A dynamic torque compensation method and application for high-precision tightening operation

Abstract

The application discloses a dynamic torque compensation method and application for high-precision tightening operation, and relates to the technical field of automatic control, which comprises the following steps: a multi-modal sensing module array is assembled to collect multi-modal physical signals in a tightening operation process; a cross-modal spatio-temporal feature of the multi-modal physical signals is extracted by using a space-time convolution network, the cross-modal spatio-temporal feature is subjected to feature fusion, and a fusion feature vector is generated; a dynamic torque compensation value is calculated based on a hierarchical near-end strategy optimization algorithm according to the fusion feature vector; and the tightening operation is compensated and adjusted according to the dynamic torque compensation value. The application can solve the problems of the prior compensation method, such as lack of multi-modal information fusion capability, poor real-time performance of edge computing, and poor control precision caused by dependence on a fixed threshold control algorithm.

Description

Technical Field

[0001] This application relates to the field of automation control technology, and in particular to a dynamic torque compensation method, apparatus, electronic device, and computer-readable storage medium for high-precision tightening operations. Background Technology

[0002] Traditional industrial tightening equipment relies on fixed threshold control (such as PID algorithms), which is difficult to cope with dynamic disturbances such as material deformation and thread wear, leading to the risk of over-tightening or under-tightening. Existing AI compensation methods (such as CN117698442A) are mostly based on single sensor data, lack multimodal information fusion capabilities, and have not solved the real-time problem of edge computing. Patent CN117195996A discloses a torque prediction method based on deep neural networks, but its application is limited to vehicle steer-by-wire and does not cover the micromechanical characteristics of industrial tightening. Patent CN119017317A discloses a tightening system that improves safety through fastener length detection, but lacks adaptive adjustment of process parameters. Summary of the Invention

[0003] To overcome the shortcomings of the prior art, this invention provides a dynamic torque compensation method and application for high-precision tightening operations, aiming to combine multimodal sensing, reinforcement learning and edge computing dynamic compensation schemes to achieve sub-microsecond response and cross-scene adaptation.

[0004] On one hand, embodiments of the present invention propose a dynamic torque compensation method for high-precision tightening operations, comprising: constructing a multimodal sensing module array to acquire multimodal physical signals during the tightening operation; using a spatiotemporal convolutional network to extract cross-modal spatiotemporal features of the multimodal physical signals, performing feature fusion on the cross-modal spatiotemporal features to generate a fused feature vector; calculating a dynamic torque compensation value based on the fused feature vector using a hierarchical near-end strategy optimization algorithm; and adjusting the tightening operation according to the dynamic torque compensation value.

[0005] In one embodiment of the present invention, the step of using a spatiotemporal convolutional network to extract cross-modal spatiotemporal features of the multimodal physical signals, and performing feature fusion on the cross-modal spatiotemporal features to generate a fused feature vector includes: independently inputting vibration signals, thermal imaging data, and strain data using a spatiotemporal separable convolution method; preprocessing the three signals independently to obtain spatiotemporal feature maps of each modality; generating trainable weight matrices for each modality from the spatiotemporal feature maps through linear layers; calculating a similarity matrix between modalities based on the trainable weight matrices; calculating attention interactions between modalities based on the similarity matrix; and outputting and concatenating the attention interactions to generate the fused feature vector through a fully connected layer.

[0006] In one embodiment of the present invention, the calculation formula for the attention interaction is expressed as follows: ;in, Represents the query vector. Represents the key vector. Represents a value vector. The feature dimension of the key vector is represented; the spatiotemporal association weights of the query vector of the vibration mode and the key vector of the thermal imaging mode are calculated, and the value vectors are weighted and fused to obtain the attention interaction between the vibration mode and the thermal imaging mode.

[0007] In one embodiment of the present invention, the spatiotemporal convolutional network includes a temporal convolutional kernel and a spatial / spectral convolutional kernel. The temporal convolutional kernel slides along the time axis to extract dynamic evolution features; the spatial / spectral convolutional kernel processes the spatial distribution of thermal imaging or the energy concentration region of vibration spectrum.

[0008] In one embodiment of the present invention, the dynamic torque compensation method for high-precision tightening operations further includes: introducing a dynamic convolution kernel generator to adjust the convolution kernel parameters in real time according to the characteristics of the input data, wherein the expression of the dynamic convolution kernel generator is: ;in, It is a lightweight LSTM network, and the input is the current sensor state and historical context.

