Self-adaptive optical polishing system and method based on dynamic temperature control and multi-modal detection

Abstract

The invention discloses a self-adaptive optical polishing system and method based on dynamic temperature control and multi-modal detection, and relates to the technical field of intelligent manufacturing and precision machining.The optical polishing system comprises a dynamic temperature control and multi-modal detection module, a flexible execution mechanism, a closed-loop control and optimization module and a cross-scale collaboration and generalization module; the method has the advantages that the temperature field, surface appearance and microcrack signals of a polishing area are monitored in real time through the dynamic temperature control and multi-mode detection technology, thermal deformation is rapidly compensated in combination with a flexible execution mechanism, the problem that the surface shape precision is reduced due to temperature rise in a traditional technology is effectively solved, and the polishing quality is improved. Multi-dimensional defects in the polishing process can be comprehensively captured through cooperative work of the multi-mode sensor, the defect detection sensitivity and the process stability are remarkably improved, and the machining quality of the high-precision optical element is ensured.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing and precision machining technology, and specifically to an adaptive optical polishing system and method based on dynamic temperature control and multimodal detection. Background Art

[0002] With the widespread application of high-precision optical components in semiconductors, aerospace and other fields, traditional optical polishing technology faces many bottlenecks: First, the local temperature rise during the polishing process can easily cause thermal deformation of the material, resulting in a decrease in surface accuracy; Second, single-mode detection is difficult to capture multi-dimensional defects such as temperature fields, surface roughness, and microcracks in real time, and process parameter adjustments are delayed; Third, existing actuators lack flexibility, making it difficult to adapt to complex curved surfaces and heterogeneous materials. They rely on manual experience to adjust parameters, resulting in low efficiency. To address the above problems, existing technologies attempt to alleviate thermal deformation through fixed temperature control strategies or offline compensation methods, but the dynamic response is insufficient and it is impossible to coordinate multi-sensor data for global optimization. At the same time, the rule-based control model has weak generalization ability and is difficult to adapt to new materials or process change requirements. The present invention proposes an adaptive polishing system that integrates dynamic temperature control, multimodal detection and intelligent optimization. It realizes real-time multi-physical field monitoring by integrating infrared thermal imagers, laser interferometers and acoustic emission sensors, dynamically compensates for thermal deformation by combining shape memory alloy polishing heads and microfluidic cooling systems, and adopts a hierarchical closed-loop control architecture (PID fast loop + LSTM + PPO reinforcement learning slow loop) to achieve self-optimization of process parameters. It adapts to cross-material process migration through a transfer learning framework, significantly improving polishing accuracy, efficiency and process generalization ability. Summary of the Invention

[0003] The object of the present invention is to provide an adaptive optical polishing system and method based on dynamic temperature control and multimodal detection.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an adaptive optical polishing system based on dynamic temperature control and multimodal detection, the optical polishing system comprising: Dynamic temperature control and multimodal detection module, integrating infrared thermal imager, laser interferometer and acoustic emission sensor, is used to collect temperature field, surface morphology and microcrack signals of the polishing area in real time; A flexible actuator comprising a shape memory alloy polishing head and a multi-channel microfluidic cooling system, wherein: The polishing head is driven by current to achieve local deformation compensation, and the deformation amount satisfies the relationship: ; Where, is the deformation coefficient, is the driving current.

[0005] The cooling system regulates the coolant flow and temperature through gradient, and the flow distribution satisfies the exponential decay formula: ; Where, is the radius from the center of the high temperature zone.

[0006] The closed-loop control and optimization module adopts a hierarchical control architecture, including a fast loop layer (FPGA) to adjust temperature and pressure parameters in real time, and a slow loop layer (GPU acceleration) to run the LSTM+PPO reinforcement learning model for global process optimization; The cross-scale collaboration and generalization module is based on the process parameter mapping table and transfer learning framework to adapt to different materials and surface types.

