A method and system for optimizing cutter wear and energy consumption based on shield tunneling

Abstract

The application discloses a kind of cutterhead wear and energy consumption optimization method and system based on shield machine tunneling, method includes: the operating data of TBM is collected, and data preprocessing is carried out, wherein, operating data includes adjustable operating parameter;Based on the data after preprocessing, LightGBM is used to establish TBM efficiency prediction model, wherein, with Bayesian optimization to LightGBM is carried out hyperparameter optimization, with prediction error as loss function to evaluate TBM efficiency prediction model, until the TBM efficiency prediction model that meets preset requirement is obtained;With the minimization of cutterhead wear and cutterhead energy consumption as target, establish multi-objective optimization model based on NSGA-II, obtain Pareto frontier;Solve Pareto frontier, obtain the optimal adjustable operating parameter of TBM.The application can reduce tool wear and cutterhead energy consumption, improve the performance of tunnel boring machine, guide tunnel construction, promote the construction of tunnel engineering project.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of shield tunneling optimization, and more particularly, to a cutterhead wear and energy consumption optimization method and system based on shield tunneling. BACKGROUND

[0002] TBM has attracted widespread attention with its powerful ability of low construction cost, fast advancing speed and high safety, along with large-scale underground construction. However, due to the underground operation of TBM, buried in the daily line of sight, the geological conditions and construction process have great uncertainty and complexity, and still face many challenges. With the expansion of the city and the relief of traffic pressure, underground construction activities are increasing, and tunnel boring machines (TBM) have been widely used in tunnel excavation mechanization, replacing the traditional drill and blast. The key component of TBM is the disc cutter, which rotates continuously and inserts itself into the rock, while the broken rock soil is transported and moved forward through a series of equipment cooperation. That is, the TBM-based tunneling relies on the direct interaction between the cutterhead and the hard rock, and the cutting of hard rock is carried out by the thrust and torque generated by the cutterhead rolling. Generally, the cutterhead is overused without attention, which is easy to cause mechanical accidents, thereby affecting the tunneling progress. Therefore, the wear of the cutterhead determines the safety, thrust speed and low cost of the TBM to a great extent. However, the wear of the cutterhead is affected by many factors such as rock mass properties and installation position, which is a complex evolution process, and poses a challenge to measure and understand the wear of the cutterhead in tunnel construction.

[0003] In the field of tunnel engineering, on-site inspection based on expert knowledge and experience still plays a crucial role in TBM operation, but inevitably has subjectivity and randomness, and errors may induce potential risks. On-site inspection of cutterhead wear requires TBM to stop, which will affect the progress of the tunnel and increase the cost. As an alternative, the monitoring system is installed on the TBM, which can generate massive data records of key parameters of TBM operation in real time, providing a reference for safe construction of tunnel engineering. However, the data obtained is usually nonlinear and fluctuates greatly with noise, which brings challenges to cutterhead wear prediction and optimization. Therefore, it is necessary to develop a reliable method for tunnel engineering, data mining, performance prediction and optimization.

[0004] For cutterhead wear prediction, in recent years, it can be divided into three categories, namely empirical method, theoretical method and statistical method. The empirical method is based on full-size laboratory test to analyze the performance of TBM. The empirical method is based on the field observation of TBM performance and geological conditions to evaluate the performance. The statistical method mainly relies on mathematical rules to construct performance prediction method. These methods help to predict the performance of TBM and enrich the knowledge base of the field. However, the empirical method needs to design and conduct multiple experiments, which is costly and time-consuming, and requires given test conditions and some physical assumptions. In this case, the results may only be applicable to specific situations. Theoretical methods are often suitable for simple scenarios, with many assumptions that may not be suitable for actual situations, and the reliability of the results is questionable. The robustness of statistical methods for nonlinear and complex systems is inconsistent, and their predictive ability may be weakened by outliers and extreme values. That is, these methods have their shortcomings and may face challenges in practical applications.

[0005] Due to the limitations of assumptions and method capabilities, empirical, theoretical and statistical methods cannot provide accurate and reliable predictions. On the contrary, machine learning methods can provide reliable predictions by modeling patterns through data mining, but existing research has focused little on cutterhead wear prediction.

[0006] Therefore, the present application aims to develop a machine learning method to accurately predict and optimize cutterhead wear and cutterhead energy consumption. SUMMARY

[0007] In view of the above defects or improvement needs of the prior art, the present application provides a cutterhead wear and energy consumption optimization method and system based on shield tunneling, which accurately predicts and optimizes cutterhead wear and cutterhead energy consumption.

[0008] To achieve the above-mentioned purpose, according to one aspect of the present application, a TBM cutterhead wear and energy consumption prediction and optimization method is provided, characterized by comprising:

[0009] Collecting the running data of the TBM and performing data preprocessing, wherein the running data includes adjustable running parameters and fixed running parameters;

[0010] Based on the preprocessed data, a TBM performance prediction model is established using LightGBM, wherein the hyperparameters of LightGBM are optimized using Bayesian optimization, and the TBM performance prediction model is evaluated using prediction error as the loss function until a TBM performance prediction model that meets the preset requirements is obtained;

[0011] Taking the minimization of cutterhead wear and cutterhead energy consumption as the goal, a multi-objective optimization model based on NSGA-II is established to obtain the Pareto front;

[0012] Solving the Pareto front, the optimal adjustable running parameters of the TBM are obtained.

