An AI-based bolt fastening device and its method

Abstract

The present invention discloses a bolt tightening device and method based on artificial intelligence, which solves the problem that traditional bolt tightening methods are difficult to ensure the tightening to the most ideal state and have low tightening accuracy. The key points of its technical solution include a tightening wrench, a tightening driving device, and a tightening monitoring device. By preparing tightening data, an artificial intelligence-based tightening monitoring model is established, and the monitoring model is trained and verified through the tightening data. After the training is completed, the tightening monitoring model is deployed to the tightening monitoring device for prediction and judgment. Through the real-time monitoring of the tightening monitoring device, the start and stop actions of the tightening driving device are controlled. The bolt tightening device and method based on artificial intelligence of the present invention can monitor the tightening project status in real time, and can perform prediction and judgment through the artificial intelligence monitoring model to accurately stop the tightening work, which can effectively improve the accuracy of bolt tightening preload control and is more reliable.

Description

Technical Field

[0001] The present invention relates to bolt fastening technology, and particularly to an artificial intelligence-based bolt fastening device and method. Background Art

[0002] Large bolts are widely used in wind power, nuclear industry, ships and other large mechanical equipment, which is a prerequisite for the normal operation of the equipment. A large number of bolts are used in large equipment, and there are high requirements for the uniformity of bolt fastening. Traditional fastening methods cannot ensure the uniformity of bolt fastening or cannot exert the maximum fastening efficiency of the bolts. In addition, for some bolts in harsh service environments, higher requirements are placed on the reliability of fastening. Traditional methods are difficult to ensure the fastening to the most ideal state and cannot directly predict the fastening pre-tightening force. Summary of the Invention

[0003] The purpose of the present invention is to provide an artificial intelligence-based bolt fastening device and method, which can monitor the state of the fastening project in real time, and can make predictive judgments through the monitoring model of artificial intelligence to accurately stop the fastening work, effectively improve the control accuracy of the pre-tightening force of bolt fastening, and make the fastening more reliable.

[0004] The above technical object of the present invention is achieved through the following technical solutions:

[0005] An artificial intelligence-based bolt fastening method includes the following steps:

[0006] Data preparation: Based on actual engineering requirements, perform fastening processing on bolts of different specifications, and collect and obtain actual fastening data;

[0007] Model construction: Design a monitoring model suitable for the characteristics of bolt fastening projects, preprocess the collected actual fastening data, perform time-frequency transformation on the collected sequence data and standardize it to adapt to the corresponding model, and train and verify the monitoring model with the processed fastening data;

[0008] Model application: Encode and deploy the trained monitoring model on the bolt fastening monitoring device, and communicatively connect the fastening torque measurement module, the torsional angle measurement module and the human-computer interaction module;

[0009] Fastening work: Set fastening parameters through the human-computer interaction module, start the fastening drive device, connect the bolt fastening wrench, the fastening monitoring device receives the fastening data fed back by the monitoring in real time, determines in real time whether the fastening process meets the fastening requirements through the monitoring model, and feeds back to control the fastening drive device to stop when the requirements are met, completing the fastening work.

[0010] An artificial intelligence-based bolt fastening device includes: a fastening wrench, a fastening drive device, and a fastening monitoring device;

[0011] The fastening monitoring device includes a torque measurement module, a torsional angle measurement module, a human-machine interaction module, and a data communication module, and further includes a monitoring model embedded in the fastening monitoring device;

[0012] The fastening wrench is controlled by the fastening driving device, and the fastening driving device is coupled to and responds to the fastening monitoring device;

[0013] The human-machine interaction module sets the fastening parameters for the fastening project. The torque measurement module measures and obtains the fastening torque, and the torsional angle measurement module measures and obtains the fastening torsional angle. The monitoring model is connected to the torque measurement module and the torsional angle measurement module through the data communication module, and outputs a predicted fastening stop instruction when it is determined that the predetermined fastening requirement is met.

[0014] In summary, the present invention has the following beneficial effects:

[0015] By training the constructed monitoring model with the actually collected fastening data, the prediction and judgment of the monitoring model can be more practical and accurate; through the setting of the fastening monitoring device, relevant data during the fastening process can be obtained in real time, and through the monitoring model, prediction and judgment can be carried out in real time. Through the communication connection between the fastening monitoring device and the fastening driving device, real-time and precise control of the fastening action of the fastening wrench can be achieved, the state during the fastening process can be monitored in real time, and when the fastening requirement is met, the fastening work can be stopped in a timely and accurate manner, effectively improving the bolt fastening accuracy and the reliability of the fastening work. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a structural schematic block diagram of a bolt fastening device;

[0017] Figure 2 is a flow schematic diagram of a bolt fastening method;

[0018] Figure 3 is a framework diagram of a monitoring model based on a convolutional neural network;

[0019] Figure 4 is a framework diagram of a monitoring model based on a recurrent neural network;

[0020] Figure 5 is a framework diagram of a monitoring model based on traditional machine learning. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The present invention will be further described in detail below with reference to the accompanying drawings.

