Shield tunneling machine energy consumption prediction and optimization method

Abstract

The invention relates to the technical field of shield tunneling optimization, and discloses a shield tunneling machine energy consumption prediction and optimization method, which comprises the following steps: S1, carrying out comprehensive energy consumption analysis on a cutterhead system, a propulsion system and a muck discharge system of a shield tunneling machine, and calculating the total energy consumption of the cutterhead system, the propulsion system and the muck discharge system of the shield tunneling machine; s2, preliminarily screening related parameters influencing shield energy consumption, performing abnormal value processing, noise reduction processing, standardization processing and correlation analysis of characteristic variables on the data, and finally outputting characteristic parameters related to shield machine energy consumption prediction; s3, constructing a shield tunneling machine energy consumption prediction model; according to the method, the energy consumption mathematical model of each system is established from the angle of force acting through comprehensive analysis of the cutter head system, the propelling system and the muck discharging system of the shield tunneling machine, the energy consumption of the shield tunneling machine can be accurately quantified, and energy consumption prediction of the shield tunneling machine is more accurate and higher in reliability.

Description

Technical Field

[0001] The invention relates to the technical field of shield tunneling optimization, and more specifically to a shield machine energy consumption prediction and optimization method. Background Art

[0002] Competition within the construction industry is becoming increasingly fierce, and cost control has become the key to success. Companies of the same type are actively exploring ways to save energy and increase efficiency. If they do not follow up in a timely manner and fall behind in intelligent energy consumption management, they will face the risk of being eliminated by the market. The traditional energy consumption management methods of shield machines have many drawbacks, such as energy waste, shield machine process parameters working at full load, serious energy consumption of equipment idling, delayed and inaccurate energy consumption data statistics, and it is difficult to accurately locate key energy-saving points and realize the analysis and application of post-data. Based on the above internal and external pressures and opportunities, it is imperative to introduce an intelligent energy consumption management and control system for shield machines.

[0003] Disadvantages of existing technology:

[0004] 1. Lack of data collection capabilities: Currently, shield machine equipment has not yet carried out any form of energy consumption data collection, resulting in the energy consumption status of the equipment being completely unknown, and it is impossible to obtain basic data to support energy consumption supervision and analysis;

[0005] 2. Blind spot in energy consumption management: Due to the lack of monitoring of equipment energy consumption, it is impossible to identify high-energy-consuming equipment or abnormal energy consumption links, resulting in difficulty in discovering and controlling energy waste;

[0006] 3. Lack of data analysis and optimization methods: Without energy consumption data, the existing system cannot carry out energy consumption trend analysis, energy consumption forecasting, energy consumption evaluation or parameter optimization, and it is difficult to improve the energy consumption management level of the shield machine;

[0007] 4. Disconnection between process and energy consumption data: Failure to combine energy consumption data with construction process parameters, inability to fully understand the relationship between equipment operation status and energy consumption, and difficulty in optimizing construction process and energy consumption allocation;

[0008] To this end, the present invention provides a shield machine energy consumption prediction and optimization method. Summary of the invention

[0009] In order to overcome the above-mentioned defects of the prior art, the present invention provides a shield machine energy consumption prediction and optimization method to solve the problems existing in the above-mentioned background technology.

[0010] The present invention provides the following technical solution: a method for predicting and optimizing energy consumption of a shield machine, comprising the following steps:

[0011] S1. Conduct a comprehensive energy consumption analysis on the cutterhead system, propulsion system and slag discharge system of the shield machine, and calculate the total energy consumption of the cutterhead system, propulsion system and slag discharge system of the shield machine;

[0012] S2. Preliminarily screen the relevant parameters that affect the shield machine energy consumption, and perform outlier processing, noise reduction, standardization and correlation analysis of characteristic variables on the data, and finally output the characteristic parameters related to the shield machine energy consumption prediction;

[0013] S3, constructing a shield machine energy consumption prediction model with characteristic parameters related to shield machine energy consumption prediction as input, shield machine energy consumption as prediction target, and LSTM as basic structure;

[0014] S4. GA is used to optimize the hyperparameters of LSTM, and an energy consumption prediction model based on GA-LSTM is established to predict the energy consumption of the shield machine.

[0015] Preferably, the comprehensive energy consumption analysis of the cutter head system, propulsion system and slag discharge system of the shield machine in step S1 specifically includes:

[0016] The total energy consumption expression for constructing the shield machine cutter system, propulsion system and slag discharge system is as follows:

[0017] E=T a Vt+T p w β t+T r w r t①

[0018] In formula ①, E is the total energy consumption of the shield machine, T a is the thrust of the shield machine, V is the propulsion speed, T p is the cutter head torque, w β is the cutter head speed, T r is the screw conveyor torque, w r is the speed of screw conveyor, and t is the excavation time.

