AI robot polishing quality prediction system based on digital twin drive

Abstract

The invention relates to the technical field of robot polishing quality prediction, and discloses an AI robot polishing quality prediction system based on digital twin drive, and the system comprises a process parameter collection module which can dynamically collect parameters such as processing track offset, judge quality deviation risk and trigger collaborative analysis; the surface topography analysis module extracts surface contour features through a white light interferometer, and quantitatively evaluates texture uniformity; the thermal distribution analysis module collects temperature field data and deduces a thermal coupling effect trend; the quality prediction fusion module is combined with the multi-dimensional data to generate a quality performance fusion index and track compensation and cooling adjustment signals; a process optimization generation module generates robot attitude compensation and cooling intensity adjustment parameters according to the parameters; the digital twin modeling module monitors virtual-real mapping consistency and outputs model confidence to correct a fusion index; and the dynamic optimization execution module forms closed-loop control. According to the system, accurate prediction and dynamic optimization of the polishing quality are achieved, and the machining quality and the production efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot polishing quality prediction, and specifically to an AI robot polishing quality prediction system driven by digital twins. Background Art

[0002] In modern manufacturing, robotic polishing, a key surface treatment process, is widely used in high-end fields such as aerospace, automotive manufacturing, and precision instruments. Its processing quality directly impacts product performance, lifespan, and appearance. However, the polishing process is influenced by multiple factors, posing numerous challenges to quality control.

[0003] From a process parameter perspective, the dynamic changes in parameters such as machining path offset, tool wear, and material hardness during robotic polishing operations are difficult to accurately and in real time. Traditional methods often fail to comprehensively and dynamically collect and analyze these parameters, making it difficult to identify quality deviation risks in advance. Quality issues are often addressed only after they arise, impacting production efficiency and increasing costs.

[0004] In terms of surface topography, there is a lack of effective technical means for extracting surface profile characteristic parameters and quantitatively evaluating surface texture uniformity in the polished area of ​​a workpiece. While the use of equipment such as white-light interferometers can obtain certain data, the weak link in traditional technologies is how to accurately extract key information from height distribution cloud maps and power spectrum density maps and fuse them to obtain a reliable surface topography consistency index.

[0005] Thermal distribution analysis also has its shortcomings. Extracting the workpiece surface temperature gradient and residual stress distribution parameters is difficult, and the constructed thermal coupling prediction model is not highly accurate. This makes it difficult to accurately deduce the trend of the material's thermal coupling effects and to obtain reliable thermal stability assessments in a timely manner. This makes it difficult to effectively control thermal deformation and other issues during the polishing process.

[0006] In terms of quality prediction, traditional quality prediction methods usually consider certain factors in isolation and are unable to synergistically predict factors such as surface morphology, thermal distribution, and tool wear. This results in inaccurate compensation and adjustment signals being generated, making it difficult to achieve precise control of polishing quality.

[0007] During process optimization, polishing strategy matching lacks flexibility and real-time performance. When trajectory compensation or cooling adjustments are required, traditional methods cannot quickly generate accurate robot posture compensation parameters and cooling intensity adjustment parameters based on real-time signals, impacting polishing results and production efficiency.

[0008] The application of digital twin technology in the polishing field is still in the development stage. The monitoring of the mapping consistency between the physical polishing system and the virtual model is not perfect, and the calculation and processing of data synchronization errors and state response delays are not accurate enough, resulting in inaccurate model confidence assessment values ​​and unable to provide a reliable reference for quality prediction. Summary of the Invention

[0009] The purpose of the present invention is to provide an AI robot polishing quality prediction system driven by digital twins to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention provides the following technical solution: an AI robot polishing quality prediction system driven by digital twins, the system comprising: A process parameter acquisition module is used to dynamically collect the machining trajectory offset, tool wear, and material hardness parameters of the robot polishing operation to obtain a process state parameter set. Based on the process state parameter set, the module determines and analyzes the quality deviation risk, generates a quality anomaly signal, triggers a collaborative analysis instruction based on the generated quality anomaly signal, and executes the surface morphology analysis module and the thermal distribution analysis module based on the triggered collaborative analysis instruction; Surface topography analysis module, used to extract surface profile characteristic parameters of the workpiece polishing area, quantitatively evaluate the surface texture uniformity, and obtain the surface topography consistency index; Thermal distribution analysis module, used to extract the surface temperature gradient and residual stress distribution parameters of the workpiece, deduce the trend of the material's thermal-mechanical coupling effect, and obtain the thermal stability evaluation value; The quality prediction fusion module is used to receive the surface topography consistency index and thermal stability evaluation value, make collaborative predictions on the polishing quality performance, and generate trajectory compensation signals and cooling adjustment signals; The process optimization generation module is used to receive the trajectory compensation signal and the cooling adjustment signal, perform polishing strategy matching, and generate the robot posture compensation parameters and cooling intensity adjustment parameters; The digital twin modeling module is used to monitor the mapping consistency between the physical polishing system and the virtual model, extract data synchronization errors and state response delays, and generate model confidence assessment values; The quality prediction fusion module further combines the model confidence evaluation value to dynamically correct the quality performance fusion index.

[0011] Preferably, the determination and analysis of the quality deviation risk includes: By collecting the robot machining trajectory offset in real time, the Euclidean distance between it and the theoretical path is calculated and marked as the trajectory offset; Extract the radial wear depth and end face wear area from the tool wear parameters, calculate their geometric mean, and mark them as tool wear characteristic values; Collect the Rockwell hardness value and microhardness dispersion of the material hardness parameters, perform normalized weighted calculation, and obtain the comprehensive material hardness index; The trajectory deviation, tool wear characteristic value and material hardness comprehensive index are compared with the preset thresholds respectively. When any parameter exceeds the corresponding threshold, a quality abnormality signal is generated.

[0012] Preferably, the quantitative evaluation of surface texture uniformity includes: Acquire three-dimensional surface topography data through white light interferometry to generate height distribution cloud map and power spectrum density map; Extract the contour arithmetic mean deviation and peak-to-valley height difference from the height distribution cloud map, calculate the ratio of the two and take the logarithm to obtain the texture uniformity coefficient; Extract the mid-frequency energy concentration and spectrum symmetry from the power spectrum density diagram, calculate the harmonic mean of the two, and mark them as the spectrum characteristic index; The texture uniformity coefficient and the spectral characteristic index are weightedly fused to obtain the surface morphology consistency index.

[0013] Preferably, the trend deduction of the thermal-mechanical coupling effect of the material includes: Collect workpiece surface temperature field data in real time, extract temperature gradient amplitude and thermal cycle characteristic parameters; Construct a thermal-mechanical coupling prediction model, input the temperature gradient amplitude into the model for Gaussian process regression processing, and output the probability of thermal deformation in the future time interval; Extract the heating rate and cooling rate from the thermal cycle characteristic parameters, and calculate the absolute difference between the two to obtain the thermal cycle stability factor; The thermal deformation probability and the thermal cycle stability factor are nonlinearly combined to obtain the thermal stability evaluation value.

[0014] Preferably, the collaborative prediction of polishing quality performance includes: Retrieve tool wear characteristic value data, set its influence weight, and obtain the wear influence correction value through calculation and processing; The surface morphology consistency index, thermal stability evaluation value and wear effect correction value are normalized and calculated to generate the quality performance fusion index; A quality performance judgment threshold is set. If the fusion index is lower than the threshold, a trajectory compensation signal is generated; if it is higher than the threshold, a cooling adjustment signal is generated.

