Corner torque wrench calibration method

Abstract

The invention provides a corner torque wrench calibration method, which comprises the steps of simulating and executing accurate control operation aiming at dynamic resistance torque through an optimized response path, and synchronously matching torque output with angle change to obtain a simulation result after calibration precision is enhanced; obtaining deviation data in a simulation result, carrying out iterative correction on the control system by adopting a feedback algorithm, and determining a corrected torque output model to improve the overall efficiency; according to the corrected torque output model, generating a resistance torque adjustment scheme at multiple angles, and processing complex working conditions through an integrated real-time response mechanism to obtain a final dynamic simulation framework; and outputting a calibration precision index from the final dynamic simulation framework, and confirming that the index meets industrial requirements by adopting a verification algorithm, thereby completing comprehensive optimization of the wrench calibration process.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for calibrating a torque wrench. Background Technology

[0002] In the fields of mechanical manufacturing and precision calibration, the performance of the angle torque wrench directly affects product quality and safety, and its calibration process is crucial to ensuring tool accuracy.

[0003] With the advancement of industrial automation and intelligent manufacturing, calibration equipment needs greater flexibility and accuracy to meet diverse production demands.

[0004] Traditional calibration methods typically rely on manual operation or mechanical devices to provide resistance torque, but these methods are difficult to adapt to the dynamic requirements under complex working conditions, especially in scenarios that require simulating resistance torque changes at different angles, thus revealing obvious limitations.

[0005] A major drawback of existing calibration methods is their inadequacy in simulating dynamic drag torque.

[0006] Traditional equipment often uses fixed mechanical structures or simple hydraulic systems. When faced with scenarios that require precise control of resistance torque changes at different angles, these systems often lack sufficient response speed and accuracy.

[0007] For example, during calibration, the wrench needs to withstand different resistance torques at multiple angle points to verify whether its performance meets the standards. However, existing equipment cannot quickly and accurately adjust the resistance torque, which leads to a decrease in the reliability of the calibration results.

[0008] Furthermore, the control systems of these methods are often unable to adapt to dynamic changes during the calibration process in real time, which limits calibration efficiency and accuracy.

[0009] The core technical challenge lies in how to achieve accurate simulation and rapid response of dynamic resistance torque.

[0010] The primary challenge is to achieve precise control of the drag torque at different angles.

[0011] The resistance torque needs to switch rapidly and remain stable at a specific angle point, which requires the calibration equipment to have high-precision torque output capability and rapid dynamic adjustment capability.

[0012] If this problem cannot be solved, the calibration equipment will have difficulty simulating the complex torque changes under real-world operating conditions.

[0013] Secondly, the need for rapid response further exacerbates the complexity of control systems.

[0014] During calibration, the equipment needs to adjust the torque output according to the angle change within milliseconds, which places extremely high demands on the real-time performance and stability of the control system.

[0015] Unresolved issues with precise control can lead to torque output deviations, which in turn directly affect the achievement of rapid response. These two issues are interconnected and together restrict the improvement of calibration equipment performance.

[0016] Therefore, in the process of calibrating the angle torque wrench, how to achieve precise control and rapid dynamic adjustment of the resistance torque at different angles through equipment has become a key issue in improving calibration accuracy and efficiency.

[0017] For example, when calibrating a torque wrench used for automotive assembly, the equipment needs to simulate a continuously varying resistance torque from 5 Nm to 50 Nm within a rotation range of 0° to 90°, and the torque adjustment time after each angle change must not exceed 100 milliseconds; otherwise, the calibration data will be distorted, affecting the actual performance of the wrench.

[0018] Solving this problem is directly related to whether calibration equipment can meet the high-precision and dynamic industrial requirements.

[0019] The limitations of dynamic drag torque simulation and the response bottleneck of the control system make existing calibration methods difficult to adapt to calibration tasks under complex operating conditions.

[0020] Therefore, how to achieve precise control and rapid adjustment of drag torque at different angles has become a key issue in the field of calibration technology. Summary of the Invention

[0021] This invention provides a method for calibrating an angle torque wrench, mainly comprising:

[0022] The dynamic resistance torque data of the angle torque wrench during the calibration process is obtained. The torque output signal under different angle changes is collected in real time by the sensor, and the signal is converted into a digital sequence for subsequent processing to obtain the preliminary resistance torque distribution. Based on the initial resistance torque distribution, a preset simulation algorithm is used to analyze the angle change, determine the target torque value corresponding to each angle point, and generate an adjustment command sequence for control system input. The power of the control system is provided by a servo motor. Extract real-time response requirements from the adjustment command sequence. If the requirements exceed a preset threshold, activate the fast adjustment module; otherwise, maintain the current torque output to obtain an optimized response path. By optimizing the response path, precise control operations are performed for the simulation of dynamic drag torque, and the torque output is synchronized with the angle change to obtain simulation results with enhanced calibration accuracy. Obtain deviation data from the simulation results, use a feedback algorithm to iteratively correct the control system, and determine the corrected torque output model to improve overall efficiency; Based on the modified torque output model, a multi-angle resistance torque adjustment scheme is generated. By integrating a real-time response mechanism to handle complex working conditions, the final dynamic simulation framework is obtained. The calibration accuracy index is output from the final dynamic simulation framework, and the verification algorithm is used to confirm that the index meets industrial requirements, thereby completing the comprehensive optimization of the wrench calibration process.

[0023] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0024] This invention discloses a dynamic resistance torque optimization method for calibrating angle torque wrenches. Addressing the problems of inaccurate dynamic resistance torque data acquisition, low matching degree between torque output and angle changes, and insufficient response efficiency under complex operating conditions in traditional calibration processes, this invention proposes a complete solution. This invention collects torque signals at different angles in real time using sensors, converts them into digital sequences to generate a preliminary resistance torque distribution, and uses simulation algorithms to analyze angle changes, determine the target torque value, and generate an adjustment command sequence. For real-time response requirements exceeding a threshold, this invention activates a rapid adjustment module to optimize the response path, ensuring synchronous matching between torque output and angle. Through iterative correction of deviations using a feedback algorithm, this invention generates multi-angle resistance torque adjustment schemes, integrates a real-time response mechanism to handle complex operating conditions, and ultimately outputs calibration accuracy indicators that meet industrial requirements. This invention significantly improves calibration accuracy and efficiency, achieving precise control of dynamic resistance torque and stable response under complex operating conditions, providing efficient and accurate technical support for the calibration of angle torque wrenches. Attached Figure Description

[0025] Figure 1 This is a flowchart of a method for calibrating an angle torque wrench according to the present invention; Figure 2 This is a schematic diagram of a torque wrench calibration method according to the present invention; Figure 3 This is another schematic diagram of a torque wrench calibration method according to the present invention; Figure 4 This is a schematic diagram showing the interaction between the control system and the angle torque wrench in this invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] like Figure 1-3 The method for calibrating an angle torque wrench in this embodiment may specifically include:

[0028] S101. Obtain the dynamic resistance torque data of the angle torque wrench during the calibration process. Collect the torque output signal under different angle changes in real time through the sensor, and convert the signal into a digital sequence for subsequent processing to obtain the preliminary resistance torque distribution.

[0029] The torque output signal of a torque wrench at different angles is acquired in real time by a sensor, resulting in an original analog signal sequence. If the original analog signal sequence contains noise, a mean filtering algorithm is used to smooth the signal, resulting in a denoised analog signal sequence. Based on the denoised analog signal sequence, analog-to-digital conversion technology is used to convert it into a digital sequence, obtaining processable digital signal data. For the digital signal data, the torque output value corresponding to each angle is calculated, resulting in an angle-torque correspondence dataset. If outliers exist in the angle-torque correspondence dataset, they are filtered out using a preset threshold, resulting in an optimized torque distribution dataset. Based on the optimized torque distribution dataset, a linear interpolation algorithm is used to calculate the resistance torque distribution at continuous angles, obtaining a preliminary resistance torque distribution curve.

[0030] Furthermore, by using the preliminary drag torque distribution curve, the torque characteristic values ​​at key angle points are extracted to obtain the dynamic drag torque distribution characteristics during the calibration process.

