Design method of torque wrench with positioning function

Abstract

The application discloses a design method of a torque wrench with a positioning function, and comprises the following steps: first, a 3-axis coordinate system is established, and an attitude angle is set; second, a torque wrench motion attitude matrix is determined, a space rotation relationship between a torque wrench self-coordinate system and an initial coordinate system is analyzed, and positioning of the torque wrench is realized; third, the attitude angle is calculated according to the motion attitude matrix; fourth, the existence of errors is considered, and Kalman filtering is used to reduce the errors; and fifth, a process of simulating fastening of a bolt in the field is carried out, a motion track of the torque wrench is saved to a storage, and then the track is sent to an upper computer by using a communication module, and then the track of each bolt is analyzed to check whether the torque wrench reaches each point, so that accurate control is realized. The application realizes real-time collection of a track and time record during assembly of a bolt, and solves the problem of whether each bolt at a special position is assembled in place according to requirements.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bolt assembly, in particular to a design method of a torque wrench with positioning function. BACKGROUND

[0002] Torque wrenches are widely used in the power and transportation industries to tighten different bolts with a certain pre-tightening force to ensure the normal connection of some electrical mechanical equipment and normal operation. Because of its simple structure, flexible assembly, reliable connection, and detachable and reusable characteristics, bolt fasteners are also widely used in the shipbuilding, railway, and aviation industries. Wireless torque wrenches apply corresponding torque to the bolts, increasing the firmness and tightness of the thread connection between them and preventing gaps and relative slipping between the connecting parts. The German Association of Engineers classifies standard fasteners according to the instruction VDI2862, and for bolt connection positions where connection failure can cause injury or death, both torque and angle of rotation must be monitored. When tightening the bolts of the brake system, the torque and angle of rotation must be determined in advance. For key threaded connections that require sufficient and uniform pre-tightening force on ships, the torque and angle method must be used to assemble each bolt to its maximum tightening effect to ensure higher safety.

[0003] In actual production, the torque and angle method is also one of the key development directions of tightening technology, so the production tool should best meet the new requirements: it can measure both torque and angle. However, most torque wrenches on the domestic market cannot measure the angle, and the torque wrenches that can measure the angle from abroad are expensive, and most small and medium-sized enterprises are reluctant to spend a lot of money to purchase such products in order to reduce costs, which makes the space for the torque and angle method tightening process to play narrow and the assembly quality of the product difficult to improve. In order to improve the assembly quality of domestic products, reduce production costs, and improve the competitiveness of Chinese enterprises in the market, it is of great significance to design an intelligent torque wrench that can measure the angle and meet the current production needs.

[0004] In the existing bolt assembly, the bolts are generally fixed according to a certain percentage of the total number of bolts in a plane, and after the position and posture of the connected structure are adjusted, the remaining bolts are tightened. When tightening the bolts, there are still some problems hidden in the assembly with a torque wrench. The sequence of bolt installation is often arbitrary, especially when it needs to be operated according to the percentage of the total number of bolts, the position and sequence of bolt installation can only be completed by manual experience, which leads to unguaranteed bolt installation quality and low construction efficiency, which is extremely harmful to large warships. SUMMARY

[0005] Invention purposes: The purpose of the present application is to provide a torque wrench with positioning function, which aims to improve the construction efficiency of bolts and the fine management of torque wrenches. Real-time collection of trajectory and time record during bolt assembly solves the problem of whether each bolt at special positions is assembled in place according to the requirements, and embodies the traversal of the positioning torque wrench.

