A method for improving the calibration accuracy of a digital torque wrench

Abstract

This invention provides a method for improving the calibration accuracy of digital torque wrenches, relating to the field of precision torque measurement technology. This method involves acquiring the output voltages of two sensors to construct a force application point characteristic value K, selecting calibration force application points at equal intervals, acquiring the full-range gradient standard torque and corresponding voltage signals, obtaining the signal strength value S, forming a calibration dataset, establishing a torque calculation model, solving the coefficient vector using the least squares method, and fixing it. During operation, the voltage is acquired in real time to calculate the current S and K, which are then substituted into the fixed model to output the final torque value. This invention, by constructing a full-range torque calculation model with an orthogonal coordinate system, decouples force value and position information, compensates for shear interference and geometric nonlinearity errors under short-arm conditions using a quadratic correction term, eliminates the dead zone of traditional piecewise fitting models, and significantly improves the measurement accuracy and stability of the digital torque wrench across the entire range.

Description

Technical Field

[0001] This invention relates to the field of precision torque measurement technology, specifically a method for improving the calibration accuracy of digital torque wrenches. Background Technology

[0002] Digital torque wrenches, as precision fastening tools, are widely used in precision mechanical assembly in fields such as automotive manufacturing and aerospace. Their measurement accuracy directly affects the reliability and safety of bolted connections. Traditional digital torque wrenches typically use resistance strain gauge sensors to calculate torque values ​​by measuring the deformation of an elastic body. However, in actual use, operators must adhere to fixed force application points. Any deviation in hand position (i.e., a change in the force application point) will lead to a change in the length of the lever arm, thus causing deviations in the torque measurement readings.

[0003] Patent application CN112629737A proposes a torque wrench torque measurement method based on dual bending moment sensors. This method involves installing two sets of bending moment sensors (with output voltages of [missing information]) at different positions on the lever arm. and By adjusting the amplification factor of the signal conditioning circuit or setting the correction factor, , Construct a linear difference model (or its equivalent) An attempt was made to counteract the lever arm through linear superposition. Impact on measurement results.

[0004] Although the aforementioned existing technologies have reduced the errors caused by changes in the point of force application to some extent, the following technical shortcomings still exist in the pursuit of higher accuracy full-range measurement:

[0005] Physical limitations of linear models: Existing technologies are based on the ideal Euler-Bernoulli beam theory, assuming that the sensor output is linearly related only to the bending moment. However, in actual physical processes, especially in short-arm conditions (where the force application point is close to the bending moment sensor), the additional strain generated by shear force and the stress concentration caused by the St. Venant's Principle cannot be ignored. The existing single linear subtraction model cannot fit these high-order nonlinear errors caused by physical mechanisms, making it difficult to further improve the measurement accuracy in the short-arm region.

[0006] Underfitting propagation in linear models: When using a single-plane model to perform least-squares fitting on full-range data, it is often impossible to simultaneously maintain accuracy at the long and short lever arm ends. To accommodate the nonlinear distortion at the short lever arm end, the fitting plane will deflect, causing the long lever arm region, which should have good linearity, to exhibit a large systematic deviation, making it impossible to achieve consistent high accuracy across the entire range.

[0007] Insufficient robustness of signal processing: Existing technologies mainly rely on the difference between two signals. As the core quantity for torque calculation, the difference between signals (differential mode signal) has a smaller amplitude than the sum of signals (common mode signal) in engineering practice. It is also more susceptible to random noise and circuit noise floor, resulting in a lower signal-to-noise ratio of the system. The stability of the measurement results still needs to be improved under high sensitivity requirements.

