CN114036696B - Cutterhead torque prediction method and system based on neural network model fine adjustment - Google Patents
Cutterhead torque prediction method and system based on neural network model fine adjustment Download PDFInfo
- Publication number
- CN114036696B CN114036696B CN202111451855.2A CN202111451855A CN114036696B CN 114036696 B CN114036696 B CN 114036696B CN 202111451855 A CN202111451855 A CN 202111451855A CN 114036696 B CN114036696 B CN 114036696B
- Authority
- CN
- China
- Prior art keywords
- ring segment
- data set
- neural network
- tunneled
- prediction model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003062 neural network model Methods 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000005641 tunneling Effects 0.000 claims abstract description 105
- 238000007781 pre-processing Methods 0.000 claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 230000008014 freezing Effects 0.000 claims abstract description 7
- 238000007710 freezing Methods 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 44
- 238000012360 testing method Methods 0.000 claims description 35
- 239000002689 soil Substances 0.000 claims description 10
- KJONHKAYOJNZEC-UHFFFAOYSA-N nitrazepam Chemical compound C12=CC([N+](=O)[O-])=CC=C2NC(=O)CN=C1C1=CC=CC=C1 KJONHKAYOJNZEC-UHFFFAOYSA-N 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 8
- 238000010219 correlation analysis Methods 0.000 claims description 7
- 230000003213 activating effect Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 238000010220 Pearson correlation analysis Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Hardware Design (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to a cutterhead torque prediction method and a cutterhead torque prediction system based on neural network model fine tuning, wherein the method comprises the following steps: establishing an original data set of the tunneled ring segment according to the equipment operation state parameters; determining a cutter torque prediction model of the tunneled ring segment according to the original data set and a preset neural network model; determining an original data set of a new tunneling ring section according to the equipment operation state parameters, and determining a data set of the new tunneling ring section after pretreatment according to the original data set of the new tunneling ring section; and fine-tuning the neural network prediction model of the cutterhead torque of the tunneling ring segment according to the data set after the preprocessing of the new tunneling ring segment. According to the neural network model fine tuning-based method, the generated cutter torque prediction model is finely tuned on the data which is recorded in real time and is preprocessed by the new tunneling ring segment in a form of freezing part of parameters of the existing model, so that the problem that the cutter torque prediction model on the new tunneling ring segment is reduced in performance can be effectively avoided.
Description
Technical Field
The invention relates to the technical field of shield construction, in particular to a cutterhead torque prediction method and system based on neural network model fine tuning.
Background
The shield is heavy huge-load equipment widely applied to underground tunnel construction. The control of the cutter torque of the shield tunneling machine is critical to tunneling efficiency, safety of constructors and stability of ground environment, however, the control of the existing cutter torque mainly depends on the control experience of a main driver, and the phenomenon of unreasonable torque setting exists. When the torque of the cutterhead is set improperly, serious engineering accidents are possibly caused. The torque of the shield cutterhead is predicted in real time in the tunneling process, and important references can be provided for the control of a main driver.
At present, a great deal of research work is done by a plurality of domestic and foreign scholars and specialists in the aspect of predicting the torque of the shield cutterhead. The geology is complicated and changeable, and the equipment running state is changeable, so that the real-time recorded airborne parameters often do not meet the independent and same-distribution conditions, and the predictability of a cutter torque prediction model established on the tunneled ring segment on the subsequent ring segment often decreases.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a cutterhead torque prediction method and system based on neural network model fine tuning.
In order to achieve the above object, the present invention provides the following solutions:
A cutterhead torque prediction method based on neural network model fine tuning comprises the following steps:
Establishing an original data set of the tunneled ring segment according to equipment operation state parameters recorded by an airborne system;
Determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model;
Determining an original data set of a new tunneling ring segment according to the equipment operation state parameters, and preprocessing the original data set of the new tunneling ring segment;
And fine tuning a neural network prediction model of the cutterhead torque of the tunneling ring segment according to the data set of the new tunneling ring segment after pretreatment.