[0006] Technical solutions are as follows:

[0007] A design method of a torque wrench with positioning function, characterized by comprising the following steps:

[0008] Firstly, a 3-axis coordinate system is established, and the attitude angle includes: the angle of rotation around the X-axis is called roll angle, the angle of rotation around the Y-axis is called pitch angle, and the angle of rotation around the Z-axis is called yaw angle;

[0009] Secondly, the nine-axis gyroscope is analyzed to realize real-time collection of 3-axis acceleration, 3-axis angular acceleration and 3-axis magnetometer, determine the motion attitude matrix of the torque wrench, analyze the space rotation relationship between the torque wrench coordinate system and the initial coordinate system, and realize the positioning of the torque wrench;

[0010] Thirdly, the attitude angle is calculated according to the motion attitude matrix;

[0011] Fourthly, the Kalman filter is used to reduce errors considering the existence of errors;

[0012] Fifthly, the process of tightening the bolt is simulated on site, the motion trajectory of the torque wrench is saved to the storage, then the communication module is used to send the trajectory to the upper computer, and then the trajectory of each bolt is analyzed to see whether the torque wrench reaches each point, so as to realize accurate control.

[0013] Further, the direction cosine method is used to determine the motion attitude matrix of the torque wrench, and the direction cosine method is that in the motion process of the torque wrench, the acceleration sensor will follow the target object motion, at this time the gravitational acceleration will produce displacement components in the X, Y and Z three axes of the coordinate; first, rotate the roll angle γ around the X-axis, rotate the pitch angle θ around the Y-axis, and rotate the yaw angle around the Z-axis Each rotation is represented by a direction cosine, as shown in the following formula:

[0014] Rotate the X-axis by γ angle:

[0015]

[0016] Rotate the Y-axis by θ angle:

[0017]

[0018] Rotate the Z-axis by angle:

[0019]

[0020] The transformation from the self-coordinate system to the initial state coordinate system is represented by the product of the direction cosines of the three rotation processes:

[0021]

[0022] In the formula, is a motion attitude matrix, and is represented by the product of the direction cosines of the three rotation processes; when calculating the offset component of the gravity acceleration on the coordinate axis, the motion attitude matrix is specifically:

[0023]

[0024] Each element in the motion attitude matrix is represented by a character T with a subscript:

[0025]

[0026] The heading angle, the pitch angle and the roll angle can be obtained from the above motion attitude matrix, as follows:

[0027] θ 主 = arcsin (-T 13 )

[0028]

[0029]

[0030] Further, the determination of the torque wrench motion attitude matrix adopts a quaternion method, the quaternion method is data solving by three-axis speed collected by the inertial sensor module, initially, the torque wrench is horizontally static, the attitude angle and the angular velocity are zeroed, and the initial quaternion is obtained; then the real-time updating of the quaternion is performed by using the fourth-order Runge-Kutta method, and the three-axis angular velocity read by the gyroscope is solved into the attitude angle.

[0031] Further, after the attitude angle is obtained in the third step, programming is performed by using C language in the MDK environment, the motion trajectory of the torque wrench is obtained, saved to the storage and sent to the host computer through the communication module, and then the trajectory of each bolt is observed to see whether each point of the torque wrench reaches.

[0032] Further, in the fourth step, considering the existence of errors, Kalman filtering is used to reduce the errors, the Kalman filtering algorithm observes the input and output of the system, and on this basis, the state of the system is optimally estimated, and the system equation is as follows:

[0033] X(k) = AX(k-1) + BU(k) + w(k)

[0034] Z(k) = HX(k) + v(k)

[0035] where X(k) is the state vector of the system; Z(k) is the observation vector of the system; A and B are system matrices; H is an observation matrix; w(k) is state noise; v(k) is observation noise;

[0036] According to the above formula, we can obtain the state estimation formula and the covariance of the previous time as follows:

[0037]

[0038] P(k|k-1) = AP(K-1|K-1)A T +Q

[0039] wherein: represents the state estimation of the current time obtained by the previous time prediction, and P(k|k-1) represents the covariance of the current time obtained by the previous time prediction; Q is the covariance matrix of w(k);

[0040] From the above two formulas, we can calculate the Kalman gain formula as

[0041]

[0042] wherein R is the covariance matrix of v(k);

[0043] From the predicted state estimation and the actual measurement value, the Kalman filter state estimation and covariance of the system can be calculated as shown in the following formula:

[0044]

[0045] P(k) = [I-K(k)H]P(k-1)

[0046] The Kalman filter continuously circulates through time update and state update to make the optimal estimation value for each time; the Kalman filter algorithm first predicts the value of the next state according to the initial value, then determines the covariance, calculates the Kalman gain, and then calculates the optimal estimation value according to the predicted value and the measured value, and then updates the covariance; and then according to the new covariance, the results are selected for use.

