Optical lens curved surface polishing method based on femtosecond laser

Abstract

The application relates to a kind of optical lens curved surface polishing method based on femtosecond laser applied to the technical field of optical lens precision machining, and the core step is: the hard synchronization trigger connection of femtosecond laser and coaxial white light OCT is established;After laser pulse, OCT is triggered to collect three-dimensional data with delay, and coaxial micro air flow is swept to resist interference simultaneously;The qualified judgment of processing depth, surface roughness and subsurface damage is carried out by using the multi-parameter end point model of weighted voting mechanism;If unqualified, then according to the deviation and material removal rate, 3D galvanometer scanning parameters and femtosecond laser parameters are dynamically adjusted, and predictive thermal management based on OCT reflectance monitoring is executed.The application solves the four big problems of monitoring interference, single quality criterion, efficiency and precision imbalance and thermal management lag in the prior art through the closed loop of anti-interference monitoring, multi-parameter intelligent judgment, dynamic adaptive execution and thermal management, and realizes the high-precision, high-efficiency and high-qualified polishing of high-end optical lens curved surface.

Description

Technical Field

[0001] This invention relates to the field of precision machining technology for optical lenses, specifically to a method for polishing curved surfaces of optical lenses based on femtosecond lasers. Background Technology

[0002] In large numerical aperture lithography systems and high-precision optical imaging systems, the application of optical lenses with special surface shapes (such as aspherical and freeform surfaces) is essential to eliminate aberrations and ensure image quality. The processing quality of these lenses directly depends on real-time controllable high-precision polishing technology. Currently, mainstream lens polishing technologies are mainly divided into two categories, both of which have significant drawbacks:

[0003] 1. Limitations of mechanical polishing technology: Mechanical polishing removes surface material through cutting and plastic deformation, requiring the use of various grinding tools and abrasives. For free-form curved lenses, repeated manual polishing and surface shape inspection are necessary. For example, when processing a 200mm diameter sapphire aspherical lens, at least 5 rounds of polishing and 3 rounds of inspection are required, with a single processing cycle exceeding 48 hours. This results in low production efficiency (daily capacity ≤ 2 pieces), a yield rate of less than 75%, and an inability to meet mass production requirements. Furthermore, grinding tool wear leads to poor processing consistency, making it difficult to exceed 1.2λ (λ=632.8nm) in surface shape accuracy, which is unsuitable for the lithography machine's requirement of ≤0.8λ for lens surface shape accuracy.

[0004] 2. Deficiencies of non-mechanical polishing technologies: Although non-mechanical polishing (such as electrochemical polishing, plasma polishing, liquid jet polishing, etc.) has improved the level of automation, it has two major problems: First, the cost is high (the cost of a single plasma polishing equipment exceeds 5 million yuan, and the replacement cycle of consumables is short, with a single replacement cost exceeding 100,000 yuan); Second, it requires the use of harmful chemicals such as hydrofluoric acid and nitric acid, and about 20L of fluoride-containing wastewater is generated for every 100 lenses processed, causing irreversible pollution to the environment, which is not in line with the current industrial trend of green manufacturing.

[0005] 3. Shortcomings of existing femtosecond laser polishing technology: Although femtosecond lasers, as a "cold processing" method, have the advantages of wide material adaptability and a small heat-affected zone (≤1.5μm), existing technologies still face four industry-recognized technical bottlenecks when achieving batch processing and high consistency of optical lenses of arbitrary surface shapes:

[0006] Monitoring interference problem: The plasma generated during the processing (lasting about 3-8μs) ​​and material debris seriously interfere with the detection signal of the coaxial OCT, resulting in a detection signal-to-noise ratio of only 10:1 and a depth detection error of ≥0.5μm, which makes the subsequent closed-loop control lose a reliable data basis;

[0007] The problem of single quality judgment: Existing technologies mostly use "processing depth" as the end point of judgment, which cannot conduct online evaluation of surface roughness (Ra is usually ≥1.8nm) and subsurface damage (such as the 915nm characteristic phase transition peak of quartz lenses). This leads to hidden scrapping of "geometric dimensions meet the standards but optical performance is not up to standard", resulting in low yield.

