Sand table type factory production simulation prediction method and systemTechnical Field
The invention belongs to the technical field of digital manufacturing management, and particularly relates to a sand table type factory production simulation prediction method and system.
Background
The market competition of the manufacturing industry is more and more vigorous, with the arrival of new technological waves, manufacturing enterprises face huge impact and challenges of information technology, and most of manufacturing enterprises which rely on an infinite supply mode of labor force in early stages reach the obsolete edge. This promotes the utilization efficiency of resources, the automation level of production and the improvement of the production management flow of enterprises, and changes from early manual work to the modern industry of automatic production and intelligent management. The manufacturing industry is a foundation stone developed in the future of China and is an important industry of China, so that the intellectualization and the digitization of a factory are unprecedented.
However, at present, the production of the traditional factory enables the production of the equipment layer to work according to the production plan, the control layer feeds back the data such as equipment state, production progress and the like to the data layer, the factory production is adjusted according to the data of the data layer, the traditional factory production system has a virtual simulation model without a manufacturing workshop, large data information taking the model as a carrier cannot be formed, and virtual simulation analysis in the production stage cannot be supported, so that the data analysis and production management efficiency is low, the production equipment failure cannot be predicted, and the production running risk is increased.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a sand table type factory production simulation prediction method and system, which acquire factory equipment specification information, factory actual operation data and equipment rule constraint information, acquire a three-dimensional equipment model through 3D modeling and model light weight processing according to the factory equipment specification information, acquire an equipment relationship topological graph according to the equipment rule constraint information, acquire a factory sand table model through a digital twin system according to the three-dimensional equipment model and the equipment relationship topological graph, acquire deisoed actual operation data according to the factory actual operation data through processing, map the deisoed actual operation data through the factory sand table model to acquire virtual factory operation data, acquire a fault early warning model through neural network training according to the equipment historical data, acquire equipment fault early warning information through the fault early warning model processing according to the virtual factory operation data, start acousto-optic positioning alarm, generate a work log according to the equipment fault early warning information and upload to a terminal, thereby realizing the construction of a factory sand table, improving the information interaction capability and the data analysis capability, further improving the production management efficiency and realizing the equipment fault prediction early warning.
The aim of the invention can be achieved by the following technical scheme:
a sand table type factory production simulation prediction method comprises the following steps:
S1: acquiring plant equipment specification information, actual plant operation data and equipment rule constraint information;
S2: obtaining a three-dimensional equipment model through 3D modeling and model light weight processing according to the plant equipment specification information, obtaining a virtual entity and entity relation through neural network word segmentation model processing according to the equipment rule constraint information, obtaining an equipment relation topological graph according to the virtual entity and the entity relation, and obtaining a plant sand table model through a digital twin system according to the three-dimensional equipment model and the equipment relation topological graph;
S3: obtaining abnormal actual operation data through data processing according to the actual operation data of the factory, and inputting the abnormal actual operation data into the factory sand table model through a pre-constructed data transmission path to perform data mapping to obtain virtual factory operation data;
S4: acquiring equipment historical data, training the equipment historical data through a neural network to obtain a fault early warning model, and processing the virtual factory operation data through the fault early warning model to obtain equipment fault early warning information;
s5: uploading the equipment fault early warning information to the factory sand table model, starting sound and light positioning alarm, generating a work log according to the equipment fault early warning information, and uploading the work log to a terminal.
Preferably, the step S2 specifically includes the following steps:
S201: the model lightening process comprises the steps of storing an initial three-dimensional equipment model obtained through 3D modeling as an STL format three-dimensional equipment model, analyzing redundant point lines and redundant patches through light weight tools according to the STL format three-dimensional equipment model, and removing the redundant point lines and the redundant patches to obtain the three-dimensional equipment model;
S202: obtaining a factory equipment database, performing word segmentation processing on the equipment rule constraint information through a BiLSTM neural network word segmentation model to obtain an equipment information corpus, performing corpus labeling on the equipment information corpus according to a predefined entity type to obtain labeling information, obtaining the virtual entity through correlation with the factory equipment database according to the labeling information, processing the equipment information corpus through a BERT model to obtain deep semantic information, performing feature extraction on the deep semantic information to obtain a feature vector, performing convolution operation on the feature vector through a convolution neural network to obtain a convolution result, performing maximum pooling on the convolution result to obtain a maximum feature value, obtaining the entity relation through the convolution neural network according to the maximum feature value, and obtaining the equipment relation topological graph according to the virtual entity and the entity relation.
