CN-116222482-B - Corner in-situ calibration device and method based on artificial intelligence

CN116222482BCN 116222482 BCN116222482 BCN 116222482BCN-116222482-B

Abstract

The invention discloses an artificial intelligence-based corner in-situ calibration device and method. The device is characterized in that the device comprises a loading mechanism (1), an adjusting mechanism (2), a calibration connecting piece (3), a shaft to be calibrated (4) and an angle sensor to be calibrated (5), wherein the upper end and the lower end of the adjusting mechanism (2) are respectively provided with the loading mechanism (1) and the calibration connecting piece (3), the loading mechanism (1) is respectively connected with the adjusting mechanism (2) and the calibration connecting piece (3), the calibration connecting piece (3) is connected with the shaft to be calibrated (4), and the angle sensor to be calibrated (5) is fixed on the shaft to be calibrated (4). The method is characterized by comprising the steps of constructing a bidirectional long and short time memory network BLSTM model, generating a loss function, optimizing the model and obtaining a function relation. The device method can eliminate the influence of the on-line calibration interference factor, solves the problem that the sensor cannot be directly calibrated, and has the advantages of high precision and easiness in operation.

Inventors

NI BO
WENG JUN
ZHAO JILIANG
TIAN FANGYI
MAO ZHIYONG
XIONG TAO
CHEN WEI

Assignees

上海精密计量测试研究所

Dates

Publication Date: 20260508
Application Date: 20221219

Claims (6)