[0013] Further, the data preprocessing comprises:

[0014] Determine the actual cutter wear according to the operation data of the TBM:

[0015]

[0016] Wherein, O1 is the wear of single disc cutter; K is the wear coefficient; D is the diameter of the cutting path of the disc cutter; N is the rotation speed of the cutter; L is the thrust distance of the TBM; V is the thrust speed of the TBM.

[0017] Further, the data preprocessing comprises:

[0018] Determine the actual cutter energy consumption according to the operation data of the TBM:

[0019] O2 = E1 + E2

[0020]

[0021]

[0022] Wherein, O2 is the total energy consumption power of the cutter, E1 is the energy power consumed by the hydraulic thrust system, E2 is the energy power consumed by the cutter system, F i is the thrust, T i is the cutter torque, t i is the time of sequence occurrence, ω i is the rotation speed of the cutter.

[0023] Further, the hyperparameter optimization of LightGBM by Bayesian optimization comprises:

[0024] The Bayesian optimization algorithm based on sequential model optimization technology is used to optimize the hyperparameters by using the Parzen estimator tree algorithm; wherein the Bayesian optimization algorithm takes the EI criterion as the function of the optimization target:

[0025]

[0026] Wherein, x is the hyperparameter candidate; y is the output of the target function; y * is the threshold value of the target function, and p(y|x) is a proxy model representing the probability of y occurring under the condition of given x;

[0027] The optimal hyperparameter combination under the proxy function p(y|x) is searched to maximize the EI with respect to x.

[0028] Furthermore, based on the preprocessed data, a TBM performance prediction model is established using LightGBM. This involves Bayesian optimization of LightGBM hyperparameters, and using prediction error as the loss function to evaluate the TBM performance prediction model, until a TBM performance prediction model meeting preset requirements is obtained. This includes:

[0029] The prediction error is obtained based on the prediction values ​​of the actual cutterhead wear, actual cutterhead energy consumption, and TBM efficiency prediction model.

[0030] If the prediction error meets the preset requirements, the TBM performance prediction model training is complete.

[0031] Furthermore, with the goal of minimizing tool head wear and tool head energy consumption, a multi-objective optimization model based on NSGA-II is established to obtain the Pareto front, including:

[0032] Determine the optimization objective of the multi-objective optimization model:

[0033] Minimize F(x) = [f1(x), f2(x), f3(x), ..., f m (x)]

[0034] Where F(x) is based on n independent variables x1 to x2. n A set of m objective functions;

[0035] Determine the constraints for the multi-objective optimization model:

[0036] g j (x)≤0, j=1,2,...,k

[0037]

[0038] Where, x l and x u It is each variable x i The lower and upper limits of g(x); g(x) is an inequality constraint;

[0039] The Pareto front is obtained based on the optimization objective, constraints, and NSGA-II.

[0040] Furthermore, the process of solving for the Pareto front to obtain the optimal adjustable operating parameters of the TBM includes:

[0041] The scores of all candidate solutions in the Pareto front are determined using the optimal solution superiority distance method:

[0042]

[0043] Among them, S i Let be the score of the i-th solution, with a value in the range [0,1]. It is the Euclidean distance from the i-th solution to the most negative solution. It is the Euclidean distance from the i-th solution to the ideal solution;

[0044] With S i The largest solution is the optimal solution in the Pareto front, which yields the optimal adjustable operating parameters of the TBM.

[0045] According to a second aspect of the present invention, a cutterhead wear and energy consumption optimization system based on tunnel boring machine excavation is provided, comprising:

[0046] The first main module is used to collect the TBM's operating data and perform data preprocessing. The operating data includes adjustable operating parameters and fixed operating parameters.

[0047] The second main module is used to build a TBM performance prediction model based on the preprocessed data using LightGBM. In this module, Bayesian optimization is used to optimize the hyperparameters of LightGBM, and the mean squared error is used as the loss function to evaluate the TBM performance prediction model until a TBM performance prediction model that meets the preset requirements is obtained.

[0048] The third main module is used to establish a multi-objective optimization model based on NSGA-II with the goal of minimizing tool head wear and tool head energy consumption, and to obtain the Pareto front.

[0049] The fourth main module is used to solve the Pareto front to obtain the optimal adjustable operating parameters of the TBM.

[0050] According to a third aspect of the present invention, an electronic terminal is provided, comprising:

[0051] At least one processor, at least one memory, a communication interface, and a bus; wherein,

[0052] The processor, memory, and communication interface communicate with each other through the bus;

[0053] The memory stores program instructions that can be executed by the processor, which calls the program instructions to implement the method.

[0054] According to a fourth aspect of the present invention, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing computer instructions that cause the computer to implement the method described herein.

[0055] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0056] 1. The present invention provides a method for optimizing cutterhead wear and energy consumption. It develops an improved LightGBM based on Bayesian optimization and predicts TBM performance based on TBM operating parameters under specific geological conditions. The method employs Non-Dominated Sorting Genetic Algorithm II (NSGA-II) to solve a Pareto front search multi-objective optimization (MOO) problem, and then uses the optimal solution superiority-inferiority distance (TOPSIS) to obtain the optimal solution for all objectives. This method can reduce cutter wear and cutterhead energy consumption, improve the performance of tunnel boring machines, guide tunnel construction, and promote the construction of tunnel engineering projects.

[0057] 2. The tool head wear and energy consumption optimization method of the present invention optimizes the hyperparameters of LightGBM through Bayesian method. The improved LightGBM has better fitting accuracy and more accurate prediction in predicting TBM tool wear and tool power compared with other machine learning methods such as random forest (RF) and support vector regression (SVR).

