Torsion detection device and detection method for winding formed pipe

Abstract

The invention discloses a torsion detection device and detection method for a winding forming pipe, and belongs to the technical field of profile performance detection. The device comprises a fixer, a torque wrench, a box cage, a machine head, a torque tester, a visual sensor and a control system. According to the invention, through precise installation of the fixer, uniform torque transmission is realized, cooperative detection of the torque wrench, the torque tester and the visual sensor is combined, the problems of large installation deviation, high measurement error and the like in traditional detection are effectively solved, and the detection precision and reliability are significantly improved; in addition, by integrating a visual sensor and a control system, displacement data of the spiral pipe can be collected and analyzed in real time, synchronous data analysis is conducted through the control system, efficient and accurate detection of the torque performance of the profile is achieved, and technological parameters are automatically optimized so as to achieve the optimal production quality of the spiral pipe.

Description

Technical Field

[0001] This invention belongs to the field of profile performance testing technology, primarily applied to profile production quality control, and relates to the field of torque detection within material mechanical property testing technology. Specifically, it provides a torque detection device and method for winding tubing. By integrating mechanical fixation, precise measurement, and intelligent analysis, this device and method achieve standardized testing of profile torsional performance, providing reliable technical support for quality control and process optimization of profile products. Background Art

[0002] In pipeline construction, for ease of transportation, winding pipe technology is used. Profiles are produced on-site using a winding machine (such as the winding equipment described in Publication No. CN212004760U) to create pipes. During on-site construction, torque testing of the pipes is a critical step in ensuring product quality. Traditional torque testing methods rely primarily on manual operation of simple torque wrenches, which present numerous technical drawbacks. First, the lack of standardized fixtures makes it easy for profiles to shift or slip during testing, resulting in distorted measurement data. Second, traditional methods cannot guarantee the stability of the torque applied during each test, resulting in poor repeatability of measurement results. Furthermore, these testing methods lack a real-time data acquisition system, relying entirely on the operator's subjective judgment. This makes it difficult to generate objective and traceable quality records. This extensive testing method no longer meets the high standards of product quality control required by modern manufacturing. Furthermore, this type of profile performance testing is destructive, posing certain safety risks, and improper manual operation can lead to accidental injuries.

[0003] The torque detection devices in the prior art generally have deficiencies in their structural design. The fixing devices mostly use a simple clamping method, which cannot ensure precise concentric positioning with the profile, resulting in uneven torque transmission and systematic deviations in the measurement data. During the test process, the fixing stability of the profile is insufficient, and micro-displacement is easily generated when torque is loaded, which seriously affects the measurement accuracy. In addition, conventional detection devices lack an effective limiting mechanism, which can neither guarantee the standard test length of the profile nor control the force application trajectory of the torque wrench. These structural defects cause large fluctuations in the measurement results, and the test data of different operators or different batches lack comparability, which brings difficulties to product quality assessment. Although some automated testing equipment is already on the market, its high cost and complicated operating procedures limit its popular application on the production line.

[0004] With the advancement of intelligent manufacturing technology, the industry has put forward higher requirements for torque detection. On the one hand, it is necessary to standardize and automate the testing process to reduce interference from human factors; on the other hand, the detection system is required to have data acquisition and analysis functions to provide data support for process optimization. Most of the existing detection devices lack intelligent monitoring modules and cannot record the deformation characteristics and mechanical response of the profile under the action of torsion in real time. Especially in the research and development of new composite material profiles, traditional detection methods are difficult to accurately capture the nonlinear deformation characteristics of the material. Therefore, the development of a new torque detection device for winding forming pipes that integrates mechanical fixation, precise measurement and intelligent analysis has become a key technical breakthrough direction for improving the quality control level of profile products. Summary of the Invention

[0005] (1) Purpose of the invention In view of the problems in the existing technology of inaccurate torque data collection during the construction testing phase, and the fact that skilled workers still need to rely on experience to observe the degree of pipe deformation with the naked eye to determine whether the destructive torque has been reached and adjust the process parameters accordingly, without specific implementation standards, large manual errors, and certain dangers in such destructive material mechanical property tests, the present invention aims to provide a device for intelligently detecting the maximum tolerable torque of wound forming pipes, which is used to solve the problems of low measurement accuracy, incomplete data collection, insufficient operation standardization, and safety risks in the existing profile torque detection.