Preferably, the step S3 specifically includes the following steps:
S301: obtaining data abnormal statistics through data de-skew model processing according to the actual operation data of the factory, wherein the calculation formula is as follows: Wherein Oi represents the ith actual operation data of the plant, Ki represents the data anomaly statistics, and alpha and beta represent weight coefficients;
S302: and when the data abnormal statistic is smaller than the statistic critical value, the data abnormal statistic is marked as normal data, and abnormal actual operation data is obtained through normalization processing according to the normal data.
Preferably, the step S4 specifically includes the following steps:
S401: acquiring a data upper bound, a data lower bound and a data dimension of equipment history data, presetting that the neural network comprises N neurons, and obtaining N neuron weight vectors according to the N neurons through a random formula, wherein the random formula is expressed as: aij=rand (Uj-Lj) +lj, i=1, 2, 3..n, wherein Aij represents the neuron weight vector, U represents the upper data bound, L represents the lower data bound, j represents the data dimension;
S402: obtaining an optimized neural network according to the neuron weight vector, taking the equipment history data as a training set X= { X1, X2, x3...xM }, wherein M represents M pieces of equipment history data, inputting the training set into the optimized neural network, and obtaining the fault early warning model through training;
S403: and constructing a test data set according to the virtual factory operation data, obtaining a test neuron of the test data set, obtaining a test neuron weight vector and a test neuron distance according to the test neuron, obtaining an abnormal early warning coefficient according to the test data set and the test neuron weight vector through the fault early warning model, and obtaining the equipment fault early warning information according to the abnormal early warning coefficient.
Preferably, the step S403 specifically includes the following steps:
calculating the abnormal early warning coefficient, wherein the calculation formula is as follows: wherein i=1, 2, 3..p, D (O1, O2) represents similarity of the test dataset to the test neuron weight vector, O1 represents the test neuron weight vector, O2 represents the test dataset, P represents a test dataset dimension, wi represents the anomaly early warning coefficient, YD represents the test neuron distance;
And generating the equipment fault early warning information when the abnormality early warning coefficient is larger than or equal to a preset threshold value, and generating equipment non-abnormality information when the abnormality early warning coefficient is smaller than the preset threshold value.
Preferably, the device rule constraint information includes a device connection mode text description, a device task text description and a device function text description.
A sand table type factory production simulation prediction system, comprising:
The data acquisition module is used for acquiring the specification information of the equipment of the factory, the actual running data of the factory and the constraint information of the equipment rules;
The sand table modeling module is used for obtaining a three-dimensional equipment model through 3D modeling and model weight reduction according to the equipment specification information of the factory, obtaining a virtual entity and entity relation through neural network word segmentation model processing according to the equipment rule constraint information, obtaining an equipment relation topological graph according to the virtual entity and the entity relation, and obtaining a factory sand table model through a digital twin system according to the three-dimensional equipment model and the equipment relation topological graph;
the data processing module is used for obtaining abnormal actual operation data through data processing according to the actual operation data of the factory, inputting the abnormal actual operation data into the factory sand table model through a pre-constructed data transmission path, and performing data mapping to obtain virtual factory operation data;
the fault prediction module is used for acquiring equipment historical data, obtaining a fault early warning model through neural network training according to the equipment historical data, obtaining equipment fault early warning information through fault early warning model processing according to the virtual factory operation data, uploading the equipment fault early warning information to the factory sand table model, starting acousto-optic positioning alarm, generating a work log according to the equipment fault early warning information, and uploading the work log to a terminal.
The beneficial effects of the invention are as follows:
1. The three-dimensional equipment model is obtained through 3D modeling and model light weight processing according to the equipment specification information of the factory, virtual entity and entity relation are obtained through neural network word segmentation model processing according to the equipment rule constraint information, equipment relation topological diagram is obtained according to the virtual entity and the entity relation, factory sand table model is obtained through a digital twin system according to the three-dimensional equipment model and the equipment relation topological diagram, and factory sand table construction is achieved, wherein the transmission and loading efficiency of the three-dimensional model is effectively improved through model light weight processing, the display effect of the system is improved, and the stability of the factory sand table model is improved;
2. The fault prediction module is arranged, the fault early warning model is obtained through neural network training according to the equipment historical data, the equipment fault early warning information is obtained through the fault early warning model processing according to the virtual factory operation data, the equipment fault early warning information is uploaded to the factory sand table model, the acousto-optic positioning alarm is started, the working log is generated according to the equipment fault early warning information and is uploaded to the terminal, the production management efficiency is improved, and the equipment fault prediction early warning is realized.
Drawings
The present invention is further described below with reference to the accompanying drawings for the convenience of understanding by those skilled in the art.
