Target cutter cylinder bolt detection method and system based on deep learning and data simulation

Abstract

The embodiment of the invention provides a target cutter cylinder bolt detection method and system based on deep learning and data simulation, and the method comprises the steps: obtaining the pre-tightening force data of each fastening bolt on a target cutter cylinder, and the configuration data of the target cutter cylinder; generating a first target knife cylinder stress simulation model corresponding to the pre-tightening force data according to the pre-tightening force data and the configuration data, and inputting the pre-tightening force data, the configuration data and the first target knife cylinder stress simulation model into a pre-trained neural network depth self-learning algorithm to obtain an algorithm verification result. And if the algorithm verification result is that the verification is passed, the loosening condition of each fastening bolt is determined according to the first target cutter cylinder stress simulation model, so that the efficiency and effect of monitoring the loosening of the fastening bolts of the target cutter cylinder of the normal-pressure cutter head are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more particularly, to a method and system for detecting target cutterhead bolts based on deep learning and data simulation. Background Art

[0002] In the field of shield tunneling construction, shield machines with an atmospheric pressure cutterhead design are widely used and play an important role in underground engineering construction. However, during actual construction, the rolling cutters and scraping cutters in the target cutterhead of the shield machine are subjected to the acting force between the cutters and the rock ahead during operation. When the thrust is uneven, vibration is likely to occur. This vibration poses a risk of loosening or even falling off the fastening bolts of the target cutterhead, which may lead to problems such as the target cutterhead retracting and water ingress, and in severe cases, major accidents may occur, posing a great threat to the safety and progress of the project.

[0003] Currently, the existing solution to the above problem is to manually enter the chamber to check the looseness of the fastening bolts. However, this method has many deficiencies. On the one hand, entering the chamber for inspection affects the safety of the operators, who have to face a complex and dangerous underground environment. On the other hand, it increases the construction operation cost.

[0004] Therefore, how to improve the efficiency and effectiveness of monitoring the looseness of the fastening bolts of the target cutterhead of the atmospheric pressure cutterhead is an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the purpose of the present application is to provide a method and system for detecting target cutterhead bolts based on deep learning and data simulation to improve the efficiency and effectiveness of monitoring the looseness of the fastening bolts of the target cutterhead of the atmospheric pressure cutterhead.

[0006] Combined with the first aspect of the present application, a method for detecting target cutterhead bolts based on deep learning and data simulation is provided. The method includes: Obtain the pre-tightening force data of each fastening bolt on the target cutterhead and the configuration data of the target cutterhead; Generate a first target cutterhead stress simulation model corresponding to the pre-tightening force data according to the pre-tightening force data and the configuration data; Input the pre-tightening force data, the configuration data, and the first target cutterhead stress simulation model into a pre-trained neural network deep self-learning algorithm to obtain an algorithm verification result. The algorithm verification result includes a data verification result for indicating the verification result of the pre-tightening force data and the configuration data, and a simulation model verification result for indicating the verification result of the first target cutterhead stress simulation model; If the algorithm verification result is verified to pass, determine the looseness of each fastening bolt according to the first target cutterhead stress simulation model.

[0007] Optionally, inputting the pre-tightening force data, the configuration data, and the first target tool cylinder stress simulation model into a pre-trained neural network deep self-learning algorithm to obtain an algorithm verification result includes: Inputting the pre-tightening force data, the configuration data, and the first target tool cylinder stress simulation model into the pre-trained neural network deep self-learning algorithm. The pre-trained neural network deep self-learning algorithm calculates the pre-tightening force data, the configuration data, and the first target tool cylinder stress simulation model based on pre-learned model parameters to obtain predicted pre-tightening force data and predicted configuration data. The model parameters are obtained by training based on sample data including sample pre-tightening force data, sample configuration data, and sample simulation models; Matching the predicted pre-tightening force data with the pre-tightening force data and matching the predicted configuration data with the configuration data to obtain the data verification result; If the target data with a certainty less than or equal to a preset certainty threshold is included in the predicted pre-tightening force data and / or the predicted configuration data, re-simulate according to the target data to obtain a second target tool cylinder stress simulation model. The certainty is determined by the confidence or variance of the predicted pre-tightening force data and / or the predicted configuration data, and the corresponding confidence threshold or variance threshold; Matching the first target tool cylinder stress simulation model and the second target tool cylinder stress simulation model to obtain the simulation verification result; Obtain the algorithm verification result according to the simulation verification result and the data verification result.

[0008] Optionally, the matching the predicted pre-tightening force data with the pre-tightening force data and matching the predicted configuration data with the configuration data to obtain the data verification result includes: Obtain the first relative error and the first absolute error corresponding to the predicted pre-tightening force data and the pre-tightening force data; Determine the first data verification result according to the first relative error, the first absolute error, the first preset relative error threshold, and the first preset absolute error threshold; Obtain the second relative error and the second absolute error corresponding to the predicted configuration data and the configuration data; Determine the second data verification result according to the second relative error, the second absolute error, the second preset relative error threshold, and the second preset absolute error threshold; Obtain the data verification result according to the first data verification result and the second data verification result.

[0009] Optionally, matching the first target tool cylinder stress simulation model and the second target tool cylinder stress simulation model to obtain the simulation verification result includes: Extracting a first stress distribution feature in the first target tool cylinder stress simulation model and a second stress distribution feature in the second target tool cylinder stress simulation model, where the stress distribution feature includes maximum stress, average stress, and stress distribution; Calculating a difference index between the first stress distribution feature and the second stress distribution feature, where the difference index includes maximum stress difference, average stress difference, and stress distribution difference; Determining the verification result of the maximum stress difference according to the maximum stress difference threshold, determining the verification result of the average stress difference according to the average stress difference threshold, and determining the verification result of the stress distribution difference according to the overlap degree threshold; Obtaining the simulation verification result based on the verification result of the maximum stress difference, the verification result of the average stress difference, and the verification result of the stress distribution difference.

[0010] Optionally, determining the loosening condition of each fastening bolt according to the first target tool cylinder stress simulation model includes: Obtaining a set of stress thresholds corresponding to the loosening condition of each fastening bolt, where the set of stress thresholds is pre-trained by a deep learning model based on stress distribution sample data under different tightening torques in a stress distribution database, and the set of stress thresholds is used to indicate the loosening stress thresholds of each point near the corresponding fastening bolt; Generating a loosening detection stress reference model according to the set of stress thresholds corresponding to the loosening condition of each fastening bolt; Determining the loosening condition of each fastening bolt according to the change of each pixel point area in the first target tool cylinder stress simulation model and the stress threshold corresponding to the corresponding pixel point area in the loosening detection stress reference model.

[0011] Optionally, generating the first target tool cylinder stress simulation model corresponding to the pre-tightening force data according to the pre-tightening force data and the configuration data includes: Converting the initial pre-tightening force parameters corresponding to each fastening bolt in the pre-tightening force data into a pre-tightening force parameter conversion result with a unified dimension, where the unified dimension is Newton-meter; Converting the material property parameters, geometric parameters, and assembly parameters included in the configuration data into configuration parameter conversion results with a unified dimension, where the material property parameters include elastic modulus and Poisson's ratio, the geometric parameters include bolt diameter and contact surface size, and the assembly parameters include installation angle and torque direction; Generate an initial target cutter barrel mesh model based on finite element mesh division according to the material property parameters and geometric parameters in the pre-tightening force parameter conversion result and the configuration parameter conversion result; Apply corresponding pre-tightening force loads to the initial target cutter barrel mesh model according to the pre-tightening force values corresponding to the fastening bolts in the pre-tightening force parameter conversion result to generate a target cutter barrel mesh model after load application; Set contact surface constraint conditions and boundary conditions in the target cutter barrel mesh model after load application according to the assembly parameters in the configuration parameter conversion result to generate a target cutter barrel mesh model after constraint loading; Perform stress distribution calculation on the target cutter barrel mesh model after constraint loading to obtain the stress values of each mesh node and generate a stress distribution data set; Generate a three-dimensional stress distribution diagram in the first target cutter barrel stress simulation model according to the stress values of each mesh node and the corresponding coordinate positions in the stress distribution data set, where the three-dimensional stress distribution diagram contains the stress values of each pixel position area.

[0012] Optionally, generating a loosening detection stress reference model according to the stress threshold set corresponding to the loosening conditions of the fastening bolts includes: Obtain the loosening stress threshold data corresponding to the identification numbers of the fastening bolts in the stress threshold set, where the loosening stress threshold data includes the stress thresholds of each pixel position in a preset area around each fastening bolt under different tightening torques; Extract the coordinate ranges of the corresponding bolt installation areas in the first target cutter barrel stress simulation model according to the identification numbers of the fastening bolts in the stress threshold set to generate a bolt area division data set; Extract the stress thresholds of each pixel position in the corresponding area from the stress threshold set according to each coordinate range in the bolt area division data set to generate a bolt area stress threshold data set; Convert the stress thresholds of each pixel position in the bolt area stress threshold data set into the same dimension unit as the stress values in the first target cutter barrel stress simulation model to generate a normalized stress threshold data set; Generate the loosening detection stress reference model corresponding to the first target cutter barrel stress simulation model according to the stress thresholds and their coordinate positions of each pixel position in the normalized stress threshold data set, where the loosening detection stress reference model contains the normalized stress thresholds of each pixel position area.