[0009] In one embodiment of the present invention, the expression for calculating the dynamic torque compensation value based on the hierarchical proximal strategy optimization algorithm according to the fused feature vector is as follows: ; in, For temperature parameters, For vibration term parameters, For parameters of cross-coupling terms; Among the temperature parameters , This represents the actual measured temperature, where, Indicates the reference temperature. Indicates the temperature-torque coupling coefficient. This represents the sensitivity coefficient to temperature change rate. Among the vibration term parameters ,in, This represents the correction factor for the deformation strain ε on the vibration-temperature coupling. Based on the vibration-temperature gain coefficient, For spectral energy weighting factor, The nonlinear attenuation coefficient is... It is the vibration acceleration; ,in, The deformation-torque conversion coefficient, This is a dynamic strain rate correction factor; Among the cross-coupling term parameters, ,in, For temperature gradient and mechanical stress coefficient, The dynamic rate of change coupling factor, second derivative This refers to strain acceleration.

[0010] In one embodiment of the present invention, the dynamic torque compensation method for high-precision tightening operations further includes: For metallic materials exhibiting a clear yielding phenomenon, the formula for calculating the yield strength is as follows: ; Where Re represents the yield strength, and Fe represents the constant force maintained during the yielding stage. Indicates the original cross-sectional area of ​​the sample; For metallic materials without a clear yield phenomenon, the specified non-proportional elongation strength is used as the yield strength, expressed as: ; in, This represents the force required for a material to undergo 0.2% plastic deformation.

[0011] On the other hand, this invention also proposes a dynamic torque compensation device for high-precision tightening operations, comprising: a multimodal signal acquisition module for constructing a multimodal sensor module array to acquire multimodal physical signals during the tightening operation; a cross-modal feature fusion module for extracting cross-modal spatiotemporal features of the multimodal physical signals using a spatiotemporal convolutional network, performing feature fusion on the cross-modal spatiotemporal features, and generating a fused feature vector; a dynamic torque compensation value calculation module for calculating a dynamic torque compensation value based on the fused feature vector using a hierarchical near-end strategy optimization algorithm; and a compensation adjustment module for adjusting the tightening operation according to the dynamic torque compensation value.

[0012] In another aspect, embodiments of the present invention also provide an electronic device, comprising: a memory and one or more processors connected to the memory, the memory storing a computer program, and the processors executing the computer program to implement the dynamic torque compensation method for high-precision tightening operations as described in any of the above embodiments.

[0013] In another aspect, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions for performing the dynamic torque compensation method for high-precision tightening operations as described in any of the above embodiments.

[0014] As can be seen from the above, the embodiments of the present invention, compared with the prior art, can have at least one or more of the following beneficial effects: The proposed dynamic torque compensation method for high-precision tightening operations uses a multimodal sensing module to collect multimodal physical signals during the tightening process, extracts cross-modal spatiotemporal features using a spatiotemporal convolutional network to generate a high-dimensional state vector, and dynamically calculates the dynamic torque compensation value online based on a hierarchical near-end policy optimization (PPO) algorithm. This achieves multimodal information fusion, solves the problem of spatiotemporal asynchrony of multimodal sensor data, and improves the control accuracy and real-time performance of edge computing of industrial tightening equipment compared to methods using fixed thresholds. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart illustrating a dynamic torque compensation method for high-precision tightening operations, provided as an embodiment of the present invention; Figure 2 This is a schematic diagram of the dynamic torque compensation system provided in an embodiment of the present invention; Figure 3 This is a flowchart of multimodal feature fusion provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of PPO model training provided in an embodiment of the present invention; Figure 5 A schematic diagram of the structure of a dynamic torque compensation device for high-precision tightening operations provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a computer-readable storage medium provided in an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described with reference to the accompanying drawings and embodiments.

[0017] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments, and should all fall within the protection scope of the present invention.

[0018] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are applicable in distinguishing similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or applicable to such processes, methods, products, or apparatus.

[0019] It should also be noted that the division of multiple embodiments in this invention is only for the convenience of description and should not constitute a special limitation. Features in various embodiments can be combined and referenced in each other without contradiction.