[0022] According to one or more embodiments, an artificial intelligence-based bolt fastening device is disclosed, as Figure 1 shown, including a fastening wrench, a fastening driving device, and a fastening monitoring device.

[0023] The fastening wrench is connected to the fastening driving device through a fastening sleeve, is controlled by the fastening driving device, and performs a fastening action after being controlled by the fastening driving device.

[0024] The fastening driving device is coupled to and responsive to the fastening monitoring device, and receives instructions or signals output by the fastening monitoring device for driving control. The fastening driving device is hydraulically driven or pneumatically driven, and is connected to the fastening wrench through a hydraulic / pneumatic pipeline to provide a driving force to the fastening wrench.

[0025] The fastening monitoring device includes a torque measurement module, a torsional angle measurement module, a human-machine interaction module, and a data communication module, and also includes a monitoring model embedded in the fastening monitoring device.

[0026] The torque measurement module measures the torque in real time during the bolt fastening process to obtain the fastening torque. The torsional angle measurement module measures the torsional angle in real time during the bolt fastening process to obtain the fastening torsional angle. The human-machine interaction module sets the fastening parameters of the fastening project. According to the torque and torsional angle data measured during the fastening process, the fastening friction coefficient can be calculated in real time to accurately calculate the actual pre-tightening force of the bolt.

[0027] The human-machine interaction module includes a touch screen and a remote controller; the remote controller is a handle remote control scanner, which includes control buttons, a display screen, and a scanning component for engineering records. The scanning operation is performed through the control buttons, the fastening bolt information (bolt position, number, etc.) is scanned and recorded through the scanning component, and is displayed through the display screen. The human-machine interaction module sets the fastening parameters of the fastening project, which can be manually input through the touch screen, and the relevant parameter data of the fastening wrench can be obtained by scanning through the scanning component.

[0028] The monitoring model is constructed based on any one of a convolutional neural network, a recurrent neural network, and a traditional machine learning algorithm. After being trained and verified, it is used for fastening prediction.

[0029] For example, the monitoring model is constructed based on a deep convolutional neural network. The deep convolutional neural network includes a first-level input layer, a second-level input layer, a hidden layer, and an output layer. The first-level input layer is the multi-modal data in the time domain and frequency domain of the torque and torsional angle, the second-level input layer is the pre-determined friction coefficient of the screw pair, and the hidden layer is a series of standard layers, and each standard layer includes convolution, activation function, batch-normalization, and pooling steps.

[0030] The monitoring model is connected to the torque measurement module and the torsional angle measurement module through the data communication module, responds to the fastening data and makes a judgment, and outputs a predicted fastening stop instruction when it is determined that the predetermined fastening requirement is met. The data communication module can communicate and connect between various components through cables or wirelessly.

[0031] Connect the fastening drive device to the fastening wrench, and connect the fastening monitoring device to the fastening drive device and the fastening wrench through the data communication module. The fastening monitoring device starts the fastening drive device according to the fastening parameters input by the touch screen of the human-computer interaction module and the relevant parameters obtained by scanning through the scanning component. During the fastening process, the torque measurement module and the torsional angle measurement module perform real-time measurements and feedback the measured fastening data to the monitoring model. The monitoring module judges the fastening process according to the preset fastening requirements and outputs a predicted fastening stop instruction when the predetermined requirements are met. The fastening monitoring device outputs the corresponding instruction to the fastening drive device, thereby stopping the fastening action and completing the fastening work.

[0032] According to one or more embodiments, a bolt fastening method based on artificial intelligence is disclosed, as Figure 2 shown, including the following steps:

[0033] S1. Data preparation: Based on actual engineering requirements, perform fastening processing on bolts of different specifications and collect actual fastening data.

[0034] Specific data preparation includes the following steps:

[0035] S11. Use the fastening drive device and the fastening wrench to fasten bolts of different specifications and stop when the predetermined torque is reached.

[0036] S12. Fasten until the bolt fails in a manual control manner, and collect the actual fastening data during the fastening process, including the friction coefficient of the screw pair, the fastening torque, the fastening twist angle, and the bolt pre-tightening force returned by the force sensor. Use the bolt pre-tightening force returned by the force sensor as the gold standard for model training.