[0058] 3. The tool head wear and energy consumption optimization method of the present invention uses TOPSIS's NSGA-II to search for the optimal solution of the output parameters. After multi-objective optimization, the total wear of the TBM tool is significantly reduced by 16.80%, and the tool head energy consumption is significantly reduced by 14.91%, thus improving the performance of the TBM. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of a framework for predicting and optimizing TBM performance indicators according to an embodiment of the present invention.

[0060] Figure 2 The positions and views of the four edge hobs (38, 39, 40a, 40b) in this embodiment of the invention are shown in ((a) the layout of the cutters, (b) the wear information of 40b, (c) the wear information of 40a, (d) the wear information of 38, and (e) the wear information of 39).

[0061] Figure 3 This diagram shows the specific location of tool wear based on the calculations and field measurements of an embodiment of the invention. (The pentagram marks indicate the actual field measurements of tool head wear).

[0062] Figure 4 The TBM performance indicators (a) cutter head wear and (b) cutter head power consumption are shown as line graphs in this embodiment of the invention.

[0063] Figure 5 The diagram shows the cumulative wear value of the total cutterhead and the cumulative energy consumption value of the total cutterhead before and after optimization in an embodiment of the present invention.

[0064] Figure 6 This is a flowchart of a method for optimizing cutterhead wear and energy consumption based on tunnel boring machine excavation, according to an embodiment of the present invention.

[0065] Figure 7 This is a schematic diagram of a cutterhead wear and energy consumption optimization system based on tunnel boring machine excavation, according to an embodiment of the present invention.

[0066] Figure 8 This is a schematic diagram of the physical structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0068] In the description of this invention, unless otherwise explicitly specified and limited, the terms "connected," "linked," and "fixed" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0069] Those skilled in the art will understand that, unless otherwise stated, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the word “comprising” as used in the specification of this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof.

[0070] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as in the embodiments of this application.

[0071] The method for optimizing cutterhead wear and energy consumption based on tunnel boring machine excavation of the present invention can be applied to the technical fields of tunnel boring, construction and tunnel excavation.

[0072] This invention provides a method for optimizing cutterhead wear and energy consumption in tunnel boring machine (TBM) excavation. It applies Bayesian optimization to the hyperparameter optimization of LightGBM, utilizing an improved LightGBM for TBM performance prediction. Based on the Multi-Objective Optimization (MOO) method and NSGA-II, candidate solutions for optimizing cutterhead wear and energy consumption are obtained. The optimal solution is then obtained using the optimal solution superiority distance method. This method can reduce cutterhead wear and energy consumption, improve the performance of tunnel boring machines, guide tunnel construction, and promote the construction of tunnel engineering projects.

[0073] like Figure 1 and Figure 6 As shown, the method for predicting and optimizing TBM cutterhead wear and energy consumption according to the present invention includes the following steps S100 to S400:

[0074] Step S100: Collect the TBM's operating data and perform data preprocessing;

[0075] The operational data includes adjustable operational parameters and fixed operational parameters.

[0076] Electronic sensors installed on the TBM can record real-time operating data and monitor its construction status. The real-time operating data collected by the electronic sensors includes: total thrust, cutterhead torque, average soil pressure on the left side, average soil pressure on the right side, average advance speed, average pressure of group A, average pressure of group B, screw conveyor pressure, gear oil temperature, gear oil pressure, conveyor mud delivery flow rate, and net excavation time.

[0077] By cleaning the TBM's operational data, raw data is obtained. The raw data cannot directly reveal TBM cutterhead wear and energy consumption; these must be calculated from the TBM's operational data.

[0078] The calculation method for cutter head wear is as follows:

[0079]

[0080] Where O1 is the wear amount of the single-disc cutter; k is the wear coefficient; D is the cutting path diameter of the disc cutter; N is the disc rotation speed; L is the TBM thrust distance; and v is the TBM thrust speed.

[0081] The thrust velocity of a TBM is: V (mm / min) = P (mm / r) × N (r / min)

[0082] Where P is the permeability (mm / r) and N is the rotational speed of the cutter head.

[0083] In tunnel construction, the wear coefficient k is determined by soil conditions, the wear resistance of the cutting tools, and the tool arrangement of the cutterhead. The wear resistance of the cutting tools and the cutterhead arrangement are specific to certain types of TBMs; therefore, the difference in the wear coefficient k is mainly related to soil conditions. Based on this, this invention determines the wear coefficient k by relying on real-time TBM operating parameters, actual cutterhead wear values ​​measured by on-site personnel, and the experience of tunnel experts.

[0084] In another embodiment of the invention, the TBM's cutterhead has multiple cutters working together to cut soil or rock. With multiple cutters mounted on the same cutting path diameter, the cutting thickness is typically distributed among the cutters, requiring the wear coefficient k to be modified. The wear of the cutter head, O1, is:

[0085]

[0086] Where, k q To correct for the wear coefficient; Q is the number of tools under the same cutting diameter; D is the cutting path diameter of the disc cutter; L is the TBM thrust distance (km); P is the penetration rate (mm / r).

[0087] TBM cutterhead energy consumption is also an important indicator affecting the total project cost and carbon emissions for environmental protection. The energy consumption of the cutterhead is closely related to the generated load, including the thrust and torque. The calculation method for cutterhead energy consumption is as follows:

[0088] O2 = E1 + E2

[0089]

[0090]

[0091] Where O2 is the total power consumption of the cutter head, E1 is the power consumption of the hydraulic thrust system, E2 is the power consumption of the cutter system, and F... i For thrust, T i The torque of the cutter head is t. i ω represents the time at which the sequence occurs. i This represents the rotational speed of the cutter head.