[0031] In one possible implementation, the control system acquires torque signals via a torque sensor, and simultaneously via an encoder ( Figure 4 Item 2) Acquire the rotation angle signal to obtain the original analog signal sequence. If the original analog signal sequence contains noise, a mean filtering algorithm is used to smooth the signal, resulting in a denoised analog signal sequence. Based on the denoised analog signal sequence, the digital sequence after analog-to-digital conversion (A / D conversion) is indexed according to the sampling timestamp, so that each torque value... Each has a corresponding angle value. The initial formation The coordinate point sequence is a dataset representing the correspondence between angles and torques.

[0032] In one possible implementation, to ensure the calculated torque value is accurate and valid, upper and lower torque thresholds are set. and If the torque value at a certain angle in the angle-torque correspondence dataset Values ​​exceeding the upper and lower threshold ranges (such as sudden spike pulses from the sensor) are identified as outliers and directly discarded, resulting in an optimized torque distribution dataset. The upper and lower torque thresholds are dynamically set based on the specifications of the wrench being calibrated and industry standards; for example, they are typically set to 1% to 5% of the wrench's rated full-scale range as a fluctuation tolerance.

[0033] In one possible implementation, the drag torque distribution at continuous angles is calculated using a linear interpolation algorithm based on the optimized torque distribution dataset to obtain a preliminary drag torque distribution curve. If a specific angle needs to be obtained... (at a known sampling point) and Torque value between The calculation formula is:

[0034]

[0035] The algorithm calculates the resistance torque distribution at continuous angles, ultimately obtaining a preliminary resistance torque distribution curve. Further, based on this preliminary resistance torque distribution curve, torque characteristic values ​​at key angle points are extracted to obtain the dynamic resistance torque distribution characteristics during the calibration process. Specifically, according to calibration specifications (e.g., every 5° or at specific 30°, 60°, and 90° points), the corresponding precise torque values ​​are extracted from the interpolated curve. These characteristic values ​​collectively constitute the dynamic resistance torque distribution characteristics of the wrench during the calibration process.

[0036] S102. Based on the preliminary resistance torque distribution, the angle change is analyzed using a preset simulation algorithm to determine the target torque value corresponding to each angle point, thereby generating an adjustment command sequence for control system input. The power of the control system is provided by a servo motor.

[0037] The connection between the control system and the angle torque wrench is as follows: Figure 4 As shown:

[0038] Figure 4 In the diagram, item 1 represents the angle torque wrench; item 2 represents the encoder; item 3 represents the torque sensor; item 4 represents the servo motor; 5 represents the torque output end of the angle torque wrench; 6 represents the torque input at the upper end of the torque sensor; 7 represents the torque input at the lower end of the torque sensor; and 8 represents the handle of the angle torque wrench.

[0039] In use, the torque sensor has two torque inputs, an upper one and an lower one. The torque output end of the angle torque wrench is connected to the upper torque input of the torque sensor, and the motor shaft of the servo motor is connected to the lower torque input of the torque sensor. During use, the operator turns the handle of the angle torque wrench, and the servo motor experiences a resistance torque. The torque sensor measures this resistance torque, and an encoder is used to measure the rotation angle of the angle torque wrench. In other embodiments of the invention, a multiplier can also be connected in series between the motor shaft of the servo motor and the torque sensor. The multiplier is essentially a speed reducer, which can amplify the output torque of the servo motor.

[0040] The process begins by acquiring initial resistance torque distribution data. An angle change is processed using a pre-defined simulation algorithm to calculate the torque value at each angle point, resulting in a target torque set. Based on this target torque set, a linear interpolation algorithm is used to smooth the torque values, generating a continuous torque change curve, thus obtaining a smoothed torque sequence. If outliers exist in the smoothed torque sequence, a pre-defined threshold detection algorithm identifies these outliers and replaces them with the average value of neighboring points, resulting in a corrected torque sequence. Based on the corrected torque sequence, a gradient descent algorithm is used to optimize the torque distribution at angle points, generating a preliminary adjustment command sequence, thus obtaining an optimized command set. For the optimized command set, the continuity of the command sequence is analyzed. If discontinuities exceed a pre-defined threshold, spline interpolation is used to fill these discontinuities, resulting in a continuous command sequence. This continuous command sequence is then mapped to the input parameters of the control system, generating the final control command sequence, thus obtaining the system input sequence. Finally, the system input sequence is used to verify the matching degree between the torque distribution and the initial resistance torque, determining the accuracy of the control command sequence and obtaining the verification results.

[0041] In one possible implementation, the initial drag torque distribution data is acquired, and the angle change is processed through a preset simulation algorithm to calculate the torque value at each angle point, thus obtaining the target torque set. The preset simulation algorithm is a logical algorithm based on target prediction and command generation, typically including objective function modeling, dynamic compensation algorithm, and smoothing algorithm. The steps for calculating the torque value at each angle point are as follows: The dynamic drag torque distribution characteristics extracted by S101 are obtained; the algorithm sets the slope of torque increase with angle according to the calibration standard (e.g., an increase of 0.5 Nm for every 1° rotation), and uses the formula... (Where K is the stiffness coefficient determined by the simulation algorithm), the algorithm will calculate at a specific angle In order to compensate for the dynamic resistance of the wrench itself and achieve the standard required tightening state, the servo motor should output the theoretical target torque, and these calculated torque values ​​are converted into control commands for the servo motor (such as current or torque percentage signals).

[0042] In one possible implementation, if outliers exist in the smoothed torque sequence, the outliers are identified using a preset threshold detection algorithm and replaced with the average value of neighboring points to obtain a corrected torque sequence. The preset threshold detection algorithm typically employs a dynamic deviation detection algorithm. For example, the algorithm sets a sliding window (e.g., 5-10 consecutive sampling points), calculates the average value μ and standard deviation σ of the data within the window, and if the torque value of a certain sampling point... If the absolute value of the difference from the window mean μ exceeds k times the standard deviation (usually k=3, i.e., the 3σ principle), the point is defined as an outlier. Since this sequence is a smoothed moment sequence, the algorithm also monitors the slope dT / d between adjacent points. If the slope suddenly increases illogically (e.g., far exceeding the torque response limit of the servo motor), it is determined to be an anomaly caused by sensor noise or sporadic interference. Once the angle is detected... torque value at For outlier points, the system will extract the torque values ​​from n normal sampling points before and after them (usually n=1 or 2):

[0043]

[0044] Alternatively, a weighted average can be used:

[0045]

[0046] The threshold detection algorithm here uses statistical regularities to "smooth out" subtle anomalies and fills in the gaps with neighboring values, resulting in a smooth and continuous corrected torque sequence used to generate the final synchronization control signal. This method ensures the continuity of the torque sequence on the time axis, which is crucial for the "Proportional-Integral-Derivative (PID) Control Algorithm" mentioned in S104, because PID control is very sensitive to the continuity of the signal and the first derivative (rate of change).

[0047] In one possible implementation, the torque allocation at angle points is optimized using a gradient descent algorithm based on the corrected torque sequence, generating a preliminary adjustment command sequence to obtain an optimized command set. The optimization objective is to minimize the deviation between the servo motor's command output value and the actual required torque characteristic value. Specifically, the system first pre-assigns an initial motor control command (usually a current signal or torque percentage) to each angle point based on the corrected torque sequence. Then, the partial derivative of the objective function with respect to each command parameter (i.e., the gradient) is calculated. If the gradient is positive, it indicates that lowering the command reduces the error; if it is negative, the command is raised. The torque allocation value at each angle point is updated in the opposite direction of the gradient according to a set step size (learning rate). (in Let ∠J be the step size and ∠J be the gradient. The optimization process will be repeated until one of the following conditions is met:

[0048] 1. Convergence: The change in error between two consecutive iterations is lower than the preset minimum value.

[0049] 2. Meets the standard: The matching error of the calculated instruction sequence in simulation verification is lower than the accuracy threshold required by the industry standard.

