CN-122018435-A - Terminal vibration suppression method for strong impact load operation robot

CN122018435ACN 122018435 ACN122018435 ACN 122018435ACN-122018435-A

Abstract

The invention discloses a method for inhibiting terminal vibration of a strong impact load operation robot, which is characterized in that according to data of a transmission mechanism of the robot in the impact operation process, terminal vibration measured by a combined force-vision sensor is used as input, a model driving and data driving method is adopted after preprocessing, and a prediction model of the terminal vibration of the robot is established. And Model Predictive Control (MPC) is adopted to carry out closed-loop control on the vibration of the tail end of the robot in each control period, and Active Disturbance Rejection Control (ADRC) is introduced to inhibit nonlinear discontinuous vibration generated in the impact process, so that active vibration inhibition in the impact process is realized. And detecting the vibration of the tail end of the robot after the suppression, and compensating the suppression error in proportion in the next corresponding impact process by using a self-learning method. After multiple compensations tend to stabilize, the compensation curve is recorded and recalled later in the impact process. The invention can effectively solve the problems of non-linear and discontinuous vibration of the tail end of the impact load operation robot and improve the operation stability of the impact robot.

Inventors

CHEN XIANZHONG
HUANG SHUO
HOU QINGWEN
WANG LIJUN
ZHANG JIE
Xie Chaoda

Assignees

北京科技大学

Dates

Publication Date: 20260512
Application Date: 20241111

Claims (7)

