Shield cutter wear prediction method and system based on dynamic load and frictional work

Abstract

The invention discloses a shield cutter wear prediction method and system based on dynamic load and frictional work, and the method comprises the steps: obtaining parameters and operation parameters of a to-be-predicted shield cutter, inputting the parameters and operation parameters into a pre-trained wear prediction model, and obtaining the normal dynamic load, slippage and wear loss of the cutter, thereby achieving wear prediction; a training process of the wear prediction model comprises the following steps: step 1, based on different parameters of the shield cutter and shield operation parameters, adopting finite element simulation to obtain a normal dynamic load and a slip amount of the corresponding shield cutter, further calculating to obtain friction work in the shield operation, and adopting discrete element wear simulation to obtain a wear amount of the friction work to the cutter; 2, repeating the step 1 to obtain a plurality of groups of normal dynamic loads, slippage amounts and wear amounts corresponding to different shield cutter operation parameters, and constructing a training set after preprocessing; and 3, training a neural network by using the training set to obtain a wear prediction model meeting training requirements.

Description

Technical Field

[0001] The present invention belongs to the technical field of shield machines, and particularly relates to a method and system for predicting shield cutter wear based on dynamic load and friction work. Background Technique

[0002] A shield machine is a device used for tunnel under-excavation construction, which has a metal shell and is equipped with a complete machine and auxiliary equipment inside. The shield machine performs operations such as soil excavation, soil slag transportation, overall machine propulsion, and segment installation under the cover of the shell, so that the tunnel is formed at one time.

[0003] As an important device for tunnel excavation, the wear condition of the cutters of the shield machine is directly related to the excavation efficiency and engineering safety. In actual operation, it is usually determined whether to replace according to the wear amount of the cutters. Generally speaking, when the wear of the cutters reaches 1 / 3 to 1 / 2 of the original thickness, a new cutter head should be considered for replacement. Excessive wear will affect the cutting efficiency and stability of the cutters, and may even lead to engineering accidents. Therefore, accurately predicting the wear degree of the cutters and replacing the cutter head in time is crucial.

[0004] Currently, the main methods for predicting the wear of shield hob include the derivation of cutter wear prediction models, experimental methods, and numerical simulation methods.

[0005] Regarding the derivation of the prediction model for the cutters, from a macroscopic perspective, Dong-Jie Ren et al. derived a prediction formula for the hob life from the perspective of friction work based on the energy method. From a microscopic perspective, Yandong Yang et al. derived a prediction model for hob wear based on the abrasive wear model and combined with the CSM rock-breaking force prediction formula. However, the loads of the above prediction models come from empirical formulas, and the loads are static loads, while in reality they are dynamic loads. Moreover, the undetermined coefficients in the prediction model need to be determined according to the engineering actual situation, lacking the foresight of wear prediction and not meeting the requirements for optimizing the cutterhead design in advance.

[0006] In terms of numerical simulation, Zhang Jiaqi, Han Meidong et al. from Tianjin University performed numerical simulation on the shield cutterhead using abaqus finite element software. Although the dynamic load of the cutters can be calculated, only the revolution of the hob is considered and the rotation of the hob is not considered, so the slip amount of the hob cannot be calculated, and thus the friction work cannot be calculated. Dong Yong et al. from Huazhong University of Science and Technology performed numerical simulation on the wear of the cutters using discrete element software, but did not focus on the relationship between friction work and wear amount, and the discrete element numerical simulation has complex modeling and requires too much time.

[0007] In summary, the existing problems in the prediction of shield cutter wear include: 1. The prediction of the wear model is based on static load, while in reality it is dynamic load. 2. Lack of forward-looking in prediction. 3. Lack of the slip distance required for calculating the friction work and the factors affecting the slip distance. 4. At present, the time required for the wear prediction of the shield model is too long, and it is difficult to make a quick prediction for different situations. Summary of the Invention

[0008] The purpose of the present invention is to overcome the defects of the existing technology and propose a shield cutter wear prediction method and system based on dynamic load and friction work.