[0092] Step S200: Based on the preprocessed data, a TBM performance prediction model is established using LightGBM. In this model, Bayesian optimization is used to optimize the hyperparameters of LightGBM, and the prediction error is used as the loss function to evaluate the TBM performance prediction model until a TBM performance prediction model that meets the preset requirements is obtained.

[0093] Lightweight Gradient Boosting Machine (LightGBM) is a distributed gradient boosting method based on decision tree algorithms. Its capabilities lie in supporting efficient parallel training, faster training speed, lower memory consumption, and higher accuracy when processing massive amounts of data quickly. LightGBM is based on gradient-based one-sided gradient sampling (GOSS) and mutually exclusive feature binding (EFB) techniques. GOSS first sorts the training instances in descending order and divides them into two subsets based on their absolute gradient values. Subset A has a larger gradient, with a gradient value of A×100%, while subset B consists of the remaining instances with lower gradients. C The random sample consists of a size of b × |A c The gradient value is (1-a)×100%. Subsequently, the instance features j are evaluated by their estimated variance gain on subset A∪B. The segment is divided, where V′j(d) is calculated by the following formula.

[0094]

[0095] Where A l =[x i ∈A:x ij <d],A r =[x i ∈A:x ij >d];B l =[x i ∈B:x ij <d];B r =[x i ∈B:x ij >d],gi is the negative gradient of the loss function.

[0096] Secondly, EFB aims to reduce the feature dimensionality of a dataset by regrouping it into new bundles, thereby largely avoiding unnecessary computation time and the cost of zero feature values, thus improving data processing speed. EFB mainly consists of two data processing steps. First, EFB needs to determine which features should be bound to the original data features. It treats this as a graph coloring problem, using features as vertices and adding edges for every two non-mutually exclusive features, thus finding the optimal binding result using a greedy algorithm. Next, EFB needs to create new bundles, which is a crucial step in the method. By setting the offset of the original values ​​of the dataset features using a histogram-based algorithm, it can store exclusive features in different bundles.

[0097] Hyperparameters of machine learning models are externally configured values ​​and cannot be estimated from the input data. To obtain an accurate, trained predictive model, machine learning algorithms require hyperparameter tuning. Bayesian optimization algorithms can search for candidate hyperparameters based on probability estimates, providing optimal solutions to the hyperparameter tuning problem. It offers a computational cost advantage over automatic tuning in machine learning, especially for automatic tuning based on large-scale data. Specifically, for the LightGBM method, this invention employs Bayesian optimization for hyperparameter selection. A Bayesian optimization algorithm based on Sequential Model Optimization (SMBO) is used, utilizing the Parzen Estimator Tree (TPE) algorithm for hyperparameter optimization. The Bayesian optimization algorithm uses the Expected Improvement (EI) criterion as a function of the optimization objective. EI is given in the following equation.

[0098]

[0099] Where x is a candidate hyperparameter; y is the output of the objective function; y * Let p(y|x) be the threshold of the objective function, and p(y|x) be a surrogate model representing the probability of y occurring given x. The goal is to maximize EI with respect to x by searching for the optimal combination of hyperparameters under the surrogate function p(y|x).

[0100] For the TBM performance prediction model established by LightGBM after Bayesian optimization and parameter tuning, the preprocessed running data obtained from S100 is divided into training set and test set for model testing.

[0101] Specifically, step S200 includes steps S201 to S202:

[0102] Step S201: Obtain the prediction error based on the actual tool head wear, actual tool head energy consumption, and the prediction value of the TBM performance prediction model;

[0103] Step S202: If the prediction error meets the preset requirements, the TBM performance prediction model training is complete.

[0104] Step S200 also includes step S203:

[0105] If the prediction error meets the preset requirements, the hyperparameters are optimized using a Bayesian method, a TBM performance prediction model is built based on LightGBM, and then trained.

[0106] This invention uses prediction error as the loss function to evaluate the training effect of the model. The prediction error includes: mean absolute error (MAE), root mean square error (RMSE), and R0. 2 Three evaluation metrics. The closer the MAE and RMSE values ​​are to zero, the closer the machine learning predictions are to the actual values ​​in the dataset, while R... 2The closer the value is to 1, the higher the correlation between the TBM performance index and the operating parameters, meaning the more accurate the machine learning model prediction.

[0107]

[0108]

[0109]

[0110] Where N is the total number of samples used for testing, f i (x)) is the predicted value of the TBM performance prediction model, y i Actual cutter head wear, actual cutter head energy consumption;

[0111] Step S300: With the goal of minimizing tool head wear and tool head energy consumption, establish a multi-objective optimization model based on NSGA-II to obtain the Pareto front;

[0112] Specifically, step S300 includes steps S301 to S303.

[0113] Step S301: Determine the optimization objective of the multi-objective optimization model:

[0114] Minimize F(x) = [f1(x), f2(x), f3(x), ..., f m (x)]

[0115] Where F(x) is based on n independent variables x1 to x2. n A set of m objective functions;

[0116] Step S302: Determine the constraints of the multi-objective optimization model:

[0117] g j (x)≤0, j=113,...,k

[0118]

[0119] Where, x l and x u It is each variable x i The lower and upper limits of g(x); g(x) is an inequality constraint;

[0120] Step S303: Obtain the Pareto front based on the optimization objective, constraints, and NSGA-II.