1. The corner in-situ calibration device based on artificial intelligence is characterized by comprising a loading mechanism (1), an adjusting mechanism (2), a calibration connecting piece (3), a shaft to be calibrated (4) and an angle sensor to be calibrated (5), wherein the upper end and the lower end of the adjusting mechanism (2) are respectively provided with the loading mechanism (1) and the calibration connecting piece (3), the loading mechanism (1) is respectively connected with the adjusting mechanism (2) and the calibration connecting piece (3), the calibration connecting piece (3) is connected with the shaft to be calibrated (4), and the angle sensor to be calibrated (5) is fixed on the shaft to be calibrated (4); The loading mechanism (1) comprises a circular grating encoder, a bearing (8), an inner shaft (9) and a torque motor (10), wherein the circular grating disk (6) is connected with the reading head (7) to form the circular grating encoder, the circular grating encoder is fixedly connected with the inner shaft (9) through the bearing, signals generated by rotation of the inner shaft (9) are measured, and the torque motor (10) is used for generating power to enable the inner shaft (9) to rotate; The adjusting mechanism (2) comprises a disc (11), a protruding coaxial adjusting ring (16) is arranged on the disc (11), holes are formed in the side face of the coaxial adjusting ring (16), long screws (15) penetrate through the holes to be in contact with the loading mechanism (1), and the positions of the loading mechanism (1) can be adjusted by rotating the long screws (15); The edge of the disc (11) is provided with an adjusting block (12), the adjusting block (12) is connected with a supporting shaft (17), a spherical groove is formed in the adjusting block (12), the levelness of the adjusting mechanism (2) can be adjusted by rotating the adjusting block (12), so that the levelness of the whole mechanism is adjusted, the angle of the inner shaft (9) is adjusted, and the coaxiality requirements of the shaft (4) to be calibrated and the inner shaft (9) are further ensured; The calibrating connector (3) is used for connecting the loading mechanism (1) and the shaft (4) to be calibrated to rotate together, and comprises a round block (13) and a curved surface thin plate (14), wherein the upper end of the round block (13) is connected with the loading mechanism (1), the lower end of the round block is connected with the curved surface thin plate (14), the curved surface thin plate (14) is connected with the shaft (4) to be calibrated, the round block (13) is connected with the inner shaft (9) through a bearing, and the curved surface thin plate (14) is driven to rotate through the rotation of the inner shaft (9), so that the shaft (4) to be calibrated is driven to rotate.
2. An artificial intelligence-based corner in-situ calibration method, which is characterized by using the artificial intelligence-based corner in-situ calibration device as claimed in claim 1, and comprising the following steps: Firstly, constructing a bidirectional long and short time memory network BLSTM model, taking data acquired by an angle sensor (5) to be calibrated, a return stroke mark, a zero angle difference between the angle sensor (5) to be calibrated and a circular grating encoder and an angle difference between the angle sensor (5) to be calibrated and the circular grating encoder under fixed sampling time as model inputs, and fitting an output value through an artificial intelligent algorithm; Step 2, generating a loss function, namely performing static step measurement of a corner, taking data acquired by a circular grating encoder as a standard value, and constructing the loss function by utilizing the difference between the output value of the BLSTM model and the standard value; Step 3, optimizing the model, namely carrying out dynamic continuous measurement of the turning angle, minimizing a loss function through a gradient descent method, and reversely optimizing parameters of the BLSTM model; and 4, obtaining a functional relation, namely establishing the functional relation between the angle sensor (5) to be calibrated and the circular grating encoder.
3. The corner in-situ calibration method based on artificial intelligence according to claim 2, wherein in step 1, firstly, a two-way long and short time memory network model, namely a BLSTM model, is established, the BLSTM model is input through an input layer, the middle of the model is composed of a BLSTM layer, a full connection layer and a hidden layer between the two layers, the BLSTM model outputs a calculation result through an output layer as an output value, The zero angle difference between the angle sensor (5) to be calibrated and the circular grating encoder is obtained by the following steps: when the loading mechanism (1) and the adjusting mechanism (2) are connected, and the connection between the calibration connecting piece (3) and the internal shaft (9) is disconnected, the servo of the torque motor (10) is started, and the circular grating encoder searches for a zero position; The calibration connecting piece (3) is connected with the internal shaft (9), and the levelness of the loading mechanism (1) is adjusted by using the level gauge through an adjusting block (12) of the adjusting mechanism (2), so that the internal shaft (9) in the center of the loading mechanism (1) is coaxial with the shaft (4) to be calibrated below; The torque motor (10) is closed to servo, the motor below the shaft (4) to be calibrated is controlled to rotate, the shaft (4) to be calibrated drives the inner shaft (9) to rotate so that the circular grating encoder rotates, the circular grating encoder with the upper part being successfully found to be zero can be driven to rotate, the angle sensor (5) to be calibrated returns to zero, and the difference between the measured values of the circular grating encoder serving as the standard sensor and the angle sensor (5) to be calibrated is the interval angle between zero positions of the two sensors, namely the zero angle difference.
4. The method for calibrating the rotation angle in situ based on the artificial intelligence according to claim 2, wherein in the step 2, the rotation angle static step measurement comprises the steps of selecting a rotation direction and a rotation angle gradient, wherein the rotation angle gradient is in a step A, controlling a torque motor (10) to rotate at a selected speed for a certain time, stopping the rotation angle A, controlling the motor to rotate for a certain time after a certain period of time, controlling the motor to rotate for a next gradient A until the motor rotates for one circle, finishing the holding for a certain time, and recording the angle value of a circular grating encoder, the angle value of an angle sensor (5) to be calibrated, and progress or return marks.
5. The artificial intelligence based corner in-situ calibration method according to claim 4, wherein in step 3, the dynamic continuous measurement of the corner comprises selecting a rotation speed and a rotation direction, controlling the torque motor (10) to rotate continuously at the selected rotation speed, and keeping for a period of time after one rotation, and recording the angle value of the circular grating encoder, the angle value of the angle sensor (5) to be calibrated, the progress or the return mark.
6. The artificial intelligence based corner in-situ calibration method according to claim 2, wherein the optimizing the model step comprises: 1) A BLSTM (bidirectional long and short time memory network) model is built, and an angle value of an angle sensor (5) to be calibrated at the moment t, a return stroke mark, a zero angle difference between the angle sensor (5) to be calibrated and a circular grating encoder and an angle difference between the angle sensor (5) to be calibrated at the moment t and the circular grating encoder are read to be taken as model input quantities ; 2) Calculating a forgetting door, and selecting information to be forgotten: ; 3) Calculating a memory gate, selecting information to be memorized: And wherein: In order to provide a memory gate which is capable of being operated by a user, Is a temporary cellular state; 4) Calculating the state of the cell at the current moment: ; 5) Calculating an output door and a hidden layer at the current moment: , And wherein: in order to output the door, the door is provided with a door opening, The cell state is the current time; 6) By forgetting and memorizing the information, information useful for calculation at a subsequent time is transferred, and useless information is discarded, and hidden layers are outputted at each time step Wherein forgetting, memorizing and outputting are performed by hidden state at the last moment And current input Calculated forgetting door Memory door Output door To control, wherein: an activation function for each node; 7) The data of the moment of the output gate t and the output values of the data of the first 20 moments and the data of the last 20 moments corresponding to the current moment are obtained ; 8) Setting a loss function E: Wherein N is the number of output layer nodes, Output for model A standard value; 9) Reverse calculation modifies weights layer by layer according to gradient descent algorithm And deviation of : , Alpha and beta are learning efficiency and deviation adjusting coefficient respectively, and N is the number of j layers of nodes; 10 Repeating the dynamic continuous measurement of the corners for a plurality of times, re-reading new time input data, repeating the steps 1) to 9) until all recorded data are processed, and realizing the optimization of the connection weight and the deviation among the neurons of each layer through continuous loop iteration, so that the model can be finally fitted with a standard encoder with ideal precision, thereby completing the corner in-situ calibration based on artificial intelligence.

