Error compensation method and system for an industrial robot

Abstract

The present application belongs to the technical field of industrial robot control, and provides an error compensation method and system for an industrial robot, which comprises the following steps: collecting a composite sensing signal of the industrial robot and a pre-tightening force, and constructing a contribution matrix; extracting an error transmission link, constructing a space-time correlation function based on the link, and determining a pre-tightening force attenuation dominant area through a dynamic cloud chart algorithm; extracting gradient change characteristics and attenuation trend coefficients of the attenuation dominant area, screening a to-be-compensated interval, and judging a risk of saturation compensation of an actuator; if the risk of saturation exists, screening a pre-tightening force sensitive joint to construct a compensation distribution model, outputting a segmented compensation demand, and simultaneously dynamically calibrating the segmented compensation amount based on a real-time compensation result. The system comprises a contribution analysis module, an attenuation positioning module, a saturation analysis module, and a compensation calibration module. The present application is beneficial to positioning an error source dominated by pre-tightening force attenuation, and is beneficial to improving compensation accuracy and stability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial robot control, and particularly relates to an error compensation method and system of an industrial robot. BACKGROUND

[0002] In high-precision operation scenes such as welding and assembly, the end positioning accuracy of an industrial robot directly affects the product processing quality, and joint pre-tightening force decay is the key to causing positioning error accumulation. With the increase of the running time of the industrial robot, factors such as impact of rolling elements of joint bearings, temperature fluctuation, and contact stress change will gradually cause pre-tightening force decay, which is then transmitted and amplified through a joint kinematic chain, and finally causes the end positioning accuracy to deviate from the allowable threshold, affecting the operation effect.

[0003] In the prior art, when the error region corresponding to the pre-tightening force decay is determined, the prior art divides the region based on static error values, does not construct an error transmission model in combination with the joint topological structure of the robot (such as the error transmission characteristics of a serial joint), cannot present the spatiotemporal distribution law of the pre-tightening force decay in the work space, and is prone to cause blind selection of compensation objects and overcompensation or undercompensation.

[0004] Therefore, the application provides an error compensation method and system of an industrial robot. SUMMARY

[0005] In order to make up for the deficiencies of the prior art and solve at least one technical problem proposed in the background.

[0006] The technical scheme adopted by the application to solve the technical problems is: an error compensation method and system of an industrial robot, comprising:

[0007] An error compensation method of an industrial robot: comprising the following steps:

[0008] Collecting a composite sensing signal and a pre-tightening force of the industrial robot, extracting time domain parameters of the composite sensing signal and performing correlation analysis in combination with the pre-tightening force to obtain independent interference parameters, performing weight contribution quantization processing on the independent interference parameters and constructing a contribution degree matrix;

[0009] Based on the contribution degree matrix, an error transmission model is established in combination with the joint topological structure of the robot to extract error transmission links, a spatiotemporal correlation function is constructed based on the error transmission links, and a decay dominant region of the pre-tightening force is determined by inputting the spatiotemporal correlation function in combination with a dynamic cloud chart algorithm;

[0010] Extracting gradient change characteristics and decay trend coefficients of the decay dominant region, performing interval screening according to the gradient change characteristics to obtain a to-be-compensated interval, and performing saturation compensation analysis on the decay trend coefficients to determine whether the actuator of the industrial robot has a saturation compensation risk;

[0011] If there is a risk of actuator saturation compensation, the pre-tightening force sensitive joint in the error transmission link is extracted, a compensation distribution model and a compensation distribution dataset are constructed, the compensation distribution dataset is input into the compensation distribution model to obtain the segmented compensation demand of the pre-tightening force sensitive joint, the real-time segmented compensation result is obtained, and the segmented compensation amount is dynamically calibrated.

[0012] Further, the weight contribution quantification processing is performed in the following manner:

[0013] The numerical values of the independent disturbance parameters under each pre-tightening force position of each working condition combination are collected, and the decay measured values corresponding to each pre-tightening force position are collected.

[0014] Based on the decay measured values and the numerical values of the independent disturbance parameters under each working condition combination, a parameter decay dataset is constructed.

[0015] A multiple linear regression model is constructed, and the weights of the independent disturbance parameters are solved.

[0016] The determination coefficient of the multiple regression model is calculated based on the weights, and whether the weights are reliable is determined based on the determination coefficient. If the weights are reliable, the weights are retained, otherwise the weights under the corresponding working condition are discarded.

[0017] The contribution degrees of each independent disturbance parameter are obtained by normalizing the reliable weights of all independent disturbance parameters under the same working condition.

[0018] Further, the independent disturbance parameters are obtained in the following manner:

[0019] The time domain parameters of each signal in the sliding window in the composite sensing signal are extracted, and a time domain parameter set is constructed.

[0020] The variation coefficients of the same time domain parameter of each signal in the time domain parameter set and the pre-tightening force of different positions are calculated by setting the pre-tightening force of the industrial robot bearing under the same working condition and different positions.

[0021] The time domain parameters of each signal are screened based on the variation coefficients, and the signals with strong correlation are extracted.

[0022] A blind source separation model is constructed, the signals with strong correlation are input into the blind source separation model, and the independent disturbance parameters are obtained.

[0023] Further, the error transmission model is constructed in the following manner:

[0024] The mechanical physical quantities corresponding to the independent disturbance parameters are extracted, and a mapping table of independent disturbance parameters-mechanical physical quantities is established.

[0025] The joint topology of the robot is obtained, and the mapping table of independent disturbance parameters-mechanical physical quantities is combined to extract an error transmission path diagram.

[0026] Based on the contribution degree matrix, the amplification determination coefficient of each joint for each independent interference parameter in the error transmission path graph is calculated, whether the independent interference parameter of the joint exists amplification effect is judged based on the amplification determination coefficient, and if exists, the error transmission link guided by the pre-tightening force attenuation is extracted.

