Search

CN-121608169-B - Visual-guide positioning control system of rotary die transferring machine

CN121608169BCN 121608169 BCN121608169 BCN 121608169BCN-121608169-B

Abstract

The invention relates to the technical field of manipulator control, in particular to a visual-guided positioning control system of a rotary die transferring machine. The system comprises a joint driving unit, a flexible transmission assembly and a vision module. The method comprises the steps of firstly decoupling generalized disturbance torque representing nonlinear resistance of a flexible component from quadrature axis current fed back by a driving unit based on a joint dynamics model, secondly, utilizing an LSTM network to learn mechanical hysteresis loop characteristics of the torque in different rotary motion states, predicting terminal dynamic position compensation quantity of transmission delay time of a coverage system, and finally, carrying out undisturbed fusion on the dynamic position compensation quantity and visual positioning coordinates based on smooth weight generated by real-time running speed. By combining physical priori with data driving, the invention effectively overcomes the accumulated errors caused by nonlinear interference of cables and signal transmission hysteresis in slip ring-free rotation operation, and realizes high-dynamic precision positioning.

Inventors

  • WANG RUREN
  • BI XINGGUO

Assignees

  • 美机科技集团有限公司

Dates

Publication Date
20260508
Application Date
20260203

Claims (7)

  1. 1. A vision-guided rotary die transfer machine positioning control system, comprising: The resistance observation module is used for acquiring the cross-axis current fed back by the joint driving unit, and the joint angular velocity and the angular acceleration fed back by the encoder, calculating the rigid body theoretical moment required by the joint to overcome the moment of inertia and viscous friction based on the preset mechanical arm rigid body kinetic parameters, subtracting the rigid body theoretical moment from the electromagnetic torque converted by the cross-axis current, and decoupling to obtain generalized disturbance torque; the error prediction module takes the generalized disturbance torque and the current joint motion state sequence as joint characteristics to be input into a long-short-term memory network, learns the mechanical hysteresis loop characteristics of the generalized disturbance torque under different rotation angles, and outputs the terminal dynamic position compensation quantity covering the future moment of the system transmission delay time; The fusion control module is used for obtaining the real-time running speed of the end effector, calculating a feedforward weight through a smooth weighting function based on the real-time running speed, and carrying out weighted fusion on the dynamic position compensation quantity of the end and the visual positioning coordinate by utilizing the feedforward weight to generate a final position control instruction, wherein the final position control instruction is equal to the product of the feedforward weight and the dynamic position compensation quantity of the end, the product of the complement of the feedforward weight and the corrected visual coordinate after delay compensation is added, the sum of the product of the complement of the feedforward weight and the corrected visual coordinate is the value after delay correction of the coordinate output by the visual feedback module, and the complement of the feedforward weight is one minus the feedforward weight.
  2. 2. The vision-guided rotational-transfer-molding machine positioning control system of claim 1, further comprising a flexible transmission assembly for connecting the joint driving unit and the end effector, wherein the resistance observation module is configured to perform a segmentation processing step of reading absolute position data of the encoder in real time, calculating a unidirectional cumulative rotation angle of the joint driving unit relative to a zero point, dividing the unidirectional cumulative rotation angle into a linear resistance region and a nonlinear tension region, wherein the linear resistance region corresponds to an angle range in which the flexible transmission assembly is in a free bending state, the nonlinear tension region corresponds to an angle range in which the flexible transmission assembly is in a tension deformation state, marking generalized disturbance torque as first type characteristic data when the unidirectional cumulative rotation angle is in the linear resistance region, marking generalized disturbance torque as second type characteristic data when the unidirectional cumulative rotation angle is in the nonlinear tension region, and activating a corresponding pre-trained long-short-term memory network submodel according to an input characteristic type.