[0006] (2) Technical solution In order to achieve the purpose of the invention and solve the technical problems, the present invention adopts the following technical solutions: In one aspect, the present invention provides a torque detection device for winding a tubular product, comprising: The die head has a cage installed on one side. The cage is used to wind the profile fed by the die head into a spiral tube. A torque tester is also installed on the die head. A retainer, mounted on the spiral tube; A torque wrench, one end of which is mounted in cooperation with the fixed end of the holder, and a torque sensor is provided to sense the torque of the torque wrench; A visual sensor is installed on one side of the spiral tube and is used to collect image data of the spiral tube in real time; The control system consists of an on-site control device and a cloud server. The on-site control device is connected to the visual sensor, torque sensor, and torque tester. The control system is equipped with an image recognition module, a deformation discrimination module, and a process parameter adjustment module, wherein: The image data collected by the visual sensor is calculated and analyzed through the image recognition module and the deformation discrimination module to obtain the cumulative strain value, full-field deformation acceleration, and local strain gradient of the sample area. A three-level state discrimination mechanism is used to determine whether the spiral tube has undergone irreversible plastic deformation. The real-time data of the torque tester and torque wrench obtained when plastic deformation occurs is used as the basis for judging whether the spiral tube meets the standards. The process parameters are improved and optimized according to the deep learning model of the process parameter adjustment module.

[0007] On the other hand, the present invention also provides a method for detecting torque of a wound tube, which uses the aforementioned torque detection device for a wound tube, and the method comprises the following steps: S1. Test Preparation: The profile is continuously fed into the die head and placed into a matching cage, where it is wound into a spiral tube of a fixed length. S2. Torque wrench installation: Drill a hole in the spiral tube and install the holder, ensuring the holder and spiral tube are concentric. Attach the torque wrench to the holder and install the torque wrench with the limiter. S3. Torque Test: The visual sensor is activated to capture images in real time. The die begins applying a uniformly accelerated feed pressure. The real-time image data collected by the visual sensor is continuously uploaded to the cloud server. The image recognition module and deformation discrimination module calculate and analyze the cumulative strain value of the sample area, the global deformation acceleration, and the local strain gradient, which are fed back to the on-site control device. The on-site control device uses a three-level state discrimination mechanism to determine whether the spiral tube has undergone irreversible plastic deformation. If plastic deformation is determined to have occurred, the die immediately stops applying force and slowly releases the force. At the same time, real-time data from the torque tester and torque wrench are intercepted as the breaking torque. S4. Parameter adjustment: The on-site control device determines whether the strength of the spiral tube meets the standard based on the project requirements. If not, the process parameters and breaking torque of the spiral tube can be uploaded to the cloud server. The cloud server uses the deep learning model of the process parameter adjustment module to provide improvement suggestions for the process parameters. Based on the improvement suggestions, the process parameters are improved and optimized, and a new spiral tube is remade. The above steps are repeated until the strength of the spiral tube meets the project requirements.

[0008] (3) Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: (1) The torque wrench of the present invention is connected to the spiral tube through a fixture, so that the torque applied by the torque wrench can be evenly transmitted to the spiral tube, avoiding local stress concentration, thereby improving the accuracy of the measurement data. Through the coordinated work of the torque wrench and the torque tester, dual detection and comparison of torque data are achieved, effectively reducing the system error of a single measuring device.

[0009] (2) The present invention ensures that the fixture and the spiral tube are connected concentrically with high precision by drilling holes in the cross direction, combined with the fastening structure of the screw and nut, effectively avoiding torque measurement errors caused by installation deviations. The setting of the limit plate can accurately limit the feeding length of the spiral tube, ensure that the initial conditions of each test are consistent, reduce human operation errors, and improve the reliability of the test. The limiting effect of the bracket on the torque wrench can prevent it from deflecting during the test process, ensure the stability of the torque application direction, and thus improve the measurement accuracy.

