Industrial robot automatic control method, device and computer readable storage medium

Abstract

The present invention proposes an industrial robot automatic control method, device and computer-readable storage medium, wherein the method includes: using image recognition technology to extract the surface features of the workpiece to be polished, and controlling the initial force and torque of the industrial robot end effector based on the surface features of the workpiece to be polished; monitoring the surface temperature fluctuation of the workpiece during the polishing process through the temperature sensor of the industrial robot end effector, and then adjusting the force and torque of the industrial robot end effector; real-time acquisition of power and state data during the polishing process, using deep learning technology to predict whether there is an abnormal condition, and once an abnormality is found, immediately terminate the polishing operation. Through this method and the corresponding device, the industrial robot can realize intelligent automatic adjustment during the polishing and polishing process, and maintain high performance and high reliability under complex working conditions.

Description

Technical Field

[0001] The present invention provides an industrial robot automatic control method, a device and a computer-readable storage medium, and relates to the technical field of industrial robot control. Background Art

[0002] During manual grinding, the operator can reasonably control the grinding feed rate based on his own experience and actual conditions by relying on his perception of grinding force and visual observation. In addition, when encountering strong impact, the human body has its own buffering ability, which can effectively alleviate the impact and play a shock-absorbing effect, thereby protecting the body of the worker to a certain extent. Robot grinding, on the other hand, is different from manual grinding. The force exerted by mechanical grinding far exceeds that of manual grinding, which results in greater impact and vibration. Because the robot structure is rigid and cannot play a shock-absorbing role like the human body, during the processing process, the severe vibration will be transmitted to the entire robot mechanism through the grinding wheel and manipulator. In the long run, it will inevitably affect the service life of the robot. At the same time, as the grinding process continues to advance, the accumulated thermal effect will cause the surface of the workpiece material to change, so the same control strategy cannot be used throughout the entire grinding process. In addition, the current robot grinding technology still relies on human supervision and cannot achieve the goal of full automation. Summary of the invention

[0003] The present invention proposes an industrial robot automatic control method, device and computer-readable storage medium to solve the above-mentioned problems:

[0004] The industrial robot automatic control method proposed by the present invention comprises:

[0005] Use image recognition technology to extract the surface features of the workpiece to be polished, and control the initial force and torque of the end effector of the industrial robot based on the surface features of the workpiece to be polished;

[0006] Monitoring the surface temperature fluctuation of the workpiece during the grinding process by a temperature sensor of the industrial robot end effector, thereby adjusting the force and torque of the industrial robot end effector;

[0007] The power and status data of the grinding operation are collected in real time, and deep learning technology is used to predict whether there are abnormal conditions. Once an abnormality is found, the grinding operation is terminated immediately.

[0008] Furthermore, the surface features of the workpiece to be polished are extracted using image recognition technology, and the initial force and torque of the end effector of the industrial robot are controlled based on the surface features of the workpiece to be polished, including:

[0009] Under uniform lighting conditions, a high-definition camera mounted on the end effector of an industrial robot is used to obtain multi-view images of the surface of the workpiece to be polished;

[0010] Using filtering methods to remove potential noise points in the image, the filtering methods include median filtering and Gaussian filtering algorithms, and optimizing the contrast performance of the image by adaptively adjusting histogram equalization;

[0011] After the image contrast is enhanced, the edge details in the image are captured by the edge detection operator, and then the image is reconstructed in three dimensions. After the three-dimensional reconstruction, the surface features of the workpiece are extracted using deep learning technology. The surface features of the workpiece include: the surface roughness of the workpiece to be polished, the Vickers hardness of the surface material of the workpiece to be polished, and the corrosion depth of the surface of the workpiece to be polished;

[0012] After acquiring the surface features of the workpiece, the initial force and torque conditions are set for the end effector of the industrial robot according to the pre-established end effector force model and torque model.

[0013] Further, specifically, the pre-established end effector force model is:

[0014]

[0015] in, represents the contact length between the end effector of the industrial robot and the workpiece to be polished, t represents the processing time, m represents the mass of the grinding head of the end effector of the industrial robot, represents the initial force of the end effector of the industrial robot, represents the surface roughness of the workpiece to be polished, K represents the stiffness coefficient of the end effector of the industrial robot, represents the feed speed of the grinding head of the industrial robot end effector, β represents the deflection of the grinding head of the industrial robot end effector, represents the instantaneous value of the cutting force, η represents the stable machining coefficient, H represents the Vickers hardness of the surface material of the workpiece to be processed, D represents the corrosion depth of the surface of the workpiece to be polished, μ represents the friction coefficient between the surface of the workpiece to be polished and the grinding head of the end effector of the industrial robot, A represents the contact area between the grinding head of the end effector of the industrial robot and the surface of the workpiece to be polished, Indicates the relative movement speed of the industrial robot's grinding head contact point with the workpiece during linear motion.