Preferably, the determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model includes:
Carrying out data preprocessing on the original data set of the tunneling ring segment to obtain a preprocessed data set;
carrying out Pelson correlation analysis on cutter torque and airborne parameters on the preprocessed data set to obtain input characteristics;
Determining a training set and a testing set according to the preprocessed data set and the input features;
training and testing the preset neural network model according to the training set and the testing set to obtain a cutter torque prediction model of the tunneled ring segment.
Preferably, the data preprocessing is performed on the original data set of the tunneled ring segment to obtain a preprocessed data set, including:
removing non-tunneling data in the original data set of the tunneling ring segment to obtain a removed data set;
and carrying out normalization processing on the removed data set to obtain the preprocessed data set.
Preferably, the pearson correlation analysis of the cutter torque and the on-board parameter is performed on the preprocessed data set, so as to obtain an input feature, which includes:
Calculating the pearson correlation coefficient of the data in the preprocessed data set;
sequencing the pearson correlation coefficients of the data from large to small, and selecting the first eight parameters with the maximum pearson correlation coefficients as the input characteristics; the input characteristics comprise cutter head rotation speed, high-voltage transformer secondary current, screw machine pressure, screw machine cutter head torque, propulsion pressure, average soil pressure, soil discharge amount and screw machine rotation speed.
Preferably, the training and testing the preset neural network model according to the training set and the testing set to obtain a cutterhead torque prediction model of the tunneled ring segment, including:
Determining a first prediction model by using the preset neural network model based on the training set;
testing the first prediction model according to the test set to obtain a tested cutter torque prediction model of the tunneled ring segment; the input of the cutter torque prediction model of the tunneled ring segment is the input characteristic; and outputting the cutter torque prediction model of the tunneled ring segment as the cutter torque.
Preferably, the preset neural network model is a four-layer neural network based on an error back propagation algorithm.
Preferably, the step of the error back propagation algorithm comprises:
when the cutter torque prediction model of the tunneled ring segment is established, the weight coefficient of the neural network model is randomly initialized, and then the neural network model is trained according to the training set to obtain the weight coefficient of the cutter torque prediction model of the tunneled ring segment; when a load prediction model of a new tunneling ring segment is generated, loading a cutterhead torque prediction model of the tunneling ring segment and a weight coefficient of the cutterhead torque prediction model of the tunneling ring segment, which is completed by training the tunneling ring segment, on a data set of the new tunneling ring segment; the weight coefficient is a parameter to be finely adjusted in the cutter torque prediction model;
the input data in the training set of the cutter torque prediction model of the tunneled ring segment is transmitted forwards along the cutter torque prediction model of the tunneled ring segment, so that a predicted value is obtained;
calculating an error value between the predicted value and the true value by using a mean square error function;
And updating the weight coefficient by using a gradient descent algorithm back propagation until the error value converges.
Preferably, the fine tuning of the neural network prediction model of the cutterhead torque of the tunneled ring segment is performed according to the data set after the pretreatment of the new tunneled ring segment, including:
freezing weight parameters of a first layer network and a second layer network of the cutterhead torque prediction model of the tunneled ring segment based on a neural network model fine tuning method;
activating weight coefficients of a third layer network and a fourth layer network of the tunneling ring segment cutter torque prediction model;
and fine-tuning the weight coefficients of the third layer network and the fourth layer network by combining the preprocessed data set so as to generate a cutter torque prediction model applicable to the new tunneling ring segment.
A cutterhead torque prediction system based on neural network model fine tuning, comprising:
the data acquisition module is used for establishing an original data set of the tunneled ring segment according to equipment running state parameters recorded by the airborne system;
The model building module is used for determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model;
The sample data determining module is used for determining an original data set of a new tunneling ring segment according to the equipment running state parameters and determining a data set of the new tunneling ring segment after pretreatment according to the original data set of the new tunneling ring segment;
and the fine tuning module is used for fine tuning the neural network prediction model of the cutterhead torque of the tunneling ring segment according to the data set after the pretreatment of the new tunneling ring segment.
