Sand blasting mechanical arm based on multi-sensor sensing, control system and method

Abstract

The invention relates to the technical field of steel structure surface treatment, in particular to a sand blasting mechanical arm based on multi-sensor sensing and a control system and method. A multi-sensor sensing module based on a sand blasting mechanical arm is constructed. The multi-sensor sensing module comprises a pressure sensor, a distance measuring sensor, a joint motor encoder and a depth camera which are located on the sand blasting mechanical arm. Collecting pressure data in the sand blasting pipe in real time through a pressure sensor, and performing abnormality judgment in combination with a preset normal pressure interval; the swing angular velocity of the spray gun is detected through a joint motor encoder, and motion parameters of the spray gun are dynamically adjusted in combination with sand blasting surface images collected by a depth camera in real time; according to the method, the depth cameras which are symmetrically arranged are used for carrying out image collection on a sand blasting area, the surface cleanliness is judged through a pre-trained image classification model, the surface roughness Sa value is calculated through point cloud residual errors, the quality fluctuation problem caused by blind adjustment of traditional technological parameters is solved, and the sand blasting uniformity is improved by 40% or above.

Description

Technical Field

[0001] This invention relates to the field of steel structure surface treatment technology, specifically to a sandblasting robotic arm, control system, and method based on multi-sensor perception. Background Technology

[0002] In the field of sandblasting, the position and oscillation speed control of traditional spray guns are mostly manually adjusted or fixed parameters, which cannot be dynamically adjusted. This leads to uneven sandblasting, and large-sized workpieces are prone to problems such as difficulty in locating local defects, over-blasting or under-blasting of complex curved surfaces, affecting the processing pass rate and hindering the development of the industry. Taking the collaborative control method of UAV-robotic arm system for precision spraying provided by Chinese Patent Publication No. CN120479640A as an example, when processing complex curved workpieces, the sandblasting quality often fluctuates greatly due to the lack of effective perception and precise control. Specifically, it is difficult to accurately grasp the pressure inside the tube in real time during sandblasting. When the pressure is abnormal, it is not possible to find the root cause in time. Problems such as poor material supply, nozzle wear, or abrasive caking can cause the pressure to be too low, slowly decrease, or fluctuate violently, affecting the stability of sandblasting dynamics and resulting in inconsistent results. In addition, manual quality inspection is inefficient and has large errors, making it difficult to accurately assess surface cleanliness and roughness. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a sandblasting robotic arm, control system and method based on multi-sensor perception.

[0004] The technical solution adopted in this invention is as follows: A sandblasting method based on multi-sensor perception includes the following steps: S1. Construct a multi-sensor perception module based on a sandblasting robotic arm. The multi-sensor perception module includes a pressure sensor, a range sensor, a joint motor encoder, and a depth camera located on the sandblasting robotic arm. S2. Real-time pressure data inside the sandblasting pipe is collected via a pressure sensor, and anomalies are determined based on a preset normal pressure range: If the pressure value is abnormal, perform a pressure abnormality type diagnosis, generate the corresponding diagnostic results, and trigger a system early warning prompt for maintenance. If the pressure value is normal, the distance sensor is activated to detect the distance between the spray gun and the workpiece surface in real time, and the telescopic arm is driven to automatically adjust to the preset distance range. S3. The spray gun swing angular velocity is detected by the joint motor encoder, and the spray gun motion parameters are dynamically adjusted by combining the sandblasting surface image collected in real time by the depth camera. S4. Use symmetrically arranged depth cameras to acquire images of the sandblasting area. After filtering, cropping and feature extraction, use a pre-trained image classification model to determine the surface cleanliness and calculate the surface roughness Sa value through point cloud residuals. If the cleanliness / roughness meets the standards, record the quality data and archive it; If the cleanliness / roughness does not meet the standard, a quality alarm will be triggered and the workpiece position will be marked. After adjusting the sandblasting parameters, a re-blasting operation will be performed.