[0008] The problem of imbalance between efficiency and precision: The material removal rate varies significantly in different areas of complex curved surfaces. Fixed processing parameters (such as scanning line spacing of 50μm and laser energy of 5μJ) cannot simultaneously meet the high precision requirements of steep sections and the high efficiency requirements of planar sections, resulting in low overall processing efficiency (≤15mm² / min).

[0009] The problem of lagging thermal management: Although femtosecond laser is a cold processing method, under the requirements of high repetition frequency and high efficiency processing, the heat accumulation effect can still cause local micro-deformation of the lens (protrusion ≥0.3μm). Traditional thermal management is mostly post-compensation, with a delayed response, which seriously affects the consistency of the surface shape of the entire aperture.

[0010] To address the aforementioned issues, this invention proposes a collaborative intelligent polishing method involving "femtosecond laser-3D galvanometer-coaxial OCT," which achieves high-precision, high-efficiency, and high-consistency processing of high-end optical lenses through hard synchronization triggering, multi-parameter judgment, dynamic execution, and thermal management. Summary of the Invention

[0011] The purpose of this invention is to overcome the shortcomings of the existing technology and solve the four major industry bottlenecks of "monitoring interference problem, single quality judgment problem, efficiency and accuracy imbalance problem, and thermal management lag problem", so as to achieve high precision, high efficiency and high pass rate grinding of high-end optical lenses of any shape and multiple materials.

[0012] To achieve the above objectives, this invention provides a method for polishing curved surfaces of optical lenses based on femtosecond lasers. The system upon which this method relies includes (see...). Figure 2 ):

[0013] Femtosecond laser: pulse width 100-300fs, output power 1-5W, pulse energy 1-10μJ, repetition frequency adjustable from 100kHz to 2MHz;

[0014] Programmable delay circuit: Model AD9516, trigger signal delay accuracy ±0.1μs, CLK_IN pin is connected to the pulse synchronization signal of the femtosecond laser, and DELAY_OUT pin outputs the trigger signal to the OCT detection unit to achieve hard synchronization;

[0015] Coaxial white light OCT detection unit: axial resolution ≤5μm, lateral resolution ≤10μm, field of view 100×100mm, sampling frequency ≥1kHz, wavelength range 800-1000nm, resolution 1nm;

[0016] 3D galvanometer: scanning range ≥300×300mm, positioning accuracy ±5μm, maximum scanning speed 1000mm / s, supports scanning paths such as spirals and gratings;

[0017] Coaxial micro-airflow nozzle: made of 316L stainless steel, with an angle of 30°±5° to the laser beam path, and an inner diameter of 0.5-1mm;

[0018] Inert gas supply unit: provides nitrogen / argon gas with a purity ≥99.99%, pressure regulation range 0.1-0.5MPa, and flow rate controlled by a mass flow controller;

[0019] Industrial control computer: CPU is Intel i7-13700K, memory is 32GB, runs multi-parameter endpoint model algorithm, and is linked with 3D galvanometer and OCT detection unit through RS485 communication protocol (response time ≤1ms).

[0020] Processing platform: vacuum adsorption type, with repeatability accuracy of ±2μm, used to fix the lens to be processed.

[0021] The femtosecond laser-based method for polishing curved surfaces of optical lenses provided by this invention is based on a logic chain of "anti-interference monitoring - multi-parameter determination - dynamic execution - mode adaptation," and specifically includes the following steps (see...). Figure 1 ):

[0022] Step S1: System startup and hard synchronization establishment

[0023] The optical lens substrate to be polished (such as a sapphire aspherical lens with a diameter of 200mm and a curvature of R=50mm) is fixed on the processing platform. The coaxial white light OCT detection unit, femtosecond laser, and 3D galvanometer are started. The programmable delay circuit AD9516 is configured through the industrial control computer to establish a hard synchronization trigger connection between the femtosecond laser pulse signal and the OCT detection unit. The femtosecond laser outputs one pulse synchronization signal (frequency 100kHz-2MHz) to the CLK_IN pin of AD9516. After delay calculation, AD9516 outputs a trigger signal from the DELAY_OUT pin to the OCT acquisition control module. The delay accuracy is ±0.1μs, ensuring accurate timing matching of "laser pulse output-delay-OCT acquisition".