[0016] Further, specifically, the pre-established end effector torque model is:

[0017]

[0018] in, represents the initial torque, Indicates the reference value of the initial torque, represents the torque constant, M represents the load mass, r represents the length of the lever arm, and J represents the moment of inertia. represents the damping coefficient, represents the nonlinear torque response adjustment parameter, represents the environmental impact correction factor, represents the reference speed of the end effector, V represents the linear speed of the end effector, The coefficient representing the rate of change of power, represents the power change rate, ϕ represents the power response nonlinear adjustment parameter, It represents the center angle of the circle corresponding to the grinding path of the end effector, and t represents the processing time.

[0019] Furthermore, the temperature fluctuation of the workpiece surface during the grinding process is monitored by a temperature sensor of the industrial robot end effector, and the force and torque of the industrial robot end effector are adjusted, including:

[0020] Deploy temperature sensors on the robot's end effector to monitor temperature fluctuations in real time during operation;

[0021] After obtaining the temperature change data, the specially constructed force regulation model is used to optimize the force parameters, and then the torque regulation model is used to make corresponding adjustments to the torque parameters.

[0022] Further, specifically, the force regulation model is:

[0023]

[0024] in, represents the adjusted force at time t, represents the initial force of the end effector of the industrial robot at time t, μ represents the friction coefficient between the surface of the workpiece to be polished and the grinding head of the end effector of the industrial robot, Indicates instantaneous temperature fluctuations, represents the displacement of the end effector of the industrial robot at time t, represents the preset maximum displacement, m represents the mass of the grinding head of the industrial robot end effector, c represents the specific heat capacity of the workpiece to be ground, and D represents the temperature fluctuation coefficient. represents the temperature accumulation coefficient, represents the average temperature over time τ.

[0025] Further, specifically, the torque regulation model is:

[0026]

[0027] in, represents the torque at the current time t, r represents the arm length of the end effector of the industrial robot, represents the adjusted force at time t, represents the temperature influence coefficient, represents the force at time τ, and τ represents the integral variable, which is used to calculate the cumulative effect of the force from the starting time 0 to time t.

[0028] Furthermore, the power and status data of the grinding process are collected in real time, and deep learning technology is used to predict whether there are abnormal conditions. Once an abnormality is found, the grinding operation is terminated immediately, including:

[0029] A variety of sensors are deployed on the grinding and polishing robots, including force sensors, torque sensors, temperature sensors, and vibration sensors, to comprehensively collect power and status information during the processing. The collection frequency is pre-set according to the characteristics of the grinding process and the frequency of data changes;

[0030] Use filtering algorithms to remove noise from the data collected during the processing;

[0031] For power data, features are extracted from power data such as grinding force and torque, including the change trend of force, peak and valley values ​​of torque, and fluctuation frequency of force and torque. For state data, for temperature and vibration state data, the temperature change rate, temperature non-uniformity index, and spectral characteristics of vibration and changes in vibration amplitude are extracted;

[0032] Construct an initial deep learning model, i.e., a long short-term memory network model, collect polishing operation data under normal and abnormal conditions, divide the data into training set, validation set, and test set according to a preset ratio, and annotate the training set data into normal state data and abnormal state data;

[0033] The constructed initial long short-term memory network model is trained using the training set data, and the model parameters, including weights and biases, are adjusted so that the model can learn the characteristic differences between normal and abnormal state data, and the prediction accuracy of the model is improved by minimizing the cross entropy loss function. During the training process, the model is verified using the validation set data to prevent the model from overfitting;

[0034] Use the test set data to evaluate the trained model, calculate the model's accuracy, recall, and F1 value, evaluate the model's performance, and optimize the model based on the evaluation results, including adjusting the model structure, hyperparameters, and increasing the amount of training data;

[0035] Deploy the trained and evaluated deep learning model to the real-time monitoring system of the polishing operation;

[0036] The polishing operation data collected and preprocessed in real time is input into the deployed deep learning model. The deep learning model predicts whether there is any abnormal condition in the current polishing operation based on the learned feature pattern and outputs the prediction result.

[0037] Once the model predicts an abnormality, a stop command is immediately sent to the grinding equipment, and the grinding operation is terminated through the control interface to prevent the abnormal situation from further deteriorating and prevent greater damage to the equipment and workpiece.