[0010] (3) The present invention can collect and analyze the displacement data of the spiral tube in real time by integrating the visual sensor and the control system, and perform data synchronous analysis through the control system, thereby realizing automatic monitoring of the detection process, reducing human intervention, and improving detection efficiency and objectivity. By setting an image recognition module, a deformation discrimination module, and a process parameter adjustment module in the control system, the image data collected by the visual sensor is transmitted to the control system, and the image recognition module and the deformation discrimination module are used to calculate and analyze the cumulative strain value, full-field deformation acceleration, and local strain gradient of the sample area. The three-level state discrimination mechanism can accurately capture the tiny displacement changes of the spiral tube under the action of torsion, and the deep learning model of the process parameter adjustment module can realize automatic adjustment of the process parameters. The system automatically completes data statistics, analysis, and recording, significantly reducing errors introduced by human operation, realizing efficient and accurate detection of the profile torque performance, and automatically optimizing the process parameters to achieve the best production quality of the spiral tube. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a front view of a torque detection device for a winding tube according to an embodiment of the present invention; Figure 2 A side view of a torque detection device for a winding tube according to an embodiment of the present invention; Figure 3 is a structural diagram of a fixer according to an embodiment of the present invention; Figure 4 Schematic diagram of the installation of the bracket and torque sensor according to an embodiment of the present invention; Figure 5 Flowchart of a method for detecting torque of a wound tube according to an embodiment of the present invention; Explanation of the accompanying reference numerals: 1- spiral tube; 2- torque wrench; 3- holder; 4- cage; 5- machine head; 6- torque tester; 7- limit plate; 8- visual sensor; 9- bracket; 10- screw rod; 11- nut; 12- torque sensor; 13- control system. DETAILED DESCRIPTION

[0012] To make the purposes, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention.

[0013] Example 1: Torque detection device like Figures 1 to 4 In the embodiment of the torque detection device for winding formed pipes of the present invention shown, the device mainly includes a torque wrench 2, a holder 3, a cage 4, a head 5, and a torque tester 6. The holder 3 is mounted on the spiral tube 1, and the torque wrench 2 is connected to the spiral tube 1 through the holder 3. The cage 4 is mounted on one side of the head 5, and the torque tester 6 is provided on the head 5.

[0014] The torque wrench 2 is connected to the spiral tube 1 through the holder 3, so that the torque applied by the torque wrench 2 can be evenly transmitted to the spiral tube 1, avoiding local stress concentration, thereby improving the accuracy of the measurement data. The cage 4 is used to wrap the fed profile into the spiral tube 1. The limiting effect of the cage 4 on the spiral tube 1 also enhances the stability of the detection process and prevents the spiral tube 1 from shifting during the torque loading process. The main part of the holder 3 is a disc, with four screws 10 and several nuts 11 symmetrically arranged around the disc. The center of the disc is provided with a hexagonal head. The screw 10 of the holder 3 is connected to the outer end of the spiral tube 1 through a cross-shaped drilling. The holder 3 uses a fastening structure of screws 10 and nuts 11 to achieve a reliable connection with the spiral tube 1. The holder 3 is preferably set concentrically with the spiral tube 1 with high precision, effectively avoiding torque measurement errors caused by installation deviations. The fixed end of the holder 3 is installed in conjunction with the torque wrench 2, so that the torque applied by the torque wrench 2 can be evenly transmitted to the spiral tube 1, ensuring the accuracy of the measurement data. The connection end of the torque wrench 2 is preferably provided with a hexagonal sleeve, which matches the hexagonal head at the center of the disc of the holder 3, and the two are installed by plugging the hexagonal head and the hexagonal sleeve.

[0015] A limit plate 7 is installed on one side of the spiral tube 1 to limit the feeding length of the spiral tube 1. This ensures that the initial position of the spiral tube 1 is consistent during each test, reducing human error and improving test reliability. The cage 4 includes a guide rail frame, forming rollers, and guide rollers. The profile is folded and spirally wound on the forming rollers within the cage 4 to form a tubular shape. The male and female locking buckles on both sides of the profile are self-locking, while the guide grooves move the tubular spiral tube 1 forward. The cage 4 structure also enhances the limiting effect on the spiral tube 1, preventing it from shifting during torsional loading and ensuring accurate measurement results.

