Shield tunnel settlement prediction method based on artificial intelligence

Abstract

The present invention discloses a shield tunnel settlement prediction method based on artificial intelligence, including data collection, data optimization, construction of a shield tunnel settlement prediction model, model optimization and settlement prediction. The present invention relates to the field of tunnel construction monitoring technology, and specifically refers to a shield tunnel settlement prediction method based on artificial intelligence. This scheme cleans data by defining a dimension-refined distance function and calculating gravity between data, deeply excavates the potential relationship between data, reduces noise interference, and improves the accuracy and reliability of data; by designing a mapping function, a loop control layer, a deep correlation layer and a loss function, the model's recognition and prediction accuracy of complex settlement patterns are improved; by designing intervention factors, defining adaptive step sizes, bidirectional iterative searches and progressively adjusting searches to dynamically search model parameters, the search efficiency is improved, the model training time consumption and loss value are balanced, and the model is made more efficient and stable.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel construction monitoring, and in particular to a shield tunnel settlement prediction method based on artificial intelligence. Background Art

[0002] The shield tunnel settlement prediction method based on artificial intelligence uses artificial intelligence technology and big data analysis technology to predict the settlement phenomenon during the shield tunnel construction process. It has high accuracy and reliability, and can provide a scientific basis for settlement control in shield construction, ensuring construction safety and the stability of surrounding buildings. However, the traditional data optimization method has problems such as poor dimension processing and insufficient data relationship mining when processing shield tunnel settlement prediction data; the traditional shield tunnel settlement prediction model has problems such as low prediction accuracy, poor adaptability to complex settlement patterns, and difficulty in capturing deep correlations in time series data; the traditional model optimization method has problems such as low search efficiency, easy to fall into local optimal solutions, and difficulty in balancing model training time and loss value. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a shield tunnel settlement prediction method based on artificial intelligence. In view of the problems that the traditional data optimization method has poor data optimization effect due to the imprecise dimension processing and insufficient data relationship mining when processing shield tunnel settlement prediction data, this scheme performs data cleaning by defining dimension refinement distance function, calculating gravity between data, cluster center evolution and test clustering, which more accurately reflects the characteristics and differences of data in different dimensions, deeply mines the potential relationship between data, reduces noise interference, and improves the accuracy and reliability of data; in view of the low prediction accuracy, poor adaptability to complex settlement patterns and difficulty in capturing time in the traditional shield tunnel settlement prediction model, the present invention provides a shield tunnel settlement prediction method based on artificial intelligence. In view of the problems that the traditional data optimization method has in processing shield tunnel settlement prediction data, such as imprecise dimension processing and insufficient data relationship mining, this scheme performs data cleaning by defining dimension refinement distance function, calculating gravity between data, cluster center evolution and test clustering, and more accurately reflects the characteristics and differences of data in different dimensions, deeply mines the potential relationship between data, reduces noise interference, and improves the accuracy and reliability of data; in view of the low prediction accuracy, poor adaptability to complex settlement patterns and difficulty in capturing time in the traditional shield tunnel settlement prediction model, the present invention provides a shield tunnel settlement prediction method based on artificial intelligence. In order to solve the problem of deep correlation in time sequence data, this scheme improves the recognition and prediction accuracy of the model for complex settlement patterns by designing mapping functions, loop control layers, deep correlation layers and loss functions, better adapts to various complex patterns that may appear in the settlement process of shield tunnels, and improves the prediction accuracy and practicality of the model; in order to solve the problems of low search efficiency, easy to fall into local optimal solutions, and difficulty in balancing model training time and loss values ​​in traditional model optimization methods, this scheme dynamically searches model parameters by designing intervention factors, defining adaptive step sizes, bidirectional iterative search and progressive adjustment search, which improves search efficiency, avoids falling into local optimal solutions, balances model training time and loss values, and makes the model more efficient and stable.

[0004] The technical solution adopted by the present invention is as follows: a shield tunnel settlement prediction method based on artificial intelligence, the method comprising the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data optimization;

[0007] Step S3: constructing a shield tunnel settlement prediction model;

[0008] Step S4: model optimization;

[0009] Step S5: Settlement prediction.

