CN116663402A - Mixing station concrete quality prediction method based on digital twin - Google Patents

Mixing station concrete quality prediction method based on digital twin Download PDF

Info

Publication number
CN116663402A
CN116663402A CN202310557712.2A CN202310557712A CN116663402A CN 116663402 A CN116663402 A CN 116663402A CN 202310557712 A CN202310557712 A CN 202310557712A CN 116663402 A CN116663402 A CN 116663402A
Authority
CN
China
Prior art keywords
concrete
mixing station
production
machine learning
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310557712.2A
Other languages
Chinese (zh)
Inventor
张晓晖
白文奇
赵力
谷振东
刘青
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian University of Technology
Original Assignee
Xian University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian University of Technology filed Critical Xian University of Technology
Priority to CN202310557712.2A priority Critical patent/CN116663402A/en
Publication of CN116663402A publication Critical patent/CN116663402A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06395—Quality analysis or management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Software Systems (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Computer Hardware Design (AREA)
  • Educational Administration (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Medical Informatics (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Physics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Manufacturing & Machinery (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Preparation Of Clay, And Manufacture Of Mixtures Containing Clay Or Cement (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a mixing station concrete quality prediction method based on digital twin, which comprises the following steps: 1) Establishing a geometric model of the mixing station digital twin body; 2) Constructing a data model of the mixing station digital twin body; 3) Establishing a mixing station digital twin body; 4) Generating an electronic account for concrete production; 5) Processing the concrete production electronic account by a digital twin body of the mixing station to obtain a production virtual image of the mixing station; 6) And excavating historical data from the concrete production electronic ledger, training a machine learning model on line, and predicting the quality of concrete produced subsequently by using the trained machine learning model. The method can improve the visibility of operators of the mixing station to each production link, thereby being beneficial to formulating a more accurate concrete proportioning scheme, realizing the fine management of the production process and solving the problems of blind production and rough management existing in the current concrete mixing station.