[0008] Therefore, there is an urgent need for a digital torque wrench calibration method that can overcome the defects of the above-mentioned linear model, effectively compensate for the nonlinear error of the short lever arm, and improve the signal-to-noise ratio of the system. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a method for improving the calibration accuracy of digital torque wrenches, solving problems such as physical limitations, linear underfitting, and poor signal robustness in existing linear models.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a method for improving the calibration accuracy of a digital torque wrench, wherein the digital torque wrench includes an elastic body, and a first strain sensor and a second strain sensor respectively fixed at different axial positions of the elastic body, specifically including the following steps:

[0011] S1. Acquire the output voltage of the first strain sensor. Output voltage of the second strain sensor Construct dimensionless force application point eigenvalues And within the effective change range, according to the preset step size Equal spacing selection Each target feature value is used as the calibration point for force application;

[0012] S2. At each calibrated force application point, apply multiple gradient standard torque values ​​covering the full range of the digital torque wrench using a standard torque calibrator. and synchronously collect each and the corresponding two sensor voltage signals Obtain signal strength value This forms a calibration dataset;

[0013] S3. with For strength shaft, Using the characteristic axis, a full-range torque calculation model with an orthogonal coordinate system is established. The calibration dataset is used to perform regression analysis on the torque calculation model using the least squares method to obtain the coefficient vector, and the coefficient vector is stored in the torque calculation model for solidification.

[0014] S4. Real-time data collection during operation and And calculate the current signal strength value. and the characteristic value of the point of force application Substitute the solidified torque calculation model to obtain the final torque value currently applied to the bolt.

[0015] Preferably, in step S1, each target feature value corresponds to a specific physical location as a calibration force application point, and the calibration force application points exhibit an error-matching distribution in physical space, with dense short lever arm regions and sparse long lever arm regions.

[0016] Preferably, the first strain sensor is installed on the side of the elastic body near the force application handle, and the second strain sensor is installed on the side of the elastic body near the working head; the characteristic value of the force application point in step S1 The calculation method is as follows:

[0017] ,

[0018] in, This is the output voltage of the first strain sensor; This is the output voltage of the second strain sensor.

[0019] Preferably, in step S1 The effective range of variation is determined as follows: a constant torque is applied to the digital torque wrench, and the point of application is continuously slid from the innermost end to the outermost end of the effective range of the handle. The characteristic values ​​during the sliding process are calculated and recorded in real time. maximum value and Minimum value Thus determine Effective range of change .

[0020] Preferably, the signal strength value in step S3 It is calculated by summing the voltage signals from the first and second strain sensors, representing torque information. The calculation formula is as follows:

[0021] .

[0022] Preferably, the torque calculation model in step S3 is based on... As a benchmark, introduce The linear and nonlinear correction terms achieve complete decoupling of force value information from the location information of the force application point. The specific calculation formula is as follows:

[0023] ,

[0024] in, The basic gain coefficient; This is a linear correction coefficient used to compensate for geometric linear errors caused by differences in lever arm length due to changes in the position of the force application point; It is a second-order nonlinear correction coefficient, used to compensate for the nonlinear error caused by shear force interference and Saint-Venant effect under short lever arm conditions; It is a linear correction term; It is a quadratic nonlinear correction term.

[0025] Preferably, the least squares solution process in step S3 includes:

[0026] Each group in the calibration dataset Convert to three model input items: , , ;

[0027] Constructing a design matrix Each row corresponds to a set of calibration data, and the row elements are... ;

[0028] Constructing response vectors Each element corresponds to a standard torque value of a set of calibration data. ;

[0029] The coefficient vector can be obtained directly by solving the analytical formula:

[0030] .

[0031] Preferably, the torque calculation model in step S4 has an adaptive operating condition adjustment mechanism, specifically:

[0032] When the point of force application is located in the long lever arm region, the eigenvalue Approaching 0, When the weights approach 0, the model automatically degenerates into an approximately linear model. To maintain the highest possible measurement signal-to-noise ratio;

[0033] When the point of force application is located in the short lever arm region, the eigenvalue Significantly increased, The weights are automatically increased to compensate for nonlinear errors.

[0034] This invention provides a method for improving the calibration accuracy of digital torque wrenches. It has the following beneficial effects:

[0035] 1. This invention utilizes difference ratios and eigenvalues. To eliminate the interference of force magnitude on position recognition, intensity value is used. By carrying the main torque information, the force value and position are decoupled.

[0036] 2. This invention introduces... Higher-order terms construct a continuous nonlinear model, replacing the traditional segmented threshold switching, eliminating numerical jumps at model junctions, and ensuring the continuity and smoothness of the full-range output.

