Method for accurately predicting wall cutting thrust of cutterhead when shield tunneling machine passes through reinforced concrete diaphragm wall
Through the BO-BiGRU-Attention hybrid prediction model, combined with the working state parameters of the shield machine and the changes in the position of the hob, the precise prediction of the cutting-edge wall thrust when the shield machine passes through the reinforced concrete land and connects the wall, solving the existing technology The problem of limited thrust prediction effect improves construction safety and efficiency.
Patent Information
- Application Number
- CN202411331212.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-24
- Publication Date
- 2024-12-10
AI Technical Summary
In tunnel engineering, when the shield machine crosses the reinforced concrete wall connection, it is difficult to achieve accurate prediction of the cutting wall thrust of the cutter wheel, resulting in low construction safety and efficiency. The existing methods fail to effectively utilize the information in the monitoring data and predict the effect limited.
Using the BO-BiGRU-Attention hybrid prediction model, the mapping relationship between the total shield thrust and the cutting wall cutting thrust was constructed by collecting and pre-processing the working state parameters of the shield machine. Combining the position change of the hob and Gaussian fitting, the construction was constructed. Hybrid prediction models perform real-time thrust prediction.
Improve data utilization, provide accurate prediction of cutting-edge wall thrust, improve construction safety and efficiency, and reduce construction costs.
Smart Images
Figure CN119102655A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of tunnel engineering, and in particular to a method for accurately predicting the wall-cutting thrust of a cutter head when a shield machine passes through a reinforced concrete ground-connected wall. Background Art
[0002] In the construction of urban rail transit lines, it has become a common construction scenario for new section tunnels to pass under existing stations at close range to achieve efficient transfers. In this process, the shield machine needs to pass through the unreserved reinforced concrete ground-connected walls on both sides of the station. The manual obstacle removal method has gradually been replaced by the shield direct cutting method due to its high cost, long construction period and high risk. However, when the shield machine cuts reinforced concrete, due to its inadaptability to cutting materials, problems such as steel bars being pulled or twisted, cutterheads being stuck or screw conveyors being blocked often occur, which seriously affect construction safety and efficiency. The cutterhead wall cutting thrust is a force that directly acts on the ground-connected wall. Its accurate prediction is crucial to improving engineering efficiency and safety. However, since the cutterhead thrust data is difficult to obtain directly, there is currently no mature method that can effectively predict the cutterhead thrust. Although scholars have conducted some research on the prediction of shield thrust, these methods are often oversimplified and fail to make full use of the rich information in the monitoring data, resulting in limited prediction effects. Summary of the invention
[0003] Purpose of the invention: The purpose of the present invention is to provide a method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall, thereby solving the problem of difficulty in obtaining data on the engineering site.
[0004] Technical solution: The invention discloses a method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall, comprising the following steps: (1) Collect and pre-process the working state parameters of the shield machine when it passes through the reinforced concrete ground-connected wall; (2) Analyze the data from field tests of actual projects to obtain the mapping relationship between the total thrust of the shield and the thrust of the cutterhead; (3) Construct a BO-BiGRU-Attention hybrid prediction model to predict the total thrust of the shield machine; (4) The model is applied to actual projects. The mapping relationship and the shield thrust prediction model are combined to predict the cutterhead wall thrust in real time.
[0005] Furthermore, in step (1), the working state parameters include: excavation speed, cutter head speed, shield total thrust, cutter head wall cutting thrust; the preprocessing is as follows: the collected working parameters are stored as an Excel file, and then the mean substitution method is used to process the abnormal values; the formula is as follows: ; in, represents the abnormal data value, i represents the row number where the abnormal data is located, and j represents the column number where the abnormal data is located.