[0019] Preferably, use formula ② to calculate the thrust T of the shield machine: a ;

[0020]

[0021] In formula ②, T a is the thrust of the shield machine, μ1 is the friction coefficient between the shield shell and the soil, L is the length of the shield machine, D is the outer diameter of the shield machine, γ is the soil volume, H f is the cover length, R β is the cutterhead diameter, k is the lateral earth pressure coefficient, G is the weight of the shield machine, R g is the outer diameter of the shield segment, μ g is the friction coefficient of each meter of the segment, G c is the weight of the rear matching vehicle, and μ2 is the friction coefficient between the rear matching vehicle and the track.

[0022] Preferably, the cutter head torque T is calculated using formula ③ p ;

[0023]

[0024] In formula ③, is the cutter head opening ratio, μ β Indicates the friction coefficient between the front of the cutterhead and the soil, μ β The range is 0.07-0.1, P s is the horizontal earth pressure formed when the cutterhead rotates, ρ is the soil density around the cutterhead, H is the vertical burial depth of the shield, W is the average length of the shield, K a is the active earth pressure coefficient of the surrounding soil, Q1 is the compressive strength of the cutter head, h is the cutting depth of the cutter head, D0 is the external radius of the cutter head, K i is the soil shear strength, D1 is the outer diameter of the cutter head support beam, D2 is the inner diameter of the cutter head support beam, L z is the length of the cutter head support beam, μ α is the friction coefficient when the cutter head rotates, G β is the mass of the shield cutterhead, R1 is the contact radius of the cutterhead bearing, and σ is the cutterhead sealing ratio.

[0025] Preferably, the screw conveyor torque T is calculated using formula ④ r ;

[0026]

[0027] In formula ④, D l is the diameter of the screw conveyor of the shield machine, D g is the diameter of the screw conveyor, α is the inclination angle of the screw conveyor, L a is the length of the screw conveyor, λ is the density of the soil in the screw conveyor, f t is the friction coefficient between the screw conveyor housing and the surrounding soil, and β is the spiral rise angle.

[0028] Preferably, in step S2, the relevant parameters affecting shield machine energy consumption are preliminarily screened, and outlier processing, noise reduction processing, standardization processing and correlation analysis of characteristic variables are performed on the data, and finally the characteristic parameters related to shield machine energy consumption prediction are output, which specifically include:

[0029] S21. Screening of relevant parameters affecting shield energy consumption

[0030] The parameters that affect the energy consumption of shield machines include: shield machine cutter system parameters, propulsion system parameters, and parameters related to the slag discharge system; they are: total thrust X1, cutter torque X2, screw conveyor torque X3, penetration X4, cutter speed X5, excavation speed X6, i-group main push cylinder thrust X i, i = 7, ..., n, excavation chamber j pressure X j , j = n + 1, ..., m;

[0031] S22. Outlier processing: Use the 3σ principle to detect and remove outliers;

[0032] S23, denoising processing: using a wavelet threshold denoising method to denoise the original characteristic parameter signal;

[0033] S24, dimensionless processing: using Min-Max normalization to process the characteristic parameter data after noise reduction;

[0034] S25. Characteristic parameter correlation coefficient analysis: Use the Pearson statistical method to quantitatively analyze the dimensionless characteristic parameter data. The size of the linear correlation coefficient reflects the strength of the correlation between the two variables, and selects characteristic parameters with strong correlation with energy consumption. These parameters are used as characteristic parameters related to the energy consumption prediction of the shield machine.

[0035] Preferably, in step S25, a correlation coefficient r∈[-1,1] is defined, r>0 indicates a positive correlation, r<0 indicates a negative correlation, and r=0 indicates a nonlinear relationship, and a characteristic parameter with a correlation coefficient r>0.5 is selected as a characteristic parameter with a strong correlation with energy consumption.

[0036] Preferably, in step S4, the GA is used to optimize the hyper parameters of the LSTM, specifically including:

[0037] Input and Output

[0038] Input variables: X1, X2, X3, ..., X i ; X1, X2, X3, ..., X i They represent characteristic parameters related to the prediction of shield machine energy consumption;

[0039] Input variables: Y1 total energy consumption per unit time;

[0040] Hyperparameter Definition

[0041] The LSTM hyperparameters that need to be optimized include:

[0042] Number of hidden layer neurons: [32, 64, 128, 256]

[0043] Learning rate: [0.1, 0.2, 0.3, 0.4, 0.5]

[0044] Dropout rate: [0.1, 0.2, 0.3, 0.4, 0.5]

[0045] Batch size: [16,32,64,128]

[0046] Number of LSTM layers: [1,2,3]

[0047] coding

[0048] The hyperparameters are encoded as chromosomes, with 64 hidden layer neurons, 0.001 learning rate, 0.2 dropout rate, 32 batch size, and 2 LSTMs;

[0049] GA parameter settings

[0050] Set the basic parameters of GA:

[0051] Population size: 20-100;

[0052] Crossover probability: 0.6-0.9;

[0053] Mutation probability: 0.01-0.1;

[0054] Maximum number of iterations: 50-100 or fitness convergence threshold: 10 -4 -10 -6 ;

[0055] Initialize the population

[0056] Randomly generate the initial weights and thresholds of the neural network as population individuals;

[0057] Fitness function

[0058] Use the training data to train the LSTM model and evaluate the accuracy on the validation set, using the mean square error (MSE) as the fitness value.