[0010] Furthermore, in step S1, the data collection is to collect geological condition data, shield machine parameters, environmental data and settlement data of the historical construction section; the geological condition data include stratum type, groundwater level, soil density, compressive strength and internal friction angle; the shield machine parameters include shield machine type, excavation speed, shield machine posture, propulsion force and torque; the environmental data includes time series, ground load, rainfall and temperature; the settlement data refers to the settlement of the shield tunnel.

[0011] Furthermore, in step S2, the data optimization specifically includes the following steps:

[0012] Step S21: define the dimension refinement distance function, which is expressed as follows:

[0013] ;

[0014] Among them, i and j represent the index of the data point, represents the dimension refinement distance function, represents the dimensionally refined distance between data point i and data point j, represents the square root, d represents the dimension index of the data point, and D represents the total number of dimensions of the data point. Indicates taking the absolute value, and Represent the eigenvalues ​​of the dth dimension of the i-th data point and the j-th data point, respectively. Indicates taking the maximum value, represents the maximum eigenvalue of the dth dimension, represents the dimension weight, Represents the indexes of the three data points closest to data point i, represents the eigenvalue of the dth dimension of the kth data point, represents the indexes of the three data points closest to data point j, Indicates The eigenvalue of the d-th dimension of the data point;

[0015] Step S22: Calculate the attraction between data, expressed as follows:

[0016] ;

[0017] in, represents the gravitational function between data, represents the gravitational strength between data point i and data point j, e represents a natural constant, represents the standard deviation of the eigenvalues ​​of the dth dimension, Indicates taking the minimum value, represents the minimum value between the eigenvalue of the dth dimension of the i-th data point and the eigenvalue of the dth dimension of the j-th data point, Represents the maximum value of the eigenvalue of the dth dimension of the i-th data point and the eigenvalue of the dth dimension of the j-th data point, Represents the difference between the maximum and minimum eigenvalues ​​of the dth dimension;

[0018] Step S23: Cluster center evolution, expressed as follows:

[0019] ;

[0020] Where l represents the index of the cluster center, represents the position of the lth cluster center after evolution, It means taking the closest data point. represents the position of the lth cluster center before evolution, represents the number of cluster centers, g represents the index of the data point in the lth cluster, Indicates the Euclidean distance. represents the Euclidean distance between the lth cluster center and the gth data point, represents the Euclidean distance between the g-th data point in the l-th cluster and the farthest data point, represents evolutionary weight;

[0021] Step S24: Test clustering, randomly select from all data points Initial cluster centers are calculated, the gravitational attraction strength between each data point and each cluster center is calculated, and the data points are assigned to the cluster with the largest gravitational attraction strength. The cluster centers evolve, and the assignment and evolution process is repeated until the cluster centers no longer change.

[0022] Step S25: Data cleaning, calculating the average number of data points in all clusters, setting the clusters with less than 50% of the average number of data points as abnormal clusters, and clearing the data points of the abnormal clusters.

[0023] Furthermore, in step S3, the construction of the shield tunnel settlement prediction model specifically includes the following steps:

[0024] Step S31: Establishing a model infrastructure, the shield tunnel settlement prediction model includes a pre-processing layer, a loop control layer, a depth association layer and a prediction layer, and setting the settlement data as the label data of the model;

[0025] Step S32: Design a mapping function, which is expressed as follows:

[0026] ;

[0027] in, represents the input value of the mapping function, represents the mapping function, represents the mapping scaling factor, It means taking logarithm;

[0028] Step S33: Design the pre-processing layer, as shown below:

[0029] ;

[0030] in, represents the output of the pre-processing layer, , and Represent the weight, input and bias of the pre-processing layer respectively;

[0031] Step S34: Design a loop control layer, which consists of a reset unit, an update unit, a state reference unit, and a state output unit, and specifically includes the following steps:

[0032] Step S341: Design a reset unit, as shown below:

[0033] ;

[0034] Among them, t represents the index of the time point, represents the output of the reset unit at time t, represents the Sigmoid function, and Represent the weight and bias of the reset unit respectively, represents the input of the reset unit at time t, represents the state of the loop control layer at time t-1, represents the state of the loop control layer at time t-2, It means taking the square of L2 norm;

