Intelligent shield tunneling control system based on machine learning

Abstract

The invention discloses an intelligent shield tunneling control system based on machine learning, and relates to the technical field of shield tunneling. The method comprises the following steps: acquiring waste residue parameters, tunneling parameters and equipment parameters, constructing a digital twin model, acquiring the hardness of waste residues, acquiring tunneling coefficients of shield tunneling machines by combining the hardness of the waste residues with the waste residue parameters, constructing a first calculation model to acquire real-time tunneling coefficients and variation trends thereof, and performing simulation adjustment on the shield tunneling machines under different variation trends in the digital twin model, simulation tunneling parameters and simulation equipment parameters are obtained, an adjustment coefficient is obtained, an optimal adjustment scheme is generated, and a second calculation model is constructed; a targeted control mechanism can be formed according to the real-time state of the shield tunneling machine, and the overall efficiency of shield tunneling machine control can be obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of shield tunneling, and specifically to a shield intelligent tunneling control system based on machine learning. Background Art

[0002] A shield machine is an important device for tunnel construction. Its tunneling process is affected by various factors, including geological conditions, tunneling parameters, equipment status, etc. Traditional shield tunneling control mainly relies on manual experience, making it difficult to adapt to complex and changeable construction environments, resulting in low construction efficiency, poor safety, and high energy consumption. As tunnel projects develop towards complex terrains, the limitations of traditional control methods become increasingly prominent; In the prior art, means for controlling a shield machine using artificial intelligence technology have emerged, but there are still the following problems. First, it fails to comprehensively judge the difficulty of shield machine tunneling based on different data, so a targeted control mechanism cannot be formed. Second, the large amount of data processing leads to a low overall efficiency of controlling the shield machine. In view of the deficiencies of the prior art, the present invention provides a shield intelligent tunneling control system based on machine learning. Summary of the Invention

[0003] The purpose of the present invention is to provide a shield intelligent tunneling control system based on machine learning.

[0004] The purpose of the present invention can be achieved through the following technical solutions: A shield intelligent tunneling control system based on machine learning includes the following modules: A data acquisition module, which is used to obtain the waste residue parameters, tunneling parameters, and equipment parameters of the shield machine during its tunneling process, construct a corresponding digital twin model, sample the waste residue generated during the tunneling process and obtain the waste residue hardness, and combine the waste residue parameters to obtain the tunneling coefficient of the shield machine; A first calculation module, which is used to construct a corresponding first calculation model based on the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficient at different times; A data analysis module, which is used to obtain the real-time tunneling coefficient using the first calculation model, obtain the corresponding change trend, perform simulation adjustment on the shield machine under different change trends in the digital twin model, and obtain the adjusted simulation tunneling parameters and simulation equipment parameters; A data evaluation module, which is used to obtain the corresponding adjustment coefficient based on the simulation equipment parameters under different simulation tunneling parameters and generate an optimal adjustment plan; A second calculation module, configured to construct corresponding second calculation models according to the tunneling parameters, equipment parameters, real-time tunneling coefficients, and simulated tunneling parameters of different optimal adjustment schemes, continuously input the current tunneling parameters, equipment parameters, and real-time tunneling coefficients into the second calculation models to obtain the current simulated tunneling parameters, and control the shield machine to tunnel according to the current simulated tunneling parameters.

[0005] Further, the process of obtaining the waste residue parameters, tunneling parameters, and equipment parameters of the shield machine during its tunneling process and constructing the corresponding digital twin model includes: The waste residue parameters include the amount of waste residue and the rock block size distribution. The rock block size distribution includes the large block rate and the small block rate. The tunneling parameters include the thrust, cutterhead rotation speed, and tunneling speed. The equipment parameters include the temperature, energy consumption, vibration signal, and hydraulic pressure; The waste residue parameters, tunneling parameters, and equipment parameters all have their corresponding acquisition times. The digital twin model is constructed according to each parameter using digital twin technology, and the digital twin model can simulate each parameter of the shield machine during its tunneling process synchronously.