[0007] As a further solution of the present invention: the multimodal sensor meets the following conditions: The infrared thermal imager has a spatial resolution of ≤0.1°C, a sampling frequency of 100Hz, and can locate local hot spots with a temperature difference of >5°C and a mesh diameter of >2mm. The laser interferometer has a wavelength of 632.8nm and generates a 3D height map with a resolution of 10nm every 5 seconds to identify the surface error PV value and roughness Ra; The acoustic emission sensor has a frequency band of 50-500 kHz, and the energy surge signal in the 200 kHz frequency band is extracted by short-time Fourier transform (STFT).

[0008] As a further solution of the present invention: the temperature-pressure coupling model includes: Physical driving layer: Establish thermal expansion equation based on finite element simulation: ; Where, is the coefficient of thermal expansion, is the temperature rise, is the characteristic size; Data-driven layer: uses LSTM network to input real-time temperature and pressure sequences to predict thermal deformation trends within the next 3 seconds; Instruction generation layer: Integrates physical and data prediction results to output high-temperature zone coolant flow instructions (120ml / min to 200ml / min) and temperature setting values .

[0009] As a further solution of the present invention: the shape memory alloy polishing head is a nickel titanium alloy (NiTiNOL) grid structure, the grid nodes are embedded with micro-resistance heating plates, and the curvature of the workpiece surface is mapped to the node target displacement through the inverse kinematics algorithm.

[0010] As a further embodiment of the present invention, the multi-channel microfluidic cooling system includes 16 independently controlled piezoelectric micropumps and semiconductor refrigeration chips, wherein: The coolant flow rate in the high temperature zone is 200 ml / min and the temperature is set at 5°C; The flow in the transition zone is distributed according to the exponential decay formula, and the temperature following strategy is .

[0011] As a further solution of the present invention: the layered architecture of the closed-loop control module includes: Fast cycle layer (execution cycle 1ms): PID algorithm is used to control the coolant flow rate and PWM is used to modulate the polishing head current; Slow cycle layer (execution period 1s): input the temperature, pressure and defect data sequence of the past 60 seconds, and output the pressure distribution, speed and cooling gradient optimization parameters.

[0012] As a further solution of the present invention: the process parameter mapping table construction rules include: Material Property Dimension: Coefficient of Thermal Expansion , Young's modulus ,hardness ; Process parameter dimensions: coolant gradient mode, deformation compensation coefficient, allowable temperature rise ; The corresponding temperature control strategy and pressure threshold are loaded in real time according to the workpiece material type (such as fused quartz, calcium fluoride).

[0013] As a further solution of the present invention: the transfer learning framework includes: Pre-trained ResNet-34 backbone network to extract multimodal data features; The adaptation layer is a 256-node fully connected layer, fine-tuned based on small sample data (such as 10 sets of free-form surface processing data), and the loss function is: ; Where, is the surface roughness, For processing efficiency.

[0014] The present invention also provides an adaptive optical polishing method based on dynamic temperature control and multimodal detection, the optical polishing method comprising the following steps: Data acquisition and preprocessing: Synchronously trigger the infrared thermal imager (100Hz), laser interferometer (10Hz), pressure sensor (1kHz), and acoustic emission sensor (500kHz). Align the timing using the dynamic time warping (DTW) algorithm, perform wavelet threshold denoising on the acoustic emission signal, and use Kalman filtering on the temperature data. Thermal-mechanical coupling analysis and command generation: Based on finite element simulation and LSTM network, thermal deformation trend is predicted and the polishing head deformation compensation current is output. and coolant gradient parameters ; Real-time defect detection and parameter correction: When the surface roughness Ra is greater than 0.5nm or the acoustic emission energy exceeds three standard deviations of the baseline, the PPO reinforcement learning model is triggered to correct the pressure (±10%), speed (±20%) and cooling strategy within 50ms; Cross-cycle iterative optimization: Store historical process data, train the LSTM model to output the optimal parameter combination, and adapt to new workpiece types through transfer learning.

[0015] As a further solution of the present invention: the data preprocessing includes: Temperature data is normalized to the range of 0-1, and pressure is normalized to the maximum range; Laser interferometer data is interpolated and aligned to a high-frequency temperature data stream.