[0016] In addition, the detection device also includes a bracket 9, which is arranged on one side of the spiral tube 1 to limit the torque wrench 2. The bracket 9 can effectively prevent the torque wrench 2 from shifting during the detection process, ensuring the stability of the torque application direction, thereby improving measurement accuracy. The bracket 9 is preferably provided with a torque sensor 12 for collecting the torque of the torque wrench 2 and synchronously transmitting the information to the control system. In other embodiments, the torque sensor can also be provided at the connection end that is mounted to the fixture.

[0017] Furthermore, the detection device is preferably provided with a visual sensor 8, which is mounted on one side of the spiral tube 1 via a bracket, near the detection area, and is used to collect image data of the spiral tube 1 in real time. The visual sensor 8 can accurately capture the tiny displacement changes of the spiral tube 1 under the action of torsion, and the data is transmitted to the control system via a high-speed data interface. Specifically, the visual sensor mentioned in this embodiment is composed of a high-resolution industrial camera and a polarization filter lens, which can reduce the metal reflection interference of the profile; a shock-absorbing pan-tilt head is also installed between the industrial camera and the bracket to reduce disturbances caused by factors such as on-site construction and human operation; an external light source is also provided on the bracket to ensure stable brightness in the sampling area; the image data acquisition rate is greater than or equal to 30fps to ensure sufficient sampling.

[0018] In one embodiment, a control system 13 is further included. The control system 13 is composed of an on-site control device and a cloud server. The on-site control device is connected to the visual sensor, torque sensor, and torque tester to achieve real-time collection, storage, and analysis of detection data. The on-site control device is connected to the machine head 5 to control the feeding force. In addition, the control system is equipped with an image recognition module, a deformation discrimination module, and a process parameter adjustment module, wherein: The image recognition module transmits the image data collected by the visual sensor 8 to the image recognition module, performs dimensionality reduction processing through pooling operation in the image recognition module, and uploads the reduced-dimensionality data to the cloud server in real time. The cloud server performs polarization noise reduction processing on the reduced-dimensionality data through the polarization light response layer to obtain reduced-dimensionality data. The reduced-dimensionality data is convolved to extract speckle features to obtain a speckle feature set, thereby determining the tracking mark. The displacement field quantity is calculated based on the pixel displacement and edge brightness change of the tracking mark. The displacement field quantity is mapped into 10 consecutive frames of strain field data with a resolution of 128×128 per frame according to the time-space relationship, and is further analyzed by the deformation discrimination module.

[0019] The image recognition module determines the displacement field by tracking changes in speckle brightness, achieving high accuracy under stable light sources. To improve algorithm response speed, the algorithm uses a layered deployment: a local basic preprocessing layer and a cloud server-based core algorithm layer. Raw data undergoes local pooling preprocessing and is transmitted to the cloud server. The cloud server's high-performance computing power rapidly computes the displacement field information into 10 consecutive frames of strain field data, each with a resolution of 128×128 pixels. This data is then sent to the deformation discrimination module for further analysis. The image recognition module also includes a polarization response layer to enhance texture recognition in reflective metal areas and further eliminate the effects of metal reflections.

[0020] The deformation discrimination module, including a 3D convolutional layer, an LSTM neural network layer, and a fully connected layer, inputs the 128×128×10 strain field data into the 3D convolutional layer and simultaneously discriminates the temporal and spatial variation characteristics of the displacement field to obtain a spatiotemporal feature sequence. After receiving the spatiotemporal feature sequence extracted by the 3D convolutional layer, the LSTM neural network layer analyzes the dynamic evolution path of the strain field based on it. After filtering through a forget gate and state updating, the feature vector of the evolution information is input into the fully connected layer. The fully connected layer, as the output layer, maps the feature vector of the evolution information to three deformation indicators through weighted summation and nonlinear transformation. The three deformation indicators include the cumulative strain value of the sample area, the full-field deformation acceleration, and the local strain gradient: The cumulative strain value of the sample area directly reflects the degree of plastic deformation in that area. The global deformation acceleration is used to predict the trend (acceleration) of the deformation rate of the entire strain field during the observation period. This value is used to determine whether the deformation is accelerating and is an important early warning signal of instability risk. The local strain gradient is used to predict the spatial rate of change (gradient) of strain near a specific location or area. High strain gradients often indicate stress concentration and are dangerous areas for crack initiation or propagation.

