Control method of intelligent robot for steel body welding seam polishing and derusting

Abstract

The invention discloses a control method of an intelligent robot for steel body weld joint polishing and derusting. The control method comprises the following steps that S1, the robot is attached to the surface of a steel body through a magnetic wheel adsorption mechanism; s2, adopting a multispectral visual system to track the welding seam in real time; s3, extracting a weld ridge line based on a Hessian matrix, wherein the ridge line direction is determined by a feature vector corresponding to the maximum feature value; s4, generating a welding seam three-dimensional model through line structured light three-dimensional reconstruction and binocular stereoscopic vision fusion; s5, performing real-time path planning by using a dynamic window method (DWA); s6, carrying out smooth optimization on the path by adopting a B spline curve; s7, an attitude angle theta of the MEMS gyroscope and feedback data Fmeasure of the six-dimensional force sensor are collected in real time; s8, when the visual system detects the surface smoothness, an automatic repair mechanism is triggered, and a path is regenerated; and S9, realizing multi-sensor data synchronization through the DDS middleware. The invention belongs to the field of weld joint grinding machines, and particularly relates to a control method of an intelligent robot for steel body weld joint grinding and rust removal.

Description

Technical Field

[0001] The invention belongs to the field of weld grinding machines, and in particular relates to an intelligent robot control method for grinding and removing rust from steel welds. Background Art

[0002] In the field of weld seam treatment after steel welding, traditional grinding and rust removal operations mainly rely on manual operation or semi-automatic equipment, which has problems such as low efficiency, poor precision, and high labor intensity. In recent years, fully automatic walking grinding robots have gradually been used in this field, but they still have the following technical defects:

[0003] 1. Insufficient automation: Existing equipment lacks a high-precision navigation system and cannot achieve fully autonomous operation. It still requires manual intervention for path planning or remote control (e.g., CN112847057B). Manual operation is particularly difficult and unsafe in confined spaces (e.g., spherical tanks and pipelines).

[0004] 2. To meet the inspection requirements (such as surface finish Ra ≤ 3.2μm), the existing equipment needs to reduce the travel speed, resulting in low efficiency; the grinding material and process parameters (such as speed and pressure) are poorly matched, and repeated tests and adjustments are required (for example, the use of a single abrasive in CN113146532A leads to insufficient adaptability).

[0005] 3. Poor adaptability to dynamic working conditions: When working at an angle (such as welding spherical tanks), the robot's grinding pressure fluctuates (±30%) due to changes in the gravity component, and consistency cannot be guaranteed; the existing force control system relies on preset parameters and lacks real-time sensor feedback compensation (such as the lack of gyroscope and force fusion control). Summary of the Invention

[0006] The technical problems to be solved by the present invention are low degree of automation, inability to simultaneously ensure grinding speed and quality, and poor adaptability to working conditions.

[0007] To solve the above problems, the technical solution adopted by the present invention is as follows: The intelligent robot control method for grinding and derusting steel welds proposed by the present invention comprises the following steps:

[0008] S1: The robot is attached to the steel surface through the magnetic wheel adsorption mechanism, and the magnetic attraction force is ≥500N;

[0009] S2: Real-time tracking of welds using a multispectral vision system. The system includes an infrared laser profiler (Basler acA2000-165 μm), an anti-arcing filter (OD6), and a binocular camera. Adaptive dual-threshold edge detection is performed using an improved Canny operator. The threshold ranges are T1 = 0.05 × Imax and T2 = 0.15 × Imax, where Imax is the maximum grayscale value of the image.