Preferably, the model building module specifically includes:
the preprocessing unit is used for preprocessing data of the original data set of the tunneled ring segment to obtain a preprocessed data set;
The analysis unit is used for carrying out Pelson correlation analysis on cutter torque and airborne parameters on the preprocessed data set to obtain input characteristics;
A data set determining unit for determining a training set and a testing set according to the preprocessed data set and the input features;
the model construction unit is used for training and testing the preset neural network model according to the training set and the testing set to obtain the cutter torque prediction model of the tunneled ring segment.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
The invention provides a cutterhead torque prediction method and a cutterhead torque prediction system based on neural network model fine tuning, wherein the method comprises the following steps: establishing an original data set of the tunneled ring segment according to equipment operation state parameters recorded by an airborne system; determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model; determining an original data set of a new tunneling ring segment according to the equipment operation state parameters, and determining a data set of the new tunneling ring segment after pretreatment according to the original data set of the new tunneling ring segment; and fine tuning the cutterhead torque neural network prediction model of the tunneling ring segment according to the data set of the new tunneling ring segment after pretreatment. According to the neural network model fine tuning-based method, the generated cutter torque prediction model is finely tuned on the data which is recorded in real time and is preprocessed by the new tunneling ring segment in a form of freezing part of parameters of the existing model, so that the problem that the cutter torque prediction model on the new tunneling ring segment is reduced in performance can be effectively avoided.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a method for predicting cutter torque in an embodiment of the present invention;
FIG. 2 is a schematic diagram of implementation steps in an embodiment provided by the present invention;
fig. 3 is a block diagram of a cutterhead torque prediction system according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
The terms "first," "second," "third," and "fourth" and the like in the description and in the claims and drawings are used for distinguishing between different objects and not necessarily for describing a particular sequential or chronological order. Furthermore, the terms "comprise" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, inclusion of a list of steps, processes, methods, etc. is not limited to the listed steps but may alternatively include steps not listed or may alternatively include other steps inherent to such processes, methods, products, or apparatus.
The invention aims to provide a cutterhead torque prediction method and a cutterhead torque prediction system based on neural network model fine adjustment, which can effectively avoid the problem that the performance of a cutterhead torque prediction model on a new tunneling ring segment is reduced.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Fig. 1 is a flowchart of a method in an embodiment provided by the present invention, and as shown in fig. 1, the present invention provides a cutterhead torque prediction method based on neural network model fine tuning, including:
step 100: establishing an original data set of the tunneled ring segment according to equipment operation state parameters recorded by an airborne system;
Step 200: determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model;
Step 300: determining an original data set of a new tunneling ring segment according to the equipment operation state parameters, and determining a data set of the new tunneling ring segment after pretreatment according to the original data set of the new tunneling ring segment;
Step 400: and fine tuning a neural network prediction model of the cutterhead torque of the tunneling ring segment according to the data set of the new tunneling ring segment after pretreatment.
Fig. 2 is a schematic diagram of implementation steps in the embodiment provided by the present invention, as shown in fig. 2, in this embodiment, the prediction method is implemented through seven steps, specifically:
S1: establishing an original data set of the tunneled ring segment according to equipment operation state parameters recorded by an airborne system;
S2: preprocessing the established tunneling ring segment original data set, removing non-tunneling data, and carrying out normalization processing;
s3: carrying out Pelson correlation analysis on cutter torque and other airborne parameters on the preprocessed data set in the step S2, sequencing the data set according to a mode that correlation coefficients are from large to small, and taking the first 8 features as input features of a model;
S4: acquiring a data set which is preprocessed by the tunneling ring segment according to the steps S1, S2 and S3, and dividing the data set into a training set and a testing set;
S5: based on a training set, a neural network model is utilized to establish a cutter torque pre-training model of the tunneled ring segment, and a test set is input to test a cutter torque prediction model of the tunneled ring segment;
s6, executing the step S1 and the step S2, and establishing a data set after preprocessing the new tunneling ring segment;
And S7, freezing weight parameters of the first two layers of networks of the cutterhead torque prediction model of the tunneled ring segment by utilizing a neural network model fine tuning method, activating the second two layers of networks in the load prediction model of the tunneled ring segment in the step S5, and executing the step S5 by combining the preprocessed data set to generate the cutterhead torque prediction model of the new tunneled ring segment.