[0024] Step S2: Anti-interference 3D data acquisition

[0025] Through a hard-synchronization trigger connection, a 5-10μs delay is made after the femtosecond laser pulse output (avoiding the peak period of plasma interference; experimental verification shows that the OCT signal-to-noise ratio is ≥32:1 with a 5μs delay and ≥35:1 with an 8μs delay), triggering the coaxial white light OCT detection unit to acquire three-dimensional data (including depth data, intensity data, and spectral data) of the area to be processed on the lens. Simultaneously, the industrial control computer outputs control signals to the coaxial micro-airflow nozzle, spraying nitrogen gas (purity ≥99.99%) supplied by the inert gas supply unit at a flow rate of 0.3-0.8L / min, instantly dispersing debris with a particle size ≤1μm in the area to be processed. Experimental data shows that this operation can reduce the detection error caused by debris obstruction from 0.6μm to below 0.1μm.

[0026] The acquired 3D data is preprocessed for denoising using denoising algorithms (including but not limited to wavelet denoising, Gaussian denoising, and median denoising). This invention preferably uses a wavelet denoising algorithm of "db4 wavelet basis + 3-level decomposition + soft thresholding". Specifically, the original intensity data is decomposed into 3 levels of high-frequency coefficients and 1 level of low-frequency coefficients. The high-frequency coefficients are truncated using a soft threshold of "threshold = 0.1 × max (absolute value of high-frequency coefficient)". The signal-to-noise ratio of the reconstructed data is ≥35:1, and the depth data accuracy reaches ±0.1μm.

[0027] Step S3: Multi-parameter quality assessment

[0028] The preprocessed 3D data is input into a preset multi-parameter endpoint model, and the processing area is judged according to the following logic to determine whether it meets the preset processing target (see...). Figure 3 ):

[0029] Processing depth determination: Compare the OCT depth data with the target depth (e.g., target depth 50μm, allowable error ≤ ±0.3μm). If the target depth is met, 1 point is awarded for the depth item; otherwise, 0 points are awarded.

[0030] Surface roughness determination: Surface roughness Ra is inverted through the standard deviation of intensity data. The calculation logic is as follows: take 100 consecutive intensity sample values. Calculate the mean According to the formula Calculate Ra. If Ra ≤ 1 nm, the roughness item gets 1 point; otherwise, it gets 0 points.

[0031] Subsurface damage assessment: Analyze the spectral data. If no material phase transition characteristic peaks (such as the 915nm characteristic peak of quartz and the 850nm characteristic peak of sapphire) are found, or if the intensity of the characteristic peaks is ≤5% of the baseline, then the damage item will receive 1 point; otherwise, it will receive 0 points.

[0032] Weighted voting calculation: Weighted score = Depth score × 0.4 + Roughness score × 0.3 + Damage score × 0.3. If the weighted score ≥ 0.9 (i.e., at least two items score 1 point and the total score ≥ 0.9), the machining is deemed qualified, and step S5 is executed; otherwise, it is deemed to require remapping, and deviation data (e.g., "Depth deviation -0.4μm, Ra=1.2nm") is output, and step S4 is executed. The output signal of the multi-parameter endpoint model is directly transmitted to the control unit of the 3D galvanometer via the RS485 communication protocol, realizing closed-loop control of algorithm decision-making and hardware execution, and avoiding parameter adjustment lag.

[0033] Step S4: Dynamic Parameter Adjustment and Thermal Management. Based on the deviation data and the real-time calculated material removal rate, dynamically adjust the parameters of the 3D galvanometer and femtosecond laser, and simultaneously perform thermal management.

[0034] Material removal rate calculation: according to the formula The calculation is performed, where Δd is the difference in depth data between two adjacent OCT acquisitions (e.g., the depth of the first acquisition is 49.5μm, and the depth of the second acquisition is 49.8μm, Δd=0.3μm), and Δt is the time interval between the two acquisitions (e.g., Δt=0.6s), then R=0.5μm / s;

[0035] Galvanometer parameter adjustment: If the material removal rate is <0.5μm / s (e.g., on a steep curved surface, R=0.3μm / s), set the scanning path of the 3D galvanometer to a spiral scan (with the pitch matching the line spacing), and reduce the line spacing from the initial value of 50μm to 20-30μm, and increase the number of scans from 1 to 2-3 (experimental verification: when the line spacing is 25μm, the removal rate is increased to 0.6μm / s, and the depth deviation is ≤±0.2μm).