[0010] S3: Extract weld ridge based on Hessian matrix. The calculation formula is:

[0011] The ridge direction is determined by the eigenvector corresponding to the maximum eigenvalue;

[0012] S4: Generate a 3D model of the weld seam through line structured light 3D reconstruction and binocular stereo vision fusion, with a reconstruction accuracy of ≤0.1mm;

[0013] S5: Use the dynamic window algorithm (DWA) for real-time path planning, and its cost function is:

[0014] J(v,ω)=α·heading(v,ω)+β·dist(v,ω)+γ·velocity(v,ω),

[0015] Among them, α, β, and γ are weight coefficients;

[0016] S6: Use B-spline curve to smoothly optimize the path, and the parameterized equation is:

[0017]

[0018] Among them, P i is the control point, N i,p is the B-spline basis function;

[0019] S7: Real-time acquisition of the MEMS gyroscope attitude angle θ and the six-dimensional force sensor feedback data Fmeasured, and the establishment of a dynamic force compensation model based on finite element analysis:

[0020] Fcomp=K·(G·sinθ+Fmeasured),

[0021] Where K is the stiffness coefficient, which is calibrated by finite element simulation;

[0022] S8: When the vision system detects that the surface finish Ra>3.2μm, the automatic rework mechanism is triggered to regenerate the path and adjust the grinding pressure to 3±0.5N;

[0023] S9: Multi-sensor data synchronization is achieved through DDS middleware, with communication delay ≤ 5ms, and OPC UA protocol is used to interact with the host computer.

[0024] Furthermore, in step S2, multispectral fusion adopts weighted superposition of visible light and near-infrared bands, and the formula is:

[0025] Ifusion=0.7·Ivisible+0.3·I NIR .

[0026] Furthermore, in step S5, the search space of the dynamic window method is:

[0027] v∈[v min ,v max ],ω∈[ω min ,ω max ],

[0028] where v max =0.5m / s,ω max =1.0rad / s.

[0029] Furthermore, in step S7, the sampling frequency of the six-dimensional force sensor is ≥1kHz, and the noise is eliminated by Kalman filtering. The state equation is:

[0030] x k =Ax k-1 +Bu k +w k ,z k =Hx k +v k .

[0031] Furthermore, in step S8, the surface finish is detected using gray level co-occurrence matrix (GLCM) texture analysis, and the contrast calculation formula is:

[0032]

[0033] Where P(i,j) is the gray-level co-occurrence matrix element.

[0034] Furthermore, the magnetic wheel adsorption mechanism adopts a Halbach array permanent magnet with a magnetic flux density of ≥1.2T, and the adsorption force calculation formula is:

[0035]

[0036] Among them, B is the magnetic induction intensity, A is the magnetic pole area, μ 0 is the magnetic permeability of vacuum.

[0037] Furthermore, it also includes a grinding mechanism, which is driven by a planetary gearbox with a rotation speed of ≥30,000 rpm and a gearbox efficiency of ≥95%. The transmission ratio is calculated as follows:

[0038]

[0039] Where Zsun and Zring are the number of teeth of the sun gear and ring gear respectively.

[0040] Furthermore, the image processing of the visual system adopts a sub-pixel edge fitting algorithm, and the fitting formula is:

[0041] e(x)=ax 2 +bx+c, the edge position is given by Sure.

[0042] The beneficial effects achieved by the present invention using the above method are as follows:

[0043] 1. This proposal proposes an intelligent robot control method for grinding and derusting steel welds. The invention achieves fully autonomous operation through multimodal sensor fusion (vision + force perception + gyroscope), replacing traditional manual remote control operation and avoiding human errors and safety hazards.

[0044] 2. The intelligent robot control method for grinding and rust removal of steel welds proposed in this scheme adopts an improved Canny operator and dynamic force compensation model to achieve sub-pixel weld recognition (accuracy 0.1mm) and constant force control (3±0.5N), solving the problem of uneven grinding caused by inclination angle changes in traditional equipment.

[0045] 3. The intelligent robot control method for grinding and derusting steel welds proposed in this plan uses magnetic wheel adsorption (≥500N) and dynamic path planning (DWA+B-spline) to ensure the robot's reliable adsorption and movement on complex curved surfaces (such as inclination angles of 0-90°) without the risk of falling. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0047] The present invention provides an intelligent robot control method for grinding and removing rust from steel welds, comprising the following steps:

[0048] S1: The robot is attached to the steel surface through the magnetic wheel adsorption mechanism, and the magnetic attraction force is ≥500N;