Description

Mixing station concrete quality prediction method based on digital twin
Technical Field
The invention belongs to the technical field of digital twin, and relates to a mixing station concrete quality prediction method based on digital twin.
Background
In the existing concrete production process, the whole production line produces concrete meeting the casting strength of project parts according to project design requirements, the production line comprises a production center such as a mixing station, and also comprises departments corresponding to storage and transportation of raw materials and finished concrete and quality inspection links, and all the departments cooperate with each other to finish the production work of the concrete. To ensure that the delivered concrete is acceptable. The mixing station and each department are often mutually matched in an alternating mode, and multiple cross tests exist on the quality of the concrete to ensure that the quality of the concrete is excellent.
However, the production and quality management of the current concrete are respectively completed by different units in the production link, the relation is relatively loose, and the coordination management capability on quality problems is poor. For example, for a concrete mixing station matched with each large infrastructure construction project, as a large-scale Cheng Wangwang is built in a mountain canyon, network coverage is small, communication capacity is poor, most of work is completed in a manual mode, and the quality of concrete is unqualified due to manual negligence in any link, so that the tracing is extremely difficult.
The digital twin technology takes physical entities in a real scene as attention objects, establishes a multi-physical field and multi-scale virtual model, enables the virtual model to be used as a digital mirror image of the physical entities by means of mapping feedback of information between the physical entities, always keeps consistency with the states of the physical entities, and achieves the purposes of maintaining, optimizing and the physical entities by means of data fusion analysis, artificial intelligence and the like.
The digital twin technology can realize the management of covering the whole production link, effectively improve the utilization degree of data in the concrete production process, realize the advanced judgment of the concrete quality through the introduction of artificial intelligence, and provide basis for quality management and proportion optimization.
In view of the above technical drawbacks of the prior art, there is an urgent need to develop a mixing station concrete quality prediction method based on digital twinning.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provides a mixing station concrete quality prediction method based on digital twinning, so as to solve the problems in the aspect of concrete quality guarantee in the current production process.
In order to achieve the above object, the present invention provides the following technical solutions:
a mixing station concrete quality prediction method based on digital twinning is characterized by comprising the following steps:
1) Acquiring attribute parameters of each sub-physical entity of the mixing station, and establishing a geometric model of a digital twin body of the mixing station;
2) Acquiring production state data of the mixing station in the running period in real time, and constructing a data model of a digital twin body of the mixing station;
3) Combining the data model with the geometric model to establish a mixing station digital twin body;
4) Acquiring production record data of a mixing station in real time, establishing a relation between the production record data and the production state data according to production batches to obtain detailed process record data of concrete in each production batch, and storing the detailed process record data to generate a concrete production electronic ledger;
5) Processing the concrete production electronic ledger by the digital twin body of the mixing station to obtain a production virtual image of the mixing station;
6) And excavating historical data from the concrete production electronic ledger, training a machine learning model on line, and predicting the quality of concrete produced subsequently by using the trained machine learning model.
Preferably, the processing of the concrete production electronic ledger by the digital twin body of the mixing station further comprises: judging whether the working state of the mixing station reflected in the concrete production electronic ledger is abnormal, and if the working state is abnormal, warning the abnormal state.
Preferably, the obtaining attribute parameters of each sub-physical entity of the mixing station in the step 1) specifically includes the steps of establishing a geometric model of a digital twin body of the mixing station;
1.1 Dividing the mixing station into a plurality of sub-physical entities according to functions, wherein the sub-physical entities comprise a storage system, a metering system, a conveying system, a liquid supply system, a pneumatic system, a stirring system, a main building, a control room and a dust removal system;
1.2 Acquiring attribute parameters of the mixing station, and constructing a geometric model of a digital twin body of the mixing station based on the attribute parameters, wherein the attribute parameters comprise appearance shapes, dimension sizes, internal structures, spatial positions, postures and assembly relations of all sub-physical entities.
Preferably, the production record data of the mixing station obtained in real time in the step 4) comprises raw material proportioning information, raw material monitoring information, feeding errors, concrete strength grade, slump and production amount.
Preferably, in step 6), historical data is mined from the concrete production electronic ledger, a machine learning model is trained online, and the prediction of the concrete quality of subsequent production by using the trained machine learning model specifically comprises:
6.1 Constructing a data set by using the production record data in the concrete production electronic ledger, respectively analyzing variables which can obviously influence each strength index of the concrete by a principal component analysis method, and determining the variables as main variables of each strength index of the concrete;
6.2 Respectively establishing machine learning models of all the strength indexes of the concrete by adopting an incremental machine learning algorithm, and respectively training the machine learning models of all the strength indexes of the concrete by taking main variables of all the strength indexes of the concrete as independent variables;
6.3 And predicting the quality of the concrete produced later by using a machine learning model of each strength index of the trained concrete.
Preferably, the various strength indexes of the concrete comprise compressive strength, tensile strength, freezing resistance and impermeability.