[0050] In one possible implementation, the continuity of the instruction sequence is analyzed for the optimized instruction set. If the number of discontinuities in the sequence exceeds a preset threshold, spline interpolation is used to fill the discontinuities to obtain a continuous instruction sequence. The preset threshold is dynamically set based on the sampling integrity requirements and the control cycle. In practice, this threshold is usually set based on the following dimensions: maximum number of consecutive missing points (point threshold): usually set to 3 to 5 consecutive sampling points. If the number of consecutively lost points exceeds this value, the system considers that simple linear compensation can no longer guarantee accuracy and spline interpolation must be started; maximum time interval (time threshold): usually set according to the control frequency of the servo motor. For example, if the control cycle is 10ms, the threshold may be set to 30ms. Once the discontinuity duration exceeds this value, it is judged as a serious discontinuity; proportional threshold: the proportion of the total number of discontinuities to the total number of sampling points in a complete sampling sequence (e.g., exceeding 5%). However, when configuring a calibration system, the setting of this threshold usually depends on the signal transmission stability of the sensor. If your calibration environment has significant electromagnetic interference, it is recommended to lower the point threshold and call spline interpolation more frequently to ensure the smoothness of the simulation results.

[0051] When the number of discontinuities exceeds a threshold, the system abandons simple linear connections and instead employs cubic spline interpolation, connecting the known points using piecewise cubic polynomials. Spline interpolation not only ensures that the completed curve passes through the known data points but also guarantees the continuity of the first derivative (velocity) and second derivative (acceleration) at the connection points. This ensures that the servo motor's torque output and angular rotation transitions very smoothly when traversing data-deficient sections, meeting the requirements of precise control and synchronous matching.

[0052] In one possible implementation, the process of acquiring a continuous sequence of commands, mapping it to the input parameters of the control system, and generating a final sequence of control commands results in a system input sequence. The input parameters typically include torque and current commands. ), control voltage (0-10V), register value. The mapping process is as follows: based on the torque constant of the servo motor ( Mapping: ,in, It is the command torque. It is the motor torque coefficient. This refers to the reduction ratio of the speed reducer. Considering the frictional losses and efficiency of the mechanical transmission system, the system will correct the mapping relationship based on the dynamic resistance torque characteristics obtained from S101. For example, if the transmission efficiency is... Then the actual mapped motor command needs to be compensated to In control systems, servo motors cannot directly understand torque units such as "10.5 N·m". They only accept electrical signals such as current, voltage or pulses. Therefore, the physical torque range is mapped to the numerical range that the driver can receive (e.g., mapping 0-500 N·m to the driver's integer value of 0-10000).

[0053] Furthermore, the angle-based torque points are transformed into time-based control points. The refresh cycle of the control system (PLC or controller) is fixed (e.g., 1ms). The system calculates the angle corresponding to each 1ms time point at a specific speed, and then extracts the corresponding torque mapping value. During sequence generation, the system checks the rate of change (dT / dt) of the instruction. If the instruction changes too quickly, exceeding the motor's response limit (i.e., the real-time response requirement in S103), the system fine-tunes the sequence to ensure the instruction is physically executable. If control is via an industrial bus, the system encapsulates the mapped parameters into a message sequence in a specific format (containing control words, status words, and target torque values). Finally, the system input sequence is generated, which is a time-arranged electrical / digital signal stream that the motor can directly execute. This system input sequence is ultimately input into the Kalman filtering and smoothing module in S104 to ensure that every instruction entering the servo motor is accurate and smooth.

[0054] In one possible implementation, the process of verifying the matching degree between the torque distribution and the initial resistance torque through the system input sequence, judging the accuracy of the control command sequence, and obtaining the verification result involves performing a virtual loading simulation or pre-run comparison. Specifically, the system places the system input sequence (i.e., the generated electrical signal command stream) and the initial resistance torque distribution curve obtained in S101 in the same angular coordinate system. The system calculates the value at each angular point. Torque deviation on :

[0055]

[0056] in, It is the torque expected to be generated by the instruction sequence. This is the actual resistance measured initially at that angle. In addition to point-to-point errors, the system also calculates the correlation coefficient between the two curves. If the trends (slope changes) of the two curves are highly consistent, it indicates that the instruction sequence has successfully reproduced the dynamic characteristics of the resistance torque.

[0057] The accuracy of the control command sequence depends on a preset error threshold, such as a hard threshold: the system will set a maximum permissible error. (Typically set according to industrial calibration levels, such as 0.5% or 1% of full scale), if the errors at all points in the error dataset are below... If the instruction sequence is accurate, the verification is successful; if any point exceeds the threshold, the system's accuracy is insufficient. Alternatively, statistical judgment (RMS error) can be used: calculate the root mean square error; even if individual points fluctuate, as long as the overall RMS error is within a very small range, it can be considered accurate.

[0058] Furthermore, if the verification result is accurate, the current control parameters are saved, and the final simulation result is generated (entering the final stage of S104); if the verification result is inaccurate, the feedback algorithm in S105 is triggered to iteratively correct the deviation based on the distribution characteristics of the deviation data (whether it is overall high or local oscillation). If the deviation is too large, the gradient descent algorithm will be called again to re-optimize the torque output model until the matching error in the verification result is lower than the threshold.

[0059] S103. Extract real-time response requirements from the adjustment command sequence. If the requirements exceed a preset threshold, activate the fast adjustment module; otherwise, maintain the current torque output to obtain an optimized response path.

[0060] The system retrieves real-time response requirements from the instruction sequence and determines their priority by parsing the instruction content. If the priority exceeds a preset threshold, the rapid adjustment module is activated to generate a dynamic adjustment instruction. Based on the dynamic adjustment instruction, the current torque output status is obtained, and the adjustment range is determined. The torque output is updated based on the adjustment range, generating an optimized response path.

[0061] Furthermore, the execution efficiency of the optimized response path is extracted to determine whether it meets real-time requirements. If the execution efficiency fails to meet the standard, the response path is adjusted through an iterative optimization algorithm to obtain an improved path. The instruction sequence is updated based on the improved path to generate the final response solution.

[0062] In one possible implementation, real-time response requirements are obtained from the instruction sequence. The priority of these requirements is calculated by analyzing the rate of change and error rate within the instruction content. Specifically, the torque difference between adjacent angle points in the instruction sequence is analyzed; if the torque increase per unit angle is extremely large (e.g., simulating a rigid connection tightening), the priority of that instruction is increased. The difference between the actual torque transmitted from the current sensor and the target value of the instruction is analyzed; the larger the difference, the higher the risk of system loss of control, and the higher the priority. Whether the instruction involves key feature points (such as yield points or preset calibration points) is analyzed; instructions involving core accuracy points have the highest priority.

[0063] In one possible implementation, if the priority exceeds a preset threshold, the fast adjustment module is activated to generate a dynamic adjustment command. For example, when the rate of change of the torque command exceeds 80% of the servo motor's rated response slope, the priority threshold is triggered. If the time for the standard calculation logic to complete one command issuance exceeds 5ms-10ms (depending on the system frequency), it is determined that the demand priority exceeds the threshold, and the regular logic must be skipped, activating the fast adjustment module. Once activated, the fast adjustment module generates dynamic commands to replace the original static sequence by calculating a dynamic correction coefficient based on the current priority value. The original instructions When combined with the deviation feedback value, the formula is similar to:

[0064]

[0065] Then, by adjusting the pulse width modulation (PWM) adjustment, the duty cycle of the PWM waveform output to the servo driver is directly modified to achieve instantaneous increase or decrease in torque. These instructions are no longer queued through the standard instruction queue, but are directly sent to the execution register of the servo motor through a fast channel, achieving microsecond-level synchronous matching of dynamic resistance torque.

[0066] The rapid adjustment module is a hardware-accelerated or high-priority interrupt control unit, whose structure mainly includes: a feedforward control unit: which does not wait for feedback signals but directly predicts the required current compensation based on the target torque, compensating for system inertia in advance; a high-speed cache register: specifically used to store key parameters in the optimization instruction set, reducing the latency of reading data from main memory; and a hardware interrupt trigger: a high-priority interrupt is set in the control chip (such as a DSP or FPGA). Once triggered, the system immediately suspends non-core tasks (such as interface refresh and log recording) and executes torque compensation calculations at full speed.

[0067] In one possible implementation, the step of obtaining the current torque output state based on the dynamic adjustment command and determining the adjustment range can be summarized as: final adjustment range = (real-time deviation × dynamic weight) + feedforward compensation value + PID derivative correction. Specifically:

[0068] 1. Calculate the instantaneous deviation (baseline amplitude) between the target and actual values.