[0006] Furthermore, step (2) includes the following steps: (21) The positional relationship between the shield machine cutterhead and the ground-connected wall can be decomposed into: y 1 、y 2 、y 3 The thickness of the ground-connected wall is B The cutter head diameter is D , construct a coordinate system; (22) Combined with the position coordinates of the hob, the trend diagram of the number of hobs changing with time is obtained. After Gaussian fitting, a continuous and accurate model of the change of the number of tools with time is obtained. According to the model, the number of tools that actually cut the wall at the same time during the entire cutting process is obtained; (23) Calculate the ratio of the cutterhead wall thrust to the total shield thrust using the following formula: ; ; ; ; ; Among them, the thrust data at the same pushing speed in the field test are taken respectively; among them, the pushing speed is mm / min; the values of other parameters are all from actual projects: n is the number of cutters acting on the wall at this time; N is Total number of tools, The average contact area between the cutter head and the ground-connected wall during the cutting process; L is the main length of the shield machine, in meters; is the soil bin pressure; c is Reduction of unconfined compressive strength of MJS; is the reduction factor, is the friction factor between the MJS stratum and the shield shell; η is Cutting head opening rate; It is the total weight of the supporting equipment behind the shield machine; is the proportional coefficient between the cutterhead thrust and the shield thrust; F 总 is the total thrust of the shield machine, which is composed of the cutting resistance and friction resistance of the cutter head F 2 Composition; Among them, the cutting resistance of the cutter head includes: the resistance of passing through the wall F 11 and resistance to cross MJS reinforcement area F 12。
[0007] Furthermore, step (3) includes the following steps: (31) Standardize the data; the formula is as follows: ; in, represents the standardized total thrust data, x represents the original data of the total thrust of the shield, represents the maximum total thrust of the shield machine, Indicates the minimum total thrust of the shield; (32) Feature selection: Use Spearman correlation analysis to select key features with high correlation with the total thrust of the shield machine from the preprocessed data; and divide the data into training set and validation set; (33) The BO-BiGRU-Attention hybrid prediction model is constructed, including: input layer, BIGRU layer, Attention layer, Bayesian optimization BO and output layer.
[0008] Furthermore, the BIGRU layer is specifically as follows: The BIGRU bidirectional gated recurrent unit network is used to capture the temporal dependencies in the data.
[0009] Furthermore, the Attention layer is specifically as follows: the attention mechanism is applied to the output layer of the BiGRU layer. The attention mechanism enables the model to obtain key information in the data by calculating the weight of each time step.
[0010] Furthermore, the Bayesian optimization BO includes the following steps: S1 initialization: select the initial sample point set; S2 builds a probability model: builds a Gaussian regression probability model based on the existing sample points; S3 selects the next sample point: uses the probability model to make inferences and selects the next sample point based on the inference results; S4 evaluates the objective function MAE at the selected sample points and obtains new observations; S5 updates the probability model: adds new sample points and observations to the existing sample point set and updates the probability model; S6 checks the termination condition: determines whether to terminate the algorithm based on whether the preset objective function converges and the maximum number of iterations.
[0011] Furthermore, the output layer is as follows: after the data is denormalized, the predicted total thrust value of the shield is output, and the Loss function is used to evaluate the model effect; the formula is as follows: The denormalization formula is as follows: ; The formula for calculating the Loss value is as follows: .
[0012] An electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, it implements any one of the methods described in the item to accurately predict the wall cutting thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall.
[0013] A storage medium described in the present invention stores a computer program, and when the computer program is executed by a processor, it implements any method for accurately predicting the wall cutting thrust of a cutter head when a shield machine passes through a reinforced concrete ground-connected wall.
[0014] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: improved data utilization; revealed the mapping relationship between the total thrust of the shield and the thrust of the cutterhead cutting the wall, providing a theoretical basis for accurate prediction; the constructed BO-BiGRU-Attention hybrid prediction model has high prediction accuracy, can predict the thrust of the cutterhead cutting the wall in real time, and provide accurate guidance for construction; improved the construction safety and efficiency of the shield machine passing through the reinforced concrete ground-connected wall, and reduced the construction cost. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 The coordinate system of the present invention is constructed based on the positional relationship between the cutter head surface of the shield machine and the ground-connected wall.
[0016] Figure 2 It is the change of theoretical number of cutters and Gauss fitting curve in the wall cutting process of the present invention.
