CN114742260A - Method, system and medium for predicting large-volume concrete temperature monitoring data - Google Patents

Method, system and medium for predicting large-volume concrete temperature monitoring data Download PDF

Info

Publication number
CN114742260A
CN114742260A CN202210190145.7A CN202210190145A CN114742260A CN 114742260 A CN114742260 A CN 114742260A CN 202210190145 A CN202210190145 A CN 202210190145A CN 114742260 A CN114742260 A CN 114742260A
Authority
CN
China
Prior art keywords
data
temperature
model
value
lstm
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
CN202210190145.7A
Other languages
Chinese (zh)
Inventor
吴红刚
陈浩
孔庆祥
李永强
袁荣涛
赵忠虎
张俊德
杨波
周威扬
朱兆荣
牌立芳
赵守全
康万鹏
陈康帅
程飞
杨景川
黄强斌
李在胜
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.)
Northwest Research Institute Co Ltd of CREC
Original Assignee
Northwest Research Institute Co Ltd of CREC
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 Northwest Research Institute Co Ltd of CREC filed Critical Northwest Research Institute Co Ltd of CREC
Priority to CN202210190145.7A priority Critical patent/CN114742260A/en
Publication of CN114742260A publication Critical patent/CN114742260A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Abstract

The invention discloses a method, a system and a medium for predicting large-volume concrete temperature monitoring data, which relate to the technical field of concrete temperature monitoring and solve the problems of poor prediction precision and low reliability of the conventional prediction method; then, after correlation analysis is carried out to screen important variables, a three-dimensional data set is constructed according to the total length of data, time step and the number of the variables, 70% of the reconstructed data set is used as a training set, and the rest of data is used as a test set; and then, through a training set optimization model, a test set is repeatedly adjusted and trained to obtain the MAE value of the optimal LSTM prediction model, so that a final LSTM prediction model is obtained, and finally, required input variable data are imported into the optimal LSTM model to carry out temperature prediction of target point positions.