[0059]

[0060] In formula ⑤, N is the number of samples, y i ′ is the predicted value of the i-th sample, y i is the true value of the i-th sample;

[0061] Genetic manipulation

[0062] Selection, crossover, and mutation:

[0063] Select excellent individuals based on fitness values;

[0064] Perform crossover operation on the selected individuals to generate new individuals;

[0065] Perform mutation operations on some new individuals to increase population diversity;

[0066] Update population

[0067] Replace the parent generation with the newly produced offspring to form a new generation of population;

[0068] Termination Condition

[0069] Check whether the termination condition is reached: if the maximum number of iterations or fitness value is not reached, perform genetic operations, otherwise, enter the output optimal solution;

[0070] Output the optimal solution

[0071] Output the LSTM hyperparameters corresponding to the individual with the highest fitness;

[0072] Training the final model

[0073] The final shield machine energy consumption prediction model is trained using the optimal hyperparameters.

[0074] Technical effects and advantages of the present invention:

[0075] 1. Compared with the prior art, the present invention establishes a mathematical model of energy consumption of each system from the perspective of force work through a comprehensive analysis of the shield machine's cutter head system, propulsion system and slag discharge system, covering all major sources of energy consumption. This method can not only accurately quantify the energy consumption of the shield machine, but also identify the main energy consumption links, providing a scientific basis for energy-saving optimization.

[0076] 2. Compared with the prior art, the present invention extracts characteristic parameters related to energy consumption through Pearson correlation coefficient analysis method, and then establishes an energy consumption prediction model, thereby overcoming the problem that the energy consumption calculation model in the prior art is incomplete or lacks accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a flow chart of the present invention.

[0078] Figure 2 It is the pressure diagram of the shield machine surface of the present invention.

[0079] Figure 3 This is a flow chart of feature value extraction of the present invention.

[0080] Figure 4 It is the Pearson correlation heat map of the present invention.

[0081] Figure 5 This is a structural diagram of the LSTM neural network of the present invention.

[0082] Figure 6 Flowchart for optimizing LSTM hyperparameters for GA of the present invention.

[0083] Figure 7 It is the fmincon solution process diagram of MATLAB of the present invention. DETAILED DESCRIPTION

[0084] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are only examples. The shield machine energy consumption prediction and optimization method involved in the present invention is not limited to the various structures recorded in the following embodiments. All other implementations obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0085] like Figure 1 As shown in the present invention, a method for predicting energy consumption of a shield machine is provided, comprising the following steps:

[0086] S1. Conduct a comprehensive energy consumption analysis on the cutterhead system, propulsion system and slag discharge system of the shield machine, and calculate the total energy consumption of the cutterhead system, propulsion system and slag discharge system of the shield machine.

[0087] S2. Preliminary screening of relevant parameters affecting shield machine energy consumption, and data processing of outliers, noise reduction, standardization and correlation analysis of characteristic variables, and finally output of characteristic parameters related to shield machine energy consumption prediction.

[0088] S3. A shield machine energy consumption prediction model is constructed with characteristic parameters related to shield machine energy consumption prediction as input, shield machine energy consumption as prediction target, and LSTM as the basic structure.

[0089] S4. GA is used to optimize the hyperparameters of LSTM, and an energy consumption prediction model based on GA-LSTM is established to predict the energy consumption of the shield machine.

[0090] In step S1, a comprehensive energy consumption analysis is performed on the cutter head system, propulsion system and slag discharge system of the shield machine, and the total energy consumption of the cutter head system, propulsion system and slag discharge system of the shield machine is calculated, specifically including:

[0091] The total energy consumption expression for constructing the shield machine cutter system, propulsion system and slag discharge system is as follows:

[0092] E=T a Vt+T p w β t+T r w r t①

[0093] In formula ①, E is the total energy consumption of the shield machine, T a is the thrust of the shield machine, V is the propulsion speed, T p is the cutter head torque, w β is the cutter head speed, T r is the screw conveyor torque, w r is the speed of screw conveyor, and t is the excavation time.

[0094] The derivation process of formula ① is as follows:

[0095] The total energy consumption of the shield machine cutter system, propulsion system and slag discharge system is as follows:

[0096] E=E 推力 +E 扭矩 +E 渣土排放

[0097] formula Where E is the total energy consumption of the shield machine, E 推力 is the propulsion system energy consumption, E 扭矩 Energy consumption of the cutter system, E 渣土排放 Energy consumption of the slag discharge system.

