Online prediction method for temperature of concrete in arch dam pouring bin

Abstract

The invention belongs to the field of civil engineering, and relates to an arch dam pouring bin concrete temperature online prediction method, which comprises the following steps: 1) collecting arch dam pouring bin concrete temperature data; (2) the arch dam pouring bin concrete temperature data are preprocessed; (3) data resampling is conducted on the preprocessed arch dam pouring bin concrete temperature data, and a sample data set after dimension reduction is obtained; and 4) establishing an SSA-ISVR concrete temperature online prediction model according to the sample data set after dimension reduction, and predicting the concrete temperature of the arch dam pouring bin according to the SSA-ISVR concrete temperature online prediction model. According to the online prediction model for the temperature state of the arch dam after concrete pouring, boundary conditions and the time-varying characteristics of concrete properties are considered in the model, and compared with an offline model, the online prediction model has higher prediction precision and can provide support for concrete temperature control.

Description

Technical Field

[0001] The invention belongs to the field of civil engineering and relates to a concrete temperature prediction method, in particular to an online prediction method for concrete temperature in an arch dam casting bin. Background Art

[0002] Arch dam concrete temperature control has an important impact on dam construction and safety, which requires the ability to timely and accurately perceive and predict the concrete temperature state and dynamically adjust temperature control measures. The concrete temperature predicted based on arch dam temperature simulation has a relatively accurate performance in its long-term trend prediction, but in the early stage of construction, affected by the heat of cement hydration and environmental and construction conditions, its predicted value often shows that it is difficult to fit the monitoring temperature curve.

[0003] In this regard, an online prediction model is proposed, which can quickly and accurately predict the temperature state of concrete after pouring. The model can be updated in time according to the specific boundary conditions and the time-varying characteristics of the concrete's own properties. Compared with the offline model, it has higher prediction accuracy and can provide support for concrete temperature control. Summary of the invention

[0004] In order to solve the technical problems existing in the background technology, the present invention provides an online prediction model that can provide concrete temperature state after pouring for concrete temperature control, and uses an online training method of machine learning to perform temperature prediction while considering boundary conditions and the time-varying characteristics of concrete properties itself.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] An online prediction method for the concrete temperature of an arch dam casting bin, characterized in that the online prediction method for the concrete temperature of an arch dam casting bin comprises the following steps:

[0007] 1) Collect concrete temperature data of arch dam pouring bin;

[0008] 2) preprocessing the concrete temperature data of the arch dam pouring bin obtained in step 1);

[0009] 3) Resample the preprocessed arch dam pouring bin concrete temperature data to obtain a reduced-dimensional modeling input sample data set;

[0010] 4) Establish an SSA-ISVR concrete temperature online prediction model based on the reduced dimension modeling input sample data set obtained in step 3), and predict the average temperature of concrete in the arch dam casting bin based on the SSA-ISVR concrete temperature online prediction model.

[0011] Preferably, the concrete temperature data of the arch dam casting bin in step 1) includes concrete age, concrete grade, average temperature of concrete 0.5 days before, cooling water temperature, cooling water flow, interlayer interval time and bin surface heat dissipation area;

[0012] The concrete temperature data of the arch dam casting bin is obtained by measuring the concrete temperature of the arch dam casting bin by a point thermometer and a distributed optical fiber.

[0013] Preferably, the preprocessing method in step 2) is missing value interpolation, abnormal value deletion, outlier interpolation and outlier deletion.

[0014] Preferably, the method for data resampling in step 3) is database list slicing and sorting and reorganizing.

[0015] Preferably, the specific implementation method of data resampling in step 3) is:

[0016] 3.1) extracting the concrete temperature data of the arch dam pouring bin pre-processed in step 2) through the SQL database;

[0017] 3.2) Perform a nested query on the data obtained in step 3.1), resample by data screening and sorting, generate new data samples, and obtain a reduced-dimensional modeling input sample data set.