Preferably, after the concrete production electronic ledger stores the detailed process record data of each production batch of concrete, the main variables of each strength index of the new concrete are extracted from the new detailed process record data, and the machine learning model of each strength index of the concrete is trained by utilizing the main variables of each strength index of the new concrete, so that the parameter updating of the machine learning model of each strength index of the concrete is completed.
Preferably, the incremental machine learning algorithm is a Huo Fuding tree algorithm, and the machine learning model of each strength index of the concrete is formed by adding a sliding window on the basis of a Huo Fuding tree model, meanwhile, setting Huo Fuding tree models to train the standby subtrees by utilizing data in the sliding window in the background of each node, and completing node splitting when the node splitting gain of the standby subtrees is significantly larger than that of the current subtrees.
Preferably, the mixing station concrete quality prediction method based on digital twin further comprises displaying the prediction results of the production virtual image and the concrete quality of subsequent production through a user front end.
Compared with the prior art, the mixing station concrete quality prediction method based on digital twin has one or more of the following beneficial technical effects:
1. according to the invention, the digital twin body of the mixing station is used for processing the electronic account data operated by the mixing station and the acquired current state data to obtain the virtual image data of the mixing station, the virtual image data can be fed back to the front end of a user and directly displayed to operators on a concrete production site, the operators can grasp the production state of the mixing station in real time, the abnormal state generated in the mixing station is reflected by the acquired state data, and the operators can find faults in time conveniently.
2. According to the invention, by means of the flight data stored in the electronic account data of the mixing station, the strength prediction model of the concrete is trained on line by using an incremental machine learning method, the prediction model is tightly combined with the digital twin body, the strength index of the newly mixed concrete is predicted, operators are assisted to judge the qualification of the concrete quality in a contrasting manner, and the feeding proportioning scheme can be continuously optimized according to errors existing between the theoretical mixing strength and the predicted strength.
Drawings
FIG. 1 is a flow chart of a digital twinning-based mixing station concrete quality prediction method of the present invention.
Fig. 2 is a flow chart of a Huo Fuding tree algorithm employed by the present invention.
Fig. 3 is a graph showing the comparison of the predicted index results of the Huo Fuding tree algorithm after 300 iterations, taking the concrete compressive strength prediction as an example.
Detailed Description
The invention is further illustrated by the following examples in conjunction with the accompanying drawings, which are not to be construed as limiting the scope of the invention.
Aiming at the problems in the aspect of concrete quality guarantee in the current production process, the invention provides a mixing station concrete quality prediction method based on digital twinning, which can improve the visibility of mixing station operators to each production link, thereby being beneficial to formulating a more accurate concrete proportioning scheme, realizing the fine management of the production process and solving the problems of blind production and rough management of the current concrete mixing station.
FIG. 1 shows a flow chart of the mixing station concrete quality prediction method based on digital twinning of the invention. As shown in fig. 1, the mixing station concrete quality prediction method based on digital twin of the invention comprises the following steps:
1. and obtaining attribute parameters of each sub-physical entity of the mixing station, and establishing a geometric model of the digital twin body of the mixing station.
Specifically, establishing a geometric model of the mixing station digital twin body comprises the following steps of;
1. the mixing station is functionally divided into a number of sub-physical entities. The sporophores include material storing system, metering system, conveying system, liquid supplying system, pneumatic system, stirring system, main building, control room, dust eliminating system, etc. to complete the storing, metering, conveying, stirring, discharging, controlling and other tasks of concrete material.
2. And acquiring attribute parameters of the mixing station, and constructing a geometric model of the digital twin body of the mixing station based on the attribute parameters. The attribute parameters comprise the appearance shape, the size, the internal structure, the space position, the gesture and the assembly relation of each sub-physical entity.
The property parameters of the mixing station can be obtained from design documents of the mixing station, can be obtained by measurement and the like.
2. Production state data of the mixing station during operation are collected in real time, and a data model of a digital twin body of the mixing station is constructed.
In the invention, the production state data of the concrete mixing station during operation can be acquired by utilizing the sensors of the sub-physical entities arranged at the mixing station, and the data model of the mixing station digital twin body can be constructed by utilizing the production state data.
3. And combining the data model with the geometric model to establish a mixing station digital twin body.
Substituting the collected production state data into the geometric model according to a preset mapping relation, and dynamically updating the state of the geometric model to obtain the mixing station digital twin body.
4. And acquiring production record data of a mixing station in real time, establishing a relation between the production record data and the production state data according to production batches to obtain detailed process record data of concrete in each production batch, and storing the detailed process record data to generate the concrete production electronic ledger.
In the invention, the production record data of the mixing station obtained in real time comprises raw material proportioning information, raw material monitoring information, feeding errors, concrete strength grade, slump, production amount and the like. By producing the recorded data, the production capacity of the mixing station can be obtained.
5. And processing the concrete production electronic ledger by the digital twin body of the mixing station to obtain a production virtual image of the mixing station.
And inputting the concrete production electronic ledger into the digital twin body of the mixing station in real time, and updating the production state and the production record of the digital twin body of the mixing station in real time, so that the production virtual image of the mixing station can be obtained.
Of course, the processing of the concrete production electronic ledger by the digital twins of the mixing station further comprises: judging whether the working state of the mixing station reflected in the concrete production electronic ledger is abnormal, and if the working state is abnormal, warning the abnormal state.