[0036] Laser parameter adjustment: If Ra > 1nm (e.g. Ra = 1.2nm), increase the repetition frequency of the femtosecond laser from the initial value of 1MHz to 2MHz, and decrease the pulse energy from the initial value of 5μJ to 2μJ (refinement stage parameters). Through "high frequency and low energy", optical finishing is achieved. Experiments show that this adjustment can reduce Ra to below 0.8nm.

[0037] Thermal management steps: The reflectivity change rate at the edge of the processing area is monitored in real time using OCT (sampled every 10ms). If the change rate is ≥3% / ms (predicting heat accumulation), the 3D galvanometer scanning sequence is immediately adjusted, prioritizing the processing of cold areas ≥5mm from the current area. The cooling time for these cold areas is ≥20ms (experimental verification: after 20ms cooling, the area temperature drops from 50℃ to 25℃, and the thermal deformation protrusion decreases from 0.3μm to 0.1μm). If the OCT detects a local thermal deformation protrusion ≥0.2μm, the femtosecond laser energy is reduced to below the preset thermal damage safety threshold. The preset thermal damage safety threshold is 80% of the laser damage threshold of the substrate material of the lens to be processed. For example, the laser damage threshold for quartz lenses is 5μJ / μm², for sapphire lenses it is 8μJ / μm², and for infrared glass lenses it is 7μJ / μm². Step S5: The 3D galvanometer is moved to the next processing area, and steps S2-S4 are repeated until the entire curved surface of the lens is polished.

[0038] 3. Processing mode adaptation

[0039] Before step S1, a preset processing mode can be selected according to processing requirements (see [reference]). Figure 4 The parameters and effects of each mode are as follows:

[0040] Quality-priority mode: In step S3, the Ra threshold is adjusted to ≤0.8nm, and in step S4, the reflectivity change rate monitoring threshold is adjusted to ≥2% / ms; the laser parameters are 2-4μJ / 1.5-2MHz, the 3D galvanometer scanning speed is 300-500mm / s, and the line spacing is 20-30μm; it is suitable for aspherical lenses of lithography machines, with finished product Ra≤0.8nm and surface accuracy≤0.5λ;

[0041] Efficiency-first mode: In step S3, the depth error threshold is adjusted to ≤±0.5μm, and in step S4, the reflectivity change rate monitoring threshold is adjusted to ≥5% / ms; the laser parameters are 4-8μJ / 0.5-1MHz (roughing stage), the 3D galvanometer scanning speed is 600-1000mm / s, and the line spacing is 40-60μm; it is suitable for infrared detection lenses (with slightly lower accuracy requirements), and the processing efficiency reaches 22mm² / min, which is 1.2 times that of the standard mode;

[0042] Standard mode: Parameters are as described in steps S1-S5, balancing accuracy and efficiency, suitable for most general optical lenses, finished surface accuracy 0.6λ, pass rate 95%, efficiency 18mm² / min.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. Strong anti-interference capability, solving the problem of monitoring interference: By using "hard synchronization delay (5-10μs) + coaxial micro airflow to remove debris", the signal-to-noise ratio of OCT detection is improved from 10:1 to 35:1, and the depth detection error is ≤±0.1μm, providing a reliable data source for closed-loop control.

[0045] 2. Comprehensive quality assessment, solving the problem of single quality assessment: The multi-parameter endpoint model covers "depth-roughness-subsurface damage", reducing the scrap rate of finished products from 25% to below 5%, meeting the requirements of high-end lenses for all-dimensional quality.

[0046] 3. Efficiency and precision are balanced, solving the problem of the imbalance between efficiency and precision: Dynamic parameter adjustment increases the processing efficiency from 15mm² / min to 22mm² / min (efficiency priority mode) and the surface accuracy from 1.2λ to 0.6λ, breaking through the bottleneck of "efficiency and precision cannot be achieved at the same time".