[0015] Preferably, the polishing strategy matching includes: If the trajectory compensation signal is captured, the posture adjustment command is triggered, and the robot's axial displacement and angle offset are dynamically configured according to the command to generate the robot's posture compensation parameters; If a cooling adjustment signal is captured, the thermal management instruction is triggered, and the flow parameters and injection coverage range of the cooling medium are optimized according to the instruction to generate the cooling intensity adjustment parameters.

[0016] Preferably, monitoring the mapping consistency between the physical polishing system and the virtual model includes: The physical system state data and the virtual model output data are collected by comparing the timestamps, and the root mean square error between the two is calculated and marked as the data synchronization error; Extract the physical response time and virtual response time at the time the key event is triggered, calculate the absolute difference between the two, and mark it as the state response delay; The data synchronization error and the state response delay are weighted and summed to generate a model confidence evaluation value.

[0017] Preferably, as described above, the AI ​​robot polishing quality prediction system based on digital twin drive further includes: The dynamic optimization execution module is used to adjust the feed speed of the polishing operation in real time according to the generated robot posture compensation parameters and cooling intensity adjustment parameters, and feed back the execution results to the process parameter acquisition module to form a closed-loop control link.

[0018] Preferably, the specific execution process of the dynamic optimization execution module includes: If the robot posture compensation parameters involve axial displacement adjustment, the sampling frequency of the force sensor should be increased simultaneously; If the cooling intensity adjustment parameters involve changes in flow parameters, the spatial density of temperature monitoring points will be increased simultaneously; Input the adjusted parameters into the process parameter acquisition module and restart the prediction process.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The process parameter acquisition module can dynamically collect parameters such as machining trajectory offset, tool wear and material hardness to form a process status parameter set. By analyzing these parameters, it can determine the quality deviation risk, generate quality abnormality signals in a timely manner and trigger collaborative analysis instructions, providing an accurate basis for subsequent surface morphology and thermal distribution analysis, so that quality problems can be discovered early, avoiding the expansion of problems and improving the timeliness and effectiveness of quality control.

[0020] The surface topography analysis module uses a white light interferometer to obtain three-dimensional surface topography data. By extracting parameters such as the arithmetic mean deviation of the profile and the peak-to-valley height difference and performing calculations and fusion, it obtains the surface topography consistency index, thereby achieving a quantitative assessment of the surface texture uniformity. This provides key data support for accurately judging the quality of the polished surface and helps to improve the consistency and reliability of the polished surface.

[0021] The thermal distribution analysis module collects workpiece surface temperature field data in real time, constructs a thermal coupling prediction model for Gaussian process regression processing, and combines it with thermal cycle characteristic parameter analysis to obtain a thermal stability evaluation value. It can accurately deduce the trend of the material's thermal-mechanical coupling effect, providing a scientific basis for controlling thermal deformation and other problems, and improving the thermal stability of the workpiece during the polishing process.

[0022] The quality prediction fusion module combines multiple data such as surface morphology consistency index, thermal stability evaluation value, tool wear characteristic value, etc. to generate a quality performance fusion index, and generates trajectory compensation signal and cooling adjustment signal based on the index, realizing the coordinated prediction of polishing quality performance, making the prediction results more comprehensive and accurate, and being able to take corresponding measures in time for different quality conditions, thereby improving the prediction accuracy and control ability of polishing quality.

[0023] The process optimization generation module quickly generates robot posture compensation parameters and cooling intensity adjustment parameters based on the trajectory compensation signal and cooling adjustment signal, achieving precise matching of polishing strategies and enabling the robot to dynamically adjust posture and cooling intensity according to actual conditions, thereby improving the adaptability and flexibility of the polishing process and further ensuring polishing quality.

[0024] The digital twin modeling module monitors the mapping consistency between the physical polishing system and the virtual model, calculates data synchronization error and state response delay, and generates a model confidence assessment value. The quality prediction fusion module dynamically corrects the quality performance fusion index based on this assessment value, thereby improving the reliability of the digital twin model and the accuracy of quality prediction, enabling the virtual model to more accurately reflect the state of the physical system.

[0025] The dynamic optimization execution module adjusts the feed speed of the polishing operation in real time according to the generated compensation and adjustment parameters, and feeds back the execution results to the process parameter acquisition module to form a closed-loop control link, realizing continuous optimization and dynamic adjustment of the polishing process, enabling the entire polishing system to continuously adapt to various changes, further improving polishing quality and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a working principle diagram of the AI ​​robot polishing quality prediction system based on digital twin drive described in the present invention; Figure 2Design diagram for surface texture uniformity assessment; Figure 3 Design diagram for deduction of thermal-mechanical coupling effect; Figure 4 Design drawings for collaborative prediction of quality performance; Figure 5 Mapping monitoring designs for digital twins. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] See also Figure 1-Figure 5 The present invention relates to an AI robot polishing quality prediction system based on digital twins. The system includes a process parameter acquisition module, a surface morphology analysis module, a thermal distribution analysis module, a quality prediction fusion module, a process optimization generation module, a digital twin modeling module, and may also include a dynamic optimization execution module. The specific implementation of each module is described in detail below: The process parameter acquisition module is used to dynamically collect the machining trajectory offset, tool wear, and material hardness parameters of the robot polishing operation to obtain a set of process state parameters. Multiple sensors installed on the robot's motion path collect the machining trajectory offset in real time, calculate its Euclidean distance from the theoretical path, and mark it as the trajectory offset. Image recognition technology is used to extract the radial wear depth and end face wear area from the tool wear parameters, and the geometric mean of the two is calculated and marked as the tool wear characteristic value. A hardness testing instrument is used to collect the Rockwell hardness value and microhardness dispersion from the material hardness parameters, and a normalized weighted calculation is performed to obtain a comprehensive material hardness index. Based on the process state parameter set, the trajectory offset, tool wear characteristic value, and material hardness comprehensive index are compared with preset thresholds. When any parameter exceeds the corresponding threshold, a quality anomaly signal is generated. Based on the generated quality anomaly signal, a collaborative analysis instruction is triggered, and the surface morphology analysis module and the thermal distribution analysis module are executed.

[0029] The surface topography analysis module is used to extract surface profile characteristic parameters from the polished area of ​​the workpiece, quantitatively assess the surface texture uniformity, and generate a surface topography consistency index. Three-dimensional surface topography data is acquired using a white-light interferometer to generate a height distribution cloud map and a power spectrum density map. The arithmetic mean deviation and peak-to-valley height difference are extracted from the height distribution cloud map. The ratio of the two is calculated and taken logarithmically to obtain the texture uniformity coefficient. The mid-frequency energy concentration and spectral symmetry are extracted from the power spectrum density map, and the harmonic mean of the two is calculated and labeled as the spectral characteristic index. The texture uniformity coefficient and the spectral characteristic index are weighted and fused to obtain the surface topography consistency index.

[0030] The thermal distribution analysis module is used to extract workpiece surface temperature gradients and residual stress distribution parameters, analyze trends in the material's thermal-mechanical coupling effects, and derive a thermal stability assessment. Specifically, it collects workpiece surface temperature field data in real time, extracts the temperature gradient amplitude and thermal cycle characteristic parameters. A thermal-mechanical coupling prediction model is constructed, and the temperature gradient amplitude is input into the model for Gaussian process regression, outputting the probability of thermal deformation within a future time interval. The heating rate and cooling rate are extracted from the thermal cycle characteristic parameters, and the absolute difference between them is calculated to derive the thermal cycle stability factor. The thermal deformation probability and the thermal cycle stability factor are nonlinearly combined to obtain the thermal stability assessment value.

