A method and system for quantitatively predicting the risk of TBM jams

Abstract

The present invention relates to the field of tunnel excavation technology, and more particularly to a method and system for quantitatively predicting the risk of a TBM stuck. The method comprises the following steps: determining a TBM stuck prediction index and then collecting TBM stuck prediction index data; constructing a stuck prediction data set based on the TBM stuck prediction index data; supplementing the stuck prediction data set using a sample supplementation model to obtain a complete stuck prediction data set; constructing an index prediction model based on the complete stuck prediction data set, and using the index prediction model to obtain a TBM stuck prediction index prediction value; and calculating a stuck index based on the TBM stuck prediction index prediction value to determine the likelihood of the TBM stuck. The present invention can improve the real-time nature of TBM stuck prediction while ensuring the accuracy of TBM stuck prediction, thereby ensuring the safety of the tunnel excavation process. It can also reduce the requirements for professional knowledge in prediction and increase the flexibility of prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel excavation, and in particular to a method and system for quantitatively predicting the risk of a TBM stuck. Background Art

[0002] The Full-Face Double Shield Tunnel Boring Machine (TBM) is a large-scale tunnel excavation system that integrates mechanical, electrical, hydraulic, sensor, and information technology. It is primarily used for tunneling and construction in rocky geological structures for railways, highways, water conservancy and hydropower diversion tunnels, subways, and underground engineering projects. It enables continuous excavation while simultaneously performing rock breaking, slag removal, and support operations, enabling factory-based construction with high excavation speed and efficiency. However, in complex working environments, TBMs are prone to jamming, where the cutterhead or shield becomes compressed or blocked by the surrounding rock, preventing proper rotation or advancement. This can pose serious safety risks and significant economic losses. Therefore, it is crucial to predict jams during TBM operation to prevent these events.

[0003] Existing TBM jam predictions are primarily based on geological surveys and forecasts, numerical simulations, and machine learning algorithms. However, geological surveys and forecasts have limited accuracy and scope, making accurate predictions of TBM jams difficult. While numerical simulations offer high predictive accuracy, they require a high level of expertise and are complex and computationally intensive, making real-time predictions difficult. Machine learning algorithms offer real-time TBM jam predictions, but due to the scarcity of monitoring samples, the generalization capabilities of these algorithms are limited, making accurate predictions of TBM jams difficult under diverse circumstances. Summary of the Invention

[0004] In view of the defects in the prior art, the present invention provides a method and system for quantitatively predicting the risk of TBM jams.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for quantitatively predicting the risk of TBM jams, comprising the following steps: determining TBM jam prediction indicators and collecting TBM jam prediction indicator data; constructing a jam prediction dataset based on the TBM jam prediction indicator data; supplementing the jam prediction dataset using a sample supplementation model to obtain a complete jam prediction data set; constructing an indicator prediction model based on the complete jam prediction data set, and using the indicator prediction model to obtain TBM jam prediction indicator prediction values; and calculating a jam index based on the TBM jam prediction indicator prediction values ​​to determine the likelihood of TBM jams. While ensuring the accuracy of TBM jam predictions, the present invention can improve the real-time nature of TBM jam predictions, thereby ensuring the safety of the tunnel excavation process. It can also reduce the requirements for professional knowledge and increase the flexibility of predictions.

[0006] Optionally, the TBM jam prediction indicators include shield pressure, cutter head torque, penetration and excavation speed.

[0007] Optionally, constructing a card machine prediction data set based on the TBM card machine prediction indicator data includes the following steps:

[0008] Preprocessing the TBM card machine prediction index data, and then cutting the various TBM card machine prediction index data of each construction site into index data sequences of the same length;

[0009] A card machine prediction dataset is constructed using all the above indicator data series.

[0010] Optionally, the method of using a sample supplementation model to supplement the card machine prediction data set to obtain a complete card machine prediction data set includes the following steps:

[0011] Obtaining a sample card machine feature map based on the TBM card machine prediction index data, and then obtaining a card machine prediction image set based on the card machine prediction data set;

[0012] Based on the card machine prediction image set, a sample supplement model is constructed using a self-attention generative adversarial neural network;

[0013] The sample supplement model is used to supplement the card machine prediction data set to obtain the complete card machine prediction data set.