Description

Method, system and medium for predicting large-volume concrete temperature monitoring data
Technical Field
The invention relates to the technical field of concrete temperature monitoring, in particular to a method, a system and a medium for predicting large-volume concrete temperature monitoring data.
Background
In the building standard regulation of 'large-volume concrete construction code' GB 50496-2009: mass concrete with the minimum geometric dimension of concrete structure bodies not less than 1m, or concrete which is expected to cause harmful crack generation due to temperature change and shrinkage caused by hydration of cementing materials in the concrete, and is called large-volume concrete; large-volume concrete construction is often involved in modern buildings, such as high-rise building foundations, large equipment foundations, water conservancy dams and the like; when the temperature difference between the inside and the outside of the concrete is large, the concrete can generate temperature cracks, and the structural safety and the normal use are influenced; therefore, it must be fundamentally analyzed to ensure the construction quality, and thus, the temperature of the concrete needs to be measured in real time during the pouring and curing of the mass concrete, thereby preventing cracks from being generated due to too large temperature difference, which affects the construction quality and the working performance of the concrete structure.
The problem of temperature cracks caused by overlarge internal and external temperature difference easily occurs in the pouring construction of the large-volume concrete, so that the change of the internal temperature of the concrete is an important factor influencing the quality of the structure, and the monitoring of the internal temperature of the large-volume concrete is an important means for ensuring that the temperature does not exceed the specified requirement.
A large number of methods are also accumulated for the temperature prediction of the mass concrete, for example, formula fitting is adopted according to monitoring data, and then the predicted temperature at a certain moment is obtained by a Lagrange difference method; correcting formula parameters in real time by adopting a concrete temperature change theoretical calculation formula and combining with actually acquired data types, and further predicting the concrete temperature; acquiring a concrete initial temperature curve, acquiring a dam temperature field simulation curve through a concrete dam thermal parameter set, performing inversion correction on the initial concrete dam thermal parameter set, calculating a measured point temperature field simulation curve again, and performing temperature prediction through the simulation curve.
However, the machine learning algorithm is applied less in temperature prediction, and common calculation methods are mostly an averaging method, a high-order polynomial fitting method and an Ar ima method, but the temperature prediction is time-sequential, and these methods cannot well solve the problems of poor prediction accuracy, low reliability and the like caused by the problem.
Disclosure of Invention
The invention aims to: the invention solves the technical problems of poor prediction precision and low reliability of the existing prediction method, provides a temperature sensor arrangement mode and introduces an LSTM recurrent neural network algorithm to predict the temperature monitoring data of mass concrete; the method is different from the traditional method for predicting the temperature by fitting a temperature curve with single data, fully considers the time sequence characteristics of the temperature data and the influence of multiple positions and multiple factors on the temperature of the target variable, and improves the accuracy and the effectiveness of temperature prediction.
The technical scheme adopted by the invention is as follows: a method for predicting large-volume concrete temperature monitoring data comprises the following steps:
step one, data acquisition: arranging sensors according to monitoring point positions, setting sampling time intervals, acquiring ambient temperature and humidity data in a long period of time and temperature monitoring data at different positions of the large-volume concrete, and combining the data to form an initial data set;
step two, data preprocessing: carrying out outlier inspection processing on the initial data set acquired in the step one, carrying out Savitzky-Golay filtering smoothing denoising processing after the outlier processing is finished, and carrying out dimension reduction and normalization processing on the data by adopting a Z-score standardization method after the data denoising is finished to obtain target variable data (namely target measuring point temperature data) conforming to standard normal distribution and input variable data;
step three, correlation analysis: analyzing the target variable and the input variable which are in accordance with the standard normal distribution after being preprocessed in the step two by adopting a Pearson correlation coefficient method, judging the correlation degree of the environmental temperature, the humidity data and the temperature data of other measuring points and the temperature change of the target measuring point according to the calculation result, and further screening important variables;
fourthly, reconstructing the data set: constructing a three-dimensional data set by using the important variables screened in the step three, wherein the formats are total data length, time step and variable number, randomly disordering the data set by taking the time step as a unit, taking 70% of the reconstructed data set as a training set, and taking the rest data as a test set;