[0013] Optionally, determining the loosening condition of each of the fastening bolts according to the change of each pixel point area in the first target tool cylinder stress simulation model and the stress threshold corresponding to the corresponding pixel point area in the loosening detection stress reference model includes: Dividing the three-dimensional stress distribution map in the first target tool cylinder stress simulation model into a plurality of pixel point areas, and the size of each pixel point area is the same as that of the pixel point area in the loosening detection stress reference model; Extracting the current stress value of each pixel point area from the first target tool cylinder stress simulation model to generate a current stress value set; Extracting the normalized stress threshold corresponding to each pixel point area from the loosening detection stress reference model to generate a normalized stress threshold set; Comparing the current stress value of each pixel point area in the current stress value set with the normalized stress threshold of the corresponding area in the normalized stress threshold set to generate the stress difference of each pixel point area; Counting the number of pixel point areas in the bolt installation area corresponding to each fastening bolt where the stress difference exceeds the preset loosening threshold to generate a loosening times statistical result; Determining the loosening level of each fastening bolt according to the exceeding times of each bolt installation area in the loosening times statistical result and the preset loosening times threshold, where the loosening level includes not loosened, slightly loosened, and severely loosened.

[0014] Combined with the second aspect of the present application, a target tool cylinder bolt detection system based on deep learning and data simulation is provided. The target tool cylinder bolt detection system based on deep learning and data simulation includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the target tool cylinder bolt detection system based on deep learning and data simulation implements the foregoing target tool cylinder bolt detection method based on deep learning and data simulation.

[0015] Combined with the third aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are executed, the foregoing target tool cylinder bolt detection method based on deep learning and data simulation is implemented.

[0016] Combined with the fourth aspect of the present application, a computer program product is provided. When the computer program is executed by a processor, the foregoing target tool cylinder bolt detection method based on deep learning and data simulation is implemented.

[0017] Combined with any of the above aspects, by obtaining the pre-tightening force data of each fastening bolt on the target tool cylinder and the configuration data of the target tool cylinder, according to the pre-tightening force data and the configuration data, a first target tool cylinder stress simulation model corresponding to the pre-tightening force data is generated. Input the pre-tightening force data, the configuration data, and the first target tool cylinder stress simulation model into a pre-trained neural network deep self-learning algorithm to obtain an algorithm verification result. If the algorithm verification result is verified to pass, then determine the loosening condition of each fastening bolt according to the first target tool cylinder stress simulation model, thereby improving the efficiency and effect of detecting the loosening condition of the fastening bolts of the target tool cylinder. Brief Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained in combination with these drawings.

[0019] Figure 1 Flow chart of the target tool cylinder bolt detection method based on deep learning and data simulation provided by the embodiments of the present application; Figure 2 Structural schematic diagram of a mechanical disk provided by an embodiment of the present application. Detailed Embodiments

[0020] In order to enable those skilled in the art of the present technology to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0021] The terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product, or terminal that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or terminals.

[0022] References to "embodiments" in this specification mean that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0023] Figure 1 The flowchart of the target cutterhead bolt detection method based on deep learning and data simulation provided by the embodiments of the present application is shown. It should be understood that in other embodiments, the order of some steps of the target cutterhead bolt detection method based on deep learning and data simulation in this embodiment can be shared according to actual needs, or some of the steps can also be omitted or maintained. The details of the target cutterhead bolt detection method based on deep learning and data simulation include: S101. Obtain the pre-tightening force data of each fastening bolt on the target cutterhead and the configuration data of the target cutterhead.

[0024] Among them, the pre-tightening force in the pre-tightening force data refers to the axial tension applied to make the connection of the fastening bolt reach reliable fastening when tightening the fastening bolt. For the fastening bolts on the target cutterhead, an appropriate pre-tightening force can ensure the stability and sealing of the target cutterhead structure. For example, taking the fastening bolts on the target cutterhead of a shield machine as an example, if the pre-tightening force is insufficient, the stability of the target cutterhead will be poor, and if the pre-tightening force is too large, the fastening bolt or the target cutterhead may be damaged. In subsequent embodiments, the fastening bolt can also be simply referred to as a bolt, and its meaning is the same.

[0025] The configuration data includes various characteristic information of the target cutterhead, such as the material characteristics of the target cutterhead (such as material, hardness, etc.), dimensional specifications (diameter, length, etc.), and the installation method, position, and bolt size of the bolts.

[0026] In this step, the pre-tightening force data can be obtained by using sensors. For example, for the fastening bolts of the target cutterhead, sensor sensing holes can be customized according to the number of fastening bolts and the positions of the bolt mounting holes of the fastening bolts, and sensing sensors can be embedded in each sensor sensing hole. Among them, the sensor can be, for example, a pressure sensor, which can detect the corresponding pre-tightening force data when the fastening bolt deforms due to the pre-tightening force of the fastening bolt when the target cutterhead is locked at a fixed torque.

[0027] Exemplarily, Figure 2 is a schematic structural diagram of a mechanical disc provided by the embodiments of the present application. As Figure 2As shown in the figure, the mechanical disk includes bolt mounting holes with the same number as the fastening bolts. In each bolt mounting hole, in order to ensure that the sensing sensor can accurately sense the stress change, a concave hole is designed on the side close to the target cutter barrel, and the sensing sensor is embedded in the groove. When the fastening bolt locks the target cutter barrel under a fixed torque, the sensing sensor will deform due to the pre-tightening force of the fastening bolt, and then detect the corresponding pre-tightening force data.

[0028] The configuration data can be pre-determined configuration-related data obtained from materials such as the design drawings, product manuals, or production records of the target cutter barrel.

[0029] S102. Generate a first target cutter barrel stress simulation model corresponding to the pre-tightening force data according to the pre-tightening force data and the configuration data.

[0030] The stress simulation model is a virtual model that simulates the internal stress distribution of the target cutter barrel under the action of the pre-tightening force through computer software. With the help of this model, the stress magnitude and distribution law of each part of the target cutter barrel can be intuitively understood.

[0031] In this step, the obtained pre-tightening force data and configuration data can be input into professional simulation software (such as COMSOL Multiphysics). The simulation software can classify and integrate the stress ring data according to the input pre-tightening force data and configuration data according to physical laws, and then perform data fitting simulation on the stress ring. During the simulation process, the simulation software takes into account factors such as the material properties of the target cutter barrel and the action mode of the pre-tightening force of the fastening bolt, and finally presents the stress distribution of the entire stress ring in the form of an image to generate the first target cutter barrel stress simulation model.

[0032] S103. Input the pre-tightening force data, the configuration data, and the first target cutter barrel stress simulation model into a pre-trained neural network deep self-learning algorithm to obtain an algorithm verification result.

[0033] Among them, the algorithm verification result includes a data verification result for indicating the verification result of the pre-tightening force data and the configuration data, and a simulation model verification result for indicating the verification result of the first target cutter barrel stress simulation model.

[0034] Among them, the neural network deep self-learning algorithm is a machine learning algorithm based on neural networks, which can automatically learn features and laws from a large amount of data and improve the prediction accuracy by continuously adjusting its own parameters. In this solution, the neural network deep self-learning algorithm is used to verify the pre-tightening force data, the configuration data, and the stress simulation model.

[0035] Data verification refers to comparing the collected pre-tightening force data and configuration data with the pre-learned patterns through a neural network deep self-learning algorithm to determine the accuracy and reliability of the data. Simulation verification refers to re-performing simulations on the data that is uncertain after being compared by the neural network deep self-learning algorithm, and comparing the results of the two simulations before and after to verify the accuracy of the first target barrel stress simulation model.

[0036] In this step, the pre-tightening force data, configuration data, and the first target barrel stress simulation model can be input into a pre-trained neural network deep self-learning algorithm. This neural network deep self-learning algorithm can calculate and analyze these inputs based on the pre-learned model parameters. For data verification, the neural network deep self-learning algorithm can simulate and compare the input pre-tightening force data and configuration data with the standard data patterns it has learned to determine the accuracy of the data and obtain the data verification result. For simulation verification, if the neural network deep self-learning algorithm identifies uncertain data, it can re-perform simulations on this data, and then compare the layer display results of the two simulations before and after to obtain the simulation model verification result.

[0037] Finally, based on whether the accuracy of the pre-tightening force data collected by the sensor characterized by the data verification result meets the expectations, and whether the first target barrel stress simulation model characterized by the simulation model verification result conforms to the actual situation, the algorithm verification result is obtained.

[0038] For example, if both the data verification result and the simulation model verification result meet the expectations and the actual situation, the algorithm verification result passes, and the loosening conditions of each fastening bolt can be determined based on this first target barrel stress simulation model to monitor the loosening conditions of each fastening bolt in real time; if either the data verification result or the simulation model verification result does not meet the expectations or the actual situation, the algorithm verification result fails, and the loosening conditions of each fastening bolt cannot be determined based on this first target barrel stress simulation model to monitor the loosening conditions of each fastening bolt in real time.

[0039] S104. If the algorithm verification result is verified to pass, determine the loosening conditions of each fastening bolt according to the first target barrel stress simulation model.