[0009] In view of this, the present invention proposes a shield cutter wear prediction method based on dynamic load and friction work, including:

[0010] Obtain the parameters of the shield cutter to be predicted and the operation parameters, input them into the pre-trained wear prediction model, and obtain the normal dynamic load, slip amount and cutter wear amount of the cutter, so as to realize wear prediction;

[0011] The training process of the wear prediction model includes:

[0012] Step 1: Based on the parameters of different shield cutters and the shield operation parameters, use finite element simulation to obtain the normal dynamic load and slip amount of the corresponding shield cutter, and then calculate the friction work in the shield operation. Use discrete element wear simulation to obtain the wear amount of the cutter caused by the friction work;

[0013] Step 2: Repeat Step 1 to obtain multiple groups of normal dynamic loads, slip amounts and wear amounts corresponding to different shield cutter operation parameters, and construct a training set after preprocessing;

[0014] Step 3: Use the training set to train the neural network to obtain a wear prediction model that meets the training requirements.

[0015] Preferably, the parameters of the shield cutter to be predicted include the hob radius, and the operation parameters include: rock and soil density, Young's modulus, internal friction angle, hob installation radius, cutter head rotation speed and tunneling speed.

[0016] Preferably, in Step 1, based on the parameters of different shield cutters and the shield operation parameters, using finite element simulation to obtain the normal dynamic load and slip amount of the corresponding shield cutter; including:

[0017] Model the shield cutter and the rock and soil respectively, perform mesh division, and then assemble them to form a numerical model;

[0018] Set the hob radius, rock and soil density, Young's modulus, internal friction angle, cutter installation radius, cutter head rotation speed and tunneling speed, and set the boundary conditions of the finite element simulation and the operation mode of the shield cutter;

[0019] Set the working modes of the revolution and rotation of the hob;

[0020] Obtain the normal dynamic load F according to finite element calculation n ;

[0021] Obtain the slip amount s of the shield cutter according to the following formula:

[0022]

[0023] wherein, R is the hob installation radius of the shield cutter, r is the hob radius, θ is the revolution angle of the hob, is the rotation angle.

[0024] Preferably, the normal dynamic load F n is time series data.

[0025] Preferably, in the step 1, the frictional work w in the shield operation satisfies the following formula:

[0026] w = F n ·s.

[0027] Preferably, in the step 1, discrete element wear simulation is adopted to obtain the wear amount of the cutter caused by the frictional work, including:

[0028] Generate a granular soil bed and import the hob model;

[0029] Set the bong bonds between the granular soils;

[0030] Complete the assembly between the granular soil bed and the hob model, and set the load and slip amount applied based on finite element;

[0031] Set the wear parameters between the granular soil and the hob material;

[0032] Export the wear amount of the hob.

[0033] Preferably, in the step 2, the preprocessing includes:

[0034] Data storage: Store each group of data appropriately according to the setting;

[0035] Data cleaning: Conduct data cleaning. For outliers beyond the reasonable range, use the interpolation method to correct or directly remove them, and perform filtering processing on the time series of the normal dynamic load; for missing values, use the mean filling or interpolation method to complete them;

[0036] Feature construction: Extract statistical features and time-frequency features from the time series of the normal dynamic load of the cutter, and perform normalization processing.

[0037] On the other hand, the present invention discloses a shield cutter wear prediction system based on dynamic load and frictional work, including:

[0038] A wear prediction module, which is used to obtain the parameters of the shield cutter to be predicted and the operation parameters, input them into a pre-trained wear prediction model, and obtain the normal dynamic load, slip amount and cutter wear amount of the cutter, so as to realize wear prediction;

[0039] The training process of the wear prediction model includes:

[0040] Step 1: Based on the parameters of different shield cutters and the shield operation parameters, use finite element simulation to obtain the normal dynamic load and slip amount of the corresponding shield cutter, and then calculate the friction work during shield operation. Use discrete element wear simulation to obtain the wear amount of the cutter caused by the friction work;

[0041] Step 2: Repeat Step 1 to obtain multiple groups of normal dynamic loads, slip amounts and wear amounts corresponding to different shield cutter operation parameters. After preprocessing, construct a training set;

[0042] Step 3: Use the training set to train a neural network to obtain a wear prediction model that meets the training requirements.