[0121] To find the optimal solution for multi-objective optimization, it is necessary to generate the Pareto front. This invention uses the NSGA-II algorithm to search for the Pareto front. The NSGA-II process is as follows:

[0122] An initial population of size P is randomly generated. A genetic algorithm based on non-dominant ranking is used to generate the first offspring population through selection, crossover, and mutation. Starting from the second generation, the parent and offspring populations are merged for rapid non-dominant ranking, and the crowding degree (CD) of individuals in each non-dominant layer is calculated. Suitable individuals are selected based on their non-dominant relationship and CD to generate a new parent population. A new offspring population is then generated using a genetic algorithm. These steps are repeated until the population size or maximum number of offspring meets a preset termination condition, resulting in a Pareto front.

[0123] Step S400: Solve for the Pareto front to obtain the optimal adjustable operating parameters of the TBM.

[0124] The solution to the Pareto front yields the optimal adjustable operating parameters of the TBM, including:

[0125] The scores of all candidate solutions in the Pareto front are determined using the optimal solution superiority distance method:

[0126]

[0127] Among them, S i Let be the score of the i-th solution, with a value in the range [0,1]. It is the Euclidean distance from the i-th solution to the most negative solution. It is the Euclidean distance from the i-th solution to the ideal solution;

[0128] With S i The largest solution is the optimal solution in the Pareto front, which yields the optimal adjustable operating parameters of the TBM.

[0129] This invention proposes a method combining Bayesian optimization with an improved LightGBM and a TOPSIS approach based on NSGA-II, aiming to improve the performance of a TBM (Tunnel Boring Machine), including reducing cutterhead energy consumption and wear. The method utilizes Bayesian optimization to assist the improved LightGBM in setting hyperparameters for prediction, employs NSGA-II to obtain the Pareto tip, and then applies TOPSIS to select the optimal solution for optimization. The innovation and advantages of the proposed method highlight its accuracy and efficiency in prediction and optimization for long-term and large-scale datasets. In practical applications, cutterhead hob wear is reduced from 0.001806 mm to 0.001502 mm, and cutterhead power consumption is reduced from 469.2546 kW to 400.3490 kW. The reduction in cutterhead energy consumption decreases carbon emissions, while the optimization of cutter wear reduces the frequency of tool replacements. This contributes to lower project costs and improve project profitability.

[0130] Example 1

[0131] To make the implementation process of the method of the present invention clearer, this example takes the external tunnel project of a certain subway line 6 as an example to specifically illustrate the implementation process of the method of the present invention.

[0132] The process of this embodiment is implemented on a computer platform with the following device parameters. The specific parameter information is as follows: [Operating System: Windows 10 Professional Edition 21H2; CPU: 12th Generation Intel® Core™ i7-12700F 2.10GHz; Memory: 16GB; GPU: NVIDIA GeForce RTX3070 8GB].

[0133] Step 1: Collect TBM operating data and perform data preprocessing;

[0134] Step 1.1 Data Acquisition. Electronic sensors installed at key locations on the TBM can record real-time construction data and monitor its construction status. This example uses a subway line 6 external tunnel project as an example to verify the proposed method. The dataset consists of a section approximately 640 tunnel rings long. Data units are obtained from the sensors, approximately every 20mm. After data cleaning and preprocessing, a total dataset of [39040x12] is obtained. The data is divided according to the tunnel excavation time, with the training set accounting for 70% of the total dataset and the test set accounting for 30%. The sizes of the training and test sets are [27328x12] and [11712x12], respectively.

[0135] Step 1.2 Determine the cutter head wear value. To calculate cutter wear using the JTS formula, the wear coefficient k needs to be determined. The wear coefficient k is generally determined by geological conditions, as shown in Table 1. It can be seen that geological conditions are mainly divided into three categories, and the range of values ​​for the wear coefficient k corresponds to these categories. In this example, the geological conditions of the project are mainly SIII type soil, with fewer geological conditions; the soil type is SIV, where SIII and SIV are moderately to highly weathered rocks (i.e., weathering grades III and IV). These geological features belong to columns such as gravel and rock. Based on this, tunnel experts, using extensive practical engineering experience and geological data, determined the wear coefficient k = 0.04. This example selects four edge cutters (38, 39, 40a, 40b) for verification. Their positions and views are referenced... Figure 2 As shown, the cutting path diameter of the four edge hobs is the same as the diameter of the TBM (D = 6.68m). The wear condition of the hobs can be accurately determined from the figure.

[0136] Table 1. Wear coefficient values ​​k under different geological conditions.

[0137] Clay, silt, etc. Sand, loam, etc. Sand and gravel, rock, etc. - SII and SIV types 0.001～0.003 0.01～0.02 0.03～0.05

[0138] Based on on-site manual measurements, the actual wear values ​​of the four edge hobs were measured at rings 103, 271, 353, and 556, respectively. Based on this, and according to tunnel experts, the number of cutting teeth q = 4, and the real-time operating data of the TBM, the wear coefficient k can be approximately determined to be 0.04. (Refer to...) Figure 3 Tables 1 and 2 show the actual measured values ​​and calculated results based on k, respectively. The calculated values ​​are close to the measured values, with an average error of 0.21%, indicating that the wear coefficient k determined by the experts is relatively accurate and consistent with the geological conditions of the site in this example. This estimated value can be further used for tool wear calculations in the engineering projects where this example is applied.

[0139] Table 2. Comparison of calculated and measured wear values ​​of the cutting tool at specific locations.

[0140]

[0141] This example selects 12 key operating parameters x1 to x2 collected by the TBM sensor. 12 As input data for training the model, tool wear (O1) and tool power (O2) are used as output targets, detailed in Table 3. Furthermore, Table 4 shows the data distribution associated with these parameters. Detailed trends of the tunnel boring machine's output parameters during the tunneling process are shown in Table 4. Figure 4 As shown.