[0035] Step S342: Design an update unit, which is expressed as follows:

[0036] ;

[0037] in, represents the output of the update unit at time t, and Represent the weight and bias of the update unit respectively, Represents the input of the update unit at time t;

[0038] Step S343: Design state reference unit, expressed as follows:

[0039] ;

[0040] in, represents the output of the state reference unit at time t, represents the hyperbolic tangent function, and denote the weight and bias of the state reference unit, respectively, represents the element-by-element multiplication symbol, represents the input of the state reference unit;

[0041] Step S344: Design status output unit, expressed as follows:

[0042] ;

[0043] in, Represents the state of the loop control layer at time t;

[0044] Step S35: Design a depth association layer, which is expressed as follows:

[0045] ;

[0046] in, represents the output of the deep correlation layer, , and Represent the weight, input and bias of the deep association layer respectively;

[0047] Step S36: Design the prediction layer, which is expressed as follows:

[0048] ;

[0049] in, represents the output of the prediction layer, , and Represent the weight, input and bias of the prediction layer respectively, represents the normalized exponential function;

[0050] Step S37: Design a loss function, which is expressed as follows:

[0051] ;

[0052] in, Represents the loss value of the model, Represents the total number of model training data, represents the true value of the i-th data, Represents the model's predicted value for the i-th data. represents the loss weight, represents the skewness, Indicates kurtosis.

[0053] Furthermore, in step S4, the model optimization specifically includes the following steps:

[0054] Step S41: define the parameter performance function, which is expressed as follows:

[0055] ;

[0056] in, represents the model parameters, Indicates that the parameter is Parameter performance value when and represents the parameter performance weight, Indicates that the parameter is The model training time is 200 ms. Indicates that the parameter is The model loss value at ;

[0057] Step S42: Initialization, creating a parameter search space, the search parameters include mapping expansion and contraction factors, weights and biases in the shield tunnel settlement prediction model; randomly generating initial parameter search points in the parameter search space, calculating the parameter performance values ​​and parameter performance averages of the initial parameter search points, and obtaining the parameter positions with maximum and minimum parameter performance;

[0058] Step S43: Design intervention factors, which are expressed as follows:

[0059] ;

[0060] in, Indicates the number of parameter searches. Indicates The intervention factor in the parameter search, Indicates the maximum number of parameter searches, Represents a random number ranging from -1 to 1;

[0061] Step S44: define the adaptive step size, which is expressed as follows:

[0062] ;

[0063] in, Indicates The adaptive step size during the parameter search, represents the average position of the initial parameter search points, Indicates the position with the highest current parameter performance value. Indicates the current highest parameter performance value. represents the average parameter performance value of the initial parameter search point, Indicates the current lowest parameter performance value;

[0064] Step S45: bidirectional iterative search, expressed as follows:

[0065] ;

[0066] in, Indicates The parameter position obtained by bidirectional iterative search, Indicates The parameter position obtained by bidirectional iterative search, represents a random number that follows a standard normal distribution, and r1 represents a random number ranging from 0 to 1;

[0067] Step S46: progressively adjust the search, as shown below:

[0068] ;

[0069] in, Indicates The parameter position obtained by the search is gradually adjusted. Indicates The parameter position obtained by the search is gradually adjusted, and r2 represents a random number ranging from 0 to 1;

[0070] Step S47: Dynamic search, set the parameter performance threshold and the maximum number of parameter searches, first perform a bidirectional iterative search, if the parameter performance of more than 50% of the parameter search points in the parameter positions obtained in a bidirectional iterative search is improved, directly perform the next bidirectional iterative search, otherwise perform a progressive adjustment search first, and then perform the next bidirectional iterative search; if the performance threshold of the parameter obtained by the search is greater than the parameter performance threshold, the search ends, and the parameter with the largest parameter performance is set as the model parameter; if , search again; otherwise continue searching.

[0071] Furthermore, in step S5, the settlement prediction is performed by collecting geological condition data, shield machine parameters, and environmental data at the construction site in real time, and inputting them into a shield tunnel settlement prediction model, and the model predicts the settlement of the tunnel.