[0098] Energy consumption of propulsion system E 推力 , Energy consumption of cutter system E 扭矩 , Energy consumption of slag discharge system E 渣土排放 The total energy consumption expression of the shield machine cutter head system, propulsion system and slag discharge system is obtained by analyzing them one by one.

[0099] In the shield tunneling process, the shield machine is taken as the research object, and its physical process can be expressed as the following formula:

[0100] T a -(F+M)=ma

[0101] Among them, T a is the thrust of the shield machine, F is the resistance formed during the excavation of the shield machine and its surrounding soil, M is the forward driving force of the shield machine, m is the weight of the shield machine, and a is the acceleration of the shield machine.

[0102] During the construction process, in order to maintain the stability and safety of the excavation face, the advancement speed is very slow, so the acceleration is almost zero, and the formula can be simplified to:

[0103] T a =F+M

[0104] That is, the thrust is equal to the sum of the resistance generated by the shield machine and the surrounding soil and the resistance required for the shield machine to move forward and cut.

[0105] The resistance F formed during the excavation of the shield machine and its surrounding soil includes the following four parts: the friction F1 between the shield machine and the surrounding soil, the resistance F2 caused by the friction between the rear of the shield machine and the segments, the resistance F3 caused by the supporting vehicles transporting the slag behind the shield machine, and the friction resistance F4 caused by the weight of the shield machine itself.

[0106] The pressure on the shield machine surface is Figure 2As shown in the figure, the calculation formula of the resistance F generated by the shield machine and its surroundings is:

[0107] F=F1+F2+F3+F4

[0108] The formula for the friction resistance F1 caused by the shield machine and the surrounding soil is:

[0109] F1=μ1LD(2(P v +P l ))

[0110] Where μ1 is the friction coefficient between the shield shell and the soil, L is the length of the shield machine, D is the outer diameter of the shield machine, and P v is the vertical earth pressure of the shield machine, P l is the horizontal earth pressure of the shield machine.

[0111] According to ground stress theory, the vertical earth pressure at a certain point is related to the lateral earth pressure coefficient at that point. Based on this relationship, the horizontal earth pressure on the shield surface can be obtained.

[0112] P l =P3+P4=kP v =k(P1+P2)

[0113] Among them, P3 is the lateral earth pressure at the top of the shield machine, P4 is the lateral earth pressure at the bottom of the shield machine, and k is the lateral earth pressure coefficient.

[0114] Therefore, the friction resistance F1 between the shield machine and the surrounding soil can also be obtained:

[0115] F1=2μ1LD[(γ-10)(2H f +R β )(1+k)]

[0116] Resistance F2 caused by friction between the rear of the shield machine and the segments:

[0117] F2=πR g μ g

[0118] Among them, R g is the outer diameter of the shield segment, μ g is the friction coefficient of the segment per meter.

[0119] Resistance F3 caused by the transportation of slag by the supporting vehicles behind the shield machine:

[0120] F3=G c μ2

[0121] Among them, G c is the weight of the rear matching vehicle, and μ2 is the friction coefficient between the rear matching vehicle and the track.

[0122] The friction resistance F4 caused by the weight of the shield machine itself; during the cutting and advancing process of the slurry shield machine, the cutting edge of the cutting ring will cut into the soil in front of the shield, thus generating penetration resistance at the cutting edge. The calculation formula is:

[0123] F4=μ1G

[0124] Among them, G is the weight of the shield machine, and μ1 is the friction coefficient between the shield shell and the soil.

[0125] The resistance generated by the shield machine and its surroundings is as follows:

[0126] F=μ1{2LD[(γ-10)(2H f +R β )(1+k)]+G}+πR g μ g +G c μ2

[0127] Where L is the length of the shield machine and D is the outer diameter of the shield machine.

[0128] The tunneling resistance M required for the shield machine to advance forward:

[0129]

[0130] Among them, P t is the earth pressure on the front of the shield machine, P s is the water pressure on the front of the shield machine.

[0131] Earth pressure P on the front of the shield machine t It can be expressed as: t =P5+P6

[0132] Among them, P5 is the soil pressure on the upper side of the shield machine, and P6 is the soil pressure on the lower side of the shield machine.

[0133] Shield machine front water pressure P s It can be expressed as:

[0134] P s =P7+P8

[0135] Among them, P7 is the water pressure on the upper side of the shield machine, and P8 is the water pressure on the lower side of the shield machine.

