Large casting laser adaptive rough opening method and system based on multi-modal perception

Abstract

The application relates to the technical field of casting surface treatment, in particular to a large casting laser self-adaptive rough opening method and system based on multi-modal perception, wherein the method comprises the following steps: performing global scanning on a casting in a lying state by using a 3D visual sensor to obtain a measured point cloud; adopting an iterative closest point algorithm to register the measured point cloud with a theoretical CAD model, calculating the residual error of the two to generate a sand sticking thickness thermal map; reading the joint angle fed back by a mechanical arm controller in real time, and solving the normal vector of the flange at the end of the mechanical arm through forward kinematics transformation; obtaining the included angle between the laser beam exit direction and the normal of the casting surface; dividing the casting area to be processed into zones; based on the zoning information, generating a laser processing path by using a scanning track algorithm and processing parameters corresponding to the zones, and executing cleaning on the corresponding zones based on the processing parameters and the laser processing path. The application has the technical characteristics of high cleaning efficiency and quality and the ability to self-adaptively adjust process parameters.

Description

Technical Field

[0001] This application relates to the field of casting surface treatment technology, specifically to a laser adaptive coarse opening method and system for large castings based on multimodal sensing. Background Technology

[0002] In the field of heavy-duty gas turbines and steam turbines, large cast steel parts (such as cylinders made of ZG15Cr1Mo1V material) need to have about 2mm of oxide scale and sintered sand removed from their surface after casting and before final heat treatment.

[0003] Existing technologies mainly employ manual grinding or mechanical grinding using a robotic arm holding a flap wheel. Manual grinding suffers from high labor intensity, severe dust pollution, unstable cleaning quality, and is prone to over-grinding of the substrate, as well as high safety risks. Mechanical grinding, on the other hand, has disadvantages such as extremely high material costs, the potential to introduce mechanical stress or micro-cracks into the surface of castings, and the inability to clean corners and small-radius areas due to limitations in the size of the grinding wheel (the overall grindable area is less than 60%).

[0004] Both manual and mechanical polishing have the following drawbacks: 1. Blindness of open-loop control: Using fixed laser parameters cannot adapt to the non-uniform distribution of sand thickness on the casting surface, resulting in the ablation of the substrate in thin sand areas or incomplete cleaning in thick sand areas.

[0005] 2. Lack of real-time energy compensation for complex curved surfaces: When the cylinder is horizontally processed, the pitch and flipping posture of the end of the robotic arm changes continuously, and the angle between the laser beam and the normal of the workpiece surface changes in real time. Existing equipment cannot identify the normal of the curved surface and compensate for the laser energy in real time. The cleaning effect of the top surface is up to standard, but the cleaning effect of the vertical surface and small radius curved surface is poor.

[0006] 3. Lack of quantitative quality control closed loop: After cleaning, the inspection relies on manual visual inspection, and there is no means to quantitatively detect the thickness of residual sand; once it is missed, the residual sand will become more difficult to remove after subsequent heat treatment, and will seriously affect the uniformity of the 5mm machining allowance.

[0007] 4. Lack of software collaboration among multiple devices and low level of intelligence: 3D vision, robotic arm and purging system are controlled independently without millisecond-level timing linkage. They cannot adaptively match processing and purging strategies according to the thickness of sand and the shape of curved surfaces, resulting in low processing efficiency and high rework rate.

[0008] In summary, given the problems of high labor costs, uneven surface cleaning, uncontrollable quality, and low level of intelligence in existing technologies, there is an urgent need for a surface treatment process for cast steel parts that can improve cleaning efficiency and quality and adaptively adjust process parameters. Summary of the Invention

[0009] In view of this, this application provides a method and system for adaptive coarse beam opening of large castings using laser based on multimodal sensing.

[0010] Firstly, a laser adaptive coarse-opening method for large castings based on multimodal sensing is provided, which includes the following steps: A 3D vision sensor mounted on the end effector of a robotic arm is used to perform a full-area scan of the casting in a horizontal position to obtain the measured point cloud Preal.

[0011] The measured point cloud Preal and the theoretical CAD model Mcad are registered using the iterative nearest point algorithm, and the residual ΔZ between the two is calculated to generate a heat map T(x,y) of the sand adhesion thickness.

[0012] The joint angles fed back by the robotic arm controller are read in real time, and the normal vector n of the end flange of the robotic arm is calculated through forward kinematic transformation.

[0013] Based on a high-power pulsed laser generator mounted on the end effector of a robotic arm, the angle θ between the laser beam emission direction and the normal to the surface of the casting is obtained, wherein the angle θ ranges from 0° to 90°.

[0014] Based on the heat map of the sand adhesion thickness T(x,y), the normal vector n at the end of the robotic arm, and the included angle θ, the area to be processed in the casting is divided into zones.

[0015] Based on the partition information, a laser processing path is generated using the corresponding partition scanning trajectory algorithm and processing parameters. Then, based on the processing parameters and laser processing path, the corresponding area is cleaned until all partitions of the workpiece have completed rough opening.

[0016] The above method also includes: After completing the laser processing of an area, the area is re-scanned using a 3D vision sensor. The three-dimensional topography data obtained from the re-scan is compared with the theoretical CAD model Mcad, and the surface roughness Ra and height residual ΔH after cleaning are calculated.

[0017] Set a residual judgment threshold. If the height residual ΔH is greater than the residual judgment threshold, then store the coordinates of the area in the supplementary scan queue.

[0018] After the current cycle ends, the robotic arm performs the supplementary scanning based on the supplementary scanning queue information, with reduced power and speed.