Preferably, the determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model includes:
Carrying out data preprocessing on the original data set of the tunneling ring segment to obtain a preprocessed data set;
carrying out Pelson correlation analysis on cutter torque and airborne parameters on the preprocessed data set to obtain input characteristics;
Determining a training set and a testing set according to the preprocessed data set and the input features;
training and testing the preset neural network model according to the training set and the testing set to obtain a cutter torque prediction model of the tunneled ring segment.
Preferably, the data preprocessing is performed on the original data set of the tunneled ring segment to obtain a preprocessed data set, including:
removing non-tunneling data in the original data set of the tunneling ring segment to obtain a removed data set;
and carrying out normalization processing on the removed data set to obtain the preprocessed data set.
Specifically, in step S2, the data preprocessing includes the following steps:
1. identifying tunneling data and non-tunneling data, and deleting the non-tunneling data;
2. Mapping the data to between 0 and 1 based on an extremum normalization method.
Optionally, the data preprocessing includes, but is not limited to, the steps of:
1. The data recorded by the shield machine-mounted system simultaneously comprise data and non-tunneling data in the tunneling process of the equipment, the non-tunneling data are identified according to the following discrimination formula, and then the non-tunneling data are removed;
P＝h(T)·h(v)·h(w)
wherein T represents cutter torque, v represents tunneling speed, and w represents cutter rotational speed;
2. In order to eliminate the difference between the magnitudes of the measured data, carrying out extremum normalization on all data according to the following formula, and mapping the data recorded by an airborne system to between 0 and 1;
Wherein x pre is a dimensionless form after normalization processing, x is original data before normalization, x min is a minimum value in the parameter record data, and x max is a maximum value in the parameter record data.
Preferably, the pearson correlation analysis of the cutter torque and the on-board parameter is performed on the preprocessed data set, so as to obtain an input feature, which includes:
calculating a Pelson correlation coefficient between the cutter torque and the airborne parameters in the preprocessed data set;
sequencing the obtained pearson correlation coefficients from large to small, and selecting the first eight parameters with the largest pearson correlation coefficients as the input characteristics; the input characteristics comprise cutter head rotation speed, high-voltage transformer secondary current, screw machine pressure, screw machine cutter head torque, propulsion pressure, average soil pressure, soil discharge amount and screw machine rotation speed.
Specifically, the step S3 specifically includes:
And analyzing the Person correlation coefficient of cutter torque and other airborne parameters of the established preprocessed tunneling ring segment data set according to the following formula, and selecting the first 8 features with higher Person correlation coefficient as input parameters, wherein the parameters are the cutter rotational speed, the secondary current of the high-voltage transformer, the screw machine pressure, the screw machine cutter torque, the propelling pressure, the average soil pressure, the soil discharge amount and the screw machine rotation speed. The output parameter is cutter torque.
In the method, in the process of the invention,Is the mean value of the variable X,/>The linear correlation degree between the variable X and the variable Y can be calculated by the above equation as the average value of the variable Y.
Preferably, the training and testing the preset neural network model according to the training set and the testing set to obtain a cutterhead torque prediction model of the tunneled ring segment, including:
Determining a first prediction model by using the preset neural network model based on the training set;
Testing the first prediction model according to the test set to obtain a tested cutter torque prediction model of the tunneled ring segment; the input of the cutter torque prediction model of the tunneled ring segment is the input characteristic; and outputting the cutter torque prediction model of the tunneled ring segment as the cutter torque. The preset neural network model is a four-layer neural network based on an error back propagation algorithm.
Specifically, step S4 specifically includes: and (3) establishing a preprocessed tunneling ring segment data set according to the preprocessed parameters in the step (S2) and the input characteristics determined in the step (S3), and dividing the preprocessed tunneling ring segment data set into a training set and a testing set according to the proportion of 8:2.