[0005] This technical solution utilizes multimodal sensor data fusion and diagnostic algorithms to identify the root causes of abnormal sandblasting pressure and self-correct motion parameters, thereby more efficiently addressing the problems of large quality fluctuations and low efficiency of manual quality inspection in sandblasting of complex curved surfaces. Specifically, pressure sensors monitor the dynamic stability of sandblasting, distance sensors ensure the accuracy of sandblasting distance, joint encoders provide motion trajectory feedback, and depth cameras enable visual perception of surface quality. After analysis by the control and management module, the data from each sensor forms a three-dimensional monitoring network for sandblasting pressure fluctuations, spatial position deviations, and surface treatment effects. Combined with anomaly diagnosis algorithms and visual quality assessment models, a complete "perception-decision-execution-verification" control system is constructed, resolving the efficiency and quality contradictions caused by blind parameter adjustment and manual quality inspection in traditional sandblasting processes.

[0006] In addition, the sandblasting method based on multi-sensor perception proposed according to the present invention may also have the following additional technical features: According to one embodiment of the present invention, the abnormality types of the pressure sensor of S2 include low pressure, slow pressure decrease, drastic pressure fluctuation, or high pressure; the diagnostic results are divided into poor material supply, nozzle wear, abrasive caking, or nozzle blockage, wherein: low pressure corresponds to poor material supply or nozzle blockage; slow pressure decrease corresponds to abrasive caking; and drastic pressure fluctuation corresponds to nozzle wear.

[0007] This technical solution establishes a mapping relationship between abnormal pressure characteristics and equipment malfunctions by comparing real-time monitoring data from pressure sensors with preset pressure modes. For example, excessively low pressure may be due to insufficient fluid power caused by a faulty feeding system or nozzle blockage; a slow decrease in pressure reflects the gradual caking of abrasive particles in the feeding pipeline, leading to a gradual increase in flow resistance; and drastic pressure fluctuations are caused by unstable fluid jetting due to nozzle wear.

[0008] According to one embodiment of the present invention, when any of the pressure sensor and distance sensor of S2, the joint motor encoder of S3, and the depth camera of S4 are abnormal, the sandblasting operation is immediately interrupted, the abnormal data is recorded synchronously, and a traceable maintenance log is generated.

[0009] This technical solution achieves rapid response to local faults through a multi-sensor abnormal linkage interruption and data tracing mechanism, thereby more effectively solving the problem of delayed fault detection and ambiguous responsibility positioning in traditional single-point monitoring solutions, which leads to increased production losses.

[0010] According to one embodiment of the present invention, the cleanliness of S4 is evaluated by training an RGB image classification model, and the roughness is evaluated by calculating the Sa value from the point cloud residual and comparing it with a standard sample. Both are used as the basis for quality judgment.

[0011] In this technical solution, the RGB image classification model extracts features such as texture and color difference of the sandblasted surface through a convolutional neural network and maps them to cleanliness levels, such as Sa2.5 / Sa3, to achieve rapid and large-scale detection. The point cloud residual calculation uses the three-dimensional coordinate data obtained by the depth camera and quantifies the micro-geometric error through the surface height deviation statistic, i.e., the Sa value. Combined with the comparison of the digital model of the standard sample, the physical consistency of the roughness assessment is ensured. The two together make up for the lack of information from a single sensor and solve the contradiction that visual assessment is easily affected by lighting interference and three-dimensional assessment has high computational complexity.

[0012] To achieve the above objectives, the present invention also provides a sandblasting control system based on multi-sensor perception.

[0013] A sandblasting control system based on multi-sensor perception includes the following modules: The multi-sensor sensing module integrates a pressure sensor, a distance sensor, a joint motor encoder, and symmetrically arranged depth cameras to acquire real-time data on sandblasting pressure, spray gun distance, swing angular velocity, and surface image data. The anomaly diagnosis module determines the anomaly type based on pressure data, generates diagnostic results, triggers system alerts, and pushes maintenance prompts. The motion control module uses the feedback from the ranging sensor to drive the telescopic arm to adjust the position of the spray gun, and combines encoder data to dynamically adjust the angular velocity of the spray gun swing, thereby achieving closed-loop control of motion parameters. The visual inspection module filters, crops, and extracts features from images acquired by a depth camera, evaluates cleanliness using a pre-trained classification model, calculates the roughness Sa value using point cloud residuals, and compares it with a standard sample. The control and management module coordinates the operation of each module, records quality data and abnormal events, and generates traceable logs; when any sensor parameter is abnormal, the operation is immediately interrupted and the data is archived.

