Industrial robot medical rehabilitation equipment manufacturing system

Abstract

The invention relates to the field of industrial automation, and particularly provides an industrial robot medical rehabilitation equipment manufacturing system. Based on an assembly algorithm of multi-sensor fusion and visual servo, the assembly precision of precise parts is improved; a material characteristic sensing and self-adaptive control algorithm is adopted to optimize a flexible material processing technology; a multi-modal data fusion and deep learning algorithm is adopted, so that the quality detection reliability is enhanced; a reinforcement learning algorithm optimizes a processing path and energy consumption; the invention discloses a fault prediction and health management algorithm reform maintenance mode. The algorithms comprehensively improve the production efficiency and the product quality, reduce the cost, guarantee the high-performance operation of equipment, and inject strong power for the development of the medical rehabilitation equipment manufacturing industry.

Description

Technical Field

[0001] The present invention belongs to the field of industrial automation and relates to an industrial robot medical rehabilitation equipment manufacturing system.

[0002] Semiconductor photolithography is a critical step in transferring the circuit pattern on the mask onto the surface of a semiconductor wafer coated with photoresist. It determines the chip's feature size and performance. With the continuous advancement of semiconductor technology, chip integration is increasing, and the precision and quality requirements of the photolithography process are becoming increasingly stringent. Industrial robots play a vital role in the operation and control of semiconductor photolithography equipment. However, existing industrial robot technology still has many problems in the photolithography process, which seriously restricts the further development of semiconductor photolithography technology.

[0003] Lithography positioning accuracy is difficult to guarantee: During the lithography process, the pattern on the mask must be precisely aligned to a specific location on the wafer. This positioning accuracy directly impacts chip performance and yield. Traditional industrial robotic lithography positioning systems rely primarily on simple visual positioning technology and preset alignment parameters, which cannot accurately adapt to factors such as microscopic fluctuations on the wafer surface, thermal expansion, and mechanical vibration. Furthermore, due to long-term operation of the lithography equipment and changes in environmental factors, the positioning system's accuracy gradually decreases, resulting in increased alignment deviation of the lithography pattern, affecting chip performance and reliability.

[0004] Poor photolithography pattern transfer accuracy: Photolithography pattern transfer accuracy refers to the ability to accurately replicate the pattern on the mask onto the photoresist layer. Existing industrial robotic photolithography systems are susceptible to factors such as uneven photoresist thickness, inconsistent exposure dose, and fluctuations in development process parameters during the pattern transfer process, leading to deformation, blurring, and distortion of the photolithography pattern. These issues not only affect chip performance but also increase chip manufacturing costs.

[0005] Photoresist coating uniformity is difficult to control: Photoresist coating uniformity is crucial to the quality of photolithographic patterns and the stability of the photolithographic process. Currently, industrial robots primarily use fixed coating parameters and preset coating paths during the photoresist coating process, which are unable to adjust in real time based on wafer size, shape, and surface characteristics. Furthermore, due to structural and process limitations of the coating equipment, the photoresist coating thickness can easily become uneven across the wafer surface, impacting both the quality of the photolithographic patterns and the stability of the photolithographic process. Summary of the Invention

[0006] The present invention provides an industrial robot medical rehabilitation equipment manufacturing system, including a high-precision assembly algorithm for precision parts based on multi-sensor fusion and visual servoing. This algorithm integrates high-precision force sensors, laser displacement sensors, and visual sensors on an industrial robot. It utilizes multi-sensor fusion technology to fuse and process data collected by these sensors to construct a real-time state model of the parts. Combined with a visual servo algorithm, the system adjusts the robot's motion trajectory and posture in real time based on the fused state information, achieving high-precision assembly of precision parts. In the visual servo algorithm, the adjustment amount for the robot's motion trajectory based on the fused state information is calculated using the following formula:

[0007] ΔT = K·F(Sforce, Slaser, Svision), where ΔT represents the adjustment of the industrial robot's trajectory, including position and posture adjustment information. K is an adjustment coefficient matrix, whose elements are derived from a large amount of experimental data fitting different types of component assembly scenarios and are used to weight the calculation results to adapt to different assembly requirements.