[0021] The cloud server packages the three deformation indicators and the corresponding time and space coordinates as a data packet and feeds it back to the on-site control device, which enters the deformation discrimination mechanism of the on-site control device and controls the growth rate of the feed thrust of the head 5 according to the feedback data; when the regional cumulative strain value is lower than 50% of the material limit strain, it indicates that the structure is in the elastic deformation stage and has not reached the material limit. This stage is a low-risk stage and the growth rate of the preset feed thrust of the head 5 can be maintained unchanged; if the local strain gradient undergoes a mutation of more than 200% (such as a sudden increase from 0.01 / mm to 0.03 / mm), it is determined to be a stress concentration phenomenon, indicating that potential damage has begun. It is necessary to focus on monitoring the expansion trend of microcracks in this area. This stage is a medium-risk stage, and the growth rate of the preset feed thrust of the head 5 should be gradually slowed down; when the deformation acceleration continues to exceed 0.1% / ms (that is, the strain rate increases by 0.001 per millisecond) or microcracks are observed, this state represents the initiation of necking or macro fracture, and marks that the material has entered an irreversible instability and failure stage. This stage is a high-risk stage, and the head 5 force should be stopped urgently and slowly unloaded to prevent further cracking of the sample spiral tube and cause safety accidents; and the real-time data of the torque tester 6 and the torque wrench 2 should be saved as the destructive torque, and the feed thrust destructive value of the head 5 should be recorded.

[0022] In the process parameter adjustment module, the on-site control device will preliminarily judge whether the strength of the batch of spiral tubes meets the engineering requirements based on the breaking torque. If it does not meet the standards, the data such as the feed rate, feed pressure, hot melt adhesive output rate and breaking torque of the sample spiral tubes can be uploaded to the cloud server. The cloud server will give suggestions for improving the process parameters through the deep learning model of the process parameter adjustment module. After fine-tuning, new sample spiral tubes will be remade and tested repeatedly until the strength of the spiral tubes meets the engineering requirements.

[0023] Specifically, pre-trained samples for the overall control system, encompassing all three modules, were collected in the materials laboratory. The model can also be further optimized based on real-time data uploaded to the cloud server during past construction. Using visual sensors, torque sensors, and torque testers, the system collects key data, such as the displacement and force of the spiral pipe, in real time. The system automatically performs data statistics, analysis, and recording, significantly reducing errors introduced by human operators. This enables efficient and accurate testing of profile torque performance and automatically optimizes process parameters to achieve optimal spiral pipe production quality.