[0040] In this step, it can be judged from the image of the first target barrel stress simulation model. For example, according to the pre-obtained bolt loosening warning value, the stress or stress change represented by each pixel point in the image can be measured. If the stress or stress change represented by a certain pixel point exceeds the bolt loosening warning value, it can be judged that the fastening bolt corresponding to the area where the pixel point is located may be loose.

[0041] Alternatively, it is also possible to determine whether the fastening bolt corresponding to the area is loose based on the number of pixel points in the area corresponding to each fastening bolt that exceed the bolt loosening warning value, and a preset number threshold, etc.

[0042] Optionally, the bolt loosening warning value can be determined in advance according to experience, or can be obtained through training with a large amount of sample data by a deep learning algorithm, etc.

[0043] Based on the above steps, by obtaining the pre-tightening force data of each fastening bolt on the target tool holder and the configuration data of the target tool holder, according to the pre-tightening force data and the configuration data, a first target tool holder stress simulation model corresponding to the pre-tightening force data is generated. Input the pre-tightening force data, the configuration data, and the first target tool holder stress simulation model into a pre-trained neural network deep self-learning algorithm to obtain an algorithm verification result. If the algorithm verification result is verified to pass, then determine the loosening condition of each fastening bolt according to the first target tool holder stress simulation model, thereby improving the efficiency and effect of detecting the loosening condition of the fastening bolts of the target tool holder.

[0044] Next, the foregoing step S103 will be introduced in detail: S1031. Input the pre-tightening force data, the configuration data, and the first target tool holder stress simulation model into a pre-trained neural network deep self-learning algorithm.

[0045] Among them, the pre-trained neural network deep self-learning algorithm calculates the pre-tightening force data, the configuration data, and the first target tool holder stress simulation model based on the pre-learned model parameters to obtain predicted pre-tightening force data and predicted configuration data. The model parameters are obtained through training with sample data including the sample pre-tightening force data, the sample configuration data, and the sample simulation model.

[0046] In this step, a multi-layer neural network model can be constructed, for example, including an input layer, several hidden layers (such as 3 - 5 layers, each layer containing a certain number of neurons, for example, 100 - 200 neurons per layer) and an output layer. The input layer receives the pre-tightening force data (for example, input in the form of a vector containing multiple pre-tightening force values, such as [100N, 120N, 110N]), the configuration data (which can be a vector containing multiple configuration parameters, such as [material parameter A = 0.5, structural parameter B = 1.2]) and the first target tool holder stress simulation model (for example, it can be represented in a specific data structure, such as a matrix data containing stress distribution).

[0047] In the hidden layer, an activation function (such as the ReLU function) is used to perform a non-linear transformation on the input data. The model parameters (including the weights and biases between neurons in each layer) are trained based on a large amount of sample data (such as sample pre-tightening force data, sample configuration data, sample simulation models, etc.). For example, the stochastic gradient descent algorithm can be used to continuously adjust the model parameters according to the loss function (such as the mean squared error loss function) to minimize the error between the predicted results of the model for the sample data and the true results. After the model training is completed, the pre-tightening force data, configuration data, and the first target barrel stress simulation model are input into the model, and the model calculates according to the learned parameters and outputs the predicted pre-tightening force data (such as [98N, 118N, 108N]) and the predicted configuration data (such as [material parameter A = 0.48, structural parameter B = 1.18]).

[0048] Optionally, since the number of barrel bolts is usually small and the data collected by the sensing sensors (i.e., pre-tightening force data) is single, in the case of a small dataset, a two-layer neural network can be selected to construct the above multi-layer neural network model. In this case, using a two-layer neural network can converge faster compared to using a three-layer neural network. Moreover, since the response time of the three-layer neural network algorithm is long and the loosening of the fastening bolts of the barrel is real-time, the two-layer neural network algorithm is more suitable for this scenario. In addition, the gradient of the two-layer neural network algorithm is more stable, with less attenuation and higher prediction stability.

[0049] S1032. Match the predicted pre-tightening force data with the pre-tightening force data, and match the predicted configuration data with the configuration data to obtain a data verification result.

[0050] In this step, the predicted pre-tightening force data can be compared with the actually collected pre-tightening force data, and the consistency of the data can be evaluated by calculating the difference, relative error, etc. between the two. Similarly, similar comparisons are also made for the predicted configuration data and the actual configuration data. For the pre-tightening force data, if the difference is within the preset error range, it indicates that the data verification passes; for the configuration data, such as the diameter and material properties of the target barrel, the difference between the predicted value and the actual value is compared. If the error is within the acceptable range, the configuration data verification passes. Combining the comparison results of all pre-tightening force data and configuration data, the data verification result can be obtained.

[0051] Taking the detection of the target cutter barrel of a shield machine as an example, the predicted pre-tightening force data output by the neural network deep self-learning algorithm is compared one by one with the pre-tightening force data actually collected by the sensor. For example, if the predicted pre-tightening force of a certain fastening bolt is 1200 N and the actually collected pre-tightening force is 1210 N, the difference between the two is calculated as 10 N. If the preset error range is ±20 N, then the pre-tightening force data passes the verification. For configuration data, such as the predicted inner diameter of the target cutter barrel is 1500 mm and the actual measurement is 1502 mm, and the error is within the allowable range, the configuration data also passes the verification; or, if the error between the predicted position of the target cutter barrel and the actual measured position of the target cutter barrel is within the allowable range (for example, the distance from the predicted position to the center of the cutter head is d1, and the distance from the actual measured position to the center of the cutter head is d2, and the difference between d1 and d2 is within the allowable range), the configuration data also passes the verification. Based on the comparison results of the pre-tightening force data of all fastening bolts and the various configuration data of the target cutter barrel, the data verification result of the target cutter barrel of the shield machine is obtained.

[0052] S1033. If the target data with a certainty less than or equal to the preset certainty threshold is included in the predicted pre-tightening force data and / or the predicted configuration data, re-simulate according to the target data to obtain the second target cutter barrel stress simulation model.

[0053] Among them, the certainty is determined by the confidence level or variance of the predicted pre-tightening force data and / or the predicted configuration data, and the corresponding confidence level threshold or variance threshold.

[0054] In this step, the certainty can be determined by calculating the confidence level or variance of the predicted pre-tightening force data and / or the predicted configuration data. For example, for the predicted pre-tightening force data [98 N, 118 N, 108 N], assuming their confidence levels are [0.9, 0.85, 0.95] respectively, and the preset certainty threshold is 0.9, it is found that the confidence level 0.85 corresponding to 118 N in the predicted pre-tightening force data is less than the threshold, and it is taken as the target data.

[0055] According to the target data (such as 118 N in the predicted pre-tightening force data), adjust the corresponding parameters (such as adjusting the way or magnitude of the pre-tightening force application), and re-perform the stress simulation of the target cutter barrel. During the simulation process, finite element analysis software can be used to set different boundary conditions and loading conditions to obtain more accurate stress distribution results, so as to obtain the second target cutter barrel stress simulation model.

[0056] S1034. Match the first target cutter barrel stress simulation model and the second target cutter barrel stress simulation model to obtain the simulation verification result.

[0057] The stress distribution in the first target tool holder stress simulation model and the second target tool holder stress simulation model can be compared. For example, represent the stress distribution data in the two simulation models in matrix form and calculate the similarity index between them, such as the cosine similarity. Assume that the stress distribution matrix of the first target tool holder stress simulation model is A and the stress distribution matrix of the second target tool holder stress simulation model is B, then the cosine similarity is:

[0058] where represents the inner product of matrices A and B, represent the norms of matrices A and B respectively. According to the calculated similarity index, set the corresponding threshold. When the similarity index is greater than the threshold, it is considered that the matching is successful and the simulation verification result passes; otherwise, the simulation verification result fails.

[0059] S1035. Obtain the algorithm verification result according to the simulation verification result and the data verification result.

[0060] In this step, the algorithm verification result can be obtained by integrating the simulation verification result and the data verification result. For example, set the weights of the simulation verification result and the data verification result to 0.6 and 0.4 respectively. If the simulation verification result passes (denoted as 1) and the data verification result passes (denoted as 1), then the algorithm verification result is: 0.6×1 + 0.4×1 = 1.

[0061] If the simulation verification result fails (denoted as 0) and the data verification result passes (denoted as 1), then the algorithm verification result is: 0.6×0 + 0.4×1 = 0.4.

[0062] According to the calculated algorithm verification result, set the corresponding threshold. When the algorithm verification result is greater than the threshold, it is considered that the algorithm verification passes; otherwise, the algorithm verification fails.

[0063] In a possible implementation manner, step S1032 can be implemented, for example, through the following implementation manner: Obtain the first relative error and the first absolute error corresponding to the predicted pre-tightening force data and the pre-tightening force data. Determine the first data verification result according to the first relative error, the first absolute error, the first preset relative error threshold, and the first preset absolute error threshold.

[0064] Obtain the second relative error and the second absolute error corresponding to the predicted configuration data and the configuration data. Determine the second data verification result according to the second relative error, the second absolute error, the second preset relative error threshold, and the second preset absolute error threshold.

[0065] Obtain the data verification result according to the first data verification result and the second data verification result.