step five, building and optimizing an LSTM model: constructing an LSTM network model, respectively training and testing the LSTM network model by using a training set and a testing set, and continuously optimizing the LSTM model according to an output MAE value;
step six, using the historical data obtained after the processing in the step two for model training and testing to obtain an optimized LSTM model, then inputting the data collected in real time into the optimized model, and outputting the prediction result of the target monitoring point.
Step one, the monitoring point location arrangement sensors are arranged on specified measuring points, and the spacing distance between the sensors on the same horizontal plane is controlled within the interval range of 5 m; the sensors on the same axial direction are embedded along the axial direction according to different depths, and the sampling frequency of data acquisition is 1-3 min/time.
Step two, the abnormal value checking processing is that when short-time data abnormality occurs, the abnormal data processing can be filled by adopting a data substitution method and a weighted average method.
The step two, the Savitzky-Golay filtering smoothing denoising treatment is to use a least square method to regress a small window of data to a polynomial, and then use the polynomial to estimate a point at the center of the window, so that the smoothness of a data curve can be obviously improved, and the interference of data noise can be reduced; Savitzky-Golay smoothing equation:
Figure BDA0003524231000000031
in the formula: x is the number ofk，smoothThe kth data point after the smoothing processing is carried out;
Figure BDA0003524231000000032
is the kth mean data point; h is a coefficient value and H ═ 2W; w is equal to half the amount of data within the window length; x is the number ofk+iIs the kth original data point; h isiIs a smoothing coefficient used for reducing the influence of smoothing on useful information and improving the disadvantages of a smoothing and denoising algorithm,
Figure BDA0003524231000000033
polynomial fitting solution can be performed based on the principle of two multiplications.
And step two, the Z-score standardization method carries out dimension removal and normalization treatment, and the operation formula is as follows:
z-score normalization:
Figure BDA0003524231000000034
in the formula: μ is the mean of all sample data; σ is the standard deviation of all sample data; x is the number of*The data values are normalized; x is the original data value.
And step two, the target variable data are measured point temperature data needing temperature prediction, and the input variable data are collected environment temperature data, environment humidity data and other measured point temperature data.
Step three, the Pearson correlation coefficient (r)x，y) For examining the degree of correlation between two variables, the specific formula is as follows:
Figure BDA0003524231000000041
in the formula: xiThe value of any moment of the input variable is obtained;
Figure BDA0003524231000000042
as the mean value Y of the input variablesiThe value of the target variable at any moment;
Figure BDA0003524231000000043
is the mean of the target variable; r isx，yIs the Pearson correlation coefficient between two variables; wherein, Pearson's correlation coefficient (r)x，y) The degree of correlation of the variables can be judged in the following ranges: 0.0-0.2 are very weakly or not correlated; 0.2-0.4 weakly correlated; 0.4-0.6 moderate correlation; 0.6-0.8 strong correlation; 0.8-1.0 are strongly correlated.
And fifthly, the MAE value is an average value of absolute values of errors of the predicted value and the true value, and a specific formula is as follows:
Figure BDA0003524231000000044
in the formula: y isiA prediction result x at a certain moment output by the modeliIs the true result at a certain moment.
A bulk concrete temperature monitoring data prediction system comprising:
the data acquisition module is used for arranging sensors according to monitoring point positions, setting sampling time intervals, acquiring ambient temperature and humidity data in a long period of time and temperature monitoring data of different positions of mass concrete, and combining the data to form an initial data set;
the data preprocessing module is used for carrying out abnormal value inspection processing on the initial data set acquired in the step one, carrying out Savitzky-Golay filtering smoothing noise reduction processing after the abnormal value processing is finished, and carrying out dimensionless and normalization processing on the data by adopting a Z-score standardization method after the data noise reduction is finished to obtain target variable data and input variable data which accord with standard normal distribution;
the relevance analysis module is used for analyzing the target variable and the input variable which are subjected to preprocessing in the second step and conform to the standard normal distribution by using a Pearson correlation coefficient method, judging the relevance degree of the environmental temperature, the humidity data and the temperature data of other measuring points and the temperature change of the target measuring points according to the calculation result, and further screening important variables;
the data set reconstruction module constructs a three-dimensional data set by using the important variables screened in the step three, wherein the format is total data length, time step and variable number, the data set is randomly disordered by taking the time step as a unit, 70% of the reconstructed data set is used as a training set, and the rest data set is used as a test set;
