Welding robot self-adaptive control system and method based on digital twinning

Abstract

The invention discloses a welding robot self-adaptive control system and method based on digital twinning, and relates to the field of welding robots. Comprising a multi-source data acquisition module which acquires real-time information by using a multi-mode sensor array; the multi-source data fusion module is used for fusing the collected information, fusing the data by adopting an improved self-adaptive Kalman filtering algorithm, introducing a nonlinear compensation item to carry out data compensation, and determining the posture of the welding gun; the digital twin model is used for establishing a simulation model by adopting a thermal-mechanical coupling field and determining output parameters of a welding gun in combination with welding requirements; and an adaptive control module. According to the method, the state equation and the observation equation of the welding gun are established, the process noise and the observation noise are introduced, control errors caused by the noise can be prevented on the basis of achieving closed-loop feedback control over the pose of the welding gun, meanwhile, a nonlinear compensation item is added, and the control precision of the pose of the welding gun is improved.

Description

Technical Field

[0001] The present invention relates to the field of welding robots, and particularly to an adaptive control system and method for a welding robot based on digital twin. Background Art

[0002] Robots are gradually replacing humans to perform tasks in some complex and harsh environments as well as high-precision machining environments. Among them, welding robots, as a type of automated robot, are also widely used.

[0003] After retrieval, a patent with the Chinese patent publication number CN117359062B discloses an intelligent welding robot and its control system, including a system end, a control end, and an execution end. The system end is used to display the operating state of the welding robot through big data devices and computer devices to obtain a real-time display of the operating state of the welding robot. The control end includes a deviation correction module. The controller is used to control the operating state of the welding robot through big data devices and computers to obtain real-time control of the operating state of the welding robot. The deviation correction module is used to monitor the welding trajectory of the welding robot in real time and perform deviation correction control.

[0004] The above patent has the following deficiencies: It only corrects the welding path. On the one hand, its deviation correction cannot comprehensively consider various errors and noises, and the accuracy is relatively low. On the other hand, it cannot control the torch parameters in combination with the environment, and even if the path accuracy is improved, the welding quality will still be low.

[0005] Therefore, the present invention proposes an adaptive control system and method for a welding robot based on digital twin. Summary of the Invention

[0006] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose an adaptive control system and method for a welding robot based on digital twin.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions:

[0008] An adaptive control system and method for a welding robot based on digital twin, including:

[0009] A multi-source data acquisition module, which uses a multi-modal sensor array to acquire real-time information;

[0010] A multi-source data fusion module, which fuses the acquired information, performs data fusion using an improved adaptive Kalman filter algorithm, and introduces a non-linear compensation term for data compensation to determine the posture of the torch;

[0011] A digital twin model, which establishes a simulation model using a thermal-mechanical coupling field and determines the output parameters of the torch in combination with welding requirements;

[0012] An adaptive control module that controls a welding robot based on the torch pose, torch parameters, actual welding requirements, and the kinematic model of the welding robot.

[0013] Preferably, the multi-modal sensor array includes a laser vision sensor, an infrared thermometer, and a six-axis force sensor.

[0014] An adaptive control method for a welding robot based on digital twin, comprising the following steps:

[0015] S1: Multi-source data acquisition, collecting real-time information through a multi-modal sensor array;

[0016] S2: Multi-source data fusion, using an improved adaptive Kalman filter algorithm to fuse the data, introducing a non-linear compensation term for data compensation, and determining the torch pose;

[0017] S3: Digital twin modeling, establishing a simulation model using a thermo-mechanical coupling field, and determining the output parameters of the torch in combination with welding requirements;

[0018] S4: Adaptive control, controlling the welding robot according to the torch pose in step S2 and the torch parameters in step S3, combined with actual welding requirements and the kinematic model of the welding robot.