[0008] F (Sforce, Slaser, Svision) is a comprehensive calculation function based on force sensor data Sforce, laser displacement sensor data Slaser, and vision sensor data Svision. After analyzing and processing this data using multi-sensor fusion technology, this data is input into the function to quantify the motion trajectory adjustment amount. The precise adjustment amount is then calculated based on the formula, ensuring that the industrial robot can accurately adjust the motion trajectory according to the real-time status of the components, achieving high-precision assembly of precision components.

[0009] Furthermore, the multi-sensor fusion adopts a fusion algorithm based on deep learning, such as a fusion network based on an attention mechanism, and uses a convolutional neural network to analyze and process the fused data to extract key feature information for building a real-time status model of components.

[0010] Furthermore, it also includes a flexible material processing technology optimization algorithm based on real-time perception and adaptive control of material properties. The algorithm installs strain sensors, temperature sensors and pressure sensors on the processing tools of industrial robots to monitor the strain, temperature and pressure changes in the flexible material processing process in real time. By establishing a characteristic model of the flexible material and combining it with an adaptive control algorithm, the processing parameters are automatically adjusted according to the real-time perceived material characteristic data.

[0011] Furthermore, the adaptive control algorithm adjusts the processing speed and pressure according to the strain and pressure data, and adjusts the temperature of the processing tool according to the temperature data, so as to optimize the processing technology of the flexible material.

[0012] Furthermore, it also includes a medical rehabilitation equipment quality inspection algorithm based on multimodal data fusion and deep learning. The algorithm uses visual inspection equipment, ultrasonic inspection equipment and infrared inspection equipment to obtain multimodal data of the equipment, and uses multimodal data fusion technology to fuse and process the data collected by different inspection equipment to construct a comprehensive quality information model of the equipment. It also combines deep learning algorithms to analyze and learn the fused comprehensive quality information, and establish a mapping relationship model between quality characteristics and defect types.

[0013] Furthermore, the deep learning algorithm uses a combination of convolutional neural networks and recurrent neural networks to analyze and process the fused comprehensive quality information to quickly and accurately identify various quality defects of the equipment.

[0014] Furthermore, it also includes an industrial robot processing path planning and energy consumption optimization algorithm based on reinforcement learning. The algorithm defines the processing process of the industrial robot as a Markov decision process, clearly defines the state space including the robot's position, posture, processing task progress and environmental information, and the action space including various motion actions of the robot. It also designs a reward function that uses processing efficiency, processing quality and energy consumption as reward indicators, and uses the reinforcement learning algorithm to enable the robot to learn the optimal processing path planning strategy.

[0015] Furthermore, the reinforcement learning algorithm adopts a deep Q network or a proximal strategy optimization algorithm to enable the robot to optimize the processing path in continuous interaction with the environment, improve processing efficiency and quality, and reduce energy consumption.

[0016] Furthermore, it also includes an industrial robot maintenance decision algorithm based on fault prediction and health management. The algorithm installs vibration sensors, current sensors and temperature sensors on the key components of the industrial robot to collect the robot's operating status data in real time, uses big data analysis technology to process historical data and real-time data to establish a fault prediction model, combines machine learning algorithms to evaluate and predict the robot's health status, and formulates maintenance plans and decisions based on the prediction results.

[0017] Furthermore, the machine learning algorithm uses support vector machines and random forest algorithms to analyze robot operating status data, evaluate and predict the robot's health status, and perform preventive maintenance.

[0018] Beneficial effects

[0019] A high-precision assembly algorithm for precision components based on multi-sensor fusion and visual servoing has revolutionized the assembly process. Multiple sensors work together to comprehensively perceive component information and build an accurate real-time state model. This enables industrial robots to precisely grasp the position, posture, and force conditions of components during assembly, allowing for extremely precise adjustments to motion trajectory and posture. This effectively avoids assembly errors caused by positioning deviations and uneven force distribution, significantly improving assembly accuracy and laying a solid foundation for the high-performance operation of medical rehabilitation equipment.