[0047] Further, an extended Kalman filter is used, which is based on the traditional Kalman filter algorithm, expands the nonlinear part in the system function by Taylor expansion, ignores the highest term, and approximates the original nonlinear system as a linear system, and then uses the traditional Kalman filter method to filter the linearized system model;

[0048] The extended Kalman filter discretizes the system equation dynamic equation as follows:

[0049] X(k) = f[X(k-1)] + Gw(k-1)

[0050] Z(k) = h[X(k)] + v(k)

[0051] Wherein, v(k) is observation noise, the observation noise is Gaussian white noise with mean 0, G is noise distribution matrix;If the observation noise matrix and the process noise matrix are independent of each other, and there is an initial state estimate And the covariance matrix P(0|0);The state prediction equation is:

[0052] X(k|k-1) = f[X(k-1)]

[0053] The covariance prediction method is:

[0054] P(k|k-1) = AP(K-1|K-1)A T +Q

[0055] The extended Kalman gain is:

[0056]

[0057] The state update equation of the extended Kalman filter algorithm is:

[0058]

[0059] The covariance update equation of the extended Kalman filter algorithm is:

[0060] P(k) = [I-K(k)H]P(k-1)

[0061] Wherein, A, H are Jacobian determinants of nonlinear systems, which are obtained by partial derivation of f and h;

[0062] The extended Kalman filter algorithm is proposed to filter data, and the external force of the wrench is analyzed, and a correction factor is introduced to further reduce the error.

[0063] Beneficial effects: compared with the prior art, the torque wrench and inertial navigation are combined to make a torque wrench with positioning function, which not only ensures the normal connection between the sonar cover and the steel ship structure, but also solves the problem of whether each bolt at the special position is assembled in place according to the specific requirements, and embodies its traversal. Not only in the field of ships, but also in the fields of military and civilian, the torque wrench with positioning function can be used to solve the bolt assembly completeness problem of special positions. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1The hardware module diagram of the torque wrench

[0065] Figure 2 The STM32 peripheral circuit

[0066] Figure 3 The strain gauge signal quantization circuit

[0067] Figure 4 The hardware circuit design diagram of the buzzer

[0068] Figure 5 The coordinate system of the nine-axis chip is established

[0069] Figure 6 The quaternion method attitude angle solution flow

[0070] Figure 7 The flow chart of Kalman filtering work

[0071] Figure 8 The flow chart of the modified Kalman filtering algorithm DETAILED DESCRIPTION

[0072] The technical solutions of the present application will be further described below in combination with the drawings.

[0073] The technical solutions adopted by the present application include a hardware part, a torque wrench with positioning function, the hardware part including a single-chip microcomputer, a sensor, an A / D conversion module, a storage module, a power module, a wireless communication module, a display module, a key, an upper computer, an audible and visual alarm module, and a positioning module, as shown in Figure 1

[0074] The selection of the MCU is a very important link for the design of the torque wrench, which needs to ensure that the MCU has sufficient resources and GPIO ports to control the data acquisition chip and realize communication with the external, and at the same time, the low-power design principle of the torque wrench should also be considered. In addition, in the process of data acquisition, the A / D sampling rate is high, the sampling point number is large, and the data processing contains floating-point operation, so the calculation amount is large. The system also has certain requirements for the speed of data transmission, so the performance requirements of the MCU are quite high.

[0075] ​Considering the system resource requirement, the main control chip of the intelligent wrench is selected as the STM32L4 series high-performance controller based on ARM Cortex-M4. The STM32L4 is launched by STMicroelectronics to meet the market demand for low-power MCUs, and is particularly suitable for low-power product applications. The specific model of the MCU of the system is STM32L433RBT6, which can work at a maximum frequency of 80 MHz and has a supply voltage of 1.71 V to 3.6 V. Its ultra-low power performance provides a variety of working modes: low-power running mode, low-power sleep mode, stop mode, standby mode, and shutdown mode. Five I / O ports have the ability to wake up the standby / shutdown mode.