[0018] Preferably, the specific implementation of step 4) is:

[0019] 4.1) Divide the reduced-dimensional modeling input sample data set obtained in step 3) into a training set and a test set, wherein the ratio of the training set to the test set is 7:3;

[0020] 4.2) Construct the initial model of SSA-ISVR concrete temperature online prediction in python;

[0021] 4.3) Using the training set and the test set obtained in step 4.1) to train the SSA-ISVR concrete temperature online prediction initial model obtained in step 4.2) to obtain the SSA-ISVR concrete temperature online prediction model;

[0022] 4.4) Predict the average temperature of concrete in the arch dam casting bin based on the SSA-ISVR concrete temperature online prediction model.

[0023] Preferably, the specific implementation method of the training in step 4.3) is:

[0024] 4.3.1) Obtain an initial model, the expression of the initial model is:

[0025] f(x)=Wx+b

[0026] in:

[0027] f(x) is the constructed SSA-ISVR initial model for online prediction of concrete temperature;

[0028] x is a number of data attributes in the training set;

[0029] W is the weight coefficient;

[0030] b is the error term;

[0031] The solution of W and b is:

[0032]

[0033] st(W·x i +b)-y i ≤ε+ξ i

[0034]

[0035]

[0036] in:

[0037] ξ is the slack variable;

[0038] Represents the part of the predicted value that is smaller than the true value that exceeds ε from the true value;

[0039] ξ i Represents the part of the predicted value that is greater than the true value that exceeds ε from the true value;

[0040] ε is the error limit;

[0041] C is the penalty term for error;

[0042] N is the dimension of the training data in the training set;

[0043] 4.3.2) Substitute the entire training set obtained in step 4.1) and part of the test set obtained in step 4.1) as input features into the initial model of step 4.3.1), and train the initial model to obtain the SSA-ISVR concrete temperature online prediction model.

[0044] Preferably, the method for online prediction of concrete temperature in an arch dam casting bin further comprises, after step 4):

[0045] 5) Determine whether the SSA-ISVR concrete temperature online prediction model obtained in step 4) needs to be updated. If so, update the SSA-ISVR concrete temperature online prediction model obtained in step 4), and use the updated SSA-ISVR concrete temperature online prediction model to predict the average temperature of concrete in the arch dam casting bin; if not, continue to use the SSA-ISVR concrete temperature online prediction model obtained in step 4) to predict the average temperature of concrete in the arch dam casting bin.

[0046] Preferably, the specific implementation method of the judgment in step 5) is:

[0047] 5.1) inputting the partial test set obtained in step 4.1) into the SSA-ISVR concrete temperature online prediction model obtained in step 4), and outputting the test results through the SSA-ISVR concrete temperature online prediction model obtained in step 4);

[0048] 5.2) Compare the test results obtained in step 5.1) with the test set obtained in step 4.1) one by one to obtain the error between the two;

[0049] 5.3) Determine whether the error obtained in step 5.2) exceeds the set error threshold. If so, repeat step 4) to continue training the SSA-ISVR concrete temperature online prediction model obtained in step 4) to complete the update of the SSA-ISVR concrete temperature online prediction model obtained in step 4); if not, there is no need to update the SSA-ISVR concrete temperature online prediction model obtained in step 4).

[0050] An SSA-ISVR concrete temperature online prediction model for predicting the average temperature of concrete in an arch dam casting bin is constructed based on the above-mentioned arch dam casting bin concrete temperature online prediction method.

[0051] The advantages of the present invention are:

[0052] The present invention provides an arch dam concrete temperature online prediction modeling method, comprising: 1) collecting arch dam concrete temperature data; 2) preprocessing the arch dam concrete temperature data obtained in step 1); 3) resampling the preprocessed arch dam concrete temperature data to obtain a reduced-dimensional modeling input sample data set; 4) establishing an SSA-ISVR concrete temperature online prediction model according to the reduced-dimensional modeling input sample data set obtained in step 3), and predicting the average temperature of concrete in the arch dam casting bin according to the SSA-ISVR concrete temperature online prediction model. In view of the importance of arch dam concrete temperature control, the present invention needs to timely grasp the arch dam concrete temperature state, so as to adjust the temperature control measures in real time, ensure that the concrete temperature meets the design requirements, and reduce the risk of dam body cracking. Since the implementation of temperature simulation requires certain environmental parameters, however, such parameters are often obtained based on the past timeline. Although the periodicity of environmental changes can be used to approximate the current required environmental parameters, so that temperature simulation can meet the accuracy requirements of temperature prediction to a certain extent, such environmental parameters have lost their real physical meaning. They can only describe the trend of concrete temperature changes through offline models, and are limited by their large amount of calculation. The time cost of real-time prediction is often unbearable. Considering the impact of real-time changing boundary conditions and the properties of concrete itself on concrete temperature, online modeling can be carried out through machine learning. With the arrival of new samples, the model can be adaptively adjusted through prediction errors to quickly perform modeling and prediction, so as to achieve fast and accurate prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is the process flow of the online prediction method for the concrete temperature of the arch dam casting bin corresponding to the present invention;