[0069] The system first obtains two key values: expected value. (Target torque calculated by simulation algorithm in S102) and current state value (via sensor at the current angle) (Real-time torque output signal). Basic adjustment range. .

[0070] 2. Introduce a correction coefficient for "dynamic response demand".

[0071] The adjustment magnitude is weighted according to priority: High priority (rapid changes): If the current phase is in a rapid torque increase phase (such as hard-connected simulation), the adjustment magnitude will be multiplied by a coefficient greater than 1 to achieve anticipatory compensation and prevent response lag. Low priority (stable phase): The adjustment magnitude remains linear to avoid unnecessary oscillations.

[0072] 3. Predictive compensation based on "resistance torque characteristics"

[0073] The system will refer to the dynamic resistance torque distribution characteristics obtained from S101: if the current angle point is known... Subsequently, the resistance torque will enter a rapid growth zone, and the system will anticipate and increase the adjustment range of the command. This method combines feedforward control, allowing the adjustment range to preemptively cover the hysteresis caused by mechanical systems (such as speed reducers and the elasticity of the wrench handle).

[0074] 4. Determine the final electrical signal increment using the PID algorithm.

[0075] The PID (Proportional-Integral-Derivative) control algorithm is a mathematical tool for determining the final amplitude: the proportional term (P) determines the main adjustment step size based on the current ΔT; the integral term (I) eliminates long-term system steady-state error; and the derivative term (D) predicts the deviation at the next moment based on the rate of torque change, suppressing overshoot. Ultimately, the adjustment amplitude is converted into the increment of the servo motor current command ΔI or the voltage increment ΔV.

[0076] 5. Amplitude Limitations and Safety Constraints

[0077] When determining the amplitude, the system performs amplitude limiting: Physical limit: The determined amplitude cannot exceed the maximum instantaneous torque output of the servo motor. Smoothness limit: If the calculated adjustment amplitude is too large (which may cause mechanical damage), the system will break it down into several consecutive small time steps for execution. This is why dynamic adjustment commands are generated and required to be smooth.

[0078] In one possible implementation, the path execution efficiency is extracted from the optimized response path to determine whether it meets real-time requirements. If the path execution efficiency does not meet the standard, the response path is adjusted through an iterative optimization algorithm to obtain an improved path. The path execution efficiency is typically composed of the following comprehensive indicators:

[0079]

[0080] Response delay: Calculate the lag time from the generation of the dynamic adjustment command to the sensor feedback torque reaching the target. Computational overhead: Extracts the CPU / DSP cycles consumed by the fast adjustment module in processing a single instruction. Energy consumption / stability: Extracts the number of oscillations of the motor during the adjustment process. Fewer oscillations result in higher execution efficiency.

[0081] Furthermore, the extracted efficiency metrics are compared with preset industrial implementation benchmarks: the calibration process requires that the torque output be synchronized within tiny steps of angular change (e.g., 0.01°). If there is a response delay... If the torque lags behind the angle change by more than a preset phase difference, it is determined that the real-time requirements are not met. If the path execution efficiency is not up to standard, for example, the total response time exceeds 1.5 times the system control cycle (e.g., the system cycle is 10ms, but the actual response takes 1.15ms); or during rapid adjustment, insufficient dynamic amplitude compensation causes the torque deviation of 5 consecutive sampling points to exceed 1%; or the path adjustment causes the torque output to oscillate more than 3 times (overshoot), then the response path is adjusted through iterative optimization algorithms. For example, the PID parameters in S104 are modified, the activation threshold of the rapid adjustment module is adjusted, or a feedback algorithm is introduced to analyze why the execution efficiency is low. If the oscillation is caused by too abrupt adjustment, the iterative optimization will reduce the iteration step size; if it is caused by slow response, the feedforward compensation ratio will be increased. Thus, the allocation logic of the entire instruction sequence is recalculated to obtain an improved path.

[0082] S104. Through the optimized response path, precise control operations are performed for the dynamic resistance torque simulation, and the torque output is synchronously matched with the angle change to obtain the simulation results with enhanced calibration accuracy.

[0083] Real-time data of dynamic torque and angle changes are acquired using sensors to obtain an initial dataset. Based on this initial dataset, a Kalman filter algorithm is used to denoise the dynamic torque and angle changes, resulting in a smoothed dataset. If the torque fluctuation in the smoothed dataset exceeds a preset fluctuation threshold, a proportional-integral-derivative (PI-DE) control algorithm is used to adjust the response path, obtaining optimized control parameters. Based on these optimized control parameters, the synchronization between the torque output and angle changes is adjusted in real time, resulting in a synchronization control signal. Precise control of the dynamic torque is then executed using this synchronization control signal, yielding a calibrated torque output. Based on the calibrated torque output, the matching error with the angle change is calculated, resulting in an error dataset. If the matching error in the error dataset is below a preset matching error threshold, the current control parameters are saved, yielding the final simulation result.

[0084] In one possible implementation, if the torque fluctuation in the smoothed dataset exceeds a preset fluctuation threshold, the response path is adjusted using a proportional-integral-derivative (PID) control algorithm to obtain optimized control parameters. The preset fluctuation threshold refers to the allowable deviation range between the smoothed torque value and its desired target curve, typically set to ±0.2% to ±1% of the target torque value, or set to a fixed absolute torque value (e.g., 0.1 N·m). The system calculates the current torque in real time. With target torque The difference ΔT. If |ΔT| > fluctuation threshold, the current system is determined to be in an unstable or substandard state, and the PID (Proportional-Integral-Derivative) control algorithm must be activated to adjust the response path.

[0085] Proportion adjustment: The magnitude of the adjustment is directly determined by the current level of error. The larger the error, the greater the adjustment, with the aim of quickly reducing the deviation.

[0086] Points adjustment: By accumulating past errors, the static error of the system is eliminated. This ensures that, in the final stage of calibration, the output torque stops precisely at the target value, rather than oscillating within a small deviation range.

[0087] Differential adjustment: The future trend is predicted based on the rate of change of the error. If the torque fluctuations are very large, the differential term will generate a counter-resistance, suppressing overshoot and oscillation.

[0088] The system sums the results of P, I, and D to obtain a corrected electrical signal command. This command modifies the original preliminary adjustment command sequence, forming a new response path that takes into account real-time error compensation.

[0089] Furthermore, the optimized control parameters refer to a set of core weight values ​​and their generated underlying driving parameters that, after calculation, enable the control system to achieve its optimal operating state. Specifically, these include: PID gain coefficient (…). , , Servo driver configuration values ​​(including the maximum output limit of the motor, acceleration and deceleration slope, and torque constant compensation value) and time constant (used to determine the sampling frequency and feedback period of the control loop, such as refreshing once every 1ms).

[0090] In one possible implementation, the synchronous control signal is obtained by adjusting the torque output and angle change in real time according to the optimized control parameters. This synchronous control signal is generated by superimposing the target path and feedback compensation on the time axis. The generation of the synchronous signal typically follows this process: the controller acquires the encoder's current angle and the sensor's current torque at an extremely high frequency (e.g., 1kHz or higher). Then, the deviation flow between the current actual state and the ideal trajectory preset by the optimized control parameters is calculated. This deviation is input into the PID controller and combined with the optimized parameters... , , The parameters are used to calculate the required control increment for the next moment. The calculated abstract control quantity is mapped to a PWM waveform recognizable by the servo driver. By changing the duty cycle, the current flowing to the servo motor coil is directly controlled. The system utilizes a unified clock bus (such as an EtherCAT redundant clock) to ensure that changes in the torque signal and feedback of the angle signal are aligned within microseconds, thus obtaining a synchronous control signal. This synchronous control signal typically contains all the instructions required for the driver to execute, such as torque command signals, speed limit signals, and phase lock signals.

[0091] The process of obtaining the synchronization control signal is essentially the process of fusing static optimization parameters with dynamic feedback data in real time and modulating them into drive commands. This signal is the direct drive source for the servo motor, ensuring that the torque output perfectly matches the preset simulation model in every extremely small angular step.