[0020] A flexible material processing optimization algorithm based on real-time sensing and adaptive control of material properties has revolutionized flexible material processing. Sensors monitor material strain, temperature, and pressure changes in real time, and dynamically adjust processing parameters based on characteristic models and adaptive control algorithms. This enables industrial robots to adapt to the complex and ever-changing properties of flexible materials, effectively preventing problems such as deformation and tearing, significantly improving product quality, and reducing material waste and production delays caused by processing errors, significantly increasing production efficiency.

[0021] A medical rehabilitation equipment quality inspection algorithm based on multimodal data fusion and deep learning has significantly enhanced the reliability and accuracy of quality inspections. By integrating multimodal data acquired by various testing devices and analyzing it through deep learning, this algorithm can comprehensively and deeply explore the quality characteristics of the equipment and accurately identify potential quality defects. Compared to traditional single-method inspection methods, this algorithm can detect more subtle and critical quality issues, effectively ensuring product quality, reducing the risk of substandard products entering the market, and safeguarding the health and safety of patients.

[0022] Reinforcement learning-based industrial robot machining path planning and energy optimization algorithms significantly improve machining efficiency and resource utilization. By defining a comprehensive state space, action space, and reward function, industrial robots optimize machining path planning strategies through continuous practice. This not only improves machining efficiency and ensures efficient task completion, but also reduces energy consumption, achieving energy-saving production and fully utilizing the equipment's maximum performance.

[0023] Industrial robot maintenance decision-making algorithms based on fault prediction and health management are revolutionizing industrial robot maintenance. Sensors collect real-time operating status data on key components. Through big data analysis and machine learning algorithms, they can accurately predict potential failure risks in advance. This data allows companies to develop preventive maintenance plans and proactively address issues, significantly reducing downtime, extending equipment lifespan, lowering maintenance costs, and ensuring continuous and stable production. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure (1) Flowchart of system operation principle. DETAILED DESCRIPTION

[0025] Example 1

[0026] 1. Background of the Embodiments

[0027] A professional medical rehabilitation equipment manufacturer specializes in producing high-end rehabilitation robots, prosthetics, and other products. As market demands for product precision and quality continue to rise, the company faces challenges in production, including low precision assembly accuracy for precision components, inconsistent processing quality for flexible materials, and poor reliability in equipment quality testing. These challenges severely impact product competitiveness and production efficiency. To address these challenges, the company introduced an industrial robotic medical rehabilitation equipment manufacturing system.

[0028] 2. System Implementation Process

[0029] (1) Implementation of high-precision assembly algorithm for precision parts based on multi-sensor fusion and visual servoing

[0030] Sensor Deployment: High-precision force sensors with an accuracy of ±0.01N are installed on the end effectors of industrial robots, precisely sensing changes in contact force during assembly. Laser displacement sensors with a measurement accuracy of ±0.001mm are installed on the workbench to accurately obtain component position and posture information. Furthermore, high-resolution vision sensors with 8000 x 6000 pixels capture detailed images of components. These sensors collect data in real time at a rate of 50 times per second and transmit it to the data processing unit via a high-speed network.

[0031] Data fusion and model building: The data processing unit uses a fusion network based on an attention mechanism to fuse data collected by force sensors, laser displacement sensors, and visual sensors. For example, when assembling key components of a prosthesis, the force sensor detects that the contact force reaches 0.1N instantaneously during assembly. The laser displacement sensor measures a deviation of 0.003mm in the X-axis direction, and the visual sensor identifies an angular deviation of 0.1°. The fusion network fuses these data based on pre-set weights to construct a real-time status model of the component. A convolutional neural network is used to analyze the fused data and extract key features of the component, such as position, posture, and surface texture.

