Attitude prediction model construction methods, attitude prediction methods, equipment and media

Abstract

This invention discloses a method for constructing an attitude prediction model, an attitude prediction method, equipment, and a medium, relating to the field of tunnel boring machine (TBM) technology. The key technical points are: obtaining historical TBM parameters from a historical time series; performing dimensionality reduction and filtering on the historical TBM parameters to obtain dimensionality-reduced TBM parameters; constructing a multi-objective optimization function that minimizes the difference between the horizontal and vertical positions of the front and rear ends of the TBM; combining the dimensionality-reduced TBM parameters, solving the multi-objective optimization function using a non-dominated genetic algorithm to obtain the optimal TBM parameters; and training a pre-constructed long short-term memory (LSTM) neural network using the optimal TBM parameters. After training, an attitude prediction model suitable for TBM attitude prediction is obtained. This solves the problems of insufficient computational efficiency and prediction accuracy in existing attitude prediction models.

Description

Technical Field

[0001] This invention relates to the field of tunnel boring machine (TBM) technology, and more specifically, to a method for constructing an attitude prediction model, an attitude prediction method, equipment, and a medium. Background Technology

[0002] With the rapid development of urbanization, the demand for subway tunnel projects in cities and between cities is increasing day by day. As a commonly used equipment for subway tunnel excavation, the shield tunneling machine is often subject to some unpredictable factors during the tunneling process, such as blade overload, changes in strata, and interference from the shield's own weight. This often causes the shield tunneling machine to deviate in attitude and position, resulting in a serpentine movement in the tunnel. Therefore, when the deviation exceeds the allowable range, the attitude and position of the shield tunneling machine must be adjusted in time to bring it back to the preset track axis.

[0003] Currently, existing technologies employ genetic algorithms and principal component analysis, combined with long short-term memory neural models, to optimize multidimensional influencing factors and predict shield attitude. However, these existing technologies are insufficient in terms of computational efficiency and prediction accuracy. Summary of the Invention

[0004] The purpose of this invention is to provide a method for constructing an attitude prediction model, an attitude prediction method, an apparatus, and a medium, which solves the problems of insufficient computational efficiency and prediction accuracy of attitude prediction models provided by the prior art.

[0005] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0006] A first aspect of the present invention provides a method for constructing an attitude prediction model, the method comprising:

[0007] Obtain historical shield tunneling parameters of the tunnel boring machine in a historical time series;

[0008] Dimensionally reduce and filter historical shield tunneling parameters to obtain the reduced shield tunneling parameters;

[0009] A multi-objective optimization function is constructed to minimize the difference between the horizontal and vertical positions of the front and rear ends of the tunnel boring machine. Combined with the dimensionality-reduced tunnel boring machine parameters, the multi-objective optimization function is solved using a non-dominated genetic algorithm to obtain the optimal tunnel boring machine parameters.

[0010] The pre-constructed long short-term memory neural network is trained using the optimal shield parameters. After training, an attitude prediction model that can be used for shield machine attitude prediction is obtained.

[0011] In one implementation, the historical shield parameters include the position parameters of the shield attitude and the shield operation parameters.

[0012] In one implementation, before performing dimensionality reduction filtering on historical shield tunneling parameters to obtain the dimensionality-reduced shield tunneling parameters, the method further includes: using the KS parameter verification method to perform normality verification on the historical shield tunneling parameters in order to remove outlier data from the historical shield tunneling parameters.

[0013] In one implementation scheme, principal component analysis is used to reduce the dimensionality of historical shield tunneling parameters and obtain the reduced shield tunneling parameters.

[0014] In one implementation, before using a non-dominated genetic algorithm to solve the multi-objective optimization function and obtain the optimal shield tunneling parameters, the method further includes: using an ant colony algorithm to adjust the search strategy of the non-dominated genetic algorithm.

[0015] In one implementation, the improved ant colony algorithm refers to a global pheromone update rule that performs probabilistic search according to state transition rules in each iteration, enhances the pheromone global update rule for the optimal search path obtained in each iteration, and performs local updates when constructing the search path, and performs a local update of the search path after each iteration.