[0136] P7=10(H f -1)

[0137] P8=10(H f -1+R β )

[0138] Therefore, the tunneling resistance M of the soil on the shield machine cutting surface can be expressed as:

[0139]

[0140] According to the above derivation, the thrust of the shield machine is T a for:

[0141]

[0142] Mathematical model of energy consumption of cutter head drive system

[0143] The shield machine is mainly composed of 6 parts: the torque T1 formed by the rotation of the cutter disc between the front and the surrounding soil, the torque T2 formed by the rotation of the cutter disc between the side and the surrounding soil, the resistance torque T3 generated by the soil when the cutter disc rotates to cut the soil, the torque T4 caused by the friction between the cutter disc surface and the soil, the torque T5 caused by the cutter disc's own weight, and the friction torque T6 generated on the back of the cutter disc when the cutter disc rotates. The specific formula is as follows:

[0144]

[0145] The torque T1 generated by the rotation of the cutterhead between the front and the surrounding soil. During the tunneling process of the shield machine, friction will occur between the front of the cutterhead and the soil in front, thus forming a friction torque at the center of the cutterhead. In order to ensure the normal operation of the cutterhead, this torque needs to be overcome, and its calculation formula is as follows:

[0146]

[0147] in, is the cutter head opening ratio, μ β Indicates the friction coefficient between the front of the cutterhead and the soil, μ β The range is 0.07-0.1; P s is the horizontal earth pressure formed when the cutterhead rotates, P s is the water pressure on the front of the shield machine.

[0148] The torque T2 generated by the rotation of the cutterhead between its side and the surrounding soil, and the friction torque T2 between the rotating side of the shield cutterhead and the surrounding soil are calculated as follows:

[0149]

[0150] Where ρ is the soil density around the cutterhead, H is the vertical burial depth of the shield, W is the average length of the shield, K a is the active earth pressure coefficient of the surrounding soil, μ β It indicates the friction coefficient between the front of the cutterhead and the soil.

[0151] The calculation formula of the resistance torque T3 generated by the soil when the cutter disc rotates and cuts the soil is as follows:

[0152]

[0153] Where Q1 is the compressive strength of the cutter disc, h is the cutting depth of the cutter disc (the ratio of the advancing speed to the cutter disc speed), and D0 is the circumscribed radius of the cutter disc.

[0154] The calculation formula of torque T4 generated by friction between the cutter disc surface and the soil is as follows:

[0155]

[0156] Where K i is the soil shear strength, D1 is the outer diameter of the cutter head support beam, D2 is the inner diameter of the cutter head support beam, L z is the length of the cutter head support beam.

[0157] The calculation formula of torque T5 caused by the weight of the cutter head is as follows:

[0158]

[0159] In the formula, μ α is the friction coefficient when the cutter head rotates, G β is the mass of the shield cutterhead, and R1 is the contact radius of the cutterhead bearing.

[0160] The calculation formula of the friction torque T6 generated on the back of the cutter disc when the cutter disc rotates is as follows:

[0161]

[0162] Where σ is the cutter disc sealing ratio.

[0163] From the above derivation, we can know that:

[0164]

[0165] Mathematical model of energy consumption of slag

[0166] The torque of the screw conveyor is mainly composed of two parts, the friction torque T7 between the screw conveyor housing and the soil, and the torque T8 between the screw conveyor and the spiral blades during the slag discharge process;

[0167]

[0168] The friction torque T7 between the screw conveyor housing and the soil is expressed as:

[0169]

[0170] Among them, D l is the diameter of the screw conveyor of the shield machine, D g is the diameter of the screw conveyor, α is the inclination angle of the screw conveyor, L a is the length of the screw conveyor, λ is the density of the soil in the screw conveyor, f t is the friction coefficient between the screw conveyor housing and the surrounding soil.

[0171] The torque T8 between the screw conveyor and the spiral blade during the slag discharge process is expressed as:

[0172]

[0173] Where β is the helical rise angle.

[0174] It can be deduced that:

[0175]

[0176] In one embodiment, in step S2, relevant parameters affecting shield machine energy consumption are preliminarily screened, and outlier processing, noise reduction processing, standardization processing and correlation analysis of characteristic variables are performed on the data. Finally, characteristic parameters related to shield machine energy consumption prediction are output, specifically including:

[0177] S21. Screening of relevant parameters affecting shield energy consumption

[0178] The relevant parameters that affect the energy consumption of the shield machine include: shield machine cutter head system parameters, propulsion system parameters, and parameters related to the slag discharge system; they are: total thrust, cutter head torque, screw conveyor torque, penetration, cutter head speed, excavation speed, thrust of the i-th group of main push cylinders, i=7,...,n, and excavation chamber j pressure, j=n+1,...,m.

[0179] S22. Outlier processing: Use the 3σ principle to detect and remove outliers.

[0180] S23, denoising processing: using a wavelet threshold denoising method to denoise the original characteristic parameter signal;

[0181] It can be summarized into the following three steps: (1) perform wavelet transform on the noisy signal; (2) perform threshold processing on the wavelet coefficients; and (3) perform inverse wavelet transform to reconstruct the signal.

[0182] S24, dimensionless processing: use Min-Max normalization to process the feature parameter data after noise reduction.