[0032] Visual servo algorithm execution: Based on the constructed real-time state model of components, the formula

[0033] ΔT=K·F(Sforce, Slaser, Svision) calculates the adjustment amount of the motion trajectory of the industrial robot. Among them, the adjustment coefficient matrix K is obtained by analyzing a large number of experimental data on the assembly of different types of prosthetic parts. Its element values ​​have been adapted for this assembly scenario. The comprehensive calculation function F(Sforce, Slaser, Svision) processes the sensor data to obtain the quantitative value of the motion trajectory adjustment amount. For example, it is calculated that the robot needs to move 0.003mm in the X-axis direction and the rotation angle needs to be adjusted by 0.1°. The robot control system adjusts the robot's motion trajectory and posture in real time based on these adjustments to complete the assembly task of precision parts. During the assembly process, sensor data is continuously monitored in real time, and the robot motion is dynamically adjusted to ensure that the assembly accuracy is within ±0.002mm. Compared with traditional assembly methods, the accuracy is improved by more than 5 times.

[0034] (2) Implementation of flexible material processing optimization algorithm based on real-time perception and adaptive control of material properties

[0035] Sensor Installation: Strain sensors with a resolution of ±0.001με are installed on the processing tools of the industrial robot to monitor strain changes during flexible material processing in real time. Temperature sensors with an accuracy of ±0.1°C precisely measure material temperature. Pressure sensors with a measurement accuracy of ±0.001MPa accurately sense pressure during processing. These sensors collect data 30 times per second and transmit it to the control system via a network.

[0036] Characteristic Model Establishment and Algorithm Implementation: The control system establishes characteristic models for flexible materials such as silicone and rubber based on extensive experimental data and material property theory. For example, for silicone, a mathematical model was developed based on the relationships between strain, temperature, and pressure and processing parameters. When the strain sensor detects that the strain of the silicone material reaches 500με during processing, the adaptive control algorithm, based on the characteristic model and real-time data, calculates that the processing speed needs to be reduced from 50mm / s to 30mm / s and the pressure needs to be adjusted from 0.5MPa to 0.4MPa. When the temperature sensor detects that the material temperature has risen to 40°C, the temperature of the processing tool is automatically reduced from 35°C to 32°C.

[0037] Optimized Processing: Industrial robots adjust processing parameters in real time based on control instructions generated by the control system. When machining the flexible joints of prosthetic limbs, this continuous, real-time adjustment of processing parameters based on changes in material properties effectively avoids problems such as material deformation and tearing. This has increased the product qualification rate from 70% to over 90%, boosting production efficiency by 30%.

[0038] (3) Implementation of medical rehabilitation equipment quality inspection algorithm based on multimodal data fusion and deep learning

[0039] Testing equipment deployment: High-resolution visual inspection equipment is deployed on the quality inspection production line of medical rehabilitation equipment, which can detect defects of 0.01mm on the surface of the equipment. Ultrasonic testing equipment can detect defects of 1mm inside the equipment. 3 Infrared detection equipment can accurately measure the surface temperature distribution of equipment with an accuracy of ±0.2°C. These detection devices collect data at a frequency of 20 times per second and transmit it to the data processing center via the network.

[0040] Data Fusion and Feature Extraction: The data processing center uses a multimodal data fusion algorithm based on deep learning to fuse data collected by visual, ultrasonic, and infrared detection equipment. For example, when inspecting a rehabilitation robot, the visual inspection equipment detected a suspected scratch on the device's surface, the ultrasonic inspection equipment detected an abnormal echo in a certain part of the device's interior, and the infrared inspection equipment indicated that the temperature in that part was slightly above the normal range. Using multimodal data fusion technology, this data is integrated to construct a comprehensive quality information model for the equipment. A convolutional neural network is used to analyze the fused data and extract quality characteristics of the equipment, such as surface defect types, internal structural anomalies, and areas of abnormal temperature distribution.