[0019] In the above method, the iterative nearest-point algorithm is used to register the measured point cloud Preal with the theoretical CAD model Mcad, and the residual ΔZ between the two is calculated to generate a heat map T(x,y) of the sand-coated thickness. Specifically: The residual ΔZ is calculated using the following formula: ΔZ=Zreal−Zcad Where Zreal represents the actual Z-coordinate of a point as measured by the 3D vision sensor; Zcad represents the theoretical Z-coordinate of that point on the CAD model Mcad; ΔZ represents the deviation between the measured value and the theoretical value. If the thickness threshold of the casting is set to δ, then: If ΔZ>δ, then the corresponding area of ​​the casting is determined to be a sand-adhering area, and a heat map of sand adhesion thickness T(x,y) is generated.

[0020] In the above method, the step of reading the joint angles fed back by the robotic arm controller in real time and calculating the normal vector n of the robotic arm end flange through forward kinematics transformation is specifically as follows: The normal vector n of the end flange is calculated using the following forward kinematics transformation formula: T =A1(θ1)·A2(θ2)·A3(θ3)·...·An(θn) Where T represents the coordinate transformation from the base to the end flange; A1-An represents the length of the 1-n segments of the robotic arm; and θ1-θn represents the joint angle of the 1-n segments of the robotic arm.

[0021] In the above method, the angle θ between the laser beam emission direction and the normal to the casting surface is obtained based on the high-power pulsed laser generator mounted on the end effector of the robotic arm; based on the heat map of sand adhesion thickness T(x,y), the normal vector n of the end effector of the robotic arm, and the angle θ, the area to be processed in the casting is divided into zones, specifically: The collected point cloud data is filtered and denoised, and the surface normal vector of each point is calculated by neighborhood fitting.

[0022] Based on the surface normal vector and the laser emission direction, the angle θ between the laser emission direction and the workpiece surface normal is obtained.

[0023] Based on the included angle θ, the tolerance of the normal included angle of 5°~8°, and the proximity distance of 2~3 times the laser spot diameter, the discrete point cloud is merged into a connected processing region, and the processing region is divided as follows using a region growing algorithm: When θ < 30°, it is determined to be a flat region.

[0024] When 30°≤θ<70°, it is determined to be a steep region.

[0025] When θ ≥ 70°, it is determined to be a near-vertical region.

[0026] Among them, fragmented areas with an area less than 10 times the square of the laser spot diameter are removed and merged into the nearest processing area or marked as an independent processing area.

[0027] In the above method, the step of generating a laser processing path based on partition information using the corresponding partition's scanning trajectory algorithm and processing parameters, and then performing cleanup on the corresponding area based on the processing parameters and the laser processing path, specifically involves: Execute based on partition information: A. If the area is flat, then the raster scan mode will be used for cleaning, specifically: Set the overlap ratio between two adjacent raster scan lines to 30-40%.

[0028] Generate reciprocating parallel scan lines and project the scan lines back onto the 3D surface.

[0029] Set the feed speed of the high-power pulsed laser generator to the reference speed V.

[0030] B. If the area is steep, the contour line scanning mode will be used for cleaning, specifically: Determine the height direction information of the high-power pulsed laser generator, wherein the height direction is the Z-axis of the world coordinate system or the normal of the fitted plane.

[0031] Based on the average pre-compression layer height at the included angle θ within the region, the laser path spacing is set to 0.65 × spot diameter × cosθ.

[0032] The slices are layered along the height direction, and the cross-sectional profile of each layer is offset inward to form a concentric ring scanning path. The scanning direction of adjacent rings is set to alternate between forward and reverse scanning to reduce thermal deformation. The feed speed of the high-power pulsed laser generator is set to 60% of the reference speed V.

[0033] C. If the area is near vertical, a spiral / strip scanning mode will be used for cleaning, specifically: Use spiral or strip composite scanning; the area overlap rate between two adjacent scanning paths is 50%, and the feed speed is adjusted to 40% to 50% of the reference speed V.

[0034] In this process, at the junction of each processing area, the boundary is expanded outward by 1 to 2 times the spot area as a transition zone. A linear interpolation strategy is used within the transition zone to increase the regional overlap rate by 50%, and the feed speed and laser power are smoothly transitioned from the parameters of adjacent areas.

[0035] And during execution: If the maximum residual value within the partition does not exceed 0.5mm, the power is adjusted to 50% of the rated power of the laser beam of the high-power pulsed laser generator, the speed is adjusted to 90% of the rated operating speed of the robotic arm, and the overlap ratio between two adjacent grating scan lines is set to 20%~30%.

[0036] If the maximum residual value within the partition is greater than 0.5 and less than or equal to 2mm, the power is adjusted to 75% of the rated power of the laser beam, the speed is adjusted to 60% of the rated operating speed of the robotic arm, and the overlap ratio between two adjacent grating scan lines is set to 30%~40%.

[0037] If the maximum residual value within the partition is greater than 2mm and less than 4mm, the power is adjusted to 90% of the rated power of the laser beam, the speed is adjusted to 30% of the rated operating speed of the robotic arm, and the overlap ratio between two adjacent grating scan lines is set to 40%~50%.

[0038] The above method also includes a dynamic parameter adaptive adjustment step, which includes: Based on the angle θ between the laser beam and the normal to the casting surface calculated in real time, the laser output power Pout is dynamically adjusted through the normal energy compensation model to maintain a constant energy input per unit area.

[0039] The normal energy compensation model is as follows: Pout=Pbase×(1+k×|cos(θ)|⁻¹) Where Pout is the actual output power, Pbase is the base power of the laser beam, and k is the compensation coefficient.

[0040] The above method also includes a positive pressure purging coordination step, which specifically includes: When the laser beam enters the non-re-spraying zone, the positive pressure purging system is triggered to start at the first pressure and purge the laser-polished area.