[0031] The quality prediction fusion module is used to receive the surface morphology consistency index and thermal stability evaluation value, make a collaborative prediction of the polishing quality performance, and generate a trajectory compensation signal and a cooling adjustment signal. The specific operation is to retrieve the tool wear characteristic value data, set its influence weight, and obtain the wear influence correction amount through calculation and processing; normalize the surface morphology consistency index, thermal stability evaluation value and wear influence correction amount to generate a quality performance fusion index; set the quality performance judgment threshold, if the fusion index is lower than the threshold, a trajectory compensation signal is generated, if it is higher than the threshold, a cooling adjustment signal is generated. In addition, this module also combines the model confidence evaluation value generated by the digital twin modeling module to dynamically correct the quality performance fusion index.

[0032] The process optimization generation module receives trajectory compensation signals and cooling adjustment signals, performs polishing strategy matching, and generates robot posture compensation parameters and cooling intensity adjustment parameters. If a trajectory compensation signal is captured, a posture adjustment command is triggered, dynamically configuring the robot's axial displacement and angular offset based on the command, generating the robot posture compensation parameters. If a cooling adjustment signal is captured, a thermal management command is triggered, optimizing the cooling medium's flow rate parameters and spray coverage based on the command, generating the cooling intensity adjustment parameters.

[0033] The digital twin modeling module monitors the mapping consistency between the physical polishing system and the virtual model, extracts data synchronization errors and state response delays, and generates a model confidence assessment. Specifically, the module collects physical system state data and virtual model output data by comparing timestamps, calculates the root mean square error (RMS) between the two, and labels it as the data synchronization error. It also extracts the physical and virtual response times at the moment a key event is triggered, calculates their absolute difference, and labels it as the state response delay. Finally, it performs a weighted sum of the data synchronization error and state response delay to generate a model confidence assessment.

[0034] The dynamic optimization execution module adjusts the polishing feed rate in real time based on the generated robot posture compensation parameters and cooling intensity adjustment parameters. It then feeds the results back to the process parameter acquisition module, forming a closed-loop control chain. Specifically, if the robot posture compensation parameters involve adjusting axial displacement, the force sensor sampling frequency is increased; if the cooling intensity adjustment parameters involve changing flow parameters, the spatial density of temperature monitoring points is increased. The adjusted parameters are then fed into the process parameter acquisition module, restarting the prediction process.

[0035] Example 1: This example describes in detail the determination and analysis process of the quality deviation risk in the process parameter acquisition module. When the robot is polishing, the processing trajectory offset is collected in real time with the help of a variety of sensors installed on the joints and processing areas of the robot. Specifically, high-precision encoders are used to record the actual motion position of each axis of the robot. These encoders can accurately capture the rotation angle or linear displacement of the robot joints and convert them into electrical signals. Then, these actual trajectory coordinates are compared one by one with the theoretical path coordinates pre-set in the control system. The theoretical path coordinates mentioned here are the ideal motion trajectory coordinates planned according to the polishing process requirements and the shape of the workpiece. During the comparison process, the Euclidean distance between the actual trajectory coordinates and the theoretical path coordinates is calculated. The distance is calculated by squaring the difference in each coordinate dimension, adding the square values ​​of all dimensions and then taking the square root to use this as the trajectory deviation. This trajectory deviation can intuitively reflect the degree of deviation between the actual processing trajectory of the robot and the theoretical trajectory.

[0036] To collect tool wear parameters, machine vision technology is used to monitor the polishing tool in real time. Specifically, a high-resolution industrial camera is installed near the polishing tool. This camera captures images of the tool at a constant frequency, capturing images of the tool surface. These images are then analyzed using a specialized image processing algorithm to extract two key parameters: radial wear depth and end face wear area. To extract radial wear depth, the algorithm identifies the wear boundary on the tool surface and calculates the maximum radial depth of the wear region. To extract end face wear area, the algorithm outlines the end face wear region and calculates the area enclosed by this outline. After extracting these two parameters, their geometric mean is calculated—multiplying the two parameters and taking the square root to obtain the tool wear characteristic value. This characteristic value comprehensively considers both radial and end face wear, providing a comprehensive reflection of the extent of tool wear.

[0037] To collect material hardness parameters, a Rockwell hardness tester is used to measure the material's Rockwell hardness value. During operation, the indenter of the Rockwell hardness tester is pressed into the material surface with a specified test force, held for a specified period, and then unloaded. The Rockwell hardness value is determined by measuring the indentation depth. Simultaneously, a microhardness tester is used to obtain the material's microhardness dispersion data. A microhardness tester applies a small test force to create a tiny indentation on the material surface. The hardness values ​​at different indentations are then measured to determine the microhardness distribution and, from there, the microhardness dispersion. After obtaining the Rockwell hardness value and microhardness dispersion, both data are normalized. Normalization maps the data to the same range, such as between 0 and 1, to eliminate the influence of different dimensions on the calculation. After normalization, the Rockwell hardness value and microhardness dispersion are weighted according to pre-defined weights based on material properties and polishing process requirements to produce a comprehensive material hardness index. This index comprehensively reflects the material's overall hardness level and the uniformity of its hardness distribution.

[0038] After completing the calculation of the three parameters of trajectory deviation, tool wear characteristic value and material hardness comprehensive index, they are compared with their corresponding preset thresholds. These preset thresholds are determined through a large number of process tests and historical data accumulation, and represent the reasonable range of each parameter under normal polishing operations. When any parameter exceeds its corresponding threshold, the system determines that there is a risk of quality deviation in the current polishing operation and immediately generates a quality anomaly signal. This quality anomaly signal will be transmitted to the system's control center, which will trigger a collaborative analysis instruction based on the signal, and then start the surface morphology analysis module and the thermal distribution analysis module to further analyze the surface morphology and thermal distribution of the workpiece to comprehensively evaluate the polishing quality.

[0039] Throughout the data collection and analysis process, each sensor and detection instrument operates continuously at the set sampling frequency to ensure real-time and continuous data. Simultaneously, the system filters the collected data to remove noise and ensure data accuracy. Furthermore, to accommodate different polishing workpieces and process requirements, preset thresholds and weighting parameters can be adjusted within the system's parameter settings interface to enhance system adaptability and flexibility. Through this series of operations, the process parameter acquisition module effectively identifies and analyzes the risk of quality deviation in robotic polishing operations, providing a reliable basis for subsequent quality prediction and process optimization.

[0040] For example, consider a model of industrial robot performing curved surface polishing on aircraft aluminum alloy components. The robot has six degrees of freedom, and its end effector is equipped with a polishing tool. Before polishing begins, the machining trajectory must be planned, generating theoretical path coordinates and storing them in the robot control system.

[0041] To implement the coordinate transformation algorithm, high-precision encoders are first installed at each joint of the robot to collect real-time angular data. The encoders used are incremental photoelectric encoders with a resolution of 10,000 pulses per revolution, which meets the requirements for high-precision trajectory acquisition. The robot's base coordinate system serves as the global coordinate system, and the coordinate systems of each joint are established using the Denavit-Hartenberg (D-H) parameter method. By establishing a D-H parameter table for each joint, the transformation relationship between adjacent joint coordinate systems is determined. Specifically, for the i-th joint, its D-H parameters include the joint angle θi, the link length ai, the link offset di, and the torsion angle αi. Using these parameters, a homogeneous transformation matrix Ti is constructed to describe the transformation from the i-1th coordinate system to the i-th coordinate system.