Preferably, the step of the error back propagation algorithm comprises:
when the cutter torque prediction model of the tunneled ring segment is established, the weight coefficient of the neural network model is randomly initialized, and then the neural network model is trained according to the training set to obtain the weight coefficient of the cutter torque prediction model of the tunneled ring segment; when a load prediction model of a new tunneling ring segment is generated, loading a cutterhead torque prediction model of the tunneling ring segment and a weight coefficient of the cutterhead torque prediction model of the tunneling ring segment, which is completed by training the tunneling ring segment, on a data set of the new tunneling ring segment; the weight coefficient is a parameter to be finely adjusted in the cutter torque prediction model;
the input data in the training set of the cutter torque prediction model of the tunneled ring segment is transmitted forwards along the cutter torque prediction model of the tunneled ring segment, so that a predicted value is obtained;
calculating an error value between the predicted value and the true value by using a mean square error function;
And updating the weight coefficient by using a gradient descent algorithm back propagation until the error value converges.
Optionally, the error back propagation algorithm includes the following steps:
1. Randomly initializing a weight coefficient theta of a cutter torque prediction model on a tunneling ring segment data set when the cutter torque prediction model of the tunneling ring segment is established; when a load prediction model of a new tunneling ring segment is generated, loading a network model which is trained by the tunneling ring segment and a weight coefficient theta thereof on a new tunneling ring segment data set, wherein the weight coefficient theta is a parameter to be finely adjusted in a cutter torque prediction model;
2. the input data x in the training set is propagated forward along the neural network model to obtain a predicted value F function is forward transfer function of cutter torque prediction model;
3. Calculating a predicted value using a mean square error function MSE An error from the true value y;
the mean square error function MSE is calculated as:
Where m represents the sample size of the training set, y i represents the true value of the ith sample of the training set, A predicted value representing an ith sample of the training set;
4. The weight coefficient in the step a is updated by back propagation through a gradient descent algorithm until the error value converges;
the calculation formula of the gradient descent algorithm is as follows:
In the following, the following steps are taken: =update the left value with the right value of the symbol, α denotes the learning rate, J (θ) is the loss function, i.e. the mean square error function MSE.
Specifically, step S5 specifically includes: establishing a cutter torque prediction model of the tunneled ring segment: based on the training set, utilizing a neural network model to establish a cutter torque prediction model of the tunneled ring segment, and inputting a test set to test the cutter torque prediction model of the tunneled ring segment; the neural network model is a 4-layer neural network model based on a back propagation algorithm, the input of the model is 8 characteristic parameters determined in the step S4, the 8 characteristic parameters comprise cutter head rotation speed, high-voltage transformer secondary current, screw machine pressure, screw machine cutter head torque, propulsion pressure, average soil pressure, soil discharge amount and screw machine rotation speed, and the output of the model is cutter head torque.
Optionally, step S6 specifically includes: and executing the steps S1-S2, and establishing a data set after preprocessing the new tunneling ring segment.
Preferably, the fine tuning of the neural network prediction model of the cutterhead torque of the tunneled ring segment according to the data set after the pretreatment of the new tunneled ring segment includes:
freezing weight parameters of a first layer network and a second layer network of the cutterhead torque prediction model of the tunneled ring segment based on a neural network model fine tuning method;
activating weight coefficients of a third layer network and a fourth layer network of the tunneling ring segment cutter torque prediction model;
and fine-tuning the weight coefficients of the third layer network and the fourth layer network by combining the preprocessed data set so as to generate a cutter torque prediction model applicable to the new tunneling ring segment.
Further, step S7 specifically includes: and (3) freezing the front 2-layer weight parameters of the cutter torque prediction model of the tunneled ring segment by using a neural network model fine tuning method, activating the weight coefficient of the rear 2-layer network of the cutter torque prediction model of the tunneled ring segment in the step (S5), and executing the weight coefficient of the rear 2-layer network after fine tuning in the step (S5) by combining the preprocessed data set so as to generate the cutter torque prediction model applicable to the new tunneled ring segment.
The fine tuning mode of the network parameters of the rear 2 layers is determined according to the performance of a cutter torque prediction model obtained by fine tuning on the preprocessed new tunneling ring segment data set on the new tunneling ring segment data set, namely, the Root Mean Square Error (RMSE) and the decision coefficient (R 2) on a training set and a testing set of the new tunneling ring segment, and the fine tuning mode which enables the model to perform optimally is selected, wherein the optimal mode means that the RMSE is relatively smaller, and the R 2 is relatively higher;
the calculation formula of the root mean square error RMSE is as follows:
The calculation formula of the decision coefficient R 2 is as follows:
In the method, in the process of the invention, Is the average of y for all true values.