[0041] A preset time is elapsed before the laser beam enters the re-spraying zone, triggering the positive pressure purging system to start with a second pressure, using high-pressure airflow to loosen the dense sand layer.

[0042] in, The second pressure is greater than the first pressure; The respray zone is defined as: Based on the heat map T(x,y) of the sand thickness, areas with a thickness exceeding the set threshold are marked as re-spraying zones; when the laser beam enters the re-spraying zone, the rated power of the laser beam for the re-spraying zone is automatically increased to 100%-110%, and the speed is adjusted to 15% of the robot's rated operating speed.

[0043] The above method also includes an AI incremental learning step: A differential closed-loop mechanism is established between the measured point cloud (Preal) and the completed clean point cloud. For complex curved surfaces, a high-precision differential model of the initial state point cloud and the completed clean point cloud is constructed.

[0044] The cleaned 3D topography is compared with the initial state using a normal alignment algorithm to automatically generate an RGB pseudo-color residual heatmap.

[0045] Based on residual heatmaps, under-cleaned and over-cleaned regions are identified and marked, and the energy density threshold is adaptively fine-tuned through multidimensional correlation analysis.

[0046] Based on the residual data, iterate the energy density thresholds of under-cleaned and over-cleaned regions, construct a cloud-based process library, and automatically optimize the machining parameters of complex curved surfaces.

[0047] Secondly, a multimodal sensing-based adaptive laser coarse-opening system for large castings is provided to implement the aforementioned adaptive laser coarse-opening method for large castings, comprising: The actuator, including a five-axis industrial robotic arm and an external travel axis, is used to drive the end effector to any position on the surface of the casting.

[0048] The processing unit includes a high-power pulsed laser generator and a laser galvanometer mounted on the end effector, used to generate high-energy pulsed lasers and focus them to scan, break up and peel off oxide scale and adhering sand; The sensing unit includes a 3D vision sensor mounted on the end effector, used to acquire three-dimensional topography and texture information of the casting surface.

[0049] The auxiliary unit includes a positive pressure purging system mounted on the actuator for blowing away loose sand particles that have been broken up by the laser on the surface of the casting.

[0050] The system includes a control unit, comprising an industrial control computer and a robotic arm controller. The industrial control computer is used for motion control, parameter adjustment, data processing, and issuing control commands to the actuators, processing units, sensing units, and auxiliary units. The robotic arm controller is used to receive commands from the industrial control computer and control the movement of the actuators.

[0051] This application embodiment precisely controls the laser energy density between the sand removal threshold and the substrate ablation threshold, and performs real-time normal energy compensation to ensure that the energy density fluctuation across the entire surface is less than ±10%, the heat-affected zone is less than 5μm, and the single-sided substrate removal amount is less than 0.05mm. Through intelligent zoning and path optimization, the effective processing time of the robotic arm is increased (to over 78%), the rework rate is reduced (to below 3%), and the overall cycle time is improved by more than 40% compared to traditional pneumatic grinding. Simultaneously, 3D point cloud data is used as the acceptance carrier to achieve quantitative judgment of the cleaning effect and full-process data traceability. Relying on the AI ​​incremental learning engine, the process parameters are automatically optimized through the differential closed-loop mechanism, which greatly reduces the cost of introducing and debugging new processes and the workload of on-site manual debugging. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating the laser adaptive coarse-opening method for large castings based on multimodal sensing provided in this application embodiment; Figure 2 The logic control diagram of the laser adaptive coarse opening method for large castings based on multimodal sensing provided in the embodiments of this application; Detailed Implementation

[0054] To better understand the above technical solutions, the technical solutions of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this application and the specific features in the embodiments are detailed descriptions of the technical solutions of this application, rather than limitations on the technical solutions of this application. In the absence of conflict, the embodiments of this application and the technical features in the embodiments can be combined with each other.

[0055] Please see Figure 1 , Figure 2 This paper presents a laser adaptive coarse-opening method for large castings based on multimodal sensing, which includes the following steps: Step 100: Use the 3D vision sensor mounted on the end effector of the robotic arm to perform a full-area scan of the horizontally placed casting and obtain the measured point cloud Preal.

[0056] Step 200: The measured point cloud Preal is registered with the theoretical CAD model Mcad using the iterative nearest point algorithm, and the residual ΔZ between the two is calculated to generate a heat map T(x,y) of the sand adhesion thickness.

[0057] Step 300: Read the joint angles fed back by the robotic arm controller in real time, and calculate the normal vector n of the robotic arm end flange through forward kinematic transformation.

[0058] Step 400: Based on the high-power pulsed laser generator mounted on the end effector of the robotic arm, obtain the angle θ between the laser beam emission direction and the normal of the casting surface, wherein the angle θ is in the range of 0°~90°.

[0059] In this step, if the included angle θ exceeds 90°, it is considered an invalid incident or requires flipping.

[0060] Step 500: Based on the heat map of the sand adhesion thickness T(x,y), the normal vector n at the end of the robotic arm and the included angle θ, the area to be processed of the casting is divided into zones.

[0061] Step 600: Based on the partition information, generate a laser processing path using the scanning trajectory algorithm and processing parameters of the corresponding partition, and perform cleaning on the corresponding area based on the processing parameters and laser processing path until all partitions of the workpiece have completed rough opening.

[0062] Step 600 also includes a rescanning step 601, which specifically includes: Step 6011: After completing the laser processing of a region, the region is re-scanned using a 3D vision sensor. The three-dimensional topography data obtained from the re-scan is compared with the theoretical CAD model Mcad, and the surface roughness Ra and height residual ΔH after cleaning are calculated.