[0006] Further, the process of sampling the waste residue generated during the tunneling process and obtaining the waste residue hardness, and obtaining the tunneling coefficient of the shield machine in combination with the waste residue parameters includes: Sampling the waste residue corresponding to the waste residue parameters at a single acquisition time to obtain a number of rock samples, processing them into regular shapes, and using the Brinell hardness test method to obtain the hardness values of each rock sample respectively. Taking the average value of the hardness values of each rock sample as the waste residue hardness at this acquisition time; Denote the amount of waste residue, large block rate, small block rate, and waste residue hardness at the same acquisition time as Z a 、Z b 、Z c 、Z d , and obtain the tunneling coefficient S of the shield machine at this acquisition time; ; are the preset weight values for the amount of waste residue, large block rate, small block rate, and waste residue hardness respectively. The tunneling coefficient is used to reflect the difficulty of the shield machine tunneling at the corresponding acquisition time, and the tunneling coefficients of the shield machine at different acquisition times are obtained respectively.

[0007] Further, the process of constructing the corresponding first calculation model according to the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficients at different times includes: Generating a first calculation set according to the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficients at different acquisition times, and dividing it into a first training set and a first test set; Construct a first convolutional neural network, take different waste residue parameters, tunneling parameters, and equipment parameters in the first training set as the input data of the first convolutional neural network, and take the corresponding tunneling coefficient in the first training set as the output data of the first convolutional neural network; Train the first convolutional neural network to obtain an initial first convolutional neural network, use the first test set to verify the model of the initial first convolutional neural network, and output the initial first convolutional neural network with an error less than or equal to the preset first test error threshold as the first calculation model.

[0008] Further, the process of using the first calculation model to obtain the real-time tunneling coefficient and obtain the corresponding change trend includes: Continuously input the current waste residue parameters, tunneling parameters, and equipment parameters into the first calculation model to obtain the corresponding real-time tunneling coefficient, and sequentially record the n most recently obtained real-time tunneling coefficients as S i , i = 1, 2,..., n, and obtain the change coefficient B of each real-time tunneling coefficient i ; ; Obtain the corresponding change trend according to the change coefficient of the obtained real-time tunneling coefficient, including an upward trend, a downward trend, and a constant trend.

[0009] Further, the process of respectively simulating and adjusting the shield machine under different change trends in the digital twin model and obtaining the adjusted simulated tunneling parameters and simulated equipment parameters includes: When the real-time tunneling coefficient shows an upward trend, increase the thrust and cutter head speed of the shield machine in the digital twin model. When the real-time tunneling coefficient shows a downward trend, decrease the thrust and cutter head speed of the shield machine in the digital twin model; When the real-time tunneling coefficient shows a constant trend, do not adjust it. The simulation adjustment has different adjustment amplitudes, and take the adjusted thrust and cutter head speed as the simulated tunneling parameters; Obtain the temperature, energy consumption, and tunneling speed of the shield machine under different adjusted simulated tunneling parameters in the digital twin model, and record them as simulated equipment parameters. Take a single simulated tunneling parameter and its simulated equipment parameters as an adjustment plan.

[0010] Further, the process of obtaining the corresponding adjustment coefficient according to the simulated equipment parameters under different simulated tunneling parameters and generating an optimal adjustment plan includes: Record the temperature, energy consumption, and tunneling speed of the simulated equipment parameters in a single adjustment plan as H a 、H b 、H c , and record the temperature, energy consumption, and tunneling speed of the shield machine before adjustment as Q a 、Qb and Q c to obtain the adjustment coefficient R of this adjustment scheme; ; The adjustment coefficient is used to reflect the operating conditions of the shield machine under the simulation tunneling coefficients with different adjustment amplitudes. The adjustment coefficients of different adjustment schemes are obtained at the same acquisition moment, and the adjustment scheme with the largest adjustment coefficient is used as the optimal adjustment scheme.