[0066] In this embodiment, the first absolute error calculation refers to the absolute value of the difference between the predicted value and the true value. For each value in the predicted pre-tightening force data, the difference between it and the corresponding true value in the pre-tightening force data is calculated respectively, and then the absolute value is taken. For example, the predicted pre-tightening force data is F p1 , F p2 , …, F pn , and the pre-tightening force data is F r1 , F r2 , …, F rn , then the calculation formula of the first absolute error E a1 , E a2 , …, E an is , where .

[0067] The first relative error calculation is the ratio of the absolute error to the true value, which can be expressed in percentage form. After obtaining the first absolute error, calculate the first relative error E r1 , E r2 , …, E rn corresponding to each predicted pre-tightening force data. The calculation formula is .

[0068] After determining the first relative error and the first absolute error, the first relative error and the first absolute error can be judged respectively according to the preset first preset relative error threshold and the first preset absolute error threshold. For each predicted pre-tightening force data, if its first absolute error is less than or equal to the first absolute error threshold and its first relative error is less than or equal to the first relative error threshold, it indicates that the predicted pre-tightening force data passes the verification. Based on this, if all the predicted pre-tightening force data meet the above conditions, or more than a specific number of the predicted pre-tightening force data meet the above conditions, the first data verification result is passed, otherwise it is not passed.

[0069] The second absolute error calculation is similar to the first absolute error calculation method. For each parameter value in the predicted configuration data, the difference between it and the corresponding true parameter value in the configuration data is calculated respectively, and then the absolute value is taken. Similarly, the second relative error corresponding to each predicted configuration data can be calculated in a similar manner.

[0070] After determining the second relative error and the second absolute error, the second relative error and the second absolute error can be judged respectively according to a preset second preset relative error threshold and a second preset absolute error threshold. For each parameter value in the prediction configuration data, if its second absolute error is less than or equal to the second absolute error threshold and the second relative error is less than or equal to the second relative error threshold, it indicates that the prediction configuration data passes the verification. Based on this, if all parameter values meet the above conditions, or if more than a specific number of parameter values meet the above conditions, the second data verification result is passed; otherwise, it is not passed.

[0071] Then, according to a preset comprehensive judgment rule, based on the first data verification result and the second data verification result, the data verification result can be obtained. For example, the data verification result is passed only when both the first data verification result and the second data verification result are passed; otherwise, the data verification result is not passed. Alternatively, corresponding weights can be set for each predicted pre-tightening force data and each parameter value in the prediction configuration data, and weighted fusion is performed according to each predicted pre-tightening force data, each parameter value, and their respective corresponding weights, and the data verification result is determined according to the fusion result.

[0072] In a possible implementation manner, step S1033 can be implemented, for example, through the following implementation manner: Extract the first stress distribution feature in the first target tool holder stress simulation model and the second stress distribution feature in the second target tool holder stress simulation model. Among them, the stress distribution feature includes the maximum stress, the average stress, and the stress distribution.

[0073] Calculate the difference indexes between the first stress distribution feature and the second stress distribution feature. Among them, the difference indexes include the maximum stress difference, the average stress difference, and the stress distribution difference.

[0074] Determine the verification result of the maximum stress difference according to the maximum stress difference threshold, determine the verification result of the average stress difference according to the average stress difference threshold, and determine the verification result of the stress distribution difference according to the overlap degree threshold.

[0075] Based on the verification result of the maximum stress difference, the verification result of the average stress difference, and the verification result of the stress distribution difference, obtain the simulation verification result.

[0076] In this embodiment, in the stress simulation model of the first target tool cylinder, the stress values of all nodes in the model can be comprehensively traversed. Taking two-dimensional grid division as an example, starting from the upper left corner node, the stress value of each node can be compared with the recorded maximum stress value row by row and column by column. If the stress value of the current node is larger, the maximum stress value is updated until all nodes are traversed, so as to determine the maximum stress in the first stress distribution characteristic. For the stress simulation model of the second target tool cylinder, the same traversal and comparison method can be adopted.

[0077] To extract the average stress, the node range involved in the calculation in the model can be first determined, for example, the nodes within the effective stress region are selected. In the stress simulation model of the first target tool cylinder, the stress values of all nodes within this range can be added up to obtain the total stress, and then the number of nodes is counted. Dividing the total stress by the number of nodes can obtain the average stress in the first stress distribution characteristic. In the stress simulation model of the second target tool cylinder, the same effective stress region as the first model needs to be determined, and the average stress is calculated in the same way to ensure comparability between the two.

[0078] To extract the stress distribution, the model can be divided into regions. For example, it can be divided according to the structural characteristics of the target tool cylinder, such as different parts, the area around the fastening bolts, etc. For the stress simulation model of the first target tool cylinder, the stress value range and distribution frequency in each region are counted. The stress values can be divided into several intervals, and the number of nodes in each interval is counted to obtain the stress distribution frequency. In the stress simulation model of the second target tool cylinder, the same region division method is used to facilitate comparing the stress distributions of the two models in the same region.

[0079] After extracting the stress distribution characteristics of the two models, the difference indexes between them are calculated, such as the maximum stress difference, average stress difference, and stress distribution difference.

[0080] The maximum stress difference can be obtained by calculating the absolute value of the difference between the maximum stress in the first stress distribution characteristic and the maximum stress in the second stress distribution characteristic. For example, if the maximum stress of the first model is 180 MPa and the maximum stress of the second model is 185 MPa, then the maximum stress difference is |180 - 185| = 5 MPa. This difference reflects the change of the maximum stress in the two models. A larger difference may mean that the change of the target data during re-simulation has a greater impact on the maximum stress.

[0081] The calculation method of the mean stress difference is similar. Subtract the mean stress in the second stress distribution characteristic from the mean stress in the first stress distribution characteristic, and then take the absolute value. For example, if the mean stress of the first model is 100 MPa and the mean stress of the second model is 102 MPa, then the mean stress difference is |100 - 102| = 2 MPa. This difference reflects the change in the mean stress level, and a smaller difference indicates that the overall stress conditions of the two models are relatively close.

[0082] For the stress distribution difference, for each divided area, methods such as the Euclidean distance can be used to compare the stress distribution frequencies of the first model and the second model in this area. For example, in a certain area, the stress distribution frequency vector of the first model is [0.1, 0.2, 0.3, 0.4], and that of the second model is [0.15, 0.18, 0.32, 0.35]. The stress distribution difference in this area is represented by calculating the Euclidean distance between the two. Then, the stress distribution differences of all areas are summarized. A weighted average method can be used, and different weights are assigned according to the importance of each area to obtain a comprehensive stress distribution difference index.

[0083] After obtaining the difference index, the verification results of the maximum stress difference, mean stress difference, and stress distribution difference can be determined according to the corresponding thresholds. For example, the maximum stress difference threshold can be a pre-set standard value, and the calculated maximum stress difference is compared with it. If the maximum stress difference is less than or equal to this threshold, the verification result is passed, indicating that the two models are in good agreement in terms of the maximum stress; the mean stress difference threshold is also pre-set, and the calculated mean stress difference is compared with this threshold. If the mean stress difference is less than or equal to this threshold, the verification result is passed, indicating that the two models are relatively close in terms of the overall stress level; the verification of the stress distribution difference is carried out through the overlap threshold, and this overlap can be obtained by calculating the overlap area of the stress distribution frequencies of the two models in each area, etc. The calculated overlap is compared with the overlap threshold. If the overlap is greater than or equal to this threshold, the verification result is passed, indicating that the two models have a high similarity in stress distribution. For example, the overlap threshold is 0.8, and the calculated overlap is 0.85, and the verification result is passed.

[0084] Finally, based on the verification results of the maximum stress difference, mean stress difference, and stress distribution difference, a comprehensive judgment is made to obtain the simulation verification result.

[0085] For example, when the verification results of the above three difference indicators are all passed, the simulation verification result is passed. This indicates that the first target barrel stress simulation model and the second target barrel stress simulation model have good consistency in terms of maximum stress, average stress, and stress distribution. The model obtained from re-simulation is relatively close to the initial model, and the influence of the uncertainty of the target data on the model is within an acceptable range. The first target barrel stress simulation model can be continuously used for subsequent work such as judging the loosening of fastening bolts.

[0086] If any one of the verification results is not passed, the simulation verification result is not passed. For example, if the verification result of the maximum stress difference is not passed, while the verification results of the average stress difference and stress distribution difference are passed, the overall simulation verification result is still not passed. Feedback and handling can be carried out according to the specific non-passing situation, such as re-checking the target data, evaluating the structural design of the target barrel or the bolt installation position, etc., to improve the accuracy and reliability of the simulation model.

[0087] Next, a detailed introduction will be given on how to determine the loosening situation of each fastening bolt according to the first target barrel stress simulation model in the aforementioned step S104.

[0088] S1041. Obtain the set of stress thresholds corresponding to the loosening situation of each fastening bolt.

[0089] Among them, the set of stress thresholds is pre-trained by a deep learning model based on the stress distribution sample data under different tightening torques in the stress distribution database. The set of stress thresholds is used to indicate the loosening stress thresholds of each point near the corresponding fastening bolt.