Fig. 3 is a block connection diagram of a cutterhead torque prediction system in an embodiment provided by the present invention, as shown in fig. 3, and this embodiment further provides a cutterhead torque prediction system based on neural network model fine tuning, including:
the data acquisition module is used for establishing an original data set of the tunneled ring segment according to equipment running state parameters recorded by the airborne system;
The model building module is used for determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model;
The sample data determining module is used for determining an original data set of a new tunneling ring segment according to the equipment running state parameters and determining a data set of the new tunneling ring segment after pretreatment according to the original data set of the new tunneling ring segment;
and the fine tuning module is used for fine tuning the neural network prediction model of the cutterhead torque of the tunneling ring segment according to the data set after the pretreatment of the new tunneling ring segment.
Preferably, the model building module specifically includes:
the preprocessing unit is used for preprocessing data of the original data set of the tunneled ring segment to obtain a preprocessed data set;
The analysis unit is used for carrying out Pelson correlation analysis on cutter torque and airborne parameters on the preprocessed data set to obtain input characteristics;
A data set determining unit for determining a training set and a testing set according to the preprocessed data set and the input features;
the model construction unit is used for training and testing the preset neural network model according to the training set and the testing set to obtain the cutter torque prediction model of the tunneled ring segment.
The beneficial effects of the invention are as follows:
According to the invention, the data established in real time by the shield machine-mounted system is preprocessed, the tunneling ring segment data set is established, the cutter head torque prediction model established in the existing tunneling ring segment is applied to the new tunneling ring segment after being finely tuned by a neural network model fine tuning method, the self-adaptive updating of the model in the tunneling process is realized, the problem of the performance degradation of the model when the working condition changes is effectively solved, and the prediction precision is improved.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (6)
1. A cutterhead torque prediction method based on neural network model fine tuning is characterized by comprising the following steps:
Establishing an original data set of the tunneled ring segment according to equipment operation state parameters recorded by an airborne system;
determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model, wherein the cutter torque prediction model comprises the following specific steps of:
Carrying out data preprocessing on the original data set of the tunneling ring segment to obtain a preprocessed data set;
carrying out Pelson correlation analysis on cutter torque and airborne parameters on the preprocessed data set to obtain input characteristics;
Determining a training set and a testing set according to the preprocessed data set and the input features;
Training and testing the preset neural network model according to the training set and the testing set to obtain a cutter torque prediction model of the tunneled ring segment, wherein the cutter torque prediction model specifically comprises the following steps:
Determining a first prediction model by using the preset neural network model based on the training set;
Testing the first prediction model according to the test set to obtain a tested cutter torque prediction model of the tunneled ring segment; the input of the cutter torque prediction model of the tunneled ring segment is the input characteristic; the output of the cutter torque prediction model of the tunneled ring segment is the cutter torque;
Determining an original data set of a new tunneling ring segment according to the equipment operation state parameters, and determining a data set of the new tunneling ring segment after pretreatment according to the original data set of the new tunneling ring segment;
Finely adjusting a neural network prediction model of the cutterhead torque of the tunneling ring segment according to the data set of the tunneling ring segment after pretreatment;
the preset neural network model is a four-layer neural network based on an error back propagation algorithm; the error back propagation algorithm comprises the following steps:
when the cutter torque prediction model of the tunneled ring segment is established, the weight coefficient of the neural network model is randomly initialized, and then the neural network model is trained according to the training set to obtain the weight coefficient of the cutter torque prediction model of the tunneled ring segment; when a load prediction model of a new tunneling ring segment is generated, loading a cutterhead torque prediction model of the tunneling ring segment and a weight coefficient of the cutterhead torque prediction model of the tunneling ring segment, which is completed by training the tunneling ring segment, on a data set of the new tunneling ring segment; the weight coefficient is a parameter to be finely adjusted in the cutter torque prediction model;
the input data in the training set of the cutter torque prediction model of the tunneled ring segment is transmitted forwards along the cutter torque prediction model of the tunneled ring segment, so that a predicted value is obtained;
calculating an error value between the predicted value and the true value by using a mean square error function;
And updating the weight coefficient by using a gradient descent algorithm back propagation until the error value converges.
