Method for judging stability of tower footing in permafrost region based on IABC-RF model

Abstract

The invention discloses a permafrost region tower footing stability discrimination method based on an IABC-RF model, and the method achieves the fitting of factors affecting the stability of a tower footing in a permafrost region through the use of an improved IABC algorithm and an RF model, and obtains the IABC-RF model. Therefore, on-site workers can effectively predict the horizontal displacement and vertical settlement of the tower footing by detecting the stress condition of the tower footing, the environment temperature and the ice content of frozen soil on site, and finally the stability of the tower footing is judged. According to the method, the search efficiency of the IABC algorithm is fully utilized to optimize the hyper-parameters, so that the fitting effect of the random forest algorithm prediction model is relatively good, prediction and judgment of the stability of the tower footing can be relatively accurately realized, a reliable basis is provided for construction and maintenance of the tower footing in the permafrost region, and the maintenance time and labor cost of the tower footing are reduced.

Description

Technical Field

[0001] The invention relates to a method for detecting the stability of a transmission line tower foundation in geotechnical engineering, and in particular to a method for distinguishing the stability of a tower foundation in a permafrost region based on an artificial bee colony (IABC)-random forest (RF) model. Background Art

[0002] The permafrost area in my country accounts for about 22.4% of the country's land area. Permafrost has the disadvantages of poor thermal stability and is very sensitive to external temperature changes. The permafrost in the permafrost area is divided into active layer permafrost and permafrost. Under the influence of external temperature, the water in the active layer permafrost melts, while the water in the permafrost does not melt. Therefore, the thawing and frost heaving of permafrost caused by temperature changes, which in turn causes uneven settlement and even cracking of the foundation, are the main engineering diseases faced by permafrost areas.

[0003] In order to speed up the construction of UHV backbone power grid, the number of planned areas and industrial areas along the lines is gradually increasing, and the optional path corridors are becoming fewer and fewer. UHV and other transmission line projects will inevitably pass through permafrost areas. The frost heave problem in permafrost areas can easily cause the foundation of high-voltage transmission and transformation line towers to become unstable, and the stability of the tower foundation has a great impact on the safety of the entire regional power grid. Summary of the invention

[0004] In order to solve the shortcomings of the background technology, the present invention provides a method for judging the stability of tower foundation in permafrost areas based on the IABC-RF model. The method can perform intelligent perception and judgment on the stability of tower foundation, provide a reliable basis for the construction and maintenance of tower foundation in permafrost areas, and reduce the time and labor cost of tower foundation maintenance.

[0005] The objective of the present invention is achieved through the following technical solutions:

[0006] A method for determining the stability of a tower foundation in permafrost areas based on the IABC-RF model comprises the following steps:

[0007] Step 1: Based on the existing cases of tower foundation instability in permafrost areas, collect parameters that may affect the stability of the tower foundation: foundation burial depth, backfill soil conditions, tower foundation stress conditions, surrounding environment, ambient temperature, frozen soil ice content, foundation type as independent variables, and horizontal displacement and vertical settlement during the tower foundation instability process as dependent variables;

[0008] Step 2: Create an RF model, extract the data collected in step 1 to generate corresponding training sets and test sets, and normalize the data before inputting it into the RF model;

[0009] Step 3: Use the improved IABC algorithm to optimize the hyperparameters in the RF model: the number of decision trees, the maximum number of features, the minimum number of samples required for leaf nodes, the minimum number of samples required to split internal nodes, the maximum depth of the tree, and the maximum number of leaf nodes;

[0010] Step 4: Substitute the hyperparameters optimized by the IABC algorithm into different RF models to obtain the IABC-RF prediction model of tower foundation stability. Compare the test set data with the predicted values ​​of the IABC-RF prediction model, and use the root mean square error (RMSE) and R 2 Evaluate the fitting effect of the IABC-RF prediction model and finally obtain a more accurate IABC-RF model;

[0011] Step 5: Bring the foundation burial depth, backfill soil conditions, tower foundation stress conditions, surrounding environment, ambient temperature, frozen soil ice content, and foundation type into the IABC-RF model obtained in step 4 to obtain the predicted horizontal displacement and vertical settlement, and determine whether the deformation reaches the amount that causes tower foundation instability.