[0016] In one implementation scheme, the expression for the state transition rule is:

[0017] Where J represents the set of nodes that are directly reachable from the node and are not already visited in the node sequence. For nodes i To the node j pheromones, For nodes i To the node j Heuristic functions, As a pheromone-stimulating factor, β Let q be the expected heuristic factor, q represent a random number, q0 represent a random number between 0 and 1, and S represent the probability calculated by the probability calculation function.

[0018] The expression for the global pheromone update rule is: Where C represents the length of the path chosen by the ant. Indicates the global volatile factor of pheromones. This represents the amount of pheromone released by an ant as it passes through the edge formed by nodes i and j.

[0019] The expression for the pheromone local update rule is: ,in, This indicates the local volatile factor of pheromones. The chaotic factor generated by the chaotic mapping function is represented by k, which represents the current iteration number, and K represents the maximum iteration number. This represents the initial value of the pheromone.

[0020] A second aspect of the present invention provides an attitude prediction method, the method comprising:

[0021] Obtain the shield machine parameters in the current time series;

[0022] The shield tunneling parameters are input into the attitude prediction model constructed by the attitude prediction model construction method provided in the first aspect of the present invention, and the attitude prediction result of the shield tunneling machine in the next time series is output.

[0023] A third aspect of the present invention provides an electronic device, including a memory and a processor;

[0024] A memory for storing computer programs, the computer programs including program instructions;

[0025] A processor is configured to execute the program instructions to cause the electronic device to perform an attitude prediction model construction method as provided in the first aspect of the present invention and an attitude prediction method as provided in the second aspect of the present invention.

[0026] A fourth aspect of the present invention provides a computer program product comprising program instructions that, when executed by an electronic device, cause the electronic device to perform an attitude prediction model construction method as provided in the first aspect of the present invention and an attitude prediction method as provided in the second aspect of the present invention.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] In the attitude prediction model construction method provided by this invention, an improved ant colony algorithm is used to adjust the search strategy of the non-dominated genetic algorithm. This adjustment improves the efficiency and accuracy of the non-dominated genetic algorithm in solving the multi-objective optimization function that minimizes the difference between the horizontal and vertical positions of the shield front and rear ends. Then, based on the optimal shield parameters, a pre-constructed long short-term memory neural network is trained. Upon completion of training, an attitude prediction model that can be used for shield machine attitude prediction is obtained. Since the adjusted non-dominated genetic algorithm improves the solution accuracy, the prediction accuracy of the attitude prediction model trained based on the solved optimal shield parameters is also improved accordingly. This solves the problems of insufficient computational efficiency and prediction accuracy in the attitude prediction methods provided by the prior art. Attached Figure Description

[0029] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0030] Figure 1A flowchart illustrating a method for constructing an attitude prediction model according to an embodiment of the present invention;

[0031] Figure 2 This is a flowchart illustrating an attitude prediction method provided in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are only for explaining this invention and are not intended to limit this invention.

[0033] It should be noted that the terms "comprising" or "may include" used in the various embodiments of this application indicate the presence of the claimed function, operation, or element, and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in the various embodiments of this application, the terms "comprising," "having," and their cognates are intended only to indicate a specific feature, number, step, operation, element, component, or combination of the foregoing, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or adding one or more combinations of the foregoing.

[0034] It should be understood that terms such as "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0035] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for constructing a pose prediction model according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0036] S101, obtain the historical shield tunneling parameters of the tunnel boring machine in the historical time series.

[0037] In this embodiment, historical tunnel boring machine (TBM) parameters include positional parameters of the TBM's attitude and operational parameters. For example, the positional parameters of the TBM's attitude include displacement parameters, angle parameters, cut angle mileage, and three-dimensional coordinates. The displacement parameters can be further divided into horizontal and vertical deviations at the TBM's front end and rear end. Specifically, the horizontal deviation at the TBM's front end refers to the distance the TBM's head deviates from the tunnel's design axis in the horizontal direction, and the horizontal deviation at the TBM's front end refers to the distance the TBM's head deviates from the tunnel's design axis in the vertical direction.