[0092] In one possible implementation, the process involves calculating the matching error with the angle change based on the calibrated torque output to obtain an error dataset. If the matching error in the error dataset is lower than a preset matching error threshold, the current control parameters are saved to obtain the final simulation result. In this process, the matching error refers to the deviation between the calculated actual output torque value of the servo motor after calibration and the preset target torque model at the same angle. The error is typically calculated from two dimensions: point-to-point deviation and overall trend.

[0093] Instantaneous matching error: Calculate for each sampling angle The absolute error below:

[0094] Relative error percentage:

[0095] Root Mean Square Error (RMSE): Used to assess the overall level of fit across the entire angular range.

[0096] Based on the calibration level of the torque wrench, the preset matching error threshold is typically set according to the following standards: High precision level (e.g., 0.5 or 1 grade wrench): The matching error threshold is typically set to ±0.5% or ±1.0%. Industrial application level: The matching error threshold may be relaxed to ±2% to ±3%. Dynamic response threshold: In addition to torque deviation, it may also include an angle lag threshold (e.g., the angle deviation when the torque meets the standard must not exceed ±0.1°). S105. Obtain the deviation data from the simulation results, use a feedback algorithm to iteratively correct the control system, and determine the corrected torque output model to improve overall efficiency.

[0097] Obtain deviation data from the simulation results, and determine the distribution characteristics of the deviation data through data analysis to obtain a deviation dataset. Extract key control parameters from the deviation dataset, including core gain parameters (PID parameters), dynamic compensation parameters (feedforward and inertial compensation), and system constraint and threshold parameters. Iteratively update the control parameters using a feedback algorithm such as Proportional-Integral-Derivative (PID), determining the updated control parameter set. Adjust the torque output model based on the updated control parameter set, and verify the adjusted torque output through simulation to obtain the torque output value. If the deviation between the torque output value and the target efficiency exceeds a preset threshold, optimize the torque output model using a gradient descent algorithm to obtain an optimized torque model. Generate new control commands using the optimized torque model, and verify the execution effect of the control commands using real-time simulation to determine system performance parameters, including response bandwidth, steady-state error, settling time, and calculated gain margin. Analyze the efficiency improvement based on the system performance parameters. If the efficiency improvement does not meet expectations, adjust the iteration step size of the feedback algorithm to obtain a new deviation dataset. Extract control parameters from the new deviation dataset, repeat the iterative optimization process, and determine the final torque output model.

[0098] In one possible implementation, the deviation data obtained from the simulation results includes the torque numerical deviation generated by S104, i.e., the error dataset; deviations on the time axis, such as response delay time; and deviations in the control process, such as overshoot generated by PID regulation and oscillation frequency. Determining the distribution characteristics through data analysis aims to identify the regularity of the system deviations, such as finding that the deviations are mainly concentrated in the "initial stage of force application" or exhibit a "hysteresis distribution," thereby obtaining the deviation dataset. Analysis methods typically include:

[0099] Statistical distribution analysis: Mean and variance: Determine whether the deviation is generally too high / too low (systematic deviation) or fluctuates wildly (random deviation). Normality test: Observe whether the deviation conforms to a normal distribution. If the deviation is concentrated in a specific angular interval, it indicates that the resistance simulation model for that interval has a defect.

[0100] Spatial / Angle Correlation Analysis: Analyzes the relationship between deviation and angle. For example: Does the error increase when the angle is large? This could reflect that the nonlinear resistance characteristics of the mechanical structure have not been accurately captured.

[0101] Frequency domain analysis: If the deviation exhibits periodic fluctuations, analyze the fluctuation frequency using methods such as Fourier transform to determine whether it is caused by improper frequency matching of the servo motor or mechanical resonance.

[0102] Trend extraction: Regression analysis is used to determine the evolution trend of the deviation and to judge whether the system is stabilizing or gradually getting out of control.

[0103] In one possible implementation, the torque output model is adjusted based on the updated control parameter set, and the adjusted torque output is calculated through simulation to obtain the torque output value. Specifically, the torque output model adjustment includes: function fitting adjustment: if the deviation data shows insufficient resistance simulation at a specific angle, the system will modify the local slope of the "angle-torque" function curve to make the model more closely match the actual resistance characteristics. Time-domain lead adjustment: if the system has response lag, the adjustment method will include increasing the feedforward, i.e., issuing an enhanced torque command milliseconds before the angle reaches the target value. Dynamic gain correction: based on the distribution characteristics of the deviation, the sensitivity of the control algorithm is reset, for example, increasing it in the torque mutation range. Adjust the price lower within a stable range to prevent fluctuations.

[0104] Simulation verification ensures that the modified scheme can accurately counteract resistance while improving overall operating efficiency. Before actually driving the servo motor, the system runs the program in a "virtual environment" to ensure that the adjusted model is safe and efficient. The specific operation process is as follows: the dynamic resistance torque distribution characteristics extracted by S101 are input into the simulation system as a virtual load, the control sequence generated based on the new parameters is input into the simulator, the result of the interaction between the virtual torque of the servo motor and the simulated resistance torque under the current control command is calculated, and the simulated torque output value is calculated on the time axis and angle axis. The system will then obtain a set of torque sequences after the simulation operation.

[0105] Furthermore, the system compares this simulated value (torque output value) with the target efficiency or standard torque requirement, calculates the deviation, and if the deviation still exceeds a preset threshold (set to 3% to 5% of the target efficiency), it triggers the subsequent gradient descent algorithm to perform deeper global optimization of the torque output model, obtaining an optimized torque model. If the deviation meets the target, the control parameter set is confirmed to be valid, and preparation is made to generate the final control signal. When the deviation exceeds the target, the gradient descent algorithm optimizes the parameters of the torque output model, specifically by defining a loss function. This represents the mean square error between the actual torque output and the target torque. This represents key parameters in the model (such as torque coefficient, damping constant, etc.). Calculating the gradient (finding direction): The algorithm calculates the partial derivative of the loss function with respect to each parameter by taking the derivative. This indicates which parameter, when changed, will reduce the bias most quickly. Fine-tune the parameters using the iteration step size (learning rate), following the opposite direction of the gradient: Using the updated parameters The mathematical model for torque output is revised. For example, if torque is found to be consistently lagging at high angles, the algorithm will automatically increase the pre-compensation gain in that range. Through continuous iteration, the deviation between the calculated torque output value and the target efficiency is reduced to within the threshold. S106. Based on the revised torque output model, a multi-angle resistance torque adjustment scheme is generated. Complex working conditions are handled by integrating a real-time response mechanism to obtain the final dynamic simulation framework.

[0106] Torque data and operating parameters are collected through a torque output model to obtain an initial torque distribution. Based on the initial torque distribution, an angle distribution calculation method is used to generate a multi-angle resistance torque adjustment scheme. If the response speed of the multi-angle resistance torque adjustment scheme is lower than a preset response speed threshold, the adjustment parameters are optimized through a real-time response mechanism to obtain a fast response scheme. The applicability of the fast response scheme under various operating conditions is analyzed through a complex operating condition processing module to determine the operating condition adaptation parameters. Based on the operating condition adaptation parameters, a dynamic feedback control mechanism is used to adjust the resistance torque scheme to obtain dynamic adjustment results. The dynamic adjustment results and operating condition adaptation parameters are integrated through a dynamic simulation framework to generate a dynamic simulation output. If the accuracy of the dynamic simulation output is lower than a preset accuracy threshold, the simulation parameters are optimized through a support vector machine algorithm to obtain the final dynamic simulation framework.

[0107] In one possible implementation, the acquisition of torque data and operating parameters through the torque output model is achieved synchronously with the sensor network via the control system's underlying drive interface: the system utilizes the clock pulse of the synchronization control signal generated by S104 to instantly read the current sensor values ​​such as temperature and motor speed while acquiring each torque value. The model maps these discrete operating data to torque data; for example, it records that "at a speed of 5° / s and an ambient temperature of 25°C, the torque corresponding to an angle of 90° is 100 N·m". The model converts the operating parameters into correction factors; if the temperature deviates from the standard value, the model automatically marks the temperature compensation status of that batch of data. The operating parameters are divided into three dimensions: physical environment parameters, mechanical and equipment state parameters, and characteristic parameters of the calibrated object.