2. The cutterhead torque prediction method based on neural network model fine tuning of claim 1, wherein the performing data preprocessing on the original data set of the tunneled ring segment to obtain a preprocessed data set comprises:
removing non-tunneling data in the original data set of the tunneling ring segment to obtain a removed data set;
and carrying out normalization processing on the removed data set to obtain the preprocessed data set.
3. The cutterhead torque prediction method based on neural network model fine tuning of claim 1, wherein the performing pearson correlation analysis of cutterhead torque and on-board parameters on the preprocessed data set to obtain input features comprises:
Calculating the pearson correlation coefficient of the data in the preprocessed data set;
sequencing the pearson correlation coefficients of the data from large to small, and selecting the first eight parameters with the maximum pearson correlation coefficients as the input characteristics; the input characteristics comprise cutter head rotation speed, high-voltage transformer secondary current, screw machine pressure, screw machine cutter head torque, propulsion pressure, average soil pressure, soil discharge amount and screw machine rotation speed.
4. The neural network model fine-tuning based cutterhead torque prediction method according to claim 1, wherein the fine-tuning of the neural network prediction model of cutterhead torque of the tunneled ring segment according to the data set after the pretreatment of the new tunneled ring segment comprises:
freezing weight parameters of a first layer network and a second layer network of the cutterhead torque prediction model of the tunneled ring segment based on a neural network model fine tuning method;
activating weight coefficients of a third layer network and a fourth layer network of the tunneling ring segment cutter torque prediction model;
and fine-tuning the weight coefficients of the third layer network and the fourth layer network by combining the preprocessed data set so as to generate a cutter torque prediction model applicable to the new tunneling ring segment.
5. A neural network model fine-tuning based cutterhead torque prediction system based on the neural network model fine-tuning based cutterhead torque prediction method as set forth in any one of claims 1 to 4, comprising:
the data acquisition module is used for establishing an original data set of the tunneled ring segment according to equipment running state parameters recorded by the airborne system;
The model building module is used for determining a cutter torque prediction model of the tunneled ring segment according to the original data set of the tunneled ring segment and a preset neural network model;
The sample data determining module is used for determining an original data set of a new tunneling ring segment according to the equipment running state parameters and determining a data set of the new tunneling ring segment after pretreatment according to the original data set of the new tunneling ring segment;
and the fine tuning module is used for fine tuning the neural network prediction model of the cutterhead torque of the tunneling ring segment according to the data set after the pretreatment of the new tunneling ring segment.
6. The cutterhead torque prediction system based on neural network model fine tuning of claim 5, wherein the model building module specifically comprises:
the preprocessing unit is used for preprocessing data of the original data set of the tunneled ring segment to obtain a preprocessed data set;
The analysis unit is used for carrying out Pelson correlation analysis on cutter torque and airborne parameters on the preprocessed data set to obtain input characteristics;
A data set determining unit for determining a training set and a testing set according to the preprocessed data set and the input features;
the model construction unit is used for training and testing the preset neural network model according to the training set and the testing set to obtain the cutter torque prediction model of the tunneled ring segment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111451855.2A CN114036696B (en) | 2021-12-01 | 2021-12-01 | Cutterhead torque prediction method and system based on neural network model fine adjustment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111451855.2A CN114036696B (en) | 2021-12-01 | 2021-12-01 | Cutterhead torque prediction method and system based on neural network model fine adjustment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114036696A CN114036696A (en) | 2022-02-11 |
| CN114036696B true CN114036696B (en) | 2024-05-17 |
Family
ID=80139451