[0072] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0073] (1) In order to solve the problem that traditional data optimization methods have poor data optimization effects due to imprecise dimension processing and insufficient data relationship mining when processing shield tunnel settlement prediction data, this scheme performs data cleaning by defining dimension refinement distance function, calculating gravity between data, cluster center evolution and test clustering, so as to more accurately reflect the characteristics and differences of data in different dimensions, deeply mine the potential relationships between data, reduce noise interference, and improve the accuracy and reliability of data.

[0074] (2) In order to address the problems of low prediction accuracy, poor adaptability to complex settlement patterns, and difficulty in capturing deep correlations in time series data in traditional shield tunnel settlement prediction models, this scheme improves the model's recognition and prediction accuracy of complex settlement patterns by designing mapping functions, loop control layers, deep correlation layers, and loss functions. This improves the model's ability to better adapt to various complex patterns that may occur during the settlement of shield tunnels, thereby improving the model's prediction accuracy and practicality.

[0075] (3) To address the problems of low search efficiency, easy to fall into local optimal solutions, and difficulty in balancing model training time and loss values ​​in traditional model optimization methods, this scheme dynamically searches model parameters by designing intervention factors, defining adaptive step sizes, bidirectional iterative search, and gradual adjustment search, thereby improving search efficiency, avoiding falling into local optimal solutions, balancing model training time and loss values, and making the model more efficient and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A schematic diagram of the flow of the shield tunnel settlement prediction method based on artificial intelligence provided by the present invention;

[0077] Figure 2 is a schematic flow chart of step S2;

[0078] Figure 3 is a schematic flow chart of step S3;

[0079] Figure 4 It is a schematic diagram of the process of step S4.

[0080] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0081] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0082] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0083] Example 1, see Figure 1 The present invention provides a shield tunnel settlement prediction method based on artificial intelligence, which comprises the following steps:

[0084] Step S1: data collection, collecting geological condition data, shield machine parameters, environmental data and settlement data of the historical construction section;

[0085] Step S2: Data optimization, optimizing data quality by defining dimension refinement distance function, calculating gravitational attraction between data, cluster center evolution, test clustering and data cleaning;

[0086] Step S3: constructing a shield tunnel settlement prediction model, by establishing a model infrastructure, designing a mapping function, a pre-processing layer, a loop control layer, a depth correlation layer, a prediction layer and a loss function to construct a shield tunnel settlement prediction model;

[0087] Step S4: Model optimization, optimizing model parameters by defining parameter performance functions, designing intervention factors, defining adaptive step sizes, bidirectional iterative search, progressive adjustment search, and dynamic search;

[0088] Step S5: Settlement prediction: by collecting the geological condition data, shield machine parameters and environmental data of the construction site in real time, inputting them into the shield tunnel settlement prediction model, the model predicts the settlement of the tunnel.

[0089] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, the data collection is to collect geological condition data, shield machine parameters, environmental data and settlement data of the historical construction section; the geological condition data include stratum type, groundwater level, soil density, compressive strength and internal friction angle; the shield machine parameters include shield machine type, excavation speed, shield machine posture, propulsion force and torque; the environmental data includes time series, ground load, rainfall and temperature; the settlement data refers to the settlement of the shield tunnel.

[0090] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S2, the data optimization specifically includes the following steps:

[0091] Step S21: define the dimension refinement distance function, which is expressed as follows:

[0092] ;

[0093] Among them, i and j represent the index of the data point, represents the dimension refinement distance function, represents the dimensionally refined distance between data point i and data point j, represents the square root, d represents the dimension index of the data point, and D represents the total number of dimensions of the data point. Indicates taking the absolute value, and Represent the eigenvalues ​​of the dth dimension of the i-th data point and the j-th data point, respectively. Indicates taking the maximum value, represents the maximum eigenvalue of the dth dimension, represents the dimension weight, Represents the indexes of the three data points closest to data point i, represents the eigenvalue of the dth dimension of the kth data point, represents the indexes of the three data points closest to data point j, Indicates The eigenvalue of the d-th dimension of the data point;

[0094] Step S22: Calculate the attraction between data, expressed as follows:

[0095] ;