[0037] 3. The model of this invention automatically adapts to physical laws. In the long lever arm region, the model is approximately linear, maintaining a high signal-to-noise ratio; in the short lever arm region, The automatic intervention effectively compensates for nonlinear errors caused by shear deformation and stress concentration. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the process of the present invention;

[0039] Figure 2 This is a schematic diagram comparing the distribution of the equidistant sampling feature of the present invention with that of traditional physical equidistant sampling in physical space;

[0040] Figure 3 This is a schematic diagram comparing the error correction capabilities of the nonlinear fusion model of this invention with those of the traditional single-plane model. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] like Figure 1-3 As shown, this embodiment of the invention provides a method for improving the calibration accuracy of a digital torque wrench. The digital torque wrench includes an elastic body, and a first strain sensor and a second strain sensor respectively fixed at different axial positions of the elastic body. The method specifically includes the following steps:

[0043] S1. Acquire the output voltage of the first strain sensor. Output voltage of the second strain sensor Construct dimensionless force application point eigenvalues And within the effective change range, according to the preset step size Equal spacing selection Each target feature value is used as the calibration point for force application;

[0044] Each target feature value corresponds to a specific physical location as the calibration force application point, and the calibration force application points exhibit an error-matching distribution in physical space, with dense short lever arm regions and sparse long lever arm regions.

[0045] In step S1 The effective range of variation is determined as follows: a constant torque is applied to the digital torque wrench, and the point of application is continuously slid from the innermost end to the outermost end of the effective range of the handle. The characteristic values ​​during the sliding process are calculated and recorded in real time. maximum value and minimum value Thus determine Effective range of change .

[0046] The first strain sensor is installed on the side of the elastic body near the force application handle, and the second strain sensor is installed on the side of the elastic body near the working head; characteristic value of the force application point. The calculation method is as follows:

[0047] ,

[0048] in, This is the output voltage of the first strain sensor; This is the output voltage of the second strain sensor.

[0049] The magnitude of constant torque does not affect the eigenvalue. The calculation results only need to satisfy that the output voltage of the first strain sensor and the second strain sensor is higher than 3 times the circuit noise floor. Usually, 5%-10% of the full scale of the digital torque wrench is selected.

[0050] Traditional methods in physical length Equal-interval sampling. This invention points out that the rate of change of error is related to the eigenvalue. Directly related. At the start of calibration, the operator slides the point of force from the innermost to the outermost end of the handle at any torque, and the microcontroller records this. The range, for example The system automatically generates targets. sequence: The operator moves the point of force application, and the real-time calculation... Data is collected when the value matches the target value.

[0051] like Figure 2 As shown, this strategy manifests in physical space as follows: in the region close to the sensor, i.e., the short lever arm, the sampling points are very dense, for example, spaced 5 mm apart; in the region far away, the sampling points are sparse, for example, spaced 15 mm apart. This perfectly matches the physical characteristics of the drastic stress field changes in the short lever arm region.

[0052] S2. At each calibrated force application point, apply multiple gradient standard torque values ​​covering the full range of the digital torque wrench using a standard torque calibrator. and synchronously collect each and the corresponding two sensor voltage signals Obtain signal strength value This forms a calibration dataset;

[0053] S3. with For strength shaft, Using the characteristic axis, a full-range torque calculation model with an orthogonal coordinate system is established. The calibration dataset is used to perform regression analysis on the torque calculation model using the least squares method to obtain the coefficient vector, and the coefficient vector is stored in the torque calculation model for solidification.

[0054] signal strength value It is calculated by summing the voltage signals from the first and second strain sensors, representing torque information. The calculation formula is as follows:

[0055] .

[0056] signal strength value It is a common-mode signal, compared to the differential-mode signal used in traditional methods. It has a larger amplitude and stronger resistance to random noise and circuit noise floor interference, which can significantly improve the signal-to-noise ratio and measurement stability of the system.