[0019] Preferably, in step S2, the improved adaptive Kalman filter algorithm model is:

[0020] Where:

[0021] x k Is the torch pose state vector, which includes three-dimensional position (x, y, z) and Euler angles (α, β, γ);

[0022] μ k Is the control input quantity, which comes from the data of the robot joint motor encoder;

[0023] ω k ~N(0, Q k ), which is the process noise, Q k Is the covariance matrix, Q k =diag([0.1 2 ,0.1 2 ,0.1 2 ,0.05 2 ,0.05 2 ,0.05 2 );

[0024] z k Is the observation vector, which is formed by fusing the data collected by the multi-modal sensor array;

[0025] H k is the observation matrix, the Jacobian matrix determined according to the sensor layout;

[0026] υ k ~N(0, R k ), which is the observation noise, R k =σ 2 I 6×6 , σ = 0.05mm, I 6×6 is the 6×6 identity matrix.

[0027] Preferably: In the step S2, the formula for non-linear compensation is: where K k is the time-varying gain coefficient, through the formula λ∈(0.75, 0.85) is the empirical coefficient, |||| F represents the Frobenius norm of the matrix.

[0028] Preferably: In the step S3, the heat transfer equation of the thermo-mechanical coupling field simulation model is: where: ρ is the material density; c p is the specific heat capacity; k is the thermal conductivity; q loss is the heat loss term; q arc is the arc heat source term.

[0029] Preferably: In the step S3, the arc heat source term q arc adopts a double ellipsoid model: where U is the welding voltage, I is the welding current, η is the heat efficiency coefficient, and a, b, c are the arc distribution characteristic parameters.

[0030] Preferably: In the step S4, the control all adopts a control method based on past error compensation, which includes the following steps:

[0031] S41: Collect the target control value F i in the past control and the actually executed value F i ' of the end observed, and calculate the deviation ΔF i =F i -F i ', which represents the value of the i-th control forward from this control;

[0032] S42: Establish a weight compensation model based on time sequence to obtain the compensation deviation ΔF0 for this time;

[0033] S43: Combine the compensation deviation ΔF0, and compensate the target value F = F0 - ΔF0 for this time according to the formula, where F is the updated target control value and F0 is the target control value before update.

[0034] Preferably, in the step S42, the weight compensation model based on time sequence is as follows: where k is the weight coefficient, k i > k i+1 .

[0035] Preferably, in the step S42, where k0 is the initial weight.

[0036] The beneficial effects of the present invention are as follows:

[0037] 1. By establishing the state equation and observation equation of the welding torch, introducing process noise and observation noise, the present invention can prevent control errors caused by noise while realizing closed-loop feedback control of the welding torch pose, and at the same time adding a non-linear compensation term, thereby increasing the control accuracy of the welding torch pose.

[0038] 2. By establishing a simulation model of the thermo-mechanical coupling field, combining the actual output of the welding torch with environmental information, and using a double-ellipsoid model to establish a heat conduction equation, the present invention can predict the temperature during welding, and thus adjust the welding torch parameters in a timely manner according to the prediction, thereby improving the welding quality.

[0039] 3. For the control of each parameter, the present invention adopts a compensation control method, uses past errors to compensate the target control value, increases the control accuracy and control sensitivity, and at the same time, the past errors adopt a weighted control form, which can make the compensation value match the working state of the recent welding torch and the entire welding robot, thereby increasing the compensation accuracy. Description of the Drawings

[0040] Figure 1 is an architecture diagram of an adaptive control system for a welding robot based on digital twin proposed by the present invention;

[0041] Figure 2 is a flowchart of an adaptive control method for a welding robot based on digital twin proposed by the present invention. Detailed Embodiments

[0042] The technical solutions of the present invention will be further described in detail below in conjunction with the specific embodiments.

[0043] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installation", "connection", "connection", and "setting" should be understood in a broad sense. For example, it can be fixedly connected, set, or detachably connected, set, or integrally connected, set. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0044] Example 1:

[0045] An adaptive control system for a welding robot based on digital twin, which includes:

[0046] A multi-source data acquisition module, which uses a multi-modal sensor array to acquire real-time information;

[0047] A multi-source data fusion module, which fuses the acquired information, performs data fusion using an improved adaptive Kalman filter algorithm, introduces a non-linear compensation term for data compensation, and determines the pose of the welding torch;

[0048] A digital twin model, which establishes a simulation model using a thermal-mechanical coupling field and determines the output parameters of the welding torch in combination with welding requirements;

[0049] An adaptive control module, which controls the welding robot according to the pose of the welding torch and the welding torch parameters in combination with the actual welding requirements and the kinematic model of the welding robot.