[0011] Furthermore, the process of constructing the corresponding second calculation model according to the tunneling parameters, equipment parameters, real-time tunneling coefficient, and simulation tunneling parameters of different optimal adjustment schemes includes: Generating a second calculation set based on the simulation tunneling parameters, the tunneling parameters before adjustment, equipment parameters, and real-time tunneling coefficient of the optimal adjustment scheme at different acquisition moments, and dividing it into a second training set and a second test set; Constructing a second convolutional neural network, using different tunneling parameters, equipment parameters, and real-time tunneling coefficient in the second training set as the input data of the second convolutional neural network, and using the corresponding simulation tunneling parameters in the second training set as the output data of the second convolutional neural network; Training the second convolutional neural network to obtain an initial second convolutional neural network, validating the model of the initial second convolutional neural network using the second test set, and outputting the initial second convolutional neural network with an error less than or equal to the preset second test error threshold as the second calculation model.

[0012] Compared with the prior art, the beneficial effects of the present invention are: By obtaining the waste residue parameters of the shield machine during its tunneling process and measuring the hardness of the generated waste residue, and combining the two to obtain the tunneling coefficient of the shield machine, it can effectively reflect the difficulty of tunneling of the shield machine at different moments. Based on this, the shield machine is simulated and adjusted in the digital twin model, and the corresponding adjustment coefficient is obtained. The largest one is used as the optimal adjustment scheme, which can form a targeted control mechanism for the real-time state of the shield machine; By using machine learning algorithms to construct the first calculation model and the second calculation model, the complex calculation process can be directly obtained through the calculation model, which is beneficial to shortening the time required for controlling the shield machine and can significantly improve the overall efficiency of shield machine control. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the modules of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0014] As Figure 1 shown, a shield intelligent tunneling control system based on machine learning includes the following modules: A data acquisition module, which is used to obtain the waste residue parameters, tunneling parameters, and equipment parameters of a shield machine during its tunneling process, construct a corresponding digital twin model, sample the waste residue generated during the tunneling process and obtain the hardness of the waste residue, and obtain the tunneling coefficient of the shield machine in combination with the waste residue parameters; A first calculation module, which is used to construct a corresponding first calculation model according to the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficient at different times; A data analysis module, which is used to obtain the real-time tunneling coefficient by using the first calculation model, obtain the corresponding change trend, perform simulation adjustment on the shield machine under different change trends in the digital twin model, and obtain the adjusted simulated tunneling parameters and simulated equipment parameters; A data evaluation module, which is used to obtain the corresponding adjustment coefficient according to the simulated equipment parameters under different simulated tunneling parameters and generate an optimal adjustment plan; A second calculation module, which is used to construct a corresponding second calculation model according to the tunneling parameters, equipment parameters, real-time tunneling coefficient, and simulated tunneling parameters of different optimal adjustment plans, continuously input the current tunneling parameters, equipment parameters, and real-time tunneling coefficient into the second calculation model to obtain the current simulated tunneling parameters, and control the shield machine to tunnel according to the current simulated tunneling parameters.