[0012] Compared with the prior art, the present invention has the following advantages:

[0013] The present invention achieves the fitting of factors affecting the stability of the tower foundation in permafrost areas by using the improved IABC algorithm and RF model, and obtains the IABC-RF model. The on-site staff can effectively predict the horizontal displacement and vertical settlement of the tower foundation by on-site detection of the changes in the stress conditions of the tower foundation, the ambient temperature, and the ice content of the frozen soil, and finally realize the discrimination of the stability of the tower foundation. The present invention makes full use of the search efficiency of the IABC algorithm to optimize the hyperparameters, so that the fitting effect of the random forest algorithm prediction model is better, and the prediction and discrimination of the stability of the tower foundation can be realized more accurately, which provides a reliable basis for the construction and maintenance of the tower foundation in permafrost areas, and reduces the time and labor costs of tower foundation maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is the overall implementation flow chart of the tower foundation stability determination method in permafrost areas based on the IABC-RF model. DETAILED DESCRIPTION

[0015] The technical solution of the present invention is further described below in conjunction with the accompanying drawings, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.

[0016] The present invention provides a method for determining the stability of tower foundations in permafrost areas based on the IABC-RF model. The method targets existing cases of tower foundation instability in permafrost areas, collects parameters that may affect the stability of the tower foundation, introduces them into the RF model, and efficiently optimizes the hyperparameters in the RF model through the IABC algorithm, thereby improving the performance of the RF model and obtaining a high-performance IABC-RF model. Finally, the horizontal displacement and vertical settlement of the tower foundation are predicted by taking into account the tower foundation stress conditions, ambient temperature, and ice content in the frozen soil that change with the construction process and operation time, thereby determining the stability of the tower foundation in permafrost areas. Figure 1 As shown, the following steps are included:

[0017] Step 1: Frost heaving in permafrost areas can cause the tower foundation to pull up, tilt, and overturn, thus making the tower foundation unstable. Based on existing cases of tower foundation instability in permafrost areas, collect parameters that may affect the stability of the tower foundation: foundation burial depth, backfill soil conditions, tower foundation stress conditions, surrounding environment, ambient temperature, frozen soil ice content, foundation type as independent variables, and horizontal displacement and vertical settlement during the tower foundation instability process as dependent variables.

[0018] Step 2: Create an RF model, extract the data collected in step 1 to generate the corresponding training set and test set, and normalize the data before inputting it into the RF model. The specific steps are as follows:

[0019] Step 21: Bootstrap sampling is performed on the original sample set to obtain K samples as the training set, and the samples are modeled into a decision tree, i.e., a weak classifier;

[0020] Step 22: Take the decision results generated by each weak classifier and select the corresponding variables to model them into a decision tree, so as to obtain a random forest with K decision trees;

[0021] Step 2 and 3: Use the obtained random forest to predict the test set, and finally use the voting mechanism to find the decision tree with the highest score and output it as the final result. The RF model can be expressed as:

[0022]

[0023] In the formula, S i is the i-th sample set; C k is the kth decision tree; η(C k ,S i ) is C k For S i The decision value of is a function of the voting mechanism.

[0024] Step 3: Use the improved IABC algorithm to optimize the hyperparameters in the RF model: the number of decision trees, the maximum number of features, the minimum number of samples required for leaf nodes, the minimum number of samples required to split internal nodes, the maximum depth of the tree, and the maximum number of leaf nodes. The specific steps are as follows:

[0025] Step 31: Identify the hyperparameters that need to be optimized. When the number of trees reaches a certain size, it almost no longer affects the accuracy of the model. Therefore, setting the number of decision trees to 50 can meet the requirements of prediction accuracy and computational efficiency. In addition, the maximum depth of the tree and the maximum number of leaf nodes both play a role in limiting the size of the tree and preventing overfitting. To simplify the analysis, only the maximum tree depth is optimized, and the maximum number of leaf nodes uses the default value. In this case, the hyperparameters that need to be optimized are the maximum number of features, the minimum number of samples required to become a leaf node, the minimum number of samples required to split an internal node, and the maximum depth of the tree.