[0057] Torque calculation model with As a benchmark, introduce The linear and nonlinear correction terms achieve complete decoupling of force value information from the location information of the force application point. The specific calculation formula is as follows:

[0058] ,

[0059] in, The basic gain coefficient; This is a linear correction coefficient used to compensate for geometric linear errors caused by differences in lever arm length due to changes in the position of the force application point; It is a second-order nonlinear correction coefficient, used to compensate for the nonlinear error caused by shear force interference and Saint-Venant effect under short lever arm conditions; It is a linear correction term; It is a quadratic nonlinear correction term.

[0060] The least squares method solution process includes:

[0061] Each group in the calibration dataset Convert to three model input items: , , ;

[0062] Constructing a design matrix Each row corresponds to a set of calibration data, and the row elements are... ;

[0063] Constructing response vectors Each element corresponds to a standard torque value of a set of calibration data. ;

[0064] The coefficient vector can be obtained directly by solving the analytical formula:

[0065] .

[0066] This result is the globally optimal solution, i.e., the coefficient vector. ,Will The torque calculation model is incorporated into the microcontroller unit. During measurement, a high-precision torque value can be obtained through simple multiplication and addition operations, with no jumps across the entire measurement range; multiple sets of data are collected. Data. Converted Form. Constructing a system of linear equations. The least squares method is used to obtain This process requires no iteration, is fast, and yields the coefficient vector of the globally optimal solution.

[0067] S4. Real-time data collection during operation and And calculate the current signal strength value. and the characteristic value of the point of force application Substitute the solidified torque calculation model to obtain the final torque value currently applied to the bolt, and then perform the force application operation.

[0068] The torque calculation model has an adaptive operating condition adjustment mechanism, specifically:

[0069] When the point of force application is located in the long lever arm region, the eigenvalue Approaching 0, When the weights approach 0, the model automatically degenerates into an approximately linear model. To maintain the highest possible measurement signal-to-noise ratio;

[0070] When the point of force application is located in the short lever arm region, the eigenvalue Significantly increased, The weights are automatically increased to compensate for nonlinear errors, thus solving the problem of insufficient accuracy of traditional linear models in the short lever arm region.

[0071] Digital torque wrenches also include a microcontroller unit responsible for signal acquisition, data calculation, and result output, as well as a matching display screen and buzzer.

[0072] The final torque value is displayed in real time on the digital torque wrench's screen, and provides tiered prompts based on the preset target torque value: when the torque reaches 90% of the target value, the buzzer emits a short warning sound; when the torque reaches the target value, the buzzer emits a long beep and the screen flashes, prompting the operator to immediately stop applying force.

[0073] This invention uses the characteristic value of the force application point. The equidistant adaptive sampling strategy automatically achieves an error-matching physical sampling distribution with dense sampling in short lever arm regions and sparse sampling in long lever arm regions. This precisely matches the stress field variation law, significantly improving the effectiveness and representativeness of the calibration data and avoiding the shortcomings of traditional physical equidistant sampling in high-error regions.

[0074] Secondly, constructing through coordinate transformation - The intensity-feature orthogonal coordinate system achieves complete decoupling of force value information and force application point location information, while using the intensity value of the sum of the two signals. As the main torque load capacity, it significantly improves the system signal-to-noise ratio compared to the traditional scheme that relies on differential mode signals, and enhances the anti-interference ability and stability of the measurement results.

[0075] Third, a full-range continuous torque calculation model with a quadratic nonlinear correction term is established, which effectively compensates for the nonlinear error caused by shear force interference and Saint-Venant effect under short lever arm conditions, completely eliminates the splicing dead zone of the traditional piecewise fitting model, solves the underfitting problem that the accuracy of long and short lever arms of the linear model cannot be taken into account, and achieves smooth, continuous and consistent high accuracy of full-range output.