[0015] It should be further noted that in the specific implementation process, the process of obtaining the waste residue parameters, tunneling parameters, and equipment parameters of the shield machine during its tunneling process and constructing the corresponding digital twin model includes: The waste residue parameters refer to the relevant parameters of the waste residue generated by the tunneling of the shield machine, including the waste residue volume and the rock block size distribution. The waste residue volume refers to the total volume of the waste residue generated per unit time, and the rock block size distribution refers to the size distribution of the rock blocks in the waste residue generated per unit time, which includes the large block rate and the small block rate; The tunneling parameters refer to the relevant parameters for controlling the shield machine during its tunneling process, including the thrust, cutter head rotation speed, and tunneling speed. The thrust refers to the power for the shield machine to move forward, the cutter head rotation speed refers to the rotation speed of the cutter head, and the tunneling speed refers to the distance the shield machine advances per unit time; The equipment parameters refer to the relevant parameters reflecting the operating state of the shield machine during its tunneling process, including the temperature, energy consumption, vibration signal, and hydraulic pressure. The temperature refers to the main bearing temperature of the shield machine, the energy consumption refers to the total amount of energy consumed by the shield machine per unit time, the vibration signal refers to the vibration frequency and vibration amplitude of the shield machine, and the hydraulic pressure refers to the pressure of the hydraulic system of the shield machine; The obtained waste residue parameters, tunneling parameters, and equipment parameters all have their corresponding acquisition times. Using digital twin technology, corresponding digital twin models are constructed based on the obtained relevant parameters. The digital twin models can simultaneously simulate various relevant parameters of the shield machine during its tunneling process; The digital twin technology is a cross - domain comprehensive technology that closely connects the physical world and the virtual world. By means of digitalization, a virtual mapping of physical entities is created to achieve real - time monitoring, analysis, prediction, and optimization of physical entities.

[0016] It should be further noted that in the specific implementation process, the process of sampling the waste residue generated during the tunneling process and obtaining the waste residue hardness, and combining the waste residue parameters to obtain the tunneling coefficient of the shield machine includes: For the waste residue parameters, the acquisition time is the end time of the unit time. The waste residue corresponding to the waste residue parameters at a single acquisition time is sampled to obtain a number of rock samples, and the rock samples need to cover different rock layers and volumes; The obtained rock samples are processed into regular shapes, such as cubes or cylinders, to meet the requirements of hardness testing. The Brinell hardness testing method is used to obtain the hardness values of each rock sample, and the average value of the hardness values of each rock sample is used as the waste residue hardness at this acquisition time; The waste residue volume, large - block ratio, small - block ratio, and waste residue hardness at the same acquisition time are respectively denoted as Z a 、Z b 、Z c 、Z d , and the tunneling coefficient of the shield machine at this acquisition time is obtained, denoted as S; ; Among them, are respectively the preset weight values of the waste residue volume, large - block ratio, small - block ratio, and waste residue hardness. The tunneling coefficient is used to reflect the difficulty of tunneling of the shield machine at the corresponding acquisition time. The smaller the tunneling coefficient, the easier it is to tunnel in the current geology. The larger the tunneling coefficient, the more difficult it is to tunnel in the current geology; In the embodiments of the present invention, the tunneling coefficient is obtained to evaluate the difficulty of tunneling of the shield machine in the face of different geologies. Therefore, the present invention performs weighted calculation on the parameters that can significantly affect or reflect the difficulty of tunneling, and then obtains a comprehensive tunneling coefficient; On the premise of not considering other parameters, only consider the parameters that can affect and reflect the difficulty of tunneling in terms of geology, such as waste residue volume, large - block ratio, small - block ratio, and waste residue hardness; First, the high or low amount of waste residue will indirectly reflect whether the geology is easy to tunnel. The easier the geology is to tunnel, the higher the amount of waste residue per unit time; the more difficult the geology is to tunnel, the lower the amount of waste residue per unit time. It can be seen that the amount of waste residue is negatively correlated with the tunneling coefficient. Second, the high or low of the large block rate and small block rate also indirectly reflects whether the geology is easy to tunnel. The easier the geology is to tunnel, the higher the small block rate and the lower the large block rate due to the soft geology; the more difficult the geology is to tunnel, the lower the small block rate and the higher the large block rate due to the hard geology. It can be seen that the large block rate is positively correlated with the tunneling coefficient, and the small block rate is negatively correlated with the tunneling coefficient. Third, the hardness of the waste residue directly reflects the hardness of the rock layer tunnelled by the shield machine, and thus will directly reflect whether the geology is easy to tunnel. The lower the hardness of the waste residue, the easier the geology is to tunnel; the higher the hardness of the waste residue, the more difficult the geology is to tunnel. It can be seen that the hardness of the waste residue is positively correlated with the tunneling coefficient. Since the present invention expects that the smaller the tunneling coefficient, the easier the current geology is to tunnel, and the larger the tunneling coefficient, the more difficult the current geology is to tunnel. Therefore, when performing weighted calculation, the parameters positively correlated with the tunneling coefficient can be directly multiplied by the corresponding weight value, while the parameters negatively correlated with the tunneling coefficient need to take their fractions and then multiply by the corresponding weight value. The above is the acquisition logic of the tunneling coefficient. In the actual application scenario, the weight values of the waste residue amount, large block rate, small block rate, and waste residue hardness are preset as 0.3, 0.2, 0.1, and 0.4 respectively. Now, 3 groups of experimental data are provided. Due to different parameter units and ranges, it is necessary to perform normalization processing on them. The process of the normalization processing will not be elaborated here. The normalized parameters are used as the input values when calculating the tunneling coefficient, that is, Z a 、Z b 、Z c 、Z d ; Sample a: waste residue amount 150 m³, large block rate 70%, small block rate 15%, waste residue hardness 6. After normalization processing, they are 0.125, 0.833, 0.1, and 0.75 respectively, and the corresponding tunneling coefficient is about 3.87. Sample b: waste residue amount 300 m³, large block rate 50%, small block rate 30%, waste residue hardness 5; after normalization processing, they are 0.5, 0.5, 0.4, and 0.5 respectively, and the corresponding tunneling coefficient is about 1.15. Sample c: waste residue amount 200 m³, large block rate 60%, small block rate 20%, waste residue hardness 6; after normalization processing, they are 0.25, 0.667, 0.2, and 0.75 respectively, and the corresponding tunneling coefficient is about 2.13.