the LSTM model building and optimizing module is used for building an LSTM network model, importing a training set for model training, continuously optimizing and adjusting the weight and the threshold value in the model, importing a test set into the optimized model, judging the precision of a model prediction result according to the MAE value output by the model, repeatedly adjusting and training to obtain the MAE value of the optimal LSTM prediction model, and further obtaining a final LSTM prediction model;
and the temperature prediction module is used for importing the required input variable data into the optimal LSTM model to predict the temperature of the target variable.
A computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method for predicting bulk concrete temperature monitoring data as described above.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
compared with other machine learning algorithms, the deep learning algorithm LSTM is introduced, so that the method has a better prediction effect on data with time sequence characteristics; when other machine learning algorithms perform data analysis, for example, the BP neural network and the SVM regard the analyzed data as independent continuous variables, and the temperature data are ignored to have the time sequence characteristic, so that great errors are brought to long-term temperature data prediction. Although a common RNN recurrent neural network has a time memory function, important information is easily lost for long-sequence data, and a large deviation of predicted data is easily caused. The LSTM neural network is provided with a memory gate, a forgetting gate and an output gate, so that important information in long data can be retained, invalid information in short-term data can be removed, the problem of loss of important information in long-term sequence data can be effectively solved, the problem of gradient explosion is effectively avoided, and prediction of temperature data is facilitated.
In addition, compared with the traditional temperature prediction method, the method fully considers the influence of multiple factors on the temperature data at a certain position, and has higher precision instead of only taking the target variable data as a unique analysis object for interpolation, fitting and prediction.
Drawings
The invention will now be described, by way of example, with reference to the accompanying drawings, in which:
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a screenshot of application case data preprocessing code;
FIG. 3 is a diagram of an application example partitioning a training set and a test set code screenshot;
FIG. 4 is a code screenshot of a part of an LSTM network output structure built by an application example;
FIG. 5 is a scatter plot of application output results;
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
Example 1
As shown in fig. 1, the present embodiment provides a method for predicting large-volume concrete temperature monitoring data, which includes the following specific steps:
1. the sensor layout method comprises the following steps: according to the actual situation of a site, sensors are arranged on a specified measuring point, and the spacing distance between the sensors on the same horizontal plane is controlled within the 5m spacing range; the sensors in the same axis direction are embedded along the axis direction according to different depths, and the arrangement mode can effectively acquire the information of other important point positions with high temperature correlation with target measuring points, so that the accuracy of temperature prediction is improved;
2. establishing an initial data set: setting a sampling time interval, setting the sampling frequency at 1 min/time, acquiring ambient temperature and humidity data in a longer period of time and temperature monitoring data at different positions of the mass concrete, and combining the data to form an initial data set;
3. and (5) checking and processing data exception. Under the action of various interference factors on site, the monitoring data may have the condition of sudden increase or loss of individual values, but because the temperature change of concrete is not obvious within a period of time, when the data in a short time is abnormal, the abnormal data can be processed by adopting a data substitution method (filling by data after the value) and a weighted average method (filling by data average value within a period of time before and after the value);
4. the method adopts a Savitzky-Golay filter algorithm to perform data noise reduction treatment, a small window of data is regressed to a polynomial by using a least square method, and then a point at the center of the window is estimated by using the polynomial, so that the smoothness of a data curve can be obviously improved, and the interference of data noise can be reduced;
Savitzky-Golay smoothing equation:
Figure BDA0003524231000000071
in the formula: x is the number ofk，smoothThe kth data point after the smoothing processing is carried out;
Figure BDA0003524231000000072
is the kth mean data point; h is a coefficient value and H ═ 2 w; w is equal to half the amount of data within the window length; x is the number ofk+iIs the kth original data point; h isiIs a smoothing coefficient used for reducing the influence of smoothing on useful information and improving the disadvantages of a smoothing and denoising algorithm,
Figure BDA0003524231000000073
polynomial fitting solving can be carried out based on a two-multiplication principle;