[0090] First, a stress distribution database can be constructed, which collects a large amount of stress distribution sample data under different tightening torques. In actual operation, these data can be obtained through experimental tests. For example, for the target barrel of the same specification, the fastening bolts are tightened with different tightening torques, and then the stress values of each point near the fastening bolts are measured using a high-precision stress sensor. The tightening torque and the corresponding stress distribution data are recorded for each experiment, and these data are continuously accumulated into the stress distribution database. At the same time, computer simulation can also be combined to simulate the stress distribution of the target barrel under different tightening torques, and the simulation data is also incorporated into the database.

[0091] Then, based on the data in the stress distribution database, a deep learning model is trained. During the training process, the data in the stress distribution database is used as the input, and the corresponding tightening torque and the loosening condition of the fastening bolt are used as the output. The model will continuously adjust its own parameters to make the mapping relationship between the input and the output more accurate. For example, a convolutional neural network model can be used to automatically extract the features in the stress distribution data. After multiple iterative trainings, the model will learn the characteristic patterns of the stress distribution under different tightening torques, and the correlations between these patterns and the bolt loosening condition.

[0092] After the training is completed, the deep learning model can predict the loosening condition of the bolt according to the input stress distribution data. By analyzing the stress distribution under different tightening torques, the stress thresholds of each point near the bolt corresponding to each tightening torque are determined, and these thresholds form a set of stress thresholds. For example, for a specific tightening torque, the deep learning model can analyze the stress value range of different points near the bolt. When the stress value is lower or higher than this range, it may indicate that the bolt is loose.

[0093] S1042. Generate a loosening detection stress reference model according to the set of stress thresholds corresponding to the loosening condition of each fastening bolt.

[0094] According to the set of stress thresholds, construct the framework of the model. For example, a three-dimensional space model can be used to represent the structures of the target tool holder and the fastening bolt in the three-dimensional space, and then map the stress thresholds of each point near each fastening bolt to the corresponding spatial positions. For example, using the method of grid division, the space around the target tool holder and the bolt is divided into small grid units, each grid unit corresponds to a specific point, and the stress threshold of this point is assigned to the corresponding grid unit.

[0095] Optionally, the constructed model framework can also be optimized and refined. In practical applications, since the stress distribution near the fastening bolt is not uniform and may be affected by factors such as the structure and material properties of the target tool holder. Therefore, the model can be optimized to consider the influence of these factors on the stress thresholds. For example, through methods such as finite element analysis, the structures of the target tool holder and the fastening bolt can be simulated, the influence laws of different factors on the stress distribution can be analyzed, and then the stress thresholds in the model can be adjusted according to these laws.

[0096] Then, verify and evaluate the generated reference model for loose detection stress. The accuracy and reliability of the reference model for loose detection stress can be checked through experimental tests or comparison with known accurate data. For example, conduct a bolt loosening experiment on the actual target tool cylinder, measure the stress values at each point near the bolt, and then compare these values with the stress thresholds in the reference model for loose detection stress. If there are large errors in the reference model for loose detection stress, further correction and improvement are required until the reference model for loose detection stress can accurately reflect the stress threshold distribution near the tightened bolt, thereby generating the reference model for loose detection stress to provide a reference for subsequent determination of the loosening condition of the tightened bolt.

[0097] S1043. Determine the loosening condition of each tightened bolt according to the changes in the pixel point regions in the first target tool cylinder stress simulation model and the stress thresholds corresponding to the pixel point regions in the reference model for loose detection stress.

[0098] First, the first target tool cylinder stress simulation model can be processed to extract the stress information of each pixel point region. In the first target tool cylinder stress simulation model, each pixel point corresponds to a specific position on the target tool cylinder and has a corresponding stress value. Image recognition and data analysis methods can be used to extract the stress information in the first target tool cylinder stress simulation model to form a data set containing the stress values of each pixel point. For example, use an image processing algorithm to segment the image of the model, identify each pixel point, and read its corresponding stress value.

[0099] At the same time, extract the stress thresholds of the corresponding pixel point regions from the reference model for loose detection stress, and find the stress thresholds corresponding to each pixel point by corresponding to the pixel points in the first target tool cylinder stress simulation model. The mapping relationship between pixel points can be established according to the coordinate systems and grid division methods of the two models, so as to accurately obtain the stress thresholds of the corresponding points.

[0100] Then, the stress values of each pixel point in the first target tool cylinder stress simulation model can be compared with the stress thresholds of the corresponding pixel points in the reference model for loose detection stress. For each pixel point, if its stress value is lower or higher than the corresponding stress threshold, it indicates that the tightened bolt near this point has loosened. For example, if the stress value of a certain pixel point is significantly lower than the stress threshold, it means that the pre-tightening force of the bolt has decreased and there are signs of loosening; if the stress value is significantly higher than the stress threshold, it indicates that the bolt is subjected to an additional external force and there may also be a risk of loosening.

[0101] After comparing the stress values and stress thresholds of all pixel points, a comprehensive analysis is carried out on the comparison results of the pixel points near each fastening bolt. The number of pixel points with abnormal stress values (lower or higher than the stress threshold) near each fastening bolt can be counted, or the stress deviation degree of these abnormal pixel points can be calculated. If the number of abnormal pixel points near a certain fastening bolt is large, or the stress deviation degree is large, it can be judged that the fastening bolt may be loose. At the same time, according to the distribution of abnormal pixel points, the degree and position of the fastening bolt loosening can be further judged. For example, if the abnormal pixel points are concentrated on one side of the fastening bolt, it may indicate that the fastening bolt is loose on that side.

[0102] Next, a detailed introduction will be given on how to generate the first target tool cylinder stress simulation model corresponding to the pre-tightening force data according to the pre-tightening force data and configuration data in the foregoing step S102.

[0103] S1021. Convert the initial pre-tightening force parameters corresponding to each fastening bolt in the pre-tightening force data into a pre-tightening force parameter conversion result with a unified dimension.

[0104] Among them, the unified dimension is Newton-meter.

[0105] In this step, the pre-tightening force data can be comprehensively sorted out first to clarify the initial pre-tightening force parameters corresponding to each fastening bolt and their current dimension forms. Then, according to the conversion relationship between different dimensions, each initial pre-tightening force parameter is converted. For example, if the dimension of the initial pre-tightening force parameter is kilonewton, since 1 kilonewton is equal to 1000 Newtons, the kilonewton value needs to be multiplied by 1000 to be converted into Newtons.

[0106] During the conversion process, multiple calculations or different calculation methods can be used for verification to ensure the accuracy of the converted pre-tightening force parameters. After converting the pre-tightening force parameters of all fastening bolts into Newton-meters, they are sorted into an ordered data set as the pre-tightening force parameter conversion result for use in subsequent steps.

[0107] S1022. Convert the material property parameters, geometric parameters, and assembly parameters included in the configuration data into configuration parameter conversion results with a unified dimension.

[0108] Among them, the material property parameters include elastic modulus and Poisson's ratio, the geometric parameters include bolt diameter and contact surface size, and the assembly parameters include installation angle and torque direction.

[0109] For the elastic modulus among the material property parameters, first determine its initial dimension, for example, megapascals. Since 1 megapascal is equal to 1,000,000 pascals, and the pascal is the basic unit of pressure in the International System of Units, and its relationship with newtons and meters is 1 pascal = 1 newton / square meter, it is necessary to convert the elastic modulus from megapascals to a unified dimension based on newtons and meters.

[0110] For geometric parameters, the fastening bolt diameter and contact surface dimensions can be given in different length units such as millimeters, centimeters, etc. They can be uniformly converted to meters.

[0111] For assembly parameters, the installation angle is generally in degrees and can be converted to radians because the radian system is more convenient in many calculations. The conversion formula is radian = angle × π / 180.

[0112] The torque direction is usually represented by a vector. By defining a positive direction and a negative direction, for example, its representation can be unified to ensure that all torque direction representations are consistent.

[0113] After completing the dimension conversion of each parameter, organize the material property parameters, geometric parameters, and assembly parameters into a complete data set of the configuration parameter conversion results.

[0114] S1023. Generate an initial target cutter barrel mesh model based on finite element mesh division according to the pre-tightening force parameter conversion results and the material property parameters and geometric parameters in the configuration parameter conversion results.

[0115] In this step, first import the pre-tightening force parameter conversion results and the material property parameters and geometric parameters in the configuration parameter conversion results into a professional finite element analysis software. The finite element analysis software can construct a geometric model of the target cutter barrel according to these parameters. The shape and size of the target cutter barrel are determined by the geometric parameters, and the mechanical properties of the material are determined by the material property parameters.

[0116] Then, the geometric model of the target cutter barrel can be subjected to finite element mesh division. In this process, the continuum of the target cutter barrel can be discretized into many small elements, and these elements can be of different shapes such as triangles, quadrilaterals, tetrahedrons, etc. When dividing the mesh, the structural characteristics of the target cutter barrel and the complexity of the stress distribution can be considered. For regions with large stress changes, such as near bolts and contact surfaces, finer meshes should be divided to improve the calculation accuracy; while for regions with small stress changes, relatively sparse meshes can be divided to reduce the calculation amount. During the mesh division process, the material properties of each unit can also be determined according to the material property parameters to ensure that the mechanical properties of each unit are consistent with the actual material. After the division is completed, an initial target cutter barrel mesh model based on finite element mesh division is obtained, and the initial target cutter barrel mesh model contains the geometric information, material information, and mesh division information of the target cutter barrel.