[0014] This technical solution achieves real-time calibration of process parameters and proactive prevention and control of quality risks through multi-sensor data fusion and modular intelligent algorithms, thereby more efficiently solving the problems of frequent manual intervention, delayed parameter adjustments, and large quality fluctuations. Specifically, process parameters are collected in real time through multi-sensor perception modules such as pressure, distance, angular velocity, and vision, constructing a digital twin mapping of the sandblasting process; the anomaly diagnosis module uses pressure feature pattern recognition to achieve fault prediction; the motion control module optimizes trajectory accuracy through position-velocity dual closed loops; the vision detection module combines image classification and 3D point cloud quantification of surface quality; and finally, the control management module achieves data integration and anomaly linkage control, solving the problems of dynamic stability, motion accuracy, and surface consistency in the sandblasting process.

[0015] According to one embodiment of the present invention, the abnormal diagnosis module has a built-in pressure threshold range [P] Low ,P High The pressure change rate threshold is used to determine whether the pressure is slowly decreasing or fluctuating drastically by analyzing the slope of the real-time pressure curve, and to predict the abrasive supply status by combining historical data.

[0016] This technical solution achieves real-time perception and fault prediction of abrasive supply status through pressure dynamic characteristic analysis and historical data fusion prediction, thereby more efficiently solving the problems of response lag and high misjudgment rate caused by the reliance on manual inspection in traditional methods.

[0017] According to one embodiment of the present invention, the visual inspection module employs a symmetrical deployment of depth cameras to cover both sides of the sandblasting area, generates a full surface quality map through image stitching technology, marks the coordinates of substandard areas, and feeds them back to the motion control module.

[0018] This technical solution achieves parallel data acquisition of the sandblasting areas on both sides through a symmetrical layout of depth cameras, and eliminates blind spots caused by single-view occlusion using stereo vision algorithms. The image stitching technology, through feature point matching and spatial transformation matrix calculation, fuses multi-view images into a full-surface quality map under a unified coordinate system, and constructs a correlation mapping between the three-dimensional geometric model of the sandblasted surface and pixel-level quality annotations. After receiving the coordinates of the substandard areas, the motion control module converts the spatial position into joint motor commands through inverse kinematics algorithms, forming a closed-loop control, which solves the problem of difficult-to-accurate positioning of local quality defects in large-sized workpieces.

[0019] According to one embodiment of the present invention, the motion control module employs a PID control algorithm, based on the real-time distance fed back by the ranging sensor and the preset range [D]. Min D Max To compensate for any deviation, dynamically adjust the telescopic arm's moving speed to ensure a stable distance between the spray gun and the workpiece surface.

[0020] This technical solution achieves dynamic adaptive adjustment of the spray gun distance through the collaboration of PID algorithm and distance measurement feedback, thereby more efficiently solving the problems of uneven sandblasting and surface quality fluctuation caused by manual adjustment lag or fixed parameter control.

[0021] To achieve the above objectives, the present invention also provides a sandblasting robotic arm based on multi-sensor perception.

[0022] A sandblasting robotic arm based on multi-sensor perception includes the following structure: The drive arm structure includes a fixed arm and a telescopic arm arranged in sequence. The telescopic arm is driven by a ball screw and a servo motor, and slides along the cavity of the fixed arm to achieve axial extension and retraction. The three-degree-of-freedom joint includes joint I, joint II and joint III arranged in sequence. Joint I is installed at the end of the telescopic arm and supports ±90° rotation. Joint II is installed at the end of joint I and supports 360° continuous rotation. Joint III is installed at the end of joint II and supports ±90° swing. Each joint has a built-in joint motor encoder. Depth cameras are symmetrically arranged on both sides of joint I. A distance sensor is integrated at the front of joint III. The spraying structure includes a left sandblasting pipe and a right sandblasting pipe, as well as a left spray gun and a right spray gun; the outlet of the left sandblasting pipe is clamped to the left side of joint III by a left clamp, and the outlet of the right sandblasting pipe is clamped to the right side of joint III by a right clamp. The left spray gun and the right spray gun are respectively connected to the left sandblasting pipe and the right sandblasting pipe through quick connectors.

[0023] This technical solution achieves high-precision axial extension and retraction through ball screw transmission of the drive arm, realizes rotation-slewing-oscillation control through a three-degree-of-freedom joint, and provides angle feedback to form a position closed loop through the joint motor encoder. Depth cameras are symmetrically arranged on both sides of joint I to acquire the three-dimensional shape of the workpiece surface using stereo vision. A distance sensor is integrated into the front of joint III to realize real-time compensation of the distance between the spray gun and the workpiece. The symmetrical layout of the dual sandblasting pipes is merged into a single spray gun through a quick connector. The data from the robotic arm and multiple sensors are fused to calculate the optimal sandblasting trajectory to adapt to complex curved surfaces, solving the over-spraying / under-spraying problem caused by the lack of perception in traditional robotic arms.