[0042] When the robot moves, each joint encoder outputs pulse signals in real time. The data acquisition system collects these signals at a sampling frequency of 1000 Hz and converts them into the rotation angle value θi of each joint. Then, based on the D-H parameters of each joint, the homogeneous transformation matrices T1, T2, ..., T6 of each joint coordinate system relative to the base coordinate system are calculated in sequence. The transformation matrix Tend of the end effector coordinate system relative to the base coordinate system can be obtained by multiplying the transformation matrices of each joint, that is: ; Through this transformation matrix, the coordinates of the end effector in each joint coordinate system can be converted into Cartesian coordinates (x, y, z) in the base coordinate system, thereby obtaining the actual trajectory coordinates of the robot end effector.

[0043] To generate theoretical path coordinates, the polishing surface of the aluminum alloy component is first modeled in 3D using computer-aided design (CAD) software. Based on the polishing process requirements, the polishing path is determined using a specific tool path, such as parallel or circular. Taking parallel tool paths as an example, a series of parallel machining paths are planned on the CAD model, with several theoretical points evenly distributed along each path. For each theoretical point, the CAD software retrieves its Cartesian coordinates (x0, y0, z0) in the base coordinate system. These coordinates are then stored in the robot control system's memory in the order in which they are processed, forming a sequence of theoretical path coordinates.

[0044] During the actual polishing operation, the data acquisition system acquires the actual trajectory coordinates (x, y, z) of the robot end effector in real time and compares them one by one with the pre-stored theoretical path coordinates. This is achieved through timestamp synchronization, ensuring that the actual and theoretical coordinates are compared at the same processing moment. When the robot reaches a theoretical point, the control system records the timestamp t0 at that moment and simultaneously acquires the actual coordinates (x, y, z) at that moment.

[0045] To calculate the Euclidean distance, consider two points in three-dimensional space with theoretical coordinates (x0, y0, z0) and actual coordinates (x, y, z). First, calculate the difference in each coordinate dimension: Δx = x - x0, Δy = y - y0, and Δz = z - z0. Then, square each difference to obtain Δx², Δy², and Δz². Add these squared values ​​to obtain Δx² + Δy² + Δz², and finally take the square root of this sum to obtain the Euclidean distance d between the two points: ; This distance is the trajectory deviation, which is used to represent the degree of deviation between the actual trajectory and the theoretical path.

[0046] In terms of software implementation, the trajectory offset calculation module is developed using the C++ programming language. This module contains the following key functions: Coordinate transformation function: Inputs the angle values ​​of each joint and outputs the Cartesian coordinates of the end effector. This function calculates the transformation matrix of each joint by calling a predefined D-H parameter table, and then obtains the end effector transformation matrix through matrix multiplication, thereby extracting the end effector coordinates.

[0047] Theoretical path acquisition function: Based on the current processing time, the corresponding theoretical point coordinates are read from the control system memory. This function uses linear interpolation to fit the trajectory between adjacent theoretical points to improve the comparison accuracy.

[0048] Distance calculation function: Input the actual coordinates and theoretical coordinates to calculate the Euclidean distance. This function uses the square root method mentioned above and controls the accuracy of the calculation result to four decimal places.

[0049] In terms of hardware implementation, the data acquisition system utilizes a high-performance industrial computer equipped with a multi-channel data acquisition card to receive encoder signals and send control commands. The computer's processor clock speed should be no less than 3.0 GHz, and its memory should be no less than 8 GB to ensure real-time data processing. The data acquisition card's sampling frequency is software-configurable; in this example, it is set to 1000 Hz, which meets the trajectory acquisition requirements of high-speed polishing operations.

[0050] To verify the accuracy of this implementation, multiple experiments were conducted. Taking a circular trajectory with a radius of 100mm as an example, the theoretical path was a standard circle, and the actual trajectory was obtained by the robot. Using the aforementioned method to calculate trajectory deviation, the results showed a maximum deviation of 0.08mm, meeting the precision requirements for polishing aviation aluminum alloy components (≤0.1mm). Computational time was also tested. At a sampling frequency of 1000Hz, the calculation time for each coordinate point was approximately 0.5ms, far less than the sampling period of 1ms, ensuring real-time performance.

[0051] In practical applications, this implementation has the following advantages: High precision: By adopting the D-H parameter method for coordinate transformation and combining it with a high-precision encoder, sub-millimeter track offset measurement can be achieved.

[0052] Real-time: Based on high-performance hardware and efficient algorithms, it can meet the real-time monitoring needs of high-speed polishing operations.

[0053] Flexibility: By modifying the theoretical path coordinates and adjusting the D-H parameters, it can adapt to the polishing operations of workpieces of different shapes and different types of robots.

[0054] Reliability: Through multiple experimental verifications and industrial field tests, this implementation method has high reliability and stability.

[0055] In practical applications, factors such as robot kinematic errors and encoder installation errors can lead to errors in trajectory offset calculation. To improve accuracy, the robot can be regularly calibrated to correct the D-H parameters and calibrate the encoder. Furthermore, algorithms such as Kalman filtering can be used to filter the collected trajectory data and reduce noise.

[0056] Example 2: This example describes in detail the process of quantitatively evaluating the surface texture uniformity in the surface morphology analysis module. After the workpiece polishing operation is completed or in progress, a white light interferometer is used to scan the polished area of ​​the workpiece. The white light interferometer emits a white light beam to the surface of the workpiece, and the light beam interferes with the reference beam after being reflected by the surface, thereby acquiring three-dimensional morphology data of the surface. Specifically, the optical system of the interferometer decomposes white light into light of different wavelengths. After these lights are irradiated on the surface of the workpiece, the optical path of the reflected light will be different due to the different surface heights, thereby forming interference fringes. By collecting and analyzing the interference fringes, the three-dimensional contour of the workpiece surface can be reconstructed, and a height distribution cloud map and a power spectrum density map can be generated.

[0057] In the height distribution cloud map, it is necessary to extract two key parameters: the profile arithmetic mean deviation and the peak-to-valley height difference. The extraction process of the profile arithmetic mean deviation (Ra) is to sample the surface profile along a specific cross-sectional direction in the selected analysis area to obtain a series of profile height values. Then, the arithmetic mean of the absolute deviations of these height values ​​relative to the profile centerline is calculated, which is the profile arithmetic mean deviation. Peak-to-valley height difference ( ) locates the highest and lowest points of the surface profile within the same analysis area and calculates the height difference between them. Once these two parameters are obtained, their ratio is calculated: the arithmetic mean deviation of the profile divided by the peak-to-valley height difference. The logarithm of this ratio is then taken to determine the texture uniformity coefficient. This calculation is based on the principle that when the surface texture is uniform, the ratio of the arithmetic mean deviation of the profile to the peak-to-valley height difference remains relatively stable. Taking the logarithm of this ratio makes it more sensitive to subtle changes in texture uniformity.

[0058] For the analysis of power spectrum density graph, the two main parameters are extracted: mid-frequency energy concentration and spectrum symmetry. Power spectrum density graph is a graph obtained by converting the spatial frequency characteristics of the surface profile. It reflects the contribution of different spatial wavelength components to the surface morphology. Mid-frequency band (usually set to The method for extracting the energy concentration of the spatial frequency range is to define the mid-frequency band in the power spectrum density diagram, calculate the sum of the energies of all frequency components in the range, and compare it with the total energy of the entire spectrum to obtain the energy concentration of the mid-frequency band. The spectrum symmetry is determined by analyzing the distribution of the power spectrum density diagram on the left and right sides of the mid-frequency band. Specifically, the symmetry index of the spectrum about the center frequency in the mid-frequency band can be calculated. For example, its symmetry can be characterized by calculating the difference or correlation coefficient between the energy distribution on the left and right sides. After obtaining the mid-frequency band energy concentration and spectrum symmetry, the harmonic mean of the two is calculated and marked as the spectral characteristic index. The harmonic mean is calculated by first taking the reciprocals of the two numbers, finding their average, and then taking the reciprocal. This calculation method can highlight the influence of smaller values ​​and is more suitable for characterizing the comprehensive characteristics of the two parameters.