[0017] Figure 3 It is the ratio of the wall cutting thrust to the total thrust of the present invention.
[0018] Figure 4 It is the Spearman correlation analysis diagram of the present invention.
[0019] Figure 5 It is the BiGRU structure diagram of the present invention.
[0020] Figure 6 This is a schematic diagram of the Attention structure of the present invention.
[0021] Figure 7 It is a flow chart of the Bayesian optimization steps of the present invention.
[0022] Figure 8 This is the flow chart of the BO-BiGRU-Attention algorithm of the present invention.
[0023] Fig. 9 It is a density scatter diagram of the actual value and predicted value of the total thrust of the shield machine according to the present invention.
[0024] Fig.10 is the error value of the thrust on the training set and the validation set of the present invention.
[0025] Fig.11 It is a density scatter diagram of the theoretical value and predicted value of the wall cutting thrust of the cutter head of the present invention. DETAILED DESCRIPTION
[0026] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.
[0027] like Figure 1 As shown, an embodiment of the present invention provides a method for accurately predicting the wall cutting thrust of a cutter head when a shield machine passes through a reinforced concrete ground-connected wall, comprising the following steps: (1) The shield machine working state parameters are collected and preprocessed when the shield machine passes through the reinforced concrete ground-connected wall. The working state parameters include: excavation speed, cutter head speed, total thrust of the shield machine, and cutter head wall cutting thrust. The preprocessing is as follows: the collected working parameters are stored as an Excel file, and then the mean substitution method is used to process the outliers. The formula is as follows: ; in, represents the abnormal data value, i represents the row number where the abnormal data is located, and j represents the column number where the abnormal data is located.
[0028] (2) Analyze the data from the field test of the actual project to obtain the mapping relationship between the total thrust of the shield and the thrust of the cutterhead cutting the wall; including the following steps: (21) The positional relationship between the shield machine cutterhead and the ground-connected wall can be decomposed into: y 1 、y 2 、y 3 The thickness of the ground-connected wall is B The cutter head diameter is D ,like Figure 1 As shown, construct a coordinate system; (22) Figure 2 As shown, combined with the position coordinates of the hob, the trend diagram of the number of hobs changing with time is obtained. After Gaussian fitting, a continuous and accurate model of the number of tools changing with time is obtained. According to the model, the number of tools that actually cut the wall at the same time during the entire cutting process is obtained; (23) Calculate the ratio of the cutterhead wall thrust to the total shield thrust using the following formula: ; ; ; ; ; Among them, the thrust data at the same pushing speed in the field test are taken respectively; among them, the pushing speed is mm / min; the values of other parameters are all from actual projects: n is the number of cutters acting on the wall at this time; N is Total number of tools, The average contact area between the cutter head and the ground-connected wall during the cutting process; L is the main length of the shield machine, in meters; is the soil bin pressure; c is Reduction of unconfined compressive strength of MJS; is the reduction factor, is the friction factor between the MJS stratum and the shield shell; η is Cutting head opening rate; It is the total weight of the supporting equipment behind the shield machine; is the proportional coefficient between the cutterhead thrust and the shield thrust; F 总 is the total thrust of the shield machine, which is composed of the cutting resistance and friction resistance of the cutter head F 2 Composition; Among them, the cutting resistance of the cutter head includes: the resistance of passing through the wall F 11 and resistance to cross MJS reinforcement area F 12。 As shown in Table 1.
[0029] Table 1 ; The ratio of the wall cutting force of the cutter head to the total thrust is as follows: Figure 3 shown.
[0030] (3) Constructing a BO-BiGRU-Attention hybrid prediction model to predict the total thrust of the shield machine; including the following steps: (31) Standardize the data; the formula is as follows: ; in, represents the standardized total thrust data, x represents the original data of the total thrust of the shield, represents the maximum total thrust of the shield machine, Indicates the minimum total thrust of the shield; (32) Figure 4 As shown, feature selection: Spearman correlation analysis is used to select key features with high correlation with the total thrust of the shield from the preprocessed data; and the training set and the validation set are divided; (33) Figure 5-8 As shown in the figure, the BO-BiGRU-Attention hybrid prediction model includes: input layer, BIGRU layer, Attention layer, Bayesian optimization BO and output layer.