[0024] According to one embodiment of the present invention, the inlets of the left and right sandblasting pipes are clamped to the side of the fixed arm, and both the inlets of the left and right sandblasting pipes are equipped with pressure sensors, and both the left and right sandblasting pipes are equipped with explosion-proof steel wire braided layers.

[0025] This technical solution uses a pressure sensor and an explosion-proof pipe body to achieve real-time monitoring of sandblasting pressure and overpressure protection, thereby more efficiently solving the problem of pipe burst risk and processing interruption caused by uncontrolled pressure in traditional sandblasting pipes.

[0026] Compared with the prior art, the present invention has the following advantages: (1) The sandblasting method of the present invention achieves accurate identification of pressure abnormalities and self-correction of motion parameters through multi-sensor data fusion and intelligent diagnostic algorithm. For pressure sensors, it can distinguish fault types such as low pressure and slow drop. Combined with distance sensor and joint encoder, it dynamically adjusts the position and swing speed of spray gun, solves the quality fluctuation problem caused by blind adjustment of traditional process parameters, and improves the sandblasting uniformity by more than 40%. (2) The control system constructs a closed-loop control through modularization. The abnormal diagnosis module uses pressure threshold and change rate to predict the abrasive supply status. The visual inspection module uses RGB image classification and point cloud residual calculation to evaluate cleanliness and roughness. The control management module realizes multi-sensor abnormal linkage interruption and data traceability, reducing the frequency of manual intervention by 75% and shortening the response time of quality accidents to within 10 seconds. (3) The sandblasting robot arm achieves high-precision trajectory control through a three-degree-of-freedom joint and a ball screw drive arm. The joint motor encoder provides 0.1° level angle feedback. The symmetrically arranged depth camera covers the sandblasting area on both sides. Combined with image stitching technology, it generates a full surface quality map. The drive telescopic arm uses a PID algorithm to stabilize the spray gun distance, solves the problem of over-spraying / missed spraying on complex curved surfaces, and increases the pass rate of large-size workpieces from 65% to 92%. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the principle of the method of the present invention.

[0028] Figure 2 This is a schematic diagram of the control system of the present invention.

[0029] Figure 3 This is the front view of the robotic arm of the present invention.

[0030] Figure 4 This is a top view of the robotic arm of the present invention.

[0031] In the diagram: 1. Fixed arm; 2. Telescopic arm; 3. Joint I; 4. Joint II; 5. Joint III; 6. Left clamp; 7. Left spray gun; 8. Left sandblasting pipe; 9. Right clamp; 10. Right spray gun; 11. Right sandblasting pipe; 12. Distance sensor; 13. Depth camera; 14. Pressure sensor. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Example 1 like Figure 1 As shown, this embodiment provides a sandblasting method based on multi-sensor perception, including the following steps: S1. Construct a multi-sensor sensing module based on a sandblasting robotic arm. The multi-sensor sensing module includes a pressure sensor 14, a range sensor 12, a joint motor encoder, and a depth camera 13 located on the sandblasting robotic arm. S2. Real-time pressure data inside the sandblasting pipe is collected via pressure sensor 14, and anomalies are determined based on the preset normal pressure range: If the pressure value is abnormal, perform a pressure abnormality type diagnosis, generate the corresponding diagnostic results, and trigger a system early warning prompt for maintenance. If the pressure value is normal, the distance sensor 12 is activated to detect the distance between the spray gun and the workpiece surface in real time, and the telescopic arm 2 is driven to automatically adjust to the preset distance range. S3. The spray gun swing angular velocity is detected by the joint motor encoder, and the spray gun motion parameters are dynamically adjusted by combining the sandblasting surface image collected in real time by the depth camera 13. S4. Use symmetrically arranged depth cameras 13 to acquire images of the sandblasting area. After filtering, cropping and feature extraction, the surface cleanliness is judged by a pre-trained image classification model, and the surface roughness Sa value is calculated by point cloud residual. If the cleanliness / roughness meets the standards, record the quality data and archive it; If the cleanliness / roughness does not meet the standard, a quality alarm will be triggered and the workpiece position will be marked. After adjusting the sandblasting parameters, a re-blasting operation will be performed.