[0059] The texture uniformity coefficient and the spectral characteristic index need to be weighted and fused to obtain the surface morphology consistency index. In the process of weighted fusion, it is first necessary to set appropriate weights for the texture uniformity coefficient and the spectral characteristic index according to the requirements of the polishing process and the material properties of the workpiece. For example, for polishing processes with higher requirements for surface roughness, a larger weight may be assigned to the texture uniformity coefficient; while for processes with higher requirements for surface waviness, more attention may be paid to the weight of the spectral characteristic index. The setting of weights is usually based on process experience and a large amount of test data to ensure that the fused index can accurately reflect the consistency of the surface morphology. After determining the weight, multiply the texture uniformity coefficient by its weight, multiply the spectral characteristic index by its corresponding weight, and then add the two products together to obtain the surface morphology consistency index.

[0060] Throughout the entire operation, the scanning parameters of the white light interferometer need to be properly set based on the material of the workpiece, the size of the polishing area, and the expected accuracy requirements. For example, parameters such as the scanning step size and resolution will directly affect the accuracy of the acquired three-dimensional topography data. For different workpiece materials, the interferometer's light source intensity and wavelength range may need to be adjusted to ensure that clear interference fringes can be obtained. In addition, when extracting parameters and performing calculations, it is necessary to pay attention to the processing range and accuracy of the data to avoid inaccurate results due to improper data capture or calculation errors.

[0061] To ensure the reliability of the evaluation results, the system also pre-processes the collected 3D topography data, including denoising and smoothing, to eliminate random errors and interference that may be introduced during the measurement process. Furthermore, for complex workpieces or polishing areas with special texture requirements, it may be necessary to scan and analyze each area separately. The evaluation results for each area are then combined to fully reflect the topographic consistency of the entire polished surface.

[0062] Through the above detailed operation steps and data processing process, the surface morphology analysis module can accurately quantify the surface texture uniformity of the workpiece polishing area. The obtained surface morphology consistency index can provide an important basis for subsequent polishing quality prediction, helping the system to promptly detect possible problems in the surface morphology and take corresponding measures to adjust and optimize, thereby ensuring that the quality of the polished workpiece meets the requirements.

[0063] After the robot polishing operation is completed or in progress, a white light interferometer is used to scan the polished area of ​​the workpiece to obtain three-dimensional topography data. The working principle of the white light interferometer is to emit a white light beam to the surface of the workpiece. After the light beam is reflected by the surface, it interferes with the reference beam, thereby obtaining three-dimensional topography data of the surface. Specifically, the optical system of the interferometer decomposes white light into light of different wavelengths. After these lights are irradiated on the surface of the workpiece, the optical path of the reflected light is different due to the different surface heights, thus forming interference fringes. By collecting and analyzing the interference fringes, the three-dimensional contour of the workpiece surface can be reconstructed, thereby realizing the acquisition of three-dimensional topography data.

[0064] After acquiring 3D topography data, a height distribution cloud map and a power spectrum density map are generated. When generating the height distribution cloud map, the surface height information is visualized in the form of a cloud map based on the reconstructed 3D contour of the workpiece surface. Different colors in the cloud map represent different height values, clearly demonstrating the distribution of surface heights. To generate the power spectrum density map, the spatial frequency characteristics of the surface contour are converted. This maps the contribution of different spatial wavelength components to the surface topography. The 3D topography data is processed using a corresponding algorithm to generate the power spectrum density map.

[0065] The two key parameters, the arithmetic mean deviation of the profile and the peak-to-valley height difference, are extracted from the height distribution cloud map. The extraction process of the arithmetic mean deviation of the profile (Ra) is as follows: within the selected analysis area, the surface profile is sampled along a specific cross-sectional direction to obtain a series of profile height values. Then, the arithmetic mean of the absolute deviations of these height values ​​relative to the profile centerline is calculated, which is the arithmetic mean deviation of the profile. Specifically, the profile centerline is first determined. The profile centerline is a reference line with an average height, so that within the sampling length, the area above the profile line is equal to the area below it. Next, for each profile height value obtained by sampling, the absolute value of the difference between it and the profile centerline is calculated. All these absolute values ​​are added and divided by the number of sampling points to obtain the arithmetic mean deviation of the profile.

[0066] To extract the peak-to-valley height difference (Rz), the highest and lowest points of the surface profile are found within the same analysis area, and the height difference between them is calculated. In practice, the surface profile data within the height distribution cloud map is first traversed to identify the highest and lowest height values ​​of all sampling points. The peak-to-valley height difference is then calculated by subtracting the lowest height value from the highest height value.

[0067] When extracting these two parameters, it's important to ensure the analysis area is representative, reflecting the typical surface characteristics of the workpiece's polished area. Furthermore, the sampling density and cross-sectional orientation also influence the accuracy of parameter extraction. Sampling parameters should generally be appropriately set based on the workpiece's material, polishing process requirements, and desired accuracy. For example, for workpieces with demanding surface finishes, a smaller analysis area and higher sampling density are often chosen to more accurately capture the surface's microscopic features.

[0068] To ensure data accuracy, the white light interferometer needs to be calibrated before acquiring 3D topography data to ensure the instrument's measurement accuracy. During the scanning process, the white light interferometer's light source intensity, wavelength range, scanning step size, resolution, and other parameters also need to be adjusted according to the workpiece's material and surface characteristics. For example, for highly reflective metal materials, the light source intensity may need to be adjusted to avoid overexposure; for workpieces with large surface undulations, the scanning step size and resolution need to be reasonably set to ensure that the surface contour information can be fully acquired.

[0069] After extracting the mean profile deviation and the peak-to-valley height difference, to obtain the texture uniformity coefficient, it is necessary to calculate the ratio of the two and take the logarithm. Specifically, the mean profile deviation is first divided by the peak-to-valley height difference to obtain a ratio, and then the natural logarithm of this ratio is taken to obtain the texture uniformity coefficient. The principle of this calculation process is that when the surface texture is uniform, the ratio of the mean profile deviation to the peak-to-valley height difference will be in a relatively stable range. After taking the logarithm, it can more sensitively reflect subtle changes in texture uniformity, allowing the texture uniformity coefficient to more accurately characterize the uniformity of the surface texture.

[0070] Throughout the entire operation, data processing and calculations require specialized software. This software processes the raw data collected by the white-light interferometer, automatically reconstructing 3D profiles, generating height distribution cloud maps and power spectrum density maps, and extracting and calculating the profile's arithmetic mean deviation and peak-to-valley height difference. The software also features data filtering and denoising capabilities, eliminating random errors and interference introduced during the measurement process and improving the accuracy of parameter extraction.

[0071] Complex workpieces or polishing areas with special texture requirements may require scanning and analysis in separate regions. The parameter extraction results for each region are then combined to fully reflect the characteristics of the entire polished surface. For example, for workpieces with curved structures, different regions have different curvatures and surface texture characteristics. Regional processing can more accurately obtain the arithmetic mean deviation of the profile and the peak-to-valley height difference for each region, thereby more comprehensively evaluating the surface topography.