[0038] The industrial robot automatic control device proposed by the present invention comprises:

[0039] The preliminary control module uses image recognition technology to extract the surface features of the workpiece to be polished, and controls the initial force and torque of the end effector of the industrial robot based on the surface features of the workpiece to be polished;

[0040] A control and adjustment module, used to monitor the temperature fluctuation of the workpiece surface during the grinding process through the temperature sensor of the industrial robot end effector, and then adjust the force and torque of the industrial robot end effector;

[0041] The abnormality prediction module is used to collect power and status data during the grinding operation in real time, and use deep learning technology to predict whether there are abnormal conditions. Once an abnormality is found, the grinding operation is terminated immediately.

[0042] The present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the industrial robot automatic control method.

[0043] The beneficial effects of the present invention are as follows: through visual and temperature feedback, customized processing of different workpieces can be achieved, complex surface shapes can be adapted and grinding consistency can be maintained; real-time monitoring can improve processing efficiency and avoid damage to workpieces or tools due to overheating or abnormal vibration; high-quality polished surfaces can be ensured, material waste and subsequent processing costs can be reduced, and tool life can be extended; combined with artificial intelligence technology, the robot can have autonomous adaptation and learning capabilities, improve the level of full automation, and reduce dependence on manual adjustment; through the implementation of this integration of multiple technologies, industrial robots can achieve intelligent automatic adjustment during the grinding and polishing process, and maintain high performance and high reliability under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the automatic control method of the industrial robot according to the present invention. DETAILED DESCRIPTION

[0045] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. The embodiments described are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0048] One embodiment of the present invention provides an automatic control method for an industrial robot, the method comprising:

[0049] Use image recognition technology to extract the surface features of the workpiece to be polished, and control the initial force and torque of the end effector of the industrial robot based on the surface features of the workpiece to be polished;

[0050] Monitoring the surface temperature fluctuation of the workpiece during the grinding process by a temperature sensor of the industrial robot end effector, thereby adjusting the force and torque of the industrial robot end effector;

[0051] The power and status data of the grinding operation are collected in real time, and deep learning technology is used to predict whether there are abnormal conditions. Once an abnormality is found, the grinding operation is terminated immediately.

[0052] The working principle and effect of the above technical solution are: combining computer vision, sensor monitoring and deep learning to optimize the performance of industrial robots in the grinding and polishing process; extracting surface information with computer vision, image acquisition and analysis: using a camera or other imaging device to obtain workpiece surface information, and extracting significant features through computer vision technology (such as edge detection, texture analysis, etc.); parameter setting: according to the analysis results (such as surface roughness, shape irregularity, etc.), calculate the required initial grinding force and torque to adapt to the workpiece surface with different characteristics, and effectively set the grinding and polishing strategy; real-time temperature monitoring and adjustment, temperature sensor monitoring, installing a temperature sensor on the robot end effector to monitor the temperature change of the workpiece surface in real time; dynamic adjustment, adjusting the grinding force and torque according to temperature changes to ensure that the operation is carried out within a safe temperature range to prevent damage to the workpiece or poor performance due to temperature increase; power and state information acquisition and analysis, data acquisition: continuously collecting power information (such as grinding force, torque) and state information during the grinding process; deep learning anomaly detection: using a pre-trained deep learning model to analyze power and state information, predicting whether there are potential anomalies (such as abnormal wear, tool failure, etc.); abnormality handling, automatic stop, When the model detects an anomaly, it immediately stops the grinding process to protect the equipment and workpieces; sends an alarm notification to the operator so that further inspection or maintenance measures can be taken. Improve processing accuracy and quality. By accurately setting the initial grinding conditions and dynamically adjusting the operating parameters, the optimal processing effect corresponding to the surface features of different workpieces is achieved, and the quality and consistency of the finished product are improved; enhance process safety. Real-time temperature monitoring and anomaly detection technology reduce the risk of possible workpiece damage and equipment failure, and improve the safety of the grinding process; improve operational efficiency, reduce downtime and time loss caused by misoperation, optimize grinding cycles and speeds, and improve work efficiency; reduce operation and maintenance costs. By identifying and preventing potential anomalies, equipment wear and failure frequency are reduced, and maintenance and replacement costs are reduced; the adaptive ability is improved. Combined with vision and machine learning technology, the robot system can adapt to different workpieces and conditions more flexibly, and improve its adaptive ability in complex environments; through the coordinated use of multiple technologies, it provides an intelligent, efficient and reliable solution for automated grinding and polishing processes.