[0142] Table 3. Detailed description of the selected TBM parameters and performance indicators.

[0143]

[0144]

[0145] Table 4. Data description of selected TBM parameters and performance indicators.

[0146]

[0147] Step 2: Based on the preprocessed data, a TBM performance prediction model is established using LightGBM. In this model, Bayesian optimization is used to optimize the hyperparameters of LightGBM, and the prediction error is used as the loss function to evaluate the TBM performance prediction model until a TBM performance prediction model that meets the preset requirements is obtained.

[0148] Specifically, tool wear and tool power consumption are key performance indicators of TBMs. Accurate prediction of tool wear and tool power can provide reliable estimates, which is beneficial for TBM operational decisions. The proposed model can provide accurate predictions, demonstrating its powerful ability to map the nonlinear relationship between input and output variables. To highlight the advantages of the improved LightGBM, a series of systematic analyses were conducted. The specific analysis results are as follows:

[0149] (1) Bayesian optimization has good performance in hyperparameter search. To achieve the high performance of the improved LightGBM, appropriate hyperparameters need to be set, including the maximum depth D of the tree. m The number of leaves N l The learning rate η and the number of estimators N e Table 5 lists f1(x) i f2(x) and f2(x) i The hyperparameter search space and optimal hyperparameter results are presented. It can be seen that Bayesian optimization yields a large number of hyperparameters for model setup, and the results are high, indicating that Bayesian optimization has good performance in effectively optimizing hyperparameters and helps achieve good performance. Among all hyperparameters, the optimal hyperparameters obtained are as follows: the learning rate η for the objective O1 is 0.0268, and the number of estimators N... e The value is 660, and the number of leaves is N. i The maximum depth is 57, and the maximum depth is D. m The learning rate η for O2 is 0.0181, and the number of estimators N is 81. e The value is 710, and the number of leaves is N. l The maximum depth is 3, and the maximum depth is D. m The set value is 81. This example sets the optimal hyperparameters based on the Bayesian optimization results, enabling the model to conduct further exploration.

[0150] Table 5. Search space and optimal hyperparameters

[0151]

[0152] (2) Based on the optimal hyperparameters, the improved LightGBM achieved good results in predicting TBM tool wear and tool power. During training, the loss was recorded in real time to demonstrate the model's training performance. Clearly, the loss value decreased rapidly and then gradually flattened out, indicating that the model can be trained well to learn the features in the data. To verify the model's performance, the well-trained model was tested on the test set. The results are shown in Table 6. It can be seen that the model can provide accurate real-time predictions, with an O1 MAE of 1.2584E-04, an RMSE of 1.7237E-04, and a R... 2 The MAE for O2 was 0.9084, the RMSE was 31.0081, and the R value was 36.0126. 2 The value is 0.8044. Furthermore, overall, these results indicate that the improved LightGBM performs well, showing good agreement with actual measurements in predicting the cutting power and cutting wear of the TBM.

[0153] Table 6. Comparison results of different machine learning methods.

[0154]

[0155] (3) Compared with other machine learning methods, the improved LightGBM exhibits better performance. This study compared the improved LightGBM with other machine learning methods such as random forest (RF) and support vector regression (SVR). Table 6 shows the comparison results of different methods O1 and O2. It can be seen that the improved LightGBM performs best in predicting TBM tool power and tool wear. RF and SVR are worse than the improved LightGBM in all evaluation metrics, with SVR performing the worst among all methods. It can be seen that the trend and fluctuation of the prediction results of the improved LightGBM are basically consistent with the actual ground values, while the differences of other methods are greater. From all the comparison results, the improved LightGBM outperforms other methods in predicting TBM tool wear and tool power, indicating that the improved LightGBM has better performance and great potential for practical application.

[0156] Overall, the improved LightGBM based on Bayesian optimization performs well in predicting cutting power and cutting wear. Furthermore, the improved LightGBM exhibits better performance compared to other machine learning methods. In conclusion, all results show that the improved Bayesian-optimized LightGBM demonstrates excellent performance, effectiveness, and reliability in predicting tool wear and tool power, and possesses significant practical potential.

[0157] Step 3: With the goal of minimizing tool head wear and tool head energy consumption, establish a multi-objective optimization model based on NSGA-II to obtain the Pareto front;

[0158] Multi-objective optimization and result analysis. When performing multi-objective optimization, f1(x) is used. i f2(x) and f2(x) i As a multi-objective fitness function, its purpose is to minimize the wear of the bottom cutterhead rollers (O1) and the power consumption of the cutterhead (O2). Furthermore, during TBM tunneling, parameters x1 to x5 are considered to be adjustable through direct and indirect methods, while other parameters are considered fixed factors that cannot be directly or indirectly adjusted. This study mainly focuses on the impact of parameters x1 to x5 on the results of O1 and O2. The variation space of these five adjustable decision variables x1 to x5 in the multi-objective optimization is shown in Table 7.

[0159] Table 7. Variable space of five adjustment decision variables x1 to x5 in multi-objective optimization.

[0160]

[0161] During training with NSGA-II, the population size was set to 50 and the maximum number of offspring was set to 100 as the parameters for NSGA-II.

[0162] Step 4: Solve for the Pareto front to obtain the optimal adjustable operating parameters of the TBM.