[0038] Angular parameters include pitch angle, azimuth angle, and roll angle. Pitch angle is the angle between the tunnel boring machine's (TBM) central axis and the horizontal plane, reflecting the TBM's vertical inclination and determining the tunnel's slope. Azimuth angle, also known as yaw angle, is the angle between the TBM's projection onto the horizontal plane and the tunnel's design axis, reflecting the TBM's deviation direction in the horizontal plane. Roll angle, also called twist angle or yaw angle, is the angle of rotation of the TBM around its central axis, mainly caused by the reaction force of the cutterhead, reflecting the TBM's rotational state.

[0039] The cut-out mileage refers to the mileage position of the cut-out of the tunnel boring machine, which helps determine the longitudinal position of the tunnel boring machine in the tunnel.

[0040] Three-dimensional coordinates refer to the three-dimensional coordinate position of the center or key parts of the tunnel boring machine cutterhead, which can accurately determine the specific position of the tunnel boring machine in space.

[0041] The operating parameters of a tunnel boring machine (TBM) include tunneling speed, propulsion cylinder pressure, propulsion cylinder stroke difference, cutterhead rotation speed, and cutterhead torque. The parameters described in this embodiment are common operating parameters of TBMs, and will not be described in detail here.

[0042] Sensors were deployed on the tunnel boring machine (TBM) to record its attitude and position parameters as well as its operational parameters. Each ring was considered a recording sample, with over 1800 rings recorded. The statistical parameters are shown in Table 1. Parameters 1-14 were used as input indicators, and parameters 15-18 as output indicators. The coordinates of the TBM's attitude position parameters were based on the intersection of the design axis and the cross-section as the origin. Horizontal deviation to the left was negative, and deviation to the right was positive; vertical deviation downwards was negative, and deviation upwards was positive.

[0043] Table 1

[0044]

[0045] S102, dimensionality reduction and filtering of historical shield tunneling parameters to obtain the dimensionality-reduced shield tunneling parameters.

[0046] Specifically, principal component analysis was used to reduce the dimensionality of historical shield tunneling parameters and obtain the reduced shield tunneling parameters.

[0047] Principal component analysis (PCA) was used to analyze and screen the 18 parameters provided above. The specific implementation is as follows:

[0048] The original data is standardized by adjusting the mean of each variable to 0 and the variance to 1 to eliminate the influence of different dimensions and orders of magnitude between variables.

[0049] Calculate the covariance matrix between the variables based on the standardized data.

[0050] The covariance matrix is ​​decomposed into eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​represent the magnitude of the variance of the principal components, and the eigenvectors represent the direction of the principal components.

[0051] Based on the magnitude of the eigenvalues, the top k largest eigenvalues ​​and their corresponding eigenvectors are selected as principal components.

[0052] The raw data is projected onto the selected principal components to obtain the principal component scores.

[0053] The principal component scores were used to analyze and filter the 18 parameters, ultimately selecting 13 input features and 4 output features, totaling 17 feature parameters, as the sample set for training the Long Short-Term Memory Neural Network.

[0054] S103, construct a multi-objective optimization function that minimizes the difference between the horizontal and vertical positions of the front and rear ends of the shield. Combine the reduced shield parameters and use a non-dominated genetic algorithm to solve the multi-objective optimization function to obtain the optimal shield parameters.

[0055] In this embodiment, the expression for the multi-objective optimization function that minimizes the difference between the horizontal and vertical positions of the shield's front and rear ends is: ,in, This represents the predicted value of the i-th solution to objective n (n takes the value 1-4). The true value of target n is the coordinate value of the horizontal and vertical positions of the front and rear ends of the shield. In other words, it is the difference between the horizontal position of the front end, the vertical position of the front end, the horizontal position of the rear end, and the vertical position of the rear end. The minimum value of S is taken as the optimal shield attitude parameter for calculation.

