused track point set; Track point set based on fusionConstruction of display spaceThe subdivision model divides the display space into subareas corresponding to each display, and the track points of each display are corresponding to each observerCalculating the space distance between the locus point and each display device to determine the display device to which the locus point belongsIs defined by the sub-region of (2): Track data and corresponding exhibition tool Associating; At the same time, combine interactive behavior In a display deviceGenerating behavior event log data corresponding to each display。 The conversion process of the analog load time sequence data comprises the following steps: Constructing a display dynamics equation: ; Wherein, the In order to develop a quality matrix for the display,In order to provide a damping matrix,In the form of a matrix of stiffness,In the form of a generalized coordinate vector,Is an external excitation vector; logging behavior event generated by behavior acquisition module Inputting into a dynamic equation of the exhibition tool according to the behavior typeDuration and duration of timeDetermining a corresponding external st