[0027] Further, the way of extracting the error transmission link guided by the pre-tightening force attenuation is:

[0028] The current working condition and the topological amplification coefficient of each joint of the industrial robot are obtained, and if the independent interference parameter of the current working condition is the temperature drift parameter and the vibration harmonic parameter;

[0029] The contribution degree of the temperature drift parameter and the contribution degree of the vibration harmonic parameter of each joint are multiplied by the topological amplification coefficient respectively to obtain the amplification determination coefficient of each independent interference parameter in each joint;

[0030] If it is determined that the independent interference parameter of the joint exists amplification effect, the mechanical physical quantity corresponding to the dominant independent interference parameter of the joint is extracted;

[0031] The joint nodes without amplification effect in the error transmission path graph are removed, the error transmission link of the pre-tightening force attenuation is extracted in combination with the mechanical physical quantity, and the construction of the error transmission model is realized.

[0032] Further, the way of constructing the space-time correlation function is:

[0033] The contribution degree of the dominant independent interference parameter of the joint node, the attenuation cumulative amount of the joint node, and the topological amplification coefficient of the joint node are obtained;

[0034] The end error accumulation equation is constructed, the contribution degree of the dominant independent interference parameter of the joint node, the attenuation cumulative amount of the joint node, and the topological amplification coefficient of the joint node are input into the end error accumulation equation, and the end error accumulation amount of the error transmission link is obtained;

[0035] The topological correction coefficient is obtained, the topological correction coefficient and the end error accumulation amount are multiplied, and the construction of the space-time correlation function is realized.

[0036] Further, the way of judging whether there is a saturation compensation risk is:

[0037] The difference between the maximum positioning error value and the maximum positioning error threshold value in the core interval coordinate range during the real-time operation of the industrial robot is obtained to obtain the compensation demand;

[0038] Based on the dynamics model of the industrial robot, a joint torque and compensation amount equation is constructed to obtain a compensation torque;

[0039] Obtaining the transmission ratio of the joint reducer of the industrial robot, and performing ratio processing on the compensation demand and the transmission ratio to obtain a compensation rotation angle;

[0040] Taking the compensation torque and the compensation rotation angle as criterion factors of the industrial robot;

[0041] Saturated risk criterion is constructed, if the criterion factors meet the saturated risk criterion, it is determined that the industrial robot has an actuator saturation compensation risk.

[0042] Further, the determination method of the core interval coordinate range is:

[0043] Obtaining the key operation area of the industrial robot, calculating the included angle between the current coordinate point of the industrial robot and the key operation area, and taking the included angle as the operation motion angle;

[0044] Calculating the absolute deviation of the gradient direction angle and the operation motion angle, and determining whether the gradient direction points to the key operation area based on the absolute motion deviation;

[0045] Obtaining the gradient change rate and the gradient direction of each coordinate point in the attenuation dominant area of the pre-tightening force, and performing priority extraction to obtain a first priority point;

[0046] The first priority point and the adjacent first priority point are aggregated into a continuous coordinate range, and the coordinate range is the first priority interval;

[0047] All the first priority intervals are taken as the to-be-compensated intervals, and the core interval coordinate range of the to-be-compensated intervals is output.

[0048] Further, the method for obtaining the segmented compensation demand is:

[0049] Obtaining the amplification determination coefficient of the dominant independent interference parameter of each joint of the error transmission link, performing descending order sorting based on the amplification determination coefficient, and selecting the joint with the highest amplification determination coefficient as the pre-tightening force sensitive joint;

[0050] Obtaining the total compensation demand of the pre-tightening force sensitive joint, the spatial gradient change rate, and combining the compensation allocation data of the historical pre-tightening force sensitive joint;

[0051] The compensation allocation data and the total compensation demand and the spatial gradient change rate of the current pre-tightening force sensitive joint are constructed into a compensation allocation data set;

[0052] A compensation allocation model is constructed through a random forest regression algorithm, the compensation allocation data set is input into the compensation allocation model, and the model outputs the segmented compensation demand of the pre-tightening force sensitive joint.

[0053] An error compensation system of an industrial robot: comprising the following modules:

[0054] The contribution analysis module is used for collecting the composite sensing signals and the pretightening force of the industrial robot, extracting time domain parameters of the composite sensing signals, and performing correlation analysis on the time domain parameters in combination with the pretightening force to obtain independent interference parameters, performing weight contribution quantization processing on the independent interference parameters, and constructing a contribution matrix.

[0055] The attenuation positioning module is used for establishing an error transmission model in combination with a robot joint topological structure to extract an error transmission link, constructing a space-time correlation function based on the error transmission link, and determining a pretightening force attenuation dominant area by inputting the space-time correlation function into a dynamic cloud chart algorithm.

[0056] The saturation analysis module is used for extracting gradient change characteristics and attenuation trend coefficients of the attenuation dominant area, performing interval screening on the gradient change characteristics to obtain a to-be-compensated interval, and performing saturation compensation analysis on the attenuation trend coefficients to determine whether an actuator of the industrial robot has a saturation compensation risk.

[0057] The compensation calibration module is used for extracting a pretightening force sensitive joint in the error transmission link, constructing a compensation distribution model and a compensation distribution data set, inputting the compensation distribution data set into the compensation distribution model to obtain a segmented compensation demand amount of the pretightening force sensitive joint, obtaining a real-time segmented compensation result, and dynamically calibrating the segmented compensation amount if the actuator has the saturation compensation risk.

[0058] The industrial robot pretightening force attenuation analysis method has the following beneficial effects:

[0059] 1. By collecting composite sensing signals such as mechanical vibration, temperature change, and micro-strain of the joints of the industrial robot, performing correlation analysis on the composite sensing signals in combination with the pretightening force to extract independent interference parameters, and constructing a contribution matrix through multivariate linear regression and normalization processing, key factors that are strongly related to the attenuation of the pretightening force can be identified, the influence weight of different interference parameters on the attenuation of the pretightening force can be quantized, basic data support for subsequent positioning error sources can be provided, and the attenuation analysis requirements of the pretightening force of the industrial robot under different load and speed conditions can be met.