[0043] Compared with the prior art, the advantages of the present invention are as follows:

[0044] Compared with the traditional finite element method, the method of the present invention can calculate the slip amount of the hob breaking rock and the factors affecting the slip amount; compared with the traditional discrete element method, the method of the present invention is based on the dynamic load and slip amount calculated by the finite element method for discrete element numerical simulation, which can more directly reflect the relationship between the friction work and the wear amount, and has a simple model and less calculation time; compared with the traditional empirical formula derivation, the method of the present invention is based on numerical simulation and considers the dynamic load rather than the static load of the cutter, which is more in line with the actual situation. Description of the Drawings

[0045] Figure 1 is a structural diagram of a disc cutter, where Figure 1(a) is the cutter ring and Figure 1(b) is the cutter shaft;

[0046] Figure 2 is the time series diagram of the normal dynamic load of the hob;

[0047] Figure 3 is the EDEM model flow chart of the friction work and the wear amount;

[0048] Figure 4 is the schematic diagram of the numerical model after the assembly of the shield cutter model and the working rock soil model;

[0049] Figure 5 is the training flow chart of the wear prediction model of the shield cutter of the present invention;

[0050] Figure 6 is the schematic diagram of the interaction interface of the cutter wear prediction platform;

[0051] Figure 7 is the granular bed of rock and soil in EDEM;

[0052] Figure 8 is the bond in the granular bed of rock and soil;

[0053] Figure 9 is the assembly model of the hob and rock and soil in EDEM;

[0054] Figure 10 is the contour map of the wear amount of the hob. Specific implementation manner

[0055] The present invention provides a shield cutter wear prediction method based on dynamic load and friction work, including:

[0056] Obtain the parameters of the shield cutter to be predicted and the operation parameters, input them into a pre-trained wear prediction model, and obtain the normal dynamic load, slip amount and cutter wear amount of the cutter, so as to realize wear prediction;

[0057] The training process of the wear prediction model includes:

[0058] Step 1: Based on the parameters of different shield cutters and shield operation parameters, use finite element simulation to obtain the normal dynamic load and slip amount of the corresponding shield cutter, and then calculate the friction work during shield operation. Use discrete element wear simulation to obtain the wear amount of the cutter caused by the friction work;

[0059] Step 2: Repeat Step 1 to obtain multiple groups of normal dynamic loads, slip amounts and wear amounts corresponding to different shield cutter operation parameters. After preprocessing, construct a training set;

[0060] Step 3: Use the training set to train a neural network to obtain a wear prediction model that meets the training requirements.

[0061] The technical solution of the present invention will be described in detail below with reference to the drawings and embodiments.

[0062] Embodiment 1

[0063] Embodiment 1 of the present invention provides a shield cutter wear prediction method based on dynamic load and friction work. The specific process is as follows:

[0064] In order to obtain the dynamic load and slip amount required for calculating the friction work, abaqus finite element software is used for numerical simulation. The specific process is as follows:

[0065] 1. Model the shield hob and rock and soil respectively, perform mesh division, and assemble them.

[0066] 2. Set the material parameters of the rock and soil and the hob.

[0067] 3. Set the boundary conditions to fix the rock and soil. For the hob, set that the hob can revolve around the center of the cutterhead and can also rotate around the cutter axis. The structural diagrams of the disc hob are shown in Figures 1(a) and 1(b).

[0068] 4. Stpe1: The hob penetrates into the rock and reaches the specified penetration depth

[0069] Step2: The hob rolls on the rock, the hob rotates, and moves parallel to the rock surface. By adjusting different parameters: cutterhead rotation speed, cutterhead penetration depth, tool installation radius, Young's modulus of rock and soil, internal friction angle of rock and soil, density of rock and soil. Different dynamic loads of the tool for rock breaking and the sliding displacement of the tool can be obtained

[0070] Among them, the sliding displacement of the tool is calculated according to the following formula:

[0071]

[0072] In the formula, s is the sliding displacement, R is the installation radius of the hob, r is the radius of the hob, θ is the revolution angle of the hob, and φ is the rotation angle of the hob around the cutter axis.