5. performing data normalization, namely performing dimensionless processing on various acquired data by adopting a Z-score standardization method to ensure that the processed data conforms to standard normal distribution and is convenient for subsequent data analysis;
z-score normalization:
Figure BDA0003524231000000074
in the formula: μ is the mean of all sample data; σ is the standard deviation of all sample data; x is the number of*The data values are normalized; x is the original data value;
6. taking temperature data of a measuring point needing temperature prediction as a target variable, and taking other collected various data as input variables;
7. analyzing the degree of association, namely, in order to select an important parameter variable related to a target variable, analyzing by adopting a Pearson correlation coefficient method, judging the degree of association between the environmental temperature, humidity data and temperature data of other measuring points and the temperature change of the target measuring point according to a calculation result, and further screening the important variable;
pearson's correlation coefficient (r)x，y) Also called product-difference correlation, is used to examine the degree of correlation between two variables, and the specific formula is as follows:
Figure BDA0003524231000000081
in the formula: xiThe value of any moment of the input variable is obtained;
Figure BDA0003524231000000082
as the mean value Y of the input variablesiThe value of the target variable at any moment;
Figure BDA0003524231000000083
is the mean of the target variable; r is a radical of hydrogenx，yIs the pearson correlation coefficient between the two variables. Wherein, Pearson's correlation coefficient (r)x，y) The degree of correlation of the variables can be judged in the following ranges: 0.0-0.2 are very weakly or not correlated; 0.2-0.4 weakly correlated; 0.4-0.6 moderate correlation; 0.6-0.8 strong correlation; 0.8-1.0 are strongly correlated;
8. compared with the traditional temperature prediction method, the method takes the target variable data as an analysis object, performs temperature prediction through data fitting, considers the influence of multiple factors on the target variable, rewrites the data structure into three-dimensional data and constructs a new data set in a mutual lap joint mode; determining time steps, taking the time steps as data lap joint lengths, and constructing a three-dimensional data set according to the total length of variable data and the number of input variables in a format of (total data length, time steps and variable number);
9. taking 70% of the reconstructed data set as a training set and 30% of the data set as a test set, randomly disordering the data set by taking a time step as a unit, and improving the depth of the model during training;
10. establishing an LSTM model, determining the number of input layer nerve units according to the number of input variables, establishing an LSTM hidden layer, adjusting the number of the input layer nerve units according to actual needs, and only providing one output unit for an output layer; the specific principle of the LSTM is as follows:
the LSTM is a special recurrent neural network, and the repeating unit is called memory block (memory block), and mainly includes three gates (form gate, input gate, output gate) and a memory unit (cell), and controls the flow amount of information by adding several control-number-stage gates, and transmits the control information to the next time. The specific principle is as follows:
it＝sigmoid(Wzixt+Whiht-1+bi)
ft＝sigmoid(Wzfxt+Whfht-1+bf)
ot＝sigmoid(Wxoxt+Whoht-1+bo)
ct＝ft⊙ct-1+it⊙tanh(Wxcxt+Whcht-1+bc)
ht＝ot⊙tanh(ct)
w and b are respectively a weight matrix and an offset item;
input gate it: controlling how much information can flow into a memory cell (c)t)；
Forget to forgetDoor ft: controlling how much information in the memory cell at the previous moment can be accumulated and matched with the memory cell at the current moment;
output gate ot: controlling how much information in the memory cell at the current moment can flow into the current hidden state htPerforming the following steps;
hidden state ht: from ctAnd (4) calculating. Because c istSelf-renews in a linear manner, so that it is added first to tanh (c) having a non-linear functiont) Then by the output gate otTo obtain the current hidden state ht；
11. Optimizing the model, importing a training set to train the model, and continuously optimizing and adjusting the weight and the threshold value in the model;
12. and model evaluation, namely importing a test set into the LSTM model, judging the precision of a model prediction result according to the MAE value (average absolute error) output by the model, and repeatedly adjusting and training to obtain the MAE value of the optimal LSTM prediction model. The MAE value of the invention is the average value of the absolute value of the error between the predicted value and the true value:
Figure BDA0003524231000000091
in the formula: y isiA prediction result x at a certain moment output by the modeliIs the real result at a certain moment;
13. and importing the parameter variable data into an optimal LSTM model to predict the temperature of the target variable.
Example 2
The embodiment of the utility model also provides a bulky concrete temperature monitoring data prediction system, include:
the data acquisition module is used for arranging sensors according to monitoring point positions, setting sampling time intervals, acquiring ambient temperature and humidity data in a long period of time and temperature monitoring data of different positions of the large-volume concrete, and combining the data to form an initial data set;