[0108] In one possible implementation, the step of generating a multi-angle resistance torque adjustment scheme based on the initial torque distribution using an angle distribution calculation method specifically involves utilizing the resistance torque measured at each angle point in S101. As a benchmark, the solution first requires the servo motor to reach... At the angle, an equal and opposite force must be output to neutralize this inherent resistance. Depending on the requirements of the calibration task (e.g., simulating a 50 N·m tightening process), the target torque increment is superimposed at discrete angle points. Formula expression: The solution adjusts the torque distribution according to different calibration stages: Initial stage: sparse angle distribution, torque rises steadily; Yield / tightening stage: increased sampling frequency in key angle ranges, fine-tuning torque gain to simulate the realistic feeling of physical bolt tightening. Finally, the calculated torque value at each angle point is encapsulated into an instruction mapping table. This table contains the correspondence between "angle - target torque - expected current" and also includes a compensation mechanism: if the angle distribution shows obvious mechanical jamming (sudden increase in resistance) at 45°, the solution will preset a pulse-type adjustment instruction at that point.

[0109] In one possible implementation, if the response speed of the multi-angle resistance torque adjustment scheme is lower than a preset response speed threshold, the adjustment parameters are optimized through a real-time response mechanism to obtain a fast response scheme. The preset response speed threshold can be: a millisecond-level time threshold: usually set to 5ms-20ms. If the time from the angle encoder sensing the position change to the servo motor completing torque compensation exceeds this value, it is determined that the response is too slow; a phase difference threshold: the distance of the angle deviating from the preset point when the actual torque reaches the target value during dynamic rotation, usually set to 0.1° to 0.5°; and a frequency response threshold: the maximum command frequency (e.g., 50Hz) that the system can stably track. If the command change frequency exceeds this value, the system cannot follow in time.

[0110] The real-time response mechanism is a high-priority combination of hardware and algorithms, typically constructed as follows: Interrupt-driven architecture: Hardware interrupts are triggered by encoder pulses; once the angle changes, the CPU immediately suspends non-core tasks and prioritizes torque calculation. Feedforward control link: Based on conventional feedback (PID), a prediction path is added to directly predict the required current value for the next moment based on the current rotational speed and issue it in advance. Fast adjustment module: This is the module mentioned in S103, which contains a preset instruction cache and can skip complex model reconstruction, directly calling the mapped correction parameters.

[0111] When the response speed of the multi-angle resistance torque adjustment scheme is detected to be lower than the preset response speed threshold, the system optimizes the adjustment parameters in the following ways to improve execution efficiency: 1. Adjust the time-domain parameters of the PID controller, i.e., increase the proportional coefficient ( Increasing the slope of the initial response and shortening the rise time makes the system more responsive. Introducing differential prediction (…). 1. **Enhancing the perception of error change trends and suppressing potential overshoot in advance, thus allowing the system to operate at higher gains:** 2. **Shortening instruction computation depth, i.e., reducing the order of nonlinear fitting:** While maintaining accuracy, simplify complex spline curves or high-order polynomial fitting into locally linear piecewise calculations, reducing the floating-point operation time of the control chip. 3. **Dynamically compensating for gain updates, i.e., speed compensation:** Based on the current angular velocity w, dynamically increase the given gain of the current loop; the higher the speed, the greater the compensation force, to counteract the hysteresis generated by the motor's back electromotive force. 4. **Inertia pre-compensation:** Based on the rotational inertia of the mechanical system, pre-incorporate the driving torque parameters required to overcome inertia into the scheme. 5. **Step size compression:** In gradient descent or iterative algorithms, appropriately adjust the algebraic step size. Although this sacrifices a small amount of steady-state accuracy, it significantly improves the speed at which parameters converge to the target range, ensuring real-time performance.

[0112] In one possible implementation, a complex operating condition processing module analyzes the applicability of the fast response scheme under various operating conditions. Operating condition adaptation parameters, including temperature compensation coefficient, speed gain weight, stiffness adaptive coefficient, and filter frequency constant, are determined through coefficient optimization. Specifically, environmental variables are artificially changed in the simulation environment (e.g., the simulated temperature is increased by 10°C). The compensation point is then found, i.e., the torque deviation caused by the environmental change is calculated. Calculate the correction factor: Find a coefficient that can offset the deviation. For example, if the drag torque increases by an average of 0.1% for every 1°C increase in temperature, then determine a multiplication factor of 1.001. Integrate these correction values ​​for different variables using a lookup table or formula, forming a set of logical judgments to automatically switch parameters based on the real-time operating conditions transmitted from the sensors.

[0113] Example scenario: A wrench operates in a low-temperature (0°C) environment. Due to the increased viscosity of the lubricating grease, the resistance torque increases. Applicability analysis: The system found that the original solution responded slowly and the torque output was too low at 0°C. Determining adaptation parameters: The system extracted a "low-temperature resistance compensation value". Application result: Final control command = Original command + .

[0114] In one possible implementation, the resistance torque scheme is adjusted using a dynamic feedback control mechanism based on the operating condition adaptation parameters to obtain a dynamic adjustment result. Specifically, the multi-angle resistance torque adjustment scheme generated by S102 is used as a reference curve, and it is deformed or compensated for in real time, i.e.: 1. Proportional scaling: Based on the operating condition adaptation parameters (such as temperature compensation coefficient), the resistance torque scheme of the entire range is enlarged or reduced as a whole. Example: If the low temperature causes the system resistance to increase by 5%, the mechanism will automatically increase the torque command of all points in the scheme by 5% synchronously. 2. Local dynamic correction: During the rotation process, if a specific angle range is found... If unexpected jitter occurs, the mechanism will temporarily increase the resistance torque command for that range to smooth the output. 3. Phase advance correction: If it is detected that the torque response lags behind the angle change at high speed, the mechanism will predict and trigger the command point in the plan in advance to compensate for mechanical lag.

[0115] The dynamic feedback mechanism is a real-time closed-loop correction system, mainly composed of the following three components: 1. Real-time sensing layer: Simultaneously reads torque sensor (actual output), angle encoder (current position), and operating condition sensors (such as real-time temperature and speed) at extremely high frequencies (e.g., 1000Hz). 2. Difference comparison layer: Substitutes the real-time collected operating condition parameters into the operating condition adaptation parameter matrix to calculate the ideal expected value under the current environment and compares it with the actual feedback value. 3. Fast decision layer: Utilizes a built-in control law (such as adaptive PID or feedforward control) to instantly calculate the compensation current or torque increment required to eliminate the deviation.

[0116] In one possible implementation, if the accuracy of the dynamic simulation output is lower than a preset accuracy threshold, the simulation parameters are optimized using a support vector machine (SVM) algorithm to obtain the final dynamic simulation framework. The preset accuracy threshold typically corresponds to the allowable error range of national or industry standards (such as ISO 6789), and is usually set between ±0.5% and ±1.0% (relative to the full-scale target torque). Specifically, the passing grade is an error ≤1.0%. If the accuracy is between 0.5% and 1.0%, the system may only perform regular iterations. The SVM trigger is if the accuracy fluctuates for a long period and cannot converge to within 0.5%, or if the instantaneous error exceeds 2.0%, the system determines that the physical model (such as a linear model) has failed and must activate the support vector machine (SVM) for nonlinear regression optimization.

[0117] When the accuracy of the dynamic simulation output falls below the aforementioned threshold, it indicates that the system has encountered nonlinear disturbances (such as complex mechanical friction fluctuations or nonlinear deformation caused by temperature drift). SVM exhibits extremely high robustness when handling small-sample, nonlinear data during the calibration process. The system utilizes support vector regression to find an optimal hyperplane that can best fit the true drag torque law, thereby significantly reducing the error of the simulation output.

[0118] S107. Output the calibration accuracy index from the final dynamic simulation framework, and use the verification algorithm to confirm that the index meets industrial requirements, thereby completing the comprehensive optimization of the wrench calibration process.

[0119] Calibration accuracy index data is generated through a dynamic simulation framework. The raw data is cleaned and formatted using a data processing module to obtain a calibration accuracy dataset. If the index values ​​in the calibration accuracy dataset deviate from preset industry standards, a linear regression algorithm is used to fit the calibration parameters, resulting in an optimized calibration parameter set. Based on the optimized calibration parameter set, the configuration of the automated calibration system in the wrench calibration process is adjusted, generating an updated calibration execution plan. A verification algorithm is used to evaluate the updated calibration execution plan to determine whether the calibration accuracy index meets the industry standards, obtaining verification results. If the verification results show that the calibration accuracy index does not meet the industry standards, the data processing module analyzes the causes of the deviation and generates deviation correction parameters. Based on the deviation correction parameters, the calibration model in the dynamic simulation framework is updated, generating new calibration accuracy index data. The automated calibration system executes the new calibration accuracy index data, optimizing the wrench calibration process and obtaining the final calibration result.