  3. 3. The vision-guided rotational-transfer-molding machine positioning control system according to claim 1, wherein the acquiring of the quadrature-axis current, the joint angular velocity and the angular acceleration further comprises a preprocessing process of performing moving average filtering on the acquired quadrature-axis current and calculating a differential value of the filtered quadrature-axis current with respect to time to obtain a current change rate, monitoring the change rate of the joint angular velocity at the same time, determining that the system is in mechanical jamming and generating an abnormal blocking flag when the current change rate is greater than a preset impact gradient threshold value and the joint angular velocity is not able to be lifted with the increase of the current is monitored, and the fusion control module is used for setting the feedforward weight to zero and locking the generated final position control command to be a current position maintaining command in response to the abnormal blocking flag.
  4. 4. The positioning control system of the vision-guided rotary die machine, which is characterized in that preset mechanical arm rigid body dynamic parameters are acquired and stored through a parameter identification process, wherein under an idle working condition that a joint driving unit is not connected with a flexible transmission assembly, the joint driving unit is controlled to execute multi-frequency-band sinusoidal scanning motion, a quadrature axis current sequence, a joint angular velocity sequence and an angular acceleration sequence in the motion process are synchronously acquired, a dynamic regression equation taking rotational inertia and viscous friction coefficients as coefficients to be determined is constructed, and the acquired sequence data is subjected to iterative solution by utilizing a least square method until residual errors are converged, so that calibration values of the rotational inertia and the viscous friction coefficients are obtained.
  5. 5. The positioning control system of the vision-guided rotary die machine, which is characterized in that the long-term and short-term memory network comprises an input attention layer, first-order differences of input generalized disturbance torque sequences are calculated, zero crossing points at which differential symbols turn over are recognized, cross correlation coefficients of the input generalized disturbance torque sequences and current joint angular velocity sequences are calculated, dynamic weights are distributed for generalized disturbance torque features of a current time step based on the zero crossing points and the cross correlation coefficients, and when the zero crossing points are detected, the dynamic weights with the weight coefficients larger than a preset weight threshold value are distributed.
  6. 6. The vision-guided rotational-transfer-modeling machine positioning control system of claim 1, wherein the smoothing weighting function is an S-type membership function based on a speed interval, and calculating the feedforward weight specifically includes presetting a first speed threshold and a second speed threshold, wherein the first speed threshold is defined as a dynamic compensation lower-effective limit, the second speed threshold is defined as a vision closed-loop upper-effective limit, and the first speed threshold is greater than the second speed threshold, calculating the feedforward weight to be one if the real-time running speed is greater than the first speed threshold, wherein the final position control instruction is completely dominated by the final dynamic position compensation amount, calculating the feedforward weight to be zero if the real-time running speed is less than the second speed threshold, wherein the final position control instruction is completely dominated by the visual positioning coordinates, and smoothly decaying from one to zero according to the S-type curve as the speed decreases if the real-time running speed is between the first speed threshold and the second speed threshold.
  7. 7. The vision-guided rotational transfer molding machine positioning control system of claim 1, wherein the vision positioning coordinates are generated by a vision feedback module comprising an industrial camera mounted on an end effector, an exposure control unit, and an image processing unit, wherein generating the vision positioning coordinates comprises: According to a preset photosensitive characteristic curve, when the real-time running speed exceeds a preset motion blur threshold value, shortening the shutter time of the industrial camera according to an inverse proportion relation, synchronously and non-linearly improving the analog gain to compensate the image brightness, and collecting an original image frame; The image processing unit receives the terminal dynamic position compensation quantity output by the error prediction module, and superimposes the terminal dynamic position compensation quantity to the current joint theoretical position fed back by the encoder, and calculates to obtain the estimated projection coordinates of the characteristic points; In the region of interest, identifying template feature points by utilizing a sub-pixel edge extraction algorithm, calculating sub-pixel level coordinates of the template feature points under an image coordinate system, calling a camera internal reference matrix calibrated in advance and a hand-eye calibration external reference matrix, mapping and converting the sub-pixel level coordinates into three-dimensional space coordinates under a robot base coordinate system, carrying out distortion correction on the converted three-dimensional space coordinates, and outputting final visual positioning coordinates to a fusion control module.