[0060] 2. By constructing an error transmission model based on the contribution matrix and the topological structure of the joints of the robot, extracting an error transmission link that guides the attenuation of the pretightening force, and determining an attenuation dominant area by a space-time correlation function in combination with a dynamic cloud chart algorithm, the mechanical physical quantities corresponding to the independent interference parameters can be combined with the error transmission characteristics of the joint kinematic chain, the space-time distribution law of the attenuation of the pretightening force in the workspace can be reflected, the identification of the attenuation area can be more in line with the actual motion logic of the industrial robot, and the error transmission analysis scene of the common industrial robot structure with a series topological structure can be met.

[0061] 3、Extract the gradient change feature and the attenuation trend coefficient of the attenuation dominant area, screen the interval to be compensated according to the gradient change feature, and perform saturation compensation analysis on the attenuation trend coefficient to judge the risk of the actuator, which is beneficial to analyze the compensation object with significant error change and high correlation with the key operation area, and to evaluate the load impact of the compensation action on the actuator in advance, providing priority basis and safety reference for the development of subsequent compensation strategy, and reducing resource waste or excessive load of the mechanism caused by blind compensation.

[0062] 4、When there is a saturation risk, screen the pretension sensitive joint, construct a compensation allocation model to output the segmented compensation demand amount based on the random forest regression algorithm, and then calculate the effect coefficient based on the real-time compensation result to dynamically calibrate the compensation amount, which can concentrate the compensation resources on the joints with significant impact on the pretension attenuation error, generate a segmented scheme that adapts to the carrying capacity of the actuator, and adjust the subsequent action according to the actual compensation effect, so that the compensation process is more in line with the actual capacity of the actuator and the error change, which helps to improve the stability of the compensation precision and reduce the wear of the actuator. BRIEF DESCRIPTION OF DRAWINGS

[0063] The application will be further described below with reference to the drawings.

[0064] Figure 1 is a flowchart of an error compensation method of an industrial robot according to an embodiment of the application;

[0065] Figure 2 is a flowchart of a determination of a strong correlation between a time domain parameter and a pretension attenuation;

[0066] Figure 3 is a module architecture diagram of an error compensation system of an industrial robot according to an embodiment of the application. DETAILED DESCRIPTION

[0067] In order to make the technical means, creative features, purposes and effects of the application easy to understand, the application will be further described below with reference to the specific embodiments.

[0068] Embodiment 1

[0069] Please refer to Figure 1 The error compensation method of an industrial robot according to an embodiment of the application includes the following steps:

[0070] S1, collect the composite sensing signal and the pretension force of the industrial robot, extract the time domain parameters of the composite sensing signal and perform correlation analysis with the pretension force to obtain independent interference parameters, perform weight contribution quantization processing on the independent interference parameters and construct a contribution matrix.

[0071] The composite sensing signal of the industrial robot and the pre-tightening force are collected, and the time domain parameters of the composite sensing signal are extracted and analyzed in association with the pre-tightening force.

[0072] Preferably, the piezoelectric acceleration sensor arranged on the joint bearing seat of the industrial robot is used to collect the mechanical vibration signal reflecting the contact impact of the rolling element and the inner and outer rings.

[0073] The temperature values of different sampling points in the joint area are collected by the infrared array sensor, and the temperature change signal of each sampling point is extracted to reflect the change of the thermal pre-tightening force.

[0074] The micro-strain signal (i.e. contact strain force) reflecting the contact pressure fluctuation is collected by the strain gauge pasted on the bearing outer ring of the joint area of the industrial robot.

[0075] The mechanical vibration signal, the temperature change signal and the micro-strain signal are collectively referred to as the composite sensing signal.

[0076] The time domain parameters of each signal in the composite sensing signal in the sliding window are extracted to construct a time domain parameter set.

[0077] The variation coefficients of the same time domain parameter of each signal in the time domain parameter set and the pre-tightening force of different gears under the same working condition are calculated.

[0078] The time domain parameters of each signal include the peak value, the root mean square value and the kurtosis of the mechanical vibration signal.

[0079] The temperature change rate and the temperature fluctuation variance of the temperature change signal.

[0080] The strain peak value and the strain mean value of the micro-strain signal.

[0081] If the variation coefficient of at least one time domain parameter of each signal is higher than or equal to the preset decay correlation threshold value, and the variation coefficient increases with the decay of the pre-tightening force, it is determined that there is a strong correlation between the time domain parameter and the decay of the pre-tightening force.

[0082] The preset decay correlation threshold value is one-to-one corresponding to the time domain parameter of each signal, i.e. the variation coefficient of the pre-tightening force of one time domain parameter has a corresponding decay correlation threshold value.

[0083] As shown in Figure 2 If the variation coefficient of the time domain parameter is lower than the preset decay correlation threshold value, or shows a decreasing trend with the decay of the pre-tightening force, it is not processed.

[0084] It can be understood that the attenuation correlation threshold is preset by a person skilled in the art based on historical test data of the industrial robot under different load ratios and speed ratio conditions, and the minimum coefficient of variation when the time domain parameter is strongly correlated with the pre-tightening force attenuation in the data is taken as the attenuation correlation threshold to adapt to the judgment needs of the strong correlation relationship under actual conditions.

[0085] If the time domain parameter of each signal is strongly correlated with the pre-tightening force attenuation, the signal with the strong correlation relationship is extracted.

[0086] A blind source separation model is constructed, and the signal with the strong correlation relationship is input into the blind source separation model to obtain independent interference parameters.

[0087] For example, if the mechanical vibration signal, the temperature change signal and the micro-strain signal are all strongly correlated with the pre-tightening force attenuation, the corresponding independent interference parameters are: temperature drift parameters, vibration harmonic parameters and micro-strain pulse parameters.

[0088] It should be noted that the blind source separation model is constructed by selecting the FastICA adaptive blind source separation algorithm, and the mechanical vibration, temperature change and micro-strain signals screened out and strongly correlated with the pre-tightening force attenuation are input into the model to separate the independent components corresponding to the pre-tightening force attenuation in the mixed signal, i.e., to obtain the temperature drift, vibration harmonic and micro-strain pulse independent interference parameters.

[0089] The temperature drift parameter is used to reflect the pre-tightening force offset caused by temperature changes; the vibration harmonic parameter is used to reflect the pre-tightening force fluctuation amplitude caused by periodic impacts of the rolling body; and the micro-strain pulse parameter is used to represent the pre-tightening force pulse change caused by instantaneous contact stress mutation.