[0072] Through the above specific implementation method, it is possible to accurately generate height distribution cloud maps and power spectrum density maps from the three-dimensional morphology data obtained by the white light interferometer, and extract two key parameters: the arithmetic mean deviation of the profile and the peak-to-valley height difference. This lays the foundation for the subsequent calculation of the texture uniformity coefficient and the surface morphology consistency index, and thus provides an important basis for the evaluation and prediction of polishing quality.

[0073] Example 3: This example describes in detail the trend deduction process of the material thermomechanical coupling effect in the thermal distribution analysis module. During the robot polishing operation, the temperature field data and residual stress distribution parameters of the workpiece surface are collected in real time by means of infrared temperature sensors and stress sensors arranged on the workpiece surface. The infrared temperature sensor can measure the temperature of the workpiece surface in a non-contact manner. Its working principle is to determine the temperature by detecting the infrared radiation energy emitted by the object. It has the characteristics of fast response speed and wide measurement range. The stress sensor senses the residual stress changes on the workpiece surface in real time by being attached to the workpiece surface or embedded inside the workpiece, converts the stress signal into an electrical signal and transmits it to the data acquisition system.

[0074] From the collected temperature field data, it is necessary to extract the temperature gradient amplitude and thermal cycle characteristic parameters. The temperature gradient amplitude refers to the maximum rate at which the temperature changes with position within a certain spatial range on the workpiece surface. It reflects the difference in temperature distribution in space. When extracting the temperature gradient amplitude, the temperature field data is first spatially gridded to divide the workpiece surface into multiple small grid units, each grid unit corresponding to a temperature value. Then, the temperature difference between adjacent grid units is calculated, and the maximum temperature difference in the entire temperature field is determined, which is the temperature gradient amplitude. Thermal cycle characteristic parameters include heating rate and cooling rate. The heating rate refers to the rate at which the workpiece surface temperature increases over time during the polishing process, and the cooling rate is the rate at which the temperature decreases over time. When extracting these two parameters, it is necessary to perform time series analysis on the temperature field data, select a specific temperature change stage, and calculate the rate of change of temperature over time in this stage to obtain the heating rate and cooling rate.

[0075] Next, a thermomechanical coupling prediction model is constructed to predict the trends of the thermomechanical coupling effects of the material. The model is based on the Gaussian process regression algorithm, a non-parametric statistical model that can make probabilistic predictions about unknown data given limited data. When building the model, it is first necessary to train historical data, including temperature gradient amplitudes and thermal deformation conditions under different polishing process parameters. The temperature gradient amplitude is used as the input variable of the model, and the trained model is used to predict the probability of thermal deformation within a future time interval. The future time interval here can be set according to the requirements and actual conditions of the polishing process, for example, set to the next 10 minutes or longer.

[0076] To calculate the thermal cycle stability factor, it is necessary to extract the heating rate and cooling rate from the thermal cycle characteristic parameters and calculate the absolute difference between them. This absolute difference is calculated by subtracting the smaller rate value from the larger rate value. The result is the thermal cycle stability factor. This factor reflects the degree of difference between the heating rate and cooling rate during the thermal cycle. The smaller the difference, the more stable the thermal cycle process; the larger the difference, the more unstable the thermal cycle process.

[0077] After obtaining the thermal deformation probability and thermal cycle stability factor, they need to be nonlinearly combined to obtain the thermal stability assessment value. This nonlinear combination can take the form of various functions, such as exponential functions and polynomial functions. Taking the exponential function as an example, the thermal deformation probability and thermal cycle stability factor are first normalized so that they are in the same numerical range. Then, the normalized thermal deformation probability is used as the base of the exponential function, and the thermal cycle stability factor is used as the exponent. The exponential operation is then performed, and the final thermal stability assessment value is obtained by adjusting the appropriate coefficients. This nonlinear combination method can fully consider the interaction between the thermal deformation probability and the thermal cycle stability factor, and more accurately reflect the stability trend of the material under the action of thermal-mechanical coupling.

[0078] Throughout the data acquisition and analysis process, the placement and number of sensors significantly impact the accuracy of the results. Temperature sensors must be strategically positioned based on the workpiece's shape and the characteristics of the polishing area to ensure comprehensive and accurate measurement of the surface temperature distribution. Stress sensors also require careful consideration of their placement and orientation to ensure effective sensing of residual stress changes. Furthermore, the data acquisition system's sampling frequency must be adjusted based on the speed of the polishing process and the rate of temperature change to ensure that transient changes in temperature and stress are captured.

[0079] To improve the model's predictive accuracy, the thermomechanical coupling prediction model requires regular updates and optimization. This can be achieved by collecting new polishing data, retraining the model, and adjusting its parameters to better adapt it to different polishing conditions. Furthermore, when conducting trend analysis, it's also necessary to consider the impact of other factors on the thermomechanical coupling effect, such as polishing tool wear and changes in the machining trajectory. These factors may indirectly affect the temperature and stress distribution on the workpiece surface and therefore need to be comprehensively considered during the analysis process.

[0080] Through the detailed operation steps and data processing process described above, the thermal distribution analysis module can accurately deduce the trend of the thermal-mechanical coupling effect of the material. The obtained thermal stability evaluation value can provide key information for polishing quality prediction, helping the system to predict in advance the quality problems that may arise in the material under the action of heat, and to adjust the polishing process parameters in time to ensure the quality of the polished workpiece.

[0081] Example 4: This example describes in detail the process of collaborative prediction of polishing quality performance in the quality prediction fusion module. Taking the robot polishing operation of a metal workpiece as an example, when the workpiece enters the polishing process, the system first retrieves the tool wear characteristic value data. Assuming that the polishing tool is a ceramic grinding wheel at this time, in the early polishing process, the radial wear depth of 0.3mm and the end face wear area of ​​5mm² have been obtained through machine vision technology, and the tool wear characteristic value is calculated to be the geometric mean of the two. Based on the material properties of the ceramic grinding wheel and the polishing process requirements of the workpiece, the system presets its influence weight to 0.25, and obtains the wear influence correction amount through weighted calculation, which is used for the subsequent comprehensive evaluation of quality performance.

[0082] Next, the system receives the surface topography consistency index (SCI) from the surface topography analysis module and the thermal stability evaluation value (TSE) from the thermal distribution analysis module. Assuming this polishing scenario, the SCI is calculated to be 0.72 after scanning the workpiece surface with a white light interferometer, extracting parameters such as the arithmetic mean deviation of the profile and the peak-to-valley height difference. The TSE is calculated to be 0.65 after analyzing the workpiece surface temperature field data and residual stress parameters, performing Gaussian process regression, and calculating the thermal cycling stability factor.

[0083] The system then normalizes the values ​​of the three parameters: the surface consistency index, the thermal stability assessment, and the wear effect correction. Normalization maps each parameter to a range of 0 to 1, eliminating the effects of dimensional differences. For example, the original value of 0.72 for the surface consistency index remains relatively proportional after normalization. The same applies to the thermal stability assessment of 0.65. The wear effect correction is calculated to a normalized value of 0.28.

[0084] Subsequently, the system performs a comprehensive calculation of the normalized parameters according to the preset fusion algorithm to generate a quality-performance fusion index. The weights of the parameters in the fusion algorithm are set according to the process requirements. It is assumed that the weight of the surface morphology consistency index is 0.4, the weight of the thermal stability assessment value is 0.35, and the weight of the wear effect correction is 0.25. By weighted summation, the three parameters are multiplied by the corresponding weights and then added together to obtain the quality-performance fusion index. In this example, the calculation process is: 0.72×0.4 + 0.65×0.35 + 0.28×0.25, and the final value of the fusion index is obtained.