[0053] In one embodiment of the present invention, image recognition technology is used to extract surface features of a workpiece to be polished, and the initial force and torque of an end effector of an industrial robot are controlled based on the surface features of the workpiece to be polished, including:

[0054] Under uniform lighting conditions, a high-definition camera mounted on the end effector of an industrial robot is used to obtain multi-view images of the surface of the workpiece to be polished;

[0055] Using filtering methods to remove potential noise points in the image, the filtering methods include median filtering and Gaussian filtering algorithms, and optimizing the contrast performance of the image by adaptively adjusting histogram equalization;

[0056] After the image contrast is enhanced, the edge details in the image are captured by the edge detection operator, and then the image is reconstructed in three dimensions. After the three-dimensional reconstruction, the surface features of the workpiece are extracted using deep learning technology. The surface features of the workpiece include: the surface roughness of the workpiece to be polished, the Vickers hardness of the surface material of the workpiece to be polished, and the corrosion depth of the surface of the workpiece to be polished;

[0057] After acquiring the surface features of the workpiece, the initial force and torque conditions are set for the end effector of the industrial robot according to the pre-established end effector force model and torque model.

[0058] The working principle and effect of the above technical solution are as follows: using computer vision and deep learning technology, by analyzing the workpiece surface images taken by the industrial robot, to formulate appropriate grinding and polishing parameters, high-resolution image acquisition, image acquisition and multi-angle view: using the high-resolution camera installed on the industrial robot, to obtain multi-angle images of the workpiece surface under uniform lighting conditions to ensure that the surface features of the workpiece are fully captured; image preprocessing, denoising filtering, applying median filtering and Gaussian filtering to remove noise in the image and enhance image quality, contrast enhancement, using adaptive histogram equalization to improve the image contrast and highlight surface details; feature extraction and 3D reconstruction, edge detection: using operators (such as Canny or Sobel operators) to detect image edges and obtain surface contours; three-dimensional reconstruction, performing three-dimensional reconstruction based on multi-view images to reproduce the surface morphology of the workpiece; deep learning analysis, using deep learning models to analyze the processed image data, extracting key information on the workpiece surface, including surface roughness, Vickers hardness of the material, and corrosion depth; setting initial grinding parameters, using the extracted surface information, calculating and setting the initial force and torque of the industrial robot end effector through the initial force model and initial torque model to meet the processing requirements of specific workpieces. Improve processing adaptability and accuracy, be able to intelligently adjust processing parameters for workpiece surfaces of different types and states, and effectively improve processing accuracy and quality; optimize grinding strategies, by pre-understanding the material properties and state of the workpiece surface (such as hardness and corrosion), optimize grinding operation strategies, and reduce unnecessary wear and processing defects; enhance the level of automation, use deep learning technology to realize automatic recognition and analysis of complex information on the workpiece surface, improve the degree of production automation, and reduce manual intervention; reduce material and equipment wear, by setting appropriate initial force and torque, reduce adverse impact and wear during processing, and extend the service life of tools and equipment; improve production efficiency, faster processing preparation and parameter setting improve overall production efficiency, and reduce debugging and adjustment time; through the application of highly automated and intelligent technology, the grinding and polishing processes are greatly optimized, and the efficiency of the production line and product quality are improved.

[0059] In one embodiment of the present invention, specifically, the pre-established end effector force model is:

[0060]

[0061] in, represents the contact length between the end effector of the industrial robot and the workpiece to be polished, t represents the processing time, m represents the mass of the grinding head of the end effector of the industrial robot, represents the initial force of the end effector of the industrial robot, represents the surface roughness of the workpiece to be polished, K represents the stiffness coefficient of the end effector of the industrial robot, represents the feed speed of the grinding head of the industrial robot end effector, β represents the deflection of the grinding head of the industrial robot end effector, represents the instantaneous value of the cutting force, η represents the stable machining coefficient, H represents the Vickers hardness of the surface material of the workpiece to be processed, D represents the corrosion depth of the surface of the workpiece to be polished, μ represents the friction coefficient between the surface of the workpiece to be polished and the grinding head of the end effector of the industrial robot, A represents the contact area between the grinding head of the end effector of the industrial robot and the surface of the workpiece to be polished, Indicates the relative movement speed of the industrial robot's grinding head contact point with the workpiece during linear motion.

[0062] In one embodiment of the present invention, specifically, the pre-established end effector torque model is:

[0063]

[0064] in, represents the initial torque, Indicates the reference value of the initial torque, represents the torque constant, M represents the load mass, r represents the length of the lever arm, and J represents the moment of inertia. represents the damping coefficient, represents the nonlinear torque response adjustment parameter, represents the environmental impact correction factor, represents the reference speed of the end effector, V represents the linear speed of the end effector, The coefficient representing the rate of change of power, represents the power change rate, ϕ represents the power response nonlinear adjustment parameter, It represents the center angle of the circle corresponding to the grinding path of the end effector, and t represents the processing time.