[0183] Data dimensionless means that multiple parameters cannot be compared and analyzed due to inconsistent dimensions. Therefore, each parameter needs to be dimensionless to eliminate dimensional differences before being used for subsequent research and analysis. Min-Max normalization is also called deviation standardization, which processes the excavation data through linear transformation so that the final result of the data is standardized to the interval [0,1], where min refers to the minimum value in the data and max refers to the maximum value in the data. The original data sequence {x1, x2, x3,..., xn} is transformed as follows:

[0184]

[0185] Among them, the data sequence after dimensionless processing is ∈[0,1].

[0186] S25. Characteristic parameter correlation coefficient analysis: Use the Pearson statistical method to quantitatively analyze the dimensionless characteristic parameter data. The size of the linear correlation coefficient reflects the strength of the correlation between the two variables, and selects characteristic parameters with strong correlation with energy consumption. These parameters are used as characteristic parameters related to the energy consumption prediction of the shield machine.

[0187] In step S25, the correlation coefficient r∈[-1,1] is defined, r>0 indicates positive correlation, r<0 indicates negative correlation, and r=0 indicates nonlinear relationship, and the characteristic parameter with correlation coefficient r>0.5 is selected as the characteristic parameter with strong correlation with energy consumption.

[0188] The calculation formula is as follows:

[0189]

[0190] Where X and Y represent the two sets of parameters being analyzed; n represents the number of parameters.

[0191] Step 25 can use Python program to calculate the correlation coefficient and draw the correlation heat map, such as Figure 3 As shown, this figure can select characteristic parameters that are strongly correlated with energy consumption.

[0192] GA-LSTM is a hybrid algorithm that combines genetic algorithm (GA) and long short-term memory network (LSTM). GA is used to optimize the hyperparameters of LSTM, and LSTM is used to process stored data (such as time series). Through the global search capability of GA, the optimal LSTM hyperparameter combination can be automatically found, thereby improving the prediction ability of the model.

[0193] LSTM is a special recurrent neural network (RNN) that controls the flow of information through a gating mechanism (input gate, forget gate, output gate) and can effectively process long sequence data.

[0194] LSTM cell structure

[0195] enter:

[0196] The input x at the current time step t ;

[0197] The hidden state h at the previous time step t-1 ;

[0198] The cell state C at the previous time step t-1 ;

[0199] Output:

[0200] The hidden state h at the current time step t ;

[0201] The cell state C at the current time step t .

[0202] LSTM calculation process

[0203] Forget gate: determines which information is discarded from the cell state.

[0204] formula:

[0205] f t =σ(W f [h t-1 ,x t ]+b f )

[0206] Where: W f is the weight matrix of the forget gate, b f is the bias term of the forget gate, and σ is the sigmoid activation function.

[0207] Input gate: decides what new information will be stored in the cell state.

[0208] formula:

[0209] i t =σ(W i [h t-1 ,x t ]+b i )⑥

[0210]

[0211] Where: W i is the weight matrix of the input gate, b i is the bias term of the input gate, W C is the weight matrix of the candidate cell state, b C is a candidate cell state.

[0212] Output gate: determines what information will be output to the hidden state.

[0213] formula:

[0214] o t =σ(W o [h t-1 ,x t ]+b o ⑧

[0215] h t =o t tanh(C t ) ⑨

[0216] Where: W o is the weight matrix of the output gate, b i is the bias term of the output gate.

[0217] Genetic algorithm is an optimization algorithm based on natural selection and genetic mechanism, which can globally optimize the problem. In GA-LSTM, GA is used to optimize the number of hidden layer neurons, the number of LSTM layers, the learning rate, the dropout rate and the batch size of the LSTM neural network, thereby improving the prediction accuracy.

[0218] The GA optimization of the hyper parameters of LSTM in step S4 specifically includes:

[0219] 1. Input and Output

[0220] Input variables: X1, X2, X3, ..., Xi; they represent the excavation speed, penetration rate, total thrust, excavation chamber pressure, etc.

[0221] Input variables: Y1 Total energy consumption per unit time (e.g. kWh)

[0222] 2. Hyperparameter definition. The LSTM hyperparameters that need to be optimized include:

[0223] Number of hidden layer neurons: [32, 64, 128, 256]

[0224] Learning rate: [0.1, 0.2, 0.3, 0.4, 0.5]

[0225] Dropout rate: [0.1, 0.2, 0.3, 0.4, 0.5]

[0226] Batch size: [16,32,64,128]

[0227] Number of LSTM layers: [1,2,3]

[0228] 3. Coding

[0229] Encode hyperparameters as chromosomes. For example:

[0230] Chromosome: [64, 0.001, 0.2, 32, 2], which means: the number of hidden layer neurons is 64, the learning rate is 0.001, the Dropout rate is 0.2, the batch size is 32, and the LSTM is 2.

[0231] 4.GA parameter settings

[0232] Set the basic parameters of the genetic algorithm (GA):

[0233] Population size (usually 20-100);

[0234] Crossover probability (e.g. 0.6-0.9);

[0235] Mutation probability (e.g. 0.01-0.1);

[0236] Maximum number of iterations: 50-100 or fitness convergence threshold: 10 -4 -10 -6 .