[0056] The non-dominated genetic algorithm employs the third generation of non-dominated genetic algorithms. Non-dominated Sorting Genetic Algorithm III (NSGA-III) is an improved version of NSGA-II (hereinafter referred to as the genetic algorithm), primarily handling problems with four or more objectives (multi-objective optimization problems) (Xu et al., 2025). When the problem involves four or more objectives, the main difficulties include the non-dominance of the population, inefficient recombination operations, difficulty in representing high-dimensional objective surfaces, high computational cost of performance metrics, and increased visualization difficulty. The core improvement of NSGA-III lies in its selection operation, particularly the maintenance of diversity among population members. This characteristic is achieved through the provision of reference points and a dynamic update mechanism. Its calculation formula is as follows:

[0057]

[0058]

[0059]

[0060]

[0061] In the formula: zi represents the extreme point; D represents the spatial dimension; a i Represents the intercept between the hyperplane and the target axis; f i (x) represents individual x in the i-th population. i The value of the i-th objective function; w i Represents the weight vector; grid x Represents the position index of individual x in the grid; grid i Represents the position index of grid i; The value represents the density of the grid; d is the grid unit length; floor() represents rounding down.

[0062] The above formula is used to explore the multidimensional space. The grid congestion is dynamically adjusted according to decision preferences. The weighting method is used to screen the Pareto optimal solution in the multidimensional space to achieve optimal consideration of multiple objectives.

[0063] The above embodiments can determine the Pareto front set for optimizing the shield tunneling front and rear ends at four optimization objectives (horizontal / vertical). However, these correspond to multiple sets of operational parameter combinations, which differ to some extent. Therefore, this embodiment employs a multi-objective optimization function, incorporating the combinations of various shield tunneling operational parameters into the prediction model. The results of the four index parameters are then output, and the absolute values ​​of these differences are calculated by comparing them with the actual measured values ​​on-site. Finally, the results are summed to minimize the total difference S among the four objectives. The historical data processing described above primarily involves model training to obtain the corresponding solution set. Combined with practical engineering considerations, weights can be assigned to the four absolute values ​​to select the optimal solution from the obtained results.

[0064] In one embodiment, before solving the multi-objective optimization function using a non-dominated genetic algorithm to obtain the optimal shield tunneling parameters, the method further includes, based on the steps described in this embodiment, adjusting the search strategy of the non-dominated genetic algorithm using an ant colony algorithm.

[0065] The ant colony algorithm provided in this embodiment is an improved ant colony optimization (IACO) algorithm. It is based on the traditional ant colony optimization (ACO) algorithm, but with parameter improvements to its heuristic function, thus improving the overall computational efficiency (Moissa et al., 2021). Its basic process is as follows: Basic grid parameters are set, and a certain number of ant colonies are generated within the grid. Various features act as "food" to attract the ant colony. During the search for "food," each ant releases pheromones along its path, influencing the path selection of the next ant. Finally, the shortest path is selected by recording the total path distance. The calculation formula is as follows:

[0066] ;in, η ij Representative node i To the node j Heuristic functions; E j and E S Represents the energy level of the corresponding node; N i Representative node i Nearby node set.

[0067] The improved ant colony algorithm refers to a probabilistic search performed according to the state transition rule in each iteration, an enhanced global pheromone update rule for the pheromone of the optimal search path obtained in each iteration, and a pheromone local update rule that performs local updates when constructing the search path and performs a local update of the search path after each iteration.

[0068] Specifically, in the ant colony algorithm, ants perform probabilistic path searches with a bias based on a random proportion rule. In the improved ant colony algorithm, ants use a state transition rule for searching. This rule not only stores the information already obtained but also allows for biased searches based on the specific problem.

[0069] In the ant colony algorithm, all ants perform a global pheromone update, which can prevent ants from quickly finding the shortest path, further reducing the algorithm's search efficiency. In the improved ant colony algorithm, however, not all ants perform a global update. Instead, after each iteration, pheromone enhancement is only applied to the optimal path. Pheromones on non-optimal paths are left unenhanced, allowing them to gradually dissipate and decrease in concentration. This ultimately leads more ants to choose the optimal path, significantly improving the algorithm's search efficiency and shortening the search time.