[0041] Model Training and Testing: Quality data from over 5,000 pieces of various medical rehabilitation equipment was collected to construct a training dataset. A deep learning algorithm combining convolutional neural networks and recurrent neural networks was used to train and analyze the training dataset, establishing a quality testing model. During the actual testing process, the extracted quality features were input into the trained model, enabling rapid and accurate identification of various quality defects in the equipment. For example, a comprehensive inspection of a rehabilitation robot can be completed within one minute, achieving an accuracy rate exceeding 95%, an increase of over 25% compared to traditional single-method testing methods.

[0042] (IV) Implementation of industrial robot processing path planning and energy consumption optimization algorithm based on reinforcement learning

[0043] State, Action, and Reward Definitions: Define the state space of the industrial robot's machining process, including the robot's position accuracy (±0.01mm), attitude angle accuracy (±0.1°), machining task progress, and ambient temperature and humidity information. The action space encompasses the robot's various motions, including linear motion, rotational motion, and speed adjustment. Design a reward function that rewards the robot with a +10 reward when it completes the machining task within the specified timeframe with high quality and energy consumption below a set threshold. If quality issues or excessive energy consumption occur, a -5 penalty is applied.

[0044] Model Training and Application: Using a proximal policy optimization algorithm, the robot learns and trains the machining process of an industrial robot, establishing a machining path planning and energy optimization model. When machining the metal frame of a rehabilitation robot, the robot selects actions based on its current state, such as adjusting speed and direction, and adjusts its strategy based on reward feedback after executing the action. After a period of training, the robot learns the optimal machining path planning strategy, improving machining efficiency by 40% and reducing energy consumption by 20%.

[0045] (V) Implementation of industrial robot maintenance decision-making algorithm based on fault prediction and health management

[0046] Sensor installation and data acquisition: Vibration sensors are installed on key components of industrial robots, such as motors and reducers, which can detect vibrations of 0.01 mm / s. 2 Vibration changes are detected. A current sensor with an accuracy of ±0.1A monitors motor current. A temperature sensor with an accuracy of ±0.1°C collects the temperature of key components in real time. These sensors collect data 10 times per second and transmit it to a data analysis system via the network.

[0047] Fault prediction and health assessment: The data analysis system utilizes big data analytics to analyze and process collected historical and real-time data. For example, a motor fault prediction model was established through long-term monitoring and analysis of motor vibration, current, and temperature data. Combining support vector machines and random forest algorithms, the robot's health status is assessed and predicted. If the vibration sensor detects an abnormal increase in motor vibration, as well as unusual changes in current and temperature, the model predicts that the motor may fail within the next 24 hours.

[0048] Maintenance Decision Execution: Based on failure prediction and health status assessment results, appropriate maintenance plans and decisions are developed. When an impending motor failure is predicted, maintenance personnel are preemptively dispatched to inspect and repair the motor and replace any potentially problematic parts. This preventative maintenance approach has reduced industrial robot downtime by 60%, significantly improving equipment reliability and service life.

Claims

1. An industrial robot medical rehabilitation equipment manufacturing system, characterized in that: This includes a high-precision assembly algorithm for precision parts based on multi-sensor fusion and visual servoing. This algorithm integrates high-precision force sensors, laser displacement sensors, and visual sensors on industrial robots. It uses multi-sensor fusion technology to fuse the data collected by different sensors to build a real-time state model of the parts. Combined with the visual servo algorithm, the motion trajectory and posture of the industrial robot are adjusted in real time based on the fused state information, achieving high-precision assembly of precision parts. In the visual servo algorithm, the adjustment amount of the industrial robot's motion trajectory is calculated based on the fused state information using the following formula: ΔT = K·F(Sforce, Slaser, Svision), where ΔT represents the adjustment of the industrial robot's motion trajectory, including position and posture adjustment information. K is an adjustment coefficient matrix, whose element values ​​are derived from a large amount of experimental data fitting different types of component assembly scenarios and are used to weight the calculation results to adapt to different assembly requirements. F (Sforce, Slaser, Svision) is a comprehensive calculation function based on force sensor data Sforce, laser displacement sensor data Slaser, and visual sensor data Svision. After analyzing and processing these data using multi-sensor fusion technology, the function is input to quantify the motion trajectory adjustment amount. The precise adjustment amount is then calculated based on the formula, ensuring that the industrial robot can accurately adjust the motion trajectory according to the real-time status of the components and achieve high-precision assembly of precision components.