[0065] The working principle and effect of the above technical solution are as follows: a comprehensive initial force and initial torque model for the end effector of an industrial robot is provided, which aims to accurately control and optimize the performance of the industrial robot in processing tasks by considering multiple factors; the surface roughness in the initial force model reflects the initial surface state of the workpiece to be polished, and affects the distribution and size of the force; through the integration of K and Vw−Ve, the model considers the influence of the robot stiffness and the grinding head feed speed on the force, ensuring that the appropriate force is applied at the initial stage of processing; the cutting force Multiplying by the stable processing coefficient η indicates the instantaneous force adjustment during the processing to ensure a relatively stable processing quality; through the friction coefficient μ, contact area A and relative moving speed The influence of friction between the grinding head and the workpiece on the force output is considered; in the initial torque model, , M, r and θ, the model adapts to the torque requirements under different load masses and lever arms, and adjusts the response through nonlinear parameters; the exponential term Simulate the environment to correct the torque to ensure accurate torque prediction at different line speeds; power change rate It reflects the influence of power fluctuation on torque, and prevents torque anomalies caused by sudden power changes by considering the power change rate and its nonlinear adjustment. Enhance machining accuracy, by precisely controlling the initial force and torque, ensure that the robot can apply variable and appropriate force under the variable surface state of the workpiece, and improve machining accuracy; optimize robot performance, comprehensively consider speed, load and environmental factors, adjust the output of torque and force, so that the robot can adapt to different operating conditions and improve performance stability; improve machining efficiency, by reducing unnecessary force and torque fluctuations, achieve higher machining speed and stability, and thus improve overall operating efficiency; reduce failures and wear, precise force and torque control reduces component wear and unexpected failures caused by overload or improper operation, and extends equipment life; provide flexible response capabilities, under various environmental influences, through nonlinear response and dynamic adjustment, the model can enable the robot to maintain high-performance machining output in different workpieces and tasks; this technical solution provides a flexible and precise initial condition setting method by comprehensively considering the interaction factors between the end effector and the workpiece, laying the foundation for efficient and reliable industrial robot applications.

[0066] In one embodiment of the present invention, the temperature sensor of the end effector of the industrial robot monitors the surface temperature fluctuation of the workpiece during the grinding process, and then adjusts the force and torque of the end effector of the industrial robot, including:

[0067] Deploy temperature sensors on the robot's end effector to monitor temperature fluctuations in real time during operation;

[0068] After obtaining the temperature change data, the specially constructed force regulation model is used to optimize the force parameters, and then the torque regulation model is used to make corresponding adjustments to the torque parameters.

[0069] In one embodiment of the present invention, specifically, the force adjustment model is:

[0070]

[0071] in, represents the adjusted force at time t, represents the initial force of the end effector of the industrial robot at time t, μ represents the friction coefficient between the surface of the workpiece to be polished and the grinding head of the end effector of the industrial robot, Indicates instantaneous temperature fluctuations, represents the displacement of the end effector of the industrial robot at time t, represents the preset maximum displacement, m represents the mass of the grinding head of the industrial robot end effector, c represents the specific heat capacity of the workpiece to be ground, and D represents the temperature fluctuation coefficient. represents the temperature accumulation coefficient, represents the average temperature over time τ.

[0072] In one embodiment of the present invention, specifically, the torque adjustment model is:

[0073]

[0074] in, represents the torque at the current time t, r represents the arm length of the end effector of the industrial robot, represents the adjusted force at time t, represents the temperature influence coefficient, represents the force at time τ, and τ represents the integral variable, which is used to calculate the cumulative effect of the force from the starting time 0 to time t.

[0075] The working principle and effect of the above technical solution are: by real-time detection of the temperature change of the robot end effector, the force and torque of the actuator are dynamically adjusted to optimize the operation effect and reduce the negative impact of temperature rise on the robot performance. The temperature sensor installed at the end effector of the robot is used to monitor the change of the actuator temperature ΔT(t) in real time; the basic force in the force adjustment model