[0096] in, represents the gravitational function between data, represents the gravitational strength between data point i and data point j, e represents a natural constant, represents the standard deviation of the eigenvalues ​​of the dth dimension, Indicates taking the minimum value, represents the minimum value between the eigenvalue of the dth dimension of the i-th data point and the eigenvalue of the dth dimension of the j-th data point, Represents the maximum value of the eigenvalue of the dth dimension of the i-th data point and the eigenvalue of the dth dimension of the j-th data point, Represents the difference between the maximum and minimum eigenvalues ​​of the dth dimension;

[0097] Step S23: Cluster center evolution, expressed as follows:

[0098] ;

[0099] Among them, l represents the index of the cluster center, represents the position of the lth cluster center after evolution, It means taking the closest data point. represents the position of the lth cluster center before evolution, represents the number of cluster centers, g represents the index of the data point in the lth cluster, Indicates the Euclidean distance. represents the Euclidean distance between the lth cluster center and the gth data point, represents the Euclidean distance between the g-th data point in the l-th cluster and the farthest data point, represents evolutionary weight;

[0100] Step S24: Test clustering, randomly select from all data points Initial cluster centers are calculated, the gravitational attraction strength between each data point and each cluster center is calculated, and the data points are assigned to the cluster with the largest gravitational attraction strength. The cluster centers evolve, and the assignment and evolution process is repeated until the cluster centers no longer change.

[0101] Step S25: Data cleaning, calculating the average number of data points in all clusters, setting the clusters with less than 50% of the average number of data points as abnormal clusters, and clearing the data points of the abnormal clusters.

[0102] By performing the above operations, in order to solve the problem of poor data optimization effect caused by imprecise dimension processing and insufficient data relationship mining in traditional data optimization methods when processing shield tunnel settlement prediction data, this scheme performs data cleaning by defining dimension refinement distance function, calculating gravity between data, cluster center evolution and test clustering, which can more accurately reflect the characteristics and differences of data in different dimensions, deeply mine the potential relationship between data, reduce noise interference, and improve data accuracy and reliability.

[0103] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the construction of the shield tunnel settlement prediction model specifically includes the following steps:

[0104] Step S31: Establishing a model infrastructure, the shield tunnel settlement prediction model includes a pre-processing layer, a loop control layer, a depth association layer and a prediction layer, and setting the settlement data as the label data of the model;

[0105] Step S32: Design a mapping function, which is expressed as follows:

[0106] ;

[0107] in, represents the input value of the mapping function, represents the mapping function, represents the mapping scaling factor, It means taking logarithm;

[0108] Step S33: Design the pre-processing layer, as shown below:

[0109] ;

[0110] in, represents the output of the pre-processing layer, , and Represent the weight, input and bias of the pre-processing layer respectively;

[0111] Step S34: Design a loop control layer, which consists of a reset unit, an update unit, a state reference unit, and a state output unit, and specifically includes the following steps:

[0112] Step S341: Design a reset unit, as shown below:

[0113] ;

[0114] Among them, t represents the index of the time point, represents the output of the reset unit at time t, represents the Sigmoid function, and Represent the weight and bias of the reset unit respectively, represents the input of the reset unit at time t, represents the state of the loop control layer at time t-1, represents the state of the loop control layer at time t-2, It means taking the square of L2 norm;

[0115] Step S342: Design an update unit, which is expressed as follows:

[0116] ;

[0117] in, represents the output of the update unit at time t, and Represent the weight and bias of the update unit respectively, Represents the input of the update unit at time t;

[0118] Step S343: Design state reference unit, expressed as follows:

[0119] ;

[0120] in, represents the output of the state reference unit at time t, represents the hyperbolic tangent function, and denote the weight and bias of the state reference unit, respectively, represents the element-by-element multiplication symbol, represents the input of the state reference unit;

[0121] Step S344: Design status output unit, expressed as follows:

[0122] ;

[0123] in, Represents the state of the loop control layer at time t;

[0124] Step S35: Design a depth association layer, which is expressed as follows:

[0125] ;

[0126] in, represents the output of the deep correlation layer, , and Represent the weight, input and bias of the deep association layer respectively;

[0127] Step S36: Design the prediction layer, which is expressed as follows:

[0128] ;

[0129] in, represents the output of the prediction layer, , and Represent the weight, input and bias of the prediction layer respectively, represents the normalized exponential function;

[0130] Step S37: Design a loss function, which is expressed as follows:

[0131] ;

[0132] in, Represents the loss value of the model, Represents the total number of model training data, represents the true value of the i-th data, Represents the model's predicted value for the i-th data. represents the loss weight, represents the skewness, Indicates kurtosis.