In the present invention, the concrete mixing station operating state may be determined based on a preset rule. For example, whether the air circuit pressure and the concrete arch air pressure are in a reasonable interval or not is detected, and whether the air compressor runs normally or not can be judged; detecting the liquid flow can judge whether the operation pipelines of the water supply system and the additive system are unobstructed. By judging, the abnormal working state of each sub-physical entity of the concrete mixing station can be recorded and warned.
6. And excavating historical data from the concrete production electronic ledger, training a machine learning model on line, and predicting the quality of concrete produced subsequently by using the trained machine learning model.
In the invention, historical data is mined from the concrete production electronic ledger, a machine learning model is trained on line, and the concrete quality of subsequent production is predicted by using the trained machine learning model specifically comprises the following steps:
1. and constructing a data set by using the production record data in the concrete production electronic ledger, respectively analyzing variables which can obviously influence each strength index of the concrete by a principal component analysis method, and determining the variables as main variables of each strength index of the concrete.
Wherein, each strength index of the concrete comprises compressive strength, tensile strength, freezing resistance, impermeability and the like.
By the principal component analysis method, variables which can significantly affect the compressive strength, tensile strength, antifreeze strength, impervious strength and the like of concrete can be obtained from a dataset constructed by using production record data in the concrete production electronic ledger, and taken as principal variables.
That is, each strength index of the concrete, such as compressive strength, tensile strength, antifreeze strength or impervious strength, is used as a prediction target, and a principal component analysis method is used to analyze which data in the data set has a significant influence on the prediction target, and the data is used as a principal variable of the prediction target.
Specifically, a data set is constructed by using the existing records in the concrete production electronic ledger, the variance contribution rate of each principal component is calculated through a principal component analysis method, and the principal component with high contribution rate is selected as a principal variable.
Extracting a record existing at the current moment in the concrete production electronic ledger, constructing a sample array by using variables in the record, wherein the constructed sample array is a set of n p-dimensional vectors, and carrying out standardized change on sample array elements as shown in a formula (1):
wherein x is ij Representing the ith value in the jth feature,representing the average value of the j-th feature,a standardized array Z is obtained.
For the standardized matrix Z, the correlation coefficient matrix is calculated as shown in the formula (2):
wherein the method comprises the steps of
Solving characteristic equation |R-lambda I of sample correlation matrix R p And (3) obtaining p characteristic roots, and determining m principal components with the principal component information utilization rate reaching a set index.
2. And respectively establishing machine learning models of all the strength indexes of the concrete by adopting an incremental machine learning algorithm as a prediction model of all the strength indexes, and respectively training the machine learning models of all the strength indexes of the concrete (namely, the prediction models of all the strength indexes) by taking main variables of all the strength indexes of the concrete as independent variables.
In the invention, an incremental machine learning algorithm-Huo Fuding tree algorithm is adopted to respectively establish machine learning models of various strength indexes of concrete, namely, a compressive strength machine learning model, a tensile strength machine learning model, an anti-freezing strength machine learning model, an anti-seepage strength machine learning model and the like, and the machine learning models are used as prediction models of compressive strength, tensile strength, anti-freezing strength, anti-seepage strength and the like. Based on the existing data, namely, the main variables of each strength index of the concrete, training of a prediction model corresponding to the strength index is performed.
The Huo Fuding tree model utilizes Huo Fuding bounds to enable determination of optimal segmentation properties on nodes in relatively small scale training data. The expression of Huo Fuding is shown in formula (3):
consider a real random variable R, whose range is R, assuming that n independent observations have been made for the variable, and calculate their averageAccording to the Huo Fuding inequality, the mean value of the variables is calculated to be at least +.>
Huo Fuding has the property of generating a probability distribution independent of observations, at the cost of more observations to achieve the same delta and epsilon, with the boundary being more conservative than the boundary depending on the distribution. Let G (X) i ) Is a method for calculating information gain, wherein X is a And X b For the two attributes with the largest information gain, the information gain difference of the two attributes is expressed asAs shown in formula (4):
if the information gain difference of the two attributes is greater than Hough Ding Jie, i.eThen it can prove X a The confidence level of the attribute with the maximum information gain is 1-delta, and X can be selected at the moment a As a split point, the leaf node is changed to a branch node.
In the Huo Fuding tree model, information gain is used to select the best split attributes on the root node and the internal nodes. The information gain IG is the difference between the information entropy H (D) and the corresponding conditional entropy H (d|x). The information entropy and the conditional entropy are calculated as shown in formulas (5), (6) and (7):
IG＝H(D)-H(D|X) (7)
when constructing the Huo Fuding tree model, m are observed on one leaf nodeAfter independent object, let X a Attribute IG (X a )，X b Attribute IG (X b ) Then pass through IG (X a ) Subtracting IG (X) b ) A new variable Δig can be obtained. If ΔIG is greater than ε, then select X a As the splitting attribute, if Δig is not obvious, it takes a long time to determine the optimal splitting attribute. When the method applies Huo Fuding tree model to process regression problem, variance reduction is selected to replace information gain function in classification tree, S is data in nodes, total data amount is N, attribute A of data is selected and h is used as shown in formulas (8) and (9) A The current node can be divided into two parts for boundary, where S R And S is L Respectively data in two parts, the data quantity is N respectively R And N L I.e. s=s L +S R ，N＝N L +N R 。