[0050] The multi-modal sensor array includes a laser vision sensor, an infrared thermometer, and a six-axis force sensor.

[0051] Embodiment 2:

[0052] An adaptive control method for a welding robot based on digital twin, which includes the following steps:

[0053] S1: Multi-source data acquisition, acquiring real-time information through a multi-modal sensor array;

[0054] S2: Multi-source data fusion, performing data fusion using an improved adaptive Kalman filter algorithm, introducing a non-linear compensation term for data compensation, and determining the pose of the welding torch;

[0055] S3: Digital twin modeling, establishing a simulation model using a thermal-mechanical coupling field and determining the output parameters of the welding torch in combination with welding requirements;

[0056] S4: Adaptive control, controlling the welding robot according to the pose of the welding torch in step S2 and the welding torch parameters in step S3 in combination with the actual welding requirements and the kinematic model of the welding robot.

[0057] In the step S2, the improved adaptive Kalman filter algorithm model is:

[0058] Where:

[0059] x k is the pose state vector of the welding torch, which includes three-dimensional positions (x, y, z) and Euler angles (α, β, γ);

[0060] μ kTo control the input quantity, which comes from the data of the robot joint motor encoder;

[0061] ω k ~N(0, Q k ), which is the process noise, and Q k is the covariance matrix, and Q k = diag([0.1 2 , 0.1 2 , 0.1 2 , 0.05 2 , 0.05 2 , 0.05 2 );

[0062] z k is the observation vector, which is formed by fusing the data collected by the multi-modal sensor array;

[0063] H k is the observation matrix, the Jacobian matrix determined according to the sensor layout;

[0064] υ k ~N(0, R k ), which is the observation noise, and R k = σ 2 I 6×6 , where σ = 0.05mm and I 6×6 is the 6×6 identity matrix.

[0065] In the step S2, the formula for non-linear compensation is: where K k is the time-varying gain coefficient, and through the formula λ ∈ (0.75, 0.85) is the empirical coefficient, and |||| F represents the Frobenius norm of the matrix.

[0066] Embodiment 3:

[0067] A digital twin-based adaptive control method for a welding robot, which includes the following steps:

[0068] S1: Multi-source data acquisition, collecting real-time information through a multi-modal sensor array;

[0069] S2: Multi-source data fusion, fusing the data using an improved adaptive Kalman filtering algorithm and introducing a non-linear compensation term for data compensation to determine the posture of the welding torch;

[0070] S3: Digital twin modeling, establishing a simulation model using a thermo-mechanical coupling field and determining the output parameters of the welding torch in combination with the welding requirements;

[0071] S4: Adaptive control, based on the torch posture in step S2 and the torch parameters in step S3, combined with the actual welding requirements and the kinematic model of the welding robot, to control the welding robot.

[0072] In the said step S2, the improved adaptive Kalman filter algorithm model is:

[0073] Where:

[0074] x k is the torch pose state vector, which includes three-dimensional position (x, y, z) and Euler angles (α, β, γ);

[0075] μ k is the control input quantity, which comes from the data of the robot joint motor encoder;

[0076] ω k ~N(0, Q k ), which is the process noise, and Q k is the covariance matrix, and Q k = diag([0.1 2 , 0.1 2 , 0.1 2 , 0.05 2 , 0.05 2 , 0.05 2 );

[0077] z k is the observation vector, which is formed by fusing the data collected by the multi-modal sensor array;

[0078] H k is the observation matrix, the Jacobian matrix determined according to the sensor layout;

[0079] υ k ~N(0, R k ), which is the observation noise, and R k = σ 2 I 6×6 , σ = 0.05mm, and I 6×6 is the 6×6 identity matrix.