[0031] The BIGRU layer is specifically as follows: The BIGRU bidirectional gated recurrent unit network is used to capture the temporal dependencies in the data.
[0032] The details of the Attention layer are as follows: the attention mechanism is applied to the output layer of the BiGRU layer. The attention mechanism enables the model to obtain key information in the data by calculating the weight of each time step.
[0033] Bayesian optimization BO includes the following steps: S1 initialization: select the initial sample point set; S2 builds a probability model: builds a Gaussian regression probability model based on the existing sample points; S3 selects the next sample point: uses the probability model to make inferences and selects the next sample point based on the inference results; S4 evaluates the objective function MAE at the selected sample points and obtains new observations; S5 updates the probability model: adds new sample points and observations to the existing sample point set and updates the probability model; S6 checks the termination condition: determines whether to terminate the algorithm based on whether the preset objective function converges and the maximum number of iterations.
[0034] The optimal hyperparameters of the model finally determined by the Yesian optimization method are shown in Table 2.
[0035] Table 2 ; The output layer is as follows: after denormalizing the data, the predicted total thrust value of the shield is output, and the Loss function is used to evaluate the model effect; the formula is as follows: ; The formula for calculating the Loss value is as follows: .
[0036] The relative position distribution diagram of the actual total propulsion force and the predicted total propulsion force is shown in the figure below: Fig. 9 , Fig.10 The error curves of the model established by the present invention on the thrust and torque data sets are shown. As can be seen from the figure, with the increase in the number of iterations, the error values on the training set and the validation set gradually decrease and finally tend to be stable. When the number of iterations increases to 200, the prediction error value of the thrust drops to an extremely low and stable level, and the error values on the training set and the validation set are finally 0.002 and 0.003 respectively. The small error value indicates that the prediction model proposed by the present invention avoids the problem of overfitting and has good prediction accuracy and generalization ability.
[0037] (4) Apply the model to actual projects, combine the mapping relationship and the shield thrust prediction model, and predict the wall cutting thrust of the cutter head in real time. The specific strategy is as follows: the value of the ratio corresponds to the thrust speed. When the thrust speed value is within a certain range, it corresponds to a certain proportional coefficient. Then, the proportional coefficient can be dynamically selected to improve the accuracy of the model, such as Fig.11 .
[0038] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, it implements any one of the methods described in the method for accurately predicting the wall cutting thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall.
[0039] An embodiment of the present invention also provides a storage medium storing a computer program. When the computer program is executed by a processor, it implements any one of the methods for accurately predicting the wall cutting thrust of a cutter head when a shield machine passes through a reinforced concrete ground-connected wall.
Claims
1. An accurate prediction method for the thrust of the cutterhead when a shield machine passes through a reinforced concrete ground-connected wall. It is characterized in that The following steps are involved: (1) Collect and pre-process the working state parameters of the shield machine when it passes through the reinforced concrete ground-connected wall; (2) Analyze the data from field tests of actual projects to obtain the mapping relationship between the total thrust of the shield and the thrust of the cutterhead; (3) Construct a BO-BiGRU-Attention hybrid prediction model to predict the total thrust of the shield machine; (4) The model is applied to actual projects. The mapping relationship and the shield thrust prediction model are combined to predict the cutterhead wall thrust in real time.
2. According to the method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall as described in claim 1, It is characterized in that In step (1), the working state parameters include: excavation speed, cutter head speed, total thrust of shield machine, and cutter head wall cutting thrust; the preprocessing is as follows: the collected working parameters are stored as an Excel file, and then the mean substitution method is used to process abnormal values; the formula is as follows: ； in, represents the abnormal data value, i represents the row number where the abnormal data is located, and j represents the column number where the abnormal data is located.