[0026] Step 32: Submit the hyperparameters to be optimized into the IABC algorithm to obtain the optimal parameters, including the following steps:

[0027] (1) Determine the values ​​of each parameter and the number of bees, and calculate the initial resource fitness value:

[0028] u ij =u minj +rand(0,1)(u maxj -u minj )

[0029] In the formula, u ij is the value of the jth dimension of the i-th solution; u maxj Indicates u ij The upper bound of u minj Indicates u ij The lower bound of ; rand is a random number in the interval;

[0030] (2) Employed bee stage: Produce a food source: u ij =u ij +θ ij (u ij -u kj ), and set the solution discarded during the iteration to a i , the set of all discarded solutions is A(A={a 1 ,a 2 ,··,an}), at the end of each iteration, when the old solution a i When it is discarded, according to x ij =y minj +rand(0,1)(y maxj -y minj ) generates a new solution xi(xi 1 ,xi2 ,…,xi D ), where: θ ij is any number in the interval; maxj ,y minj are the search boundaries of the j-th dimension element of each solution; D is the dimension of the solution;

[0031] (3) Follow the bee stage: In this stage, a food source is selected and the fitness value is calculated to determine the probability P of the food source being selected. i :

[0032]

[0033] In the formula, f i t i is the fitness value, ns is the number of bees in the IABC algorithm, and the employed bee stage and the follower bee stage will be repeated until the optimal parameters are obtained.

[0034] Step 4: Substitute the hyperparameters optimized by the IABC algorithm into different random forest models to obtain the IABC-RF prediction model of the tower foundation stability. Compare the test set data with the predicted values ​​of the IABC-RF prediction model and use RMSE and R 2 Evaluate the fitting effect of the IABC-RF prediction model and finally obtain a more accurate IABC-RF model. The specific steps are as follows:

[0035] Step 41: Determine the importance of features in random forest

[0036] When constructing each tree of the random forest, Bootstrap sampling is performed on the data sample set K. The selected samples form the training data sample set, and the samples that are not selected are called the out-of-bag data sample set. Each tree is tested and verified using the out-of-bag data sample set. The error between the tree model output result and the actual result is called the out-of-bag error, denoted as Error oob The importance of features is evaluated by calculating the accuracy of out-of-bag samples. Loop through each tree and assume that the out-of-bag error of the kth tree is Error oobb,k Randomly add random noise to the nth feature of each sample in the out-of-bag data sample set of the current tree, and obtain the out-of-bag error Error again ooba,c , then the out-of-bag error of the kth tree is reduced by Error k for:

[0037] Error k =|Error ooba,k -Error oobb,k

[0038] Therefore, the average out-of-bag error of K trees in the forest is reduced by Erroravg for:

[0039]

[0040] After adding random noise, the greater the reduction in the average out-of-bag error, the more important the feature is.

[0041] Step 42: Use RMSE and R 2 The fitting effect of the IABC-RF prediction model was evaluated, and finally a more accurate IABC-RF model was obtained.

[0042] RMSE is an indicator for calculating the deviation between two data sets. The calculation formula is as follows:

[0043]

[0044] R 2 is a measure of the deviation between the predicted and monitored values, reflecting the degree of fit, and it is calculated as:

[0045]

[0046] Where y is the measured value, y′ is the predicted value, and N is the total amount of data.

[0047] Step 5: According to the physical project in the permafrost area, the foundation type, backfill soil condition, foundation burial depth, and surrounding environment are determined values. The tower foundation stress conditions, ambient temperature, and frozen soil ice content will change with the construction process and operation time. The independent variables such as foundation burial depth, backfill soil condition, tower foundation stress conditions, surrounding environment, ambient temperature, frozen soil ice content, and foundation type are brought into the IABC-RF prediction model finally obtained in step 4 to obtain the predicted horizontal displacement and vertical settlement, and determine whether it has reached the deformation amount that causes the tower foundation to become unstable.

Claims

1. A method for determining the stability of tower foundation in permafrost areas based on the IABC-RF model, characterized in that The method comprises the following steps: Step 1: Based on the existing cases of tower foundation instability in permafrost areas, collect parameters that may affect the stability of the tower foundation: foundation burial depth, backfill soil conditions, tower foundation stress conditions, surrounding environment, ambient temperature, frozen soil ice content, foundation type as independent variables, and horizontal displacement and vertical settlement during the tower foundation instability process as dependent variables; Step 2: Create an RF model, extract the data collected in step 1 to generate corresponding training sets and test sets, and normalize the data before inputting it into the RF model; Step 3: Use the improved IABC algorithm to optimize the hyperparameters in the RF model: the number of decision trees, the maximum number of features, the minimum number of samples required for leaf nodes, the minimum number of samples required to split internal nodes, the maximum depth of the tree, and the maximum number of leaf nodes; Step 4: Substitute the hyperparameters optimized by the IABC algorithm into different RF models to obtain the IABC-RF prediction model of the tower foundation stability. Compare the test set data with the predicted values ​​of the IABC-RF prediction model, and use the root mean square error and R 2 Evaluate the fitting effect of the IABC-RF prediction model and finally obtain the IABC-RF model; Step 5: Bring the foundation burial depth, backfill soil conditions, tower foundation stress conditions, surrounding environment, ambient temperature, frozen soil ice content, and foundation type into the IABC-RF model obtained in step 4 to obtain the predicted horizontal displacement and vertical settlement, and determine whether the deformation reaches the amount that causes tower foundation instability.