[0024] Example 2: Torque detection method On the basis of the above embodiment 1, embodiment 2 of the present invention further proposes a method for detecting torque of a winding tube. Based on the above torque detection device for winding tube of the present invention, as shown in FIG. Figure 5 As shown, the method mainly includes the following steps when implemented: S1. Test preparation: The profile is continuously fed into the die head 5 and sent into the matching box cage 4 to be wound into a spiral tube 1 of fixed length. To ensure the consistency of the winding length, a limit plate 7 is set to limit the feeding length; S2. Torque wrench installation: Drill holes in the cross direction at the outer end of the spiral tube 1 to install the fixture 3, and securely connect it with the screw 10 and nut 11. During installation, strictly calibrate the length of the screw 10 and the distance between the nut 11 and the screw 10 to ensure that the fixture 3 and the spiral tube 1 remain concentric. Insert the hexagonal socket of the torque wrench 2 into the hexagonal head at the center of the disc of the fixture 3 to complete the installation. Install the torque wrench 2 and the bracket 9 in a fixed position. S3. Torque test: Start the visual sensor 8 to collect image data in real time. After a period of image collection to confirm the identification and tracking of the sample area, upload the pre-processed data to the cloud server. After the cloud server confirms that the real-time transmission is normal, start the test. The head 5 begins to apply a uniformly accelerated feed pressure, and the on-site control device continuously uploads the real-time image data collected by the visual sensor 8 to the cloud server. The cloud server calculates and analyzes the force and deformation of the sample area through the image recognition module and the deformation discrimination module, and feeds back the three deformation indicators to the on-site control device in real time. The on-site control device uses a three-level state discrimination mechanism to determine whether the spiral tube has undergone irreversible plastic deformation. When it is determined that plastic deformation has occurred, the head 5 is immediately stopped from applying force and slowly unloading the force. At the same time, the real-time data of the torque tester 6 and the torque wrench 2 are intercepted as the destructive torque, and the feed thrust of the head 5 is recorded as the destructive thrust value; S4. Parameter adjustment: The on-site control device determines whether the strength of the batch of spiral tubes meets the standards based on engineering requirements. If not, the data such as the feed rate, feed pressure, hot melt adhesive discharge rate and breaking torque of the sample spiral tubes can be uploaded to the cloud server. The cloud server uses the deep learning model of the process parameter adjustment module to provide suggestions for improving the process parameters. After adjustment based on the improvement suggestions, new sample spiral tubes are remade and the above detection steps are repeated until the strength of the spiral tubes meets the engineering requirements.

[0025] Preferably, step S3 specifically includes the following sub-steps when performing the torque test: SS31. Image data collected by the visual sensor is transmitted to the image recognition module, where it is subjected to dimensionality reduction processing through pooling operations. The reduced-dimensionality data is then uploaded to the cloud server in real time. The cloud server then passes the reduced-dimensionality data through a polarization-responsive layer to perform polarization noise reduction to obtain reduced-dimensional data. The reduced-dimensional data is then convolved to extract speckle features, resulting in a speckle feature set, which is then used to identify the tracking marker. The displacement field is calculated based on the pixel displacement and edge brightness changes of the tracking marker. The displacement field is then plotted as 10 consecutive frames of strain field data, each with a resolution of 128×128 pixels, based on the time-space relationship. SS32. Input strain field data into the deformation discrimination module. The convolutional layer in the deformation discrimination module discriminates the temporal and spatial variations of the displacement field, generating a spatiotemporal feature sequence. The LSTM neural network layer receives the spatiotemporal feature sequence and analyzes the dynamic evolution path of the strain field. After filtering through a forget gate and state update, the feature vector of the evolution information is input into the fully connected layer. The fully connected layer, acting as the output layer, uses weighted summation and nonlinear transformation to map the high-dimensional features onto three deformation metrics: the cumulative strain value of the sample area, the global deformation acceleration, and the local strain gradient. SS33. The cloud server packages the three deformation indicators and the corresponding time and space coordinates as a data packet and feeds it back to the on-site control device. The data enters the deformation discrimination mechanism of the on-site control device and controls the growth rate of the head feed thrust according to the feedback data. When the regional cumulative strain value is lower than 50% of the material limit strain, the growth rate of the head feed thrust is maintained unchanged. If the local strain gradient suddenly changes, the growth rate of the head feed thrust is gradually slowed down. When the deformation acceleration continues to exceed 0.1% / ms or microcracks are observed, the head force is stopped urgently and the force is slowly unloaded. The real-time data of the torque tester and torque wrench are saved as the breaking torque, and the head feed thrust is recorded as the breaking thrust value.

[0026] The torque detection technology for wound-wrapped tubing of the present invention utilizes an integrated torque testing device and automated monitoring system to construct a high-precision profile performance detection system. Through the precise installation of the fixture 3, the coordinated detection of the torque wrench 2 and torque tester 6, the real-time monitoring of the visual sensor 8, and the intelligent analysis of the control system, accurate measurement of the profile's torsional performance is achieved. The coordinated coordination of various components, from the precise positioning of the spiral tube 1 and the uniform application of torque to the real-time data collection and analysis, effectively resolves the problems of large installation deviations, high measurement errors, and incomplete data collection that exist in traditional torque testing, providing reliable technical support for the quality control of profile products.