1. A method for inhibiting terminal vibration of a strong impact load operation robot is mainly applied to a front strong impact open-eye operation robot of a metallurgical furnace. The method is characterized by comprising the following steps of: And S1, establishing impact vibration simulation and analyzing a model to obtain vibration characteristics of various structures of the robot in the impact process. And collecting relevant data of a transmission mechanism in the operation of the robot corresponding to the strong impact load in an experiment, measuring corresponding terminal vibration by a parallel resultant force-visual sensor as input, preprocessing, combining model driving and data driving, and establishing a prediction model of the terminal vibration of the robot under the current working condition by adopting a neural network. And S2, setting a time period for vibration suppression, and reversely compensating the terminal vibration of the robot by taking deltat as one control period. The compensation process is divided into a moving and a shocking process as shown in fig. 2, and the two stages perform different vibration compensation strategies. And S3, reversely compensating the vibration of the front and back, left and right and up and down three degrees of freedom of the tail end in the running process of the robot in each control period. And detecting the compensated tail end vibration of the robot in an experiment, compensating errors of the prediction model and the control system by using a self-learning method, and compensating the difference in the next corresponding impact process. And S4, after vibration compensation is completed, recording a compensation curve for currently inhibiting impact vibration, and calling in the subsequent actual production operation process.
2. The method for suppressing the end vibration of a high impact load working robot according to claim 1, wherein the step S1 is implemented as the steps of: and S1.1, establishing an impact vibration simulation and analysis model, and analyzing the impact vibration dynamics simulation and corresponding multi-body structure each-order modal shape cloud patterns under the application scene of the robot, and obtaining the vibration characteristics of various structures of the robot in the impact process by the established impact vibration energy flow model and the mechanical arm impact stress conduction-coupling-diffusion mechanism model. And S1.2, measuring to obtain nonlinear discontinuous vibration of the tail end of the robot during operation by adopting a force sense and vision fusion sensing method according to vibration characteristics. In the moving stage of the robot, the vibration of the tail end of the robot is mainly measured in a visual positioning mode, the visual measurement in the impact stage is shielded, the vibration of the tail end is measured by adopting a force-visual sensor joint measurement formula: Vib=α×Vib F +(1-α)×Vib V (1) Wherein, vib is the final vibration result, vib F is the vibration measured by the force sensor, vib V is the vibration measured by the vision sensor, alpha is the proportion of the force sensor, and the vibration can be adjusted according to the field condition. And step S1.3, combining the measured vibration with the mathematical model in the step S1.1 and the related data of the robot measured by the existing sensor during working, and taking the combined vibration as a training set during training of the neural network. And carrying out data preprocessing and dimension reduction on the used data, and respectively establishing a robot tail end vibration prediction model according to different impact scenes by adopting a neural network.
3. The method for suppressing the end vibration of a high impact load working robot according to claim 1, wherein the step S1.3 is implemented as the steps of: And S1.3.1, collecting and arranging data required by establishing a strong impact load operation robot tail end vibration prediction model, preprocessing the data firstly, and carrying out data alignment on sensors with different sampling periods because the follow-up control takes deltat as a control period. For the method of taking average with sampling period faster than Δt, the correction formula is: Where n is the total number of Data collected by the current sensor within Δt, data i is the collected Data, and Data new is the converted Data. For the sensor with the sampling period being slower than deltat, the data alignment is carried out by adopting a data expansion mode. After data alignment is carried out, missing value supplementation and abnormal value correction are carried out on the data, then a principal component analysis method is adopted to carry out dimension reduction on high-dimension input data, important information of the data is reserved while the dimension of the data is reduced, and the establishment time of a model is shortened. And S1.3.2, shifting a predicted result of the input model for training, predicting the tail end vibration of the robot in advance by 1 data point, and reducing the execution delay of the control system. And S1.3.3, dividing the finished data into four types according to four operation modes of industrial silicon, calcium carbide, ferrosilicon and a silicomanganese furnace, dividing a training set, a testing set and a verification set, performing supervised learning by XGBoost, and respectively establishing a robot tail end vibration prediction model.
4. The method of claim 1, wherein the moving process in step S2 is a process of accelerating the impact robot to approach the impact surface, and the vibration in three directions is mainly coarse-tuned. The impact process relates to reaming of a centering open hole and repeated drill rod of a robot, and the vibration in the front-rear direction is accurately controlled and regulated.
5. The method for suppressing the end vibration of a high impact load working robot according to claim 1, wherein the step S3 is implemented as the steps of: And S3.1, reversely compensating vibration of the tail end in three directions in the running process of the robot by adopting an MPC mode according to the output of the tail end vibration prediction model of the robot, and inhibiting nonlinear discontinuous vibration disturbance caused by collision between the tail end of the robot and the furnace wall in the impact process by using ADRC. And S3.2, designing corresponding control strategies according to different driving modes of the high impact load working robot, and compensating vibration offset into the executing mechanisms of the robot for moving, impacting, pitching and rotating so as to inhibit vibration. The compensation mode calculates vibration to be suppressed through the secondary server and transmits the vibration to the PLC control unit, and the compensation value is used as an additional input to compensate the program variable of the corresponding degree of freedom of the PLC responsible control. And S3.3, detecting the vibration of the tail end of the robot again after the vibration compensation, compensating the error of the prediction model and the execution error of the control system by using self-learning, calculating the error of three degrees of freedom vibration compensation according to the detected vibration of the tail end of the robot after the vibration compensation, and correcting a compensation curve at the moment corresponding to the next impact, wherein the correction formula is as follows: Wherein N x_new 、N y_new and N z_new are vibration compensation values of three degrees of freedom after self-learning optimization, N x_old 、N y_old and N z_old are vibration compensation values of three degrees of freedom before optimization, and N x_dev 、N y_dev and N z_dev are errors of three degrees of freedom vibration compensation.
6. The method for suppressing vibration of a tip of a high impact load working robot according to claim 5, wherein the self-learning method adjusts a vibration suppression curve in an experiment. In order to prevent the fluctuation of the transmission system from being too large and affecting the stable operation of the robot due to direct compensation of all the differences, a compensation coefficient n is set in a formula. By adopting a progressive compensation strategy, the coarse adjustment stage n takes 15%, the fine adjustment stage n takes 25%, and the proportion can be adjusted according to actual production. In actual operation, the vibration precision of the visual measurement is affected by environmental factors, the compensation coefficient of the formula (2) is reduced to 10%, and the influence of measurement errors on a control system is reduced.
7. The method for suppressing the tail end vibration of a high impact load working robot according to claim 1, wherein the step S4 is to record a compensation curve for suppressing the impact vibration at present, and reversely compensate the tail end vibration of the impact robot in proportion for a plurality of times, and record the vibration compensation curve in the working scene data of the corresponding model when the envelope circle of the vibration in three directions is smaller than Δm, and directly call the vibration suppression in the subsequent production working process.