[0117] S1024. According to the pre-tightening force values corresponding to each fastening bolt in the pre-tightening force parameter conversion result, apply the corresponding pre-tightening force load to the initial target tool holder mesh model to generate the target tool holder mesh model after load application.

[0118] In this step, first, the pre-tightening force values corresponding to each fastening bolt can be extracted from the pre-tightening force parameter conversion result. Then, find the corresponding positions of each fastening bolt in the initial target tool holder mesh model. For example, positioning can be performed according to the installation position information of the fastening bolt and the coordinate system of the mesh model.

[0119] After determining the positions of the fastening bolts, apply the corresponding pre-tightening force values to the mesh model in a suitable manner. For the application of the pre-tightening force of the fastening bolts, an equivalent method can be used to convert the pre-tightening force into forces acting on the nodes around the fastening bolts. For example, the pre-tightening force can be evenly distributed to a certain number of nodes around the bolt so that the resultant force received by these nodes is equal to the pre-tightening force.

[0120] When applying the pre-tightening force load, the direction of the force should be consistent with the tightening direction of the fastening bolt, and the acting point should be accurately located in the contact area between the fastening bolt and the target tool holder.

[0121] Optionally, after applying the pre-tightening force loads of all fastening bolts, the mesh model can also be checked to ensure that the pre-tightening force of each fastening bolt is correctly applied, and the target tool holder mesh model after load application is generated. This target tool holder mesh model after load application reflects the initial state of the target tool holder under the action of the pre-tightening force.

[0122] S1025. According to the assembly parameters in the configuration parameter conversion result, set the contact surface constraint conditions and boundary conditions in the target tool holder mesh model after load application to generate the target tool holder mesh model after constraint loading.

[0123] In this step, the assembly parameters can be obtained from the configuration parameter conversion result first, including information such as the installation angle and torque direction. For the contact surface constraint conditions, the contact relationships between the target tool holder and the fastening bolts, and between different components of the target tool holder can be determined. For example, according to the installation angle and the actual assembly situation, it can be judged whether the contact surface is fully fitted, partially fitted, or has a gap, etc.

[0124] In finite element analysis, these contact relationships are simulated by setting contact pairs and contact types. For example, for a fully fitted contact surface, it can be set as a bonded contact so that the nodes on the contact surface do not have relative displacement during the calculation process; for a contact surface with a gap, it can be set as a frictional contact to consider the frictional force and relative sliding on the contact surface.

[0125] For the boundary conditions, they can be set according to the actual installation and usage of the target cutter barrel. If the target cutter barrel is fixedly installed on a certain device, then certain parts of the target cutter barrel can be set as fixed constraints, that is, the nodes of these parts will not displace during the calculation process. The torque direction information can be used to determine the rotation direction and constraint conditions of the target cutter barrel when it is under force. After completing the setting of the contact surface constraint conditions and boundary conditions, update the mesh model of the target cutter barrel after the load is applied to generate a mesh model of the target cutter barrel after constraint loading, and this mesh model of the target cutter barrel after constraint loading takes into account the actual assembly situation and force constraints of the target cutter barrel.

[0126] S1026. Calculate the stress distribution of the mesh model of the target cutter barrel after constraint loading, obtain the stress values of each mesh node, and generate a stress distribution data set.

[0127] In this step, it can start from the most basic mesh elements. For each mesh element, calculate the stress state inside the element according to its material properties and the applied loads. For example, by considering factors such as the geometric shape of the element and the displacements of the nodes, analyze the force balance and deformation coordination relationships within the element, and gradually determine the stresses at each point within the element.

[0128] After determining the stresses of each mesh element, transfer this stress information to the nodes corresponding to the connection points between the mesh elements. Through methods such as weighted averaging, summarize the stress information of adjacent elements to the nodes to obtain the stress values of each node. Then, through multiple iterations, continuously adjust the displacements and stresses of the nodes to make the calculation results gradually converge to a stable value. During each iteration, according to the results obtained from the previous iteration, recalculate the stresses of the elements and the stresses of the nodes until the convergence conditions are met.

[0129] Finally, collect the stress values of all mesh nodes and organize them according to the node numbers and coordinate positions to generate a stress distribution data set containing the stress information of each node.

[0130] S1027. Generate a three-dimensional stress distribution map in the first target cutter barrel stress simulation model according to the stress values of each mesh node and the corresponding coordinate positions in the stress distribution data set.

[0131] Among them, the three-dimensional stress distribution map contains the stress values of each pixel point area.

[0132] First, a three-dimensional space can be constructed to represent the shape and structure of the target cutter barrel. According to the coordinate positions of each mesh node in the stress distribution data set, place these nodes at the corresponding positions in the three-dimensional space to form the basic framework of the target cutter barrel, and the distribution and connection relationships of the nodes reflect the geometric shape of the target cutter barrel.

[0133] Next, in order to visually display the stress distribution, the stress value of each node can be associated with a visual representation method. For example, the method of color mapping can be used to map stress values of different magnitudes to different colors. For example, it can be set that a smaller stress value corresponds to blue, and as the stress value increases, the color gradually transitions to green, yellow, and finally to red for a larger stress.

[0134] After determining the color mapping rule, for each grid node, find the corresponding color according to its stress value. Then, starting from this node, perform color transition to its adjacent nodes. Since the actual stress distribution changes continuously, smooth color transition is required between nodes to more accurately reflect the stress change. Methods such as linear interpolation can be used to achieve this color transition, making the color change between adjacent nodes natural and smooth, thus obtaining the three-dimensional stress distribution map in the first target barrel stress simulation model.

[0135] In one embodiment, the aforementioned S1042 can generate a loosening detection stress reference model through the following method.

[0136] Obtain the loosening stress threshold data corresponding to the identification numbers of each fastening bolt in the stress threshold set. Among them, the loosening stress threshold data includes the stress thresholds of each pixel point in a preset area around each fastening bolt under different tightening torques.

[0137] According to the identification numbers of each fastening bolt in the stress threshold set, extract the coordinate range of the corresponding bolt installation area in the first target barrel stress simulation model to generate a bolt area division data set.

[0138] According to each coordinate range in the bolt area division data set, extract the stress thresholds of each pixel point in the corresponding area from the stress threshold set to generate a bolt area stress threshold data set.

[0139] Convert the stress thresholds of each pixel point in the bolt area stress threshold data set into the same dimension unit as the stress values in the first target barrel stress simulation model to generate a normalized stress threshold data set.

[0140] According to the stress thresholds of each pixel point in the normalized stress threshold data set and their coordinate positions, generate a loosening detection stress reference model corresponding to the first target barrel stress simulation model, where the loosening detection stress reference model contains the normalized stress thresholds of each pixel point area.

[0141] In this embodiment, the loosening stress threshold data corresponding to the identification numbers of each fastening bolt in the stress threshold set can be obtained first. The stress threshold set is obtained by training a deep learning model based on a large number of stress distribution sample data under different tightening torques. Conduct a comprehensive inspection of the database storing this set to ensure the integrity and accuracy of the data. Use indexing technology to query the corresponding data one by one according to the identification numbers of each fastening bolt. Check the numbers carefully during the query to avoid errors. After finding the data, compare it with the original sample data or check the statistical features to confirm its accuracy. The loosening stress threshold data covers the stress thresholds of each pixel point in the preset area around each fastening bolt under different tightening torques. Sort and organize the queried data according to the identification numbers, unify the data format and make a backup to provide an accurate and orderly data basis for subsequent operations.

[0142] Next, according to the identification numbers of each fastening bolt in the stress threshold set, extract the coordinate range of the corresponding bolt installation area in the first target tool cylinder stress simulation model to generate a bolt area division data set. Combine the model coordinate system and the designed installation position of the bolt, and automatically search for the bolt installation position according to the identification number. After determining the installation position, set the size and shape of the preset area around the bolt according to the actual situation, such as setting a radius range with the bolt center as the center of the circle. Use a coordinate judgment algorithm to traverse the model coordinate data, find all the coordinate points that fall within the preset area, and extract their coordinate ranges. Organize these coordinate ranges, record the corresponding bolt identification numbers, form a bolt area division data set, and verify its accuracy through a visualization method.

[0143] Then, according to each coordinate range in the bolt area division data set, extract the stress thresholds of each pixel point in the corresponding area from the stress threshold set to generate a bolt area stress threshold data set. Analyze the bolt area division data set in detail to clarify the bolt installation area corresponding to each coordinate range. Use a coordinate matching algorithm to compare the coordinate range with the pixel point coordinates in the stress threshold set, and considering the coordinate accuracy, accurately find all the pixel points that fall within this coordinate range. Extract the stress thresholds corresponding to these pixel points under different tightening torques and record them completely. Sort the data according to the coordinate range and the order of pixel points to form a bolt area stress threshold data set, and check the data quality through statistical analysis, and correct or eliminate outliers.

[0144] After that, convert the stress thresholds of each pixel point in the bolt area stress threshold dataset into the same dimension unit as the stress values in the first target barrel stress simulation model to generate a normalized stress threshold dataset. For example, the dimension unit of the stress values in the model can be determined by referring to the model documentation, and the dimension unit of the bolt area stress threshold dataset can be analyzed to determine the conversion relationship between the two. For each pixel point in the dataset, convert the stress threshold according to the conversion relationship. Organize the converted data, maintaining the original order and structure, and ensure that all stress thresholds are correctly converted through sampling inspection to form a normalized stress threshold dataset.