[0016] By adopting the above technical solution, compared with the prior art, the beneficial effects of the present invention are: 1. This invention uses dynamic temperature control and multimodal detection technology to monitor the temperature field, surface morphology, and microcrack signals in the polishing area in real time. Combined with a flexible actuator to quickly compensate for thermal deformation, it effectively avoids the problem of surface accuracy degradation caused by temperature rise in traditional processes. The collaborative work of multimodal sensors can fully capture multi-dimensional defects in the polishing process, significantly improving defect detection sensitivity and process stability, and ensuring the processing quality of high-precision optical components. 2. The system of the present invention adopts a layered closed-loop intelligent control architecture. The fast cycle layer controls the temperature and pressure parameters in real time. The slow cycle layer optimizes the process strategy globally based on the reinforcement learning model to achieve adaptive adjustment and dynamic balance of process parameters. Compared with the traditional fixed rule control model, this architecture can autonomously learn and optimize complex nonlinear process relationships, reduce dependence on manual experience, improve polishing efficiency and adaptability to different materials and surface types. 3. The cross-material process adaptation capability of the present invention is based on the transfer learning framework, which enables the system to quickly migrate historical process data to new workpiece types through pre-trained models, and realize automatic loading of cross-material temperature control strategies and pressure thresholds in combination with process parameter mapping tables. This technology breaks through the limitations of traditional processes such as long parameter adjustment cycles and weak generalization capabilities for new materials, significantly shortens process development time, and takes into account processing accuracy and efficiency at the same time, meeting the efficient production needs of multi-variety and small-batch optical components. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Flowchart of the adaptive optics polishing system architecture; Figure 2Flow chart of the steps of the optical polishing method. DETAILED DESCRIPTION

[0018] The specific embodiments of the present invention will be further described below in conjunction with the accompanying drawings. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0019] In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0020] Please see the attached Figure 1 The present invention is an adaptive optical polishing system based on dynamic temperature control and multimodal detection. The optical polishing system includes: Dynamic temperature control and multimodal detection module, integrating infrared thermal imager, laser interferometer and acoustic emission sensor, is used to collect temperature field, surface morphology and microcrack signals of the polishing area in real time; A flexible actuator comprising a shape memory alloy polishing head and a multi-channel microfluidic cooling system, wherein: The polishing head is driven by current to achieve local deformation compensation, and the deformation satisfies the relationship: ; Where, is the deformation coefficient, is the driving current.

[0021] The cooling system regulates the coolant flow and temperature through gradients, and the flow distribution satisfies the exponential decay formula: ; Where, is the radius from the center of the high temperature zone.

[0022] The closed-loop control and optimization module adopts a hierarchical control architecture, including a fast loop layer (FPGA) to adjust temperature and pressure parameters in real time, and a slow loop layer (GPU acceleration) to run the LSTM+PPO reinforcement learning model for global process optimization; The cross-scale collaboration and generalization module is based on the process parameter mapping table and transfer learning framework to adapt to different materials and surface types.

[0023] In one embodiment of the present invention, the multimodal sensor satisfies the following conditions: The infrared thermal imager has a spatial resolution of ≤0.1°C, a sampling frequency of 100Hz, and can locate local hot spots with a temperature difference of >5°C and a mesh diameter of >2mm. The laser interferometer has a wavelength of 632.8nm and generates a 3D height map with a resolution of 10nm every 5 seconds to identify the surface error PV value and roughness Ra; The acoustic emission sensor has a frequency band of 50-500 kHz, and the energy surge signal in the 200 kHz frequency band is extracted by short-time Fourier transform (STFT).

[0024] In one embodiment of the present invention: the temperature-pressure coupling model includes: Physical driving layer: Establish thermal expansion equation based on finite element simulation: ; Where, is the coefficient of thermal expansion, is the temperature rise, is the characteristic size; Data-driven layer: uses LSTM network to input real-time temperature and pressure sequences to predict thermal deformation trends within the next 3 seconds; Instruction generation layer: Integrates physical and data prediction results to output high-temperature zone coolant flow instructions (120ml / min to 200ml / min) and temperature setting values .

[0025] In one embodiment of the present invention, the shape memory alloy polishing head is a nickel-titanium alloy (NiTiNOL) grid structure, the grid nodes are embedded with micro-resistance heating plates, and the curvature of the workpiece surface is mapped to the target displacement of the node through the inverse kinematics algorithm.