[0034] This technical solution utilizes multimodal sensor data fusion and diagnostic algorithms to identify the root causes of abnormal sandblasting pressure and self-correct motion parameters, thereby more efficiently addressing the problems of large quality fluctuations and low efficiency of manual quality inspection in sandblasting of complex curved surfaces. Specifically, pressure sensor 14 monitors the dynamic stability of sandblasting, distance sensor 12 ensures the accuracy of sandblasting distance, joint encoder provides motion trajectory feedback, and depth camera 13 enables visual perception of surface quality. After analysis by the control and management module, the data from each sensor forms a three-dimensional monitoring network for sandblasting pressure fluctuations, spatial position deviations, and surface treatment effects. Combined with anomaly diagnosis algorithms and visual quality assessment models, a complete "perception-decision-execution-verification" control system is constructed, resolving the efficiency and quality contradictions caused by blind parameter adjustment and manual quality inspection in traditional sandblasting processes.

[0035] In addition, the sandblasting method based on multi-sensor perception proposed according to the present invention may also have the following additional technical features: According to one embodiment of the present invention, the abnormality types of the pressure sensor 14 of S2 include low pressure, slow pressure decrease, drastic pressure fluctuation, or high pressure; the diagnostic results are divided into poor material supply, nozzle wear, abrasive caking, or nozzle blockage, wherein: low pressure corresponds to poor material supply or nozzle blockage; slow pressure decrease corresponds to abrasive caking; and drastic pressure fluctuation corresponds to nozzle wear.

[0036] This technical solution establishes a mapping relationship between abnormal pressure characteristics and equipment failures by comparing real-time monitoring data from pressure sensor 14 with preset pressure modes. For example, excessively low pressure may be due to insufficient fluid power caused by a fault in the feeding system or nozzle blockage; a slow decrease in pressure reflects the gradual caking of abrasive particles in the feeding pipeline, leading to a gradual increase in flow resistance; and drastic pressure fluctuations are caused by unstable fluid jetting due to nozzle wear.

[0037] According to one embodiment of the present invention, when any of the pressure sensor 14 and distance sensor 12 of S2, the joint motor encoder of S3, and the depth camera 13 of S4 are abnormal, the sandblasting operation is immediately interrupted, the abnormal data is recorded synchronously, and a traceable maintenance log is generated.

[0038] This technical solution achieves rapid response to local faults through a multi-sensor abnormal linkage interruption and data tracing mechanism, thereby more effectively solving the problem of delayed fault detection and ambiguous responsibility positioning in traditional single-point monitoring solutions, which leads to increased production losses.

[0039] According to one embodiment of the present invention, the cleanliness of S4 is evaluated by training an RGB image classification model, and the roughness is evaluated by calculating the Sa value from the point cloud residual and comparing it with a standard sample. Both are used as the basis for quality judgment.

[0040] In this technical solution, the RGB image classification model extracts features such as sandblasted surface texture and color difference through a convolutional neural network and maps them to cleanliness levels, such as Sa2.5 / Sa3, to achieve rapid and large-scale detection. The point cloud residual calculation uses the three-dimensional coordinate data obtained by the depth camera 13 and quantifies the micro-geometric error through the surface height deviation statistic, i.e., the Sa value. Combined with the comparison of the digital model of the standard sample, the physical consistency of the roughness assessment is ensured. The two together make up for the lack of information from a single sensor and solve the contradiction that visual assessment is easily affected by lighting interference and three-dimensional assessment has high computational complexity.

[0041] Example 2 Based on Example 1, such as Figure 2 As shown, this embodiment provides a sandblasting control system based on multi-sensor perception, including the following modules: The multi-sensor sensing module integrates a pressure sensor 14, a distance sensor 12, a joint motor encoder, and a symmetrically arranged depth camera 13 to acquire sandblasting pressure, spray gun distance, swing angular velocity, and surface image data in real time. The anomaly diagnosis module determines the anomaly type based on pressure data, generates diagnostic results, triggers system alerts, and pushes maintenance prompts. The motion control module, based on the feedback from the ranging sensor 12, drives the telescopic arm 2 to adjust the position of the spray gun, and dynamically adjusts the angular velocity of the spray gun swinging in combination with encoder data to achieve closed-loop control of motion parameters; The visual inspection module filters, crops, and extracts features from images acquired by the depth camera 13, evaluates cleanliness using a pre-trained classification model, calculates roughness Sa value using point cloud residuals, and compares it with a standard sample. The control and management module coordinates the operation of each module, records quality data and abnormal events, and generates traceable logs; when any sensor parameter is abnormal, the operation is immediately interrupted and the data is archived.