[0037] S13. Mark the optimal stop moment of each bolt fastening process and obtain relevant data for model training and verification. Manually mark the ideal stop moment of each bolt fastening process for use in fastening model training.

[0038] Under laboratory conditions, use the fastening drive device and the corresponding wrench, combined with the torque (pressure) measurement device, the torsional angle measurement device, and the pre-tightening force measurement device, to perform fastening operations on bolts of different specifications, different strength grades, different flange materials, and different lubrication conditions, and complete the recording of complete fastening data.

[0039] S2. Model construction: Design a monitoring model suitable for the engineering characteristics of bolt fastening, preprocess the collected fastening data, perform time-frequency transformation on the collected sequence data and standardize it to adapt to the corresponding model, and train and verify the monitoring model with the processed fastening data.

[0040] The training and simulation testing of AI model construction are specifically as follows:

[0041] S21. Preprocess the actually collected fastening data. Perform time-frequency transformation on the data and standardize the corresponding model. Preprocess the data according to the obtained data characteristics and extract features. For example, based on the periodic characteristics of the fastening process of the hydraulic torque wrench, perform time-frequency transformation on the pressure data to capture the characteristics of different fastening stages. At the same time, record the materials of the flange and bolts, and give the friction force level in combination with the actual lubrication condition.

[0042] S22. Train the constructed monitoring model with the actually collected fastening data and continuously verify it until the model converges.

[0043] S23. For the trained monitoring model, perform simulated input based on the fastening torque and fastening twist angle in the actually collected fastening data, and test the accuracy of the output result of the monitoring model.

[0044] S3. Model deployment and application: Recode and deploy the trained monitoring model based on the characteristics of embedded hardware to the bolt fastening monitoring device, and communicate and connect the fastening torque measurement module, torsion angle measurement module and human-computer interaction module.

[0045] S4. Fastening work: Set the fastening parameters through the human-computer interaction module, start the fastening drive device, connect the bolt torque wrench, the fastening monitoring device receives the fastening data of the monitoring feedback in real time, determines in real time whether the fastening process meets the fastening requirements through the monitoring model, and feeds back to control the fastening drive device to stop when the requirements are met, and completes the fastening work.

[0046] The fastening data received by the fastening monitoring device in real time includes: the fastening torque measured by the fastening torque measurement module, the fastening twist angle measured by the torsion angle measurement module, and the fastening parameters set by the human-computer interaction module. After the fastening monitoring device continues to segment and preprocess the received data, it makes a prediction and judgment through the monitoring model. The monitoring model calculates the fastening friction coefficient based on the torque and twist angle data during the fastening process to facilitate the accurate calculation of the actual pre-tightening force of the bolt.

[0047] The designed monitoring model can be established based on convolutional neural network, recurrent neural network, gradient boosting machine, etc.

[0048] Example 1. Construct a deep learning model based on convolutional neural network:

[0049] Such as Figure 3The constructed model has two data input ports. After the pressure data undergoes time-frequency transformation (such as wavelet transformation), it is converted into two-dimensional image data and input into the model through input port 1. For the image model, feature extraction is performed through multiple convolution-activation-pooling operations. The torsional angle data is input through data input port 2, and multiple one-dimensional convolution operations are performed on it. Then, the above two feature tensors are merged in the middle. Subsequently, convolution-activation-pooling operations continue, and following the idea of multi-task learning, the fastening friction coefficient is output, and the fastening pre-tightening force is output, and it is predicted whether the fastening process should end. This model analyzes whether the fastening is completed based on the feature differences of the pressure data at different stages of fastening, fuses the changes in the fastening pressure (derived torque) and the torsional angle to sense the friction parameters of the system, calculates the pre-tightening force, and finally outputs the pre-tightening force of the fastening process.

[0050] For the designed model, training and validation are carried out based on the fastening data set obtained in the first step. During the training process, the prediction of the pre-tightening force is trained for N epochs, and then the classification of the friction level is trained for 1 epoch, and so on alternately. Finally, the model converges to a performance that can meet the actual engineering applications.

[0051] For the trained model, based on the pressure and torsional angle data that can be obtained during the actual fastening process, simulated input is performed to verify the accuracy of the model output results. According to the characteristics of the hydraulic fastening device and the model prediction results, the corresponding control device hardware is designed, and control instructions are sent to the fastening drive device based on the model results to control the start and termination of the fastening of the equipment.