[0133] By performing the above operations, in order to address the problems of low prediction accuracy, poor adaptability to complex settlement patterns and difficulty in capturing deep correlations in time series data in traditional shield tunnel settlement prediction models, this scheme improves the model's recognition and prediction accuracy of complex settlement patterns by designing mapping functions, loop control layers, deep correlation layers and loss functions, better adapts to various complex patterns that may occur in the shield tunnel settlement process, and improves the model's prediction accuracy and practicality.

[0134] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S4, the model optimization specifically includes the following steps:

[0135] Step S41: define the parameter performance function, which is expressed as follows:

[0136] ;

[0137] in, represents the model parameters, Indicates that the parameter is Parameter performance value when and represents the parameter performance weight, Indicates that the parameter is The model training time is 200 ms. Indicates that the parameter is The model loss value at ;

[0138] Step S42: Initialization, creating a parameter search space, the search parameters include mapping expansion and contraction factors, weights and biases in the shield tunnel settlement prediction model; randomly generating initial parameter search points in the parameter search space, calculating the parameter performance values ​​and parameter performance averages of the initial parameter search points, and obtaining the parameter positions with maximum and minimum parameter performance;

[0139] Step S43: Design intervention factors, which are expressed as follows:

[0140] ;

[0141] in, Indicates the number of parameter searches. Indicates The intervention factor in the parameter search, Indicates the maximum number of parameter searches, Represents a random number ranging from -1 to 1;

[0142] Step S44: define the adaptive step size, which is expressed as follows:

[0143] ;

[0144] in, Indicates The adaptive step size during the parameter search, represents the average position of the initial parameter search points, Indicates the position with the highest current parameter performance value. Indicates the current highest parameter performance value. represents the average parameter performance value of the initial parameter search point, Indicates the current lowest parameter performance value;

[0145] Step S45: bidirectional iterative search, expressed as follows:

[0146] ;

[0147] in, Indicates The parameter position obtained by bidirectional iterative search, Indicates The parameter position obtained by bidirectional iterative search, represents a random number that follows a standard normal distribution, and r1 represents a random number ranging from 0 to 1;

[0148] Step S46: progressively adjust the search, as shown below:

[0149] ;

[0150] in, Indicates The parameter position obtained by the search is gradually adjusted. Indicates The parameter position obtained by the search is gradually adjusted, and r2 represents a random number ranging from 0 to 1;

[0151] Step S47: Dynamic search, set the parameter performance threshold and the maximum number of parameter searches, first perform a bidirectional iterative search, if the parameter performance of more than 50% of the parameter search points in the parameter positions obtained in a bidirectional iterative search is improved, directly perform the next bidirectional iterative search, otherwise perform a progressive adjustment search first, and then perform the next bidirectional iterative search; if the performance threshold of the parameter obtained by the search is greater than the parameter performance threshold, the search ends, and the parameter with the largest parameter performance is set as the model parameter; if , search again; otherwise continue searching.

[0152] By performing the above operations, in order to address the problems of low search efficiency, easy to fall into local optimal solutions, and difficulty in balancing model training time and loss values ​​in traditional model optimization methods, this solution dynamically searches model parameters by designing intervention factors, defining adaptive step sizes, bidirectional iterative search, and progressive adjustment search, thereby improving search efficiency, avoiding falling into local optimal solutions, balancing model training time and loss values, and making the model more efficient and stable.