[0047] 4. Proactive thermal management to solve the problem of lagging thermal management: Predictive thermal management based on OCT reflectivity change rate + thermal deformation compensation reduces the heat-affected zone from 1.5μm to 0.6μm, ensuring the surface shape consistency of large-format lenses.

[0048] 5. Wide material compatibility and environmentally friendly: Compatible with materials such as quartz, sapphire, and infrared glass, with lens diameters ranging from 50 to 500 mm; requires no harmful chemicals, uses only inert gases, and produces no wastewater or waste residue. It also eliminates mold wear, reducing consumable costs by more than 30%. Attached Figure Description

[0049] Figure 1 This is a flowchart of the method for polishing curved surfaces of optical lenses based on femtosecond lasers according to the present invention.

[0050] Figure 2 This is a structural block diagram of the system upon which the femtosecond laser-based optical lens surface polishing method of this invention depends.

[0051] Figure 3 This is a flowchart of the multi-parameter endpoint determination logic in this invention.

[0052] Figure 4 This is a flowchart of the processing mode and adaptive control in this invention. Detailed Implementation

[0053] To make the technical solution of the present invention clearer, the following detailed description is provided with reference to four specific embodiments and one comparative example. All embodiments are based on the above system architecture and steps, with only the parameters adjusted to adapt to different scenarios, thus verifying the technical progress of the present invention.

[0054] Example 1: Standard pattern processing of sapphire aspherical lenses

[0055] Processing object: Sapphire aspherical lens, diameter 200mm, curvature R=50mm, target depth 50μm, material laser damage threshold 8μJ / μm² (thermal damage safety threshold 6.4μJ / μm²).

[0056] System parameters:

[0057] Femtosecond laser: pulse width 200 fs, initial parameters 5 μJ / 1 MHz;

[0058] 3D galvanometer: The scanning path is a spiral (pitch 50μm), the initial value of the scanning line spacing is 50μm, and the scanning speed is 600mm / s;

[0059] OCT detection: delay time 8μs, micro-airflow nozzle flow rate 0.5 L / min (nitrogen, purity 99.999%), wavelet denoising (db4 basis, 3-layer decomposition).

[0060] Multi-parameter thresholds: depth error ≤ ±0.3μm, Ra ≤ 1nm, spectral characteristic peaks ≤ 5% baseline;

[0061] Thermal management: Reflectivity change rate threshold ≥3% / ms, cold zone cooling time ≥20ms.

[0062] Processing procedure and results:

[0063] Step S2: The depth data acquired by OCT is preprocessed to achieve an accuracy of ±0.1μm and a signal-to-noise ratio of 36:1;

[0064] Step S3: The first judgment result is "Depth 49.7μm (score 1), Ra=1.1nm (score 0), no damage peak (score 1)", weighted score = 1×0.4+0×0.3+1×0.3=0.7<0.9, and it is determined that further processing is needed;

[0065] Step S4: Calculate the removal rate R = 0.3μm / 0.6s = 0.5μm / s (no need to adjust the galvanometer), adjust the femtosecond laser parameters to 2μJ / 2MHz, and perform one additional scan; secondary judgment Ra = 0.9nm (score 1), weighted score = 1.0, qualified; finished product effect: processing cycle 12 hours, surface accuracy 0.6λ (λ = 632.8nm), Ra = 0.9nm, heat-affected zone 0.6μm, pass rate 95%.

[0066] Example 2: Quality-first processing of quartz aspherical lenses

[0067] Processing object: Quartz aspherical lens (for lithography machine), 100mm in diameter, curvature R=30mm, target depth 30μm, laser damage threshold 5μJ / μm² (thermal damage safety threshold 4μJ / μm²).

[0068] System parameters:

[0069] Femtosecond laser parameters: 2-4 μJ / 1.5-2 MHz;

[0070] 3D galvanometer: scanning path is a spiral (pitch 20μm), scanning speed is 300mm / s, and line spacing is 20μm;

[0071] OCT detection: delay time 5μs, micro-flow nozzle flow rate 0.3 L / min (argon, purity 99.999%).