[0052] Example 2: Constructing a model based on a recurrent neural network:

[0053] For example, a long short-term memory network (LSTM), and the model design block diagram is as Figure 4 shown. In this example, a fastening drive device and a fastening wrench based on hydraulic power are designed based on a recurrent neural network, and a new intelligent fastening control system is developed. Due to the excellent time series data prediction ability of the recurrent neural network, a method combining the recurrent neural network with traditional empirical formulas can be adopted. The subsequent pressure curve and angle curve are predicted through the recurrent neural network to analyze the friction coefficient of the bolt pair. Both the subsequent pre-tightening force can be calculated using traditional formulas, and the yield limit of the bolt can be analyzed through the trend of the curve to ensure the safety performance of the bolt.

[0054] The construction and training of the model are specifically as follows:

[0055] Based on the basic LSTM architecture and combined with the idea of seq2seq, a separate pressure data prediction branch and an angle data prediction branch are designed, and the hidden variables of the two parts are merged in the middle layer for analyzing the friction level of the bolt. The output of this model is the prediction curves of the friction level, pressure, and angle data. Based on the above data and combined with the empirical formula of the bolt engineering, the bolt pre-tightening force and its yield limit are calculated, and it is judged whether the shutdown condition is reached. The number of LSTM units at the input end of the recurrent neural network is (m - l), and the number of LSTM units at the output end is (n - m). In the actual use process, the curve information of the latest (m - l) time intervals will be selected as the input.

[0056] For the designed model, training and validation are carried out based on the fastening data set obtained under laboratory conditions. During the data processing, the known pressure-time curve and angle-time curve are segmented at a specified time interval and input in batches according to the given recurrent neural network specifications for training. During the training process, independent prediction of the pressure-time curve and angle-time curve is carried out for N epochs of training, and then 1 epoch of training is carried out for the classification of the friction level, and so on alternately. Finally, the model converges to a performance that can meet the actual engineering applications. Through the calculation of the empirical formula, the pre-tightening force data of the subsequent model and the yield limit of the bolt are finally obtained.

[0057] Example 3: Construct a model based on traditional machine learning algorithms:

[0058] For example, models such as random forest and XGBoost, and the design block diagram is as Figure 5 shown. Since traditional machine learning algorithms have good accuracy and high computational efficiency on small sample data, this example intends to adopt a similar idea to Example 2, that is, using a machine learning model to analyze the friction coefficient of the bolt, predict the subsequent trends of the pressure-time curve and angle-time curve, and the yield strength of the bolt, and calculate the pre-tightening force of the bolt through the empirical formula.

[0059] The calculation of the friction level and the prediction of the pressure-time curve and angle-time curve are carried out independently. The friction level is calculated based on the random forest algorithm. By inputting the pressure and angle sensor information into the model through one channel, the friction level is finally classified. Based on the XGBoost algorithm to predict the pressure-time curve and angle-time curve, two independent XGBoost models need to be trained, which are respectively used for the prediction of the two curves.

[0060] According to one or more embodiments, an artificial intelligence-based bolt fastening system is disclosed, including a server, the server having a memory; and

[0061] A processor coupled to a memory, the processor being configured to execute instructions stored in the memory, and the processor performs the following operations:

[0062] Data preparation: Based on actual engineering requirements, perform fastening processing on bolts of different specifications, and collect and obtain actual fastening data;

[0063] Model construction: Design a monitoring model suitable for the engineering characteristics of bolt fastening, preprocess the collected fastening data, perform time-frequency transformation on the collected sequence data and standardize it to adapt to the corresponding model, and train and verify the monitoring model with the processed fastening data;

[0064] Model application: Encode and deploy the trained monitoring model on the bolt fastening monitoring device, and communicatively connect the fastening torque measurement module, the torsional angle measurement module, and the human-computer interaction module;

[0065] Fastening work: Set fastening parameters through the human-computer interaction module, start the fastening drive device, connect the bolt fastening wrench, the fastening monitoring device receives the fastening data monitored and fed back in real time, determines in real time whether the fastening process meets the fastening requirements through the monitoring model, and feeds back to control the fastening drive device to stop when the requirements are met, thus completing the fastening work.

[0066] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0067] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0068] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed couplings, direct couplings, or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices, or units, and can also be electrical, mechanical, or other forms of connection.

[0069] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiments of the present invention.

[0070] In addition, each functional unit in various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0071] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0072] This specific embodiment is only an interpretation of the present invention and does not limit the present invention. Those skilled in the art can make modifications to this embodiment without creative contributions according to needs after reading this specification, but as long as they are within the scope of the claims of the present invention, they are protected by the patent law.