[0073] The dynamic load of the tool is the load after considering the speed and acceleration, and it changes with time. The schematic diagram of the dynamic load is as Figure 2 shown.

[0074] After obtaining the sliding displacement and dynamic load of the tool, the friction work can be calculated according to the following formula.

[0075] w = F n ·s

[0076] Furthermore, the influence relationship between various parameters and friction work and dynamic load can be obtained. In the formula, w is the friction work, and F n is the normal dynamic load, and s is the sliding distance.

[0077] After obtaining the dynamic load and sliding distance of the hob for rock breaking, the friction work required for the hob to break rock can be obtained, and then the relationship between the friction work and the wear amount can be obtained.

[0078] The flow chart of the EDEM model for establishing the friction work and wear amount is as Figure 3 shown.

[0079] Through the above process, the relationships between various parameters and dynamic load, sliding displacement, and wear amount can be obtained. Then machine learning is used for prediction.

[0080] Use a data-driven neural network to predict tool wear. The data comes from the above process. The input parameters are: density of rock and soil, Young's modulus, internal friction angle, tool installation radius, cutterhead rotation speed, tunneling speed. The output parameters are the normal dynamic load of the tool, sliding displacement, and tool wear amount. AsFigure 4 It is a schematic diagram of the numerical model after the assembly of the shield tool model and the working geotechnical model; Figure 5 It is the training flow chart of the wear prediction model of the shield tool of the present invention.

[0081] The specific process is as follows:

[0082] 1. Diversify the simulation scenarios, design multiple groups of simulation scenarios, and change different parameters for each group;

[0083] 2. Data collection and storage

[0084] 2.1 Collect output data

[0085] Dynamic load time series: Store as high-resolution time series data;

[0086] Slip amount: Extract the total slip amount at each time step;

[0087] Wear amount: Wear volume.

[0088] 2.2 Store data

[0089] The data is stored in tabular form, where each row corresponds to a set of input parameters and the corresponding output data.

[0090] 2.3 Export time series

[0091] The dynamic load time series can be exported separately in CSV or HDF5 format for subsequent time series analysis.

[0092] 3. Data processing

[0093] 3.1 Data cleaning

[0094] 1) Process outliers

[0095] Check whether there are unreasonable output values, such as the abnormal increase or decrease of the normal dynamic load (beyond the theoretical calculation range).

[0096] Outlier processing method: Use interpolation method to correct or directly eliminate abnormal samples.

[0097] 2) Data smoothing

[0098] Perform filtering on the time series of the normal dynamic load (such as low-pass filtering or moving average) to reduce the influence of high-frequency noise.

[0099] 3) Missing value processing

[0100] If there is data loss during the simulation process, use mean filling or interpolation method to complete the missing data.

[0101] 3.2 Feature construction

[0102] 1) Direct features of input parameters

[0103] Retain the original input parameters as the input features of the model, including: rock and soil density, Young's modulus, internal friction angle, cutter installation radius, cutter head rotation speed, tunneling speed.

[0104] 2) Time series feature extraction

[0105] Extract the following features from the time series of the normal dynamic load of the cutter:

[0106] Statistical features: mean, maximum, minimum, standard deviation, peak interval time, etc.

[0107] Time-frequency features: Perform Fourier transform on the dynamic load to extract the main frequency components and energy distribution.

[0108] Feature standardization: Normalize all input features (including statistical features and time-frequency features).

[0109] 4 Dataset division

[0110] Division of training set, validation set and test set:

[0111] Training set (70%): Used for model training.

[0112] Validation set (15%): Used to adjust the hyperparameters of the model.

[0113] Test set (15%): Used to evaluate the model performance.