[0017] It should be further noted that in the specific implementation process, the process of constructing the corresponding first calculation model according to the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficient at different times includes: Since the waste residue parameters, tunneling parameters, and equipment parameters can all be directly obtained through sensors, while the waste residue hardness needs to be determined through experiments, the tunneling coefficient cannot be directly obtained; However, when the tunneling parameters are certain, the tunneling conditions under different tunneling coefficients will be directly reflected in the waste residue parameters and equipment parameters, that is, there is a certain corresponding relationship, so that the tunneling coefficient can be directly obtained without obtaining the waste residue hardness; Obtain the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficient at different collection times. At this time, the obtained tunneling coefficient is obtained by using the waste residue hardness. Generate the first calculation set according to the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficient at different collection times, and divide it into a first training set and a first test set; Construct a first convolutional neural network, use the different waste residue parameters, tunneling parameters, and equipment parameters in the first training set as the input data of the first convolutional neural network, and use the corresponding tunneling coefficient in the first training set as the output data of the first convolutional neural network; Train the first convolutional neural network to obtain an initial first convolutional neural network, use the first test set to verify the model of the initial first convolutional neural network, and output the initial first convolutional neural network with an error less than or equal to the preset first test error threshold as the first calculation model; The convolutional neural network is a deep learning model specifically used to process complex data. It belongs to the field of artificial intelligence technology. By constructing and training neural network models with different levels, it can automatically learn complex patterns and features from a large amount of data. It is one of the most widely used models at present. In the embodiments of the present invention, the constructed convolutional neural network is respectively marked as the first convolutional neural network and the second convolutional neural network described below to achieve distinction.

[0018] It should be further noted that in the specific implementation process, the process of using the first calculation model to obtain the real-time tunneling coefficient and obtain the corresponding change trend includes: Continuously input the current waste residue parameters, tunneling parameters, and equipment parameters into the first calculation model to obtain the corresponding real-time tunneling coefficient, and record the n most recently obtained real-time tunneling coefficients in the order of acquisition as S i , i = 1, 2,..., n, and obtain the change coefficient B of each real-time tunneling coefficient i ; ; When B i > B i-1 And , its corresponding real-time tunneling coefficient is marked as an upward trend. When B i < B i-1 and , its corresponding real-time tunneling coefficient is marked as a downward trend; When B i > B i-1 and , its corresponding real-time tunneling coefficient is marked as a downward trend. When B i < B i-1 and , its corresponding real-time tunneling coefficient is marked as an upward trend; The real-time tunneling coefficients in other cases except the above are all marked as a constant trend. The change trends include an upward trend, a downward trend, and a constant trend.