[0076] Furthermore, the least squares method is used to solve the global optimal model parameters in one step without iterative calculation, resulting in high computational efficiency. The model also has an adaptive working condition adjustment mechanism, which automatically degenerates into a high signal-to-noise ratio linear model when the lever arm is long, and automatically intervenes to compensate for the quadratic correction term when the lever arm is short, thus comprehensively improving the measurement reliability and safety of the digital torque wrench in precision assembly scenarios.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for improving the calibration accuracy of a digital torque wrench, the digital torque wrench comprising an elastic body, and a first strain sensor and a second strain sensor fixed at different axial positions of the elastic body, respectively, characterized in that, Specifically, the following steps are included: S1. Acquire the output voltage of the first strain sensor. Output voltage of the second strain sensor Construct dimensionless force application point eigenvalues And within the effective change range, according to the preset step size Equal spacing selection Each target feature value is used as the calibration point for force application; S2. At each calibrated force application point, apply multiple gradient standard torque values ​​covering the full range of the digital torque wrench using a standard torque calibrator. and synchronously collect each and the corresponding two sensor voltage signals Obtain signal strength value This forms a calibration dataset; S3. with For strength shaft, Using the characteristic axis, a full-range torque calculation model with an orthogonal coordinate system is established. The calibration dataset is used to perform regression analysis on the torque calculation model using the least squares method to obtain the coefficient vector, and the coefficient vector is stored in the torque calculation model for solidification. S4. Real-time data collection during operation and And calculate the current signal strength value. and the characteristic value of the point of force application Substitute the solidified torque calculation model to obtain the final torque value currently applied to the bolt.

2. The method for improving the calibration accuracy of a digital torque wrench according to claim 1, characterized in that: In step S1, each target feature value corresponds to a specific physical location as a calibration force application point, and the calibration force application points exhibit an error-matching distribution in the physical space, with dense short lever arm regions and sparse long lever arm regions.

3. The method for improving the calibration accuracy of a digital torque wrench according to claim 1, characterized in that: The first strain sensor is installed on the side of the elastic body near the force application handle, and the second strain sensor is installed on the side of the elastic body near the working head; the characteristic value of the force application point in step S1 The calculation method is as follows: ， in, This is the output voltage of the first strain sensor; This is the output voltage of the second strain sensor.

4. The method for improving the calibration accuracy of a digital torque wrench according to claim 1, characterized in that: In step S1 The effective range of variation is determined as follows: a constant torque is applied to the digital torque wrench, and the point of application is continuously slid from the innermost end to the outermost end of the effective range of the handle. The characteristic values ​​during the sliding process are calculated and recorded in real time. maximum value and minimum value Thus determine Effective range of change .

5. The method for improving the calibration accuracy of a digital torque wrench according to claim 1, characterized in that: The signal strength value in step S3 It is calculated by summing the voltage signals from the first and second strain sensors, representing torque information. The calculation formula is as follows: 。 6. The method for improving the calibration accuracy of a digital torque wrench according to claim 1, characterized in that: The torque calculation model in step S3 is based on As a benchmark, introduce The linear and nonlinear correction terms achieve complete decoupling of force value information from the location information of the force application point. The specific calculation formula is as follows: ， in, The basic gain coefficient; This is a linear correction coefficient used to compensate for geometric linear errors caused by differences in lever arm length due to changes in the position of the force application point; It is a second-order nonlinear correction coefficient, used to compensate for the nonlinear error caused by shear force interference and Saint-Venant effect under short lever arm conditions; It is a linear correction term; It is a quadratic nonlinear correction term.

7. The method for improving the calibration accuracy of a digital torque wrench according to claim 1, characterized in that: The least squares solution process in step S3 includes: Each group in the calibration dataset Convert to three model input items: , , ; Constructing a design matrix Each row corresponds to a set of calibration data, and the row elements are... ; Constructing response vectors Each element corresponds to a standard torque value of a set of calibration data. ; The coefficient vector can be obtained directly by solving the analytical formula: 。 8. The method for improving the calibration accuracy of a digital torque wrench according to claim 1, characterized in that: The torque calculation model in step S4 has an adaptive operating condition adjustment mechanism, specifically: When the point of force application is located in the long lever arm region, the eigenvalue Approaching 0, When the weights approach 0, the model automatically degenerates into an approximately linear model. To maintain the highest possible measurement signal-to-noise ratio; When the point of force application is located in the short lever arm region, the eigenvalue Significantly increased, The weights are automatically increased to compensate for nonlinear errors.

Images