[0026] In one embodiment of the present invention, a multi-channel microfluidic cooling system includes 16 independently controlled piezoelectric micropumps and semiconductor refrigeration chips, wherein: The coolant flow rate in the high temperature zone is 200 ml / min and the temperature is set at 5°C; The flow in the transition zone is distributed according to the exponential decay formula, and the temperature following strategy is .

[0027] In one embodiment of the present invention, the layered architecture of the closed-loop control module includes: Fast cycle layer (execution cycle 1ms): PID algorithm is used to control the coolant flow rate and PWM is used to modulate the polishing head current; Slow cycle layer (execution period 1s): input the temperature, pressure and defect data sequence of the past 60 seconds, and output the pressure distribution, speed and cooling gradient optimization parameters.

[0028] In one embodiment of the present invention, the process parameter mapping table construction rules include: Material Property Dimension: Coefficient of Thermal Expansion , Young's modulus ,hardness ; Process parameter dimensions: coolant gradient mode, deformation compensation coefficient, allowable temperature rise ; The corresponding temperature control strategy and pressure threshold are loaded in real time according to the workpiece material type (such as fused quartz, calcium fluoride).

[0029] In one embodiment of the present invention: the transfer learning framework includes: Pre-trained ResNet-34 backbone network to extract multimodal data features; The adaptation layer is a 256-node fully connected layer, fine-tuned based on small sample data (such as 10 sets of free-form surface processing data), and the loss function is: ; Where, is the surface roughness, For processing efficiency.

[0030] Please see the attached Figure 2 The present invention also provides an adaptive optical polishing method based on dynamic temperature control and multimodal detection, the optical polishing method comprising the following steps: Data acquisition and preprocessing: Synchronously trigger the infrared thermal imager (100Hz), laser interferometer (10Hz), pressure sensor (1kHz), and acoustic emission sensor (500kHz). Align the timing using the dynamic time warping (DTW) algorithm, perform wavelet threshold denoising on the acoustic emission signal, and use Kalman filtering on the temperature data. Thermal-mechanical coupling analysis and command generation: Based on finite element simulation and LSTM network, thermal deformation trend is predicted and the polishing head deformation compensation current is output. and coolant gradient parameters ; Real-time defect detection and parameter correction: When the surface roughness Ra is greater than 0.5nm or the acoustic emission energy exceeds three standard deviations of the baseline, the PPO reinforcement learning model is triggered to correct the pressure (±10%), speed (±20%) and cooling strategy within 50ms; Cross-cycle iterative optimization: Store historical process data, train the LSTM model to output the optimal parameter combination, and adapt to new workpiece types through transfer learning.

[0031] In one embodiment of the present invention, data preprocessing includes: Temperature data is normalized to the range of 0-1, and pressure is normalized to the maximum range; Laser interferometer data is interpolated and aligned to a high-frequency temperature data stream.

[0032] Example 1 Technical Background: In the high-precision polishing of fused silica aspheric lenses, the traditional constant-pressure polishing process is difficult to solve the problem of material thermal deformation caused by local temperature rise. The present invention uses dynamic temperature control and multimodal detection technology to achieve precise control of the polishing process.

[0033] Implementation steps: 1. Multimodal data acquisition and synchronization: The infrared thermal imager monitors the temperature field of the polishing area in real time at a sampling frequency of 100Hz, with a spatial resolution of 0.08℃. It detects a local hot spot area with a diameter of 3.2mm, where the temperature at the center is higher than the surrounding area. .

[0034] The laser interferometer scans the entire surface every 5 seconds to generate a three-dimensional height map (resolution 10nm). The measured surface error PV value is / 25( =632.8nm), surface roughness Ra=0.32nm.

[0035] The acoustic emission sensor captured high-frequency stress wave signals at a sampling rate of 500kHz. Through short-time Fourier transform (STFT) analysis, it was found that the energy in the 200kHz frequency band suddenly increased to 3.2 times the baseline value in the 12th minute of processing, triggering a microcrack warning.