[0063] In this embodiment, the surface roughness must meet the standard in order for the height residual to be up to standard. Therefore, the surface roughness Ra is calculated first. If it does not meet the standard, it means that the height residual must not meet the standard. Then step 6012 can be executed directly for judgment. If it meets the standard, the height residual ΔH needs to be judged to determine whether to add a sweep.

[0064] Step 6012: Set a residual judgment threshold. If the height residual ΔH is greater than the residual judgment threshold, then store the coordinates of the area in the supplementary scan queue. Specifically, the residual judgment threshold can be 0.2mm.

[0065] Step 6013: After the current cycle ends, based on the supplementary scanning queue information, the robotic arm performs supplementary scanning with reduced power and speed.

[0066] In this step, the power can be reduced to 20-40% of the original power, and the speed can be reduced to 40-60% of the original speed. Specifically, the power is reduced to 30% of the original power, and the speed is reduced to 50% of the original speed.

[0067] In step 200, the iterative nearest-point algorithm is used to register the measured point cloud Preal with the theoretical CAD model Mcad, and the residual ΔZ between the two is calculated to generate a heat map T(x,y) of the sand-coated thickness. Specifically: The residual ΔZ is calculated using the following formula: ΔZ=Zreal−Zcad Where Zreal represents the actual Z-coordinate of a point as measured by the 3D vision sensor; Zcad represents the theoretical Z-coordinate of that point on the CAD model Mcad; and ΔZ represents the deviation between the measured value and the theoretical value.

[0068] If the thickness threshold of the casting is set to δ, then: If ΔZ>δ, then the corresponding area of ​​the casting is determined to be a sand-adhering area, and a heat map of sand adhesion thickness T(x,y) is generated.

[0069] Specifically, the value range of δ is 0~5mm.

[0070] In step 300, the real-time reading of the joint angle information fed back by the robotic arm controller, and the calculation of the normal vector n of the robotic arm end flange through forward kinematics transformation, specifically involves: The normal vector n of the end flange is calculated using the following forward kinematics transformation formula: T =A1(θ1)·A2(θ2)·A3(θ3)·...·An(θn) Where T represents the coordinate transformation from the base to the end flange; A1-An represents the length of the 1-n segments of the robotic arm; and θ1-θn represents the joint angle of the 1-n segments of the robotic arm.

[0071] Steps 400 and 500 are specifically as follows: The collected point cloud data is filtered and denoised, and the surface normal vector of each point is calculated by neighborhood fitting.

[0072] Based on the surface normal vector and the laser emission direction, the angle θ between the laser emission direction and the workpiece surface normal is obtained, where θ = arccos(laser normal ⋅ surface normal).

[0073] Based on the included angle θ, the tolerance of the included normal angle of 5°~8°, and the proximity distance of 2~3 times the laser spot diameter, the discrete point cloud is merged into a connected processing region, and the following processing region is divided using a region growing algorithm.

[0074] When θ < 30°, it is determined to be a flat region.

[0075] When 30°≤θ<70°, it is determined to be a steep region.

[0076] When θ ≥ 70°, it is determined to be a near-vertical region.

[0077] Among them, fragmented areas with an area less than 10 times the square of the laser spot diameter are removed and merged into the nearest processing area or marked as an independent processing area.

[0078] In step 600, the step of generating a laser processing path based on the partition information using the corresponding partition's scanning trajectory algorithm and processing parameters, and then performing cleanup on the corresponding area based on the processing parameters and the laser processing path, specifically involves: Execute based on partition information: A. If the area is flat, then the raster scan mode will be used for cleaning, specifically: Set the overlap ratio between two adjacent raster scan lines to 30-40%; Generate reciprocating parallel scan lines and project the scan lines back onto the 3D surface; Set the feed speed of the high-power pulsed laser generator to the reference value speed V; Specifically, the overlap ratio between two adjacent grating scan lines refers to the following: before laser beam processing, the grating scan lines are projected onto the workpiece surface for positioning. Two adjacent scan lines will partially overlap, forming an overlap area. The overlap ratio of this area is calculated as: twice the width of the scan line minus the total width of the two scan lines, then divided by twice the width of the scan line. For example, if one scan line is 0.6mm wide and the total width of the two scan lines is 1mm, the overlap ratio is (0.6*2-1) / (0.6*2) = 16.7%.

[0079] It should be noted that when setting the overlap ratio between two adjacent grating scan lines, a ratio less than 20% can easily result in slight streaks on the workpiece surface, leading to rework. An overlap ratio exceeding 50% can cause heat concentration on the workpiece surface, potentially forming melting pits. Therefore, in this embodiment, a laser scan line overlap ratio of 30-40% is more reasonable. Specifically, the laser scan line overlap ratio is set to 35%.

[0080] B. If the area is steep, the contour line scanning mode will be used for cleaning, specifically: Determine the height direction information of the high-power pulsed laser generator, wherein the height direction is the Z-axis of the world coordinate system or the normal of the fitted plane; Based on the average pre-compression layer height of the included angle θ in the region, the laser row spacing is set to 0.65 × spot diameter × cosθ. The average pre-compression layer height specifically refers to the following: in steep areas, the row spacing will be 'stretched' due to the slope. The layer height is pre-corrected by multiplying the average inclination angle θˉ in the region by cosθˉ so that the actual overlap rate on the slope is still maintained at 35%.