[0019] It should be further noted that in the specific implementation process, the process of separately simulating and adjusting the shield machines under different change trends in the digital twin model and obtaining the adjusted simulated tunneling parameters and simulated equipment parameters includes: When the real-time tunneling coefficient is in an upward trend, it means that the current geology is not easy to tunnel. In the digital twin model, the thrust and cutter head speed of the shield machine are both increased; When the real-time tunneling coefficient is in a downward trend, it means that the current geology is easy to tunnel. In the digital twin model, the thrust and cutter head speed of the shield machine are both decreased; When the real-time tunneling coefficient is in a constant trend, its thrust and cutter head speed are not adjusted. The simulation adjustment includes increasing, decreasing, and not adjusting the tunneling parameters. The simulation adjustment has different adjustment amplitudes, and the adjustment amplitude is used to reflect the numerical value of increasing or decreasing the tunneling parameters; The adjusted thrust and cutter head speed are used as the simulated tunneling parameters. In the digital twin model, the temperature, energy consumption, and tunneling speed of the shield machine under different adjustment amplitudes of the simulated tunneling parameters are obtained and recorded as the simulated equipment parameters. A single simulated tunneling parameter and its simulated equipment parameters are used as an adjustment plan.

[0020] It should be further noted that in the specific implementation process, the process of obtaining the corresponding adjustment coefficient according to the simulated equipment parameters under different simulated tunneling parameters and generating the optimal adjustment plan includes: Taking any adjustment plan as an example, the temperature, energy consumption, and tunneling speed in its simulated equipment parameters are respectively recorded as H a , H b , H c , and the temperature, energy consumption, and tunneling speed of the shield machine before adjustment are respectively recorded as Q a , Q b , Q c, obtain the adjustment coefficient of this adjustment plan, denoted as R; ; The adjustment coefficient is used to reflect the operating conditions of the shield machine under the simulation tunneling coefficients with different adjustment amplitudes. The larger the adjustment coefficient, the more stable the operating conditions before and after adjustment; the smaller the adjustment coefficient, the more unstable the operating conditions before and after adjustment; Adopt the same method to obtain the adjustment coefficients corresponding to different adjustment plans, and take the adjustment plan with the largest adjustment coefficient as the optimal adjustment plan; In the embodiments of the present invention, by obtaining the adjustment coefficient to evaluate the stability degree of the operating conditions of the shield machine under different adjustment plans, therefore, the present invention calculates the weighted sum of the differences of the various parameters that can reflect the operating conditions before and after adjustment, and then obtains a comprehensive adjustment coefficient; If the difference in the temperature of the shield machine before and after adjustment is smaller, it means that the operating conditions of the shield machine under the corresponding adjustment plan are more stable. Therefore, the temperature difference of the shield machine is negatively correlated with the adjustment coefficient. Similarly, the energy consumption difference and the tunneling speed difference of the shield machine before and after adjustment are also negatively correlated with the adjustment coefficient; Since the present invention expects that the larger the adjustment coefficient, the more stable the operating conditions before and after adjustment, and the smaller the adjustment coefficient, the more unstable the operating conditions before and after adjustment. Therefore, when performing weighted calculation, the difference that is negatively correlated with the adjustment coefficient needs to be taken as its fraction and then multiplied by the corresponding weight value. The above is the acquisition logic of the adjustment coefficient.