[0072] Multi-parameter thresholds: Ra≤0.8nm, spectral characteristic peaks≤3% baseline;

[0073] Thermal management: Reflectivity change rate threshold ≥2% / ms.

[0074] Finished product results: processing cycle of 10 hours, surface accuracy of 0.5λ, Ra=0.7nm, no subsurface damage peak (915nm characteristic peak intensity ≤2% baseline), fully meeting the requirements of lithography machine lenses, with a pass rate of 98%.

[0075] Example 3: Processing Infrared Glass Freeform Surface Lenses in an Efficiency-First Mode

[0076] Processing target: Infrared glass freeform surface lens, 300mm in diameter, 60μm in target depth, laser damage threshold 7μJ / μm² (thermal damage safety threshold 5.6μJ / μm²).

[0077] System parameters:

[0078] Femtosecond laser parameters: roughing 4-8μJ / 0.5-1MHz, finishing 2μJ / 2MHz (only activated when Ra is out of tolerance);

[0079] 3D galvanometer: scanning path is a spiral (pitch 50μm), scanning speed is 800mm / s, and line spacing is 50μm;

[0080] OCT detection: delay time 10μs, micro-airflow nozzle flow rate 0.8 L / min (nitrogen, purity 99.99%).

[0081] Multi-parameter thresholds: depth error ≤ ±0.5μm, Ra ≤ 1.2nm;

[0082] Thermal management: Reflectivity change rate threshold ≥ 5% / ms.

[0083] Processing procedure and results:

[0084] Processing area calculation: The effective processing area (material removal area) of the freeform surface lens is approximately 19800 mm² (based on the material removal volume conversion at a target depth of 60 μm).

[0085] Efficiency calculation: Processing cycle 15 hours = 900 minutes, processing efficiency = 19800mm² ÷ 900min = 22mm² / min (meets the efficiency index of the efficiency priority mode).

[0086] Finished product results: surface accuracy 0.8λ, Ra=1.1nm, pass rate 92%, meeting the requirements of infrared detection system.

[0087] Example 4: Adaptation verification of a 500mm diameter infrared glass lens

[0088] Processing object: 500mm diameter infrared glass lens, spherical surface type (curvature R=200mm), target depth 80μm, laser damage threshold 7μJ / μm² (thermal damage safety threshold 5.6μJ / μm²).

[0089] System parameter adaptation:

[0090] 3D galvanometer: Employs "regional spiral scanning" (each region has a diameter of 100mm, matching the OCT detection field of view (100×100mm) to avoid decreased detection accuracy due to an excessively large field of view; regions overlap by 5mm to prevent processing breaks at region junctions).

[0091] Micro-airflow nozzle: The flow rate is increased to 0.8 L / min to ensure complete removal of debris during the processing of large-diameter lenses;

[0092] Femtosecond laser: repetition frequency set to 1MHz, pulse energy 6μJ (balancing processing efficiency and heat accumulation).

[0093] Thermal management: Cooling time in cold zones is extended to 25ms, adapting to the thermal diffusion characteristics of large-diameter lenses.

[0094] Finished product results: Processing cycle of 30 hours, surface accuracy of 0.9λ, Ra=1.3nm, and pass rate of 88%, proving that this method can be adapted to the processing of lenses with a full range of diameters from 50 to 500mm.

[0095] Comparative example: Processing effect without enhancement

[0096] Using a baseline system (femtosecond laser + 3D galvanometer + ordinary OCT, without delayed triggering, micro-blowing, and multi-parameter model) lacking the "anti-interference monitoring + multi-parameter judgment + dynamic thermal management" functions of this invention, the same sapphire aspherical lens as in Example 1 was processed. The results are as follows: the OCT signal-to-noise ratio was only 10:1, and the depth detection error was ≥0.5μm; only the depth judgment endpoint was passed, the finished product Ra=1.8nm, and the intensity of the 915nm characteristic peak on the subsurface reached 12% of the baseline (damage exists); without thermal management, the heat-affected zone was 1.5μm, and the surface accuracy was 1.2λ; the processing cycle was 24 hours, and the pass rate was 75%. The comparison shows that this invention achieves significant improvements in accuracy, efficiency, and pass rate, demonstrating clear technological advancement.