[0076] The STM32L433RBT6 also has rich peripheral resources, including 83 GPIO ports, 3 IIC, 3 SPI, 3 USART, and 1 low-power UART, 1 SAI (serial audio interface), and 1 CAN. The analog components include a 12-bit AD converter, 2 12-bit DA converter channels, 2 analog comparators, and 1 operational amplifier. The memory includes 128 KB of flash memory and 64 KB of static random access memory. The minimum system circuit composed of the MCU is shown in Figure 2

[0077] ​The power supply circuit provides reliable working power for the system, ensuring the normal and stable operation of the intelligent wrench, and is an important component of the torque wrench. Considering the portability of the torque wrench, the system uses rechargeable lithium batteries for power supply. Compared with traditional zinc-manganese dry batteries, rechargeable lithium batteries can avoid frequent battery replacement, have low self-discharge rate, long cycle life, high energy density, and are more environmentally friendly. The standard voltage provided by the lithium battery is 4.2V, and a 3.3V control power supply is directly generated from the battery for the MCU only. The specific model of the power management chip is LD39015M33R, which has ultra-low voltage drop, low static current and low noise characteristics, making it suitable for low-power battery-powered applications. That is, this ceramic capacitor can ensure that the battery still works normally when the battery is low. The input voltage range of the chip LD39015M33R is 1.5V to 5.5V, and the maximum current that can be provided is 150mA, and the typical voltage difference is 80mV. The power supply rejection at low frequency is 65dB, and starts to decline at 10kHz. The chip can be controlled to start and stop mode by enabling logic, and the total current consumption in the off mode is less than 1μA. The power module also introduces a capacitor, which has good filtering effect on the charging and discharging characteristics of the capacitor, making the input and output more stable. The system uses lithium batteries for power supply, and the hardware design also needs to consider the charging problem. In this system, a single-input, single-section lithium battery, 800mA charging management chip with model BQ21040 is selected. BQ21040 has high integration, and is a linear charger suitable for portable applications. The charging voltage accuracy of the chip is 1%, and the leakage current is low, which is 1μA. In addition, the chip also has a 6.6V overvoltage protection function.

[0078] The torque measurement module selects a resistance strain gauge. The torque detection of the torque tool is to detect the strain of the metal handle through the strain gauge fixed on the handle of the tool, and to convert the torque value, and the strain gauge is generally connected in the form of a bridge. The voltage signal collected in the bridge circuit is an analog signal, which needs to be A / D converted. This design selects the 4-channel, 24-bit analog-to-digital converter ADS1220 of Texas Instruments. ADS1220 supports a wide operating voltage range of 2.3V-5.5V and a low current consumption of 120μA. The chip contains a programmable gain amplifier, a 2% precision oscillator and a 0.5℃ precision temperature sensor. The sampling rate of the chip is as high as 2KHz, supporting two differential inputs or four single-ended inputs, suitable for small signal measurement applications. The strain gauge signal quantization circuit is shown in Figure 3

[0079] ​For the selection of the positioning module, there are many products on the market that can meet the measurement angle requirements, however, to realize its positioning function, only relying on six-axis sensor is not enough, nine-axis can be more accurate than six-axis, and considering the performance index, price, product maturity, market supply, therefore, the design selects MPU9250 or ICM-20948, because they have 3-axis MEMS gyroscope, 3-axis MEMS accelerometer and 3-axis magnetometer. The precision is improved by adding a magnetic field meter to assist the calculation.

[0080] In order to make the wrench exchange data with the PC, WIFI is used for communication. A high integration and mature technology ESP8266 is selected as the communication module of the design. The module has powerful functions, supports three modes, and has simple and efficient AT commands, which is convenient for development. In addition, the module is small in size and low in cost, and is convenient for embedding into various electronic products.