[0090] The contribution of the independent interference parameters is quantified and a contribution matrix is constructed in the following manner:

[0091] Based on the actual operation scene of the industrial robot, N kinds of working condition combinations of different load ratios and speed ratios are designed, and K kinds of pre-tightening force positions are set for each working condition combination.

[0092] It should be noted that the load ratio is the ratio of the actual load of the working condition to the rated load, and the speed ratio is the ratio of the actual speed of the working condition to the rated speed.

[0093] The values of the independent interference parameters under each pre-tightening force position of each working condition combination are collected, and the measured values of the attenuation corresponding to each pre-tightening force position are collected. ;

[0094] Based on the measured values of the attenuation and the values of the independent interference parameters under each working condition combination, a parameter attenuation data set is constructed.

[0095] Through the formula: A multiple linear regression model is constructed to solve the weight w of the independent interference parameter;

[0096] wherein, , , is each independent interference parameter, for example is a temperature drift parameter, is a vibration harmonic parameter, is the nth independent interference parameter;

[0097] , , is the weight to be solved corresponding to each independent interference parameter, is an error term of the multiple linear regression model;

[0098] The parameter attenuation data set is input into the multiple linear regression model, and the weight to be solved under each group of working conditions is solved by the least square method;

[0099] The determination coefficient of the multiple regression model is calculated based on the weight to be solved, and whether the weight is reliable is judged based on the determination coefficient, if reliable, the weight is retained, otherwise the weight under the corresponding working condition is discarded;

[0100] For example, the adjusted determination coefficient is calculated , it is considered that the weight is reliable;

[0101] The contribution degree of each independent interference parameter is obtained based on the reliable weight of all independent interference parameters under the same working condition;

[0102] Based on the contribution degree of each independent interference parameter under each group of working conditions, a contribution degree matrix is constructed.

[0103] S2, based on the contribution degree matrix, an error transmission model is established to extract an error transmission link combined with the robot joint topology structure, a space-time correlation function is constructed based on the error transmission link, and the attenuation dominant area of the pre-tightening force is determined by inputting the space-time correlation function combined with the dynamic cloud algorithm;

[0104] wherein, the error transmission model is constructed in the following way:

[0105] S201, the mechanical physical quantity corresponding to the independent interference parameter is extracted, and a mapping table of independent interference parameter-mechanical physical quantity is established;

[0106] Preferably, the mechanical physical quantity of the independent interference parameter is associated and processed to establish the mapping table of independent interference parameter-mechanical physical quantity;

[0107] The skilled in the art can understand that, if the independent interference parameter is any one or more of the temperature drift parameter, the vibration harmonic parameter and the micro-strain pulse parameter;

[0108] The temperature drift corresponds to the thermal deformation of the joint of the industrial robot, the thermal deformation of the joint causes the joint gap to increase, and the mechanical physical quantity associated with the temperature drift parameter is the joint gap value;

[0109] The vibration harmonic corresponds to the impact of the rolling element, the impact of the rolling element causes the change of the bearing stiffness, and the mechanical physical quantity associated with the vibration harmonic parameter is the change amount of the bearing stiffness;

[0110] The micro-strain pulse corresponds to the fluctuation of the contact stress, the fluctuation of the contact stress causes the fluctuation of the stiffness, and the mechanical physical quantity associated with the micro-strain pulse parameter is the fluctuation amount of the stiffness;

[0111] S202, acquire the joint topology of the robot, and extract the error transmission path graph combined with the mapping table of the independent interference parameter-mechanical physical quantity;

[0112] Preferably, the topology type of the industrial robot is acquired, and the error transmission path is extracted according to the motion sequence of the industrial robot;

[0113] Illustratively, if the industrial robot is a series topology, the error transmission path is acquired according to the joint motion sequence, i.e. joint 1 to joint 2 to the final end effector, the error of the front-stage joint is transmitted to the rear-stage joint through the force-displacement coupling relationship, and the end execution error is amplified;

[0114] Mark the joint node of the mechanical physical quantity corresponding to the independent interference parameter in the error transmission path, and construct the error transmission path graph;

[0115] S203, calculate the amplification determination coefficient of each independent interference parameter in each joint in the error transmission path graph based on the contribution matrix, judge whether the independent interference parameter of the joint exists amplification effect based on the amplification determination coefficient, and if exists, extract the error transmission link guided by the pre-tightening force attenuation;

[0116] Preferably, the current working condition and the topology amplification coefficient of each joint of the industrial robot are acquired, and if the current working condition of the independent interference parameter is the temperature drift parameter and the vibration harmonic parameter;

[0117] The contribution of the temperature drift parameter and the contribution of the vibration harmonic parameter of each joint are multiplied by the topology amplification coefficient respectively to obtain the amplification determination coefficient of each independent interference parameter in each joint;

[0118] If the amplification determination coefficient of the independent interference parameter is higher than or equal to the preset amplification determination threshold, it is determined that the independent interference parameter of the joint exists amplification effect;

[0119] It can be understood that the amplification determination threshold is preset by the skilled person in the art based on the historical error transmission test data of the industrial robot under typical load, rotational speed working conditions, and the minimum contribution of the joint independent disturbance parameter (such as temperature drift, vibration harmonic) causing significant error amplification and the product value of the topological amplification coefficient, which will screen out the joints that significantly affect the end error;

[0120] If the amplification determination coefficient of the independent disturbance parameter is lower than the preset amplification determination threshold, it is determined that the amplification effect of the independent disturbance parameter of the joint is not within the expectation;

[0121] If the independent disturbance parameter of the joint has an amplification effect, the mechanical physical quantity corresponding to the dominant independent disturbance parameter of the joint is extracted;

[0122] It should be noted that if the amplification determination coefficient of the temperature drift parameter is higher than that of the vibration harmonic parameter, the mechanical physical quantity corresponding to the temperature drift parameter is extracted. Since the higher the amplification determination coefficient, the more significant the amplification effect of the mechanical physical quantity corresponding to the parameter on the end error, the mechanical physical quantity corresponding to the independent disturbance parameter with a higher amplification determination coefficient is extracted first;