[0076] The initial base force plus an adjustment term based on temperature changes ensures that the force does not fail due to temperature changes; the output force is adjusted considering the friction coefficient μ and the temperature effect over time (1−β⋅ΔT(t)); the integral term The effect of the cumulative average temperature on the force reflects the weakening of the force output due to the gradual increase in temperature; the torque in the torque adjustment model Calculated based on the adjusted force F(t) and the lever arm length r; Temperature compensation adjusts the torque by (1+α⋅ΔT(t)) to offset the change in mechanical properties caused by temperature increase; Compensation for the cumulative effect of torque, integral term Used to offset the change of torque caused by the cumulative thermal effect at the previous time point. High-fidelity operation, through real-time adjustment, reduces the negative impact of temperature changes on the force and torque accuracy of the robot's end effector, and improves the fidelity of operation; performance stability, through a highly adaptable adjustment model, it can maintain stable operating performance during temperature rise, and will not cause the actuator to move inaccurately due to temperature changes; reduce heat-induced failures, by accurately compensating for mechanical changes caused by thermal expansion and contraction, reduce mechanical fatigue and failure risks caused by temperature fluctuations, and extend equipment life; improve efficiency and safety, while ensuring efficiency, maintain high mechanical safety, and prevent failures and safety hazards that may be caused by overheating operations; enhance adaptability, whether in high temperature environments or long-term operation, the system can automatically adjust force and torque to keep the robot in the best operating state; through real-time monitoring and dynamic compensation, not only the efficient operation of the robot is maintained, but also its autonomous adjustment capabilities in different environments and operating conditions are enhanced.

[0077] In one embodiment of the present invention, power and status data during the grinding operation are collected in real time, and deep learning technology is used to predict whether there is an abnormal condition. Once an abnormality is found, the grinding operation is immediately terminated, including:

[0078] A variety of sensors are deployed on the grinding and polishing robots, including force sensors, torque sensors, temperature sensors, and vibration sensors, to comprehensively collect power and status information during the processing. The collection frequency is pre-set according to the characteristics of the grinding process and the frequency of data changes;

[0079] Use filtering algorithms to remove noise from the data collected during the processing;

[0080] For power data, features are extracted from power data such as grinding force and torque, including the change trend of force, peak and valley values ​​of torque, and fluctuation frequency of force and torque. For state data, for temperature and vibration state data, the temperature change rate, temperature non-uniformity index, and spectral characteristics of vibration and changes in vibration amplitude are extracted;

[0081] Construct an initial deep learning model, i.e., a long short-term memory network model, collect polishing operation data under normal and abnormal conditions, divide the data into training set, validation set, and test set according to a preset ratio, and annotate the training set data into normal state data and abnormal state data;

[0082] The constructed initial long short-term memory network model is trained using the training set data, and the model parameters, including weights and biases, are adjusted so that the model can learn the characteristic differences between normal and abnormal state data, and the prediction accuracy of the model is improved by minimizing the cross entropy loss function. During the training process, the model is verified using the validation set data to prevent the model from overfitting;

[0083] Use the test set data to evaluate the trained model, calculate the model's accuracy, recall, and F1 value, evaluate the model's performance, and optimize the model based on the evaluation results, including adjusting the model structure, hyperparameters, and increasing the amount of training data;

[0084] Deploy the trained and evaluated deep learning model to the real-time monitoring system of the polishing operation;

[0085] The polishing operation data collected and preprocessed in real time is input into the deployed deep learning model. The deep learning model predicts whether there is any abnormal condition in the current polishing operation based on the learned feature pattern and outputs the prediction result.

[0086] Once the model predicts an abnormality, a stop command is immediately sent to the grinding equipment, and the grinding operation is terminated through the control interface to prevent the abnormal situation from further deteriorating and prevent greater damage to the equipment and workpiece.