[0081] If the pre-tightening force loaded on the bolt is too large, the bolt and the connecting surface will be deformed, and even the bolt will be pulled off. In order to avoid such situation, an alarm process should be used to remind the user of the wrench when the preset pre-tightening force is about to be reached during the actual use of the wrench. The intelligent wrench designed in the subject uses sound and light alarm to jointly alarm. The sound prompt module is composed of a fixed frequency buzzer, which responds differently in different processes of the wrench loading, so as to remind the user. Figure 4 The hardware circuit design of the buzzer is shown in the figure. The buzzer is driven by NPN triode Q1, and is connected with the I / O pin of MCU through BELL. The buzzer makes sound when the high level is high, and different sound effects can be designed by arranging the length of high level, so as to prompt different working states. In this way, the operator can operate directly according to the sound prompt without looking at the display screen.

[0082] The technical scheme adopted by the application includes a software part, which is written in C language under MDK environment, and is debugged and downloaded by LINK simulator. The specific steps are as follows:

[0083] Firstly, the coordinate system is established according to the selected nine-axis chip, the angle of rotation around the X-axis is called roll, the angle of rotation around the Y-axis is called pitch, and the angle of rotation around the Z-axis is called yaw.

[0084] Secondly, the application mainly uses the analysis of the nine-axis gyroscope to realize the real-time collection of 3-axis acceleration, 3-axis angular acceleration and 3-axis magnetometer. It is to calculate the first point, upload it to the computer through the communication module to record the first point (initial point), and then calculate the second point through displacement, and so on. In this way, the positioning can be recorded. At this time, we need to understand the space rotation relationship between the torque wrench coordinate system and the initial coordinate system. In order to complete the positioning of the torque wrench, the motion attitude matrix needs to be determined. Next, two methods are introduced:

[0085] (1) Direction cosine method: as above Figure 5 The nine-axis chip establishes a coordinate system. During the movement of the torque wrench, the acceleration sensor will move with the target object. At this time, the gravity acceleration will produce displacement components in the X, Y and Z axes of the coordinate. First, rotate the roll angle γ around the X axis, then rotate the pitch angle θ around the Y axis, and finally rotate the heading angle around the Z axis And each rotation can be represented by a direction cosine, as shown in the following formula:

[0086] Rotate the X axis by γ angle:

[0087]

[0088] Rotate the Y axis by θ angle:

[0089]

[0090] Rotate the Z axis by angle:

[0091]

[0092] From the self-coordinate system to the initial state coordinate system, the direction cosine of the three rotation processes can be represented by the product of the direction cosines:

[0093]

[0094] In the formula, is the attitude matrix, which represents the product of the direction cosines of the three rotation processes. Therefore, when calculating the offset component of the gravity acceleration on the coordinate axis, we know the specific number of the attitude matrix:

[0095]

[0096] Each element in the attitude matrix is represented by a character T with a subscript:

[0097]

[0098] The heading angle, pitch angle and roll angle can be obtained from the above attitude matrix as follows:

[0099] θ 主 = arcsin(-T 13 )

[0100]

[0101]

[0102] (2) The quaternion method. This method is based on the three-axis velocity collected by the inertial sensor module. Initially, the torque wrench is horizontal and static, the attitude angle and angular velocity are set to zero, and the initial quaternion is obtained. Then the real-time updating of the quaternion is carried out by using the fourth-order Runge-Kutta method, and the three-axis angular velocity read by the gyroscope is calculated into the attitude angle. The specific steps are as follows Figure 6 :

[0103] Thirdly, according to the above two methods, we can obtain the required attitude matrix, and according to the calculation, we can obtain the required attitude angle. Then we use C language to program in MDK environment to obtain the required result and analyze it.

[0104] Fourthly, considering the existence of errors, Kalman filtering is used to reduce errors. The Kalman filtering algorithm observes the input and output of the system, and optimally estimates the state of the system on this basis. The system equation is as follows:

[0105] X(k) = AX(k-1) + BU(k) + w(k)

[0106] Z(k) = HX(k) + υ(k)

[0107] Where: X(k) is the state vector of the system; Z(k) is the observation vector of the system; A and B are system matrices; H is the observation matrix; w(k) is the state noise; υ(k) is the observation matrix.