The Huo Fuding tree algorithm is based on the premise that the data flow analyzed and processed by the algorithm is distributed smoothly, and the Huo Fuding tree model itself does not complete the design of model self update for the outdated samples, so that the Huo Fuding tree model cannot process the sample data characteristic offset problem in the data flow. The invention adds a sliding window structure on the basis of the Huo Fuding tree model, so that the Huo Fuding tree model can continuously update the concerned data range, and meanwhile, the Huo Fuding tree model is arranged on the background of each node to train the standby subtree by utilizing the data in the sliding window, and when the node splitting gain of the standby subtree is obviously larger than that of the current subtree (for example, three times that of the current subtree), the splitting of the nodes is completed. Fig. 2 shows a Huo Fuding tree algorithm flow chart of the present invention.
In the improved Huo Fuding tree model of the present invention, the tree isEach internal node is provided with a replacement subtree list, the subtree starts training when the concept drift of the data stream occurs, namely, when the information gain of one other attribute on the node is found to be better than the current attribute gain, the generation of a replacement subtree is started, and the new attribute gain difference satisfies the following conditionsAnd->The replacement of the subtrees is then completed and the modified Huo Fuding tree model protects the model from older outdated data generation while maintaining the lightweight of the predictive model. Taking a concrete compressive strength prediction model as an example, the prediction index result pairs of the model before and after improvement after 300 iterations are shown in fig. 3.
In summary, the Huo Fuding tree model is a classical incremental online learning algorithm. Because online learning is different from offline learning by a learning strategy of adjusting model parameters by batch data, the model is easier to be unstable, the Huo Fuding tree algorithm adopts Huo Fuding world to ensure that node samples can approach to overall distribution with arbitrary precision, the Huo Fuding tree is applied to classification problems, and the Huo Fuding tree model after being changed can be used for predicting the regression problems by changing the information gain function of the original algorithm.
Meanwhile, the Huo Fuding tree model is not designed by pruning, which means that the algorithm can only process data streams which are distributed smoothly, and when the data streams are changed dynamically, the Huo Fuding tree model cannot adjust outdated parts in the tree structure. The improvement of the algorithm mainly comprises two aspects, namely, firstly, a concept drift detector (sliding window structure) is added in a data stream, a model can only consider recent data contained in a sliding window, and the previous model needs to consider all historical data on nodes, so that the model is protected from outdated data. Secondly, when the gain of the current attribute in the node is reduced, the model starts to train a standby subtree for the corresponding node in the background, and when the gain of the selected attribute of the standby subtree is obviously higher than the gain of the current attribute, the replacement of the subtree is completed. This is very similar to the process of new node splitting, but the requirements are more stringent, effectively preventing the generation of excessive replacement subtrees. The splitting growth mode of the replacement subtrees is the same as that of the decision tree, each replacement subtree can acquire the data flow acquired by the corresponding node of the replacement subtree to train the subtree, the adaptation of the prediction model is substantially enhanced through the improvement, the light weight of the model is maintained, and a better prediction effect is obtained in the test of using the existing data to take the concrete compressive strength as an example.
In addition, in the present invention, preferably, after the concrete production electronic ledger stores detailed process record data of each production batch of concrete, the main variables of each strength index of the new concrete are extracted from the new detailed process record data, and the machine learning model of each strength index of the concrete is trained by using the main variables of each strength index of the new concrete, so as to complete parameter updating of the machine learning model of each strength index of the concrete.
Therefore, after the concrete production electronic ledger finishes the complete record of each concrete production, the main component feature vector is extracted from the new record to serve as a new sample, the new sample is used for training the prediction model, and parameter updating of the model is completed, so that the model is more in line with the actual situation.
3. And predicting the quality of the concrete produced subsequently by using a machine learning model of each strength index of the trained concrete.
After the machine learning model of each strength index of the concrete is established and trained, main variables of each strength index of the concrete can be extracted from the concrete production electronic ledger generated in real time in the production process and input into the corresponding machine learning model, and then the concrete quality prediction result of the mixing station can be obtained.
To realize the mixing station concrete quality prediction method based on digital twin, the mixing station concrete quality prediction system based on digital twin is matched. The prediction system comprises a data acquisition end, a server end, a user front end and a background feedback end.
The data acquisition end comprises sensors and related intelligent terminal equipment distributed in each sub-physical entity by the concrete mixing station. The data acquired by the data acquisition end is transmitted to the server end in real time, the real-time mapping of the data in the digital twin body is completed by the server end according to the pre-established mapping relation of the digital twin body, the abnormal production state of the mixing station is warned, and the background feedback end sends an abnormal signal to the mixing station main control console. And the server side also correlates the production state data with the production record data and stores the production state data and the production record data in an electronic ledger for concrete production, and extracts the data in the electronic ledger to train a prediction model. Operators can check the production state in the current mixing station and the prediction result of the strength index of the concrete in the front end of the user, so that the problems of blind production and rough management in the traditional concrete production are solved to a certain extent.
The above examples of the present invention are merely illustrative of the present invention and are not intended to limit the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. Not all embodiments are exhaustive. All obvious changes or modifications which come within the spirit of the invention are desired to be protected.