[0085] The system has a preset quality performance threshold, determined based on the workpiece's polishing quality standards and historical process data (assuming it's set at 0.6). If the calculated quality performance fusion index falls below the threshold, the system determines that the polishing quality may not meet requirements and generates a trajectory compensation signal. If it exceeds the threshold, a cooling adjustment signal is generated. In this example, if the calculated fusion index is 0.58, which is below the preset threshold of 0.6, a trajectory compensation signal is generated, indicating that the robot's polishing trajectory needs to be adjusted.

[0086] In addition, the quality prediction fusion module also needs to receive the model confidence assessment value generated by the digital twin modeling module to dynamically correct the quality performance fusion index. The digital twin modeling module generates a model confidence assessment value by comparing the data synchronization error and state response delay between the physical polishing system and the virtual model. Assuming that in this example, the model confidence assessment value is 0.85, the system will correct the fusion index based on this assessment value. For example, through weighted adjustment, the fusion index can be corrected from 0.58 to 0.61. If the corrected value is higher than the threshold, the system may re-determine the generation of a cooling adjustment signal instead of a trajectory compensation signal.

[0087] Throughout the collaborative prediction process, real-time and accurate data are crucial. Tool wear characteristic values ​​must be continuously updated as the polishing process progresses. For example, after each polishing cycle, tool wear data is recollected and a new wear correction is calculated. Surface topography and thermal distribution data must also be dynamically collected during the polishing process to ensure that the parameters input into the quality prediction fusion module accurately reflect the current workpiece state.

[0088] The system's parameter settings are flexible and can be adjusted based on different workpiece materials, polishing tool types, and process requirements. For example, for hard metals, the weight of the thermal stability assessment may need to be increased; for workpieces with extremely high surface finish requirements, the weight of the surface topography consistency index needs to be increased. The weight of the wear effect correction can also be dynamically adjusted based on the tool's wear resistance and polishing time.

[0089] Once a trajectory compensation signal or cooling adjustment signal is generated, it is transmitted to the process optimization generation module to trigger subsequent polishing strategy matching operations. For example, the trajectory compensation signal prompts the system to dynamically configure the robot's axial displacement and angular offset to generate posture compensation parameters; the cooling adjustment signal guides the system to optimize the cooling medium flow parameters and spray coverage to generate cooling intensity adjustment parameters.

[0090] Through the operational process of the above specific examples, the quality prediction fusion module realizes the collaborative prediction of polishing quality performance. Combined with the confidence assessment of the digital twin model, it dynamically corrects the prediction results to provide accurate guidance signals for subsequent process optimization, ensuring that the polishing operation can continue to meet quality requirements, avoiding misjudgments caused by single parameter assessment, and improving the stability and reliability of the entire polishing system.

[0091] Example 5: This example uses the robotic polishing of an aluminum alloy workpiece as an example to describe in detail the polishing strategy matching process within the process optimization generation module and the specific execution process of the dynamic optimization execution module. Assuming that during the polishing process, the quality prediction fusion module generates a trajectory compensation signal or cooling adjustment signal based on a comprehensive calculation of the surface topography consistency index, thermal stability assessment value, and wear effect correction value. The following describes a specific implementation method.

[0092] When the process optimization generation module captures the trajectory compensation signal, it triggers the posture adjustment instruction. For example, in the curved surface polishing scenario of aluminum alloy workpieces, because the robot processing trajectory offset exceeds the preset threshold, the system determines that the trajectory needs to be compensated. At this time, the system dynamically configures the robot's axial displacement and angular offset according to the posture adjustment instruction. Specifically, through the parameter adjustment interface of the robot control system, the adjustment parameters of the axial displacement are input, such as increasing the displacement of 0.5mm in the X-axis direction to correct the insufficient polishing area caused by the trajectory offset; at the same time, the angular offset of the robot end effector is adjusted, such as increasing the rotation angle of the Y-axis by 2°, so that the polishing tool maintains a better contact angle with the workpiece surface. After the adjustment is completed, the robot posture compensation parameters are generated. The parameters contain detailed information such as the displacement adjustment value and angle offset value of each axis, which are used to guide the robot's motion control.

[0093] If a cooling adjustment signal is captured, taking the excessively high surface temperature of the workpiece during polishing as an example, the system triggers a thermal management instruction. At this point, it is necessary to optimize the flow parameters and spray coverage of the cooling medium. For example, the original cooling medium is a water-based cutting fluid, the flow rate is set to 5L / min, and the spray range covers 60% of the workpiece surface. Since the thermal stability evaluation value shows that the temperature gradient amplitude of the workpiece surface is large and the probability of thermal deformation is high, the system decides to increase the cooling flow rate to 8L / min, and expand the spray coverage to 85% by adjusting the installation position and angle of the nozzle. During specific operation, the opening of the flow control valve is adjusted to increase the flow rate through the control panel of the cooling system, and the position of the nozzle is adjusted manually or automatically to ensure that the cutting fluid can cover the high-temperature area more comprehensively, and generate cooling intensity adjustment parameters, which include information such as the adjusted flow value, nozzle position coordinates and spray angle.

[0094] The dynamic optimization execution module adjusts the feed speed of the polishing operation in real time based on the generated robot posture compensation parameters and cooling intensity adjustment parameters. For example, when the robot posture compensation parameters involve axial displacement adjustment, in order to ensure the accuracy of force feedback, the sampling frequency of the force sensor is simultaneously increased. Assuming that the original sampling frequency is 100Hz, it is increased to 200Hz after adjustment to collect contact force data between the polishing tool and the workpiece more frequently to avoid force control inaccuracy caused by displacement adjustment. The specific operation is to modify the sampling frequency parameters in the setting interface of the force sensor and restart the sensor to make it take effect.

[0095] If the cooling intensity adjustment parameters involve changes to flow parameters, such as increasing the cooling medium flow rate from 5L / min to 8L / min, the spatial density of temperature monitoring points is increased simultaneously. For example, in addition to the original five temperature monitoring points on the workpiece surface, one monitoring point is added every 15mm, bringing the total number of monitoring points to eight, to more comprehensively monitor temperature changes. During implementation, new temperature sensors are installed in appropriate locations based on the shape of the workpiece and the distribution of high-temperature areas. These sensors are then connected to the data acquisition system to ensure compatibility with the existing system.

[0096] The adjusted parameters are input into the process parameter acquisition module, and the prediction process is restarted. Taking axial displacement adjustment and flow rate change as an example, after adjustment, the robot performs polishing according to the new posture compensation parameters. The force sensor collects contact force data at a higher sampling frequency, and the temperature monitoring points collect temperature data at a greater spatial density. This data is transmitted to the process parameter acquisition module in real time. The process parameter acquisition module re-collects parameters such as machining trajectory offset, tool wear, and material hardness. It calculates the new trajectory offset, tool wear characteristic value, and material hardness composite index, and compares them with preset thresholds to determine whether there is a risk of quality deviation.

[0097] During the polishing process of an aluminum alloy workpiece, if the quality-performance fusion index generated after the initial adjustment still fails to reach the preset threshold, the system will trigger the polishing strategy matching and dynamic optimization execution process again. For example, if the thermal stability evaluation value after trajectory compensation still indicates temperature issues, the system may further adjust the cooling parameters, increasing the flow rate to 10L / min or adjusting the injection angle to optimize the cooling effect. After each adjustment, the dynamic optimization execution module will feed back the execution results to the process parameter acquisition module, forming a closed-loop control chain.