[0080] In the said step S2, the formula for non-linear compensation is: Where K k is the time-varying gain coefficient, through the formula λ ∈ (0.75, 0.85) is the empirical coefficient, |||| F represents the Frobenius norm of the matrix.

[0081] In the said step S3, the heat transfer equation of the thermal-mechanical coupling field simulation model is: Where: ρ is the material density; c p is the specific heat capacity; k is the thermal conductivity; q loss is the heat loss term; q arc is the arc heat source term.

[0082] In the step S3, the arc heat source term q arc adopts a double ellipsoid model: Where U is the welding voltage, I is the welding current, η is the thermal efficiency coefficient, and a, b, c are the arc distribution characteristic parameters.

[0083] Example 4:

[0084] A digital twin-based adaptive control method for a welding robot, which includes the following steps:

[0085] S1: Multi-source data acquisition, collecting real-time information through a multi-modal sensor array;

[0086] S2: Multi-source data fusion, using an improved adaptive Kalman filter algorithm to fuse data and introducing a non-linear compensation term for data compensation to determine the posture of the welding torch;

[0087] S3: Digital twin modeling, establishing a simulation model using a thermal-structural coupling field and determining the output parameters of the welding torch in combination with welding requirements;

[0088] S4: Adaptive control, controlling the welding robot according to the posture of the welding torch in step S2 and the parameters of the welding torch in step S3 in combination with the actual welding requirements and the kinematic model of the welding robot.

[0089] In the step S2, the improved adaptive Kalman filter algorithm model is:

[0090] Where:

[0091] x k is the pose state vector of the welding torch, which includes three-dimensional positions (x, y, z) and Euler angles (α, β, γ);

[0092] μ k is the control input quantity, which comes from the data of the robot joint motor encoder;

[0093] ω k ~N(0, Q k ), which is the process noise, and Q k is the covariance matrix, and Q k = diag([0.1 2 , 0.1 2 , 0.1 2 , 0.05 2 , 0.05 2, 0.05 2 );

[0094] z k is the observation vector, which is formed by fusing the data collected by the multi-modal sensor array;

[0095] H k is the observation matrix, the Jacobian matrix determined according to the sensor layout;

[0096] υ k ~N(0, R k ), which is the observation noise, R k =σ 2 I 6×6 , σ = 0.05mm, I 6×6 is the 6×6 identity matrix.

[0097] In the step S2, the formula for non-linear compensation is: where K k is the time-varying gain coefficient, through the formula λ ∈ (0.75, 0.85) is the empirical coefficient, |||| F represents the Frobenius norm of the matrix.

[0098] In the step S3, the heat transfer equation of the thermo-mechanical coupling field simulation model is: where: ρ is the material density; c p is the specific heat capacity; k is the thermal conductivity; q loss is the heat loss term; q arc is the arc heat source term.

[0099] In the step S3, the arc heat source term q arc adopts the double ellipsoid model: where U is the welding voltage, I is the welding current, η is the thermal efficiency coefficient, and a, b, c are the arc distribution characteristic parameters.

[0100] In the step S4, the control all adopts the control method based on past error compensation, which includes the following steps:

[0101] S41: Collect the target control value F i in the past control and the actually executed value F i ' of the end observed, and calculate the deviation ΔF i =F i -F i ', which represents the value of the i-th control forward from this control;

[0102] S42: Establish a weight compensation model based on time sequence to obtain the compensation deviation ΔF0 for this time;

[0103] S43: Combine the compensation deviation ΔF0, and compensate the target value of this time according to the formula F = F0 - ΔF0, where F is the updated target control value and F0 is the target control value before update.

[0104] In the step S42, the weight compensation model based on time sequence is as follows: where k is the weight coefficient, k i > k i+1 .

[0105] Example 5:

[0106] An adaptive control method for a welding robot based on digital twin, which includes the following steps:

[0107] S1: Multi-source data acquisition, and collect real-time information through a multi-modal sensor array;

[0108] S2: Multi-source data fusion, use an improved adaptive Kalman filtering algorithm to fuse the data, and introduce a non-linear compensation term to compensate the data, and determine the pose of the welding torch;

[0109] S3: Digital twin modeling, use a thermal-mechanical coupling field to establish a simulation model, and determine the output parameters of the welding torch in combination with welding requirements;

[0110] S4: Adaptive control, control the welding robot according to the pose of the welding torch in step S2 and the parameters of the welding torch in step S3 in combination with the actual welding requirements and the kinematic model of the welding robot.