3. According to the method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall as described in claim 1, It is characterized in that Step (2) includes the following steps: (21) The positional relationship between the shield machine cutterhead and the ground-connected wall can be decomposed into: y 1 、y 2 、y 3 The thickness of the ground-connected wall is B The cutter head diameter is D , construct a coordinate system; (22) Combined with the position coordinates of the hob, the trend diagram of the number of hobs changing with time is obtained. After Gaussian fitting, a continuous and accurate model of the change of the number of tools with time is obtained. According to the model, the number of tools that actually cut the wall at the same time during the entire cutting process is obtained; (23) Calculate the ratio of the cutterhead wall thrust to the total shield thrust using the following formula: ； ； ； ； ； Among them, the thrust data at the same pushing speed in the field test are taken respectively; among them, the pushing speed is mm / min; the values of other parameters are all from actual projects: n is the number of cutters acting on the wall at this time; N is Total number of tools, The average contact area between the cutter head and the ground-connected wall during the cutting process; L is the main length of the shield machine, in meters; is the soil bin pressure; c is Reduction of unconfined compressive strength of MJS; is the reduction factor, is the friction factor between the MJS formation and the shield shell; η is Cutting head opening rate; It is the total weight of the supporting equipment behind the shield machine; is the proportional coefficient between the cutterhead thrust and the shield thrust; F 总 is the total thrust of the shield machine, which is composed of the cutting resistance and friction resistance of the cutter head F 2 Composition; Among them, the cutting resistance of the cutter head includes: the resistance of passing through the wall F 11 and resistance to cross MJS reinforcement area F 12。 4. According to the method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall as described in claim 1, It is characterized in that Step (3) includes the following steps: (31) Standardize the data; the formula is as follows: ； in, represents the standardized total thrust data, x represents the original data of the total thrust of the shield, represents the maximum total thrust of the shield machine, Indicates the minimum total thrust of the shield; (32) Feature selection: Use Spearman correlation analysis to select key features with high correlation with the total thrust of the shield machine from the preprocessed data; and divide the data into training set and validation set; (33) The BO-BiGRU-Attention hybrid prediction model is constructed, including: input layer, BIGRU layer, Attention layer, Bayesian optimization BO and output layer.
5. According to the method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall as described in claim 4, It is characterized in that The BIGRU layer is specifically as follows: The BIGRU bidirectional gated recurrent unit network is used to capture the temporal dependencies in the data.
6. According to the method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall as described in claim 4, It is characterized in that The details of the Attention layer are as follows: the attention mechanism is applied to the output layer of the BiGRU layer. The attention mechanism enables the model to obtain key information in the data by calculating the weight of each time step.
7. According to claim 4, a method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall, It is characterized in that Bayesian optimization BO includes the following steps: S1 initialization: select the initial sample point set; S2 builds a probability model: builds a Gaussian regression probability model based on the existing sample points; S3 selects the next sample point: uses the probability model to make inferences and selects the next sample point based on the inference results; S4 evaluates the objective function MAE at the selected sample points and obtains new observations; S5 updates the probability model: adds new sample points and observations to the existing sample point set and updates the probability model; S6 checks the termination condition: determines whether to terminate the algorithm based on whether the preset objective function converges and the maximum number of iterations.
8. According to claim 4, a method for accurately predicting the thrust of the cutter head when a shield machine passes through a reinforced concrete ground-connected wall, It is characterized in that The output layer is as follows: After the data is denormalized, the predicted total thrust value of the shield is output, and the Loss function is used to evaluate the model effect; the formula is as follows: The denormalization formula is as follows: ； The formula for calculating the Loss value is as follows: 。 9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the computer program is loaded into the processor, it realizes the method for accurately predicting the wall cutting thrust of the cutter head when the shield machine passes through the reinforced concrete ground-connected wall according to any one of claims 1 to 8.
10. A storage medium storing a computer program. It is characterized in that When the computer program is executed by a processor, a method for accurately predicting the wall cutting thrust of a cutter head when a shield machine passes through a reinforced concrete ground-connected wall according to any one of claims 1 to 8 is implemented.