[0036] Data synchronization uses the dynamic time warping (DTW) algorithm to interpolate the laser interferometer low-frequency data (10Hz) to the temperature data stream (100Hz) to eliminate timing deviations.

[0037] 2. Thermal-mechanical coupling modeling and instruction generation: The physical driver layer calculates the thermal expansion of fused silica based on finite element simulation: ; Combined with the LSTM network, the thermal deformation trend within the next 3 seconds is predicted and the deformation compensation amount of the polishing head is output. m corresponds to the driving current =3.5A.

[0038] Coolant gradient control command is generated: the coolant flow rate in the high temperature zone (radius r≤5mm) is increased to 200ml / min, the temperature is set to T_hotspot-15℃=20℃, the flow rate in the transition zone is distributed according to the exponential decay formula, and the flow rate at the radius r=15mm is reduced to =200 =110ml / min.

[0039] 3. Closed-loop control and real-time optimization: The fast loop layer (FPGA) executes PID control in 1ms cycle to control the temperature fluctuation in the high temperature area. within 0.5℃; When the acoustic emission energy exceeds three standard deviations of the baseline, the slow loop layer (GPU) calls the PPO reinforcement learning model to reduce the polishing head speed from 1200 rpm to 1020 rpm (a 15% reduction) within 50 ms, and prioritizes allocating 80% of cooling resources to the high-temperature area.

[0040] Technical effects: PV value of the surface accuracy of the fused silica lens after processing≤ / 20( =632.8nm), the surface roughness Ra is stable at 0.30-0.35nm, which is better than the traditional process Ra=0.5nm; The micro-crack defect rate was reduced from 1.2% in the traditional process to 0.08%, processing efficiency was increased by 22%, and the working time per piece was shortened to 45 minutes.

[0041] Example 2 Technical background: For calcium fluoride free-form surface lens (curvature radius R = 150mm, aspheric coefficient )’s low thermal conductivity requires rapid process adaptation across materials and complex surfaces.

[0042] Implementation steps: 1. Cross-scale parameter mapping and transfer learning: Call the process parameter mapping table and load the calcium fluoride material properties: thermal expansion coefficient , Young's modulus =80GPa, hardness =158.

[0043] According to the rule base, the automatic matching of coolant flow is increased by 20%, the high temperature zone flow is adjusted from 200m / / min to 240ml / min, and the maximum allowable temperature rise is Set to 5°C.

[0044] The transfer learning framework loads a pre-trained fused silica processing model (ResNet-34 backbone network), adds a 256-node fully connected layer, inputs 10 sets of calcium fluoride free-form surface processing data (including temperature field, pressure distribution, and defect records), freezes the backbone network weights, and only fine-tunes the adaptation layer loss function: ; After 2 hours of training, the model outputted the optimal pressure distribution, which was 12% pressure increase in the edge area and 8% pressure reduction in the center area.

[0045] 2. Dynamic deformation compensation and surface adaptation: The inverse kinematics algorithm converts curvature Mapped to the target displacement of the polishing head mesh node, the calculated The driving current of the corresponding node is 1=2.8A, which triggers the austenite phase transformation of the nickel-titanium alloy and realizes local deformation compensation.

[0046] The multi-channel microfluidic system distributes the nozzles in concentric circles. The coolant temperature in the high temperature zone (curvature mutation zone) is set to 5°C and the flow rate is 240ml / min; the flow rate in the transition zone is set to Distribution, the flow rate at radius r=40mm drops to .

[0047] 3. Defect suppression and process verification: The acoustic emission sensor detected energy fluctuations in the 250kHz frequency band and identified it as subsurface damage, triggering the PPO model to dynamically adjust the polishing head tilt angle by 2° to reduce local stress concentration. The surface roughness Ra fed back by the laser interferometer is 0.45 nm, which is close to the processing level of fused quartz, verifying the effectiveness of transfer learning.

[0048] Technical effects: Free-form surface adaptation error ≤ 1.8μm, surface consistency deviation < ; The training cycle of the transfer learning model has been shortened from the traditional 72 hours to 2 hours, the generalization ability has been improved by 35%, and the compatible material types have been expanded to more than 5 types.