[0114] Platform construction:

[0115] Use MATLAB, relying on its rich GUI development tools and powerful data processing and machine learning functions, to build a cutter wear prediction platform.

[0116] The following is the detailed plan:

[0117] Overall function:

[0118] The user inputs 6 parameters: rock and soil density, Young's modulus, internal friction angle, cutter installation radius, cutter head rotation speed, tunneling speed.

[0119] After clicking the button "Calculate", call the pre-trained neural network model to predict the output: normal dynamic load (display the time series graph), slip amount, wear amount. As Figure 6 shown. As Figure 7 shown is the granular bed of rock and soil in EDEM; Figure 8 are the bond bonds in the granular bed of rock and soil; Figure 9 is the assembly model of the hob and rock and soil in EDEM; Figure 10 is the wear amount contour map of the hob.

[0120] Detailed steps:

[0121] 1. Data preparation

[0122] Pre-trained neural network model

[0123] Use the Deep Learning Toolbox in MATLAB to train a neural network model and save it as a MAT file (such as model.mat).

[0124] Model input: 6 parameters (rock and soil density, Young's modulus, etc.).

[0125] Model output: 3 prediction results (normal dynamic load time series, slip amount, wear amount).

[0126] Prepare a set of test data to ensure the accuracy of the model prediction results for subsequent use.

[0127] 2. MATLAB platform design

[0128] 2.1 Create a GUI interface

[0129] Use the App Designer or GUIDE tool in MATLAB to develop an interactive interface.

[0130] Open App Designer:

[0131] Design the GUI interface and add the following controls in the designer:

[0132] Input area: 6 parameters;

[0133] EditField control for entering 6 parameters (rock and soil density, Young's modulus, etc.).

[0134] Attach a Label (tag) in front of each input box to describe the parameter name.

[0135] Buttons:

[0136] Button control named "Calculate" to trigger model prediction.

[0137] Button control named "Save Results" to save the output to a file.

[0138] Result display area:

[0139] Axes control for plotting the time series graph of the normal dynamic load.

[0140] EditField control for displaying the slip amount and wear amount.

[0141] Message prompt box:

[0142] A Label control for displaying prompt messages such as input parameter verification and prediction completion.

[0143] 2.2 Design the GUI layout

[0144] The left side is the input parameter area: there are 6 input boxes, with each parameter on a separate line.

[0145] The right side is the output display area: a time series graph of the normal dynamic load is shown at the top.

[0146] The numerical values of the slip amount and wear amount are shown at the bottom.

[0147] Example 2

[0148] Example 2 of the present invention discloses a shield cutter wear prediction system based on dynamic load and friction work, including:

[0149] A wear prediction module for obtaining the parameters of the shield cutter to be predicted and the operation parameters, inputting them into a pre-trained wear prediction model, and obtaining the normal dynamic load, slip amount, and cutter wear amount of the cutter, thereby realizing wear prediction;

[0150] The training process of the wear prediction model includes:

[0151] Step 1: Based on the parameters of different shield cutters and shield operation parameters, use finite element simulation to obtain the corresponding normal dynamic load and slip amount of the shield cutter, and then calculate the friction work during shield operation. Use discrete element wear simulation to obtain the wear amount of the cutter due to friction work;

[0152] Step 2: Repeat Step 1 to obtain multiple sets of normal dynamic loads, slip amounts, and wear amounts corresponding to different shield cutter operation parameters. After preprocessing, construct a training set;

[0153] Step 3: Use the training set to train a neural network to obtain a wear prediction model that meets the training requirements.