[0081] The slices are layered along the height direction, and the cross-sectional profile of each layer is offset inward to form a concentric ring scanning path. The scanning direction of adjacent rings is set to alternate between forward and reverse scanning to reduce thermal deformation. The feed rate of the high-power pulsed laser generator is set to 60% of the reference speed V. C. If the area is near vertical, a spiral / strip scanning mode will be used for cleaning, specifically: Use a combination of spiral or strip scanning; the overlap rate between two adjacent scanning paths is 50%, and the feed rate is adjusted to 40%–50% of the reference speed V. As a specific implementation method, the following table can be used as a reference to perform the cleaning using the corresponding process parameters during application:

[0082] Table 1: Partition Parameter Settings Table In this process, at the junction of each processing area, the boundary is expanded outward by 1 to 2 times the spot area as a transition zone. A linear interpolation strategy is used within the transition zone to increase the regional overlap rate by 50%. The feed speed and laser power are smoothly transitioned from the parameters of adjacent areas to avoid obvious processing marks.

[0083] It also includes a dynamic parameter adaptive adjustment step 602, which includes: Based on the angle θ between the laser beam and the normal of the casting surface calculated in real time, the laser output power Pout is dynamically adjusted through the normal energy compensation model to maintain a constant energy input per unit area. The normal energy compensation model is as follows: Pout=Pbase×(1+k×|cos(θ)|⁻¹) Where Pout is the actual output power, Pbase is the base power of the laser beam, and k is the compensation coefficient. Specifically, the value of K ranges from 1.1 to 1.6.

[0084] This embodiment also includes a positive pressure purging coordination step 603, which specifically includes: Step 6031: When the laser beam enters the non-re-spraying area, the positive pressure purging system is triggered to start with the first pressure to purge the laser polishing area. Step 6032: Before the laser beam enters the re-spraying zone, a preset time (e.g., 50ms) is elapsed, the positive pressure purging system is triggered to start with the second pressure, using high-pressure airflow to loosen the dense sand layer; in, The second pressure is greater than the first pressure; specifically, the first pressure can be 0.2 MPa and the second pressure can be 0.8 MPa.

[0085] The respray zone is defined as: Based on the heat map T(x,y) of the sand adhesion thickness, areas with thicknesses exceeding a set threshold are marked as re-spraying zones. When the laser beam enters the re-spraying zone, the pulse frequency and pulse width of the laser beam are automatically increased for the re-spraying zone. Specifically, when the laser beam enters the re-spraying zone, the rated power of the laser beam for the re-spraying zone is automatically increased to 100%-110%, and the speed is adjusted to 15% of the robot's rated operating speed.

[0086] Specifically, the set threshold is 2mm.

[0087] This embodiment also includes an AI incremental learning step 700, which specifically includes: Step 701: Establish a differential closed-loop mechanism between the measured point cloud (Preal) and the finished clean point cloud, and construct a high-precision differential model between the initial state point cloud and the finished clean point cloud for complex curved surfaces. Step 702: The cleaned 3D shape is compared with the initial state using a normal alignment algorithm to automatically generate an RGB pseudo-color residual heat map; Step 703: Identify and mark under-cleaned and over-cleaned regions based on residual heatmaps, and adaptively fine-tune the energy density threshold through multidimensional correlation analysis.

[0088] Step 704: Based on the residual data, iterate the energy density thresholds of the under-cleaned and over-cleaned regions, construct a process cloud library, and automatically optimize the machining parameters of complex curved surfaces.

[0089] Furthermore, based on the above-mentioned multimodal sensing-based adaptive coarse beam opening method for large castings, the present invention provides a second embodiment of a multimodal sensing-based adaptive coarse beam opening system for large castings, used to implement the above-described adaptive coarse beam opening method for large castings, comprising: The actuator, including a five-axis industrial robotic arm and an external travel axis, is used to drive the end effector to any position on the surface of the casting. The processing unit includes a high-power pulsed laser generator and a laser galvanometer mounted on the end effector, used to generate high-energy pulsed lasers and focus them to scan, break up and peel off oxide scale and adhering sand; The sensing unit includes a 3D vision sensor mounted on the end effector, used to acquire three-dimensional topography and texture information of the casting surface; The auxiliary unit includes a positive pressure purging system mounted on the actuator for blowing away loose sand particles that have been broken up by the laser on the surface of the casting; The system includes a control unit, comprising an industrial control computer and a robotic arm controller. The industrial control computer is used for motion control, parameter adjustment, data processing, and issuing control commands to the actuators, processing units, sensing units, and auxiliary units. The robotic arm controller is used to receive commands from the industrial control computer and control the movement of the actuators.

[0090] In summary, based on the above scheme, by precisely controlling the laser energy density between the sand removal threshold and the substrate ablation threshold, and performing real-time normal energy compensation, the energy density fluctuation across the entire surface is ensured to be less than ±10%, the heat-affected zone to be less than 5μm, and the substrate removal amount on one side to be less than 0.05mm. Through intelligent zoning and path optimization, the effective processing time of the robotic arm is increased (to over 78%), the rework rate is reduced (to below 3%), and the overall cycle time is improved by more than 40% compared to traditional pneumatic grinding. Simultaneously, 3D point cloud data is used as the acceptance carrier to achieve quantitative judgment of the cleaning effect and full-process data traceability. Relying on the AI ​​incremental learning engine, the process parameters are automatically optimized through a differential closed-loop mechanism, which significantly reduces the cost of introducing and debugging new processes and the workload of on-site manual debugging.

[0091] To better understand and implement the technical solution of this application, the present invention will be further described in detail below with reference to specific embodiments. Specific implementation examples:

[0093] 1. Object: High-pressure outer cylinder of 1000MW steam turbine (ZG15Cr1Mo1V).

[0094] 2. Operating Condition Settings: Material: ZG15Cr1Mo1V (low alloy heat-resistant steel); Cleaning area: Approximately 18m²; Machining allowance on one side: 5mm; Objective: To remove a 2mm thick oxide scale and zircon sand sintered layer.