[0042] This technical solution achieves real-time calibration of process parameters and proactive prevention and control of quality risks through multi-sensor data fusion and modular intelligent algorithms, thereby more efficiently solving the problems of frequent manual intervention, delayed parameter adjustments, and large quality fluctuations. Specifically, process parameters are collected in real time through multi-sensor perception modules such as pressure, distance, angular velocity, and vision, constructing a digital twin mapping of the sandblasting process; the anomaly diagnosis module uses pressure feature pattern recognition to achieve fault prediction; the motion control module optimizes trajectory accuracy through position-velocity dual closed loops; the vision detection module combines image classification and 3D point cloud quantification of surface quality; and finally, the control management module achieves data integration and anomaly linkage control, solving the problems of dynamic stability, motion accuracy, and surface consistency in the sandblasting process.

[0043] According to one embodiment of the present invention, the abnormal diagnosis module has a built-in pressure threshold range [P] Low ,P High The pressure change rate threshold is used to determine whether the pressure is slowly decreasing or fluctuating drastically by analyzing the slope of the real-time pressure curve, and to predict the abrasive supply status by combining historical data.

[0044] This technical solution achieves real-time perception and fault prediction of abrasive supply status through pressure dynamic characteristic analysis and historical data fusion prediction, thereby more efficiently solving the problems of response lag and high misjudgment rate caused by the reliance on manual inspection in traditional methods.

[0045] According to one embodiment of the present invention, the visual inspection module employs a depth camera 13 symmetrically deployed to cover the sandblasting area on both sides, generates a full surface quality map through image stitching technology, marks the coordinates of the substandard areas and feeds them back to the motion control module.

[0046] This technical solution achieves parallel data acquisition of the sandblasting areas on both sides through a symmetrical layout of 13 depth cameras, and eliminates blind spots caused by single-view occlusion using stereo vision algorithms. The image stitching technology, through feature point matching and spatial transformation matrix calculation, fuses multi-view images into a full-surface quality map under a unified coordinate system, and constructs a correlation mapping between the three-dimensional geometric model of the sandblasted surface and pixel-level quality annotations. After receiving the coordinates of the substandard areas, the motion control module converts the spatial position into joint motor commands through inverse kinematics algorithms, forming a closed-loop control, which solves the problem of difficult-to-accurate positioning of local quality defects in large-sized workpieces.

[0047] According to one embodiment of the present invention, the motion control module employs a PID control algorithm, based on the real-time distance fed back by the ranging sensor 12 and the preset range [D]. Min D Max To compensate for any deviation, dynamically adjust the moving speed of the telescopic arm 2 to ensure a stable distance between the spray gun and the workpiece surface.

[0048] This technical solution achieves dynamic adaptive adjustment of the spray gun distance through the collaboration of PID algorithm and distance measurement feedback, thereby more efficiently solving the problems of uneven sandblasting and surface quality fluctuation caused by manual adjustment lag or fixed parameter control.

[0049] Example 3 Based on Example 2, such as Figure 3 and Figure 4 As shown, this embodiment provides a sandblasting robotic arm based on multi-sensor perception, including the following structure: The drive arm structure includes a fixed arm 1 and a telescopic arm 2 arranged in sequence. The telescopic arm 2 is driven by a ball screw and a servo motor, and slides along the cavity of the fixed arm 1 to achieve axial extension and retraction. The three-degree-of-freedom joint includes joint I3, joint II4 and joint III5 arranged sequentially. Joint I3 is installed at the end of the telescopic arm 2 and supports ±90° rotation. Joint II4 is installed at the end of joint I3 and supports 360° continuous rotation. Joint III5 is installed at the end of joint II4 and supports ±90° swing. Each joint has a built-in joint motor encoder. Depth cameras 13 are symmetrically arranged on both sides of joint I3. A distance sensor 12 is integrated at the front of joint III5. The spraying structure includes a left sandblasting pipe 8 and a right sandblasting pipe 11, as well as a left spray gun 7 and a right spray gun 10. The outlet of the left sandblasting pipe 8 is clamped to the left side of joint III 5 by a left clamp 6, and the outlet of the right sandblasting pipe 11 is clamped to the right side of joint III 5 by a right clamp 9. The left spray gun 7 and the right spray gun 10 are respectively connected to the left sandblasting pipe 8 and the right sandblasting pipe 11 through quick connectors.