[0145] Finally, construct a three-dimensional space framework identical to the first target barrel stress simulation model, ensuring that the dimensions and coordinate systems are consistent. Accurately place the pixel points in the framework according to their coordinate positions, and use data mapping to assign the normalized stress thresholds to each point to ensure data integrity. Smooth the stress thresholds of adjacent pixel points, for example, use interpolation algorithms to achieve continuous and natural transitions. Visualize the stress thresholds through color mapping, with different stress magnitudes corresponding to different colors. Validate and evaluate the generated model by comparing it with the first target barrel stress simulation model, checking the correspondence of the coordinate system and pixel points and the rationality of the stress threshold distribution, and make timely corrections and adjustments to generate a loosening detection stress reference model containing the normalized stress thresholds of each pixel point area, providing an important basis for accurately judging the loosening of bolts.

[0146] In one embodiment, the loosening condition of each fastening bolt in the foregoing S1043 can be determined by the following method.

[0147] Divide the three-dimensional stress distribution map in the first target barrel stress simulation model into multiple pixel point areas. Among them, the size of each pixel point area is the same as that in the loosening detection stress reference model.

[0148] Extract the current stress values of each pixel point area from the first target barrel stress simulation model to generate a current stress value set; Extract the normalized stress thresholds corresponding to each pixel point area from the loosening detection stress reference model to generate a normalized stress threshold set; Compare the current stress values of each pixel point area in the current stress value set with the normalized stress thresholds of the corresponding areas in the normalized stress threshold set to generate the stress difference of each pixel point area; Count the number of pixel point areas whose stress differences in the bolt installation area corresponding to each fastening bolt exceed the preset loosening threshold to generate a loosening frequency statistical result; Determine the loosening grades of each fastening bolt according to the exceeding times of each bolt installation area in the loosening times statistical result and the preset loosening times threshold. Among them, the loosening grades include not loosened, slightly loosened, and severely loosened.

[0149] In this embodiment, first, the three-dimensional stress distribution diagram in the first target barrel stress simulation model can be divided into multiple pixel point regions, and the size of each pixel point region should be consistent with the pixel point region in the loosening detection stress reference model. This step is the basis for subsequent comparison and judgment. The three-dimensional stress distribution diagram of the first target barrel stress simulation model can be divided in the same way according to the division rule of the pixel point region in the loosening detection stress reference model. For example, if the loosening detection stress reference model divides the surface of the target barrel into several pixel point regions according to a certain grid size, then the same grid size is also used for division when processing the first target barrel stress simulation model. During the division process, it is necessary to ensure that the boundary of each pixel point region is clear and accurately aligned with the corresponding region in the loosening detection stress reference model to ensure the accuracy of subsequent data extraction and comparison.

[0150] Next, extract the current stress values of each pixel point region from the first target barrel stress simulation model to generate a set of current stress values. After the division is completed, for each pixel point region, search for and record the stress value of this region in the first target barrel stress simulation model. The corresponding stress value can be accurately extracted by traversing the stress data in the model according to the coordinate range of the pixel point region. For example, use the coordinate system and data query function to locate the position of each pixel point region and obtain its stress value. Collect the stress values of all pixel point regions and arrange them in a certain order (such as according to the number or coordinate order of the pixel points) to form a set of current stress values.

[0151] Then, extract the normalized stress thresholds corresponding to each pixel point region from the loosening detection stress reference model to generate a set of normalized stress thresholds. Since the normalized stress threshold has been determined for each pixel point region when generating the loosening detection stress reference model, the corresponding threshold can be extracted from the generated loosening detection stress reference model according to the previously divided pixel point regions. Similarly, use a method similar to that for extracting the current stress value to search for the corresponding normalized stress threshold in the loosening detection stress reference model according to the coordinate information of the pixel point region. Organize all the extracted normalized stress thresholds into a set of normalized stress thresholds.

[0152] After that, the current stress values of each pixel position area in the current stress value set are compared with the normalized stress thresholds of the corresponding areas in the normalized stress threshold set to generate the stress differences of each pixel position area. Specifically, for each pixel position area, its current stress value can be subtracted by the corresponding normalized stress threshold to obtain the stress difference. For example, by looping through the current stress value set and the normalized stress threshold set, the stress differences of each pixel position area can be calculated in sequence to generate the stress difference set of each pixel position area.

[0153] Next, count the number of pixel position areas in the bolt installation area corresponding to each fastening bolt whose stress differences exceed the preset loosening threshold to generate the loosening times statistical result. The preset loosening threshold is a standard value determined based on actual experience and experimental data, and is used to judge whether the stress change of the pixel position area is abnormal. For the installation area corresponding to each fastening bolt, check the stress differences of all pixel position areas in this area. If the stress difference of a certain pixel position area exceeds the preset loosening threshold, it is considered that a possible loosening situation has occurred in this area and is recorded as one loosening. Count the number of loosening times in each bolt installation area to form the loosening times statistical result.

[0154] Finally, determine the loosening level of each fastening bolt according to the exceeding times of each bolt installation area in the loosening times statistical result and the preset loosening times threshold. The preset loosening times threshold can be set according to the actual situation, and different thresholds correspond to different loosening levels.

[0155] For example, the loosening levels include not loosened, slightly loosened, and severely loosened. If the number of loosening times in the installation area of a certain fastening bolt is less than the lower limit of the preset loosening times threshold, it is considered that the fastening bolt is not loosened; if the number of loosening times is between the lower limit and the upper limit of the preset loosening times threshold, it is judged as slightly loosened; if the number of loosening times exceeds the upper limit of the preset loosening times threshold, it is determined as severely loosened. Record the loosening level of each fastening bolt for taking corresponding measures in time. For example, the fastening bolts that are not loosened can continue to be used normally, the slightly loosened fastening bolts can be inspected regularly, and the severely loosened fastening bolts need to be tightened immediately.

[0156] In the above embodiments, the target tool holder bolt detection system based on deep learning and data simulation for implementing the above method embodiments has at least one processor, a control module (chipset) coupled to at least one of the (at least one) processors, a memory coupled to the control module, a non-volatile memory (NVM) / storage device coupled to the control module, at least one input / output device coupled to the control module, and a network interface coupled to the control module.

[0157] The processor may include at least one single-core or multi-core processor, and the processor may include any combination of general-purpose processors or dedicated processors (such as graphics processors, application processors, baseband processors, etc.). For some alternative embodiments, the target barrel bolt detection system based on deep learning and data simulation can be used as the target barrel bolt detection system device such as the gateway described in the embodiments of the present application.

[0158] For some alternative embodiments, the target barrel bolt detection system based on deep learning and data simulation may include at least one computer-readable medium having instructions (e.g., a memory or an NVM / storage device) and at least one processor integrated with the at least one computer-readable medium and configured to execute the instructions to implement modules to perform the actions described in the present disclosure.

[0159] For one embodiment, the control module may include any suitable interface controller to provide any suitable interface to at least one of the (at least one) processors and / or any suitable device or component communicating with the control module.

[0160] The control module may include a memory controller module to provide an interface to the memory. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0161] The memory may be used, for example, to load and store data and / or instructions for the target barrel bolt detection system based on deep learning and data simulation. For one embodiment, the memory may include any suitable volatile memory, e.g., suitable DRAM.

[0162] For one embodiment, the control module may include at least one input / output controller to provide an interface to the NVM / storage device and the (at least one) input / output device.

[0163] For example, the NVM / storage device may be used to store data and / or instructions. The NVM / storage device may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (at least one) non-volatile storage device (e.g., at least one hard disk drive (HDD), at least one compact disc (CD) drive, and / or at least one digital versatile disc (DVD) drive).

[0164] The NVM / storage device may include storage resources that are physically part of the device on which the target barrel bolt detection system based on deep learning and data simulation is installed, or it may be accessible by the device without being part of the device. For example, the NVM / storage device may be accessed via the (at least one) input / output device according to a network.

[0165] (At least one) input / output device can provide an interface for the target tool holder bolt detection system based on deep learning and data simulation to communicate with any other suitable device. The input / output device can include a communication component, a pinyin component, a sensor component, etc. The network interface can provide an interface for the target tool holder bolt detection system based on deep learning and data simulation to communicate based on at least one network. The target tool holder bolt detection system based on deep learning and data simulation can wirelessly communicate with at least one component of the wireless network according to any standard and / or protocol in at least one wireless network standard and / or protocol, such as accessing a wireless network according to a communication standard.

[0166] For one embodiment, at least one of the (at least one) processors can be loaded with the logic of at least one controller of the control module (e.g., the memory controller module). For one embodiment, at least one of the (at least one) processors can be loaded with the logic of at least one controller of the control module to form a system-level load. For one embodiment, at least one of the (at least one) processors can be integrated with the logic of at least one controller of the control module on the same die. For one embodiment, at least one of the (at least one) processors can be integrated with the logic of at least one controller of the control module on the same die to form a system-on-chip (SoC).

[0167] The embodiments of the present application have been introduced in detail above. Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

[0168] An embodiment of the present invention discloses a computer-readable storage medium that stores a computer program for electronic data exchange, wherein the computer program causes the computer to execute the steps in the target tool holder bolt detection method based on deep learning and data simulation described in the foregoing embodiments.