[0095] 3. Configuration of Laser Adaptive Coarse Shot System for Large Castings Actuator: ABB IRB 6700 robotic arm, equipped with a 6-meter sixth axis travel axis; Processing unit: 1500W MOPA pulsed fiber laser, peak power 100kW, 1064nm pulsed laser; Laser galvanometer, field lens focal length F=330mm, scanning field diameter φ50mm; Sensing unit: Keyence LJ-X series 3D structured light camera, scanning accuracy ≤0.05mm. Control unit: Industrial PC (IPC / PLC) with laser process control software developed based on C++, and all adaptive algorithms of this invention are built in.

[0096] 4. Rough consecration process Step 1: Global Multimodal Sensing Scan The robotic arm, carrying a 3D camera, traverses the workpiece surface at a speed of 500 mm / s to collect complete point clouds; the industrial control computer registers the measured point cloud with the workpiece CAD model, calculates the residual ΔZ, and generates a heat map of sand adhesion. The maximum sand thickness in the center flange area is identified as 2.8 mm (re-sprayed area), and the sand thickness on the cylinder wall side is 0.6 mm (thin sand flat area); the surface normal and laser incident angle at each point are calculated in real time.

[0097] Step 2: Intelligent Zoning and Differentiated Path Planning Automatic zoning of workpiece areas: the cylinder top large plane is divided into a flat area, the cylinder side wall into a steep area, and the intake and exhaust arc into a near-vertical area; a 35% overlap grating path is generated in the flat area, with a reference speed V=150mm / s; the steep side wall area is scanned in equal-height layers at a speed of 0.6V; the arc cavity is spiral scanned at a speed of 0.45V; and the high-power parameter group is marked in the center split surface re-spraying area.

[0098] Step 3: Adaptive laser cleaning + gas-photonic synergistic purging The robotic arm performs laser ablation cleaning along the planned path: When the robotic arm is tilted and the laser incident angle θ=60°, the system automatically increases the base power from 1000W to 1180W according to the normal energy compensation model to ensure that the energy of the facade is consistent with that of the top surface. 50ms before reaching the 2.8mm re-spraying zone on the center face, the industrial control computer outputs a high level, and the purging system switches to 0.8MPa high-pressure airflow to pre-loosen the dense zircon sand; the laser synchronously increases the pulse frequency to 80kHz and the pulse width to 300ns to enhance the crushing effect of the thick sand layer; the regular cylinder wall area maintains 0.2MPa purging to promptly remove smoke and dust.

[0099] Step 4: In-situ closed-loop re-inspection and automatic re-scan After cleaning a single area, pause for 0.5 seconds and the 3D camera rescans in place. After the entire area is processed, re-inspection is performed. A 0.25mm residual sand black spot is detected. The system automatically generates a re-scanning path and performs fine cleaning at 50% of the reference power and low speed. The re-inspection shows that all residual errors are less than 0.1mm, and the process ends.

[0100] 5. Implementation Results The total time for rough polishing of a single cylinder is 4.2 hours; the base metal loss is less than 0.03mm, with no micro-melting pits or hot cracks; the surface after cleaning meets the ISO8501 Sa2.5 standard, with no residual sand adhering; no need for manual secondary polishing, the rework rate is 0, and the overall processing efficiency is 42% higher than that of traditional mechanical polishing.

[0101] It should be understood that the systems and modules described above can be implemented in various ways. For example, in some embodiments, the systems and modules can be implemented using hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated hardware. Those skilled in the art will understand that the methods and systems described above can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The systems and modules of this application can be implemented not only with hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., but also with software, for example, executed by various types of processors, or with a combination of the aforementioned hardware circuits and software (e.g., firmware).

[0102] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects may be any one or a combination of the above, or any other possible beneficial effects.

[0103] The basic concepts have been described above. Obviously, for those skilled in the art, the detailed disclosure above is merely illustrative and does not constitute a limitation of this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this application. Such modifications, improvements, and corrections are suggested in this application, and therefore remain within the spirit and scope of the exemplary embodiments of this application.

[0104] Furthermore, this application uses specific terms to describe embodiments of the application. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of the application. Therefore, it should be emphasized and noted that "an embodiment," "one embodiment," or "an alternative embodiment" mentioned twice or more in different locations in this specification do not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of the application can be appropriately combined.

[0105] Furthermore, those skilled in the art will understand that aspects of this application can be described and illustrated through several patentable types or situations, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Accordingly, aspects of this application can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. All of the above hardware or software may be referred to as a “data block,” “module,” “engine,” “unit,” “component,” or “system.” Furthermore, aspects of this application may manifest as a computer product located on one or more computer-readable media, the product including computer-readable program code.

[0106] Computer storage media may contain a propagated data signal containing computer program code, for example, on baseband or as part of a carrier wave. This propagated signal may take various forms, including electromagnetic, optical, and suitable combinations thereof. Computer storage media can be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, apparatus, or device to enable communication, propagation, or transmission of a program for use. The program code located on the computer storage medium can be propagated through any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0107] The computer program code required for the operation of each part of this application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby, and Groovy, or other programming languages. This program code can run entirely on the user's computer, or as a standalone software package on the user's computer, or partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any network, such as a local area network (LAN) or wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as Software as a Service (SaaS).

[0108] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this application are not intended to limit the order of the processes and methods of this application. Although the foregoing disclosure has discussed some currently considered useful embodiments of the invention through various examples, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the substance and scope of the embodiments of this application. For example, while the system components described above can be implemented using hardware devices, they can also be implemented solely through software solutions, such as installing the described system on existing servers or mobile devices.

[0109] Similarly, it should be noted that, in order to simplify the description of the present application and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of the embodiments of the present application sometimes combines multiple features into a single embodiment, drawing, or description thereof. However, this disclosure method does not imply that the subject matter of the application requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of the single embodiments disclosed above.