[0014] Furthermore, the present invention obtains a complete set of card machine prediction data based on the sample supplement model and the card machine prediction data set to improve the generalization performance of the indicator prediction model, thereby improving the accuracy of TBM card machine risk judgment.

[0015] Optionally, the acquiring of a sample card machine feature map according to the TBM card machine prediction index data, and further acquiring a card machine prediction image set according to the card machine prediction data set comprises the following steps:

[0016] Standardizing the TBM card machine prediction index data in the index data sequence according to the value range of the image grayscale value to obtain a standardized index data sequence;

[0017] For any construction site, obtain the corresponding sample card machine feature map, where any row of grayscale values ​​in the sample card machine feature map corresponds to a standardized indicator data sequence of the construction site;

[0018] The sample card machine feature maps of all construction sites are obtained according to the card machine prediction data set, and then a card machine prediction image set is constructed.

[0019] Furthermore, the present invention provides a data basis for the construction of a subsequent sample supplement model by converting the card machine prediction data set into a card machine prediction image set.

[0020] Optionally, the step of constructing a sample supplement model based on the card machine prediction image set using a self-attention generative adversarial neural network comprises the following steps:

[0021] Dividing the card machine prediction image set into a training set and a validation set to complete the training and validation of the self-attention generative adversarial neural network;

[0022] After completing the training and verification of the self-attention generative adversarial neural network, the generator in the self-attention generative adversarial neural network is used as the sample supplement model.

[0023] Furthermore, the present invention uses a self-attention generative adversarial neural network to construct a sample supplement model, thereby improving the intelligence level of sample supplementation and reducing the requirements for professional knowledge in sample supplementation when performing TBM card machine prediction.

[0024] Optionally, the using the sample supplementation model to supplement the card machine prediction data set to obtain the complete card machine prediction data set includes the following steps:

[0025] Generate multiple expansion card machine feature maps using the sample supplement model;

[0026] Determining a standardized indicator data expansion sequence according to the expansion card machine characteristic graph;

[0027] Destandardizing the standardized indicator data expansion sequence to obtain an indicator data expansion sequence;

[0028] The indicator data expansion sequence is supplemented to the card machine prediction data set to obtain the complete card machine prediction data set.

[0029] Furthermore, the present invention uses a sample supplementation model to supplement the card machine prediction data set to obtain a complete set of card machine prediction data, so as to improve the generalization performance of the indicator prediction model and thereby improve the accuracy of TBM card machine risk judgment.

[0030] Optionally, the step of constructing an indicator prediction model based on the complete set of card machine prediction data and obtaining TBM card machine prediction indicator prediction values ​​using the indicator prediction model comprises the following steps:

[0031] Using a long short-term memory neural network to construct an initial indicator prediction model for each of the TBM card machine prediction indicators;

[0032] Dividing the complete card machine prediction data set into multiple data complete subsets according to the TBM card machine prediction indicator type;

[0033] Using the complete subset of the TBM card machine prediction indicators to train and verify the corresponding initial indicator prediction model, to obtain the corresponding indicator prediction model;

[0034] For any of the TBM machine jam prediction indicators, the corresponding indicator prediction model is used to predict the TBM machine jam prediction indicator to obtain the corresponding TBM machine jam prediction indicator prediction value.

[0035] Furthermore, the present invention constructs an indicator prediction model based on a complete set of card machine prediction data and a long short-term memory neural network, which can not only improve the accuracy of TBM card machine prediction, but also realize real-time prediction.

[0036] Optionally, the card machine index satisfies the following relationship:

[0037]

[0038] in, is the card machine index, n is the number of TBM card machine prediction indicators, is the risk impact weight of the TBM jam prediction indicator of the i-th TBM jam, is the TBM card machine prediction index prediction value of the i-th TBM card machine prediction index, For about function.

[0039] Furthermore, the card machine index of the present invention is simple to calculate and relatively accurate, which can improve the real-time performance and accuracy of TBM card machine risk assessment.

[0040] In the second aspect, the present invention also provides a TBM card machine risk quantitative prediction system, which includes: a data acquisition device, a data output device, a processor and a storage device, the storage device includes a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer program includes program instructions, and when the program instructions are executed by the processor, the processor implements a TBM card machine risk quantitative prediction method provided by the present invention.

[0041] In summary, the present invention has at least the following beneficial effects:

[0042] 1. The present invention uses a sample supplementation model to supplement the card machine prediction data set to obtain a complete card machine prediction data set, which can improve the generalization performance of the indicator prediction model and thus improve the accuracy of TBM card machine prediction;

[0043] 2. The sample supplementation model and indicator prediction model constructed by the present invention are both constructed based on machine learning algorithms. This not only improves the accuracy and real-time performance of TBM jam prediction, but also reduces the requirements for professional knowledge in TBM jam prediction and improves the flexibility of prediction.

[0044] 3. Since the present invention has low requirements for professional knowledge and small amount of calculation when performing TBM jam prediction, it has a wider scope of application.

[0045] 4. The TBM machine risk quantitative prediction system provided by the present invention has a compact structure, stable operation, and is easy to install. It can also be well combined with the method of the present invention to improve the practicality of the method provided by the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0047] Figure 1 A schematic flow chart of a method for quantitatively predicting the risk of a TBM jammer according to an embodiment of the present invention;

[0048] Figure 2 The figure is a schematic diagram of the framework of a TBM machine risk quantification prediction system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] Specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the present invention. In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that these specific details are not necessarily required to practice the present invention. In other instances, well-known circuits, software, or methods are not specifically described to avoid obscuring the present invention.

[0050] Throughout this specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present invention. Therefore, appearances of the phrases "in one embodiment," "in an embodiment," "an example," or "an example" in various places throughout this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics may be combined in any suitable combinations and / or subcombinations in one or more embodiments or examples. Furthermore, those of ordinary skill in the art will appreciate that the figures provided herein are for illustrative purposes only and are not necessarily drawn to scale.

[0051] It should be noted in advance that, in an optional embodiment, except for independent explanations, the same symbols or letters appearing in all formulas have the same meanings and values.

[0052] In an alternative embodiment, see Figure 1 The present invention provides a method for quantitatively predicting the risk of a TBM machine jam, the method comprising the following steps:

[0053] S1. Determine TBM jam prediction indicators and then collect TBM jam prediction indicator data.

[0054] Specifically, in this embodiment, TBM jam prediction indicators include shield pressure, cutterhead torque, penetration rate, and tunneling speed. The collected TBM jam prediction indicator data includes TBM jam prediction indicator data used to construct the jam prediction dataset, as well as real-time TBM jam prediction indicator data at the target construction site. The TBM jam prediction indicator data used to construct the jam prediction dataset is primarily sourced from the internet and journals, while the real-time TBM jam prediction indicator data at the target construction site is collected during operation using the TBM's built-in data acquisition device, which typically includes pressure sensors, torque sensors, and displacement sensors.

[0055] Furthermore, in other optional embodiments, the TBM jam prediction index may not be limited to shield pressure, cutterhead torque, penetration, and tunneling speed. For example, the TBM jam prediction index may also include cutterhead wear and tunneling machine power.

[0056] S2. Construct a card machine prediction data set based on the TBM card machine prediction indicator data.

[0057] Wherein, step S2 specifically includes the following steps:

[0058] S21 , pre-processing the TBM card machine prediction index data, and then cutting the various TBM card machine prediction index data of each construction site into index data sequences with the same length.

[0059] Specifically, in this embodiment, the preprocessing of TBM machine prediction indicator data includes data discretization, outlier identification, and missing value supplementation. To facilitate subsequent model training and prediction, the various TBM machine prediction indicator data need to be cut into indicator data sequences of the same length.

[0060] More specifically, in this embodiment, the length of each indicator data sequence is 32, and the time interval between two adjacent data in the indicator data sequence is 4 minutes. In other optional embodiments, the length and time interval of the indicator data sequence can be adjusted according to actual needs.

[0061] S22. Use all the indicator data sequences to construct a card machine prediction data set.

[0062] Specifically, in this embodiment, a set of indicator data sequences in the card machine prediction data set includes a shield pressure data sequence, a cutter head torque data sequence, a penetration data sequence and a tunneling speed data sequence, and all data in a set of indicator data sequences come from the same construction site.

[0063] S3. Use a sample supplement model to supplement the card machine prediction data set to obtain a complete card machine prediction data set.

[0064] This embodiment obtains a complete set of card machine prediction data based on the sample supplement model and the card machine prediction data set, which solves the problem of scarce monitoring samples and is conducive to improving the generalization performance of the indicator prediction model constructed subsequently, thereby improving the accuracy of TBM card machine risk judgment. Step S3 specifically includes the following steps:

[0065] S31. Obtain a sample card machine feature map based on the TBM card machine prediction index data, and then obtain a card machine prediction image set based on the card machine prediction data set.

[0066] This embodiment converts the card machine prediction dataset into a card machine prediction image set, providing a data basis for the subsequent construction of a sample supplement model. Step S31 specifically includes the following steps:

[0067] S311. Standardize the TBM card machine prediction index data in the index data sequence according to the value range of the image grayscale value to obtain a standardized index data sequence.

[0068] Specifically, in this embodiment, in order to convert the indicator data sequence into an image, it is necessary to restrict the TBM card machine prediction indicator data in the indicator data sequence to the value range of the image grayscale value, that is, to perform normalization processing on the TBM card machine prediction indicator data in the indicator data sequence according to the value range of the image grayscale value, which is specifically achieved through the following relationship:

[0069]

[0070] in, is the standardized TBM jam prediction index data of the i-th TBM jam prediction index, is the TBM jam prediction index data of the i-th TBM jam prediction index, is the maximum value of the prediction index of the i-th TBM jam machine, is the minimum value of the prediction index of i TBM jam machines.

[0071] S312. For any construction site, obtain the corresponding sample card machine feature map, where any row of grayscale values ​​in the sample card machine feature map corresponds to a standardized indicator data sequence of the construction site.

[0072] Specifically, in this embodiment, the standardized TBM card machine prediction index data in the standardized index data sequence is used as the grayscale value of the pixel point in the sample card machine feature map to construct the sample card machine feature map. The constructed sample card machine feature map has a size of 64×64, that is, the length and width of the sample card machine feature map are both 64 pixels.

[0073] More specifically, since the size of the sample card machine feature map is 64×64, each row of grayscale values ​​in the sample card machine feature map is determined by the same standardized indicator data sequence, where the first 32 grayscale values ​​are respectively the 32 standardized TBM card machine prediction indicator data in the standardized indicator data sequence, and the last 32 grayscale values ​​are also respectively the 32 standardized TBM card machine prediction indicator data in the standardized indicator data sequence; the pixel values ​​of rows 1 to 4 of a sample card machine feature map are determined in sequence by four standardized indicator data sequences in a group of standardized indicator data sequences, and the subsequent four rows of grayscale values ​​are determined in sequence by the four standardized indicator data sequences in the group of standardized indicator data sequences.

[0074] Furthermore, a set of standardized indicator data series includes a standardized shield pressure data series, a standardized cutterhead torque data series, a standardized penetration data series and a standardized tunneling speed data series.

[0075] S313. Obtain the sample card machine feature maps of all construction sites according to the card machine prediction data set, and then construct a card machine prediction image set.

[0076] Specifically, in this embodiment, sample card machine feature maps of all construction sites are obtained based on the card machine prediction data set, and then the sample card machine feature maps of all construction sites are used to construct a card machine prediction image set.

[0077] S32. Based on the card machine prediction image set, a sample supplement model is constructed using a self-attention generative adversarial neural network.

[0078] This embodiment uses a self-attention generative adversarial neural network to construct a sample supplement model, improving the intelligence of sample supplementation and reducing the professional knowledge required for sample supplementation when performing TBM card machine prediction. Step S32 specifically includes the following steps:

[0079] S321. Divide the card machine prediction image set into a training set and a verification set to complete the training and verification of the self-attention generative adversarial neural network.

[0080] Specifically, in this embodiment, the card machine prediction image set is divided into a training set and a verification set in a ratio of 8:2, where the training set is used to complete the training of the self-attention generative adversarial neural network, and the verification set is used to complete the verification of the self-attention generative adversarial neural network.

[0081] More specifically, the self-attention generative adversarial neural network consists of a generator and a discriminator. The generator's task is to generate as realistic fake images as possible from random noise, making it difficult for the discriminator to distinguish whether these fake images are generated and not real. The discriminator takes fake and real images as input and is tasked with determining whether the input image comes from a real dataset or is generated by the generator. The generator and discriminator continuously improve their image generation quality and discriminative capabilities in a zero-sum game. Ultimately, the discriminator is unable to distinguish between fake and real images, meaning that the generator is now indistinguishable from the real. Therefore, the trained generator can be used to generate fake images to supplement the real images in the real dataset. Obviously, the real images in this embodiment are sample card machine feature maps.

[0082] S322. After completing the training and verification of the self-attention generative adversarial neural network, the generator in the self-attention generative adversarial neural network is used as the sample supplement model.

[0083] S33. Use the sample supplement model to supplement the card machine prediction data set to obtain the complete card machine prediction data set.

[0084] This embodiment uses a sample supplementation model to supplement the card machine prediction data set to obtain a complete card machine prediction data set, so as to obtain sufficient data for model training, thereby improving the accuracy of TBM card machine risk assessment. Step S33 specifically includes the following steps:

[0085] S331. Generate multiple expansion card machine feature maps using the sample supplement model.

[0086] Specifically, in this embodiment, the random noise input to the sample supplement model is adjusted to generate a plurality of expansion card machine feature maps.

[0087] S332. Determine a standardized indicator data expansion sequence according to the expansion card machine characteristic diagram.

[0088] Specifically, in this embodiment, according to the description of step S312, for an expansion card machine characteristic map, a set of standardized indicator data expansion sequences can be determined based on only the first 32 pixel values ​​of each row of its 1 to 4 rows of pixel values.

[0089] S333: Destandardize the standardized indicator data expansion sequence to obtain an indicator data expansion sequence.

[0090] Specifically, in this embodiment, the data in the standardized indicator data expansion sequence is standardized data and cannot be directly used as TBM card machine prediction indicator data to supplement the card machine prediction data set. Therefore, it is necessary to destandardize the data in the standardized indicator data expansion sequence to obtain the indicator data expansion sequence.

[0091] More specifically, the process of de-standardizing the expanded sequence of standardized indicator data is actually known To calculate For the sake of simplicity, the process will not be described in detail here.

[0092] S334. Supplement the indicator data expansion sequence to the card machine prediction data set to obtain the complete card machine prediction data set.

[0093] S4. Construct an indicator prediction model based on the complete set of card machine prediction data, and use the indicator prediction model to obtain TBM card machine prediction indicator prediction values.

[0094] This embodiment builds an indicator prediction model based on a complete set of card machine prediction data and a long short-term memory neural network, which can not only improve the accuracy of TBM card machine prediction, but also achieve real-time prediction. Step S4 specifically includes the following steps:

[0095] S41. Use a long short-term memory neural network to construct an initial indicator prediction model for each of the TBM card machine prediction indicators.

[0096] Specifically, in this embodiment, a long short-term memory neural network, i.e., LSTM, is used to construct initial indicator prediction models for each TBM jammer prediction indicator, specifically, an initial prediction model for shield pressure, an initial prediction model for cutterhead torque, an initial prediction model for penetration, and an initial prediction model for tunneling speed.

[0097] S42. Divide the complete card machine prediction data set into multiple data complete subsets according to the TBM card machine prediction indicator types.

[0098] Specifically, in this embodiment, since different TBM card machine prediction indicators correspond to an initial indicator prediction model respectively, it is necessary to divide the card machine prediction data complete set into multiple data complete subsets according to the type of TBM card machine prediction indicators, so as to train and verify each initial indicator prediction model.

[0099] More specifically, the data complete subset of this embodiment includes a shield pressure data complete subset, a cutterhead torque data complete subset, a penetration data complete subset, and a tunneling speed data complete subset.

[0100] S43. Use the data complete subset of the TBM card machine prediction indicators to train and verify the corresponding initial indicator prediction model to obtain the corresponding indicator prediction model.

[0101] Specifically, in this embodiment, the indicator prediction model includes a shield pressure prediction model, a cutterhead torque prediction model, a penetration prediction model, and a tunneling speed prediction model. The shield pressure initial prediction model is trained and verified using the complete subset of shield pressure data to obtain the shield pressure prediction model; the cutterhead torque initial prediction model is trained and verified using the complete subset of cutterhead torque data to obtain the cutterhead torque prediction model; the penetration initial prediction model is trained and verified using the complete subset of penetration data to obtain the penetration prediction model; and the tunneling speed initial prediction model is trained and verified using the complete subset of tunneling speed data to obtain the tunneling speed prediction model.

[0102] S44. For any of the TBM machine prediction indicators, use the corresponding indicator prediction model to predict the TBM machine prediction indicator to obtain the corresponding TBM machine prediction indicator prediction value.

[0103] Specifically, in this embodiment, a corresponding indicator data sequence is obtained based on the data of each TBM stuck machine prediction indicator collected in the target construction site, and the obtained indicator data sequence is input into the corresponding indicator prediction model to predict the future TBM stuck machine prediction indicator data, so as to obtain the TBM stuck machine prediction indicator prediction value of each TBM stuck machine prediction indicator in the target construction site.

[0104] More specifically, for an indicator data sequence of a certain TBM stuck prediction indicator in a target construction site, the last data in the sequence is the current value of the TBM stuck prediction indicator.

[0105] S5. Calculate a jam index based on the TBM jam prediction index prediction value to determine the possibility of TBM jam.

[0106] Specifically, in this embodiment, the card machine index satisfies the following relationship:

[0107]

[0108] in, is the machine jam index, n is the number of TBM machine jam prediction indicators, is the risk impact weight of the i-th TBM jam prediction indicator, is the TBM jam prediction index prediction value of the i-th TBM jam prediction index, For about function.

[0109] More specifically, It was obtained using the expert evaluation method. It is a function that reflects the degree of change of TBM jam prediction index within a certain period of time. If we use 、 、 and represents shield pressure, cutterhead torque, penetration and tunneling speed, then Specifically, the following relationships are met:

[0110]

[0111]

[0112] in, is the current value of the prediction index of the i-th TBM stuck at the target construction site, and t is the time interval described in step S21.

[0113] Furthermore, a card machine judgment threshold is set. When the calculated card machine index is greater than the card machine judgment threshold, it is determined that a TBM card machine may have occurred. In this embodiment, the card machine judgment threshold is set to 0.5.

[0114] It should be noted that, in some cases, the actions described in the specification can be performed in a different order and still achieve the desired results. In this embodiment, the order of steps given is only to make the embodiment appear clearer and easier to explain, rather than to limit it.

[0115] In an alternative embodiment, see Figure 2 This embodiment provides a system for quantitatively predicting the risk of a TBM card machine. The system comprises a data acquisition device 1, a data output device 2, a processor 3, and a memory 4. The memory 4 comprises a computer-readable storage medium storing a computer program. The computer program comprises program instructions. When executed by the processor 3, the program instructions cause the processor 3 to implement the method for quantitatively predicting the risk of a TBM card machine according to this embodiment. The system for quantitatively predicting the risk of a TBM card machine provided in this embodiment has a compact structure, stable operation, and is easy to install. It can also be well combined with the method of the present invention, making the present invention more practical and commercially valuable.

[0116] Specifically, in this embodiment, the data acquisition device 1, data output device 2, processor 3, and storage 4 are electrically connected to each other. The data acquisition device 1 includes sensing devices such as a pressure sensor, a torque sensor, and a displacement sensor mounted on the TBM. It also includes a first digital display for manually inputting TBM machine prediction index data. The data output device 2 includes a second digital display capable of displaying the machine index.

[0117] In summary, first, the method provided by the present invention uses a sample supplement model to supplement the card machine prediction data set to obtain a complete set of card machine prediction data, which can improve the generalization performance of the indicator prediction model, and thus improve the accuracy of TBM card machine prediction; second, the sample supplement model and indicator prediction model constructed by the method provided by the present invention are both based on machine learning algorithms, which can not only improve the accuracy and real-time performance of TBM card machine prediction, but also reduce the requirements for professional knowledge in TBM card machine prediction and improve the flexibility of prediction; finally, because the method provided by the present invention has low requirements for professional knowledge and small amount of calculation when performing TBM card machine prediction, it has a wider range of applications. In addition, the TBM card machine risk quantitative prediction system provided by the present invention is a system compatible with the method provided by the present invention. It has a compact structure, stable operation, is easy to install, and can improve the practicality of the method provided by the present invention.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A method for quantitatively predicting the risk of TBM jamming, characterized in that: The steps include: Determine the TBM jam prediction indicators and collect TBM jam prediction indicator data; Preprocessing the TBM card machine prediction index data, and then cutting the various TBM card machine prediction index data of each construction site into index data sequences of the same length; Use all the indicator data series to build a card machine prediction dataset; Standardizing the TBM card machine prediction index data in the index data sequence according to the value range of the image grayscale value to obtain a standardized index data sequence; The normalization of the TBM card machine prediction index data in the index data sequence is achieved through the following relationship: in, is the standardized TBM jam prediction index data of the i-th TBM jam prediction index, is the TBM jam prediction index data of the i-th TBM jam prediction index, is the maximum value of the prediction index of the i-th TBM jam machine, is the minimum value of the prediction index of i TBM jammers; For any construction site, obtain its corresponding sample card machine feature map. Any grayscale value in the sample card machine feature map corresponds to a standardized indicator data sequence for the construction site. Specifically, the standardized TBM card machine prediction indicator data in the standardized indicator data sequence is used as the grayscale value of the pixel point in the sample card machine feature map to construct the sample card machine feature map. The size of the sample card machine feature map is 64×64. Each row of grayscale values ​​in the sample card machine characteristic graph is determined by the same standardized indicator data sequence, wherein the first 32 grayscale values ​​are sequentially the 32 standardized TBM card machine prediction indicator data in the standardized indicator data sequence, and the last 32 grayscale values ​​are sequentially the 32 standardized TBM card machine prediction indicator data in the standardized indicator data sequence; The pixel values ​​of rows 1 to 4 of the sample card machine characteristic map are determined in sequence by four standardized indicator data sequences in a set of standardized indicator data sequences, and the grayscale values ​​of each of the subsequent four rows are determined in sequence by the four standardized indicator data sequences in the set of standardized indicator data sequences; Obtain the sample card machine feature maps of all construction sites according to the card machine prediction data set, and then construct a card machine prediction image set; The card machine prediction image set is divided into a training set and a validation set to complete the training and validation of the self-attention generative adversarial neural network; After completing the training and verification of the self-attention generative adversarial neural network, using the generator in the self-attention generative adversarial neural network as a sample supplement model; Generate multiple expansion card machine feature maps using the sample supplement model; Determining a standardized indicator data expansion sequence according to the expansion card machine characteristic graph; Destandardizing the standardized indicator data expansion sequence to obtain an indicator data expansion sequence; Supplementing the indicator data expansion sequence to the card machine prediction data set to obtain a complete card machine prediction data set; Using a long short-term memory neural network to construct an initial indicator prediction model for each of the TBM card machine prediction indicators; Dividing the complete card machine prediction data set into multiple data complete subsets according to the TBM card machine prediction indicator type; Using the complete subset of the TBM card machine prediction indicators, the corresponding initial indicator prediction model is trained and verified to obtain the corresponding indicator prediction model; For any of the TBM jam prediction indicators, use the corresponding indicator prediction model to predict the TBM jam prediction indicator to obtain the corresponding TBM jam prediction indicator prediction value; Calculate the jam index based on the TBM jam prediction index prediction value to determine the possibility of TBM jam; The card machine index satisfies the following relationship: ， in, is the card machine index, n is the number of TBM card machine prediction indicators, is the risk impact weight of the TBM jam prediction indicator of the i-th TBM jam, is the TBM card machine prediction index prediction value of the i-th TBM card machine prediction index, For about function; Use in sequence 、 、 and represents shield pressure, cutterhead torque, penetration and tunneling speed, then Specifically, the following relationships are met: in, is the current value of the prediction index of the i-th TBM stuck at the target construction site, and t is the time interval.

2. The method for quantitatively predicting the risk of a TBM jam according to claim 1, characterized in that: The TBM jam prediction indicators include shield pressure, cutter head torque, penetration and tunneling speed.

3. A TBM machine jam risk quantitative prediction system, characterized by: The TBM card machine risk quantitative prediction system includes: a data acquisition device, a data output device, a processor and a storage, the storage includes a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer program includes program instructions, and when the program instructions are executed by the processor, the processor implements a TBM card machine risk quantitative prediction method as described in any one of claims 1-2.

Images