the data preprocessing module is used for carrying out abnormal value inspection processing on the initial data set acquired in the step one, carrying out Savitzky-Golay filtering smoothing noise reduction processing after the abnormal value processing is finished, and carrying out dimensionless and normalization processing on the data by adopting a Z-score standardization method after the data noise reduction is finished to obtain target variable data and input variable data which accord with standard normal distribution;
the relevance analysis module is used for analyzing the target variable and the input variable which are subjected to preprocessing in the second step and conform to the standard normal distribution by using a Pearson correlation coefficient method, judging the relevance degree of the environmental temperature, the humidity data and the temperature data of other measuring points and the temperature change of the target measuring points according to the calculation result, and further screening important variables;
the data set reconstruction module constructs a three-dimensional data set by using the important variables screened in the step three, wherein the format is total data length, time step and variable number, the data set is randomly disordered by taking the time step as a unit, 70% of the reconstructed data set is used as a training set, and the rest data set is used as a test set;
the LSTM model building and optimizing module is used for building an LSTM network model, importing a training set for model training, continuously optimizing and adjusting the weight and the threshold value in the model, importing a test set into the optimized model, judging the precision of a model prediction result according to the MAE value output by the model, repeatedly adjusting and training to obtain the MAE value of the optimal LSTM prediction model, and further obtaining a final LSTM prediction model;
and the temperature prediction module is used for importing the required input variable data into the optimal LSTM model to predict the temperature of the target variable.
Example 3
The present embodiment also provides a computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the prediction method for large-volume concrete temperature monitoring data.
Wherein the computer readable storage medium stores an interface display program executable by at least one processor to cause the at least one processor to perform the steps of the method for predicting bulk concrete temperature monitoring data as described above.
Application example
Based on-site large-volume concrete test block (1 x 1 m)3) Monitoring data, specifically applying the mass concrete test block by adopting a python programming language according to the flow, respectively arranging temperature sensors on the upper part, the middle part, the lower part, the side and the corner of the mass concrete test block, and simultaneously collecting temperature and humidity data of the environment around the test block.
The method comprises the following steps of firstly, data preprocessing: as shown in fig. 2, the data sampling time is set at a fixed frequency, the sampling frequency of this example being once per minute; removing abnormal values of the data, and carrying out noise reduction and smoothing treatment on the data by adopting an SG filtering method;
step two: data normalization: normalizing the data by adopting a Z-score standardization method, and performing dimensionless operation on the data;
step three: and (3) reconstructing data: in the embodiment, data in 600 minutes are selected as an initial data set, and the data sets are constructed by various data according to time length, time step and variable number, so that the data structure of the input variable is (571, 30, 6), and the data structure of the target variable is (571, 30, 1);
step four: dividing a training set and a testing set: as shown in fig. 3, the data sets reconstructed in step three are randomly scrambled, and 70% of the data sets are taken as training sets and 30% are taken as test sets;
step five: building an LSTM network: as shown in fig. 4, the example builds a three-layer neural network, the number of input layer neural units is 6, the number of LSTM hidden layer neural units is 32, the number of output layer neural units is 1, and the variable parameters of the network are 5025 in total;
step six: optimizing and training a model: in the example, the iteration number is set to be 500, a training set is input into the network model, 20% of the training set is set as a verification set at the same time, loss value monitoring of the verification set is added, the iteration process is automatically stopped when the loss value reaches a set threshold value, model parameters are stopped from being updated, and a final LSTM model is output;
step seven: model prediction: as shown in fig. 5, the test set is input to the trained model to output the prediction result, and since the output prediction result is normalized data, the prediction result needs to be subjected to inverse normalization processing, and then compared with the true value, and the prediction performance of the model is judged according to the magnitude of the MAE value, where the MAE of the model is 0.0141, which shows that the prediction performance of the batch of data is good. In order to facilitate the display of the difference between the predicted value and the true value, a scatter diagram is adopted to visualize the result;
from the above examples, it can be seen that the LSTM model has higher accuracy and reliability in predicting temperature data with time-series characteristics, and can be used as an effective means for predicting large-volume concrete temperature monitoring data.