[0049] S2: A multispectral vision system is used to track welds in real time. The vision system includes an infrared laser profiler (Basler acA2000-165um), an anti-arcing filter (OD6 level), and a binocular camera. Adaptive dual-threshold edge detection is performed using an improved Canny operator. The threshold ranges are T1 = 0.05 × Imax and T2 = 0.15 × Imax, where Imax is the maximum grayscale value of the image. Multispectral fusion uses weighted superposition of visible light and near-infrared bands. The formula is:

[0050] Ifusion=0.7·Ivisible+0.3·INIR;

[0051] S3: Extract weld ridge based on Hessian matrix. The calculation formula is:

[0052] The ridge direction is determined by the eigenvector corresponding to the maximum eigenvalue;

[0053] S4: Generate a 3D model of the weld seam through line structured light 3D reconstruction and binocular stereo vision fusion, with a reconstruction accuracy of ≤0.1mm;

[0054] S5: Use the dynamic window algorithm (DWA) for real-time path planning, and its cost function is:

[0055] J(v,ω)=α·heading(v,ω)+β·dist(v,ω)+γ·velocity(v,ω),

[0056] Among them, α, β, γ are weight coefficients,

[0057] The search space of the dynamic window method is:

[0058] v∈[v min ,v max ],ω∈[ω min ,ω max ],

[0059] where v max =0.5m / s,ω max =1.0rad / s;

[0060] S6: Use B-spline curve to smoothly optimize the path, and the parameterized equation is:

[0061]

[0062] Among them, P i is the control point, N i,p is the B-spline basis function;

[0063] S7: Real-time acquisition of the MEMS gyroscope attitude angle θ and the six-dimensional force sensor feedback data Fmeasured, and the establishment of a dynamic force compensation model based on finite element analysis:

[0064] Fcomp=K·(G·sinθ+Fmeasured),

[0065] Where K is the stiffness coefficient. Through finite element simulation calibration, the sampling frequency of the six-dimensional force sensor is ≥1kHz, and the noise is eliminated by Kalman filtering. The state equation is:

[0066] x k =Ax k-1 +Bu k +wk ,z k =Hx k +v k ;

[0067] S8: When the visual system detects that the surface finish Ra>3.2μm, the automatic rework mechanism is triggered, the path is regenerated and the grinding pressure is adjusted to 3±0.5N. The surface finish detection adopts gray level co-occurrence matrix (GLCM) texture analysis. The contrast calculation formula is:

[0068]

[0069] Where P(i,j) is the gray level co-occurrence matrix element;

[0070] S9: Multi-sensor data synchronization is achieved through DDS middleware, with communication delay ≤ 5ms, and OPC UA protocol is used to interact with the host computer.

[0071] The magnetic wheel adsorption mechanism adopts Halbach array permanent magnet, the magnetic flux density is ≥1.2T, and the adsorption force calculation formula is:

[0072]

[0073] Among them, B is the magnetic induction intensity, A is the magnetic pole area, μ 0 is the vacuum magnetic permeability; it also includes a grinding mechanism, which is driven by a planetary gearbox with a speed of ≥30,000 rpm and a gearbox efficiency of ≥95%. The transmission ratio is calculated as follows:

[0074]

[0075] Where Zsun and Zring are the number of teeth on the sun gear and ring gear respectively; the image processing of the visual system adopts the sub-pixel edge fitting algorithm, and the fitting formula is:

[0076] e(x)=ax 2 +bx+c, the edge position is given by Sure.

[0077] Example: Grinding the weld seam on the inner wall of a spherical tank

[0078] 1. System initialization:

[0079] The magnetic wheel adsorption force is set to 600N (Equation 8);

[0080] The visual system scanning frequency is 60Hz.

[0081] 2. Weld inspection:

[0082] Improved Canny operator (T1=30, T2=100);

[0083] The Hessian matrix is used to extract the ridges (Equation 3).

[0084] 3. Path planning:

[0085] DWA parameters: α=0.6, β=0.3, γ=0.1;

[0086] The spacing between B-spline control points is 10 mm.

[0087] 4.Strength control execution:

[0088] At a 45° inclination angle, the compensation force Fcomp = 25N (Equation 7);

[0089] The feedback frequency of the six-axis force sensor is 1kHz.

[0090] 5.Quality inspection:

[0091] The GLCM contrast threshold was set to 150;

[0092] Unqualified areas will be automatically repaired 3 times.