[0153] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0154] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0155] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. The shield tunnel settlement prediction method based on artificial intelligence is characterized by: The method comprises the following steps: Step S1: data collection, collecting geological condition data, shield machine parameters, environmental data and settlement data of the historical construction section; Step S2: Data optimization, optimizing data quality by defining dimension refinement distance function, calculating gravitational attraction between data, cluster center evolution, test clustering and data cleaning; Among them, the dimension refinement distance function is defined as follows: ； Among them, i and j represent the index of the data point, represents the dimension refinement distance function, represents the dimensionally refined distance between data point i and data point j, represents the square root, d represents the dimension index of the data point, and D represents the total number of dimensions of the data point. Indicates taking the absolute value, and Represent the eigenvalues ​​of the dth dimension of the i-th data point and the j-th data point, respectively. Indicates taking the maximum value, represents the maximum eigenvalue of the dth dimension, represents the dimension weight, Represents the indexes of the three data points closest to data point i, represents the eigenvalue of the dth dimension of the kth data point, represents the indexes of the three data points closest to data point j, Indicates The eigenvalue of the d-th dimension of the data point; Step S3: constructing a shield tunnel settlement prediction model, by establishing a model infrastructure, designing a mapping function, a pre-processing layer, a loop control layer, a depth correlation layer, a prediction layer and a loss function to construct a shield tunnel settlement prediction model; Step S4: Model optimization, optimizing model parameters by defining parameter performance functions, designing intervention factors, defining adaptive step sizes, bidirectional iterative search, progressive adjustment search, and dynamic search; Step S5: Settlement prediction: by collecting the geological condition data, shield machine parameters and environmental data of the construction site in real time, inputting them into the shield tunnel settlement prediction model, the model predicts the settlement of the tunnel.

2. The method for predicting shield tunnel settlement based on artificial intelligence according to claim 1 is characterized in that: In step S2, the data optimization specifically includes the following steps: Step S21: defining a dimension refinement distance function; Step S22: Calculate the attraction between data, expressed as follows: ； in, represents the gravitational function between data, represents the gravitational strength between data point i and data point j, e represents a natural constant, represents the standard deviation of the eigenvalues ​​of the dth dimension, Indicates taking the minimum value, represents the minimum value between the eigenvalue of the dth dimension of the i-th data point and the eigenvalue of the dth dimension of the j-th data point, Represents the maximum value of the eigenvalue of the dth dimension of the i-th data point and the eigenvalue of the dth dimension of the j-th data point, Represents the difference between the maximum and minimum eigenvalues ​​of the dth dimension; Step S23: Cluster center evolution, expressed as follows: ； Where l represents the index of the cluster center, represents the position of the lth cluster center after evolution, It means taking the closest data point. represents the position of the lth cluster center before evolution, represents the number of cluster centers, g represents the index of the data point in the lth cluster, Indicates the Euclidean distance. represents the Euclidean distance between the lth cluster center and the gth data point, represents the Euclidean distance between the g-th data point in the l-th cluster and the farthest data point, represents evolutionary weight; Step S24: Test clustering, randomly select from all data points Initial cluster centers are calculated, the gravitational attraction strength between each data point and each cluster center is calculated, and the data points are assigned to the cluster with the largest gravitational attraction strength. The cluster centers evolve, and the assignment and evolution process is repeated until the cluster centers no longer change. Step S25: Data cleaning, calculating the average number of data points in all clusters, setting the clusters with less than 50% of the average number of data points as abnormal clusters, and clearing the data points of the abnormal clusters.

3. The method for predicting shield tunnel settlement based on artificial intelligence according to claim 1 is characterized in that: In step S3, the construction of the shield tunnel settlement prediction model specifically includes the following steps: Step S31: Establishing a model infrastructure, the shield tunnel settlement prediction model includes a pre-processing layer, a loop control layer, a depth association layer and a prediction layer, and setting the settlement data as the label data of the model; Step S32: Design a mapping function, which is expressed as follows: ； in, represents the input value of the mapping function, represents the mapping function, represents the mapping scaling factor, It means taking logarithm; Step S33: Design the pre-processing layer, as shown below: ； in, represents the output of the pre-processing layer, , and Represent the weight, input and bias of the pre-processing layer respectively; Step S34: Design a loop control layer, which consists of a reset unit, an update unit, a state reference unit, and a state output unit, and specifically includes the following steps: Step S341: Design a reset unit, as shown below: ； Among them, t represents the index of the time point, represents the output of the reset unit at time t, represents the Sigmoid function, and Represent the weight and bias of the reset unit respectively, represents the input of the reset unit at time t, represents the state of the loop control layer at time t-1, represents the state of the loop control layer at time t-2, It means taking the square of L2 norm; Step S342: Design an update unit, which is expressed as follows: ； in, represents the output of the update unit at time t, and Represent the weight and bias of the update unit respectively, Represents the input of the update unit at time t; Step S343: Design state reference unit, expressed as follows: ； in, represents the output of the state reference unit at time t, represents the hyperbolic tangent function, and denote the weight and bias of the state reference unit, respectively, represents the element-by-element multiplication symbol, represents the input of the state reference unit; Step S344: Design status output unit, expressed as follows: ； in, Represents the state of the loop control layer at time t; Step S35: Design a depth association layer, which is expressed as follows: ； in, represents the output of the deep correlation layer, , and Represent the weight, input and bias of the deep association layer respectively; Step S36: Design the prediction layer, which is expressed as follows: ； in, represents the output of the prediction layer, , and Represent the weight, input and bias of the prediction layer respectively, represents the normalized exponential function; Step S37: Design a loss function, which is expressed as follows: ； in, Represents the loss value of the model, Represents the total number of model training data, represents the true value of the i-th data, Represents the model's predicted value for the i-th data. represents the loss weight, represents the skewness, Indicates kurtosis.

4. The method for predicting shield tunnel settlement based on artificial intelligence according to claim 1 is characterized in that: In step S4, the model optimization specifically includes the following steps: Step S41: define the parameter performance function, which is expressed as follows: ； in, represents the model parameters, Indicates that the parameter is Parameter performance value when and represents the parameter performance weight, Indicates that the parameter is The model training time is 200 ms. Indicates that the parameter is The model loss value at ; Step S42: Initialization, creating a parameter search space, the search parameters include mapping expansion and contraction factors, weights and biases in the shield tunnel settlement prediction model; randomly generating initial parameter search points in the parameter search space, calculating the parameter performance values ​​and parameter performance averages of the initial parameter search points, and obtaining the parameter positions with maximum and minimum parameter performance; Step S43: Design intervention factors, which are expressed as follows: ； in, Indicates the number of parameter searches. Indicates The intervention factor in the parameter search, Indicates the maximum number of parameter searches, Represents a random number ranging from -1 to 1; Step S44: define the adaptive step size, which is expressed as follows: ； in, Indicates The adaptive step size during the parameter search, Indicates the Euclidean distance. represents the average position of the initial parameter search points, Indicates the position with the highest performance value of the current parameter. Indicates the current highest parameter performance value. represents the average parameter performance value of the initial parameter search point, Indicates the current lowest parameter performance value; Step S45: bidirectional iterative search, expressed as follows: ； in, Indicates The parameter position obtained by bidirectional iterative search, Indicates The parameter position obtained by bidirectional iterative search, represents a random number that follows a standard normal distribution, and r1 represents a random number ranging from 0 to 1; Step S46: progressively adjust the search, as shown below: ； in, Indicates The parameter position obtained by the search is gradually adjusted. Indicates The parameter position obtained by the search is gradually adjusted, and r2 represents a random number ranging from 0 to 1; Step S47: Dynamic search, set the parameter performance threshold and the maximum number of parameter searches, first perform a bidirectional iterative search, if the parameter performance of more than 50% of the parameter search points in the parameter positions obtained in a bidirectional iterative search is improved, directly perform the next bidirectional iterative search, otherwise perform a progressive adjustment search first, and then perform the next bidirectional iterative search; if the performance threshold of the parameter obtained by the search is greater than the parameter performance threshold, the search ends, and the parameter with the largest parameter performance is set as the model parameter; if , search again; otherwise continue searching.

5. The method for predicting shield tunnel settlement based on artificial intelligence according to claim 1 is characterized in that: In step S1, the data collection is to collect geological condition data, shield machine parameters, environmental data and settlement data of the historical construction section; the geological condition data include stratum type, groundwater level, soil density, compressive strength and internal friction angle; the shield machine parameters include shield machine type, excavation speed, shield machine posture, propulsion force and torque; the environmental data includes time series, ground load, rainfall and temperature; the settlement data refers to the settlement of the shield tunnel.

Images