[0111] In the step S2, the improved adaptive Kalman filtering algorithm model is as follows:

[0112] where:

[0113] x k is the pose state vector of the welding torch, which includes three-dimensional position (x, y, z) and Euler angles (α, β, γ);

[0114] μ k is the control input quantity, which comes from the data of the robot joint motor encoder;

[0115] ω k ~N(0, Q k ), which is the process noise, Q k is the covariance matrix, Q k =diag([0.1 2 , 0.1 2 , 0.1 2 , 0.05 2 , 0.05 2 , 0.05 2 );

[0116] z k is an observation vector, which is formed by fusing data collected by a multi-modal sensor array;

[0117] H k is an observation matrix, a Jacobian matrix determined according to the sensor layout;

[0118] υ k ~N(0, R k ), which is observation noise, and R k =σ 2 I 6×6 , σ = 0.05mm, and I 6×6 is a 6×6 identity matrix.

[0119] In the step S2, the formula for non-linear compensation is: where K k is a time-varying gain coefficient, and through the formula λ∈(0.75, 0.85) is an empirical coefficient, |||| F represents the Frobenius norm of the matrix.

[0120] In the step S3, the heat transfer equation of the thermo-mechanical coupling field simulation model is: where: ρ is the material density; c p is the specific heat capacity; k is the thermal conductivity; q loss is the heat loss term; q arc is the arc heat source term.

[0121] In the step S3, the arc heat source term q arc adopts a double ellipsoid model: where U is the welding voltage, I is the welding current, η is the heat efficiency coefficient, and a, b, c are arc distribution characteristic parameters.

[0122] In the step S4, the control all adopts a control method based on past error compensation, which includes the following steps:

[0123] S41: Collect the target control value F i in the past control and the observed actual end execution value F i ′, and calculate the deviation ΔF i =F i -F i ′, which represents the value of the i-th control forward from this control;

[0124] S42: Establish a weight compensation model based on time sequence to obtain the compensation deviation ΔF0 for this time;

[0125] S43: Combine the compensation deviation ΔF0, and compensate the target value F for this time according to the formula F = F0 - ΔF0, where F is the updated target control value and F0 is the target control value before update.

[0126] In the step S42, the weight compensation model based on time sequence is: where k is the weight coefficient, k i > k i+1 .

[0127] In the step S42, where k0 is the initial weight.

[0128] In the present invention, by establishing the state equation and the observation equation of the welding torch, and introducing process noise and observation noise, on the basis of realizing closed-loop feedback control of the welding torch pose, it can also prevent the control error caused by noise. At the same time, a non-linear compensation term is added to increase the control accuracy of the welding torch pose.

[0129] In the present invention, by establishing a simulation model of the thermo-mechanical coupling field, according to the actual output of the welding torch combined with the environmental information, a double-ellipsoid model is used to establish the heat conduction equation, so as to predict the temperature during welding, and then adjust the welding torch parameters in time according to the prediction to improve the welding quality.

[0130] In the present invention, for the control of each parameter, a compensation control method is adopted, and the past error is used to compensate the target control value, which improves the control accuracy and control sensitivity. At the same time, the past error adopts a weighted control form, which can make the compensation value match the working state of the recent welding torch and the entire welding robot, and improve the compensation accuracy.

[0131] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitution or change, and should be covered by the protection scope of the present invention.

Claims

1. An adaptive control system for a welding robot based on digital twin, characterized in that, Including: A multi-source data acquisition module that uses a multi-modal sensor array to acquire real-time information; A multi-source data fusion module that fuses the acquired information, uses an improved adaptive Kalman filtering algorithm to fuse the data, introduces a non-linear compensation term for data compensation, and determines the posture of the welding torch; A digital twin model that uses a thermo-mechanical coupling field to establish a simulation model and determines the output parameters of the welding torch in combination with welding requirements; An adaptive control module that controls the welding robot according to the posture of the welding torch and the welding torch parameters in combination with the actual welding requirements and the kinematic model of the welding robot.