[0020] like Figure 1 As shown, the first embodiment of the present invention proposes a dynamic torque compensation method for high-precision tightening operations, which includes, for example, the following steps: Step S1, constructing a multimodal sensing module array to collect multimodal physical signals during the tightening operation; Step S2, using a spatiotemporal convolutional network to extract cross-modal spatiotemporal features of the multimodal physical signals, performing feature fusion on the cross-modal spatiotemporal features to generate a fused feature vector; Step S3, calculating a dynamic torque compensation value based on the fused feature vector using a hierarchical near-end strategy optimization algorithm; Step S4, adjusting the tightening operation according to the dynamic torque compensation value.

[0021] Specifically, in combination Figure 2 As shown, in step S1, the multimodal sensing module integrates, for example, a high-frequency vibration sensor, an infrared thermal imager, and a micro quantum strain gauge. The high-frequency vibration sensor detects the thread friction frequency and measures the vibration acceleration; the infrared thermal imager monitors the thermal distribution of material deformation; and the micro quantum strain gauge performs nanoscale deformation detection.

[0022] In step S2, a spatiotemporal convolutional network (STCN) is used to extract multimodal features and generate a high-dimensional state vector. The following is a detailed explanation of the spatiotemporal convolutional network: The spatiotemporal convolutional network employs a spatiotemporally separated convolutional branch, where the input layer takes three independent inputs: vibration signal, thermal imaging data, and strain data, and performs preprocessing on each. The vibration signal is processed using a short-time Fourier transform (STFT) to generate a time-spectrum image, including (time × frequency) information; the thermal imaging data is processed to generate time-series slices, including (space × time × temperature) information; and the strain data is processed to generate a raw time-series signal, including (time × strain) information.

[0023] The spatiotemporal convolutional network includes temporal convolutional kernels and spatial / spectral convolutional kernels. The temporal convolutional kernels slide along the time axis to extract dynamic evolution features (such as the trend of vibration frequency changes). The spatial / spectral convolutional kernels process the spatial distribution of thermal imaging or the energy concentration areas of vibration spectra. The spatiotemporal convolutional network outputs spatiotemporal feature maps for each modality (with dimensions uniformly set to Batch×Time×Channel).

[0024] The spatiotemporal convolutional network employs a cross-modal attention mechanism, for example, by passing the three modal feature maps through linear layers to generate query vectors (Q), key vectors (K), and value vectors (V). The formula is expressed as: ;in, The feature map of the m-th mode. , , This is a trainable weight matrix.

[0025] Furthermore, cross-attention calculations are performed between different modalities. For example, through a query-key-value attention mechanism, the spatiotemporal correlation weights of the query vector of the vibration signal and the key vector of the thermal imaging data are calculated, and the value vectors are weighted and fused. The formula is expressed as: ; The query vector and the key vector are derived from sensors with different sampling rates, and feature alignment is achieved through a dynamic dimension projection layer.

[0026] In the above formula, the query vector The Q-vector represents the "problem" feature extracted from the vibration signal, characterizing the target of current interest (such as high-frequency anomalous vibration). For example, if an anomalous frequency component of 12kHz is detected in the vibration signal at a certain moment, the Q-vector encodes the spatiotemporal distribution of this frequency. (Key vector) This represents the "index" features extracted from thermal imaging data, characterizing potential correlations that can be queried (such as temperature gradient changes), for example, the coordinates (X, Y) of hotspot regions and the rate of temperature rise (°C / s) in the temperature field of a threaded contact surface. The K-vector encodes these spatial-temperature correlations. Value vector The V vector represents the "content" features extracted from thermal imaging data, which include the original information (such as absolute temperature values) that needs to be weighted and fused. For example, the original temperature values ​​(unnormalized gray values) of thermal imaging pixels at the same time are retained by the V vector for subsequent fusion.

[0027] Similarly, the attention interactions between other modalities (heat → strain, strain → vibration, etc.) are calculated. Finally, the attention outputs are concatenated and passed through a fully connected layer to generate a fused feature vector.

[0028] Furthermore, the spatiotemporal convolutional network employs a dynamic convolutional kernel design. Since traditional methods using fixed-size convolutional kernels struggle to adapt to the spatiotemporal scale differences in multimodal data (e.g., small kernels are needed for high-frequency vibrations, while large kernels are needed for thermal imaging to capture spatial correlations), this embodiment introduces a dynamic convolutional kernel generator. This generator adjusts the convolutional kernel parameters in real time based on the characteristics of the input data, as expressed by the formula: ; in, It is a lightweight LSTM network, and the input is the current sensor state (such as sampling rate and signal-to-noise ratio) and historical context (such as feature distribution of the previous 10 seconds).