[0070] In the ant colony algorithm, a local update is performed on all paths after each iteration. In the improved ant colony algorithm, ants not only perform local updates while constructing paths, but also again after each iteration. This makes previously visited paths increasingly less attractive to ants, indirectly encouraging them to search for unvisited paths and preventing them from converging on the same path. Without local updates, all ants would search within the narrow neighborhood of previously found optimal paths, thus preventing further searches for better solutions. Therefore, the introduction of the pheromone local update rule effectively enhances the optimization performance of the ant colony algorithm.

[0071] Specifically, the expression for the state transition rule is:

[0072] Where J represents the set of nodes that are directly reachable from the node and are not already visited in the node sequence. For nodes i To the node j pheromones, For nodes i To the node j Heuristic functions, As a pheromone-stimulating factor, β Let be the expected heuristic factor, q represent a random number, q0 represent a random number between 0 and 1, and S represent the probability calculated by the probability calculation function.

[0073] Specifically, q0 is a crucial control parameter introduced in the improved ant colony algorithm. In the state transition rules of the improved ant colony algorithm, the probability of an ant choosing the current optimal movement direction is q0. Simultaneously, the ant searches each edge with a biased probability of (1-q0). By adjusting q0, the balance between "exploration" and "biased exploration" can be effectively adjusted, thus determining whether the algorithm's next move should be to explore new areas or remain near the current optimal path. In other words, the magnitude of the parameter q0 determines the relative importance of utilizing prior knowledge and exploring new paths.

[0074] The expression for the global pheromone update rule is: Where C represents the length of the path chosen by the ant. Indicates the global volatile factor of pheromones. This represents the amount of pheromone released by an ant as it passes through the edge formed by nodes i and j.

[0075] Specifically, in the improved ant colony algorithm, only ants that have constructed the shortest path from the start of the algorithm to the current iteration are allowed to release pheromones. This means that pheromones are enhanced on that path by a strength of Q. This strategy, combined with the state transition rule, significantly enhances the algorithm's search guidance. In each iteration, the global pheromone update rule is only used after all ants have constructed their paths. Both pheromone evaporation and release only occur on edges belonging to the current best path, a significant difference from the basic ant colony algorithm. This is because the basic ant colony algorithm updates pheromones on all edges, resulting in a computational complexity of O(n^2). 2 In the improved ant colony algorithm, the computational complexity of pheromone updates is O(n), which significantly reduces the search complexity and can effectively improve the solution efficiency of the non-dominated genetic algorithm.

[0076] The expression for the pheromone local update rule is: ,in, This indicates the local volatile factor of pheromones. The chaotic factor generated by the chaotic mapping function is represented by k, which represents the current iteration number, and K represents the maximum iteration number. This represents the initial value of the pheromone.

[0077] Specifically, for the local update rule, this embodiment introduces a chaotic factor generated by a chaotic mapping function to determine the local pheromone evaporation factor, thereby improving the diversity of local search, ensuring the pheromone update performance, and reducing the probability that ants will choose the path that has been affected by the pheromone local update rule.

[0078] S104, the pre-constructed long short-term memory neural network is trained using the optimal shield parameters. After training, an attitude prediction model that can be used for shield machine attitude prediction is obtained.

[0079] In this embodiment, the gates in the Long Short-Term Memory Neural Network include an input gate, a forget gate, and an output gate, and their calculation formulas are as follows: In the formula: f t for t The output of the forget gate; σ For activation functions; W f and U f This represents the weight vector of the forget gate; x t for t The input vector of the time-series memory unit; h t-1 for t The output vector of the memory unit at time -1; b f This represents the bias vector of the forget gate.

[0080]

[0081]

[0082] In the formula: i t The output of the input gate; tanh For activation functions; W i , W c , U i , U c The input gate weight vector; C t for t Real-time status information; C’ t for t Time slot candidate information; b i and b c is the bias vector of the input gate.

[0083]

[0084] In the formula: o t for t The output of the output gate is always being output;W o and U o This is the weight vector of the output gate; b o This is the bias vector for the output gate; h t For memory units in t The value at any given moment.