[0120] In one possible implementation, the generation of calibration accuracy index data through a dynamic simulation framework involves the following specific steps:

[0121] 1. Spatiotemporal mapping and summary of dynamic errors

[0122] The dynamic simulation framework first aligns the discrete data points collected throughout the calibration cycle: placing the actual torque sequence from the S104 verification results, the target command sequence from S102, and the initial resistance reference from S101 on the same time-angle axis. At each discrete angle step... The residual deviation between the final execution result and the target value is calculated. .

[0123] 2. Multi-dimensional quantitative calculation of accuracy indicators

[0124] The framework calls its internal mathematical system module to calculate core metrics from different dimensions: relative error of the indicated value (...). ): Calculate the percentage deviation between the average torque value and the nominal value at each calibration point.

[0125]

[0126] Repeatability standard deviation(s): Calculates the consistency of force output at the same angle point using data recorded in multiple reciprocating simulations of the framework. Linearity and hysteresis index: Analyzes whether the response curves of torque as a function of angle coincide during the process from force application to force release (hysteresis effect analysis).

[0127] 3. Accuracy correction based on operating parameters

[0128] The core capability of the dynamic simulation framework lies in environmental source tracing and correction: The framework combines operating parameters (such as temperature 25°C) collected by S106 with a built-in compensation model to convert measurement data under non-standard operating conditions into accuracy performance under standard environmental conditions. Based on the system performance parameters determined by S105, it analyzes whether there are instantaneous oscillations during the dynamic adjustment process. If overshoot occurs, the framework adds an instantaneous volatility indicator to the accuracy metrics.

[0129] 4. Generate uncertainty assessment report

[0130] This is the highest level of data for calibration accuracy. The framework automatically analyzes the sources and contributions of error:

[0131] Type A uncertainty: derived from the statistical dispersion of multiple measurement data; Type B uncertainty: comprehensively considering sensor resolution, servo motor control accuracy (parameters extracted from S103 / S104), and truncation error of the simulation algorithm; Combined expanded uncertainty (U): the final output is authoritative accuracy index data such as U = 0.5%, k=2.

[0132] 5. The final generated calibration accuracy index data includes: average accuracy level: determining whether the wrench meets level 1.0, 0.5, or higher; peak error distribution: recording which angle point has the largest error in the full range, used to analyze mechanical defects; dynamic tracking coefficient: reflecting the synchronization tightness of torque and angle in the simulation frame in fast response mode; confidence interval value: providing statistical credibility support for the calibration results.

[0133] In one possible implementation, if the index values ​​in the calibration accuracy dataset deviate from a preset industry standard, a linear regression algorithm is used to fit the calibration parameters to obtain an optimized calibration parameter set. Specifically, when the accuracy deviates from the standard, the system redefines the mapping curve between the signal and the value using a linear regression algorithm such as the least squares method.

[0134] Step 1: Construct the observation dataset: The system extracts multiple data pairs from S106 and S107. For example, the independent variable x is the standard command value (ideal torque) output by the simulation framework; the dependent variable y is the actual sensor feedback and the measured value after processing.

[0135] Step 2: Establish a linear regression model: Define the linear relationship equation: Where w is the new gain coefficient and b is the new zero offset. It is the residual error.

[0136] Step 3: Solving for optimal parameters: Using the least squares method, by minimizing the residual sum of squares from all measurement points to the regression line, we calculate w and b that best approximate the true physical properties. .

[0137] Step 4: Parameter Update and Reverification: Write the calculated new w and b into the registers of the calibration controller. Rerun the data acquisition of S106 using the new parameters. If the generated index data falls within the industry standard range at this point, the calibration is considered successful.

[0138] In one possible implementation, the updated calibration execution plan is evaluated using a verification algorithm to determine whether the calibration accuracy indicators meet industry standards, resulting in a verification result. The verification algorithm is a composite algorithm based on statistical bias analysis and signal processing. It consists of the following core sub-modules: 1. Error comparison engine: compares the real-time dynamic torque data collected by S106 with preset industry standards (such as ISO 6789 or national verification procedures) point by point. 2. Uncertainty assessment operator: calculates the expanded uncertainty of the measurement results to ensure that the accuracy indicators are statistically reliable. 3. Trend stability discriminator: analyzes whether there are oscillations, hysteresis, or nonlinear abrupt changes during the calibration process.

[0139] The evaluation of the updated calibration execution scheme is mainly divided into the following three dimensions: 1. Static accuracy index evaluation (indication error): The algorithm extracts the measured values ​​at different torque points (such as 20%, 60%, and 100% of the rated range) and calculates their deviation from the standard value. If all points | If all values ​​are less than the industry standard (e.g., 1.0%), then this dimension passes the evaluation. 2. Dynamic Matching Consistency Evaluation: The algorithm uses the correlation coefficient to calculate the degree of overlap between the measured torque curve and the target simulated curve. If the correlation coefficient... This indicates that although the final torque is accurate, the resistance simulation during the intermediate process is unrealistic. Check if the curves of the loading and unloading processes highly overlap, and evaluate the idle travel error of the mechanical system. 3. Robustness and Operating Condition Adaptability Assessment: The algorithm will retrieve the performance under different operating condition parameters in S106: assess whether the operating condition adaptation parameters effectively offset the effects of the environment (such as temperature and speed). If environmental changes cause the accuracy to decrease beyond the preset range, the verification result is deemed unqualified.

[0140] Furthermore, if the verification results show that the calibration accuracy index does not meet the industry standard, the data processing module analyzes the cause of the deviation and generates deviation correction parameters. The analysis of the cause of the deviation through the data processing module includes: 1. Systematic deviation analysis (trend analysis): If the torque at all calibration points is higher than the standard value by a fixed percentage, this is usually due to sensor sensitivity drift or the influence of temperature in the calibration environment. 2. Zero-point deviation analysis: The error is large at low ranges and gradually decreases at high ranges. This is because the zero point of the sensor hardware has shifted. 3. Nonlinear deviation analysis: When the error changes curvilinearly with angle or torque (e.g., accurate in the middle, off at both ends), it usually stems from the elastic deformation of the mechanical structure or the incomplete cancellation of frictional torque changes in the reducer. 4. Dynamic hysteresis analysis: When the loading command and the feedback torque are misaligned on the time axis, this is due to improper settings of the servo controller's response step size or the parameters of the fast response module.

[0141] The process of generating the bias correction parameters is actually constructing a reverse compensation model. Specifically: 1. If the analysis conclusion is a systematic bias, the module will calculate the linear correction factor: using the slope w and intercept b fitted by linear regression (mentioned in previous steps). Parameter form: generating a new gain compensation value. 2. If the deviation exhibits a non-linear distribution with respect to angle, the module will generate a multi-point compensation matrix: in the angle range where the error is large, the compensation increment is calculated. Parameter format: Generate an "angle-torque correction" lookup table and inject it into the dynamic simulation framework. 3. If the deviation is caused by a slow response: Calculate the required increase in feedforward gain based on the response delay time. Parameter format: Update the PID parameters. Alternatively, add a time lead constant.

[0142] The above description is only 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 protection scope of the present invention.