[0021] It should be further noted that in the specific implementation process, the process of constructing the corresponding second calculation model according to the tunneling parameters, equipment parameters, real-time tunneling coefficient, and simulation tunneling parameters of different optimal adjustment plans includes: Generate a second calculation set according to the simulation tunneling parameters, the tunneling parameters before adjustment, the equipment parameters, and the real-time tunneling coefficient of different optimal adjustment plans, and divide it into a second training set and a second test set; Construct a second convolutional neural network, use the different tunneling parameters, equipment parameters, and real-time tunneling coefficient in the second training set as the input data of the second convolutional neural network, and use the corresponding simulation tunneling parameters in the second training set as the output data of the second convolutional neural network; Train the second convolutional neural network to obtain an initial second convolutional neural network, and use the second test set to verify the model of the initial second convolutional neural network, and output the initial second convolutional neural network with an error less than or equal to the preset second test error threshold as the second calculation model.

[0022] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A shield intelligent tunneling control system based on machine learning, characterized in that, It includes the following modules: A data acquisition module, which is used to obtain the waste residue parameters, tunneling parameters, and equipment parameters of the shield machine during its tunneling process, construct a corresponding digital twin model, sample the waste residue generated during the tunneling process and obtain the hardness of the waste residue, and obtain the tunneling coefficient of the shield machine in combination with the waste residue parameters; A first calculation module, which is used to construct a corresponding first calculation model according to the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficient at different times; A data analysis module, which is used to obtain the real-time tunneling coefficient by using the first calculation model, obtain the corresponding change trend, perform simulation adjustment on the shield machine under different change trends in the digital twin model, and obtain the adjusted simulated tunneling parameters and simulated equipment parameters; A data evaluation module, which is used to obtain the corresponding adjustment coefficient according to the simulated equipment parameters under different simulated tunneling parameters and generate an optimal adjustment plan; A second calculation module, which is used to construct a corresponding second calculation model according to the tunneling parameters, equipment parameters, real-time tunneling coefficient, and simulated tunneling parameters of different optimal adjustment plans, continuously input the current tunneling parameters, equipment parameters, and real-time tunneling coefficient into the second calculation model to obtain the current simulated tunneling parameters, and control the shield machine to tunnel according to the current simulated tunneling parameters.

2. The shield intelligent tunneling control system based on machine learning according to claim 1, characterized in that, The process of obtaining the waste residue parameters, tunneling parameters, equipment parameters, and constructing the digital twin model includes: The waste residue parameters include the amount of waste residue and the rock fragment size distribution, the rock fragment size distribution includes the large block rate and the small block rate, the tunneling parameters include the thrust, cutter head rotation speed, and tunneling speed, and the equipment parameters include the temperature, energy consumption, vibration signal, and hydraulic pressure; The waste residue parameters, tunneling parameters, and equipment parameters all have their corresponding acquisition times. The digital twin model is constructed according to each parameter by using digital twin technology, and the digital twin model can synchronously simulate each parameter of the shield machine during its tunneling process.

3. The shield intelligent tunneling control system based on machine learning according to claim 2, characterized in that, The process of obtaining the hardness of the waste residue and obtaining the tunneling coefficient in combination with the waste residue parameters includes: Sampling the waste residue corresponding to the waste residue parameters at a single acquisition time to obtain a number of rock samples, processing them into regular shapes, respectively obtaining the hardness values of each rock sample by using the Brinell hardness test method, and taking the average value of the hardness values of each rock sample as the hardness of the waste residue at this acquisition time; Record the waste residue volume, large block rate, small block rate, and waste residue hardness at the same collection time as Z a , Z b , Z c , Z d , and obtain the tunneling coefficient S of the shield machine at this collection time; ； The weight values preset for the amount of waste residue, large block rate, small block rate, and waste residue hardness respectively, and the tunneling coefficient is used to reflect the difficulty of tunneling by the shield machine at the corresponding acquisition moment, and the tunneling coefficients of the shield machine at different acquisition moments are obtained respectively.