Claims (10)

1. A method for predicting large-volume concrete temperature monitoring data is characterized by comprising the following steps:
step one, data acquisition: arranging sensors according to the monitoring point locations, acquiring environmental temperature data, environmental humidity data and temperature monitoring data of the monitoring points around the monitoring point locations, and merging the data to form an initial data set;
step two, data preprocessing: carrying out abnormal value inspection processing on the initial data set acquired in the step one, carrying out smooth noise reduction processing by adopting a Savitzky-Golay filtering method after the abnormal value inspection processing is finished, and carrying out de-dimension and normalization processing on the data by adopting a Z-score standardization method after the data noise reduction is finished to obtain target variable data and input variable data which accord with standard normal distribution;
step three, correlation analysis: analyzing the target variable and the input variable which are in accordance with the standard normal distribution after being preprocessed in the step two by adopting a Pearson correlation coefficient method, judging the correlation degree of the environmental temperature, the humidity data and the temperature data of other measuring points and the temperature change of the target measuring point according to the calculation result, and further screening important variables;
fourthly, reconstructing the data set: constructing a three-dimensional data set by using the important variables screened in the step three, wherein the format is total data length, time step and variable number, randomly disordering the data set by taking the time step as a unit, taking 70% of the reconstructed data set as a training set, and taking the rest data as a test set;
step five, building and optimizing an LSTM model: constructing an LSTM network model, respectively training and testing the LSTM network model by utilizing a training set and a testing set, and continuously optimizing the LSTM model according to an output MAE value;
step six, using the historical data obtained after the processing in the step two for model training and testing to obtain an optimized LSTM model, then inputting the data collected in real time into the optimized model, and outputting the prediction result of the target monitoring point.
2. The method for predicting the temperature monitoring data of the mass concrete according to claim 1, wherein in the step one, the sensors are arranged on the appointed measuring points, and the spacing distance between the sensors on the same horizontal plane is controlled within the range of 5 m; the sensors on the same axial direction are embedded along the axial direction according to different depths, and the sampling frequency of data acquisition is 1-3 min/time.
3. The method as claimed in claim 1, wherein the abnormal value checking process of step two is performed when there is a short time data abnormality, and the abnormal data process can be filled by using a data substitution method and a weighted average method.
4. The method for predicting the mass concrete temperature monitoring data according to claim 1, wherein the Savitzky-Golay filtering smoothing noise reduction processing in the second step is to use a least square method to regress a small window of the data onto a polynomial, and then use the polynomial to estimate a point at the center of the window, so that the smoothness of a data curve can be obviously improved, and the interference of data noise can be reduced; Savitzky-Golay smoothing equation:
Figure FDA0003524230990000021
in the formula: x is the number ofk，smoothThe kth data point after the smoothing processing is carried out;
Figure FDA0003524230990000024
is the kth mean data point; h is a coefficient value and H ═ 2 w; w is equal to half the amount of data within the window length; x is the number ofk+iIs the kth original data point; h isiIs a smoothing coefficient used for reducing the influence of smoothing on useful information and improving the disadvantages of a smoothing and denoising algorithm,
Figure FDA0003524230990000022
polynomial fitting solution can be performed based on the principle of two multiplications.
5. The method for predicting the temperature monitoring data of the mass concrete according to claim 1, wherein the Z-score normalization method in the second step is subjected to de-dimension and normalization processing, and the operation formula is as follows:
z-score normalization:
Figure FDA0003524230990000023
in the formula: μ is the mean of all sample data; σ is the standard deviation of all sample data; x is a radical of a fluorine atom*The data values are normalized; x is the original data value.
6. The method for predicting the temperature monitoring data of the mass concrete according to claim 1, wherein the target variable data in the second step is temperature data of a measuring point needing temperature prediction, and the input variable data are collected ambient temperature data, ambient humidity data and temperature data of other measuring points.
7. The method for predicting the temperature monitoring data of the mass concrete according to claim 1, wherein the Pearson correlation coefficient (r) in step threex，y) For investigating the degree of correlation between two variables, the particular formula being e.g.Shown below:
Figure FDA0003524230990000031
in the formula: xiThe value of any time of the input variable is obtained;
Figure FDA0003524230990000032
is the mean of the input variables; y isiThe value of the target variable at any time is obtained;
Figure FDA0003524230990000033
is the mean of the target variable; r isx，yIs the Pearson correlation coefficient between two variables; wherein, Pearson's correlation coefficient (r)x，y) The degree of correlation of the variables can be judged in the following ranges: 0.0-0.2 are very weakly or not correlated; 0.2-0.4 weakly correlated; 0.4-0.6 moderate correlation; 0.6-0.8 strong correlation; 0.8-1.0 are strongly correlated.
8. The method for predicting the temperature monitoring data of the mass concrete according to claim 1, wherein the MAE value in the fifth step is an average value of absolute values of errors between a predicted value and a true value, and a specific formula is as follows:
Figure FDA0003524230990000034
in the formula: y isiA prediction result at a certain moment is output for the model; x is a radical of a fluorine atomiIs the real result at a certain moment.
9. A system for predicting bulk concrete temperature monitoring data, comprising:
the data acquisition module is used for arranging sensors according to monitoring point positions, setting sampling time intervals, acquiring ambient temperature and humidity data in a long period of time and temperature monitoring data of different positions of the large-volume concrete, and combining the data to form an initial data set;
the data preprocessing module is used for carrying out abnormal value inspection processing on the initial data set acquired in the step one, carrying out Savitzky-Golay filtering smoothing noise reduction processing after the abnormal value processing is finished, and carrying out dimensionless and normalization processing on the data by adopting a Z-score standardization method after the data noise reduction is finished to obtain target variable data and input variable data which accord with standard normal distribution;
the relevance analysis module is used for analyzing the target variable and the input variable which are subjected to preprocessing in the second step and conform to the standard normal distribution by using a Pearson correlation coefficient method, judging the relevance degree of the environmental temperature, the humidity data and the temperature data of other measuring points and the temperature change of the target measuring points according to the calculation result, and further screening important variables;
the data set reconstruction module constructs a three-dimensional data set by using the important variables screened in the step three, wherein the format is total data length, time step and variable number, the data set is randomly disordered by taking the time step as a unit, 70% of the reconstructed data set is used as a training set, and the rest data set is used as a test set;
the LSTM model building and optimizing module is used for building an LSTM network model, importing a training set for model training, continuously optimizing and adjusting the weight and the threshold value in the model, importing a test set into the optimized model, judging the precision of a model prediction result according to the MAE value output by the model, repeatedly adjusting and training to obtain the MAE value of the optimal LSTM prediction model, and further obtaining a final LSTM prediction model;
and the temperature prediction module is used for importing the required input variable data into the optimal LSTM model to predict the temperature of the target variable.
10. A computer-readable storage medium characterized by: the readable storage medium has stored thereon a computer program which, when executed by a processor, carries out the steps of the method of predicting bulk concrete temperature monitoring data according to any one of claims 1 to 8.
CN202210190145.7A 2022-02-28 2022-02-28 Method, system and medium for predicting large-volume concrete temperature monitoring data Pending CN114742260A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210190145.7A CN114742260A (en) 2022-02-28 2022-02-28 Method, system and medium for predicting large-volume concrete temperature monitoring data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210190145.7A CN114742260A (en) 2022-02-28 2022-02-28 Method, system and medium for predicting large-volume concrete temperature monitoring data