[0050] This technical solution achieves high-precision axial extension and retraction through ball screw transmission of the drive arm, realizes rotation-slewing-oscillation control through a three-degree-of-freedom joint, and provides angle feedback to form a position closed loop through the joint motor encoder. Depth cameras 13 are symmetrically arranged on both sides of joint I3 to acquire the three-dimensional shape of the workpiece surface using stereo vision. The distance sensor 12 is integrated into the front of joint III5 to realize real-time compensation of the distance between the spray gun and the workpiece. The symmetrical layout of the dual sandblasting pipes is merged into a single spray gun through a quick connector. The robotic arm and the data perceived by multiple sensors are fused to calculate the optimal sandblasting trajectory to adapt to complex curved surfaces, solving the over-spraying / under-spraying problem caused by the lack of perception in traditional robotic arms.

[0051] According to one embodiment of the present invention, the inlets of the left sandblasting pipe 8 and the right sandblasting pipe 11 are clamped to the side of the fixed arm 1, and the inlets of the left sandblasting pipe 8 and the right sandblasting pipe 11 are both equipped with pressure sensors 14, and the left sandblasting pipe 8 and the right sandblasting pipe 11 are both equipped with explosion-proof steel wire braided layers.

[0052] This technical solution uses pressure sensor 14 and explosion-proof pipe body to realize real-time monitoring of sandblasting pressure and overpressure protection, thereby more efficiently solving the problem of pipe burst risk and processing interruption caused by pressure runaway of traditional sandblasting pipes.

[0053] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should also be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. A multi-sensor perception based sandblasting method, characterized by, The method comprises the following steps: S1, a multi-sensor perception module based on a sand blasting mechanical arm is constructed, and the multi-sensor perception module comprises a pressure sensor (14), a distance measuring sensor (12), a joint motor encoder and a depth camera (13) located on the sand blasting mechanical arm; S2, real-time acquisition of pressure data in the sand blasting pipe is performed through the pressure sensor (14), and abnormality is judged in combination with a preset normal pressure interval: If the pressure value is abnormal, pressure abnormality type diagnosis is performed, a corresponding diagnosis result is generated, and a system early warning is triggered to prompt maintenance; If the pressure value is normal, the distance measuring sensor (12) is started to detect the distance between the sand blasting gun and the workpiece surface in real time, and the telescopic arm (2) is driven to automatically adjust to a preset distance range; S3, the swing angular velocity of the sand blasting gun is detected through the joint motor encoder, and the sand blasting surface image collected by the depth camera (13) in real time is combined to dynamically adjust the motion parameters of the sand blasting gun; S4, the sand blasting area is imaged by using the symmetrically arranged depth camera (13), after filtering, cropping and feature extraction, the surface cleanliness is judged by using a pre-trained image classification model, and the surface roughness Sa value is calculated by using a point cloud residual error; If the cleanliness / roughness meets the standard, quality data is recorded and archived; If the cleanliness / roughness does not meet the standard, a quality alarm is triggered, the workpiece position is marked, and the sand blasting parameters are adjusted to perform a re-sand blasting operation.

2. The multi-sensor perception based sandblasting method of claim 1, wherein, The abnormal types of the pressure sensor (14) of S2 include pressure being too low, pressure slowly decreasing, pressure violently fluctuating or pressure being too high; The diagnosis result is divided into poor feeding, nozzle wear, abrasive agglomeration or nozzle blockage, wherein: pressure being too low corresponds to poor feeding or nozzle blockage; pressure slowly decreasing corresponds to abrasive agglomeration; Pressure violently fluctuating corresponds to nozzle wear.

3. The multi-sensor perception based sandblasting method of claim 1, wherein, The pressure sensor (14) and the distance measuring sensor (12) of S2, the joint motor encoder of S3, and the depth camera (13) of S4, when any sensor parameter is abnormal, immediately interrupt the sand blasting operation, synchronously record abnormal data and generate a traceable maintenance log.