Claims (9)

1. A mixing station concrete quality prediction method based on digital twinning is characterized by comprising the following steps:
1) Acquiring attribute parameters of each sub-physical entity of the mixing station, and establishing a geometric model of a digital twin body of the mixing station;
2) Acquiring production state data of the mixing station in the running period in real time, and constructing a data model of a digital twin body of the mixing station;
3) Combining the data model with the geometric model to establish a mixing station digital twin body;
4) Acquiring production record data of a mixing station in real time, establishing a relation between the production record data and the production state data according to production batches to obtain detailed process record data of concrete in each production batch, and storing the detailed process record data to generate a concrete production electronic ledger;
5) Processing the concrete production electronic ledger by the digital twin body of the mixing station to obtain a production virtual image of the mixing station;
6) And excavating historical data from the concrete production electronic ledger, training a machine learning model on line, and predicting the quality of concrete produced subsequently by using the trained machine learning model.
2. The method for predicting concrete quality in a mixing station based on digital twinning of claim 1, wherein processing the concrete production electronic ledger by the digital twinning of the mixing station further comprises: judging whether the working state of the mixing station reflected in the concrete production electronic ledger is abnormal, and if the working state is abnormal, warning the abnormal state.
3. The mixing station concrete quality prediction method based on digital twin according to claim 1, wherein the obtaining attribute parameters of each sub-physical entity of the mixing station in step 1) specifically comprises the steps of;
1.1 Dividing the mixing station into a plurality of sub-physical entities according to functions, wherein the sub-physical entities comprise a storage system, a metering system, a conveying system, a liquid supply system, a pneumatic system, a stirring system, a main building, a control room and a dust removal system;
1.2 Acquiring attribute parameters of the mixing station, and constructing a geometric model of a digital twin body of the mixing station based on the attribute parameters, wherein the attribute parameters comprise appearance shapes, dimension sizes, internal structures, spatial positions, postures and assembly relations of all sub-physical entities.
4. The method for predicting concrete quality of mixing station based on digital twin according to claim 1, wherein the production record data of the mixing station obtained in real time in step 4) includes raw material proportioning information, raw material monitoring information, feeding errors, concrete strength grade, slump and production quantity.
5. The method for predicting concrete quality of a mixing station based on digital twinning according to claim 1, wherein in step 6), historical data is mined from the concrete production electronic ledger, a machine learning model is trained online, and the prediction of concrete quality of subsequent production by using the trained machine learning model specifically comprises:
6.1 Constructing a data set by using the production record data in the concrete production electronic ledger, respectively analyzing variables which can obviously influence each strength index of the concrete by a principal component analysis method, and determining the variables as main variables of each strength index of the concrete;
6.2 Respectively establishing machine learning models of all the strength indexes of the concrete by adopting an incremental machine learning algorithm, and respectively training the machine learning models of all the strength indexes of the concrete by taking main variables of all the strength indexes of the concrete as independent variables;
6.3 And predicting the quality of the concrete produced later by using a machine learning model of each strength index of the trained concrete.
6. The method for predicting the quality of concrete in a mixing station based on digital twinning according to claim 5, wherein the various strength indexes of the concrete comprise compressive strength, tensile strength, freezing resistance and impermeability.
7. The method for predicting the quality of concrete in a mixing station based on digital twinning of claim 6, wherein after each concrete production electronic ledger stores detailed process record data of each production batch of concrete, extracting main variables of each strength index of new concrete from the new detailed process record data, and training a machine learning model of each strength index of the concrete by using the main variables of each strength index of the new concrete to complete parameter updating of the machine learning model of each strength index of the concrete.
8. The method for predicting the quality of concrete at a mixing station based on digital twin according to claim 7, wherein the incremental machine learning algorithm is a Huo Fuding tree algorithm, and the machine learning model of each strength index of the concrete is based on a Huo Fuding tree model, wherein a sliding window is added, and a Huo Fuding tree model is set to train a spare subtree in the background of each node by utilizing data in the sliding window, and when the node splitting gain of the spare subtree is significantly larger than that of the current subtree, the node splitting is completed.
9. The method of claim 1-8, further comprising presenting the prediction of the quality of the produced virtual image and subsequently produced concrete via a user front end.
CN202310557712.2A 2023-05-17 2023-05-17 Mixing station concrete quality prediction method based on digital twin Pending CN116663402A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310557712.2A CN116663402A (en) 2023-05-17 2023-05-17 Mixing station concrete quality prediction method based on digital twin