[0110] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are open to adaptive variation. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters are taken into account a specified number of significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of application in some embodiments of this application are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0111] For each patent, patent application, patent application publication, and other material such as articles, books, specifications, publications, and documents referenced in this application, the entire contents of that patent are incorporated herein by reference. This excludes historical application documents that are inconsistent with or conflict with the content of this application, as well as documents that limit the broadest scope of the claims of this application (currently or subsequently appended to this application). It should be noted that in the event of any inconsistency or conflict between the descriptions, definitions, and / or terminology used in the supplementary materials of this application and the content of this application, the descriptions, definitions, and / or terminology used in this application shall prevail.

[0112] Finally, it should be understood that the embodiments described in this application are merely illustrative of the principles of the embodiments of this application. Other modifications may also fall within the scope of this application. Therefore, alternative configurations of the embodiments of this application are considered as examples and not limitations, and are regarded as consistent with the teachings of this application. Accordingly, the embodiments of this application are not limited to the embodiments explicitly described and illustrated in this application.

[0113] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A laser adaptive coarse-opening method for large castings based on multimodal sensing, characterized in that, It includes the following steps: The casting in a horizontal position is scanned in its entirety using a 3D vision sensor mounted on the end effector of the robotic arm to obtain the measured point cloud Preal. The measured point cloud Preal and the theoretical CAD model Mcad are registered using the iterative nearest point algorithm, and the residual ΔZ between the two is calculated to generate a heat map T(x,y) of the sand thickness. The joint angles fed back by the robotic arm controller are read in real time, and the normal vector n of the end flange of the robotic arm is calculated through forward kinematic transformation. Based on a high-power pulsed laser generator mounted on the end effector of a robotic arm, the angle θ between the laser beam emission direction and the normal to the surface of the casting is obtained, wherein the angle θ ranges from 0° to 90°. Based on the aforementioned heat map of sand adhesion thickness T(x,y), the normal vector n at the end of the robotic arm, and the included angle θ, the area to be processed in the casting is divided into zones. Based on the partition information, a laser processing path is generated using the corresponding partition scanning trajectory algorithm and processing parameters. Then, based on the processing parameters and laser processing path, the corresponding area is cleaned until all partitions of the workpiece have completed rough opening.

2. The laser adaptive coarse beam opening method for large castings according to claim 1, characterized in that, Also includes: After completing the laser processing of an area, the area is re-scanned using a 3D vision sensor. The three-dimensional topography data obtained from the re-scan is compared with the theoretical CAD model Mcad, and the surface roughness Ra and height residual ΔH after cleaning are calculated. Set a residual judgment threshold. If the height residual ΔH is greater than the residual judgment threshold, then store the coordinates of the area in the supplementary scan queue. After the current cycle ends, the robotic arm performs the supplementary scanning based on the supplementary scanning queue information, with reduced power and speed.

3. The laser adaptive coarse beam opening method for large castings according to claim 1, characterized in that, The iterative nearest-point algorithm is used to register the measured point cloud Preal with the theoretical CAD model Mcad, and the residual ΔZ between the two is calculated to generate a heat map T(x,y) of the sand-coated thickness. Specifically: The residual ΔZ is calculated using the following formula: ΔZ=Zreal−Zcad Where Zreal represents the actual Z-coordinate of a point as measured by the 3D vision sensor; Zcad represents the theoretical Z-coordinate of that point on the CAD model Mcad; ΔZ represents the deviation between the measured value and the theoretical value. If the thickness threshold of the casting is set to δ, then: If ΔZ>δ, then the corresponding area of ​​the casting is determined to be a sand-adhering area, and a heat map of sand adhesion thickness T(x,y) is generated.

4. The laser adaptive coarse beam opening method for large castings according to claim 1, characterized in that, The process involves real-time reading of joint angle information from the robotic arm controller, followed by calculation of the normal vector n of the robotic arm's end flange using forward kinematics transformation. Specifically: The normal vector n of the end flange is calculated using the following forward kinematics transformation formula. T =A1(θ1)·A2(θ2)·A3(θ3)·...·An(θn) Where T represents the coordinate transformation from the base to the end flange; A1-An represents the length of the 1-n segments of the robotic arm; and θ1-θn represents the joint angle of the 1-n segments of the robotic arm.

5. The laser adaptive coarse beam opening method for large castings according to claim 1, characterized in that, The high-power pulsed laser generator mounted on the end effector of the robotic arm is used to obtain the angle θ between the laser beam emission direction and the normal to the casting surface. Based on the heat map of sand adhesion thickness T(x,y), the normal vector n of the robotic arm end effector, and the angle θ, the area to be processed in the casting is divided into zones, specifically: The collected point cloud data is filtered and denoised, and the surface normal vector of each point is calculated by neighborhood fitting; Based on the surface normal vector and the laser emission direction, the angle θ between the laser emission direction and the workpiece surface normal is obtained; Based on the included angle θ, the tolerance of the normal included angle of 5°~8°, and the proximity distance of 2~3 times the laser spot diameter, the discrete point cloud is merged into a connected processing region, and the processing region is divided as follows using a region growing algorithm: When θ < 30°, it is determined to be a flat region; When 30°≤θ<70°, it is determined to be a steep region; When θ ≥ 70°, it is determined to be a near-vertical region; Among them, fragmented areas with an area less than 10 times the square of the laser spot diameter are removed and merged into the nearest processing area or marked as an independent processing area.