[0027] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that a person skilled in the art can conceive within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A torque detection device for winding forming pipes, characterized in that: include: The die head has a cage installed on one side, which is used to wind the profile fed into the die head into a spiral tube, and a torque tester is set on the die head; A retainer, mounted on the spiral tube; A torque wrench is mounted in cooperation with the central fixed end of the holder and is provided with a torque sensor for detecting the torque of the torque wrench; A visual sensor is installed on one side of the spiral tube and is used to collect image data of the spiral tube in real time; The control system consists of a field control device and a cloud server. The field control device is connected to the visual sensor, the torque sensor, and the torque tester, and wherein: The control system is equipped with an image recognition module, a deformation discrimination module, and a process parameter adjustment module. The image data collected by the visual sensor is calculated and analyzed by the image recognition module and the deformation discrimination module to obtain the cumulative strain value, full-field deformation acceleration, and local strain gradient of the sample area. A three-level state discrimination mechanism is used to determine whether the spiral tube has undergone irreversible plastic deformation. The real-time data from the torque tester and torque wrench obtained when plastic deformation occurs is used as the basis for judging whether the spiral tube meets the standards. The process parameters are improved and optimized based on the deep learning model of the process parameter adjustment module.

2. The torque detection device for winding tubes according to claim 1, characterized in that: The main body of the fixer is a disc, with four screws and several nuts symmetrically arranged around the disc, a hexagonal head arranged in the center of the disc, and the screws of the fixer are connected to the outer end of the spiral tube through cross-direction drilling holes.

3. The torque detection device for winding tubes according to claim 1, characterized in that: The vision sensor consists of a high-resolution industrial camera and a polarization filter lens.

4. The torque detection device for winding tubes according to claim 1, characterized in that: The image recognition module transmits the image data collected by the visual sensor to the image recognition module, performs dimensionality reduction processing through pooling operation in the image recognition module, and uploads the dimensionality reduction data to the cloud server in real time; The cloud server performs polarization denoising on the dimensionality-reduced data through a polarized light response layer to obtain denoised data. The denoised data is then convolved to extract speckle features, resulting in a speckle feature set to identify the tracking marker. The displacement field is calculated based on the pixel displacement and edge brightness changes of the tracking marker. The displacement field is then plotted into 10 consecutive frames of strain field data, each with a resolution of 128×128 pixels, based on the time-space relationship.

5. The torque detection device for winding tubes according to claim 4, characterized in that: The deformation discrimination module includes a 3D convolutional layer, an LSTM neural network layer, and a fully connected layer. The strain field data is input into the 3D convolutional layer to simultaneously discriminate the temporal and spatial variation characteristics of the displacement field quantity, thereby obtaining a spatiotemporal feature sequence. After receiving the spatiotemporal feature sequence extracted from the 3D convolutional layer, the LSTM neural network layer analyzes the dynamic evolution path of the strain field, filters the evolution information through a forget gate, and updates the state before inputting the feature vector into the fully connected layer. The fully connected layer serves as the output layer, and the feature vector of the evolution information is mapped to three deformation indicators: the cumulative strain value of the sample area, the full-field deformation acceleration, and the local strain gradient through weighted summation and nonlinear transformation.

6. The torque detection device for winding tubes according to claim 5, characterized in that: The three-level state discrimination mechanism of the deformation discrimination module is as follows: The cloud server packages the three deformation indicators and the corresponding time and space coordinates as a data packet and feeds it back to the on-site control device. The data enters the deformation discrimination mechanism of the on-site control device and controls the growth rate of the head feed thrust according to the feedback data. When the regional cumulative strain value is lower than 50% of the material limit strain, the growth rate of the head feed thrust is maintained unchanged. If the local strain gradient suddenly changes, the growth rate of the head feed thrust is gradually slowed down. When the deformation acceleration continues to exceed 0.1% / ms or microcracks are observed, the head force is stopped urgently and the force is slowly unloaded, and the real-time data of the torque tester and torque wrench are saved as the destructive torque.