[0049] Example 3: Technical Background: Applied to the continuous processing of glass substrates with a curvature radius of R=50mm, it is necessary to solve the problem of surface consistency degradation caused by dynamic temperature rise and stress mutation.

[0050] Implementation steps: 1. Hierarchical closed-loop control and real-time response: The fast loop layer (FPGA) executes temperature control instructions in a 1ms cycle: the 16-channel piezoelectric micropump is adjusted by the PID algorithm, the cooling wave flow response delay in the high-temperature zone is ≤0.2ms, and the temperature fluctuation is controlled within +0.3°C; The slow loop layer (GPU) runs the LSTM+PPO model every 1 second, inputting the temperature, pressure, and defect sequence of the past 60 seconds (a total of 60,000 data points) and outputting the global optimization parameters: the polishing head pressure distribution is adjusted by +8%, and the rotation speed is reduced from 1500 rpm to 1350 rpm (a 10% decrease).

[0051] 2. Multimodal defect detection and suppression: The acoustic emission sensor captured a 300kHz high-frequency stress wave signal with an energy peak four times the baseline. The system identified this as subsurface damage expansion and immediately triggered the following actions: Increase the coolant flow rate in the high-temperature zone to 200 ml / min and reduce the temperature to 5°C; Polishing head deformation compensation increased to , compensate for thermal deformation error.

[0052] The laser interferometer monitors the surface roughness Ra value in real time. When Ra suddenly increases from 0.4nm to 0.62nm, the reinforcement learning model generates parameter correction instructions within 50ms, reducing the polishing pressure from 8N ​​to 7.2N (a 10% decrease).

[0053] 3. Gradient cooling and process stability assurance: The semiconductor refrigeration chip sets the coolant temperature in the transition zone to 10℃, and the flow rate in the high temperature zone (radius r≤10mm) is set to Distribution, the flow rate at radius r=30mm is ; After 8 hours of continuous processing, the system automatically starts cross-cycle optimization, calls the historical process database (storing 120 sets of parameter combinations), trains the LSTM model to predict the thermal deformation trend in the next 15 minutes, and adjusts the cooling strategy in advance.

[0054] Technical effects: The surface consistency deviation during continuous processing is ≤5nm, and there is no batch-specific deviation; The dynamic response time is ≤50ms, the process stability is improved by 42% compared with traditional methods, the daily production capacity is increased to 300 pieces, and the yield is stable at above 99.3%.

[0055] Although the present invention is disclosed above with reference to preferred embodiments, this is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent variations, and modifications made to the above embodiments in accordance with the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the scope of protection defined by the claims of the present invention.

Claims

1. Adaptive optical polishing system based on dynamic temperature control and multimodal detection, characterized by: The optical polishing system comprises: Dynamic temperature control and multimodal detection module, integrating infrared thermal imager, laser interferometer and acoustic emission sensor, is used to collect temperature field, surface morphology and microcrack signals of the polishing area in real time; A flexible actuator comprising a shape memory alloy polishing head and a multi-channel microfluidic cooling system, wherein: The polishing head is driven by current to achieve local deformation compensation, and the deformation amount satisfies the relationship: ； Where, is the deformation coefficient, is the driving current; The cooling system regulates the coolant flow and temperature through gradient, and the flow distribution satisfies the exponential decay formula: ； Where, is the radius from the center of the high temperature zone; The closed-loop control and optimization module adopts a hierarchical control architecture, including a fast-loop layer that adjusts temperature and pressure parameters in real time, and a slow-loop layer that runs an LSTM+PPO reinforcement learning model for global process optimization. The cross-scale collaboration and generalization module is based on the process parameter mapping table and transfer learning framework to adapt to different materials and surface types.

2. The adaptive optical polishing system based on dynamic temperature control and multimodal detection according to claim 1, characterized in that: The multimodal sensor meets the following conditions: The infrared thermal imager has a spatial resolution of ≤0.1°C, a sampling frequency of 100Hz, and can locate local hot spots with a temperature difference of >5°C and a mesh diameter of >2mm. The laser interferometer has a wavelength of 632.8nm and generates a 3D height map with a resolution of 10nm every 5 seconds to identify the surface error PV value and roughness Ra; The acoustic emission sensor has a frequency band of 50-500KHz, and the energy surge signal in the 200kHz frequency band is extracted by short-time Fourier transform.