[0169] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause the computer to execute the steps in the target tool holder bolt detection method based on deep learning and data simulation described in the foregoing embodiments.

[0170] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0171] Through the above specific descriptions of the embodiments, those skilled in the art can clearly understand that each implementation can be achieved by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium. The storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disk memories, tape memories, or any other medium that can be used for a computer to have or store data.

[0172] Finally, it should be noted that: the above-disclosed is only the preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or equivalently replace some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A target tool barrel bolt detection method based on deep learning and data simulation, characterized in that, Including: Obtain the pre-tightening force data of each fastening bolt on the target tool holder and the configuration data of the target tool holder; Generate a first target tool holder stress simulation model corresponding to the pre-tightening force data according to the pre-tightening force data and the configuration data; Input the pre-tightening force data, the configuration data, and the first target tool holder stress simulation model into a pre-trained neural network deep self-learning algorithm to obtain an algorithm verification result. The algorithm verification result includes a data verification result for indicating the verification result of the pre-tightening force data and the configuration data, and a simulation model verification result for indicating the verification result of the first target tool holder stress simulation model; If the algorithm verification result is verified to pass, determine the loosening condition of each fastening bolt according to the first target tool holder stress simulation model.

2. The method according to claim 1, wherein The step of inputting the pre-tightening force data, the configuration data, and the first target tool holder stress simulation model into a pre-trained neural network deep self-learning algorithm to obtain an algorithm verification result includes: Input the pre-tightening force data, the configuration data, and the first target tool holder stress simulation model into the pre-trained neural network deep self-learning algorithm. The pre-trained neural network deep self-learning algorithm calculates the pre-tightening force data and the configuration data based on the pre-learned model parameters to obtain predicted pre-tightening force data and predicted configuration data. The model parameters are trained based on sample data including sample pre-tightening force data, sample configuration data, and sample simulation models; Match the predicted pre-tightening force data with the pre-tightening force data and match the predicted configuration data with the configuration data to obtain the data verification result; If the target data with a certainty less than or equal to a preset certainty threshold is included in the predicted pre-tightening force data and / or the predicted configuration data, re-simulate according to the target data to obtain a second target tool holder stress simulation model. The certainty is determined by the confidence or variance of the predicted pre-tightening force data and / or the predicted configuration data and the corresponding confidence threshold or variance threshold; Match the first target tool holder stress simulation model and the second target tool holder stress simulation model to obtain the simulation verification result; Obtain the algorithm verification result according to the simulation verification result and the data verification result.

3. The method according to claim 2, wherein The step of matching the predicted pre-tightening force data with the pre-tightening force data and matching the predicted configuration data with the configuration data to obtain the data verification result includes: Obtain the first relative error and the first absolute error corresponding to the predicted pre-tightening force data and the pre-tightening force data; Determine the first data verification result according to the first relative error, the first absolute error, the first preset relative error threshold, and the first preset absolute error threshold; Obtain the second relative error and the second absolute error corresponding to the predicted configuration data and the configuration data; Determine the second data verification result according to the second relative error, the second absolute error, the second preset relative error threshold, and the second preset absolute error threshold; Obtain the data verification result based on the first data verification result and the second data verification result.

4. The method according to claim 2, wherein The matching of the first target barrel stress simulation model and the second target barrel stress simulation model to obtain the simulation verification result includes: Extract the first stress distribution feature in the first target barrel stress simulation model and the second stress distribution feature in the second target barrel stress simulation model. The stress distribution feature includes maximum stress, average stress, and stress distribution; Calculate the difference indexes between the first stress distribution feature and the second stress distribution feature. The difference indexes include maximum stress difference, average stress difference, and stress distribution difference; Determine the verification result of the maximum stress difference according to the maximum stress difference threshold, determine the verification result of the average stress difference according to the average stress difference threshold, and determine the verification result of the stress distribution difference according to the overlap degree threshold; Obtain the simulation verification result based on the verification result of the maximum stress difference, the verification result of the average stress difference, and the verification result of the stress distribution difference.

5. The method according to claim 1, characterized in that, The determination of the loosening condition of each fastening bolt according to the first target barrel stress simulation model includes: Obtain a set of stress thresholds corresponding to the loosening condition of each fastening bolt. The set of stress thresholds is pre-trained by a deep learning model based on stress distribution sample data under different tightening torques in a stress distribution database. The set of stress thresholds is used to indicate the loosening stress thresholds of each point near the corresponding fastening bolt; Generate a loosening detection stress reference model according to the set of stress thresholds corresponding to the loosening condition of each fastening bolt; Determine the loosening condition of each fastening bolt according to the change of each pixel point area in the first target barrel stress simulation model and the stress threshold corresponding to the corresponding pixel point area in the loosening detection stress reference model.

6. The method according to claim 1, characterized in that The generation of the first target barrel stress simulation model corresponding to the pre-tightening force data according to the pre-tightening force data and the configuration data includes: Convert the initial pre-tightening force parameters corresponding to each fastening bolt in the pre-tightening force data into a pre-tightening force parameter conversion result with a unified dimension, where the unified dimension is Newton-meter; Convert the material property parameters, geometric parameters, and assembly parameters included in the configuration data into configuration parameter conversion results with a unified dimension. The material property parameters include elastic modulus and Poisson's ratio, the geometric parameters include bolt diameter and contact surface size, and the assembly parameters include installation angle and torque direction; Generate an initial target barrel grid model based on finite element mesh division according to the material property parameters and geometric parameters in the pre-tightening force parameter conversion result and the configuration parameter conversion result; Apply the corresponding pre-tightening force load to the initial target barrel grid model according to the pre-tightening force values corresponding to each fastening bolt in the pre-tightening force parameter conversion result to generate a target barrel grid model after load application. According to the assembly parameters in the converted result of the configuration parameters, set the contact surface constraint conditions and boundary conditions in the target cutter barrel mesh model after the load is applied, and generate the target cutter barrel mesh model after constraint loading; Perform stress distribution calculation on the target cutter barrel mesh model after constraint loading, obtain the stress values of each mesh node, and generate a stress distribution data set; Generate a three-dimensional stress distribution map in the first target cutter barrel stress simulation model according to the stress values of each mesh node in the stress distribution data set and their corresponding coordinate positions, where the three-dimensional stress distribution map includes the stress values of each pixel point area.

7. The method according to claim 5, wherein The generation of the loosening detection stress reference model according to the stress threshold set corresponding to the loosening conditions of each fastening bolt includes: Obtain the loosening stress threshold data corresponding to the identification numbers of each fastening bolt in the stress threshold set, where the loosening stress threshold data includes the stress thresholds of each pixel point in a preset area around each fastening bolt under different tightening torques; According to the identification numbers of each fastening bolt in the stress threshold set, extract the coordinate range of the corresponding bolt installation area in the first target cutter barrel stress simulation model, and generate a bolt area division data set; According to each coordinate range in the bolt area division data set, extract the stress thresholds of each pixel point in the corresponding area from the stress threshold set, and generate a bolt area stress threshold data set; Convert the stress thresholds of each pixel point in the bolt area stress threshold data set into the same dimension unit as the stress values in the first target cutter barrel stress simulation model, and generate a normalized stress threshold data set; Generate the loosening detection stress reference model corresponding to the first target cutter barrel stress simulation model according to the stress thresholds of each pixel point area in the normalized stress threshold data set and their coordinate positions, where the loosening detection stress reference model includes the normalized stress thresholds of each pixel point area.

8. The method according to claim 5, characterized in that, Determine the loosening conditions of each fastening bolt according to the changes in each pixel point area in the first target cutter barrel stress simulation model and the stress thresholds corresponding to the corresponding pixel point areas in the loosening detection stress reference model, including: Divide the three-dimensional stress distribution map in the first target cutter barrel stress simulation model into multiple pixel point areas, and the size of each pixel point area is the same as that of the pixel point area in the loosening detection stress reference model; Extract the current stress values of each pixel point area from the first target cutter barrel stress simulation model, and generate a current stress value set; Extract the normalized stress thresholds corresponding to each pixel point area from the loosening detection stress reference model, and generate a normalized stress threshold set; Compare the current stress values of each pixel point area in the current stress value set with the normalized stress thresholds of the corresponding areas in the normalized stress threshold set, and generate the stress differences of each pixel point area. Count the number of stress differences exceeding the preset loosening threshold in all pixel point regions within the bolt installation area corresponding to each of the fastening bolts, and generate a loosening times statistical result; Determine the loosening level of each of the fastening bolts according to the exceeding times of each of the bolt installation areas in the loosening times statistical result and the preset loosening times threshold, where the loosening level includes not loosened, slightly loosened, and severely loosened.

9. A target cutter barrel bolt detection system based on deep learning and data simulation, characterized in that, It includes a processor and a computer-readable storage medium, and machine-executable instructions are stored in the computer-readable storage medium. When the machine-executable instructions are executed by a computer, the target tool holder bolt detection method based on deep learning and data simulation described in any one of claims 1-8 is implemented.

10. A computer program product, characterized in that, It includes a computer program. When the computer program is executed by a processor, the target tool holder bolt detection method based on deep learning and data simulation described in any one of claims 1-8 is implemented.

Images