[0123] After removing the joint nodes with amplification effects that are not within the expectation in the error transmission path graph (i.e., joint nodes without amplification effects), the error transmission link with pre-tightening force attenuation is extracted in combination with the mechanical physical quantity, and the construction of the error transmission model is realized;

[0124] It should be noted that the way to extract the error transmission link with pre-tightening force attenuation is as follows: according to the logic of pre-tightening force attenuation-existence of mechanical physical quantity with amplification effect-error transmission, the error transmission link is reconstructed as follows: pre-tightening force attenuation-increase of joint 1 gap-decrease of joint 2 stiffness-error accumulation along the path-end positioning error; The contribution ratio of the dominant independent disturbance parameter corresponding to the mechanical physical quantity of each joint node and the mechanical physical quantity is obtained from the error transmission link (extracted from the contribution matrix, such as joint 1 gap contribution ratio 35%, joint 2 stiffness attenuation ratio 45%);

[0125] Among them, the construction of the space-time correlation function is combined with the dynamic cloud chart algorithm to input the space-time correlation function to determine the pre-tightening force attenuation dominant area in the following way:

[0126] Preferably, the end error accumulation equation is as follows: The end error accumulation quantity in the error transmission link is obtained ;

[0127] Where i is the number of joint nodes in the error transmission path, is the contribution of the dominant independent disturbance parameter of the i-th joint node;

[0128] is the topology amplification coefficient of the ith joint node, and t is the running time; is the attenuation cumulative amount of the ith joint node;

[0129] It should be noted that the topology amplification coefficient is used to quantify the amplification effect of joint error along the motion chain to the end, by combining the series topology characteristics of the industrial robot (such as joint 1-joint 2-end effector structure):

[0130] Using the error injection method, a known small error (such as 0.01mm gap) is applied to the target joint, and the measured end error change is obtained, and the ratio of the two is the topology amplification coefficient of the joint (for example, joint 1 injects 0.01mm gap, and the end positioning error increases by 0.02mm, then the topology amplification coefficient is 2);

[0131] wherein the attenuation cumulative amount of the ith joint node is obtained by the formula:

[0132] is the attenuation value of the initial preload of the ith joint node, k is the preset cumulative coefficient, m is the attenuation index, and t is the running time;

[0133] It should be noted that the cumulative coefficient k and the attenuation index m are first collected industrial robot under different load ratio, speed ratio working conditions, different running time corresponding to the preload attenuation cumulative amount of measured data, and then based on the preload attenuation cumulative amount formula The collected running time and the attenuation value of the initial preload measured data are substituted into the power function model, and the k and m in the model are solved by the least square method to determine the cumulative coefficient and the attenuation index adapted to the corresponding working condition.

[0134] The end error cumulative amount and the topology correction coefficient A of the industrial robot workspace (x, y) coordinate point are obtained.

[0135] The error value of each point is calculated by the formula:

[0136] It should be noted that the topology correction coefficient A is obtained by the factory calibration data of the industrial robot.

[0137] Based on the dynamic cloud algorithm, the space-time correlation function is input into the dynamic cloud algorithm to obtain the preload attenuation dominant area.

[0138] ​​​​​Those skilled in the art will understand that after inputting the spatiotemporal correlation function into the dynamic cloud map algorithm, the algorithm first outputs the end-positioning error values ​​of all coordinate points in the industrial robot's workspace at different running times through the algorithm's analytical function. At the same time, the weight ratio of each independent interference parameter in the contribution matrix to the preload attenuation is called to calculate the proportion of the error component caused by preload attenuation in the total error of each (x,y,t) point error (i.e., the contribution ratio of preload attenuation).

[0139] The algorithm uses two preset judgment thresholds: an error value threshold (80% of the maximum allowable positioning error under working conditions to filter high error areas) and a contribution ratio threshold (e.g., 60% to ensure that the area error is mainly dominated by preload attenuation, rather than geometric errors, motor errors, or other factors). The algorithm then traverses all spatiotemporal points in the workspace, filters out coordinate points where the error value is greater than or equal to the error threshold and the contribution ratio of preload attenuation is greater than or equal to the contribution ratio threshold, and dynamically maps these points to the workspace visualization interface in a time series. The distribution of attenuation areas is presented with color gradients, ultimately forming a dynamic cloud map that changes over time. The continuously highlighted red area in the dynamic cloud map is the preload attenuation-dominated area.

[0140] Example 2

[0141] Please see Figure 1 As shown in the embodiment of the present invention, an error compensation method for an industrial robot further includes the following steps:

[0142] S3. Extract the gradient change characteristics and attenuation trend coefficient of the attenuation-dominant region, and select the interval to be compensated based on the gradient change characteristics; perform saturation compensation analysis on the attenuation trend coefficient to determine whether the actuator of the industrial robot has a risk of saturation compensation.

[0143] The method for extracting the gradient change characteristics and attenuation trend coefficient of the attenuation-dominant region is as follows:

[0144] Obtain each of the decay-dominant regions Point error value Calculate the partial derivative of the error in the x-direction partial derivatives in the y-direction ;

[0145] Through formula one: Obtain the gradient rate of change G;

[0146] Through formula two: Obtain the gradient direction angle ;

[0147] The gradient rate of change and the gradient direction angle are used as characteristics of gradient change.