[0085] The method of training a pre-constructed long short-term memory neural network using optimal shield tunneling parameters is common knowledge and will not be described in this embodiment.

[0086] In summary, this invention utilizes an improved ant colony algorithm to adjust the search strategy of the non-dominated genetic algorithm. This adjustment accelerates the efficiency and improves the accuracy of the non-dominated genetic algorithm in solving the multi-objective optimization function that minimizes the difference between the horizontal and vertical positions of the shield's front and rear ends. Then, based on this optimal shield parameters, a pre-constructed long short-term memory neural network is trained. Upon completion of training, an attitude prediction model suitable for shield machine attitude prediction is obtained. Because the adjusted non-dominated genetic algorithm improves the accuracy of the solution, the prediction accuracy of the attitude prediction model trained based on the solved optimal shield parameters is also correspondingly improved. This solves the problems of insufficient computational efficiency and prediction accuracy in existing attitude prediction methods.

[0087] In one embodiment, before performing step S102, the model building method described above further includes: using the KS parameter verification method to perform normality verification on historical shield tunneling parameters in order to remove outlier data from the historical shield tunneling parameters.

[0088] Specifically, using SPSS software, the KS statistic was calculated for each parameter. The calculated values ​​were all greater than 0.05, indicating that the parameters conform to a normal distribution at a 95% confidence interval. When the parameter sample conforms to or approximately conforms to a normal distribution, 3... σ The rules eliminate outlier data to ensure the accuracy of historical tunnel boring machine parameters.

[0089] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating an attitude prediction method provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the method includes:

[0090] S201, Obtain the shield machine parameters in the current time series;

[0091] S202, the shield parameters are input into the attitude prediction model constructed by the attitude prediction model construction method described in the above embodiment, and the attitude prediction result of the shield machine in the next time series is output.

[0092] In this embodiment, the present invention utilizes an improved ant colony algorithm to adjust the search strategy of the non-dominated genetic algorithm. This adjustment improves the efficiency and accuracy of the non-dominated genetic algorithm in solving the multi-objective optimization function that minimizes the difference between the horizontal and vertical positions of the shield front and rear ends. Then, based on this optimal shield parameters, a pre-constructed long short-term memory neural network is trained. Upon completion of training, an attitude prediction model suitable for shield machine attitude prediction is obtained. Because the adjusted non-dominated genetic algorithm improves the accuracy of solving the multi-objective optimization function, the prediction accuracy of the attitude prediction model trained based on the solved optimal shield parameters is also correspondingly improved. This solves the problems of insufficient computational efficiency and prediction accuracy in existing attitude prediction methods.

[0093] This invention also provides an electronic device. The electronic device includes a processor, a memory, a communication interface, and at least one communication bus for connecting the processor, the memory, and the communication interface. The memory includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (PROM), or portable read-only memory (CD-ROM), and is used for related instructions and data.

[0094] The communication interface is used to receive and send data. The processor can be one or more CPUs; if it is a single CPU, it can be a single-core CPU or a multi-core CPU. The processor in the electronic device reads one or more programs stored in memory and performs the following operations: obtaining historical shield machine parameters in a historical time series; performing dimensionality reduction filtering on the historical shield machine parameters to obtain dimensionality-reduced shield machine parameters; constructing a multi-objective optimization function that minimizes the difference between the horizontal and vertical positions of the shield machine's front and rear ends; combining the dimensionality-reduced shield machine parameters, solving the multi-objective optimization function using a non-dominated genetic algorithm to obtain the optimal shield machine parameters; training a pre-constructed long short-term memory neural network using the optimal shield machine parameters; after training, obtaining an attitude prediction model that can be used for shield machine attitude prediction; and also performing the following operations: obtaining the shield machine parameters in the current time series; inputting the shield machine parameters into the attitude prediction model, and outputting the attitude prediction result of the shield machine in the next time series.

[0095] It should be noted that the specific implementation of each operation can be described above. Figure 1 The corresponding descriptions of the method embodiments shown indicate that electronic devices can be used to execute an attitude prediction model construction method and an attitude prediction method according to the above method embodiments of this application, which will not be described in detail here.