2. The industrial robot medical rehabilitation equipment manufacturing system according to claim 1, characterized in that: The multi-sensor fusion adopts a fusion algorithm based on deep learning, such as a fusion network based on an attention mechanism, and uses a convolutional neural network to analyze and process the fused data to extract key feature information for building a real-time status model of components.

3. The industrial robot medical rehabilitation equipment manufacturing system according to claim 1, characterized in that: It also includes a flexible material processing technology optimization algorithm based on real-time perception and adaptive control of material properties. The algorithm installs strain sensors, temperature sensors and pressure sensors on the processing tools of industrial robots to monitor the strain, temperature and pressure changes in the flexible material processing process in real time. By establishing a characteristic model of the flexible material and combining it with an adaptive control algorithm, the processing parameters are automatically adjusted according to the real-time perceived material characteristic data.

4. The industrial robot medical rehabilitation equipment manufacturing system according to claim 3, characterized in that: The adaptive control algorithm adjusts the processing speed and pressure according to the strain and pressure data, and adjusts the temperature of the processing tool according to the temperature data, so as to optimize the processing technology of the flexible material.

5. The industrial robot medical rehabilitation equipment manufacturing system according to claim 1, characterized in that: It also includes a medical rehabilitation equipment quality inspection algorithm based on multimodal data fusion and deep learning. The algorithm uses visual inspection equipment, ultrasonic inspection equipment and infrared inspection equipment to obtain multimodal data of the equipment, and uses multimodal data fusion technology to fuse and process the data collected by different inspection equipment to build a comprehensive quality information model of the equipment. It also combines deep learning algorithms to analyze and learn the fused comprehensive quality information, and establish a mapping relationship model between quality characteristics and defect types.

6. The industrial robot medical rehabilitation equipment manufacturing system according to claim 5, characterized in that: The deep learning algorithm uses a combination of convolutional neural networks and recurrent neural networks to analyze and process the fused comprehensive quality information to quickly and accurately identify various quality defects of the equipment.

7. The industrial robot medical rehabilitation equipment manufacturing system according to claim 1, characterized in that: It also includes an industrial robot processing path planning and energy consumption optimization algorithm based on reinforcement learning. The algorithm defines the processing process of the industrial robot as a Markov decision process, clearly defines the state space that includes the robot's position, posture, processing task progress and environmental information, and the action space that includes various robot motions. It also designs a reward function that uses processing efficiency, processing quality and energy consumption as reward indicators, and uses reinforcement learning algorithms to enable the robot to learn the optimal processing path planning strategy.

8. The industrial robot medical rehabilitation equipment manufacturing system according to claim 7, characterized in that: The reinforcement learning algorithm uses a deep Q network or a proximal strategy optimization algorithm to enable the robot to optimize the processing path in continuous interaction with the environment, improve processing efficiency and quality, and reduce energy consumption.

9. The industrial robot medical rehabilitation equipment manufacturing system according to claim 1, characterized in that: It also includes an industrial robot maintenance decision algorithm based on fault prediction and health management. The algorithm installs vibration sensors, current sensors and temperature sensors on the key components of the industrial robot to collect the robot's operating status data in real time, uses big data analysis technology to process historical data and real-time data to establish a fault prediction model, combines machine learning algorithms to evaluate and predict the robot's health status, and formulates maintenance plans and decisions based on the prediction results.

10. The industrial robot medical rehabilitation equipment manufacturing system according to claim 9, characterized in that: The machine learning algorithm uses a support vector machine and a random forest algorithm to analyze the robot's operating status data, evaluate and predict the robot's health status, and perform preventive maintenance.

Images