Description

Visual-guide positioning control system of rotary die transferring machine Technical Field The invention relates to the technical field of manipulator control, in particular to a visual-guided positioning control system of a rotary die transferring machine. Background In labor-intensive industries such as clothing manufacturing and automotive interior processing, automated die handling machines have been widely used for automatic die plate replacement, handling and positioning processes. The traditional automatic die changing system is generally composed of a multi-shaft mechanical arm or truss mechanical arm matched with a conveying line, in-place detection is carried out by utilizing a photoelectric sensor or a travel switch, and grabbing and placing of the templates are completed through a preset teaching track. To accommodate the processing requirements of complex curves, the die-carrying end often needs to have 360 ° rotational capability to maintain machine-direction compliance. However, the existing rotary die transferring and positioning control technology still has obvious technical bottlenecks when facing high-frequency die changing, high-precision dynamic processing and multi-variety flexible production scenes, and is mainly characterized in that firstly, the existing die transferring machine or manipulator tail end with 360-degree infinite rotation function is widely used for realizing electric power and signal transmission between a rotating component and a fixed base by adopting a conductive slip ring structure. Because the slip ring adopts contact brush wire transmission, long-term high-speed rotation can lead to contact surface abrasion and carbon powder accumulation, thereby causing instability of signal transmission. Especially in fine control requiring millisecond-scale responses, slip ring transmissions tend to have signal delays greater than 10 ms. Second, existing visual guidance systems mostly employ an open loop control mode of "still photograph-calculate coordinates-execute actions". The method comprises the steps of performing one-time photographing and positioning before the action of the manipulator, and then blindly moving the manipulator to a target point. The method can only solve the static grabbing error, but cannot cope with the dynamic track deviation generated by the mechanical gravity center change, vibration or assembly tolerance of the die-carrying machine in the 360-degree rotation process. Third, in conventional control logic, the mold handling robot is only used as a single material handling unit, and the motion control and subsequent processing parameters are independent of each other. In automated sewing, the different templates and fabric thicknesses require the manipulator to apply different pressures when placing the templates and cooperate with specific needle lifting actions to prevent the 'needle lifting without thread feeding' fault. However, the mold transporting manipulator in the prior art lacks active sensing and adapting capability to the process environment, and cannot automatically adjust the action parameters of the end effector according to the visually recognized template characteristics at the moment of mold changing or mold transporting. The current situation of 'carrying and process separation' often requires manual intervention adjustment after die changing, and quality problems such as large difference of forward and backward stitching, broken threads or empty seams and the like are easily caused by improper needle lifting and matching. In summary, how to overcome the accumulated errors caused by the signal transmission delay and the nonlinearity of the mechanical motion in the dynamic process of 360 ° continuous rotation operation of the manipulator is a technical problem to be solved in the art. Therefore, a visual-guided positioning control system of a rotary die transferring machine is provided. Disclosure of Invention The invention aims to provide a visual-guided positioning control system of a rotary die transferring machine. The system comprises a joint driving unit, a flexible transmission assembly and a vision module. The method comprises the steps of firstly decoupling generalized disturbance torque representing nonlinear resistance of a flexible component from quadrature axis current fed back by a driving unit based on a joint dynamics model, secondly, utilizing an LSTM network to learn mechanical hysteresis loop characteristics of the torque in different rotary motion states, predicting terminal dynamic position compensation quantity of transmission delay time of a coverage system, and finally, carrying out undisturbed fusion on the dynamic position compensation quantity and visual positioning coordinates based on smooth weight generated by real-time running speed. By combining physical priori with data driving, the invention effectively overcomes the accumulated errors caused by nonlinear interference of cables and signal transmi