[0098] Throughout the entire process, the system's various modules communicate in real time via data interfaces. The posture compensation and cooling adjustment parameters generated by the process optimization generation module are transmitted to the robot control system and cooling system via industrial Ethernet. The dynamic optimization execution module sends instructions for adjusting the force sensor sampling frequency and temperature monitoring point density to the sensor control unit via the fieldbus. The real-time data collected by the process parameter acquisition module is transmitted to the central processing unit via a data acquisition card for analysis.

[0099] The extent and method of parameter adjustment vary for different workpiece types and polishing processes. For example, when polishing stainless steel workpieces, if a cooling adjustment signal is triggered, the cooling medium may need to be replaced with an emulsion, and the flow rate and injection pressure may need to be adjusted simultaneously. For trajectory compensation of complex curved workpieces, the robot's motion trajectory may need to be regenerated using offline programming software and imported into the control system.

[0100] The dynamic optimization execution module considers the interactions between parameters when adjusting them. For example, increasing the force sensor sampling frequency may increase the amount of data to be processed, and the system will automatically allocate more computing resources to the data processing unit. Increasing the density of temperature monitoring points may require expanding the number of channels on the data acquisition card, and the system will detect the hardware resources and configure them accordingly.

[0101] Through the implementation process of the above specific examples, the process optimization generation module and the dynamic optimization execution module achieve real-time adjustment and optimization of the polishing operation. Combined with the confidence assessment of the digital twin model, the prediction results and adjustment strategies are continuously revised to ensure that the polishing quality consistently meets the requirements. This closed-loop control chain can adapt to different processing conditions and workpiece characteristics, improving the flexibility and reliability of the robotic polishing system, avoiding quality fluctuations caused by single parameter adjustments, and realizing intelligent and precise control of the polishing process.

[0102] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0103] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An AI robot polishing quality prediction system based on digital twin drive, characterized by: include: A process parameter acquisition module is used to dynamically collect the machining trajectory offset, tool wear, and material hardness parameters of the robot polishing operation to obtain a process state parameter set. Based on the process state parameter set, the module determines and analyzes the quality deviation risk, generates a quality anomaly signal, triggers a collaborative analysis instruction based on the generated quality anomaly signal, and executes the surface morphology analysis module and the thermal distribution analysis module based on the triggered collaborative analysis instruction; Surface topography analysis module, used to extract surface profile characteristic parameters of the workpiece polishing area, quantitatively evaluate the surface texture uniformity, and obtain the surface topography consistency index; Thermal distribution analysis module, used to extract the surface temperature gradient and residual stress distribution parameters of the workpiece, deduce the trend of the material's thermal-mechanical coupling effect, and obtain the thermal stability evaluation value; The quality prediction fusion module is used to receive the surface topography consistency index and thermal stability evaluation value, make collaborative predictions on the polishing quality performance, and generate trajectory compensation signals and cooling adjustment signals; The process optimization generation module is used to receive the trajectory compensation signal and the cooling adjustment signal, perform polishing strategy matching, and generate the robot posture compensation parameters and cooling intensity adjustment parameters; The digital twin modeling module is used to monitor the mapping consistency between the physical polishing system and the virtual model, extract data synchronization errors and state response delays, and generate model confidence assessment values; The quality prediction fusion module further combines the model confidence evaluation value to dynamically correct the quality performance fusion index.

2. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 1 is characterized in that: The determination and analysis of quality deviation risk includes: By collecting the robot machining trajectory offset in real time, the Euclidean distance between it and the theoretical path is calculated and marked as the trajectory offset; Extract the radial wear depth and end face wear area from the tool wear parameters, calculate their geometric mean, and mark them as tool wear characteristic values; Collect the Rockwell hardness value and microhardness dispersion of the material hardness parameters, perform normalized weighted calculation, and obtain the comprehensive material hardness index; The trajectory deviation, tool wear characteristic value and material hardness comprehensive index are compared with the preset thresholds respectively. When any parameter exceeds the corresponding threshold, a quality abnormality signal is generated.

3. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 1 is characterized in that: The quantitative evaluation of surface texture uniformity includes: Acquire three-dimensional surface topography data through white light interferometry to generate height distribution cloud map and power spectrum density map; Extract the contour arithmetic mean deviation and peak-to-valley height difference from the height distribution cloud map, calculate the ratio of the two and take the logarithm to obtain the texture uniformity coefficient; Extract the mid-frequency energy concentration and spectrum symmetry from the power spectrum density diagram, calculate the harmonic mean of the two, and mark them as the spectrum characteristic index; The texture uniformity coefficient and the spectral characteristic index are weightedly fused to obtain the surface morphology consistency index.

4. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 1 is characterized in that: The trend deduction of the thermal-mechanical coupling effect of the material includes: Collect workpiece surface temperature field data in real time, extract temperature gradient amplitude and thermal cycle characteristic parameters; Construct a thermal-mechanical coupling prediction model, input the temperature gradient amplitude into the model for Gaussian process regression processing, and output the probability of thermal deformation in the future time interval; Extract the heating rate and cooling rate from the thermal cycle characteristic parameters, and calculate the absolute difference between the two to obtain the thermal cycle stability factor; The thermal deformation probability and the thermal cycle stability factor are nonlinearly combined to obtain the thermal stability evaluation value.

5. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 1 is characterized in that: The collaborative prediction of polishing quality performance includes: Retrieve tool wear characteristic value data, set its influence weight, and obtain the wear influence correction value through calculation and processing; The surface morphology consistency index, thermal stability evaluation value and wear effect correction value are normalized and calculated to generate the quality performance fusion index; A quality performance judgment threshold is set. If the fusion index is lower than the threshold, a trajectory compensation signal is generated; if it is higher than the threshold, a cooling adjustment signal is generated.

6. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 1 is characterized in that: The polishing strategy matching includes: If the trajectory compensation signal is captured, the posture adjustment command is triggered, and the robot's axial displacement and angle offset are dynamically configured according to the command to generate the robot's posture compensation parameters; If a cooling adjustment signal is captured, the thermal management instruction is triggered, and the flow parameters and injection coverage range of the cooling medium are optimized according to the instruction to generate the cooling intensity adjustment parameters.

7. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 1 is characterized in that: The monitoring of the mapping consistency between the physical polishing system and the virtual model includes: The physical system state data and the virtual model output data are collected by comparing the timestamps, and the root mean square error between the two is calculated and marked as the data synchronization error; Extract the physical response time and virtual response time at the time the key event is triggered, calculate the absolute difference between the two, and mark it as the state response delay; The data synchronization error and the state response delay are weighted and summed to generate a model confidence evaluation value.

8. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 1 is characterized in that: Also includes: The dynamic optimization execution module is used to adjust the feed speed of the polishing operation in real time according to the generated robot posture compensation parameters and cooling intensity adjustment parameters, and feed back the execution results to the process parameter acquisition module to form a closed-loop control link.

9. The AI ​​robot polishing quality prediction system based on digital twin drive according to claim 8, characterized in that: The specific execution process of the dynamic optimization execution module includes: If the robot posture compensation parameters involve axial displacement adjustment, the sampling frequency of the force sensor should be increased simultaneously; If the cooling intensity adjustment parameters involve changes in flow parameters, the spatial density of temperature monitoring points will be increased simultaneously; Input the adjusted parameters into the process parameter acquisition module and restart the prediction process.

Images