[0029] In step S3, combined Figure 3 As shown, for example, when deploying an edge computing unit, based on the PPO (Proximity Policy Optimization) algorithm principle, a dynamic torque compensation value is output. The multimodal parameter-torque coupling compensation algorithm expression introduced here is: ; in, For temperature parameters, For vibration term parameters, These are parameters for the cross-coupling term.

[0030] The temperature parameter can be further expanded as follows: ,Right now ;in, This represents the actual temperature measured by the sensor, reflecting the current temperature state of the system, and is the direct input value for dynamic compensation; The reference temperature is usually the ambient temperature or the initial temperature of the system (such as the temperature at the beginning of tightening). It serves as the zero point for temperature compensation and is used to quantify the deformation deviation caused by thermal expansion or contraction. This represents the rate of temperature change, reflecting the impact of instantaneous temperature fluctuations on the system. High-frequency temperature changes can lead to the accumulation of thermal stress within the material, requiring further investigation. Dynamically adjust the compensation amount; This represents the temperature-torque coupling coefficient, which is related to the material's thermal expansion coefficient and stiffness. It may change nonlinearly with temperature, and engineering calibration requires experimental fitting, such as measuring the relationship between torque deviation and temperature difference in a temperature step test. The temperature change rate sensitivity coefficient represents the system's response speed to transient temperature changes. During rapid temperature rise, thermal stress accumulates inside the material, requiring dynamic adjustment of the compensation amount.

[0031] Furthermore, when > At that time, the material deforms due to thermal expansion (such as bolt elongation), and needs to be repaired by... Correcting torque output, and conversely correcting contraction effect; As an ambient temperature reference, it can distinguish the impact of system self-heating from ambient temperature fluctuations. For example, if the ambient temperature drops suddenly (such as during winter outdoor operations), the system needs to prioritize compensating for external disturbances; temperature change rate and The real-time performance is directly related. For example, during rapid temperature rise (such as motor startup), the high rate of temperature change needs to be controlled by... Increase the compensation range to prevent overshoot.

[0032] Among the vibration term parameters ;in, This represents the correction factor for the deformation strain ε on the vibration-temperature coupling. Its specific form can be determined experimentally based on the material properties. The fundamental vibration-temperature gain coefficient reflects the efficiency of the material's damping characteristics in converting vibration energy into heat energy, and is related to the material's internal friction and microstructure. The spectral energy weighting factor is the proportion of vibration energy in the main frequency band that contributes to the thermal effect, which is determined by extracting the frequency domain energy distribution through STFT (Short Time Fourier Transform). The nonlinear attenuation coefficient characterizes the attenuation characteristics of high-frequency vibration energy transfer efficiency. It is related to the viscoelasticity of the material (such as the lower λ value of rubber materials). In experimental calibration, λ is often taken as 0.1~1.0 s² / m. Vibration acceleration is collected in real time by MEMS accelerometers or laser Doppler vibration meters (LDV). It needs to be combined with wavelet packet decomposition for noise reduction. High acceleration (such as |a|>10g) will lead to a significant increase in internal frictional heat generation in the material. . ;in, This is the deformation-torque conversion coefficient, which is positively correlated with the material's yield strength (e.g., η≈0.8 for steel, η≈0.6 for aluminum alloys). This is a dynamic strain rate correction factor, similar to the dynamic adjustment strategy of the aforementioned thermal deformation model.

[0033] Among the cross-coupling parameters, the thermal-strain coupling coefficient ;in, The temperature gradient and mechanical stress coefficient reflect the non-uniformity of thermal expansion of materials (such as the thermal elongation effect of lead screw). The dynamic rate of change coupling factor characterizes the nonlinear superposition of temperature change rate and strain rate (e.g., rapid temperature rise leading to increased lattice distortion); second derivative. The strain acceleration is used to provide early warning of material fatigue or fracture risks (such as triggering shutdown protection when the value exceeds a threshold).

[0034] Furthermore, the AI ​​output is corrected by loading constraints from the material physics equations (material yield strength formula) to prevent overfitting; for metallic materials with obvious yielding phenomena (such as low-carbon steel), the yield strength calculation formula is constructed as follows: ; Where Re represents the yield strength (unit: MPa); Fe represents the constant force (N) during the yielding stage, that is, the minimum stable load when the material yields; This represents the original cross-sectional area (mm²) of the sample.

[0035] For metallic materials without a clear yield phenomenon, the specified non-proportional elongation strength is used as the yield strength, expressed as: ; in, This indicates the force (N) required for the material to undergo 0.2% plastic deformation; other parameters have the same meaning as above.

[0036] Increased temperature leads to a decrease in yield strength; the formula is then revised as follows: ; in, This is the temperature correction factor.

[0037] In addition, increasing the strain rate will increase the yield strength, which can be corrected using a dynamic hardening model.

[0038] In step S4, combined Figure 4 As shown, for example, experimental data is collected for model training, model parameters are adjusted and optimized based on experimental results, and the parameter tuning results are verified in a digital twin platform. The optimized model will transmit the compensation parameters to the dynamic compensator for compensation adjustment, achieving sub-microsecond dynamic adjustment of the tightening operation, thereby improving the yield rate of the tightening operation.

[0039] In summary, the first embodiment of this invention proposes a dynamic torque compensation method for high-precision tightening operations. This method acquires multimodal physical signals during the tightening process using a multimodal sensing module, extracts cross-modal spatiotemporal features using a spatiotemporal convolutional network to generate a high-dimensional state vector, and dynamically calculates the dynamic torque compensation value online based on a hierarchical near-end policy optimization (PPO) algorithm. This achieves multimodal information fusion, solves the problem of spatiotemporal asynchrony of multimodal sensor data, and improves the control accuracy and real-time performance of edge computing of industrial tightening equipment compared to methods using fixed thresholds.

[0040] In addition, such as Figure 5As shown, the second embodiment of the present invention also proposes a dynamic torque compensation device for high-precision tightening operations, which includes, for example, a multimodal signal acquisition module 201, a cross-modal feature fusion module 202, a dynamic torque compensation value calculation module 203, and a compensation adjustment module 204.

[0041] The multimodal signal acquisition module 201 is used to construct a multimodal sensor module array to acquire multimodal physical signals during the tightening operation; the cross-modal feature fusion module 202 is used to extract the cross-modal spatiotemporal features of the multimodal physical signals using a spatiotemporal convolutional network, perform feature fusion on the cross-modal spatiotemporal features, and generate a fused feature vector; the dynamic torque compensation value calculation module 203 is used to calculate the dynamic torque compensation value based on the fused feature vector using a hierarchical near-end strategy optimization algorithm; and the compensation adjustment module 204 is used to perform compensation adjustment on the tightening operation based on the dynamic torque compensation value.

[0042] The dynamic torque compensation device for high-precision tightening operations disclosed in the second embodiment of the present invention implements the dynamic torque compensation method for high-precision tightening operations as described in the first embodiment above, and therefore will not be described in detail here. Optionally, each module and the other operations or functions described above in the second embodiment are respectively for implementing the method described in the first embodiment, and the beneficial effects of the dynamic torque compensation device for high-precision tightening operations provided in this embodiment are the same as the beneficial effects of the dynamic torque compensation method for high-precision tightening operations provided in the first embodiment above, and for the sake of brevity, they will not be repeated here.

[0043] like Figure 6 As shown, the third embodiment of the present invention also proposes an electronic device, for example including: at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit performs the method described in the first embodiment, and the beneficial effects of the electronic device provided in this embodiment are the same as the beneficial effects of the dynamic torque compensation method for high-precision tightening operations provided in the first embodiment.

[0044] like Figure 7 As shown, the fourth embodiment of the present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method. The beneficial effects of the computer-readable storage medium provided in this embodiment are the same as those of the dynamic torque compensation method for high-precision tightening operations provided in the first embodiment.

[0045] The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0046] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0047] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0048] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0049] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0050] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0051] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, 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. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0052] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0053] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