[0054] Figure 2 It is a schematic diagram of the layout of the thermocouple temperature sensor bin surface of the arch dam corresponding to the present invention;

[0055] Figure 3 Schematic diagram of the layout of the distributed optical fiber temperature sensor for arch dams corresponding to the present invention;

[0056] Figure 4 It is a cross-sectional view in the thickness direction of the optical fiber bin surface of the arch dam casting bin corresponding to the present invention;

[0057] Figure 5 is a data processing and analysis flow chart of the present invention;

[0058] Figure 6 is a flow chart of the online modeling algorithm of the present invention;

[0059] Figure 7 It is a schematic diagram of the model update of the present invention. DETAILED DESCRIPTION

[0060] See also Figure 1 The present invention provides an online prediction method for the concrete temperature of an arch dam casting bin, which specifically comprises the following steps:

[0061] 1) Data collection of concrete temperature in the arch dam casting bin; specifically: the temperature in the concrete casting bin is measured mainly through point thermometers and distributed optical fibers, and the average value is used as the concrete temperature measurement value in the bin; the arch dam casting bin temperature data includes concrete age, concrete grade, average temperature of the concrete 0.5 days before, cooling water temperature, cooling water flow, interlayer interval time, and bin surface heat dissipation area. The temperature in the concrete casting bin is measured mainly through point thermometers and distributed optical fibers, and the average value is used as the concrete temperature measurement value in the bin;

[0062] 1.1) Generally, three point thermometers are buried in the concrete casting bin. The buried method of distributed optical fiber is designed according to the shape of the casting bin. Generally, the casting bin is laid out in I-type, L-type, and Z-type optical fiber (in this method, the casting bin optical fiber is laid out in Z-type); Figure 2 The figure shows the layout of the point thermometer in a casting bin. Figure 3 The figure shows the Z-shaped plane layout of optical fiber in a casting bin.

[0063] 1.2) Both temperature sensors are installed in the cooling water pipe interlayer; optical fibers are buried intermittently between the construction layers of the casting warehouse, such as Figure 4 As shown;

[0064] 1.3) Select the density of measuring points according to the size of the casting bin and the temperature measurement requirements. Generally, measuring points are taken at intervals of 1m inside the concrete. The interval between measuring points can be adjusted according to the importance of the casting bin.

[0065] 2) Perform data preprocessing on the data obtained in step 1), the process is as follows: Figure 5 As shown; specifically: filling missing values ​​in data and identifying and processing outliers, constructing multi-feature concrete temperature data; the data cleaning and processing methods in step 2) include missing value processing, outlier identification and processing, data classification, and resampling;

[0066] 2) The data cleaning and processing methods include missing value processing, outlier identification and processing, data classification, and resampling. The specific implementation process is as follows:

[0067] 2.1) First, download the monitoring data from the cloud, import it into the database, filter missing values ​​and outliers through relational database query statements, filter through queries in the database and process missing values ​​through interpolation to obtain preliminary cleaned data; the interpolation method can refer to the following formula:

[0068]

[0069] x is the data element, i is the missing data sequence value

[0070] 2.2) Observe data features through data visualization (mainly using Python to connect to the SQL database and plot the data), interpolate or delete outliers (delete outliers with larger intervals);

[0071] 3) resampling the data preprocessed in step 2); specifically, resampling the data obtained in step 2) to obtain a reduced-dimensional modeling input sample data set; the resampling method in step 3) is database list slicing and sorting and reorganization;

[0072] Step 3) The data can be resampled according to the model requirements to extract data at appropriate concrete age intervals.