[0093] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A method for controlling an intelligent robot for grinding and derusting steel welds, characterized in that: The following steps are involved: S1: The robot is attached to the steel surface through the magnetic wheel adsorption mechanism, and the magnetic attraction force is ≥500N; S2: A multispectral vision system was used to track welds in real time. The system included an infrared laser profiler (Basler acA2000-165 μm), an anti-arcing filter (OD6), and a binocular camera. Adaptive dual-threshold edge detection was performed using an improved Canny operator. The threshold ranges were T1 = 0.05 × Imax and T2 = 0.15 × Imax, where Imax is the maximum grayscale value of the image. S3: Extract weld ridge based on Hessian matrix. The calculation formula is: The ridge direction is determined by the eigenvector corresponding to the maximum eigenvalue; S4: Generate a 3D model of the weld seam through line structured light 3D reconstruction and binocular stereo vision fusion, with a reconstruction accuracy of ≤0.1mm; S5: Use the dynamic window algorithm (DWA) for real-time path planning, and its cost function is: J(v,ω)=α·heading(v,ω)+β·dist(v,ω)+γ·velocity(v,ω), Among them, α, β, and γ are weight coefficients; S6: Use B-spline curve to smoothly optimize the path, and the parameterized equation is: Where Pi is the control point and Ni,p is the B-spline basis function; S7: Real-time acquisition of the MEMS gyroscope attitude angle θ and the six-dimensional force sensor feedback data Fmeasured, and the establishment of a dynamic force compensation model based on finite element analysis: Fcomp=K·(G·sinθ+Fmeasured), Where K is the stiffness coefficient, which is calibrated by finite element simulation; S8: When the vision system detects that the surface finish Ra>3.2μm, the automatic rework mechanism is triggered to regenerate the path and adjust the grinding pressure to 3±0.5N; S9: Multi-sensor data synchronization is achieved through DDS middleware, with communication delay ≤ 5ms, and OPC UA protocol is used to interact with the host computer.

2. The intelligent robot control method for grinding and derusting steel welds according to claim 1, characterized in that: In step S2, multispectral fusion uses weighted superposition of visible light and near-infrared bands, and the formula is: Ifusion=0.7·Ivisible+0.3·I NIR 。 3. The intelligent robot control method for grinding and derusting steel welds according to claim 2, characterized in that: In step S5, the search space of the dynamic window method is: v∈[v min ,v max ],ω∈[ω min ,oh max ], where v max =0.5m / s,ω max =1.0rad / s.

4. The intelligent robot control method for grinding and derusting steel welds according to claim 3, characterized in that: In step S7, the sampling frequency of the six-dimensional force sensor is ≥1kHz, and the noise is eliminated by Kalman filtering. The state equation is: x k =Ax k-1 +Bu k +w k ,Z k =Hx k +v k 。 5. The intelligent robot control method for grinding and derusting steel welds according to claim 4, characterized in that: In step S8, the surface finish is detected using gray level co-occurrence matrix (GLCM) texture analysis, and the contrast calculation formula is: Where P(i,j) is the gray-level co-occurrence matrix element.

6. The intelligent robot control method for grinding and derusting steel welds according to claim 5, characterized in that: The magnetic wheel adsorption mechanism adopts Halbach array permanent magnets with a magnetic flux density of ≥1.2T, and the adsorption force calculation formula is: Among them, B is the magnetic induction intensity, A is the magnetic pole area, μ 0 is the magnetic permeability of vacuum.

7. The intelligent robot control method for grinding and derusting steel welds according to claim 6, characterized in that: It also includes a grinding mechanism, which is driven by a planetary gearbox with a rotation speed of ≥30,000 rpm and a gearbox efficiency of ≥95%. The transmission ratio is calculated as follows: Where Zsun and Zring are the number of teeth of the sun gear and ring gear respectively.

8. The intelligent robot control method for grinding and derusting steel welds according to claim 7, characterized in that: The image processing of the visual system adopts a sub-pixel edge fitting algorithm, and the fitting formula is: e(x)=ax 2 +bx+c, the edge position is given by Sure.

Images