[0163] To find the optimal result, this example examines four samples (samples 7000, 17000, 27000, and 37000). Furthermore, to find the optimal solution from these Pareto fronts, this example employs the Top-Rank Distance of Optimal Solutions (TOPSIS) method. Based on the results, the analysis is as follows:

[0164] (1) The NSGA-II algorithm using TOPSIS performs well in searching for the optimal solution of the output parameters, which helps to improve the performance of the TBM. Clearly, NSGA-II performs well, extracting several Pareto fronts and thus obtaining the optimal solution. Statistical analysis of the original and optimized data is shown in Table 8. It can be seen that the average tool wear decreased from 0.001806 to 0.001502, and its standard deviation decreased from 0.000571 to 0.0001417. The average tool power decreased from 469.2546 to 400.3490, and its standard deviation decreased from 96.2401 to 83.9914. Compared with the original data, tool wear decreased by 16.83%, and tool power decreased by 14.68%. Experimental results show that this method has a good effect on optimizing TBM performance. Comparison between the optimized results and the original data shows that tool wear and tool power are significantly reduced, indicating that this method has superior performance and outstanding optimization results. The results show that this method can optimize the output parameters of the TBM and improve its performance.

[0165] Table 8. Statistical analysis of data between original values ​​and optimization results.

[0166]

[0167] Note: 1. μ is the mean, σ is the standard deviation. 2. Based on percentage formula. (The same applies below) the percentage of optimization, where μ I To improve the average value of the results, μ O This represents the average value of the original dataset.

[0168] (2) The proposed optimization method can optimize the operating parameters of TBM, which is beneficial to the performance of TBM. The original data has a wide distribution, while the optimized data has a relatively clustered distribution and a narrower range, indicating that the optimization method proposed in this paper improves the five operating parameters and is beneficial to the performance of TBM.

[0169] (3) After multi-objective optimization, the total tool wear of the TBM was significantly reduced by 16.80%, and the power consumption of the tool head was significantly reduced by 14.91%, thus improving the performance of the TBM. (Refer to...) Figure 5 As shown, the cumulative values ​​of total cutter head wear and total cutter head power consumption are presented under the original and optimized data. In the original data, the cumulative total cutter head hob wear was 70.502 mm, while after optimization, the total cutter head hob wear was 58.657 mm, a reduction of 16.80%. Furthermore, the cumulative total cutter head power consumption in the original data was 453057 kWh, while the optimized cutter head power consumption was 385485 kWh, a reduction of 14.91%. Based on these results, it can be concluded that TBMs can save significant amounts of energy and reduce carbon emissions. Additionally, based on practical project experience, assuming the average wear limit of the four edge hobs is 15 mm, the number of tool replacements based on the original data is... The optimized tool change frequency is [number] times; while the optimized tool change frequency is [number]. The reduction in cutterhead energy consumption decreases carbon emissions, while the optimization of tool wear reduces replacement frequency. All of these contribute to the efficiency of tunnel engineering. In other words, the optimization method proposed in this study significantly improves the output parameters, which is beneficial for improving TBM performance, reducing project costs, and increasing project profitability.

[0170] Based on the results of this example, the conclusions are summarized as follows: (1) The improved LightGBM shows better prediction performance in predicting tool head wear and tool head power consumption. The MAE of O1 is 1.2584E-04, the RMSE is 1.7237E-04, and the R 2 The MAE for O2 was 0.9084, the RMSE was 31.0081, and the R value was 36.0126. 2 The value is 0.8044. (2) The NSGA-II algorithm combined with TOPSIS has good solution performance for multi-objective optimization. The NSGA-II algorithm has the ability to first search the Pareto front and then apply the TOPSIS algorithm to select the optimal solution, which helps to improve the performance of TBM and is beneficial to the construction of tunnel engineering projects. The wear of the cutterhead hob decreased from 0.001806 (mm) to 0.001502 (mm), and the power consumption of the cutterhead decreased from 469.2546 (Kw) to 400.3490 (Kw). (3) The comparison results show that Bayesian optimization has good performance and can help the improved LightGBM and NSGA-II obtain better results.

[0171] The implementation of the various embodiments of the present invention is based on programmed processing by a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of the present invention can be encapsulated into various modules. Based on this reality, and building upon the above embodiments, the present invention provides a device for optimizing cutterhead wear and energy consumption during tunnel boring machine (TBM) excavation. This device is used to execute a method for optimizing cutterhead wear and energy consumption during TBM excavation as described in the above method embodiments.

[0172] like Figure 7 As shown, the present invention provides a cutterhead wear and energy consumption optimization system based on tunnel boring machine excavation, comprising:

[0173] The first main module is used to collect the TBM's operating data and perform data preprocessing, including adjustable and fixed operating parameters. The second main module is used to establish a TBM performance prediction model based on the preprocessed data using LightGBM. This model uses Bayesian optimization to optimize the hyperparameters of LightGBM and evaluates the TBM performance prediction model using the error as the loss function until a TBM performance prediction model that meets the preset requirements is obtained. The third main module is used to establish a multi-objective optimization model based on NSGA-II with the goal of minimizing toolhead wear and toolhead energy consumption, and obtain the Pareto front. The fourth main module is used to solve for the Pareto front to obtain the optimal adjustable operating parameters of the TBM.

[0174] The method in this embodiment of the invention is implemented using an electronic device; therefore, it is necessary to introduce the relevant electronic device. For this purpose, this embodiment of the invention provides an electronic device, such as... Figure 8 As shown, the electronic device includes at least one processor, a communications interface, at least one memory, and a communications bus, wherein the at least one processor, the communications interface, and the at least one memory communicate with each other via the communications bus. The at least one processor can invoke logical instructions stored in the at least one memory to execute all or part of the steps of the methods provided in the foregoing method embodiments.