Claims

1. A method for calibrating an angle torque wrench, characterized in that, The method includes: The dynamic resistance torque data of the angle torque wrench during the calibration process is obtained. The torque output signal under different angle changes is collected in real time by the sensor, and the signal is converted into a digital sequence for subsequent processing to obtain the preliminary resistance torque distribution. Based on the initial resistance torque distribution, a preset simulation algorithm is used to analyze the angle change, determine the target torque value corresponding to each angle point, and generate an adjustment command sequence for control system input. The power of the control system is provided by a servo motor. Extract real-time response requirements from the adjustment command sequence. If the requirements exceed a preset threshold, activate the fast adjustment module; otherwise, maintain the current torque output to obtain an optimized response path. By optimizing the response path, precise control operations are performed for the simulation of dynamic drag torque, and the torque output is synchronized with the angle change to obtain simulation results with enhanced calibration accuracy. Obtain deviation data from the simulation results, use a feedback algorithm to iteratively correct the control system, and determine the corrected torque output model to improve overall efficiency; Based on the modified torque output model, a multi-angle resistance torque adjustment scheme is generated. By integrating a real-time response mechanism to handle complex working conditions, the final dynamic simulation framework is obtained. The calibration accuracy index is output from the final dynamic simulation framework, and the verification algorithm is used to confirm that the index meets industrial requirements, thereby completing the comprehensive optimization of the wrench calibration process.

2. The method for calibrating an angle torque wrench according to claim 1, characterized in that, The process of acquiring dynamic resistance torque data of the angle torque wrench during calibration involves using sensors to collect torque output signals under different angle changes in real time, and converting these signals into digital sequences for subsequent processing to obtain a preliminary resistance torque distribution. The torque output signal of the torque wrench at different angles is collected in real time by the sensor to obtain the original analog signal sequence; If the original analog signal sequence contains noise, the mean filtering algorithm is used to smooth the signal to obtain a denoised analog signal sequence. Based on the denoised analog signal sequence, analog-to-digital conversion technology is used to convert it into a digital sequence to obtain processable digital signal data. For digital signal data, calculate the torque output value corresponding to each angle to obtain an angle-torque correspondence dataset; If there are outliers in the angle-torque correspondence dataset, the outlier data points are filtered out using a preset threshold to obtain an optimized torque distribution dataset. Based on the optimized torque distribution dataset, the drag torque distribution at continuous angles is calculated using a linear interpolation algorithm to obtain the preliminary drag torque distribution curve.

3. The method for calibrating an angle torque wrench according to claim 1, characterized in that, The step of analyzing angle changes using a preset simulation algorithm based on the initial resistance torque distribution to determine the target torque value corresponding to each angle point, thereby generating an adjustment command sequence for control system input, also includes: A linear interpolation algorithm is used to smooth the target torque value, generating a continuous torque variation curve and obtaining a smooth torque sequence. If there are outliers in the smooth torque sequence, the outliers are identified by a preset threshold detection algorithm and replaced with the average value of the neighboring points to obtain the corrected torque sequence. Based on the corrected torque sequence, the gradient descent algorithm is used to optimize the torque distribution at the angle points, generating a preliminary adjustment command sequence and obtaining an optimized command set; For the optimized instruction set, the continuity of the instruction sequence is analyzed. If the discontinuities in the sequence exceed a preset threshold, the discontinuities are filled using a spline interpolation algorithm to obtain a continuous instruction sequence. Obtain a continuous sequence of instructions, map it to the input parameters of the control system, generate the final sequence of control instructions, and obtain the system input sequence. By inputting the system sequence, the degree of matching between the torque distribution and the initial resistance torque is verified, the accuracy of the control command sequence is judged, and the verification results are obtained.

4. The method for calibrating an angle torque wrench according to claim 1, characterized in that, The process of extracting real-time response requirements from the adjustment command sequence, activating a rapid adjustment module if the requirements exceed a preset threshold, and maintaining the current torque output to obtain an optimized response path, includes: Obtain real-time response requirements from the instruction sequence and determine the priority of the requirements by parsing the instruction content; If the priority exceeds the preset threshold, the quick adjustment module is activated to generate dynamic adjustment instructions; Based on the dynamic adjustment command, obtain the current torque output status and determine the adjustment range; An optimized response path is generated by updating the torque output by adjusting the amplitude.

5. The method for calibrating an angle torque wrench according to claim 1, characterized in that, The optimized response path performs precise control operations for dynamic drag torque simulation, synchronously matching torque output with angle changes to obtain simulation results with enhanced calibration accuracy, including: The initial dataset is obtained by acquiring real-time data on dynamic torque and angle changes through sensors. Based on the initial dataset, the Kalman filter algorithm is used to denoise the dynamic torque and angle changes to obtain a smooth dataset; If the torque fluctuation in the smoothed dataset exceeds the preset threshold, the response path is adjusted using a proportional-integral-derivative control algorithm to obtain optimized control parameters. Based on the optimized control parameters, the torque output and angle change are synchronized in real time to obtain a synchronous control signal; By using synchronous control signals, precise control of dynamic torque is performed to obtain calibrated torque output; Based on the calibrated torque output, the matching error with the angle change is calculated to obtain the error dataset; If the matching error in the error dataset is lower than the preset threshold, the current control parameters are saved, and the final simulation result is obtained.

6. The method for calibrating an angle torque wrench according to claim 1, characterized in that, The process of acquiring deviation data from simulation results, using a feedback algorithm to iteratively correct the control system, and determining the corrected torque output model to improve overall efficiency includes: Obtain deviation data from the simulation results, determine the distribution characteristics of the deviation data through data analysis, and obtain the deviation dataset; Key control parameters are extracted from the deviation dataset, and a feedback algorithm is used to iteratively update the control parameters to determine the updated set of control parameters. The torque output model is adjusted based on the updated control parameter set, and the adjusted torque output is calculated and verified through simulation to obtain the torque output value. If the deviation between the torque output value and the target efficiency exceeds a preset threshold, the gradient descent algorithm is used to optimize the torque output model to obtain the optimized torque model. New control commands are generated using the optimized torque model, and the execution effect of the control commands is verified by real-time simulation to determine the system performance parameters. The efficiency improvement is analyzed based on the system performance parameters. If the efficiency improvement does not meet expectations, the iteration step size of the feedback algorithm is adjusted to obtain a new deviation dataset. Control parameters are extracted from the new deviation dataset, and the optimization process is repeated iteratively to determine the final torque output model.

7. The method for calibrating an angle torque wrench according to claim 1, characterized in that, The process involves generating multi-angle resistance torque adjustment schemes based on the modified torque output model, and handling complex working conditions through an integrated real-time response mechanism, resulting in the final dynamic simulation framework, including: The initial torque distribution is obtained by collecting torque data and operating parameters through the torque output model; Based on the initial torque distribution, an angle distribution calculation method is used to generate a multi-angle resistance torque adjustment scheme; If the response speed of the multi-angle resistance torque adjustment scheme is lower than the preset threshold, the adjustment parameters are optimized through a real-time response mechanism to obtain a fast response scheme. Through the complex working condition processing module, the applicability of the rapid response solution under various working conditions is analyzed, and the working condition adaptation parameters are determined. Based on the operating condition adaptation parameters, a dynamic feedback control mechanism is adopted to adjust the resistance torque scheme and obtain the dynamic adjustment result. By integrating dynamic adjustment results and operating condition adaptation parameters through a dynamic simulation framework, dynamic simulation output is generated. If the accuracy of the dynamic simulation output is lower than the preset threshold, the simulation parameters are optimized using the support vector machine algorithm to obtain the final dynamic simulation framework.

8. The method for calibrating an angle torque wrench according to claim 1, characterized in that, The process involves outputting calibration accuracy indicators from the final dynamic simulation framework, and using a verification algorithm to confirm that the indicators meet industrial requirements, thereby completing a comprehensive optimization of the wrench calibration process. This includes: The calibration accuracy index data is generated through a dynamic simulation framework. The raw data is then cleaned and formatted using a data processing module to obtain the calibration accuracy dataset. If the index values ​​in the calibration accuracy dataset deviate from the preset industry standard, a linear regression algorithm is used to fit the calibration parameters to obtain an optimized calibration parameter set. Based on the optimized calibration parameter set, adjust the configuration of the automated calibration system in the wrench calibration process to generate an updated calibration execution plan; The updated calibration execution scheme is evaluated by a verification algorithm to determine whether the calibration accuracy indicators meet industry standards, and the verification results are obtained. If the verification results show that the calibration accuracy index does not meet the industry standard, the data processing module will analyze the cause of the deviation and generate deviation correction parameters. Based on the deviation correction parameters, update the calibration model in the dynamic simulation framework to generate new calibration accuracy index data; By implementing new calibration accuracy index data through an automated calibration system, the wrench calibration process is optimized to obtain the final calibration result.

Images