4. The shield intelligent tunneling control system based on machine learning according to claim 3, characterized in that, The process of constructing the first calculation model includes: Generating a first calculation set according to the waste residue parameters, tunneling parameters, equipment parameters, and tunneling coefficient at different acquisition times, and dividing it into a first training set and a first test set; Constructing a first convolutional neural network, using different waste residue parameters, tunneling parameters, and equipment parameters in the first training set as the input data of the first convolutional neural network, and using the corresponding tunneling coefficient in the first training set as the output data of the first convolutional neural network; Training the first convolutional neural network to obtain an initial first convolutional neural network, using the first test set to verify the model of the initial first convolutional neural network, and outputting the initial first convolutional neural network with an error less than or equal to the preset first test error threshold as the first calculation model.

5. The intelligent shield tunneling control system based on machine learning according to claim 4, wherein, The process of obtaining the real-time tunneling coefficient and its change trend includes: Continuously input the current waste residue parameters, tunneling parameters, and equipment parameters into the first calculation model to obtain the corresponding real-time tunneling coefficient, and sequentially record the n real-time tunneling coefficients obtained most recently as S i , where i = 1, 2,..., n, and obtain the change coefficient B of each real-time tunneling coefficient i ; ； Obtain the corresponding change trend according to the change coefficient of the acquired real-time tunneling coefficient, including the rising trend, the falling trend, and the constant trend.

6. The shield intelligent tunneling control system based on machine learning according to claim 5, characterized in that, The process of respectively performing simulation adjustment on the shield machine under different change trends and obtaining the simulation tunneling parameters and simulation equipment parameters includes: When the real-time tunneling coefficient is in an upward trend, both the thrust and cutterhead rotation speed of the shield machine are increased in the digital twin model. When the real-time tunneling coefficient is in a downward trend, both the thrust and cutterhead rotation speed of the shield machine are decreased in the digital twin model; When the real-time tunneling coefficient is in a constant trend, no adjustment is made. The simulation adjustment has different adjustment amplitudes, and the adjusted thrust and cutterhead rotation speed are used as the simulation tunneling parameters; Obtain the temperature, energy consumption, and tunneling speed of the shield machine under the simulation tunneling parameters with different adjustment amplitudes in the digital twin model, which are recorded as the simulation equipment parameters. A single simulation tunneling parameter and its simulation equipment parameters are used as an adjustment plan.

7. The intelligent shield tunneling control system based on machine learning according to claim 6, wherein The process of obtaining the adjustment coefficient and generating the optimal adjustment plan includes: Record the temperature, energy consumption, and tunneling speed of the simulation device parameters in a single adjustment plan as H a , H b , H c , record the temperature, energy consumption, and tunneling speed of the shield machine before adjustment as Q a , Q b , Q c , and obtain the adjustment coefficient R of this adjustment plan; ； The adjustment coefficient is used to reflect the operating conditions of the shield machine under the simulation tunneling coefficients with different adjustment amplitudes. The adjustment coefficients of different adjustment plans at the same acquisition moment are respectively obtained, and the adjustment plan with the largest adjustment coefficient is used as the optimal adjustment plan.

8. The shield intelligent tunneling control system based on machine learning according to claim 7, characterized in that The process of constructing the second calculation model includes: Generate a second calculation set according to the simulation tunneling parameters of the optimal adjustment plan at different acquisition moments, the tunneling parameters, equipment parameters, and real-time tunneling coefficient before adjustment, and divide it into a second training set and a second test set; Construct a second convolutional neural network. Use the different tunneling parameters, equipment parameters, and real-time tunneling coefficient in the second training set as the input data of the second convolutional neural network, and use the corresponding simulation tunneling parameters in the second training set as the output data of the second convolutional neural network; Train the second convolutional neural network to obtain an initial second convolutional neural network, use the second test set to verify the model of the initial second convolutional neural network, and output the initial second convolutional neural network with an error less than or equal to the preset second test error threshold as the second calculation model.

Images