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111451855.2A Active CN114036696B (en) | 2021-12-01 | 2021-12-01 | Cutterhead torque prediction method and system based on neural network model fine adjustment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114036696B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI825980B (en) * | 2022-09-07 | 2023-12-11 | 英業達股份有限公司 | Setting method of in-memory computing simulator |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329172A (en) * | 2020-11-05 | 2021-02-05 | 上海交通大学 | Shield tunneling machine cutter head torque prediction method and system based on parallel neural network |
| CN112347580A (en) * | 2020-11-05 | 2021-02-09 | 上海交通大学 | Shield tunneling machine cutter head torque real-time prediction method and system |
| WO2021068454A1 (en) * | 2019-10-12 | 2021-04-15 | 联合微电子中心有限责任公司 | Method for identifying energy of micro-energy device on basis of bp neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11085295B2 (en) * | 2019-01-24 | 2021-08-10 | Huaneng Tibet Yarlungzangbo River Hydropower Development Investment Co., Ltd. | Tunnel boring robot and remote mobile terminal command system |
-
2021
- 2021-12-01 CN CN202111451855.2A patent/CN114036696B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021068454A1 (en) * | 2019-10-12 | 2021-04-15 | 联合微电子中心有限责任公司 | Method for identifying energy of micro-energy device on basis of bp neural network |
| CN112329172A (en) * | 2020-11-05 | 2021-02-05 | 上海交通大学 | Shield tunneling machine cutter head torque prediction method and system based on parallel neural network |
| CN112347580A (en) * | 2020-11-05 | 2021-02-09 | 上海交通大学 | Shield tunneling machine cutter head torque real-time prediction method and system |
Non-Patent Citations (2)
| Title |
|---|
| TBM装备掘进载荷的力学分析与功率建模;成志飞;王立辉;苏翠侠;周思阳;张茜;;机械科学与技术;20171215(第12期);全文 * |
| 基于IPSO-BP混合模型的TBM掘进参数预测;侯少康;刘耀儒;张凯;;岩石力学与工程学报;20201231(第08期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114036696A (en) | 2022-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110852423A (en) | Tunnel boring machine excavation performance and control parameter prediction method based on transfer learning | |
| CN109635461A (en) | A kind of application carrys out the method and system of automatic identification Grades of Surrounding Rock with brill parameter | |
| CN109086804B (en) | Hydraulic equipment early failure prediction method based on fusion of multi-source state monitoring information and reliability characteristics | |
| CN111079342B (en) | TBM tunneling performance prediction method based on online rock mass grade classification | |
| CN111913803B (en) | Service load fine granularity prediction method based on AKX hybrid model | |
| CN109297689B (en) | Large-scale hydraulic machinery intelligent diagnosis method introducing weight factors | |
| CN109918822A (en) | The shield driving parameter error calculation method and system returned based on random forest | |
| CN106181579A (en) | A kind of Tool Wear Monitoring method based on multisensor current signal | |
| CN114036696B (en) | Cutterhead torque prediction method and system based on neural network model fine adjustment | |
| CN114662699A (en) | Shield attitude prediction method based on machine learning | |
| CN110909802A (en) | Improved PSO (particle swarm optimization) based fault classification method for optimizing PNN (portable network) smoothing factor | |
| CN114972384A (en) | Tunnel rock intelligent rapid regional grading method based on deep learning | |
| CN105184079B (en) | A kind of rock drilling state identification method of hydraulic gate | |
| CN108388689B (en) | Shield cutter head torque calculation method based on dimensional analysis | |
| CN113567159B (en) | Scraper conveyor state monitoring and fault diagnosis method based on edge cloud cooperation | |
| CN113895272A (en) | Electric vehicle alternating current charging state monitoring and fault early warning method based on deep learning | |
| CN114810100B (en) | Shield tunneling attitude prediction method based on deep neural network | |
| CN112305379A (en) | Mode identification method and system for GIS insulation defect | |
| CN115452376A (en) | Bearing fault diagnosis method based on improved lightweight deep convolution neural network | |
| CN114897007A (en) | Drilling machine health condition evaluation method of composite information layered convolution neural network | |
| CN114236332A (en) | Power cable insulation state judgment method and system | |
| CN113627268A (en) | Model training method, and method and device for detecting fault of speed reducer for mine hoist | |
| CN111598366A (en) | Real-time drilling auxiliary decision-making method and system | |
| CN115544874A (en) | Real-time monitoring system, method, terminal and medium for shield hob abrasion loss | |
| CN112598061B (en) | Clustering and grading method for tunnel surrounding rock |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |