Shield tail cavity state prediction method and device

Abstract

The invention discloses a shield tail cavity state prediction method and device. The method comprises the following steps: acquiring internal sensor data and external sensor data of a shield tail cavity; inputting external sensor data into the external feature extraction model, and outputting external feature data; the external feature extraction model is obtained by training a full-connection auto-encoder by using historical external sensor data; inputting internal sensor data into the intracavity pressure feature extraction model, and outputting internal feature data; the intracavity pressure feature extraction model comprises a plurality of types of stack blocks, each stack block comprises two base blocks, and the two base blocks are circulated for multiple times to perform reverse prediction and forward prediction of the intracavity pressure; the type comprises trend prediction and periodical prediction; and fusing the external characteristic data and the internal characteristic data by adopting a mean square error loss function, and outputting predicted data of the intracavity pressure. The prediction precision of the pressure in the shield tail cavity can be improved, and potential risks can be recognized in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel construction, and in particular to a shield tail cavity state prediction method and device. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] The sealing system commonly used in shield machines is composed of sealing brushes and sealing grease. The sealing brushes are elastic and installed on the shield body, tightly attached to the pipe segments. Grease is injected into the gap between the brushes at the tail of the shield for sealing. During the tunneling process of the shield machine, it is necessary to monitor the changes in grease in the sealing cavity at the tail of the shield in real time to prevent the mud and groundwater in the mud and water cavity from penetrating the brush at the tail of the shield, causing leakage of the seal at the tail of the shield.

[0004] The existing methods for predicting the state of the shield tail sealing cavity mainly use the original data of the shield machine or install sensors in the shield tail cavity to sense the state of the shield tail cavity. For example, the detection system is composed of the detector body, the voltage measurement structure, the current measurement structure and the pressure sensor to detect the grease resistivity and pressure in real time. However, it is difficult to cope with sudden accidents by only using the real-time acquired data for monitoring; for another example, the pressure and temperature of the shield tail sealing system are monitored in real time by the detection unit, and the data is transmitted to the monitoring display unit. The shield tail cavity state detection is also performed using the real-time acquired data, which cannot predict sudden accidents. Once the shield tail seal fails, a large amount of mud will flow into the tunnel, which will cause a lot of economic losses and casualties. Therefore, the existing methods are far from enough to judge the state of the shield tail sealing cavity in real time only through monitoring data, and it is difficult to cope with the difficulties and dangers in actual work. Summary of the invention

[0005] The embodiment of the present invention provides a shield tail cavity state prediction method, which is used to perform real-time analysis on the state in the shield tail cavity, and to perform real-time prediction of future trend changes, so as to identify potential risks in advance. The method includes:

[0006] Obtain the internal sensor data and external sensor data of the tail cavity of the shield; the internal sensor data includes the grease status data; the grease status data reflects the moisture content, temperature, and pressure of the grease; the external sensor data includes the grease injection data, tail shield posture data, equipment operation parameters, and environmental factor data;

[0007] The external sensor data is input into the external feature extraction model, and the external feature data is output; the external feature extraction model is pre-trained by using the historical external sensor data to train the fully connected autoencoder; the external feature data reflects the data characteristics of the external sensor;

[0008] Input the internal sensor data into the intracavity pressure feature extraction model, and output the internal feature data; the internal feature data reflects the data features of the predicted intracavity pressure, the intracavity pressure feature extraction model includes multiple types of stack blocks, each stack block includes two basic blocks, the two basic blocks perform reverse prediction and forward prediction of the intracavity pressure in multiple cycles, and each basic block includes a fully connected layer; the types include trend prediction and periodic prediction;

[0009] The mean square error loss function is used to fuse the external feature data and the internal feature data to output the predicted data of the intracavitary pressure.

[0010] In one embodiment, the external feature extraction model includes multiple autoencoders and fully connected layers; each autoencoder is used to encode external sensor data into low-dimensional features and construct original data using the low-dimensional features; the fully connected layer is used to fuse the original data constructed by all autoencoders into external feature data.

[0011] In one embodiment, the stack block of the trend prediction type implements backward prediction and forward prediction by a weighted linear addition method.

[0012] In one embodiment, the stack block of the periodic prediction type implements backward prediction and forward prediction through the Fourier transform method.

[0013] In one embodiment, the method further includes: inputting the external sensor parameters into the external feature extraction model and outputting the external feature data before outputting the external feature data;

[0014] Receive messages to check whether the sensor is working abnormally;

[0015] When a message indicating that the first sensor is working abnormally is received, other sensors adjacent to the first sensor are used to correct the data collected by the first sensor; the message indicating that the first sensor is working abnormally carries the position of the first sensor.

[0016] In one embodiment, after obtaining the internal sensor data and the external sensor data of the shield tail cavity, the method further includes:

[0017] The internal sensor data and the external sensor data are preprocessed by performing one or any combination of the following:

[0018] Filtering, outlier removal, and downsampling.

[0019] The embodiment of the present invention further provides a shield tail cavity state prediction device, which is used to perform real-time analysis on the state in the shield tail cavity, and to perform real-time prediction of future trend changes, and to identify potential risks in advance. The device includes:

[0020] The data acquisition module is used to acquire the internal sensor data and external sensor data of the tail cavity of the shield; the internal sensor data includes the grease status data; the grease status data reflects the moisture content, temperature and pressure of the grease; the external sensor data includes the grease injection data, tail shield posture data, equipment operation parameters and environmental factor data;

[0021] An external feature extraction module is used to input external sensor data into an external feature extraction model and output external feature data; the external feature extraction model is pre-trained by using historical external sensor data to train a fully connected autoencoder; the external feature data reflects the data features of the external sensor;

[0022] The intracavitary pressure feature extraction module is used to input the internal sensor data into the intracavitary pressure feature extraction model and output the internal feature data; the internal feature data reflects the data features of the predicted intracavitary pressure, the intracavitary pressure feature extraction model includes multiple types of stack blocks, each stack block includes two basic blocks, the two basic blocks perform reverse prediction and forward prediction of the intracavitary pressure in multiple cycles, and each basic block includes a fully connected layer; the types include trend prediction and periodic prediction;

[0023] The fusion output module is used to fuse the external feature data and the internal feature data using a mean square error loss function, and output the predicted data of the intracavitary pressure.

[0024] An embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned shield tail cavity state prediction method when executing the computer program.

[0025] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the shield tail cavity state prediction method is implemented.

[0026] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned shield tail cavity state prediction method is implemented.

[0027] In the embodiment of the present invention, internal sensor data and external sensor data of the shield tail cavity are obtained simultaneously, wherein the internal sensor data are used to extract the characteristics of the cavity pressure, and the external data characteristics of the external sensor data are extracted at the same time, and finally the external characteristic data and the internal characteristic data are fused to output the predicted data of the cavity pressure, thereby achieving multi-sensory, multi-parameter, and multi-time series feature fusion, realizing real-time analysis of the state of the shield tail cavity and real-time prediction of future trend changes, improving the prediction accuracy of the shield tail cavity pressure, and being able to identify potential risks in advance, thereby providing timely information to construction personnel, preventing safety hazards, optimizing the construction process, and improving the safety and efficiency of shield construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0029] Figure 1 It is a schematic diagram of the flow of the shield tail cavity state prediction method in an embodiment of the present invention;

[0030] Figure 2 A diagram showing a specific example of a method for predicting the state of a shield tail cavity in an embodiment of the present invention;

[0031] Figure 3 Schematic diagram of data repair principle of the shield tail cavity state prediction method in an embodiment of the present invention;

[0032] Figure 4 Schematic diagram of a cavity pressure feature extraction model in an embodiment of the present invention;

[0033] Figure 5 This is a diagram showing an example of prediction results of the shield tail cavity state prediction method in an embodiment of the present invention;

[0034] Figure 6 Schematic diagram of a shield tail cavity state prediction device in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0036] In shield tunnel construction, real-time perception and trend prediction of the shield tail cavity status are crucial to ensuring construction safety and improving construction efficiency. With the continuous development of shield technology, accurate control of the shield tail cavity status has become an urgent need in the industry. In view of the deficiencies in the prior art, the embodiments of the present invention propose a method for real-time perception and prediction of the shield tail cavity status, which solves the problem that the prior art cannot perform real-time analysis of the status in the shield tail cavity and real-time prediction of future trend changes. It aims to identify potential risks in advance, thereby optimizing construction decisions, ensuring the safe and efficient excavation of the shield machine, improving work efficiency, saving manpower and material resources, and ensuring the safe and efficient excavation of the shield machine under complex geological conditions.

[0037] Figure 1 FIG. 1 is a flow chart of a method for predicting the state of a shield tail cavity in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0038] Step 101, obtaining internal sensor data and external sensor data of the tail cavity of the shield; the internal sensor data includes grease status data; the grease status data reflects the moisture content, temperature, and pressure of the grease; the external sensor data includes grease injection data, tail shield posture data, equipment operation parameters, and environmental factor data;

[0039] Step 102: input the external sensor data into an external feature extraction model, and output external feature data; the external feature extraction model is pre-trained by using historical external sensor data to train a fully connected autoencoder; the external feature data reflects the data features of the external sensor;

[0040] Step 103, input the internal sensor data into the intracavity pressure feature extraction model, and output the internal feature data; the internal feature data reflects the data features of the predicted intracavity pressure, the intracavity pressure feature extraction model includes multiple types of stack blocks, each stack block includes two basic blocks, the two basic blocks perform reverse prediction and forward prediction of the intracavity pressure in multiple cycles, and each basic block includes a fully connected layer; the types include trend prediction and periodic prediction;

[0041] Step 104: Use a mean square error loss function to fuse the external feature data and the internal feature data, and output predicted data of the intracavitary pressure.

[0042] from Figure 1It can be seen from the shown process that in the embodiment of the present invention, the internal sensor data and the external sensor data of the shield tail cavity are obtained simultaneously, wherein the internal sensor data are used to extract the characteristics of the cavity pressure, and the external data characteristics of the external sensor data are extracted at the same time. Finally, the external characteristic data and the internal characteristic data are fused to output the predicted data of the cavity pressure, thereby achieving multi-sensory, multi-parameter, and multi-time series feature fusion, realizing real-time analysis of the state of the shield tail cavity and real-time prediction of future trend changes, improving the prediction accuracy of the shield tail cavity pressure, and being able to identify potential risks in advance, thereby providing timely information to construction personnel, preventing safety hazards, optimizing the construction process, and improving the safety and efficiency of shield construction.

[0043] In the embodiment of the present invention, a sensing monitoring module is installed in the shield tail cavity to obtain parameters such as grease pressure, grease temperature, and grease moisture content in the shield tail sealing cavity. A variety of sensors are also installed outside the shield tail cavity to obtain external data such as grease injection data, tail shield posture data, equipment operating parameters, and environmental factor data.

[0044] The real-time perception and trend prediction of the shield tail cavity status in the embodiment of the present invention combines the shield tail sealing grease injection parameters, tail shield posture parameters, equipment excavation parameters, external soil parameters, shield tail cavity sensing and monitoring parameters, etc., to perform multi-perception and multi-parameter fusion, and construct a cavity status perception prediction model to provide construction personnel with timely and accurate status information, thereby preventing potential safety hazards and optimizing the construction process.

[0045] Figure 2 FIG. 1 is a specific example of the shield tail cavity state prediction method in the embodiment of the present invention, referring to Figure 2 In the embodiment of the present invention, multi-source data are first obtained, including grease injection parameters, tail shield posture parameters, etc. Among them, the grease pressure distribution in the tail shield cavity can directly reflect the grease state in the tail shield cavity. However, in the embodiment of the present invention, the influence of various complex engineering environmental factors such as the time series coordinated change of the grease pressure in the tail shield cavity and the grease injection pressure properties of the front, middle and rear cavities, shield posture and mud pressure are taken into account at the same time, thereby constructing a real-time perception prediction model of the tail shield cavity state with multi-scale and multivariate time series analysis. Specifically, Figure 2 , Figure 1 In the method, first, the characteristics of different external influencing factors such as the front, middle and rear cavity grease injection pressure attributes, shield posture and mud pressure are extracted through the autoencoder and fully connected layer. Then, the periodic and trend time-varying characteristics of the cavity pressure drop sampling sequence are extracted through the preset multi-scale time stack unit. Finally, the shield tail seal grease pressure is fused with the deep time series change characteristics of the external complex engineering environment factors to realize the trend change prediction of the shield tail seal grease pressure and output the future prediction sequence of the cavity pressure. Among them, the real-time perception prediction model of the shield tail cavity state can be Figure 1The collection of methods shown in specifically includes an external feature extraction model, an intracavity pressure feature extraction model and a fusion processing process.

[0046] In one embodiment, to improve data quality, after obtaining the internal sensor data and the external sensor data of the shield tail cavity in step 101, the method may further include:

[0047] The internal sensor data and the external sensor data are preprocessed by performing one or any combination of the following:

[0048] Filtering, outlier removal, downsampling, missing value filling, etc.

[0049] In one embodiment, before inputting the external sensor parameters into the external feature extraction model and outputting the external feature data, the method may further include:

[0050] Receive messages to check whether the sensor is working abnormally;

[0051] When a message indicating that the first sensor is working abnormally is received, other sensors adjacent to the first sensor are used to correct the data collected by the first sensor; the message indicating that the first sensor is working abnormally carries the position of the first sensor.

[0052] Figure 3 FIG. 1 is a schematic diagram of the data repair principle of the shield tail cavity state prediction method in an embodiment of the present invention. Figure 3 As shown, the embodiment of the present invention proposes a data repair mechanism. When the original data original_data.csv is obtained, multiple features featrue are extracted, and after data preprocessing, it is detected whether the sensor module is damaged. If damaged, the data of the adjacent sensor is used for correction and replacement, and the corrected data is used for model prediction, and finally the prediction data predictions.csv is output. It is ensured that in the case of abnormal sensor data, reliable input can still be provided for the real-time perception prediction model of the shield tail cavity state, thereby improving the dynamic update capability and prediction accuracy of the real-time perception prediction model of the shield tail cavity state.

[0053] In step 102, the external sensor data is input into an external feature extraction model, and external feature data is output; the external feature extraction model is pre-trained using historical external sensor data to train a fully connected autoencoder; the external feature data reflects the data features of the external sensor.

[0054] In the embodiment, multiple mutation points may appear in the sealing chamber pressure sequence, which are likely caused by covariate external environmental factors such as grease injection pressure properties, shield posture and mud pressure. Therefore, for external environmental factors, the deep features of the covariate external environmental factors affecting the time series changes of the sealing chamber pressure are learned by stacking autoencoders, and the hidden layer features are mapped to a high-dimensional feature vector of the same dimension as the output of the stack block of the intracavity pressure feature extraction model using a fully connected layer.

[0055] In the embodiment of the present invention, the fully connected autoencoder is obtained by unsupervised training. Unsupervised learning refers to training by learning the intrinsic structure and pattern of data without labels. The fully connected autoencoder is used to learn the low-dimensional representation (encoding) of the input data, and can reconstruct the original input data from the encoding of these low-dimensional representations, thereby strengthening key features.

[0056] In one embodiment, the external feature extraction model includes multiple autoencoders and fully connected layers; each autoencoder is used to encode external sensor data into low-dimensional features and construct original data using the low-dimensional features; the fully connected layer is used to fuse the original data constructed by all autoencoders into external feature data.

[0057] refer to Figure 2 , the autoencoder encodes each external data into a low-dimensional representation through learning, and then uses this low-dimensional representation to reconstruct the original data, so that it can better learn the important features of the original data. Then, the fully connected layer is used for feature extraction, and these external codes are fused into a feature vector.

[0058] In step 103, the internal sensor data is input into the intracavity pressure feature extraction model, and the internal feature data is output; the internal feature data reflects the data features of the predicted intracavity pressure, and the intracavity pressure feature extraction model includes multiple types of stack blocks, each stack block includes two basic blocks, and the two basic blocks perform reverse prediction and forward prediction of the intracavity pressure in multiple cycles, and each basic block includes a fully connected layer; the types include trend prediction and periodic prediction.

[0059] Figure 4 FIG. 1 is a schematic diagram of a cavity pressure feature extraction model according to an embodiment of the present invention. Figure 4As shown in the figure, the intracavitary pressure feature extraction model for multi-scale time series feature extraction of intracavitary pressure is illustrated by two stack blocks, which are trend prediction stack and periodic prediction stack. Each stack block is composed of two basic blocks, and the basic blocks are mainly fully connected layers. The input of each basic block is the input of the previous basic block minus the output of the previous basic block. The advantage of this is that each basic block only needs to predict a part of the output, and the unpredicted part will be passed to the next basic block for further prediction. Each basic block is responsible for the part it is good at, such as periodicity or trend, so the accuracy will be greatly improved. The stack residual is the final output of each stack block, which represents the part that this stack block has not predicted, and will be passed to the next stack block for further prediction.

[0060] Each building block of the stack block has two outputs: a backward prediction output and a forward prediction output. The input of each building block is adjusted by the backward prediction, minus the known information, and then enters the next building block, and the process repeats. The backward prediction captures the known information through a multi-layer fully connected neural network, and the forward prediction represents the future information. The input of the stack block may contain different types of information, so each building block only processes a part of the prediction, such as the trend stack processes the trend part, and the periodic stack processes the period part. Finally, all the forward predictions are added together to get the final prediction result.

[0061] In one embodiment, the stack block of the trend prediction type implements backward prediction and forward prediction by a weighted linear addition method.

[0062] The trend prediction stack usually behaves as a monotonic or slowly changing function, while the intracavitary pressure trend may change in a short period of time, resulting in different trends for each sample. To simulate this behavior, a polynomial is used as the basis and weighted linear addition is used to achieve forward or reverse prediction. Taking forward prediction as an example, the specific polynomial basis in each basis block is as follows, that is, the trend prediction stack block performs forward prediction according to the following formula:

[0063]

[0064] In the formula, is the forward prediction output of the previous block, is the forward prediction output of the current base block. The time vector t = [0, 1, 2, ..., H-2, H-1] / H is a discretely defined grid, each grid represents a time step in the future, t i For each grid, it represents the i-th time step in the future, N pol is the maximum polynomial degree, and b is the number of the current basic block.

[0065] In one embodiment, the stack block of the periodic prediction type implements backward prediction and forward prediction through the Fourier transform method.

[0066] The Fourier transform converts the time domain signal into the frequency domain, making it easier to process periodic trend components. Forecasting can be achieved by operating in the frequency domain and then inverse transforming back to the time domain.

[0067] The periodic prediction stack is used to capture periodic features. Using a basis with Fourier transform function, taking forward prediction as an example, the stack block performs forward prediction according to the following formula:

[0068]

[0069] In the formula, is the forward prediction output of the current base block, is the i-th step after H / 2 time steps of the forward prediction of the previous block, represents the i-th step of the previous H / 2 time steps of the forward prediction of the previous block, N hr Control harmonic oscillation, H is the time step to be predicted, b is the number of the current basis block, t is the time step, i means the current is the i-th time step. Sincosine curves can represent different seasonal components, and the superposition of multiple Fourier basis blocks can simulate more complex periodicity.

[0070] When implemented, the outputs of all stacked blocks are fused together as the intracavitary pressure feature extraction model output.

[0071] For example, the seal chamber pressure trend With periodicity Time series fusion uses a custom basis to transform the time series features and output them. The specific formula is as follows:

[0072]

[0073] in, is the output of the cavity pressure feature extraction model, N c is the number of input features, is the learnable basis transformation coefficient, TCN is the timing transformation operation,

[0074] for or

[0075] In step 104, the external feature data and the internal feature data are fused using a mean square error loss function to output predicted data of the intracavitary pressure.

[0076] During implementation, the ReLU activation function is used to output the final prediction result at time t, recorded as The specific formula is as follows:

[0077]

[0078] in, is the output of the cavity pressure feature extraction model, X Ext is the output of the external feature extraction model, W Ext is the corresponding weight parameter matrix to be learned, which can adjust E t The shape and value of Matching, to avoid one of them taking up too much weight due to its relatively large value. t The ReLU activation function maps the final fusion result to [0, +∞] for the external sensor data such as the grease injection pressure attribute, shield posture and mud pressure at time t.

[0079] Regarding the design of the loss function, the mean square error (MSE) loss is used. There are n samples in the training set, and each sample has d features. Then the feature vector of sample i is recorded as The corresponding label is The loss function is as follows:

[0080]

[0081] In the formula, L() represents the function, and the input is y i and y i is the label of the sample, that is, the true value of the sample to be predicted, and f() represents the model, which can be generally regarded as a function. represents the predicted future sample value, is the real historical sample value.

[0082] Under the mean square error loss, the farther the fitting value is, the greater the penalty is, and the closer the fitting value is, the smaller the penalty is. This is consistent with the characteristics of sudden changes in intracavitary pressure and pays more attention to the mutation point.

[0083] Figure 5 FIG. 4 is a diagram showing an example of the prediction result of the shield tail cavity state prediction method in an embodiment of the present invention. Figure 5 As shown, the method finally outputs the predicted value and the true value of the shield tail seal grease pressure within the next one minute to several hours. It can be seen that the predicted value in the embodiment of the present invention is basically consistent with the true value.

[0084] In one embodiment, the dynamic update of the model is based on data preprocessing, slicing and shuffling the data set to conform to the normal distribution, and then judging the detection status of the sensor in the shield tail cavity. If the sensor is damaged, the corresponding weight update is suspended; otherwise, it is updated normally; the model includes a real-time perception prediction model of the shield tail cavity state, an external feature extraction model, an intracavity pressure feature extraction model, etc.

[0085] To sum up, the embodiment of the present invention adopts a prediction method that integrates multi-sensing, multi-parameter, and multi-time series features, which can improve the prediction accuracy of the pressure in the shield tail cavity and more accurately monitor the state of the shield tail cavity, thereby providing timely information for construction personnel, preventing safety hazards, optimizing the construction process, and improving the safety and efficiency of shield construction.

[0086] The input data obtained in the present invention include but are not limited to fat injection parameters, shield tail posture parameters, shield tail cavity state parameters, external soil pressure parameters, etc. The constructed prediction model can be replaced by a more appropriate method, and the specific scheme is obtained according to the actual situation and experiments. The main idea is to consider the changes in the cavity fat injection pressure while considering the influence of external factors on the cavity fat injection pressure, and realize intelligent prediction through machine learning.

[0087] The present invention also provides a shield tail cavity state prediction device, as described in the following embodiments. Since the principle of the device to solve the problem is similar to the shield tail cavity state prediction method, the implementation of the device can refer to the implementation of the shield tail cavity state prediction method, and the repeated parts will not be repeated.

[0088] Figure 6 FIG. 1 is a schematic diagram of a device for predicting the state of a shield tail cavity in an embodiment of the present invention. Figure 6 The device 600 comprises:

[0089] The data acquisition module 601 is used to acquire the internal sensor data and external sensor data of the tail cavity of the shield; the internal sensor data includes the grease status data; the grease status data reflects the moisture content, temperature and pressure of the grease; the external sensor data includes the grease injection data, tail shield posture data, equipment operation parameters and environmental factor data;

[0090] The external feature extraction module 602 is used to input the external sensor data into the external feature extraction model and output the external feature data; the external feature extraction model is pre-trained by using the historical external sensor data to train the fully connected autoencoder; the external feature data reflects the data features of the external sensor;

[0091] The intracavitary pressure feature extraction module 603 is used to input the internal sensor data into the intracavitary pressure feature extraction model and output the internal feature data; the internal feature data reflects the data features of the predicted intracavitary pressure, the intracavitary pressure feature extraction model includes multiple types of stack blocks, each stack block includes two basic blocks, the two basic blocks perform reverse prediction and forward prediction of the intracavitary pressure in multiple cycles, and each basic block includes a fully connected layer; the types include trend prediction and periodic prediction;

[0092] The fusion output module 604 is used to fuse the external feature data and the internal feature data using a mean square error loss function, and output the predicted data of the intracavitary pressure.

[0093] In one embodiment, the external feature extraction model includes multiple autoencoders and fully connected layers; each autoencoder is used to encode external sensor data into low-dimensional features and construct original data using the low-dimensional features; the fully connected layer is used to fuse the original data constructed by all autoencoders into external feature data.

[0094] In one embodiment, the stack block of the trend prediction type implements backward prediction and forward prediction by a weighted linear addition method.

[0095] In one embodiment, the stack block of the periodic prediction type implements backward prediction and forward prediction through the Fourier transform method.

[0096] In one embodiment, the device 600 further includes: a data correction module.

[0097] The data correction module is used to: before the external feature extraction module 602 inputs the external sensor parameters into the external feature extraction model and outputs the external feature data, receive a message indicating whether the sensor is working abnormally; when a message indicating that the first sensor is working abnormally is received, use other sensors adjacent to the first sensor position to correct the data collected by the first sensor; the message indicating that the first sensor is working abnormally carries the first sensor position.

[0098] In one embodiment, the device 600 further includes: a data preprocessing module.

[0099] The data preprocessing module is used to: after the data acquisition module 601 acquires the internal sensor data and the external sensor data of the shield tail cavity, perform one of the following or any combination of data preprocessing on the internal sensor data and the external sensor data:

[0100] Filtering, outlier removal, and downsampling.

[0101] An embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned shield tail cavity state prediction method when executing the computer program.

[0102] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the shield tail cavity state prediction method is implemented.

[0103] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned shield tail cavity state prediction method is implemented.

[0104] In the embodiment of the present invention, internal sensor data and external sensor data of the shield tail cavity are obtained simultaneously, wherein the internal sensor data are used to extract the characteristics of the cavity pressure, and the external data characteristics of the external sensor data are extracted at the same time, and finally the external characteristic data and the internal characteristic data are fused to output the predicted data of the cavity pressure, thereby achieving multi-sensory, multi-parameter, and multi-time series feature fusion, realizing real-time analysis of the state of the shield tail cavity and real-time prediction of future trend changes, improving the prediction accuracy of the shield tail cavity pressure, and being able to identify potential risks in advance, thereby providing timely information to construction personnel, preventing safety hazards, optimizing the construction process, and improving the safety and efficiency of shield construction.

[0105] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0109] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A shield tail cavity state prediction method, characterized in that: include: Acquire the internal sensor data and external sensor data of the shield tail cavity; the internal sensor data includes the grease status data; Grease status data reflects the moisture content, temperature, and pressure of the grease; external sensor data includes grease injection data, tail shield posture data, equipment operating parameters, and environmental factor data; The external sensor data is input into the external feature extraction model, and the external feature data is output; the external feature extraction model is pre-trained by using the historical external sensor data to train the fully connected autoencoder; the external feature data reflects the data characteristics of the external sensor; Input the internal sensor data into the intracavity pressure feature extraction model, and output the internal feature data; the internal feature data reflects the data features of the predicted intracavity pressure, the intracavity pressure feature extraction model includes multiple types of stack blocks, each stack block includes two basic blocks, the two basic blocks perform reverse prediction and forward prediction of the intracavity pressure in multiple cycles, and each basic block includes a fully connected layer; the types include trend prediction and periodic prediction; The mean square error loss function is used to fuse the external feature data and the internal feature data to output the predicted data of the intracavitary pressure.

2. The method according to claim 1, characterized in that The external feature extraction model includes multiple autoencoders and fully connected layers; each autoencoder is used to encode external sensor data into low-dimensional features and construct original data using the low-dimensional features; the fully connected layer is used to fuse the original data constructed by all autoencoders into external feature data.

3. The method according to claim 1, characterized in that The stack block of the trend forecast type implements backward and forward forecasts by weighted linear addition method.

4. The method according to claim 1, characterized in that The stack block of type periodic prediction implements backward prediction and forward prediction by Fourier transform method.

5. The method according to claim 1, characterized in that The external sensor parameters are input into the external feature extraction model, and before the external feature data is output, the following is also included: Receive messages to check whether the sensor is working abnormally; When a message indicating that the first sensor is working abnormally is received, other sensors adjacent to the first sensor are used to correct the data collected by the first sensor; the message indicating that the first sensor is working abnormally carries the position of the first sensor.

6. The method according to claim 1, characterized in that After obtaining the internal sensor data and external sensor data of the shield tail cavity, it also includes: The internal sensor data and the external sensor data are preprocessed by performing one or any combination of the following: Filtering, outlier removal, and downsampling.

7. A shield tail cavity state prediction device, characterized in that: include: A data acquisition module is used to acquire internal sensor data and external sensor data of the shield tail cavity; the internal sensor data includes grease status data; Grease status data reflects the moisture content, temperature, and pressure of the grease; external sensor data includes grease injection data, tail shield posture data, equipment operating parameters, and environmental factor data; An external feature extraction module is used to input external sensor data into an external feature extraction model and output external feature data; the external feature extraction model is pre-trained by using historical external sensor data to train a fully connected autoencoder; the external feature data reflects the data features of the external sensor; The intracavitary pressure feature extraction module is used to input the internal sensor data into the intracavitary pressure feature extraction model and output the internal feature data; the internal feature data reflects the data features of the predicted intracavitary pressure, the intracavitary pressure feature extraction model includes multiple types of stack blocks, each stack block includes two basic blocks, the two basic blocks perform reverse prediction and forward prediction of the intracavitary pressure in multiple cycles, and each basic block includes a fully connected layer; the types include trend prediction and periodic prediction; The fusion output module is used to fuse the external feature data and the internal feature data using a mean square error loss function, and output the predicted data of the intracavitary pressure.

8. The device according to claim 7, characterized in that The external feature extraction model includes multiple autoencoders and fully connected layers; each autoencoder is used to encode external sensor data into low-dimensional features and construct original data using the low-dimensional features; the fully connected layer is used to fuse the original data constructed by all autoencoders into external feature data.

9. The device according to claim 7, characterized in that The stack block of the trend forecast type implements backward and forward forecasts by weighted linear addition method.

10. The device according to claim 7, characterized in that The stack block of type periodic prediction implements backward prediction and forward prediction by Fourier transform method.

11. The device according to claim 7, characterized in that Also includes: The data correction module is used to receive a message about whether the sensor is working abnormally before the external feature extraction module inputs the external sensor parameters into the external feature extraction model and outputs the external feature data; When receiving a message that the first sensor is abnormally working, correcting the data collected by the first sensor using other sensors adjacent to the first sensor; The message indicating that the first sensor is working abnormally carries the position of the first sensor.

12. The device according to claim 7, characterized in that Also includes: The data preprocessing module is used to perform one of the following or any combination of data preprocessing on the internal sensor data and the external sensor data of the shield tail cavity after the data acquisition module acquires the internal sensor data and the external sensor data: Filtering, outlier removal, and downsampling.

13. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

15. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images