Description

Corner in-situ calibration device and method based on artificial intelligence Technical Field The invention belongs to the technical field of testing and calibration, and particularly relates to an artificial intelligence-based corner in-situ calibration device and method. Background With the development of equipment manufacturing industry in China, more and more devices cannot calibrate the rotation angle sensor alone due to limited space or complicated installation, and the rotation angle sensor is widely applied, so that the reliable rotation angle in-situ calibration method is one of the key problems to be solved currently. In recent years, the development of the artificial intelligence field in China is rapid, the types of the neural networks are more and more, the training efficiency is higher and higher, and the efficiency is greatly improved compared with that of manpower when the artificial intelligence field processes some repetitive data and monotonicity work, so that the neural network model is utilized to fit some nonlinear complex function relations, and the artificial intelligence field is a good scheme. The patent CN106352973A is named as an in-situ calibration method of a sensor, the standard sensor is arranged at the top end of the sensor to be measured, the internal excitation signals are used for respectively enabling the standard sensor and the angle sensor to be calibrated to generate output signals, and the calibration can be completed by comparing the two output signals. The patent CN112539876a is named as a vertical reference type extreme environment torque calibration method and device, in the method, the difference between the measured standard torque value and the torque value to be calibrated is utilized for calibration, the function curve of the torque sensor to be measured is not fitted, other influencing factors are not considered, and the calibration effect is to be improved. Disclosure of Invention The invention aims to perform in-situ calibration based on artificial intelligence on the rotation angle sensor under the condition that the rotation angle sensor cannot be dismounted, and has the advantages of high precision and convenience in operation. The technical scheme adopted for solving the technical problems is as follows: The utility model provides a corner normal position calibrating device based on artificial intelligence, its characterized in that, it includes loading mechanism 1, adjustment mechanism 2, calibration connecting piece 3, waits to calibrate axle 4, wait to calibrate angle sensor 5, and adjustment mechanism 2 upper and lower extreme is loading mechanism 1 and calibration connecting piece 3 respectively, and loading mechanism 1 is connected with adjustment mechanism 2 and calibration connecting piece 3 respectively, and calibration connecting piece 3 is connected with waiting to calibrate axle 4, waits to be fixed with on the calibration axle 4 and wait to calibrate angle sensor 5. Further, the loading mechanism 1 comprises a circular grating encoder, a bearing 8, an inner shaft 9 and a torque motor 10, the circular grating disk 6 is connected with the reading head 7 to form the circular grating encoder, the circular grating encoder is fixedly connected with the inner shaft 9 through the bearing, signals generated by rotation of the inner shaft 9 are measured, and the torque motor 10 is used for generating power to enable the inner shaft 9 to rotate. Further, the adjusting mechanism 2 comprises a disc 11, the disc 11 is provided with a protruding coaxial adjusting ring 16, the side surface of the coaxial adjusting ring 16 is provided with holes, long screws 15 pass through the holes to contact with the loading mechanism 1, and the position of the loading mechanism 1 can be adjusted by rotating each long screw 15. Further, the adjusting block 12 is arranged at the edge of the disc 11, the adjusting block 12 is connected with the supporting shaft 17, a spherical groove is formed in the adjusting block 12, the levelness of the adjusting mechanism 2 can be adjusted by rotating the adjusting block 12, the levelness of the whole mechanism is adjusted, the angle of the inner shaft 9 is adjusted, and the coaxiality requirements of the shaft 4 to be calibrated and the inner shaft 9 are further ensured. Further, the calibration connecting piece 3 is used for connecting the loading mechanism 1 and the shaft 4 to be calibrated to rotate together, and comprises a round block 13 and a curved sheet 14, wherein the upper end of the round block 13 is connected with the loading mechanism 1, the lower end of the round block is connected with the curved sheet 14, the curved sheet 14 is connected with the shaft 4 to be calibrated, the round block 13 is connected with the inner shaft 9 through a bearing, and the curved sheet 14 is driven to rotate through the rotation of the inner shaft 9, so that the shaft 4 to be calibrated is driven to rotate. The invention also provides a