[0087] The working principle and effect of the above technical solution are as follows: multi-sensor deployment, force sensors, torque sensors, temperature sensors and vibration sensors are installed on the robot to monitor and collect power and environmental status information in the grinding and polishing process in real time; signal transmission and processing, all physical quantity information (grinding force, torque, temperature and vibration force) collected by the sensors are transmitted to the central control system; the central control system performs signal processing, including denoising and standardization, to ensure the clarity and consistency of the data; feature extraction, extracting key information from the processed real-time data, the features include the change pattern of the grinding force, the temperature non-uniformity index (used to detect local overheating) and the vibration spectrum characteristics (used to identify mechanical instability); deep neural network detection, the feature data is input into the pre-trained deep neural network model, and the model is used to detect and identify potential abnormal conditions in the grinding and polishing process; the deep neural network can accurately identify which feature combinations represent abnormal conditions by using a large amount of historical data for training; abnormality handling and alarm, if the network model determines that there is a potential abnormality in the current process (such as excessive wear, tool failure, friction abnormality), the system will immediately stop the grinding process, and at the same time, the system will send an alarm message to the relevant personnel for timely response and intervention. Improve processing quality. Real-time monitoring and anomaly detection ensure that the processing process is carried out under optimal conditions, reduce processing defects, and improve the quality of the final product. Enhance equipment safety. By timely detecting and responding to abnormal situations, the risk of equipment damage and safety accidents is reduced. Optimize maintenance and operations. Through real-time analysis and feedback of data, machine maintenance is more targeted and predictive, reducing overall operating costs. Improve production efficiency, reduce downtime due to failures or abnormalities, and maintain production continuity and efficiency. Strengthen learning and adaptability. Through continuous training and optimization of deep learning models, the system can continuously improve its ability to detect new anomalies and adapt to changing environments. The integrated sensor monitoring and intelligent analysis solutions have greatly improved the operating performance and reliability of grinding and polishing robots, and are suitable for modern manufacturing environments that require high precision and high safety.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An industrial robot automatic control method, characterized in that: The method comprises: Use image recognition technology to extract the surface features of the workpiece to be polished, and control the initial force and torque of the end effector of the industrial robot based on the surface features of the workpiece to be polished; The temperature sensor of the end effector of the industrial robot is used to monitor the temperature fluctuation of the workpiece surface during the grinding process, and then adjust the force and torque of the end effector of the industrial robot, including: Deploy temperature sensors on the robot's end effector to monitor temperature fluctuations in real time during operation; Obtain temperature change data, and immediately use the specially constructed force adjustment model to optimize the force parameters, and then use the torque adjustment model to make corresponding adjustments to the torque parameters; Specifically, the force adjustment model is: in, represents the adjusted force at time t, represents the initial force of the end effector of the industrial robot at time t, μ represents the friction coefficient between the surface of the workpiece to be polished and the grinding head of the end effector of the industrial robot, Indicates instantaneous temperature fluctuations, represents the displacement of the end effector of the industrial robot at time t, represents the preset maximum displacement, m represents the mass of the grinding head of the industrial robot end effector, c represents the specific heat capacity of the workpiece to be ground, and D represents the temperature fluctuation coefficient. represents the temperature accumulation coefficient, represents the average temperature over time τ; The power and status data of the grinding operation are collected in real time, and deep learning technology is used to predict whether there are abnormal conditions. Once an abnormality is found, the grinding operation is terminated immediately.

2. The industrial robot automatic control method according to claim 1, characterized in that: Image recognition technology is used to extract the surface features of the workpiece to be polished, and the initial force and torque of the end effector of the industrial robot are controlled based on the surface features of the workpiece to be polished, including: Under uniform lighting conditions, a high-definition camera mounted on the end effector of an industrial robot is used to obtain multi-view images of the surface of the workpiece to be polished; Using filtering methods to remove potential noise points in the image, the filtering methods include median filtering and Gaussian filtering algorithms, and optimizing the contrast performance of the image by adaptively adjusting histogram equalization; After the image contrast is enhanced, the edge details in the image are captured by an edge detection operator, and then the image is reconstructed in three dimensions. After the three-dimensional reconstruction, the surface features of the workpiece are extracted using deep learning technology. The surface features of the workpiece include: the surface roughness of the workpiece to be polished, the Vickers hardness of the surface material of the workpiece to be polished, and the corrosion depth of the surface of the workpiece to be polished; After acquiring the surface features of the workpiece, the initial force and torque conditions are set for the end effector of the industrial robot according to the pre-established end effector force model and torque model.

3. The industrial robot automatic control method according to claim 2, characterized in that: Specifically, the pre-established end effector force model is: in, represents the contact length between the end effector of the industrial robot and the workpiece to be polished, t represents the processing time, m represents the mass of the grinding head of the end effector of the industrial robot, represents the initial force of the end effector of the industrial robot, represents the surface roughness of the workpiece to be polished, K represents the stiffness coefficient of the end effector of the industrial robot, represents the feed speed of the grinding head of the industrial robot end effector, β represents the deflection of the grinding head of the industrial robot end effector, represents the instantaneous value of the cutting force, η represents the stable machining coefficient, H represents the Vickers hardness of the surface material of the workpiece to be processed, D represents the corrosion depth of the surface of the workpiece to be polished, μ represents the friction coefficient between the surface of the workpiece to be polished and the grinding head of the end effector of the industrial robot, A represents the contact area between the grinding head of the end effector of the industrial robot and the surface of the workpiece to be polished, Indicates the relative movement speed of the industrial robot's grinding head contact point with the workpiece during linear motion.