[0154] It should be noted that in the embodiments of the above system, the various modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional modules are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the embodiments, those of ordinary skill in the art should understand that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A shield cutter wear prediction method based on dynamic load and friction work, comprising: Obtaining the parameters and operation parameters of the shield cutter to be predicted, inputting them into a pre-trained wear prediction model, and obtaining the normal dynamic load, slip amount and cutter wear amount of the cutter, so as to realize wear prediction; The training process of the wear prediction model includes: Step 1: Based on the parameters of different shield cutters and shield operation parameters, finite element simulation is used to obtain the normal dynamic load and slip amount of the corresponding shield cutter, and then the friction work in shield operation is calculated. Discrete element wear simulation is used to obtain the wear amount of the cutter caused by the friction work; Step 2: Repeat Step 1 to obtain multiple groups of normal dynamic loads, slip amounts and wear amounts corresponding to different shield cutter operation parameters, and construct a training set after preprocessing; Step 3: Use the training set to train the neural network to obtain a wear prediction model that meets the training requirements.

2. The shield cutter wear prediction method based on dynamic load and friction work according to claim 1, wherein The parameters of the shield cutter to be predicted include the hob radius, and the operation parameters include: rock and soil density, Young's modulus, internal friction angle, hob installation radius, cutter head rotation speed and tunneling speed.

3. The shield cutter wear prediction method based on dynamic load and friction work according to claim 1, characterized in that In Step 1, based on the parameters of different shield cutters and shield operation parameters, finite element simulation is used to obtain the normal dynamic load and slip amount of the corresponding shield cutter; it includes: Modeling and meshing the shield cutter and the rock and soil respectively, and then assembling them to form a numerical model; Setting the hob radius, rock and soil density, Young's modulus, internal friction angle, cutter installation radius, cutter head rotation speed and tunneling speed, setting the boundary conditions of the finite element simulation and the operation mode of the shield cutter; Setting the working mode of the revolution and rotation of the hob; The normal dynamic load F is obtained according to the finite element calculation n ; Obtaining the slip amount s of the shield cutter according to the following formula: Among them, R is the installation radius of the hob of the shield cutter, r is the hob radius, and θ is the revolution angle of the hob. It is the rotation angle.

4. The shield cutter wear prediction method based on dynamic load and friction work according to claim 3, characterized in that The normal dynamic load F n is time series data.

5. The shield cutter wear prediction method based on dynamic load and friction work according to claim 3, characterized in that In Step 1, the friction work w in shield operation satisfies the following formula: w = F n ·s.

6. The shield cutter wear prediction method based on dynamic load and friction work according to claim 3, characterized in that In Step 1, discrete element wear simulation is used to obtain the wear amount of the cutter caused by the friction work, including: Generating a rock and soil particle bed and importing the hob model; Setting the bong bonds between the rock and soil particles; Completing the assembly between the rock and soil particle bed and the hob model, and setting the load and slip amount applied based on the finite element; Setting the wear parameters between the rock and soil and the hob materials; Exporting the wear amount of the hob.

7. The shield cutter wear prediction method based on dynamic load and friction work according to claim 1, wherein In Step 2, the preprocessing includes: Data storage: Storing each group of data appropriately according to the setting; Data cleaning: Conducting data cleaning, for outliers beyond the reasonable range, using the interpolation method to correct or directly eliminate them, and filtering the time series of the normal dynamic load; using the mean value filling or interpolation method to complete the missing values; Feature construction: Extracting statistical features and time-frequency features from the time series of the cutter normal dynamic load and performing normalization processing.

8. A shield cutter wear prediction system based on dynamic load and friction work, characterized in that, Including: A wear prediction module, which is used to obtain the parameters and operation parameters of the shield cutter to be predicted, input them into a pre-trained wear prediction model, and obtain the normal dynamic load, slip amount and cutter wear amount of the cutter, so as to realize wear prediction; The training process of the wear prediction model includes: Step 1: Based on the parameters of different shield cutters and shield operation parameters, finite element simulation is used to obtain the normal dynamic load and slip amount of the corresponding shield cutter, and then the friction work in shield operation is calculated. Discrete element wear simulation is used to obtain the wear amount of the cutter caused by the friction work; Step 2: Repeat Step 1 to obtain multiple sets of normal dynamic loads, sliding amounts, and wear amounts corresponding to different shield cutter operation parameters, and construct a training set after preprocessing; Step 3: Use the training set to train a neural network to obtain a wear prediction model that meets the training requirements.

Images