[0175] Furthermore, when the logical instructions in at least one of the aforementioned memories can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various method embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0176] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0177] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0178] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Based on this understanding, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, or sometimes in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0179] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for optimizing cutterhead wear and energy consumption based on tunnel boring machine (TBM) excavation, characterized in that, include: Collect TBM operational data and perform data preprocessing, including adjustable operational parameters; Based on the preprocessed data, a TBM performance prediction model is established using LightGBM. In this model, Bayesian optimization is used to optimize the hyperparameters of LightGBM, and the prediction error is used as the loss function to evaluate the TBM performance prediction model until a TBM performance prediction model that meets the preset requirements is obtained. Based on the preprocessed data, a TBM performance prediction model is established using LightGBM. This involves Bayesian optimization of LightGBM hyperparameters, and using prediction error as the loss function to evaluate the TBM performance prediction model, until a TBM performance prediction model meeting preset requirements is obtained. This includes: The prediction error is obtained based on the actual toolhead wear, actual toolhead energy consumption, and the prediction value of the TBM performance prediction model. If the prediction error meets the preset requirements, the TBM performance prediction model training is complete. With the goal of minimizing tool head wear and tool head energy consumption, a multi-objective optimization model based on NSGA-II is established to obtain the Pareto front; The goal is to minimize tool head wear and tool head energy consumption. A multi-objective optimization model based on NSGA-II is established to obtain the Pareto front, including: Determine the optimization objective of the multi-objective optimization model: , in, Based on n independent variables arrive The set of m objective functions, i.e., based on n independent variables. arrive The set of cutterhead wear and cutterhead energy consumption obtained from the TBM performance prediction model that meets the preset requirements. arrive There are n adjustable operating parameters; Determine the constraints for the multi-objective optimization model: , in, and Each variable The lower and upper limits; It is an inequality constraint; The Pareto front is obtained based on the optimization objective, constraints, and NSGA-II. Solving for the Pareto front yields the optimal adjustable operating parameters for the TBM.

2. The method for optimizing cutterhead wear and energy consumption based on tunnel boring machine excavation according to claim 1, characterized in that, The hyperparameter optimization of LightGBM using Bayesian optimization includes: A Bayesian optimization algorithm based on sequential model optimization technique is adopted, and the Parzen estimator tree algorithm is used for hyperparameter optimization; wherein, the Bayesian optimization algorithm uses the EI criterion as a function of the optimization objective. , in, It is a hyperparameter candidate; The output of the objective function; The threshold of the objective function, To represent a given under conditions A proxy model for the probability of occurrence; By searching for proxy functions The optimal combination of hyperparameters, making EI relative to maximize.

3. The method for optimizing cutterhead wear and energy consumption based on tunnel boring machine excavation according to claim 1, characterized in that, The solution to the Pareto front yields the optimal adjustable operating parameters of the TBM, including: The scores of all candidate solutions in the Pareto front are determined using the optimal solution superiority distance method: , in, The score for the i-th solution is [0, 1]. It is the Euclidean distance from the i-th solution to the most negative solution. It is the Euclidean distance from the i-th solution to the ideal solution; by The largest solution is the optimal solution in the Pareto front, which yields the optimal adjustable operating parameters of the TBM.

4. A cutterhead wear and energy consumption optimization system based on tunnel boring machine (TBM) excavation, characterized in that, include: The first main module is used to collect the TBM's operating data and perform data preprocessing. The operating data includes adjustable operating parameters. The second main module is used to build a TBM performance prediction model based on the preprocessed data using LightGBM. In this module, Bayesian optimization is used to optimize the hyperparameters of LightGBM, and the prediction error is used as the loss function to evaluate the TBM performance prediction model until a TBM performance prediction model that meets the preset requirements is obtained. Based on the preprocessed data, a TBM performance prediction model is established using LightGBM. This involves Bayesian optimization of LightGBM hyperparameters, and using prediction error as the loss function to evaluate the TBM performance prediction model, until a TBM performance prediction model meeting preset requirements is obtained. This includes: The prediction error is obtained based on the actual toolhead wear, actual toolhead energy consumption, and the prediction value of the TBM performance prediction model. If the prediction error meets the preset requirements, the TBM performance prediction model training is complete. The third main module is used to establish a multi-objective optimization model based on NSGA-II with the goal of minimizing toolhead wear and toolhead energy consumption, and to obtain the Pareto front. The establishment of the multi-objective optimization model based on NSGA-II with the goal of minimizing toolhead wear and toolhead energy consumption, and the obtaining of the Pareto front, includes: Determine the optimization objective of the multi-objective optimization model: , in, Based on n independent variables arrive The set of m objective functions, i.e., based on n independent variables. arrive The set of cutterhead wear and cutterhead energy consumption obtained from the TBM performance prediction model that meets the preset requirements. arrive There are n adjustable operating parameters; Determine the constraints for the multi-objective optimization model: , in, and Each variable The lower and upper limits; It is an inequality constraint; The Pareto front is obtained based on the optimization objective, constraints, and NSGA-II. The fourth main module is used to solve the Pareto front to obtain the optimal adjustable operating parameters of the TBM.

5. An electronic terminal, characterized in that, include: At least one processor, at least one memory, a communication interface, and a bus; wherein, The processor, memory, and communication interface communicate with each other through the bus; The memory stores program instructions that can be executed by the processor, which invokes the program instructions to implement the method as described in any one of claims 1-3.

6. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to perform the method as described in any one of claims 1-3.

Images