6. The laser adaptive coarse beam opening method for large castings according to claim 5, characterized in that, The process involves generating a laser processing path based on partition information, using the corresponding partition's scanning trajectory algorithm and processing parameters, and then performing cleanup on the corresponding area based on the processing parameters and the laser processing path. Specifically: Execute based on partition information: A. If the area is flat, then the raster scan mode will be used for cleaning, specifically: Set the overlap ratio between two adjacent raster scan lines to 30-40%; Generate reciprocating parallel scan lines and project the scan lines back onto the 3D surface; Set the feed speed of the high-power pulsed laser generator to the reference value speed V; B. If the area is steep, the contour line scanning mode will be used for cleaning, specifically: Determine the height direction information of the high-power pulsed laser generator, wherein the height direction is the Z-axis of the world coordinate system or the normal of the fitted plane; Based on the average pre-compression layer height within the region with an included angle θ, the laser row spacing is set to 0.65 × spot diameter × cosθ; The slices are layered along the height direction, and the cross-sectional profile of each layer is offset inward to form a concentric ring scanning path. The scanning direction of adjacent rings is set to alternate between forward and reverse scanning to reduce thermal deformation. The feed rate of the high-power pulsed laser generator is set to 60% of the reference speed V. C. If the area is near vertical, use a spiral or strip scanning mode for cleaning, specifically: Use a combination of spiral or strip scanning; the overlap rate between two adjacent scanning paths is 50%, and the feed rate is adjusted to 40%–50% of the reference speed V. Among them, at the junction of each processing area, the boundary is expanded outward by 1 to 2 times the spot area as a transition zone. A linear interpolation strategy is adopted in the transition zone, the regional overlap rate is increased by 50%, and the feed speed and laser power are smoothly transitioned from the parameters of adjacent areas. And during execution: If the maximum residual value within the partition does not exceed 0.5mm, the power is adjusted to 50% of the rated power of the laser beam of the high-power pulsed laser generator, the speed is adjusted to 90% of the rated operating speed of the robotic arm, and the overlap ratio between two adjacent grating scan lines is set to 20%~30%. If the maximum residual value within the partition is greater than 0.5 and less than or equal to 2mm, the power is adjusted to 75% of the rated power of the laser beam, the speed is adjusted to 60% of the rated operating speed of the robotic arm, and the overlap ratio between two adjacent grating scan lines is set to 30%~40%. If the maximum residual value within the partition is greater than 2mm and less than 4mm, the power is adjusted to 90% of the rated power of the laser beam, the speed is adjusted to 30% of the rated operating speed of the robotic arm, and the overlap ratio between two adjacent grating scan lines is set to 40%~50%.

7. The laser adaptive coarse beam opening method for large castings according to claim 1, characterized in that, It also includes a dynamic parameter adaptive adjustment step, which includes: Based on the angle θ between the laser beam and the normal of the casting surface calculated in real time, the laser output power Pout is dynamically adjusted through the normal energy compensation model to maintain a constant energy input per unit area. The normal energy compensation model is as follows: Pout=Pbase×(1+k×|cos(θ)|⁻¹) Where Pout is the actual output power, Pbase is the base power of the laser beam, and k is the compensation coefficient.

8. The laser adaptive coarse beam opening method for large castings according to claim 1, characterized in that, It also includes a positive pressure purging coordination step, which specifically includes: When the laser beam enters the non-re-spraying zone, the positive pressure purging system is triggered to start with the first pressure and purge the laser polishing area. A preset time before the laser beam enters the re-spraying zone, the positive pressure purging system is triggered to start with a second pressure, using high-pressure airflow to loosen the dense sand layer. in, The second pressure is greater than the first pressure; The respray zone is defined as: Based on the heat map T(x,y) of the sand thickness, areas with a thickness exceeding the set threshold are marked as re-spraying zones; when the laser beam enters the re-spraying zone, the rated power of the laser beam for the re-spraying zone is automatically increased to 100%-110%, and the speed is adjusted to 15% of the robot's rated operating speed.

9. The laser adaptive coarse beam opening method for large castings according to claim 1, characterized in that, It also includes AI incremental learning steps: Establish a differential closed-loop mechanism between the measured point cloud (Preal) and the finished clean point cloud, and construct a high-precision differential model between the initial state point cloud and the finished clean point cloud for complex curved surfaces. The cleaned 3D topography is compared with the initial state by using a normal alignment algorithm to automatically generate an RGB pseudo-color residual heatmap. Based on residual heatmaps, under-cleaned and over-cleaned regions are identified and marked, and the energy density threshold is adaptively fine-tuned through multidimensional correlation analysis. Based on the residual data, iterate the energy density thresholds of under-cleaned and over-cleaned regions, construct a cloud-based process library, and automatically optimize the machining parameters of complex curved surfaces.

10. A laser adaptive coarse beam opening system for large castings based on multimodal sensing, used to implement the laser adaptive coarse beam opening method for large castings as described in any one of claims 1 to 9, characterized in that, include: The actuator, including a five-axis industrial robotic arm and an external travel axis, is used to drive the end effector to any position on the surface of the casting. The processing unit includes a high-power pulsed laser generator and a laser galvanometer mounted on the end effector, used to generate high-energy pulsed lasers and focus them to scan, break up and peel off oxide scale and adhering sand; The sensing unit includes a 3D vision sensor mounted on the end effector, used to acquire three-dimensional topography and texture information of the casting surface; The auxiliary unit includes a positive pressure purging system mounted on the actuator for blowing away loose sand particles that have been broken up by the laser on the surface of the casting; The system includes a control unit, comprising an industrial control computer and a robotic arm controller. The industrial control computer is used for motion control, parameter adjustment, data processing, and issuing control commands to the actuators, processing units, sensing units, and auxiliary units. The robotic arm controller is used to receive commands from the industrial control computer and control the movement of the actuators.

Images