2. The adaptive control system of a welding robot based on digital twin according to claim 1, characterized in that The multi-modal sensor array includes a laser vision sensor, an infrared thermometer, and a six-axis force sensor.

3. A digital-twin-based adaptive control method for a welding robot, which is an implementation method of the digital-twin-based adaptive control system for a welding robot as described in claim 1 or 2, characterized in that, Including the following steps: S1: Multi-source data acquisition, acquiring real-time information through a multi-modal sensor array; S2: Multi-source data fusion, using an improved adaptive Kalman filtering algorithm to fuse the data, introducing a non-linear compensation term for data compensation, and determining the posture of the welding torch; S3: Digital twin modeling, using a thermo-mechanical coupling field to establish a simulation model and determining the output parameters of the welding torch in combination with welding requirements; S4: Adaptive control, controlling the welding robot according to the posture of the welding torch in step S2 and the welding torch parameters in step S3 in combination with the actual welding requirements and the kinematic model of the welding robot.

4. The adaptive control method for a welding robot based on digital twin according to claim 3, characterized in that, In the step S2, the improved adaptive Kalman filtering algorithm model is: Wherein: x k is the pose state vector of the welding torch, which includes the three-dimensional position (x, y, z) and the Euler angles (α, β, γ); μ k To control the input quantity, which comes from the robot joint motor encoder data; ω k ~N(0, Q k ), which is process noise, and Q k is the covariance matrix, where Q k = diag([0.1 2 , 0.1 2 , 0.1 2 , 0.05 2 , 0.05 2 , 0.05 2 ); z k is an observation vector, which is formed by fusing data collected by a multi-modal sensor array; H k is the observation matrix, the Jacobian matrix determined according to the sensor layout; υ k ~N(0, R k ), which is the observation noise, R k = σ 2 I 6×6 , σ = 0.05mm, I 6×6 is a 6×6 identity matrix.

5. The adaptive control method of a welding robot based on digital twin according to claim 4, wherein In the step S2, the formula for non-linear compensation is: where K k is a time-varying gain coefficient, and through the formula λ ∈ (0.75, 0.85) is an empirical coefficient, |||| F represents the Frobenius norm of the matrix.

6. The adaptive control method of a welding robot based on digital twin according to claim 3, wherein, In the step S3, the heat transfer equation of the thermo-mechanical coupling field simulation model is as follows: Where: ρ is the material density; c p is the specific heat capacity; k is the thermal conductivity; q loss is the heat loss term; q arc is the arc heat source term.

7. A self-adaptive control method for a welding robot based on digital twin according to claim 6, characterized in that In the step S3, the arc heat source term q arc adopts a double ellipsoid model: where U is the welding voltage, I is the welding current, η is the thermal efficiency coefficient, and a, b, and c are the arc distribution characteristic parameters.

8. The adaptive control method of a welding robot based on digital twin according to claim 3, characterized in that In the step S4, the control all adopts a control method based on past error compensation, which includes the following steps: S41: Collect the target control value F in the previous control i and the actually executed value F' of the end observed, and calculate the deviation ΔF i ′, and calculate the deviation ΔF i = F i - F i ′, which represents the value of the i-th control forward from this control; S42: Establish a weight compensation model based on time sequence to obtain the compensation deviation ΔF0 for this time; S43: Combine the compensation deviation ΔF0, and compensate the target value for this time according to the formula F = F0 - ΔF0, where F is the updated target control value and F0 is the target control value before update.

9. The adaptive control method for a welding robot based on digital twin according to claim 8, characterized in that, In the step S42, the weight compensation model based on time sequence is as follows: where k is the weight coefficient, k i > k i+1 .

10. A self-adaptive control method for a welding robot based on digital twin according to claim 9, characterized in that, In the step S42, where k0 is the initial weight.

Images