FIG. 1 is a schematic flow chart of a sand table type factory production simulation prediction method of the invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention for achieving the preset aim, the following detailed description is given below of the specific implementation, structure, characteristics and effects according to the invention with reference to the attached drawings and the preferred embodiment.
Referring to fig. 1, a sand table type factory production simulation prediction method includes the following steps:
S1: acquiring plant equipment specification information, actual plant operation data and equipment rule constraint information;
S2: obtaining a three-dimensional equipment model through 3D modeling and model light weight processing according to the plant equipment specification information, obtaining a virtual entity and entity relation through neural network word segmentation model processing according to the equipment rule constraint information, obtaining an equipment relation topological graph according to the virtual entity and the entity relation, and obtaining a plant sand table model through a digital twin system according to the three-dimensional equipment model and the equipment relation topological graph;
S3: obtaining abnormal actual operation data through data processing according to the actual operation data of the factory, and inputting the abnormal actual operation data into the factory sand table model through a pre-constructed data transmission path to perform data mapping to obtain virtual factory operation data;
S4: acquiring equipment historical data, training the equipment historical data through a neural network to obtain a fault early warning model, and processing the virtual factory operation data through the fault early warning model to obtain equipment fault early warning information;
s5: uploading the equipment fault early warning information to the factory sand table model, starting sound and light positioning alarm, generating a work log according to the equipment fault early warning information, and uploading the work log to a terminal.
Step S1 specifically relates to an information acquisition module, wherein the information acquisition module is used for acquiring the specification information of equipment of a factory, the actual operation data of the factory and the constraint information of equipment rules;
The method comprises the steps of acquiring actual operation data of a factory through a preset acquisition node, wherein the actual operation data of the factory comprises equipment operation data, equipment parameter information and the like, and the equipment rule constraint information comprises equipment connection mode text description, equipment task text description and equipment function text description. The data collection provides a data basis for data analysis, ensures the accuracy of a data source and improves the decision accuracy of the data analysis.
Step S2 specifically relates to a sand table modeling module, namely, a three-dimensional equipment model is obtained through 3D modeling and model weight reduction according to the specification information of the factory equipment, virtual entities and entity relations are obtained through neural network word segmentation model processing according to the constraint information of the equipment rules, an equipment relation topological graph is obtained according to the virtual entities and the entity relations, and a factory sand table model is obtained through a digital twin system according to the three-dimensional equipment model and the equipment relation topological graph;
specifically, the model lightening process comprises the steps of saving an initial three-dimensional equipment model obtained through 3D modeling as an STL format three-dimensional equipment model, analyzing redundant dotted lines and redundant patches through lightening tools according to the STL format three-dimensional equipment model, and removing the redundant dotted lines and the redundant patches to obtain the three-dimensional equipment model;
The device relation topological graph implementation method comprises the steps of obtaining a factory device database, carrying out word segmentation processing on the device rule constraint information through a BiLSTM neural network word segmentation model to obtain a device information corpus, carrying out corpus labeling on the device information corpus according to a predefined entity type to obtain labeling information, obtaining the virtual entity through association with the factory device database according to the labeling information, processing the device information corpus through a BERT model to obtain deep semantic information, carrying out feature extraction according to the deep semantic information to obtain a feature vector, carrying out convolution operation on the feature vector through a convolution neural network to obtain a convolution result, carrying out maximum pooling on the convolution result to obtain a maximum feature value, obtaining the entity relation through the convolution neural network according to the maximum feature value, and obtaining the device relation topological graph according to the virtual entity and the entity relation.
The three-dimensional equipment model is processed through model light weight, so that the transmission and loading efficiency of the three-dimensional model is effectively improved, the display effect of the system is improved, and the stability of the factory sand table model is improved.
Step S3 specifically relates to a data processing module, wherein data anomaly statistics are obtained through data de-anomaly model processing according to the actual operation data of the factory, and a calculation formula is as follows: Wherein, Oi represents the ith actual operation data of the plant, Ki represents the data anomaly statistics, α and β represent weight coefficients, a statistics critical value is obtained, when the data anomaly statistics are greater than or equal to the statistics critical value, the actual operation data of the plant corresponding to the data anomaly statistics are marked as invalid data and deleted, when the data anomaly statistics are less than the statistics critical value, the actual operation data of the plant corresponding to the data anomaly statistics are marked as normal data, and the deisoidal actual operation data is obtained through normalization processing according to the normal data.
The actual operation data of the factory is processed through the data disagreement model to obtain disagreement actual operation data, so that errors caused by the precision of the equipment sensor in the production process are eliminated, and the data analysis precision is improved.