[0148] For all within the attenuation-dominant region Point error value , construct an error attenuation sequence, perform exponential fitting on the error attenuation sequence, obtain a fitting equation and extract a fitting coefficient as an attenuation trend coefficient;

[0149] Preferably, by the formula: Exponential fitting is performed on the error attenuation sequence, and the fitting coefficient d is solved as the attenuation trend coefficient;

[0150] It should be noted that c is the amplitude coefficient solved in the exponential fitting process, f is the constant term solved in the exponential fitting process, e is the natural constant, and t is the running time of the industrial robot;

[0151] Among them, the way to select the to-be-compensated interval according to the gradient change characteristics is:

[0152] Obtain the coordinates of the key area of the industrial robot (such as welding points, assembly points) ;

[0153] Calculate the angle of the current coordinate point of the industrial robot pointing to the key area of the work as the work motion angle;

[0154] Calculate the absolute deviation of the gradient direction angle and the work motion angle, and determine whether the gradient direction points to the key area of the work based on the absolute motion deviation;

[0155] Preferably, the way to determine whether the gradient direction points to the key area of the work based on the absolute motion deviation is: if the absolute deviation of the gradient direction angle and the work motion angle is lower than 30 degrees, it is determined that the gradient direction points to the key area (the error increase will quickly affect the work precision); otherwise, it is deviated from the key area;

[0156] Determine whether the gradient change rate and the gradient direction of each (x, y) point in the attenuation dominant area of the pre-tightening force point to the key area of the work to determine the priority of the compensation interval and screen the to-be-compensated interval;

[0157] Preferably, the way to determine the priority of the compensation interval and screen the to-be-compensated interval is:

[0158] If the gradient change rate is higher than the preset high gradient threshold (such as 0.5 mm / mm), and the gradient direction points to the key area of the work, then the corresponding first priority point;

[0159] If the gradient change rate is higher than the preset high gradient threshold, and the gradient direction deviates from the key area of the work, only one of the two conditions is met, then the corresponding second priority point;

[0160] If the gradient change rate is lower than the preset high gradient threshold, and the gradient direction deviates from the key area of the work, then the corresponding third priority point;

[0161] The first-priority points and their adjacent first-priority points are aggregated into a continuous (x,y) coordinate range, which is the first-priority interval.

[0162] All first-priority intervals are treated as intervals to be compensated, and the core interval coordinate range of the intervals to be compensated is output.

[0163] The method for determining whether the actuator of an industrial robot has a risk of saturation compensation by performing saturation compensation analysis on the attenuation trend coefficient is as follows:

[0164] Based on the attenuation trend coefficient within the compensation interval and the maximum allowable error threshold of the industrial machine, input the fitting equation and solve for the remaining time.

[0165] By obtaining the difference between the maximum positioning error value and the maximum positioning error threshold within the core coordinate range during real-time operation of the industrial robot, the compensation requirement can be determined. ;

[0166] The compensation demand rate is obtained by comparing the compensation demand with the remaining time. ;

[0167] Based on the dynamic model of industrial robots, the equations for joint torque and compensation are constructed as follows: Obtain the compensation torque T;

[0168] Where ka is the joint force-displacement coefficient and b is the joint damping coefficient;

[0169] Understandably, the joint force-displacement coefficient ka is determined by collecting measured output torque data of industrial robot joints under different compensation amounts and fitting the linear correlation characteristics between torque and compensation amount. It is used to quantify the correspondence between compensation displacement and joint output torque. The joint damping coefficient b is obtained by collecting damping torque data of joints under different compensation demand rates, substituting it into the joint torque and compensation amount equations, and then fitting and solving it using the least squares method. It is used to characterize the degree of influence of joint motion rate on damping torque.

[0170] Obtain the transmission ratio of the industrial robot joint reducer, and process the ratio of the compensation requirement to the transmission ratio to obtain the compensation angle.

[0171] Compensation torque and compensation angle are used as criteria factors for industrial robots;

[0172] A saturation risk criterion is constructed. If the criterion factor satisfies the saturation risk criterion, it is determined that the industrial robot has a risk of saturation compensation of the actuator.

[0173] Preferably, the method for constructing the saturation risk criterion is as follows:

[0174] The rated physical limit parameters of the actuator of the industrial robot, such as the maximum torque of the motor, the maximum rotation angle of the reducer, and the maximum compensation rate of the system, are reserved with a safety margin (the safety margin is usually 20% of the corresponding parameter to avoid damage to the mechanism in extreme working conditions), and a safety threshold, that is, a saturation risk criterion, is obtained;

[0175] If the compensation demand rate If any one of the compensation torque and the compensation rotation angle is higher than or equal to the corresponding preset safety threshold, it is determined that there is a saturation compensation risk of the actuator;

[0176] If the compensation demand rate If the compensation torque and the compensation rotation angle are all lower than the corresponding preset safety threshold, there is no saturation compensation risk of the actuator at the moment.

[0177] S4, if there is a saturation compensation risk of the actuator, a pre-tightness sensitive joint in the error transmission link is extracted, a compensation distribution model and a compensation distribution data set are constructed, the compensation distribution data set is input into the compensation distribution model to obtain a segmented compensation demand amount of the pre-tightness sensitive joint, a real-time segmented compensation result is obtained, and the segmented compensation amount is dynamically calibrated;

[0178] If there is a saturation compensation risk of the actuator, a pre-tightness sensitive joint in the error transmission link is extracted, a compensation distribution model and a compensation distribution data set are constructed, and the compensation distribution data set is input into the compensation distribution model to obtain a segmented compensation demand amount of the pre-tightness sensitive joint, a real-time segmented compensation result is obtained, and the segmented compensation amount is dynamically calibrated;

[0179] An amplification determination coefficient of a dominant independent interference parameter of each joint in the error transmission link is obtained, the amplification determination coefficients are sorted in descending order based on the amplification determination coefficients, and the joint with the highest amplification determination coefficient is selected as the pre-tightness sensitive joint;

[0180] The total compensation demand amount of the pre-tightness sensitive joint, the spatial gradient change rate, and the compensation distribution data of the historical pre-tightness sensitive joint are obtained;

[0181] The compensation distribution data includes the mapping relationship between the compensation demand amount, the spatial gradient change rate, and the safety split scheme of the pre-tightness sensitive joint;

[0182] The compensation distribution data set is constructed by the compensation distribution data and the total compensation demand amount and the spatial gradient change rate of the current pre-tightness sensitive joint;

[0183] The compensation distribution model is constructed by a random forest regression algorithm, the compensation distribution data set is input into the compensation distribution model, and the model outputs the segmented compensation demand amount of the pre-tightness sensitive joint;

[0184] The skilled in the art can understand that the way to build the compensation allocation model is to preprocess the compensation allocation dataset first, eliminate abnormal compensation records in historical data caused by sensor abnormalities and actuator failures (such as data with single compensation amount exceeding the maximum load capacity of the actuator), and then normalize the total compensation demand and the spatial gradient change rate;

[0185] The total compensation demand and the spatial gradient change rate of the current preload sensitive joint are used as the input label of the model, and the segmented compensation demand in the historical safe split scheme is output. Then, the random forest regression model is initialized, the number of decision trees is set to 100, the maximum tree depth is set to 8, and the parameters are optimized through 5-fold cross-validation to reduce overfitting.