[0108] According to the above formula, we can obtain the state estimation formula and covariance of the previous time as follows:

[0109]

[0110] P(k|k-1) = AP(K-1|K-1)A T +Q

[0111] Where: represents the state estimation of the current time predicted from the previous time, and P(k|k-1) represents the covariance of the current time predicted from the previous time; Q is the covariance matrix of w(k).

[0112] From the above two equations, we can calculate the Kalman gain formula as

[0113]

[0114] where R is the covariance matrix of υ(k).

[0115] From the predicted state estimate, the actual measurement, the Kalman filter state estimate and covariance of the system can be calculated as shown in the following equations:

[0116]

[0117] P(k) = [I - K(k)H]P(k-1)

[0118] Kalman filter is constantly circulating through time update and state update, so the optimal estimate value can be obtained at each time. The Kalman filter algorithm first predicts the value at the next state according to the initial value, then determines the covariance, calculates the Kalman gain, and calculates the optimal estimate value according to the predicted value and the measurement value, and then updates the covariance. Then, according to the new results, the appropriate results are selected for use. The specific process of Kalman filter work is shown in the following figure. Figure 7

[0119] Extended Kalman filter is based on the traditional Kalman filter algorithm. The nonlinear part in the system function is expanded by Taylor, and the highest term is ignored. The original nonlinear system is approximated as a linear system, and then the traditional Kalman filter method is used to filter the linearized system model.

[0120] Unlike Kalman filter, the extended Kalman filter discretizes the system equation dynamic equation as follows:

[0121] X(k) = f[X(k-1)] + Gw(k-1)

[0122] Z(k) = h[X(k)] + v(k)

[0123] where v(k) is the observation noise, the observation noise is a Gaussian white noise with mean 0, and G is the noise distribution matrix. If the observation noise matrix and the process noise matrix are independent of each other, and there is an initial state estimate and the covariance matrix P(0|0). Then the state prediction equation is:

[0124] X(k|k-1) = f[X(k-1)]

[0125] The covariance prediction method is:

[0126] P(k|k-1) = AP(K-1|K-1)A T +Q

[0127] ​The extended Kalman gain is:

[0128]

[0129] The state update equation of the extended Kalman filter algorithm is:

[0130]

[0131] The covariance update equation of the extended Kalman filter algorithm is:

[0132] P(k) = [I - K(k)H]P(k-1)

[0133] Wherein, A, H are Jacobian determinants of the nonlinear system, which are obtained by partial derivation of f and h.

[0134] As the system runs, the calculated attitude angle will have the problem of error accumulation. Therefore, the extended Kalman filter algorithm is proposed to filter the data, and the external force on the wrench is analyzed, and a correction factor is introduced to further reduce the error. Through the analysis of Figure 8 It can be seen that the improvement of the filtering algorithm is to further reduce the error in the prediction step. The improved filtering algorithm adds mechanical prediction on the basis of the original algorithm, predicts the state with two constraint conditions, and the corrected state prediction error will be smaller.

[0135] Step 5, simulate the process of tightening the bolt in the field, the motion trajectory of the torque wrench will be saved to the storage, then sent to the host computer by using the communication module, and then analyze the trajectory of each bolt, whether the torque wrench reaches every point, so as to achieve accurate control.

[0136] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method of designing a torque wrench with a positioning function, characterized in that, The method comprises the following steps: First, establish a 3-axis coordinate system and set the attitude angle, which includes the roll angle, the pitch angle and the yaw angle; Second, analyze the nine-axis gyroscope to realize real-time collection of 3-axis acceleration, 3-axis angular acceleration and 3-axis magnetometer, determine the motion attitude matrix of the torque wrench, analyze the space rotation relationship between the coordinate system of the torque wrench and the initial coordinate system, and realize the positioning of the torque wrench; Third, calculate the attitude angle according to the motion attitude matrix, and use C language programming in the MDK environment to obtain the motion trajectory of the torque wrench; Fourth, consider the existence of errors and use Kalman filtering to reduce errors; Fifth, simulate the process of fastening the bolt on site, the motion trajectory of the torque wrench is saved to the storage, then sent to the host computer through the communication module, and then the trajectory of each bolt is analyzed to see whether the torque wrench reaches each point to realize accurate control.