[0237] 5. Initialize the population

[0238] The initial weights and thresholds of the neural network are randomly generated as population individuals.

[0239] 6. Fitness Function

[0240] The LSTM model is trained using the training data and the accuracy (mean square error MSE) is evaluated on the validation set as the fitness value.

[0241] formula:

[0242] Where N is the number of samples, y i ′ is the predicted value of the i-th sample, y i is the true value of the i-th sample.

[0243] 7. Genetic manipulation

[0244] selection, crossover, and mutation;

[0245] Select excellent individuals based on fitness values;

[0246] Perform crossover operation on the selected individuals to generate new individuals;

[0247] Perform mutation operations on some new individuals to increase population diversity.

[0248] 8. Update population

[0249] The new offspring replace the parents to form a new generation of population.

[0250] 9. Termination Conditions

[0251] Check whether the termination condition is met (such as reaching the maximum number of iterations or fitness value convergence). If not, go to step 6, otherwise, go to step 10.

[0252] 10. Output the optimal solution

[0253] Output the LSTM hyperparameters corresponding to the individual with the highest fitness.

[0254] 11. Train the final model

[0255] The final LSTM model is trained using the optimal hyperparameters to perform energy consumption prediction.

[0256] The optimization goal is to minimize the energy consumption of the shield machine and maximize the tunneling efficiency of the shield machine. The optimal solution is obtained by MATLAB's fmincon (fmincon is a general nonlinear optimization tool provided by MATLAB, suitable for continuous optimization problems with constraints). The optimal solution results are shown in Table 1:

[0257] Optimal parameters Solved value Energy consumption (J) <![CDATA[2.92×10 5 ]]> <![CDATA[Thrust force T of shield machine a (kN)]]> 14800 Propulsion speed V (cm / min) 7.31 <![CDATA[Cutting head rotational speed w β (rpm)]]> 0.42 Shield vertical burial depth H (mm / rev) 60.89 Cutting depth of cutter disc h(m) 23.93 <![CDATA[Spiral conveyor speed w r (rpm)]]> 11.94 <![CDATA[Torque T of screw conveyor r (kNm)]]> 97.81

[0258] Table 1

[0259] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A shield machine energy consumption prediction and optimization method, characterized in that: The following steps are involved: S1. Conduct a comprehensive energy consumption analysis on the cutterhead system, propulsion system and slag discharge system of the shield machine, and calculate the total energy consumption of the cutterhead system, propulsion system and slag discharge system of the shield machine; S2. Preliminarily screen the relevant parameters that affect the shield machine energy consumption, and perform outlier processing, noise reduction, standardization and correlation analysis of characteristic variables on the data, and finally output the characteristic parameters related to the shield machine energy consumption prediction; S3, constructing a shield machine energy consumption prediction model with characteristic parameters related to shield machine energy consumption prediction as input, shield machine energy consumption as prediction target, and LSTM as basic structure; S4. GA is used to optimize the hyperparameters of LSTM, and an energy consumption prediction model based on GA-LSTM is established to predict the energy consumption of the shield machine.

2. A shield machine energy consumption prediction method according to claim 1, characterized in that: The comprehensive energy consumption analysis of the cutter head system, propulsion system and slag discharge system of the shield machine in step S1 specifically includes: The total energy consumption expression for constructing the shield machine cutter system, propulsion system and slag discharge system is as follows: E=T a Vt+T p w β t+T r w r t① In formula ①, E is the total energy consumption of the shield machine, T a is the thrust of the shield machine, T p is the propulsion speed, T p is the cutter head torque, w β is the cutter head speed, T r is the screw conveyor torque, w r is the speed of screw conveyor, and t is the excavation time.

3. A shield machine energy consumption prediction and optimization method according to claim 2, characterized in that: Use formula ② to calculate the thrust T of the shield machine a ; In formula ②, T a is the thrust of the shield machine, μ1 is the friction coefficient between the shield shell and the soil, L is the length of the shield machine, D is the outer diameter of the shield machine, γ is the soil volume, H f is the cover length, R β is the cutterhead diameter, k is the lateral earth pressure coefficient, G is the weight of the shield machine, R g is the outer diameter of the shield segment, μ g is the friction coefficient of each meter of the segment, G c is the weight of the rear matching vehicle, and μ2 is the friction coefficient between the rear matching vehicle and the track.

4. A shield machine energy consumption prediction and optimization method according to claim 2, characterized in that: Use formula ③ to calculate the cutter head torque T p ; In formula ③, is the cutter head opening ratio, μ β Indicates the friction coefficient between the front of the cutterhead and the soil, μ β The range is 0.07-0.1, P s is the horizontal earth pressure formed when the cutterhead rotates, ρ is the soil density around the cutterhead, H is the vertical burial depth of the shield, W is the average length of the shield, K a is the active earth pressure coefficient of the surrounding soil, Q1 is the compressive strength of the cutter head, h is the cutting depth of the cutter head, D0 is the external radius of the cutter head, K i is the soil shear strength, D1 is the outer diameter of the cutter head support beam, D2 is the inner diameter of the cutter head support beam, L z is the length of the cutter head support beam, μ α is the friction coefficient when the cutter head rotates, G β is the mass of the shield cutterhead, R1 is the contact radius of the cutterhead bearing, and σ is the cutterhead sealing ratio.