Description

Terminal vibration suppression method for strong impact load operation robot Technical Field The invention belongs to the technical field of automatic control and robots, and particularly relates to a method for inhibiting terminal vibration of a robot working under a strong impact load. Background Aiming at the problem that a stokehole operation robot of an ore-smelting furnace is unstable in impact under a strong impact load, the impact end of the robot in the operation process collides with a furnace wall to generate nonlinear discontinuous vibration, the precision of the hole opening and the hole pulling of the stokehole operation robot is affected, and the furnace hole is damaged when serious. The method for carrying out corresponding reverse compensation on the terminal vibration is an effective measure for solving the problem, in the existing compensation method, most of vibration models are pure mathematical models, the accuracy of calculation is not high, and the stability of a system is affected when the reverse compensation is carried out. Lin Xianzhong, filed by the applicant of the invention, a predictive control method for torque rate control and vibration suppression is provided, torque and torque rate in robot motion are controlled by an MPC-based method, vibration of a robot is effectively suppressed, accuracy and stability of industrial robot operation are improved, chunyang, filed by the applicant of the invention, a bidirectional buck-boost converter control method based on a reduced-order active immunity strategy is provided, a bidirectional buck-boost converter control method based on an ADRC+MPC double-closed-loop structure is provided, an ADRC is adopted to obtain a current loop reference value, PWM duty ratio voltage regulation is controlled by MPC, dynamic response capacity of the bidirectional converter is improved, zhangbin, the model predictive control method based on a neural network is provided, weight coefficients of a permanent magnet synchronous motor are predicted by the neural network, multiple indexes such as electromagnetic torque, stator current and system loss are considered, automatic motor regulation performance under different working conditions is realized, and control accuracy and efficiency under dynamic and steady state are ensured. Therefore, in the end vibration suppression compensation of the robot working under the strong impact load, a neural network is adopted to establish a model for predicting the end vibration of the robot under the current working condition according to the historical data of the robot in the impact operation experiment process. In each control period, the MPC method is adopted to reversely compensate the vibration of the front and back, left and right and up and down three degrees of freedom of the tail end of the robot. Meanwhile, the ADRC is introduced to inhibit nonlinear discontinuous vibration disturbance caused by collision between the tail end of the robot and the furnace wall in the impact process, so that active closed-loop vibration inhibition control is realized. The tail end vibration of the robot after compensation is detected, a self-learning method is used for compensating the system error, the difference value is compensated in the next corresponding impact process, the compensation value of the tail end vibration can be more accurately compensated at the corresponding moment, the problem of nonlinear discontinuous vibration of the tail end of the impact load operation robot is effectively solved, and the impact operation process of the stokehold operation robot is more stable. Disclosure of Invention Aiming at the problems, the invention provides a method for restraining the terminal vibration of a strong impact load operation robot, which solves the problems of nonlinear and discontinuous vibration of the terminal during operation of the existing impact load operation robot, and combines a neural network, model predictive control and a self-learning system to predict and reversely compensate the terminal vibration according to working conditions so as to reduce the terminal vibration during impact. In order to solve the above problems, embodiments of the present invention provide the following solutions: A method for suppressing the vibration of tail end of robot with strong impact load is used for the robot with strong impact and open hole operation in front of metallurgical furnace. And (3) through establishing a prediction model of the terminal vibration of the robot under the current working condition, suppressing the terminal vibration of the robot by adopting MPC+ADRC. And collecting relevant data of a transmission mechanism of the robot in the impact operation experiment process, measuring corresponding terminal vibration by a parallel resultant force-visual sensor as input, and establishing a robot terminal vibration prediction model. And in a set control period, the vibration of the three degrees of freedom