[0054] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0055] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic torque compensation method for high-precision tightening operations, characterized in that, include: A multimodal sensor module array was constructed to collect multimodal physical signals during the tightening operation. A spatiotemporal convolutional network is used to extract cross-modal spatiotemporal features of the multimodal physical signal, and feature fusion is performed on the cross-modal spatiotemporal features to generate a fused feature vector. The spatiotemporal convolutional network includes a time-dimensional convolutional kernel and a spatial / spectral-dimensional convolutional kernel. The time-dimensional convolutional kernel slides along the time axis to extract dynamic evolution features. The spatial / spectral-dimensional convolutional kernel processes the spatial distribution of thermal imaging or the energy concentration region of vibration spectrum. The dynamic torque compensation value is calculated based on the fused feature vector using a hierarchical near-end strategy optimization algorithm. The tightening operation is compensated and adjusted according to the dynamic torque compensation value; The expression for calculating the dynamic torque compensation value based on the hierarchical proximal strategy optimization algorithm using the fused feature vector is as follows: ； in, For temperature parameters, For vibration term parameters, For parameters of cross-coupling terms; Among the temperature parameters , This represents the actual measured temperature, where, Indicates the reference temperature. Indicates the temperature-torque coupling coefficient. This represents the sensitivity coefficient to temperature change rate. Among the vibration term parameters ,in, This represents the correction factor for the deformation strain ε on the vibration-temperature coupling. Based on the vibration-temperature gain coefficient, For spectral energy weighting factor, The nonlinear attenuation coefficient is... It is the vibration acceleration; ,in, The deformation-torque conversion coefficient, This is a dynamic strain rate correction factor; Among the cross-coupling term parameters, ,in, For temperature gradient and mechanical stress coefficient, The dynamic rate of change coupling factor, second derivative This refers to strain acceleration.

2. The dynamic torque compensation method for high-precision tightening operations according to claim 1, characterized in that, The method involves using a spatiotemporal convolutional network to extract cross-modal spatiotemporal features of the multimodal physical signal, and then performing feature fusion on these cross-modal spatiotemporal features. Generate the fused feature vector, including: The vibration signal, thermal imaging data, and strain data are independently input using a spatiotemporal separation convolution method. The three signals are input independently and preprocessed separately to obtain the spatiotemporal feature maps of each mode. The spatiotemporal feature map is used to generate trainable weight matrices for each modality through a linear layer; The similarity matrix between modalities is calculated based on the trainable weight matrix, and the attention interaction between each modality is calculated based on the similarity matrix. The attention interactions are output and concatenated, then passed through a fully connected layer to generate the fused feature vector.

3. The dynamic torque compensation method for high-precision tightening operations according to claim 2, characterized in that, The calculation formula for the attention interaction is expressed as follows: ； in, Represents the query vector. Represents the key vector. Represents a value vector. This represents the feature dimension of the key vector; The query vector of the vibration mode and the key vector of the thermal imaging mode are used to calculate the spatiotemporal correlation weights, and the value vectors are weighted and fused to obtain the attention interaction between the vibration mode and the thermal imaging mode.

4. The dynamic torque compensation method for high-precision tightening operations according to claim 1, characterized in that, Also includes: A dynamic convolution kernel generator is introduced to adjust the convolution kernel parameters in real time according to the characteristics of the input data. The expression of the dynamic convolution kernel generator is: ； in, It is a lightweight LSTM network, and the input is the current sensor state and historical context.

5. The dynamic torque compensation method for high-precision tightening operations according to claim 1, characterized in that, Also includes: For metallic materials exhibiting a clear yielding phenomenon, the formula for calculating the yield strength is as follows: ； Where Re represents the yield strength, and Fe represents the constant force maintained during the yielding stage. Indicates the original cross-sectional area of ​​the sample; For metallic materials without a clear yield phenomenon, the specified non-proportional elongation strength is used as the yield strength, expressed as: ； in, This represents the force required for a material to undergo 0.2% plastic deformation.

6. A dynamic torque compensation device for high-precision tightening operations, used to implement the dynamic torque compensation method for high-precision tightening operations as described in any one of claims 1 to 5, characterized in that, include: A multimodal signal acquisition module is used to build a multimodal sensor module array to acquire multimodal physical signals during the tightening operation. The cross-modal feature fusion module is used to extract cross-modal spatiotemporal features of the multimodal physical signal using a spatiotemporal convolutional network, perform feature fusion on the cross-modal spatiotemporal features, and generate a fused feature vector. The dynamic torque compensation value calculation module is used to calculate the dynamic torque compensation value based on the hierarchical proximal strategy optimization algorithm according to the fused feature vector. The compensation adjustment module is used to compensate and adjust the tightening operation according to the dynamic torque compensation value.

7. An electronic device, characterized in that, include: A memory and one or more processors connected to the memory, the memory storing a computer program, the processor executing the computer program to implement the dynamic torque compensation method for high-precision tightening operations as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable commands for performing the dynamic torque compensation method for high-precision tightening operations as described in any one of claims 1-5.

Images