[0073] First, the data to be resampled is queried through the database, and then these data are nested for query. New data samples are generated by resampling through data screening and sorting. Finally, this data is used as input and concrete temperature as output to construct the model training set and test set.

[0074] 4) Establish a multi-feature SSA-ISVR initial model for online prediction of concrete temperature based on the data set in step 3); specifically: divide the data set obtained in step 3), input the training set and build an SSA-ISVR (Sparrow Optimized Incremental Support Vector Regression Machine) pre-training model in python; the data set division in step 4) includes a training set and a test set, with a ratio of 7:3, and the SSA-SVR model uses the Sparrow Optimization Algorithm to optimize the penalty factor and kernel function coefficient of support vector regression.

[0075] During the training process, we first use the training set and a part of the test set to train an initial model, which can be summarized as follows:

[0076] f(x)=Wx+b

[0077] Where f(x) is the constructed high-dimensional sample fitting function, x is the input feature, w and b are the weight coefficient and error term, respectively, which can be obtained by solving the quadratic programming problem:

[0078]

[0079] st(W·x i +b)-y i ≤ε+ξ i

[0080]

[0081]

[0082] Where ξ is the slack variable and ε is the error limit.

[0083] 5) Output the prediction results in real time and compare them with the monitoring value feedback to update the model in real time. Specifically, input the test set divided in step 4) into the SSA-SVR model and use the online training method to build a dynamically updated SSA-ISVR model to output the prediction results of the concrete temperature in the casting bin; in step 5), the online training adopts the strategy of inputting and outputting data samples one by one, that is, only one sample is input at a time, and the model is updated online dynamically according to the prediction error. Then input the other part of the test set samples one by one (that is, only one data sample is input at a time), and compare the errors based on the results of the model prediction. If the error exceeds the set error threshold, it means that the initial model has not learned enough information to explain this sample, then train this sample and update the model; otherwise, it means that the teacher model already has enough knowledge to describe this sample, and the model will not be updated. The modeling algorithm flow is as follows Figure 6 As shown, the dynamic update of the model is as follows Figure 7 As shown, in the dynamic update process, the current sample data prediction will have an impact on the model, and the model may learn new knowledge, so that it will perform well next time when facing similar samples. Table 1 shows the training and prediction results of the method provided by the present invention in multiple casting bins of an arch dam. First, the training of the initial model is used as the training result, and the test set is the root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R 2 ) as the prediction result. It can be seen that the determination coefficients of the training and test results are both over 0.9, the RMSE is about 0.2, and the MAE is about 0.1℃.

[0084] Table 1 Training and prediction results of the model constructed in this paper for a certain engineering example

[0085]

[0086] The present invention provides, on the basis of providing the above-described online prediction method for concrete temperature in an arch dam casting bin, a SSA-ISVR online prediction model for concrete temperature constructed based on the method for predicting the average temperature of concrete in an arch dam casting bin. By means of the model, the average temperature of concrete in an arch dam casting bin can be conveniently and quickly predicted in real time.

Claims

1. An online prediction method for concrete temperature in an arch dam pouring bin, characterized in that: The method for online prediction of concrete temperature in an arch dam pouring bin comprises the following steps: 1) Collect concrete temperature data of arch dam pouring bin; 2) preprocessing the concrete temperature data of the arch dam pouring bin obtained in step 1); 3) Resample the preprocessed arch dam pouring bin concrete temperature data to obtain a reduced-dimensional modeling input sample data set; 4) Establish an SSA-ISVR concrete temperature online prediction model based on the reduced dimension modeling input sample data set obtained in step 3), and predict the average temperature of concrete in the arch dam casting bin based on the SSA-ISVR concrete temperature online prediction model.

2. The method for online prediction of concrete temperature in an arch dam pouring bin according to claim 1 is characterized in that: The concrete temperature data of the arch dam pouring bin in step 1) includes the concrete age, concrete grade, average temperature of the concrete 0.5 days before pouring, cooling water temperature, cooling water flow, interlayer interval time and bin surface heat dissipation area; The concrete temperature data of the arch dam casting bin is obtained by measuring the concrete temperature of the arch dam casting bin by a point thermometer and a distributed optical fiber.