[0186] The preprocessed compensation allocation dataset is divided into training set and test set in the ratio of 7:3. The training set is used to train the model, and the test set is used to verify the performance of the model. After training, the total compensation demand and the spatial gradient change rate of the current preload sensitive joint are input into the model. The model outputs the segmented compensation demand that meets the condition of single segmented compensation amount ≤ actuator dynamic safety threshold, such as 80% of the maximum torque of the motor and 80% of the maximum angle of the reducer, and the sum of the segmented compensation amounts is equal to the total compensation demand.

[0187] The way to obtain real-time segmented compensation results and dynamically calibrate the segmented compensation amount is as follows:

[0188] Based on the segmented compensation demand of the preload sensitive joint, the segmented compensation of the preload joint is performed.

[0189] The compound sensing signal after each segmented compensation is obtained, and the actual execution value of the segmented compensation amount after execution is extracted , and the error residual of the sensitive joint after execution ;

[0190] The compensation effect coefficient of the compensation is obtained by the formula: ;

[0191] It should be noted that, is the preset output value of the compensation allocation model, The actual compensation displacement is obtained by multiplying the transmission ratio of the joint reducer and the actual rotation angle collected by the encoder built-in the preload sensitive joint during segmented compensation execution; is the error residual after segmented compensation, which is obtained by calculating the current end positioning error of the sensitive joint corresponding to the to-be-compensated interval (x, y) coordinate point after compensation through the space-time correlation function E(x, y, t);

[0192] ​The compensation effect is analyzed based on the compensation effect coefficient, if the compensation effect is not up to the standard, then the ratio of the compensation demand of the next theoretical section and the compensation effect coefficient is calculated to obtain the modified compensation demand of the next section, and the next stage compensation amount is adjusted based on the modified compensation demand of the next section.

[0193] Embodiment 3

[0194] Please refer to Figure 3 As shown in the figure, the error compensation system of the industrial robot comprises the following modules:

[0195] The contribution analysis module is used for collecting the composite sensing signal and the pretightening force of the industrial robot, extracting the time domain parameters of the composite sensing signal and performing correlation analysis on the pretightening force to obtain independent interference parameters, performing weight contribution quantization processing on the independent interference parameters and constructing a contribution matrix.

[0196] The attenuation positioning module is used for establishing an error transmission model in combination with the joint topology structure of the robot to extract an error transmission link, constructing a space-time correlation function based on the error transmission link, and determining a pretightening force attenuation dominant area by inputting the space-time correlation function into a dynamic cloud chart algorithm.

[0197] The saturation analysis module is used for extracting gradient change characteristics and attenuation trend coefficients of the attenuation dominant area, performing interval screening on the gradient change characteristics to obtain a to-be-compensated interval, and performing saturation compensation analysis on the attenuation trend coefficients to determine whether the actuator of the industrial robot has a saturation compensation risk.

[0198] The compensation calibration module is used for extracting the pretightening force sensitive joint in the error transmission link, constructing a compensation distribution model and a compensation distribution data set, inputting the compensation distribution data set into the compensation distribution model to obtain the segmented compensation demand of the pretightening force sensitive joint, obtaining a real-time segmented compensation result, and dynamically calibrating the segmented compensation amount.

[0199] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An error compensation method for an industrial robot, characterized by: The method comprises the following steps: Collecting the composite sensing signal and the pre-tightening force of the industrial robot, extracting the time domain parameters of the composite sensing signal, and performing correlation analysis on the time domain parameters and the pre-tightening force to obtain independent interference parameters, performing weight contribution quantization processing on the independent interference parameters, and constructing a contribution matrix; The weight contribution quantization processing is performed in the following manner: Collecting the numerical values of the independent interference parameters under each pre-tightening force position of each working condition combination, and collecting the measured values of the attenuation corresponding to each pre-tightening force position; Based on the measured values of the attenuation and the numerical values of the independent interference parameters under each working condition combination, a parameter attenuation data set is constructed; A multiple linear regression model is constructed, and the weight of the independent interference parameter is solved; Based on the weight, the determination coefficient of the multiple regression model is calculated, and based on the determination coefficient, it is determined whether the weight is reliable. If the weight is reliable, the weight is retained; otherwise, the weight under the corresponding working condition is discarded. Based on the reliable weights of all independent interference parameters under the same working condition, normalization processing is performed to obtain the contribution of each independent interference parameter. The independent interference parameters are obtained in the following manner: Extracting the time domain parameters of each signal in the sliding window in the composite sensing signal to construct a time domain parameter set; By setting the pre-tightening force of the industrial robot bearing under different positions of the same working condition, the variation coefficient of the same time domain parameter of each signal in the time domain parameter set and the pre-tightening force under different positions is calculated; Based on the variation coefficient, the time domain parameters of each signal are screened and processed to extract signals with strong correlation; A blind source separation model is constructed, the signals with strong correlation are input into the blind source separation model, and the independent interference parameters are obtained; Based on the contribution matrix, an error transmission model is established based on the robot joint topology to extract the error transmission link, a space-time correlation function is constructed based on the error transmission link, and the attenuation dominant area of the pre-tightening force is determined by inputting the space-time correlation function into the dynamic cloud chart algorithm; The gradient change feature and the attenuation trend coefficient of the attenuation dominant area are extracted, and the interval is screened according to the gradient change feature to obtain a to-be-compensated interval; Saturation compensation analysis is performed on the attenuation trend coefficient to determine whether there is a saturation compensation risk of the actuator of the industrial robot; If there is a saturation compensation risk of the actuator, the pre-tightening force sensitive joint in the error transmission link is extracted, a compensation distribution model and a compensation distribution data set are constructed, the compensation distribution data set is input into the compensation distribution model to obtain the segmented compensation demand of the pre-tightening force sensitive joint, real-time segmented compensation results are obtained, and the segmented compensation amount is dynamically calibrated.

2. The error compensation method of an industrial robot according to claim 1, characterized in that: The error transmission model is constructed in the following manner: Extracting the mechanical physical quantity corresponding to the independent interference parameter, and establishing a mapping table of the independent interference parameter-mechanical physical quantity; Obtaining the joint topology of the robot, and extracting an error transmission path diagram by combining the mapping table of the independent interference parameter-mechanical physical quantity; Based on the contribution matrix, the amplification determination coefficient of each joint and each independent interference parameter in the error transmission path diagram is calculated, whether the independent interference parameter of the joint has an amplification effect is determined based on the amplification determination coefficient, and if so, the error transmission link guided by the pre-tightening force attenuation is extracted.

3. A method of error compensation for an industrial robot according to claim 2, characterized in that: The error transmission link guided by the pre-tightening force attenuation is extracted in the following manner: Obtain the current working condition and topological amplification coefficient of each joint of the industrial robot, and if the independent interference parameters of the current working condition are temperature drift parameters and vibration harmonic parameters; Multiply the contribution degree of each joint temperature drift parameter and the contribution degree of each joint vibration harmonic parameter with the topological amplification coefficient respectively to obtain the amplification determination coefficient of each independent interference parameter in each joint; If it is determined that the independent interference parameters of the joint exist amplification effect, extract the mechanical physical quantity corresponding to the dominant independent interference parameter of the joint; Eliminate the joint nodes without amplification effect in the error transmission path diagram, combine the mechanical physical quantity to extract the error transmission link of the pre-tightening force attenuation, and realize the construction of the error transmission model.

4. The error compensation method of an industrial robot according to claim 1, characterized in that: The construction of the space-time correlation function is as follows: Obtain the contribution degree of the dominant independent interference parameter of the joint node, the attenuation cumulative amount of the joint node, and the topological amplification coefficient of the joint node; Construct the end error accumulation equation, input the contribution degree of the dominant independent interference parameter of the joint node, the attenuation cumulative amount of the joint node, and the topological amplification coefficient of the joint node into the end error accumulation equation, and obtain the end error accumulation amount of the error transmission link; Obtain the topological correction coefficient, multiply the topological correction coefficient and the end error accumulation amount, and realize the construction of the space-time correlation function.

5. The method of claim 1, wherein: The way to determine whether there is a saturation compensation risk is as follows: Obtain the difference between the maximum positioning error value and the maximum positioning error threshold value in the core interval coordinate range during the real-time operation of the industrial robot to obtain the compensation demand; Based on the dynamics model of the industrial robot, construct the joint torque and compensation amount equation to obtain the compensation torque; Obtain the transmission ratio of the joint reducer of the industrial robot, and perform ratio processing on the compensation demand and the transmission ratio to obtain the compensation rotation angle; Take the compensation torque and the compensation rotation angle as the criterion factor of the industrial robot; Construct a saturation risk criterion, and if the criterion factor meets the saturation risk criterion, it is determined that the industrial robot has an actuator saturation compensation risk.

6. The error compensation method of an industrial robot according to claim 5, characterized in that: The determination method of the core interval coordinate range is as follows: Obtain the key region of the industrial robot operation, calculate the included angle between the current coordinate point of the industrial robot and the pointing key region as the operation motion angle; Calculate the absolute deviation of the gradient direction angle and the operation motion angle, and determine whether the gradient direction points to the key operation region based on the absolute motion deviation; Obtain the gradient change rate and gradient direction of each coordinate point in the attenuation dominant region of the pre-tightening force, and perform priority extraction to obtain the first priority point; Aggregate the first priority point and the adjacent first priority point into a continuous coordinate range, and the coordinate range is the first priority interval; Take all the first priority intervals as the to-be-compensated intervals, and output the core interval coordinate range of the to-be-compensated intervals.

7. The method of claim 1, wherein: The way to obtain the segmented compensation demand is as follows: Obtain the amplification determination coefficient of the dominant independent interference parameter of each joint of the error transmission link, sort the amplification determination coefficients in descending order, select the joint with the highest amplification determination coefficient as the pre-tightening force sensitive joint; Obtain the total compensation demand of the pre-tightening force sensitive joint, the spatial gradient change rate, and combine the compensation allocation data of the historical pre-tightening force sensitive joint; Construct the compensation allocation data set by taking the compensation allocation data, the total compensation demand of the current pre-tightening force sensitive joint, and the spatial gradient change rate. A compensation allocation model is constructed by a random forest regression algorithm, and a compensation allocation dataset is input into the compensation allocation model, and the model outputs a segmented compensation demand of the pretight force sensitive joint.

8. An error compensation system for an industrial robot for implementing the method according to any one of claims 1-7, characterized by: Comprise the following modules: A contribution analysis module is used to collect the composite sensing signal and pretight force of the industrial robot, extract the time domain parameters of the composite sensing signal, and perform correlation analysis on the pretight force to obtain independent interference parameters, and perform weight contribution quantization processing on the independent interference parameters and construct a contribution matrix; An attenuation positioning module is used to establish an error transmission model based on the contribution matrix, extract error transmission links, construct a space-time correlation function based on the error transmission links, and determine the attenuation dominant area of the pretight force by inputting the space-time correlation function into a dynamic cloud chart algorithm; A saturation analysis module is used to extract the gradient change characteristics and attenuation trend coefficient of the attenuation dominant area, select the interval to be compensated according to the gradient change characteristics to obtain the interval to be compensated; The attenuation trend coefficient is subjected to saturation compensation analysis to determine whether the actuator of the industrial robot has saturation compensation risk; A compensation calibration module is used to extract the pretight force sensitive joint in the error transmission link, construct a compensation allocation model and a compensation allocation dataset, input the compensation allocation dataset into the compensation allocation model to obtain the segmented compensation demand of the pretight force sensitive joint, obtain real-time segmented compensation results, and dynamically calibrate the segmented compensation amount.

Images