2. The design method of a torque wrench with positioning function according to claim 1, characterized in that, The determination torque wrench motion posture matrix adopts a direction cosine method, the direction cosine method is that in the torque wrench motion process, the acceleration sensor will follow the target object motion, at this time the gravity acceleration will produce displacement components in the X, Y, Z three axes of coordinates; first set the roll angle rotating around the X axis, the pitch angle rotating around the Y axis, and the heading angle rotating around the Z axis, each rotation is represented by a direction cosine, as shown in the following formula: X-axis rotation angle: ， Y-axis rotation angle: ， Rotate about Z axis Angle: ， The transformation from the self-coordinate system to the initial state coordinate system is represented by the product of the direction cosines of the three rotation processes: ， In the formula, is a motion attitude matrix, which is expressed by the product of direction cosines of three rotation processes; when calculating the offset components of the gravity acceleration on the coordinate axes, the motion attitude matrix is specifically ， Each element in the motion pose matrix is denoted with a character T with a subscript: ， The yaw angle, the pitch angle and the roll angle can be obtained from the motion attitude matrix as follows: ， ， 。 3. The design method of a torque wrench with positioning function according to claim 1, characterized in that, The four-element number method is used to determine the motion attitude matrix of the torque wrench, the four-element number method is used to solve the data of the three-axis velocity collected by the inertial sensor module, initially, the torque wrench is horizontally static, the attitude angle and the angular velocity are set to zero to obtain the initial four-element number, then the real-time update of the four-element number is carried out by using the four-order Runge-Kutta method, and the three-axis angular velocity read by the gyroscope is solved into the attitude angle.

4. The design method of a torque wrench with positioning function according to claim 1, characterized in that, In the fourth step, the existence of errors is considered, and Kalman filtering is used to reduce errors, the Kalman filtering algorithm observes the input and output of the system, and optimally estimates the state of the system based on the observation, and the system equation is as follows: ， ， wherein: is a state vector of the system; is an observation vector of the system; A and B are system matrices; H is an observation matrix; is a state noise; is an observation matrix; According to the above formula, the state estimation formula and the covariance of the last time are as follows: ， ， wherein: represents a state estimate of the current time instant predicted from the previous time instant, represents a covariance of the current time instant predicted from the previous time instant; is a covariance matrix of is a covariance matrix of From the above two formulas, the Kalman gain formula can be calculated as ， wherein R is covariance matrix of From the predicted state estimation and the actual measurement value, the Kalman filtering state estimation and the covariance of the system can be calculated as shown in the following formula: ， ， Kalman filtering continuously circulates through time update and state update to obtain the optimal estimation value at each moment; the Kalman filtering algorithm first predicts the value at the next state according to the initial value, then determines the covariance, calculates the Kalman gain, and then calculates the optimal estimation value according to the predicted value and the measured value, and updates the covariance; then, according to the new covariance, the results are compared and analyzed to select the results for use.

5. The method of designing a torque wrench with a positioning function according to claim 4, wherein, The extended Kalman filter is used, which is based on the traditional Kalman filter algorithm, expands the Taylor series of the nonlinear part in the system function, ignores the highest term, and approximates the original nonlinear system as a linear system, and then uses the traditional Kalman filtering method to filter the linearized system model; The discrete dynamic equation of the extended Kalman filter is as follows: ， ， wherein, is the observation noise, which is Gaussian white noise with mean 0, G is the noise distribution matrix; if the observation noise matrix and the process noise matrix are independent of each other, and there is an initial state estimation and the covariance matrix ; the state prediction equation is: ， The covariance prediction method is as follows: ， The extended Kalman gain is as follows: ， The state update equation of the extended Kalman filter algorithm is as follows: ， The covariance update equation of the extended Kalman filter algorithm is as follows: ， Wherein, A, H is the Jacobian of the nonlinear system, by f, h partial derivative to get; The extended Kalman filter algorithm is proposed to filter the data, and the external force on the wrench is analyzed. A correction factor is introduced to further reduce the error.

Images