3. The method for online prediction of concrete temperature in an arch dam pouring bin according to claim 2 is characterized in that: The preprocessing methods in step 2) are missing value interpolation, abnormal value deletion, outlier interpolation and outlier deletion.

4. The method for online prediction of concrete temperature in an arch dam pouring bin according to claim 3 is characterized in that: The method for data resampling in step 3) is database list slicing and sorting and reorganizing.

5. The method for online prediction of concrete temperature in an arch dam pouring bin according to claim 4 is characterized in that: The specific implementation method of data resampling in step 3) is: 3.1) extracting the concrete temperature data of the arch dam pouring bin pre-processed in step 2) through the SQL database; 3.2) Perform a nested query on the data obtained in step 3.1), resample by data screening and sorting, generate new data samples, and obtain a reduced-dimensional modeling input sample data set.

6. The method for online prediction of concrete temperature in an arch dam pouring bin according to claim 5 is characterized in that: The specific implementation method of step 4) is: 4.1) Divide the reduced-dimensional modeling input sample data set obtained in step 3) into a training set and a test set, wherein the ratio of the training set to the test set is 7:3; 4.2) Construct the initial model of SSA-ISVR concrete temperature online prediction in python; 4.3) Using the training set and the test set obtained in step 4.1) to train the SSA-ISVR concrete temperature online prediction initial model obtained in step 4.2) to obtain the SSA-ISVR concrete temperature online prediction model; 4.4) Predict the average temperature of concrete in the arch dam casting bin based on the SSA-ISVR concrete temperature online prediction model.

7. The method for online prediction of concrete temperature in an arch dam pouring bin according to claim 6 is characterized in that: The specific implementation method of the training in step 4.3) is: 4.3.1) Obtain an initial model, the expression of the initial model is: f(x)=Wx+b in: f(x) is the constructed SSA-ISVR initial model for online prediction of concrete temperature; x is a number of data attributes in the training set; W is the weight coefficient; b is the error term; The solution of W and b is: s.t.(W·x i +b)-y i ≤ε+ξ i x i ≥0, in: ξ is the slack variable; Represents the part of the predicted value that is smaller than the true value that exceeds ε from the true value; ξ i Represents the part of the predicted value that is greater than the true value and the true value that exceeds ε; ε is the error limit; C is the penalty term for error; N is the dimension of the training data in the training set; 4.3.2) Substitute the entire training set obtained in step 4.1) and part of the test set obtained in step 4.1) as input features into the initial model of step 4.3.1), and train the initial model to obtain the SSA-ISVR concrete temperature online prediction model.

8. The method for online prediction of concrete temperature in an arch dam casting bin according to any one of claims 1 to 7, characterized in that: The method for online prediction of concrete temperature in arch dam casting bin further comprises, after step 4): 5) Determine whether the SSA-ISVR concrete temperature online prediction model obtained in step 4) needs to be updated. If so, update the SSA-ISVR concrete temperature online prediction model obtained in step 4), and use the updated SSA-ISVR concrete temperature online prediction model to predict the average temperature of concrete in the arch dam casting bin; if not, continue to use the SSA-ISVR concrete temperature online prediction model obtained in step 4) to predict the average temperature of concrete in the arch dam casting bin.

9. The method for online prediction of concrete temperature in an arch dam pouring bin according to claim 8, characterized in that: The specific implementation method of the judgment in step 5) is: 5.1) inputting the partial test set obtained in step 4.1) into the SSA-ISVR concrete temperature online prediction model obtained in step 4), and outputting the test results through the SSA-ISVR concrete temperature online prediction model obtained in step 4); 5.2) Compare the test results obtained in step 5.1) with the test set obtained in step 4.1) one by one to obtain the error between the two; 5.3) Determine whether the error obtained in step 5.2) exceeds the set error threshold. If so, repeat step 4) to continue training the SSA-ISVR concrete temperature online prediction model obtained in step 4) to complete the update of the SSA-ISVR concrete temperature online prediction model obtained in step 4); if not, there is no need to update the SSA-ISVR concrete temperature online prediction model obtained in step 4).

10. An SSA-ISVR concrete temperature online prediction model for predicting the average temperature of concrete in an arch dam casting bin, constructed based on the arch dam casting bin concrete temperature online prediction method as described in any one of claims 1 to 9.

Images