[0096] This invention also provides a computer-readable storage medium, which is a memory device in a computer device for storing programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the attitude prediction model construction method and attitude prediction method in the above embodiments. Those skilled in the art should understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can 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.

[0097] This invention also provides a computer program product containing program instructions. The computer program product may be software or program products containing program instructions, capable of running on a computing device or stored on any usable medium. When the computer program product runs on at least one electronic device, it causes the at least one electronic device to execute an attitude prediction model construction method and an attitude prediction method.

[0098] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. 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 within the scope of protection of the present invention.

Claims

1. A method for constructing an attitude prediction model, characterized in that the method... include: Obtain historical shield tunneling parameters of the tunnel boring machine in a historical time series; Dimensionally reduce and filter historical shield tunneling parameters to obtain the reduced shield tunneling parameters; A multi-objective optimization function is constructed to minimize the difference between the horizontal and vertical positions of the front and rear ends of the tunnel boring machine (TBM). The TBM parameters are then combined with the reduced-dimensional TBM parameters, and a non-dominated genetic algorithm is used to solve the multi-objective optimization function to obtain the optimal TBM parameters. Before using the non-dominated genetic algorithm to solve the multi-objective optimization function to obtain the optimal TBM parameters, the method further includes: adjusting the search strategy of the non-dominated genetic algorithm using an improved ant colony algorithm. The improved ant colony algorithm refers to a system that performs probabilistic search according to state transition rules in each iteration, enhances the global pheromone update rule for the optimal search path obtained in each iteration, and performs local updates during search path construction, with a local update of the search path after each iteration; the expression for the state transition rule is as follows: Where J represents the set of nodes that are directly reachable from the node and are not already visited in the node sequence. For nodes i To the node j pheromones, For nodes i To the node j Heuristic functions, As a pheromone-stimulating factor, β Let q be the expected heuristic factor, q represent a random number, q0 represent a random number between 0 and 1, and S represent the probability calculated by the probability calculation function. The expression for the global pheromone update rule is: Where C represents the length of the path chosen by the ant. Indicates the global volatile factor of pheromones. This represents the amount of pheromone released by an ant as it passes through the edge formed by nodes i and j. The expression for the pheromone local update rule is: ,in, This indicates the local volatile factor of pheromones. This represents the chaos factor generated by the chaotic mapping function, k represents the current iteration number, and K represents the maximum iteration number. Indicates the initial value of the pheromone; The pre-constructed long short-term memory neural network is trained using the optimal shield parameters. After training, an attitude prediction model that can be used for shield machine attitude prediction is obtained.

2. The method for constructing an attitude prediction model according to claim 1, characterized in that, The historical shield tunneling parameters include the shield's positional parameters and its operational parameters.

3. The method for constructing an attitude prediction model according to claim 1, characterized in that, Before performing dimensionality reduction and filtering on historical shield tunneling parameters to obtain the dimensionality-reduced shield tunneling parameters, the method further includes: using the KS parameter verification method to perform normality verification on the historical shield tunneling parameters in order to remove outlier data from the historical shield tunneling parameters.

4. The method for constructing an attitude prediction model according to claim 1, characterized in that, Principal component analysis was used to reduce the dimensionality of historical shield tunneling parameters and obtain the reduced shield tunneling parameters.

5. A posture prediction method, characterized in that the method include: Obtain the shield machine parameters in the current time series; The shield tunneling parameters are input into the attitude prediction model constructed by the attitude prediction model construction method as described in any one of claims 1 to 4, and the attitude prediction result of the shield tunneling machine in the next time series is output.

6. An electronic device, characterized in that, Including memory and processor; A memory for storing computer programs, the computer programs including program instructions; A processor is configured to execute the program instructions to cause the electronic device to perform an attitude prediction model construction method as described in any one of claims 1 to 4 and an attitude prediction method as described in claim 5.

7. A computer program product containing program instructions, characterized in that, When the program instructions are executed by the electronic device, the electronic device performs an attitude prediction model construction method as described in any one of claims 1 to 4 and an attitude prediction method as described in claim 5.

Images