3. The adaptive optical polishing system based on dynamic temperature control and multimodal detection according to claim 1, characterized in that: The temperature-pressure coupling model includes: Physical driving layer: Establishing thermal expansion equation based on finite element simulation: ； Where, is the coefficient of thermal expansion, is the temperature rise, is the characteristic size; Data-driven layer: uses LSTM network to input real-time temperature and pressure sequences to predict thermal deformation trends within the next 3 seconds; Instruction generation layer: integrates physical and data prediction results to output coolant flow instructions and temperature setting values ​​in high-temperature areas .

4. The adaptive optical polishing system based on dynamic temperature control and multimodal detection according to claim 1, characterized in that: The shape memory alloy polishing head is a nickel-titanium alloy grid structure, with grid nodes embedded in micro-resistance heating sheets. The curvature of the workpiece surface is mapped to the node target displacement through the inverse kinematics algorithm.

5. The adaptive optical polishing system based on dynamic temperature control and multimodal detection according to claim 1, characterized in that: The multi-channel microfluidic cooling system includes 16 independently controlled piezoelectric micropumps and semiconductor refrigeration chips, wherein: The coolant flow rate in the high temperature zone is 200 ml / min and the temperature is set at 5°C; The flow in the transition zone is distributed according to the exponential decay formula, and the temperature following strategy is .

6. The adaptive optical polishing system based on dynamic temperature control and multimodal detection according to claim 1, characterized in that: The layered architecture of the closed-loop control module includes: Fast cycle layer: PID algorithm is used to control the coolant flow rate and PWM is used to modulate the polishing head current; Slow cycle layer: Input the temperature, pressure and defect data sequence of the past 60 seconds, and output the pressure distribution, speed and cooling gradient optimization parameters.

7. The adaptive optical polishing system based on dynamic temperature control and multimodal detection according to claim 1, characterized in that: The process parameter mapping table construction rules include: Material Property Dimension: Coefficient of Thermal Expansion , Young's modulus ,hardness ; Process parameter dimensions: coolant gradient mode, deformation compensation coefficient, allowable temperature rise ; The corresponding temperature control strategy and pressure threshold are loaded in real time according to the workpiece material type.

8. The adaptive optical polishing system based on dynamic temperature control and multimodal detection according to claim 1, characterized in that: The transfer learning framework includes: Pre-trained ResNet-34 backbone network to extract multimodal data features; The adaptation layer is a 256-node fully connected layer, fine-tuned based on small sample data, and the loss function is: ； Where, is the surface roughness, For processing efficiency.

9. An adaptive optical polishing method based on dynamic temperature control and multimodal detection applicable to any one of 1-8, characterized in that: The optical polishing method comprises the following steps: Data acquisition and preprocessing: Synchronously trigger the infrared thermal imager, laser interferometer, pressure sensor (1kHz), and acoustic emission sensor. Use a dynamic time warping algorithm to align the timing, perform wavelet threshold denoising on the acoustic emission signal, and use Kalman filtering on the temperature data. Thermal-mechanical coupling analysis and command generation: Based on finite element simulation and LSTM network, thermal deformation trend is predicted and the polishing head deformation compensation current is output. and coolant gradient parameters ; Real-time defect detection and parameter correction: When the surface roughness Ra is greater than 0.5nm or the acoustic emission energy exceeds three standard deviations of the baseline, the PPO reinforcement learning model is triggered to correct the pressure, speed and cooling strategy within 50ms; Cross-cycle iterative optimization: Store historical process data, train the LSTM model to output the optimal parameter combination, and adapt to new workpiece types through transfer learning.

10. The adaptive optical polishing method based on dynamic temperature control and multimodal detection according to claim 9, characterized in that: The data preprocessing includes: Temperature data is normalized to the range of 0-1, and pressure is normalized to the maximum range; Laser interferometer data is interpolated and aligned to a high-frequency temperature data stream.

Images