[0097] Additional notes:

[0098] Method for determining the laser damage threshold: The sample to be processed (such as quartz, sapphire) is pretreated (cleaned and dried), and the sample is placed in the femtosecond laser optical path. The laser energy density is gradually increased (step size 0.5 μJ / μm²), and the spectral data of the sample is detected by coaxial white light OCT. When the material phase transition characteristic peak (such as 915 nm for quartz and 850 nm for sapphire) appears in the spectrum, the energy density at this time is recorded as the laser damage threshold of the material. The thermal damage safety threshold is taken as 80% of the laser damage threshold to ensure that no new damage is generated during the rescan.

[0099] The determination of the wavelet denoising threshold is based on the following: For the original intensity signal acquired by OCT, the denoising effect of different thresholds (0.05, 0.1, 0.15, 0.2×max (absolute value of high frequency coefficient)) is verified through experiments: When the threshold = 0.1, the signal-to-noise ratio of the denoised signal is ≥35:1, the depth detection error is ≤±0.1μm, and no effective signal (such as intensity fluctuations related to surface roughness) is lost. This threshold is the optimal choice, which takes into account both denoising effect and signal integrity.

[0100] Adjustability of parameters: All parameters in this invention (such as delay time 5-10μs, airflow 0.3-0.8 L / min) have been experimentally verified and can achieve the expected results within the adjustment range. They are not unique fixed values. For example, the debris removal rate is 92% when the micro-airflow nozzle flow rate is 0.2L / min and 98% when it is 1 L / min, which can meet the detection requirements, proving the flexibility and versatility of the parameters.

[0101] Hardware compatibility: In addition to the programmable delay circuit AD9516, CDCE6214 (delay accuracy ±0.2μs, meeting the requirements) can also be selected; the 3D galvanometer can be ScanLab hurrySCAN 30 (scanning range 300×300mm, positioning accuracy ±5μm), neither of which will affect the implementation of the solution.

[0102] Algorithm reproducibility: The wavelet denoising "db4 basis + 3-level decomposition" and the multi-parameter "weighted voting" are both industry-standard algorithm frameworks that can be implemented by those skilled in the art using the MATLAB wavelet toolbox and the Python numpy library, without any algorithmic barriers.

[0103] Compatibility of denoising algorithms: In addition to wavelet denoising used in this embodiment, those skilled in the art can choose Gaussian filtering, median filtering, or other equivalent denoising algorithms according to actual needs. As long as a preprocessing effect with a signal-to-noise ratio ≥ 35:1 can be achieved, they all fall within the protection scope of this invention.

Claims

1. A method for polishing curved surfaces of optical lenses based on femtosecond lasers, characterized in that, Includes the following steps: S1: Fix the optical lens substrate to be polished on the processing platform, start the coaxial white light OCT detection unit, femtosecond laser and 3D galvanometer, and establish a hard synchronous trigger connection between the femtosecond laser pulse signal and the coaxial white light OCT detection unit through a programmable delay circuit; S2: Through the aforementioned hard synchronization trigger connection, after a 5-10 μs delay following the femtosecond laser pulse output, the coaxial white light OCT detection unit is triggered to acquire three-dimensional data of the area to be processed on the lens. Simultaneously, a coaxial micro-airflow nozzle at a 30°±5° angle to the laser path is controlled to eject inert gas at a flow rate of 0.3-0.8 L / min to disperse debris in the area to be processed. The three-dimensional data includes depth data, intensity data, and spectral data. S3: Input the three-dimensional data into a preset multi-parameter endpoint model to determine whether the area to be processed meets the preset processing target: the preset processing target includes an error between the processing depth and the target depth ≤ ±0.3μm, a surface roughness Ra ≤ 1nm based on the intensity data inversion, and no material phase transition characteristic peaks or characteristic peak intensities ≤ 5% of the baseline in the spectral data; if all three conditions are met, the processing is deemed qualified and step S5 is executed; if at least one condition is not met, the processing is deemed to be supplementary, deviation data is output, and step S4 is executed. S4: Based on the deviation data and the material removal rate calculated in real time, dynamically adjust the scanning parameters of the 3D galvanometer and the processing parameters of the femtosecond laser: if the material removal rate is <0.5μm / s, reduce the scanning line spacing of the 3D galvanometer to 20-30μm and increase the number of scan cycles; if the surface roughness Ra derived from the intensity data is >1nm, increase the repetition frequency of the femtosecond laser to 2MHz and reduce its pulse energy to 2μJ; simultaneously, perform thermal management steps: monitor the reflectivity change rate of the processing area edge in real time through the coaxial white light OCT detection unit; if the reflectivity change rate is ≥3% / ms, adjust the scanning sequence of the 3D galvanometer, prioritizing the processing of cold areas ≥5mm away from the current area; if the coaxial white light OCT detection unit detects local thermal deformation protrusions ≥0.2μm, reduce the femtosecond laser energy to below the preset thermal damage safety threshold and perform additional scanning on the protruding area; S5: Control the 3D galvanometer to move to the next processing area, repeat steps S2-S4 until the polishing of the entire lens surface is completed; In step S4, the preset thermal damage safety threshold is 80% of the laser damage threshold of the substrate material of the lens to be processed.

2. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, In step S1, the delay accuracy of the programmable delay circuit trigger signal of the hard synchronous trigger connection is ±0.1μs.

3. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, In step S2, the inert gas is nitrogen or argon with a purity ≥ 99.99%.

4. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, In step S2, the axial resolution of the coaxial white light OCT detection unit is ≤5μm, the lateral resolution is ≤10μm, and the sampling frequency is ≥1kHz. After the three-dimensional data is acquired, it is preprocessed by a denoising algorithm, and the signal-to-noise ratio of the preprocessed signal is ≥35:

1.

5. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, The output signal of the multi-parameter endpoint model is directly transmitted to the control unit of the 3D galvanometer. The control unit adjusts the scanning path parameters of the 3D galvanometer in real time according to the qualified or supplementary processing signal to realize closed-loop control of algorithm decision-making and hardware execution. The multi-parameter endpoint model adopts a weighted voting mechanism for judgment: 1 point is awarded when a single indicator meets the preset processing target, and 0 points are awarded when it does not meet the target. The weighted score is calculated as: processing depth error score × 0.4 + surface roughness score × 0.3 + spectral characteristic peak intensity score × 0.

3. Only when the weighted score is ≥ 0.9 is the processing deemed qualified.

6. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, In step S4, the material removal rate is determined by the formula... The calculation is performed, where Δd is the difference between two adjacent depth data acquisitions, Δt is the time interval between the two acquisitions, and the unit is μm / s; and the initial value of the scanning line spacing of the 3D galvanometer is 50μm, the initial value of the repetition frequency of the femtosecond laser is 1MHz, and the initial value of the pulse energy is 5μJ.

7. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, In step S4, when adjusting the 3D galvanometer scanning sequence, the cooling time of the cold zone is ≥20ms. After cooling is completed, the process returns to the original processing area to continue polishing.

8. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, Before step S1, a preprocessing step is also included: selecting a preset processing mode according to processing requirements, wherein the processing mode includes quality priority mode, efficiency priority mode and standard mode; if quality priority mode is selected, the threshold of surface roughness Ra in step S3 is adjusted to ≤0.8nm, and the monitoring threshold of reflectivity change rate in step S4 is adjusted to ≥2% / ms; if efficiency priority mode is selected, the threshold of processing depth error in step S3 is adjusted to ≤±0.5μm, and the monitoring threshold of reflectivity change rate in step S4 is adjusted to ≥5% / ms.

9. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 8, characterized in that, When selecting the efficiency-first mode, in step S4, the pulse energy of the femtosecond laser is adjusted to 4-8 μJ, the repetition frequency is adjusted to 0.5-1 MHz, and the scanning speed of the 3D galvanometer is adjusted to 600-1000 mm / s.

10. The method for polishing curved surfaces of optical lenses based on femtosecond lasers according to claim 1, characterized in that, The optical lens substrate is made of quartz, sapphire or infrared glass, and the lens surface type includes spherical, aspherical or freeform, with a lens diameter ranging from 50 to 500 mm.

Images