Claims

1. An artificial intelligence-based bolt tightening method, characterized in that It includes the following steps: Data preparation: Based on the actual engineering requirements, fasten bolts of different specifications, and collect actual fastening data; Model construction: Design a monitoring model suitable for the engineering characteristics of bolt fastening, preprocess the collected actual fastening data, perform time-frequency transformation on the collected sequence data and standardize it to adapt to the corresponding model, and train and verify the monitoring model with the processed fastening data; Model application: Encode and deploy the trained monitoring model on the bolt fastening monitoring device, and communicate and connect the fastening torque measurement module, torsion angle measurement module, and human-computer interaction module; Fastening work: Set fastening parameters through the human-computer interaction module, start the fastening drive device, connect the bolt fastening wrench, the fastening monitoring device receives the fastening data fed back by monitoring in real time, determines in real time whether the fastening process meets the fastening requirements through the monitoring model, and feeds back to control the fastening drive device to stop when the requirements are met, completing the fastening work.

2. The bolt tightening method based on artificial intelligence according to claim 1, characterized in that Specific data preparation includes the following steps: Use the fastening drive device and the fastening wrench to fasten bolts of different specifications, stop when the fastening reaches the predetermined torque, and collect the fastening data; Subsequently, fasten until the bolt fails in a manual control manner, and collect the actual fastening data. The fastening data includes the friction coefficient of the screw pair, fastening torque, fastening torsion angle, and the bolt pre-tightening force returned by the force sensor; Mark the best stop moment of each bolt fastening process, and obtain relevant data for model training and verification.

3. The bolt tightening method based on artificial intelligence according to claim 2, characterized in that, The fastening data received by the fastening monitoring device in real time includes: the fastening torque measured by the fastening torque measurement module, the fastening torsion angle measured by the torsion angle measurement module, and the fastening parameters set by the human-computer interaction module.

4. The bolt tightening method based on artificial intelligence according to claim 2, characterized in that The training and testing of model construction are specifically as follows: Preprocess the actually collected fastening data; Train the constructed monitoring model with the actually collected fastening data, and continuously verify it until the model converges; For the trained monitoring model, based on the fastening torque and fastening torsion angle in the actual fastening data, perform simulated input to verify the accuracy of the output result of the monitoring model.

5. An artificial intelligence-based bolt tightening device, characterized in that, Based on the artificial intelligence-based bolt fastening method described in any one of claims 1 to 4, it includes: a fastening wrench, a fastening drive device, and a fastening monitoring device; The fastening monitoring device includes a torque measurement module, a torsion angle measurement module, a human-computer interaction module, and a data communication module, and also includes a monitoring model embedded in the fastening monitoring device; The fastening wrench is controlled by the fastening drive device, and the fastening drive device is coupled to and responds to the fastening monitoring device; The human-computer interaction module sets the fastening parameters of the fastening project, the torque measurement module measures and obtains the fastening torque, and the torsion angle measurement module measures and obtains the fastening torsion angle; the monitoring model is connected to the torque measurement module and the torsion angle measurement module through the data communication module, and outputs a predicted fastening stop instruction when it is determined that the predetermined fastening requirements are met.

6. The bolt tightening device based on artificial intelligence according to claim 5, characterized in that: The human-computer interaction module includes a touch screen and a remote controller; the remote controller includes control buttons, a display screen, and a scanning component.

7. The bolt tightening device based on artificial intelligence according to claim 5, characterized in that: The fastening drive device is hydraulically driven or pneumatically driven.

8. An artificial intelligence-based bolt tightening system, characterized in that: The system includes a server having a memory; and a processor coupled to the memory, the processor being configured to execute instructions stored in the memory, and the processor performs the following operations: Data preparation: Based on actual engineering requirements, perform fastening processing on bolts of different specifications, and collect actual fastening data; Model construction: Design a monitoring model suitable for the engineering characteristics of bolt fastening, preprocess the collected actual fastening data, perform time-frequency transformation on the collected sequence data and standardize it to fit the corresponding model, and train and verify the monitoring model with the processed fastening data; Model application: Encode and deploy the trained monitoring model on the bolt fastening monitoring device, and communicatively connect the fastening torque measurement module, the torsional angle measurement module, and the human-machine interaction module; Fastening work: Set fastening parameters through the human-machine interaction module, start the fastening drive device, connect the bolt fastening wrench, the fastening monitoring device receives the fastening data fed back by the monitoring in real time, determines in real time whether the fastening process meets the fastening requirements through the monitoring model, and feeds back to control the fastening drive device to stop when the requirements are met, completing the fastening work.

9. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, the method described in any one of claims 1 to 4 is implemented.

Images