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310557712.2A CN116663402A (en) 2023-05-17 2023-05-17 Mixing station concrete quality prediction method based on digital twin

Publications (1)

Publication Number Publication Date
CN116663402A true CN116663402A (en) 2023-08-29

Family

ID=87714508

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310557712.2A Pending CN116663402A (en) 2023-05-17 2023-05-17 Mixing station concrete quality prediction method based on digital twin

Country Status (1)

Country Link
CN (1) CN116663402A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117648746A (en) * 2023-12-11 2024-03-05 中铁一局集团有限公司 Digital twinning-based tunnel construction concrete super-consumption data statistics method and system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117648746A (en) * 2023-12-11 2024-03-05 中铁一局集团有限公司 Digital twinning-based tunnel construction concrete super-consumption data statistics method and system

Similar Documents

Publication Publication Date Title
CN112052992B (en) Deep learning-based construction project progress prediction system and method
CN106067079B (en) System and method for predicting dust haze based on BP neural network
CN114118224B (en) Neural network-based full-system telemetry parameter anomaly detection system
CN116167668B (en) BIM-based green energy-saving building construction quality evaluation method and system
CN108873829B (en) Phosphoric acid production parameter control method based on gradient lifting decision tree
CN116663402A (en) Mixing station concrete quality prediction method based on digital twin
CN117008557B (en) Production control method and system for blending type interpenetrating network thermoplastic elastomer
CN111950585A (en) XGboost-based underground comprehensive pipe gallery safety condition assessment method
CN113807004B (en) Cutter life prediction method, device and system based on data mining
WO2021007845A1 (en) Cloud-edge collaborative forecasting system and method for aluminum oxide production indexes
CN111400811B (en) Reliability analysis method for mixed uncertainty structure
CN113222140A (en) C4.5 algorithm and BP neuron-based power distribution network fault auxiliary decision-making method
CN117689256B (en) Aluminum alloy casting product quality tracing method
CN117829382A (en) Intelligent prediction method and system for highway construction progress
CN116542502B (en) Building construction period optimization system and method based on convolutional neural network
CN112819244B (en) Meteorological factor-based RF-HW water quality index hybrid prediction method
CN113222268B (en) Multi-mode reasoning-based tobacco baking quality prediction model establishment method
CN111625525B (en) Environment data repairing/filling method and system
CN103475527A (en) Network management fault reliability analyzing system and method
CN111443686A (en) Industrial alarm design method based on multi-objective optimization and evidence iterative update
CN117455321B (en) Intelligent assessment method and system for performance of fluidized solidified soil
CN116308569B (en) Engineering cost auxiliary index management system
Sun et al. Early warning of tobacco production quality based on deep learning algorithm
CN118211811B (en) Visual digital management system and method based on gas management
Zhang et al. Research on full-process product defect traceability analysis technology based on workshop big data

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination