Shield tunneling machine propelling speed prediction method

Abstract

The invention discloses a shield tunneling machine propulsion speed prediction method which takes Copula entropy as feature selection and is based on a whale optimization algorithm and a long-short-term memory network, and the method comprises the following steps: collecting shield data; invalid shield data are deleted in a null value removing mode; denoising the propulsion speed data by adopting a mean filtering method; calculating a Copula entropy value between each shield tunneling parameter and the propulsion speed, and screening out a predictive factor meeting the requirement by using a Hampel criterion to serve as a model input feature; performing normalization processing on shield tunneling parameters, and dividing the shield tunneling parameters into a training set and a test set; and in combination with data of the training set, optimizing hyper-parameters of the long-short-term memory network prediction model by adopting a whale optimization algorithm to obtain an optimized long-short-term memory network shield propulsion speed prediction model. According to the method, the dimensionality of the input features is reduced to reduce the calculation amount, the blindness of manually selecting hyper-parameters is avoided, and the prediction performance of the long and short-term memory network is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence in shield tunneling, and in particular to a shield machine propulsion speed prediction method using Copula entropy for feature selection and based on a whale optimization algorithm and a long short-term memory network. Background Art

[0002] The shield method is widely used as a safe and efficient method in subway construction. Most domestic subways are constructed using the shield method. However, due to the lack of a model that can accurately predict the interaction between shield tunneling and strata in real time, during the shield construction process, the shield machine driver mainly sets the tunneling parameters based on the earth pressure calculation formula and adjusts the tunneling state of the shield machine in a timely manner based on subjective experience in actual operation. Therefore, there is a possibility of risk due to lack of experience. The empirical formula used in the past to predict the tunneling parameters of shield machines is affected by many factors, whether from a macro or micro perspective, and exhibits typical complex nonlinear characteristics. Therefore, simple empirical formulas or analytical methods may have large errors in predicting the tunneling parameters of shield machines.

[0003] In actual engineering projects, it is necessary to plan the progress of the project. The tunneling speed of the shield machine is a key factor affecting the progress of the project. Simple empirical formulas or analytical methods may have large errors in predicting the tunneling speed of the shield machine. Shield tunneling is a continuous process, and tunneling parameters will change over time. Therefore, more and more scholars use time series methods to model and analyze, but in the selection of model inputs, most scholars use subjective judgment or Pearson correlation analysis, and often manually adjust the model hyperparameters, which has certain limitations. Summary of the invention

[0004] The purpose of the present invention is to address the defects of the prior art and provide a shield machine propulsion speed prediction method based on Copula entropy for feature selection, whale optimization algorithm and long short-term memory network, so as to overcome the problem of low model prediction accuracy caused by unreasonable input parameter selection and improper hyperparameter setting in the traditional long short-term memory network prediction process.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A shield machine propulsion speed prediction method is provided, which uses Copula entropy as feature selection, is based on a whale optimization algorithm and a long short-term memory network, and includes the following steps:

[0007] Step S1: Collect shield data: Shield data mainly includes parameter columns that characterize the time state of the system and the tunneling state information of the shield machine, including cutter head speed, propulsion speed, total thrust, cutter head torque, etc.;

[0008] Step S2: Deleting invalid shield data by removing null values;

[0009] Step S3: De-noising the propulsion speed data using a mean filter method;

[0010] Step S4: Calculate the Copula entropy value between each shield tunneling parameter and the advancement speed, and use the Hampel criterion to select the prediction factors that meet the requirements as the model input features;

[0011] Step S5: normalize the shield tunneling parameters, and divide the screened data set into a training set and a test set in a ratio of 8:2;

[0012] Step S6: Based on the training set data, the whale optimization algorithm is used to optimize the hyperparameters of the long short-term memory network prediction model to obtain the optimized long short-term memory network shield advancement speed prediction model.

[0013] Furthermore, the step S3 specifically includes: setting a reasonable window size N, where N is the amount of data contained, and calculating the average value of the N advancement speeds in the window to form a new series when the window slides forward, thereby replacing the original advancement speed parameter series.

[0014] In the above technical solution, preferably, step S4 specifically comprises: calculating the Copula entropy between each shield tunneling parameter and the advancement speed, that is, the inverse of the mutual information, using the Hampel criterion as the corresponding test condition, and screening qualified prediction factors.

[0015] Furthermore, the step S6 specifically includes: constructing a 10→1 supervised learning data set in the form of predicting the last group by the first 10 groups.

[0016] Furthermore, the model of the long short-term memory network adopts a structure including an input layer, two LSTM layers and an output layer, and uses the mean square error as the loss function.

[0017] Furthermore, the maximum number of training rounds of the model is set to N, the number of neurons in the two hidden layers of the model N1, N2, and the batch size are optimized through the whale algorithm, the initial search range is set respectively, and then iterative optimization is performed to find the optimal hyperparameters within the search range. The number of whales, the maximum number of iterations, and the whale dimension in the whale optimization algorithm are set; the optimal parameter results are input into the long short-term memory network prediction model to obtain the optimized long short-term memory network prediction model.

[0018] By adopting the technical solution of the present invention, the beneficial effects of the present invention are:

[0019] 1. Use Copula entropy combined with Hanpel criterion for feature selection, thereby reducing the dimension of input features to reduce the amount of calculation;

[0020] 2. Traditional parameter adjustment methods have low reliability. The present invention uses the whale optimization algorithm to optimize the long short-term memory network prediction model, introduces the bubble net hunting strategy, and achieves the purpose of optimized search through the process of whale group searching, encircling, chasing and attacking prey, avoiding the blindness of artificially selecting hyperparameters, thereby effectively improving the prediction performance of the long short-term memory network. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a shield machine propulsion speed prediction method provided by the present invention;

[0022] Figure 2 It is a mean filtering schematic diagram of a shield machine advancement speed prediction method provided by the present invention;

[0023] Figure 3 It is a fitness curve diagram of the whale optimization process of a shield machine propulsion speed prediction method provided by the present invention;

[0024] Figure 4 It is a training set and test set loss curve diagram of a shield machine propulsion speed prediction method provided by the present invention;

[0025] Figure 5 It is a prediction result diagram of a shield machine propulsion speed model 1 of a shield machine propulsion speed prediction method provided by the present invention;

[0026] Figure 6 It is a prediction result diagram of shield machine propulsion speed model 2 of a shield machine propulsion speed prediction method provided by the present invention;

[0027] Figure 7 It is a prediction result diagram of shield machine propulsion speed model 3 of a shield machine propulsion speed prediction method provided by the present invention;

[0028] Figure 8 It is a prediction result diagram of shield machine propulsion speed model 1 of a shield machine propulsion speed prediction method provided by the present invention. DETAILED DESCRIPTION

[0029] The specific embodiments of the present invention are further described in conjunction with the accompanying drawings.

[0030] As shown in the figure, in order to facilitate the understanding of the present invention, the present invention will be described more comprehensively below, and preferred embodiments of the present invention are given. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0031] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0032] Taking the shield data of a single circular shield section of a subway as an example, the details are as follows.

[0033] A method for predicting the advancement speed of a shield machine comprises the following steps:

[0034] Step S1: Collect shield data: Shield data mainly includes parameter columns that characterize the time state of the system and the tunneling state information of the shield machine, including cutter head rotation speed, propulsion speed, total thrust, cutter head torque, etc.

[0035] Preferably, the step S1 specifically includes: the interval between two shield machine data obtained through on-site collection is 10 seconds, including parameter columns representing the time state of the system, the tunneling state of the shield machine and other information. Figure 1 As shown, excluding the propulsion speed, there are 33 valid characteristic parameters including cutter disc torque, total thrust, screw machine torque, cutter disc speed, screw machine speed, etc.

[0036] Step S2: Delete invalid shield data by removing null values.

[0037] Preferably, the step S2 specifically includes: deleting all shield machine data when the propulsion speed is ≤0 or the cutter head rotation speed is ≤0 or the cutter head torque is ≤0.

[0038] Step S3: De-noising the propulsion speed data using the mean filtering method.

[0039] Preferably, step S3 specifically includes: setting a reasonable window size of 10, where 10 is the amount of data contained, and calculating the average of 10 advancement speeds in the window to form a new sequence when the window slides forward, thereby replacing the original advancement speed parameter sequence, such as Figure 2 shown.

[0040] Step S4: Calculate the Copula entropy value between each shield tunneling parameter and the advancement speed, and use the Hampel criterion to screen out the prediction factors that meet the requirements as model input features.

[0041] Preferably, the step S4 specifically comprises: calculating the Copula entropy between each shield tunneling parameter and the advancing speed, as shown in the following formula:

[0042]

[0043] Where: P(xi ) and P(y j ) represent the marginal density functions of variables X and Y, respectively, P(x i ,y j ) represents the joint probability density function of variables X and Y.

[0044] Calculate the Hampel entropy value between shield tunneling parameters and advancement speed; as shown in the following formula:

[0045]

[0046] Where: H j is the Hampel distance; 1.4826 is the normalization factor; d j For mutual information and The absolute value of the median difference of mutual information; is d j of the median.

[0047] Using the Hampel criterion, the predictive factors that meet the requirements are screened out, as shown in the following formula;

[0048]

[0049] Where: C PMI is mutual information; is the median of the mutual information. According to the 3σ criterion of the standard deviation, when H j When >3, the predictors are selected. Generally, if there are more than 10 qualified factors, only the top 10 predictors in entropy ranking are selected as the final predictors; if there are less than 10, all variables are the final predictors.

[0050] Step S5: Normalize the shield tunneling parameters and divide the filtered data set into a training set and a test set in a ratio of 8:2.

[0051] Preferably, the step S5 specifically includes: normalizing the data before model training. The formula is as follows:

[0052]

[0053] Where: x max , x min is the maximum and minimum value of variable x, are the maximum and minimum values ​​of the normalized variable x, which are 1 and -1 respectively.

[0054] Step S6: Based on the training set data, the whale optimization algorithm is used to optimize the hyperparameters of the long short-term memory network prediction model to obtain the optimized long short-term memory network shield advancement speed prediction model.

[0055] Table 1 is an overview of the above model:

[0056]

[0057] Table 1

[0058] Table 2 shows the model prediction error:

[0059]

[0060] Table 2

[0061] Preferably, the step S6 specifically includes: initializing the parameters of the whale optimization algorithm, setting the number of whales to be searched to 30, the maximum number of iterations to 20, the whale dimension to 3, and randomly generating the position X according to the whale optimization algorithm. The iteration process is as follows: Figure 3 As shown in the figure, the maximum number of training rounds for the LSTM network is set to 60, and the three hyperparameters of the number of neurons in the two hidden layers of the model, N1 and N2, and the batch size, are optimized using the whale algorithm, with the initial search ranges set to [1,100], [1,100], and [0,40], respectively.

[0062] According to the following two formulas, calculate the individual fitness and save the best individual and the best position.

[0063]

[0064]

[0065] Where: is the distance between the individual whale and the optimal position; t is the current iteration number; is the position vector of the individual whale; is the optimal solution position vector; and is the coefficient vector.

[0066] Update A and C according to the following two formulas:

[0067] and The calculation formula is:

[0068]

[0069]

[0070] Where: To control the parameters, it decreases linearly from 2 to 0; is a random vector in [0,1]; T max is the maximum number of iterations.

[0071] Generate a random number p, when p ≥ 0.5, update the position according to the following formula;

[0072]

[0073] Where: is the distance between the individual whale and its prey; b is a constant that defines the shape of the spiral; l is a random number between [-1,1].

[0074] When p<0.5 and A>1; update the former position according to the following formula.

[0075]

[0076] Where: is the position vector of a randomly selected individual whale.

[0077] Determine whether the termination condition is met. If the termination condition is met, the optimal value of the optimization target can be obtained. Otherwise, return to step 6 and continue the calculation from the beginning until the termination condition is met.

[0078] Reassign the optimal hyperparameters obtained by optimization to the prediction model and adjust the training rounds, such as Figure 4 As shown in the figure, the final shield tunneling parameter prediction model is obtained.

[0079] Step S7: Select root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and coefficient of determination (R 2 ) as the evaluation index, the prediction results of shield advancement speed are obtained, such as Figure 5-8 shown.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0081] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for predicting the propulsion speed of a shield machine, characterized in that: The shield machine propulsion speed prediction method based on Copula entropy for feature selection, whale optimization algorithm and long short-term memory network includes the following steps: Step S1: Collect shield data: Shield data mainly includes parameter columns that characterize the time state of the system and the tunneling state information of the shield machine, including cutter head speed, propulsion speed, total thrust, cutter head torque, etc.; Step S2: Deleting invalid shield data by removing null values; Step S3: De-noising the propulsion speed data using a mean filter method; Step S4: Calculate the Copula entropy value between each shield tunneling parameter and the advancement speed, and use the Hampel criterion to select the prediction factors that meet the requirements as the model input features; Step S5: normalize the shield tunneling parameters, and divide the screened data set into a training set and a test set in a ratio of 8:2; Step S6: Based on the training set data, the whale optimization algorithm is used to optimize the hyperparameters of the long short-term memory network prediction model to obtain the optimized long short-term memory network shield advancement speed prediction model.

2. A method for predicting the propulsion speed of a shield machine as claimed in claim 1, characterized in that: The step S3 specifically includes: setting a reasonable window size N, where N is the amount of data contained, and calculating the average value of the N advancing speeds in the window to form a new sequence when the window slides forward, thereby replacing the original advancing speed parameter sequence.

3. A method for predicting the propulsion speed of a shield machine as claimed in claim 1, characterized in that: Preferably, in the above technical solution, step S4 specifically comprises: calculating the Copula entropy between each shield tunneling parameter and the advancement speed, that is, the inverse of the mutual information, using the Hampel criterion as the corresponding test condition, and screening qualified prediction factors.

4. A method for predicting the propulsion speed of a shield machine as claimed in claim 1, characterized in that: The step S6 specifically includes: constructing a 10→1 supervised learning data set in the form of predicting the last group by using the first 10 groups.

5. A method for predicting the propulsion speed of a shield machine as claimed in claim 1, characterized in that: The model of the long short-term memory network adopts a structure including an input layer, two LSTM layers and an output layer, and uses mean square error as the loss function.

6. A method for predicting the propulsion speed of a shield machine as claimed in claim 5, characterized in that: The maximum number of training rounds of the model is set to N, the number of neurons in the two hidden layers of the model is N1, N2, and the batch size is optimized through the whale algorithm. The initial search range is set separately, and then iterative optimization is performed to find the optimal hyperparameters within the search range. The number of whales, the maximum number of iterations, and the whale dimension in the whale optimization algorithm are set; the optimal parameter results are input into the long short-term memory network prediction model to obtain the optimized long short-term memory network prediction model.

Images