5. A shield machine energy consumption prediction and optimization method according to claim 2, characterized in that: Use formula ④ to calculate the screw conveyor torque T r ; In formula ④, D l is the diameter of the screw conveyor of the shield machine, D g is the diameter of the screw conveyor, α is the inclination angle of the screw conveyor, L a is the length of the screw conveyor, λ is the density of the soil in the screw conveyor, f t is the friction coefficient between the screw conveyor housing and the surrounding soil, and β is the spiral rise angle.

6. A shield machine energy consumption prediction and optimization method according to claim 1, characterized in that: In step S2, the relevant parameters affecting shield machine energy consumption are preliminarily screened, and outlier processing, noise reduction processing, standardization processing and correlation analysis of characteristic variables are performed on the data. Finally, the characteristic parameters related to shield machine energy consumption prediction are output, which specifically include: S21. Screening of relevant parameters affecting shield energy consumption The parameters that affect the energy consumption of shield machines include: shield machine cutter system parameters, propulsion system parameters, and parameters related to the slag discharge system; they are: total thrust X1, cutter torque X2, screw conveyor torque X3, penetration X4, cutter speed X5, excavation speed X6, i-group main push cylinder thrust X i , i = 7, ..., n, excavation chamber j pressure X j , j = n + 1, ..., m; S22. Outlier processing: Use the 3σ principle to detect and remove outliers; S23, denoising processing: using a wavelet threshold denoising method to denoise the original characteristic parameter signal; S24, dimensionless processing: using Min-Max normalization to process the characteristic parameter data after noise reduction; S25. Characteristic parameter correlation coefficient analysis: Use the Pearson statistical method to quantitatively analyze the dimensionless characteristic parameter data. The size of the linear correlation coefficient reflects the strength of the correlation between the two variables, and selects characteristic parameters with strong correlation with energy consumption. These parameters are used as characteristic parameters related to the energy consumption prediction of the shield machine.

7. A shield machine energy consumption prediction and optimization method according to claim 6, characterized in that: In the step S25, a correlation coefficient r∈[-1,1] is defined, where r>0 indicates a positive correlation, r<0 indicates a negative correlation, and r=0 indicates a nonlinear relationship, and a characteristic parameter with a correlation coefficient r>0.5 is selected as a characteristic parameter with a strong correlation with energy consumption.

8. A shield machine energy consumption prediction and optimization method according to claim 7, characterized in that: The hyper parameters of LSTM optimized by GA in step S4 specifically include: Input and Output Input variables: X1, X2, X3, ..., X i ; X1, X2, X3, ..., X i They represent characteristic parameters related to the prediction of shield machine energy consumption; Input variables: Y1 total energy consumption per unit time; Hyperparameter Definition The LSTM hyperparameters that need to be optimized include: Number of hidden layer neurons: [32, 64, 128, 256] Learning rate: [0.1, 0.2, 0.3, 0.4, 0.5] Dropout rate: [0.1, 0.2, 0.3, 0.4, 0.5] Batch size: [16,32,64,128] Number of LSTM layers: [1,2,3] coding The hyperparameters are encoded as chromosomes, with 64 hidden layer neurons, 0.001 learning rate, 0.2 dropout rate, 32 batch size, and 2 LSTMs; GA parameter settings Set the basic parameters of GA: Population size: 20-100; Crossover probability: 0.6-0.9; Mutation probability: 0.01-0.1; Maximum number of iterations: 50-100 or fitness convergence threshold: 10 -4 -10 -6 ; (1) Initialize the population Randomly generate the initial weights and thresholds of the neural network as population individuals; (2) Fitness function Use the training data to train the LSTM model and evaluate the accuracy on the validation set, using the mean square error (MSE) as the fitness value. In formula ⑤, N is the number of samples, y i ′ is the predicted value of the i-th sample, y i is the true value of the i-th sample; (3) Genetic manipulation Selection, crossover, and mutation: Select excellent individuals based on fitness values; Perform crossover operation on the selected individuals to generate new individuals; Perform mutation operations on some new individuals to increase population diversity; (4) Update population Replace the parent generation with the newly produced offspring to form a new generation of population; (5) Termination conditions Check whether the termination condition is reached: if the maximum number of iterations or fitness value is not reached, perform genetic operations, otherwise, enter the output optimal solution; (6) Output the optimal solution Output the LSTM hyperparameters corresponding to the individual with the highest fitness; (7) Training the final model The final shield machine energy consumption prediction model is trained using the optimal hyperparameters.

Images