2. The method for determining tower foundation stability in permafrost areas based on the IABC-RF model according to claim 1 is characterized in that The specific steps of step 2 are as follows: Step 21: Bootstrap sampling is performed on the original sample set to obtain K samples as the training set, and the samples are modeled into a decision tree, i.e., a weak classifier; Step 22: Take the decision results generated by each weak classifier and select the corresponding variables to model them into a decision tree, so as to obtain a random forest with K decision trees; Step 2 and 3: Use the obtained random forest to predict the test set, and finally use the voting mechanism to find the decision tree with the highest score and output it as the final result. The RF model is expressed as: In the formula, S i is the i-th sample set; C k is the kth decision tree; η(C k ,S i ) is C k For S i The decision value of is a function of the voting mechanism.

3. The method for determining tower foundation stability in permafrost areas based on the IABC-RF model according to claim 1, characterized in that The specific steps of step three are as follows: Step 31: Identify the hyperparameters that need to be optimized: the maximum number of features, the minimum number of samples required to become a leaf node, the minimum number of samples required to split an internal node, and the maximum depth of the tree; Step 32: Submit the hyperparameters to be optimized into the IABC algorithm to obtain the optimal parameters, including the following steps: (1) Determine the values ​​of each parameter and the number of bees, and calculate the initial resource fitness value: u ij =u minj +rand(0,1)(u maxj -u minj ) In the formula, u ij is the value of the jth dimension of the i-th solution; u maxj Indicates u ij The upper bound of u minj Indicates u ij The lower bound of ; rand is a random number in the interval; (2) Employed bee stage: Produce a food source: u ij =u ij +θ ij (u ij -u kj ), and set the solution discarded during the iteration to a i , the set of all discarded solutions is A (A = {a1, a2, ··, an}). At the end of each iteration, when the old solution a i When it is discarded, according to x ij =y minj +rand(0,1)(y maxj -y minj ) generates new solutions xi(xi1,xi2,…,xi D ), where: θ ij is any number in the interval; maxj ,y minj are the search boundaries of the j-th dimension element of each solution; D is the dimension of the solution; (3) Follow the bee stage: In this stage, a food source is selected and the fitness value is calculated to determine the probability P of the food source being selected. i : In the formula, f i t i is the fitness value, ns is the number of bees in the IABC algorithm, and the employed bee stage and the follower bee stage will be repeated until the optimal parameters are obtained.

4. The method for determining tower foundation stability in permafrost areas based on the IABC-RF model according to claim 1, characterized in that The specific steps of step 4 are as follows: Step 41: Determine the importance of features in random forest When constructing each tree of the random forest, Bootstrap sampling is performed on the data sample set K. The selected samples form the training data sample set, and the samples that are not selected are called the out-of-bag data sample set. Each tree is tested and verified using the out-of-bag data sample set. The error between the tree model output result and the actual result is called the out-of-bag error, denoted as Error oob , evaluate the importance of features by calculating the accuracy of out-of-bag samples, loop through each tree, and assume that the out-of-bag error of the kth (k∈K) tree is Error oobb,k , randomly add random noise to each sample nth feature in the out-of-bag data sample set of the current tree, and obtain the out-of-bag error Error again ooba,c , then the out-of-bag error of the kth tree is reduced by Error k for: Error k =|Error ooba,k -Error oobb,k The average out-of-bag error of K trees in the forest is reduced by Error avg for: After adding random noise, if the average out-of-bag error decreases more, it means that the feature is more important; Step 42: Use RMSE and R 2 The fitting effect of the IABC-RF prediction model was evaluated, and finally the IABC-RF model was obtained.

Images