Step S4 and step S5 relate to a fault prediction module specifically, step S4 includes obtaining equipment history data, obtaining a fault early warning model through neural network training according to the equipment history data, and obtaining equipment fault early warning information through processing of the fault early warning model according to the virtual factory operation data.
Specifically, the fault early warning model construction step includes obtaining a data upper bound, a data lower bound and a data dimension of equipment history data, presetting that the neural network comprises N neurons, and obtaining N neuron weight vectors according to the N neurons through a random formula, wherein the random formula is expressed as: aij=rand (Uj-Lj) +lj, i=1, 2, 3..n, wherein Aij represents the neuron weight vector, U represents the upper data bound, L represents the lower data bound, j represents the data dimension;
Obtaining an optimized neural network according to the neuron weight vector, taking the equipment history data as a training set X= { X1, X2, x3...xM }, wherein M represents M pieces of equipment history data, inputting the training set into the optimized neural network, and obtaining the fault early warning model through training;
Specifically, the specific implementation step of the equipment fault early warning information includes constructing a test data set according to the virtual factory operation data, obtaining a test neuron of the test data set, obtaining a test neuron weight vector and a test neuron distance according to the test neuron, obtaining an abnormal early warning coefficient according to the test data set and the test neuron weight vector through the fault early warning model, and calculating the abnormal early warning coefficient, wherein a calculation formula is as follows: wherein i=1, 2, 3..p, D (O1, O2) represents similarity of the test dataset to the test neuron weight vector, O1 represents the test neuron weight vector, O2 represents the test dataset, P represents a test dataset dimension, wi represents the anomaly early warning coefficient, YD represents the test neuron distance; and generating the equipment fault early warning information when the abnormality early warning coefficient is larger than or equal to a preset threshold value, and generating equipment non-abnormality information when the abnormality early warning coefficient is smaller than the preset threshold value.
Step S5 specifically comprises the steps of uploading the equipment fault early warning information to the factory sand table model, starting sound and light positioning alarm, generating a work log according to the equipment fault early warning information, and uploading the work log to a terminal.
The production management efficiency is improved and the equipment fault prediction and early warning are realized through the fault prediction module.
Still further, the present application also provides a sand table type factory production simulation prediction system, comprising:
The data acquisition module is used for acquiring the specification information of the equipment of the factory, the actual running data of the factory and the constraint information of the equipment rules;
The sand table modeling module is used for obtaining a three-dimensional equipment model through 3D modeling and model weight reduction according to the equipment specification information of the factory, obtaining a virtual entity and entity relation through neural network word segmentation model processing according to the equipment rule constraint information, obtaining an equipment relation topological graph according to the virtual entity and the entity relation, and obtaining a factory sand table model through a digital twin system according to the three-dimensional equipment model and the equipment relation topological graph;
the data processing module is used for obtaining abnormal actual operation data through data processing according to the actual operation data of the factory, inputting the abnormal actual operation data into the factory sand table model through a pre-constructed data transmission path, and performing data mapping to obtain virtual factory operation data;
the fault prediction module is used for acquiring equipment historical data, obtaining a fault early warning model through neural network training according to the equipment historical data, obtaining equipment fault early warning information through fault early warning model processing according to the virtual factory operation data, uploading the equipment fault early warning information to the factory sand table model, starting acousto-optic positioning alarm, generating a work log according to the equipment fault early warning information, and uploading the work log to a terminal.
The working principle and the using flow of the invention are as follows:
The method comprises the steps of collecting plant equipment specification information, plant actual operation data and equipment rule constraint information through an information collection module, obtaining a three-dimensional equipment model through 3D modeling and model light weight processing according to the plant equipment specification information, obtaining an equipment relationship topological diagram according to the equipment rule constraint information, obtaining a plant sand table model through a digital twin system according to the three-dimensional equipment model and the equipment relationship topological diagram, obtaining abnormal actual operation data through data processing according to the plant actual operation data, carrying out data mapping on the abnormal actual operation data through the plant sand table model to obtain virtual plant operation data, obtaining equipment historical data, obtaining a fault early warning model through neural network training according to the equipment historical data, obtaining equipment fault early warning information through the fault early warning model processing according to the virtual plant operation data, starting acousto-optic positioning alarm, generating a work log according to the equipment fault early warning information, and uploading the work log to a terminal.
The program code embodied in the methods of embodiments of the present invention may be transmitted over any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, etc., or any suitable combination of the foregoing. Computer program code for carrying out operations of the present invention may be written in one or more programming languages, including an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).
The present invention is not limited to the above embodiments, but is not limited to the above embodiments, and any technical modifications, equivalents and modifications made to the above embodiments according to the technical principles of the present invention can be made by those skilled in the art without departing from the scope of the invention.