4. The industrial robot automatic control method according to claim 2, characterized in that: Specifically, the pre-established end effector torque model is: in, represents the initial torque, Indicates the reference value of the initial torque, represents the torque constant, M represents the load mass, r represents the length of the lever arm, and J represents the moment of inertia. represents the damping coefficient, represents the nonlinear torque response adjustment parameter, represents the environmental impact correction factor, represents the reference speed of the end effector, V represents the linear speed of the end effector, The coefficient representing the rate of change of power, represents the power change rate, ϕ represents the power response nonlinear adjustment parameter, It represents the center angle of the circle corresponding to the grinding path of the end effector, and t represents the processing time.

5. The industrial robot automatic control method according to claim 1, characterized in that: Specifically, the torque adjustment model is: in, represents the torque at the current time t, r represents the arm length of the end effector of the industrial robot, represents the adjusted force at time t, represents the temperature influence coefficient, represents the force at time τ, and τ represents the integral variable, which is used to calculate the cumulative effect of the force from the starting time 0 to time t.

6. The industrial robot automatic control method according to claim 1, characterized in that: Real-time collection of power and status data during the grinding process, use deep learning technology to predict whether there are abnormal conditions, and immediately terminate the grinding operation once an abnormality is found, including: A variety of sensors are deployed on the grinding and polishing robots, including force sensors, torque sensors, temperature sensors, and vibration sensors, to comprehensively collect power and status information during the processing. The collection frequency is pre-set according to the characteristics of the grinding process and the frequency of data changes; Use filtering algorithms to remove noise from the data collected during the processing; For power data, features are extracted from power data such as grinding force and torque, including the change trend of force, peak and valley values ​​of torque, and fluctuation frequency of force and torque. For state data, for temperature and vibration state data, the temperature change rate, temperature non-uniformity index, and spectral characteristics of vibration and changes in vibration amplitude are extracted; Construct an initial deep learning model, i.e., a long short-term memory network model, collect polishing operation data under normal and abnormal conditions, divide the data into training set, validation set, and test set according to a preset ratio, and annotate the training set data into normal state data and abnormal state data; The constructed initial long short-term memory network model is trained using the training set data, and the model parameters, including weights and biases, are adjusted so that the model can learn the characteristic differences between normal and abnormal state data, and the prediction accuracy of the model is improved by minimizing the cross entropy loss function. During the training process, the model is verified using the validation set data to prevent the model from overfitting; Use the test set data to evaluate the trained model, calculate the model's accuracy, recall, and F1 value, evaluate the model's performance, and optimize the model based on the evaluation results, including adjusting the model structure, hyperparameters, and increasing the amount of training data; Deploy the trained and evaluated deep learning model to the real-time monitoring system of the polishing operation; The polishing operation data collected and pre-processed in real time is input into the deployed deep learning model, and the deep learning model predicts whether there is an abnormal condition in the current polishing operation based on the learned feature pattern and outputs the prediction result; Once the model predicts an abnormality, a stop command is immediately sent to the grinding equipment, and the grinding operation is terminated through the control interface to prevent the abnormal situation from further deteriorating and prevent greater damage to the equipment and workpiece.

7. An industrial robot automatic control device, characterized in that: The device comprises: The preliminary control module uses image recognition technology to extract the surface features of the workpiece to be polished, and controls the initial force and torque of the end effector of the industrial robot based on the surface features of the workpiece to be polished; A control and adjustment module, used to monitor the temperature fluctuation of the workpiece surface during the grinding process through the temperature sensor of the industrial robot end effector, and then adjust the force and torque of the industrial robot end effector, including: Deploy temperature sensors on the robot's end effector to monitor temperature fluctuations in real time during operation; Obtain temperature change data, and immediately use the specially constructed force adjustment model to optimize the force parameters, and then use the torque adjustment model to make corresponding adjustments to the torque parameters; Specifically, the force adjustment model is: in, represents the adjusted force at time t, represents the initial force of the end effector of the industrial robot at time t, μ represents the friction coefficient between the surface of the workpiece to be polished and the grinding head of the end effector of the industrial robot, Indicates instantaneous temperature fluctuations, represents the displacement of the end effector of the industrial robot at time t, represents the preset maximum displacement, m represents the mass of the grinding head of the industrial robot end effector, c represents the specific heat capacity of the workpiece to be ground, and D represents the temperature fluctuation coefficient. represents the temperature accumulation coefficient, represents the average temperature over time τ; The abnormality prediction module is used to collect power and status data during the grinding operation in real time, and use deep learning technology to predict whether there are abnormal conditions. Once an abnormality is found, the grinding operation is immediately terminated.

8. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 6.

Images