Publications (1)

Publication Number Publication Date
CN114742260A true CN114742260A (en) 2022-07-12

Family

ID=82274535

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210190145.7A Pending CN114742260A (en) 2022-02-28 2022-02-28 Method, system and medium for predicting large-volume concrete temperature monitoring data

Country Status (1)

Country Link
CN (1) CN114742260A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116308217A (en) * 2023-05-19 2023-06-23 中交第四航务工程勘察设计院有限公司 Concrete monitoring platform management method and system based on Internet of things
CN117130415A (en) * 2023-10-27 2023-11-28 四川信特农牧科技有限公司 Warehouse management method and system
CN117236511A (en) * 2023-09-26 2023-12-15 中交广州航道局有限公司 Big data prediction method and device for vacuum degree of underwater pump of cutter suction dredger
CN117368470A (en) * 2023-10-09 2024-01-09 南通如日纺织有限公司 Textile antibacterial detection and quality assessment system

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116308217A (en) * 2023-05-19 2023-06-23 中交第四航务工程勘察设计院有限公司 Concrete monitoring platform management method and system based on Internet of things
CN117236511A (en) * 2023-09-26 2023-12-15 中交广州航道局有限公司 Big data prediction method and device for vacuum degree of underwater pump of cutter suction dredger
CN117368470A (en) * 2023-10-09 2024-01-09 南通如日纺织有限公司 Textile antibacterial detection and quality assessment system
CN117130415A (en) * 2023-10-27 2023-11-28 四川信特农牧科技有限公司 Warehouse management method and system
CN117130415B (en) * 2023-10-27 2024-01-02 四川信特农牧科技有限公司 Warehouse management method and system

Similar Documents

Publication Publication Date Title
CN114742260A (en) Method, system and medium for predicting large-volume concrete temperature monitoring data
CN110894038B (en) Method and device for predicting running state of elevator door system
CN109492808B (en) Method for predicting remaining parking spaces of indoor parking lot
CN104182642B (en) A kind of fault detection method based on rarefaction representation
CN106529145A (en) ARIMA-BP neutral network-based bridge monitoring data prediction method
CN111030889B (en) Network traffic prediction method based on GRU model
Lin et al. Kriging based sequence interpolation and probability distribution correction for gaussian wind field data reconstruction
CN111275255A (en) Construction method of concrete dam deformation monitoring and forecasting model
CN109034225B (en) Uncertainty parameter estimation method combining random variable grayness and Bayesian model correction
CN111832790A (en) Method and system for predicting medium and long-term water demand of water supply pipe network
CN114358435A (en) Pollution source-water quality prediction model weight influence calculation method of two-stage space-time attention mechanism
CN103646013B (en) Multiple fault reconstruction method based on covariance matrix norm approximation
CN115169409B (en) Sliding window-based bridge structure self-vibration frequency identification and early warning method and equipment
CN112100574A (en) Resampling-based AAKR model uncertainty calculation method and system
CN114970688A (en) Landslide monitoring data preprocessing method based on LSTMAD algorithm and Hermite interpolation method
CN110532620B (en) Fatigue crack propagation prediction method based on recursive least square-kernel smoothing particle filtering
CN116088453A (en) Production quality prediction model training method and device and production quality monitoring method
CN115719294A (en) Indoor pedestrian flow evacuation control method and system, electronic device and medium
Zhu et al. A coupled model for dam foundation seepage behavior monitoring and forecasting based on variational mode decomposition and improved temporal convolutional network
CN112132324A (en) Ultrasonic water meter data restoration method based on deep learning model
CN114757096B (en) Bridge temperature prediction method, device, equipment and medium based on NARX neural network
Gigante et al. Calendar year effect modeling for claims reserving in HGLM
CN115687322A (en) Water quality time series missing data completion method based on encoder-decoder and autoregressive generated countermeasure network
Parzen Change PP plot and continous sample quantile function
CN115389743A (en) Method, medium and system for predicting content interval of dissolved gas in transformer oil

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