7. The torque detection device for winding tubes according to claim 1, characterized in that: A limit plate is provided on one side of the spiral tube for limiting the feeding length of the spiral tube.

8. The torque detection device for winding tubes according to claim 1, characterized in that: It also includes a bracket, which is arranged on one side of the spiral tube and is used to limit the torque wrench.

9. A method for detecting torque of a wound tube, using the torque detection device for a wound tube according to any one of claims 1 to 8, characterized in that: The following steps are involved: S1. Test Preparation: The profile is continuously fed into the die head and placed into a matching cage, where it is wound into a spiral tube of a fixed length. S2. Torque wrench installation: Drill a hole in the spiral tube and install the holder, ensuring the holder and spiral tube are concentric. Attach the torque wrench to the holder and install the torque wrench with the limiter. S3. Torque Test: The visual sensor is activated to capture images in real time. The die begins applying a uniformly accelerated feed pressure. The real-time image data collected by the visual sensor is continuously uploaded to the cloud server. The image recognition module and deformation discrimination module calculate and analyze the cumulative strain value of the sample area, the global deformation acceleration, and the local strain gradient, which are fed back to the on-site control device. The on-site control device uses a three-level state discrimination mechanism to determine whether the spiral tube has undergone irreversible plastic deformation. If plastic deformation is determined to have occurred, the die immediately stops applying force and slowly releases the force. At the same time, real-time data from the torque tester and torque wrench are intercepted as the breaking torque. S4. Parameter adjustment; The on-site control device determines whether the spiral tube strength meets the standard based on the project requirements. If not, the process parameters and breaking torque of the spiral tube are uploaded to the cloud server. The cloud server uses the deep learning model of the process parameter adjustment module to provide improvement suggestions for the process parameters. Based on the improvement suggestions, the process parameters are improved and optimized, and a new spiral tube is remade. The above steps are repeated until the strength of the spiral tube meets the project requirements.

10. The method for detecting torque of a winding tube according to claim 9, wherein: Step S3 includes the following sub-steps when implemented: SS31. The image data collected by the visual sensor is passed to the image recognition module, where it is processed through pooling operations to reduce the dimension of the image data. The reduced data is then uploaded to the cloud server in real time. The cloud server applies polarization noise reduction to the dimensionality-reduced data through a polarization-response layer to obtain noise-reduced data. The noise-reduced data is then convolved to extract speckle features, resulting in a speckle feature set, which is then used to identify the tracking marker. The displacement field is calculated based on the pixel displacement and edge brightness changes of the tracking marker. The displacement field is then mapped into strain field data based on the time-space relationship. SS32. Input the strain field data into the deformation discrimination module. The convolutional layer in the deformation discrimination module determines the temporal and spatial variation characteristics of the displacement field, generating a spatiotemporal feature sequence. The LSTM neural network layer receives the spatiotemporal feature sequence and analyzes the dynamic evolution path of the strain field. After filtering through a forget gate and state updating, the feature vector of the evolution information is input into the fully connected layer. The fully connected layer, as the output layer, maps the high-dimensional features to three deformation indicators: the cumulative strain value of the sample area, the full-field deformation acceleration, and the local strain gradient through weighted summation and nonlinear transformation of the feature vector of the evolution information; SS33. The cloud server packages the three deformation indicators and the corresponding time and space coordinates as a data packet and feeds it back to the on-site control device. The data enters the deformation discrimination mechanism of the on-site control device and controls the growth rate of the head feed thrust according to the feedback data. When the regional cumulative strain value is lower than 50% of the material limit strain, the growth rate of the head feed thrust is maintained unchanged. If the local strain gradient suddenly changes, the growth rate of the head feed thrust is gradually slowed down. When the deformation acceleration continues to exceed 0.1% / ms or microcracks are observed, the head force is stopped urgently and the force is slowly unloaded, and the real-time data of the torque tester and torque wrench are saved as the destructive torque.

Images