4. The multi-sensor perception based sandblasting method of claim 1, wherein, The cleanliness of S4 is evaluated by training an RGB image classification model, the roughness is evaluated by calculating the Sa value through a point cloud residual error and comparing with a standard sample, both of which are used as quality determination basis.

5. A multi-sensor perception based sandblasting control system employing the multi-sensor perception based sandblasting method according to any one of claims 1 to 4, characterized in that, The method comprises the following modules: A multi-sensor perception module integrating a pressure sensor (14), a distance measuring sensor (12), a joint motor encoder and symmetrically arranged depth cameras (13) is used to acquire sand blasting pressure, sand blasting gun distance, swing angular velocity and surface image data in real time; An abnormality diagnosis module is used to judge abnormal types according to pressure data, generate diagnosis results, trigger system early warning and push maintenance prompts; A motion control module is used to drive the telescopic arm (2) to adjust the position of the sand blasting gun based on the feedback of the distance measuring sensor (12), dynamically adjust the swing angular velocity of the sand blasting gun in combination with the encoder data, and realize closed-loop control of motion parameters; A visual detection module is used to perform filtering, cropping and feature extraction on the images collected by the depth camera (13), evaluate the cleanliness through a pre-trained classification model, calculate the roughness Sa value through a point cloud residual error and compare with a standard sample. Control management module coordinates the operation of each module, records quality data and abnormal events, and generates traceable logs; when any sensor parameter is abnormal, the operation is immediately interrupted and the data is archived.

6. The multi-sensor perception based sandblasting control system of claim 5, wherein, The abnormality diagnosis module has a pressure threshold interval [P Low ,P High ] and a pressure change rate threshold, and judges slow pressure drop or severe fluctuation through real-time pressure curve slope, and predicts abrasive supply state in combination with historical data.

7. The multi-sensor perception based sandblasting control system of claim 5, wherein, The visual detection module adopts symmetrical deployment of depth cameras (13) to cover the double-sided sandblasting area, generates a full-surface quality map through image stitching technology, marks the coordinates of non-compliant areas, and feeds back to the motion control module.

8. The multi-sensor perception based sandblasting control system of claim 5, wherein, The motion control module adopts a PID control algorithm, dynamically adjusts the moving speed of the telescopic arm (2) according to the deviation between the real-time distance fed back by the distance measuring sensor (12) and the preset range [D Min ,D Max ], and ensures the stable distance between the spray gun and the workpiece surface.

9. A multi-sensor perception based sandblasting robot employing a multi-sensor perception based sandblasting control system as claimed in any one of claims 5-8, characterized in that, It comprises the following structure: The driving arm structure comprises a fixed arm (1) and a telescopic arm (2) arranged in sequence, the telescopic arm (2) is driven by a ball screw and a servo motor, and slides along the cavity of the fixed arm (1) to realize axial telescoping; The three-degree-of-freedom joint comprises joint I (3), joint II (4) and joint III (5) arranged in sequence, joint I (3) is installed at the end of the telescopic arm (2) and supports ±90° rotation, joint II (4) is installed at the end of joint I (3) and supports 360° continuous rotation, joint III (5) is installed at the end of joint II (4) and supports ±90° swing, each joint is provided with a joint motor encoder; the depth cameras (13) are symmetrically arranged on both sides of joint I (3), and the ranging sensor (12) is integrated in the front of joint III (5); The spraying structure comprises left sandblasting pipe (8) and right sandblasting pipe (11), and left spray gun (7) and right spray gun (10); the outlet of the left sandblasting pipe (8) is clamped to the left side of joint III (5) through the left clamp (6), the outlet of the right sandblasting pipe (11) is clamped to the right side of joint III (5) through the right clamp (9), and the left spray gun (7) and the right spray gun (10) are respectively connected to the left sandblasting pipe (8) and the right sandblasting pipe (11) through quick connectors.

10. The multi-sensor perception based sandblasting robot of claim 9, wherein, The inlet of the left sandblasting pipe (8) and the right sandblasting pipe (11) is clamped to the side of the fixed arm (1), and the inlet of the left sandblasting pipe (8) and the right sandblasting pipe (11) is provided with a pressure sensor (14), and the left sandblasting pipe (8) and the right sandblasting pipe (11) are provided with an explosion-proof steel wire braid.

Images