A full-band error cooperative control spatiotemporal co-integration polishing method and device

Abstract

This application discloses a spatiotemporal co-processing polishing method and equipment for full-band error collaborative control, relating to the field of ultra-precision optical processing. The method includes: acquiring the low-frequency, mid-frequency, and high-frequency errors of the target sample; constructing a time-space dual-dimensional control mechanism, optimizing and correcting low-frequency errors in the time domain using dwell time, and determining surface control points in the spatial domain through a B-spline surface dimensionality reduction optimization strategy, and inversely inferring the helix angle distribution of the dwell point tool influence function. Hybridizing the scratch trajectory through the helix angle distribution suppresses mid- and high-frequency errors. Furthermore, a multi-objective adaptive genetic algorithm optimization model is established to collaboratively solve for the optimal processing parameters with the goal of minimizing full-band errors, and drives the system to execute processing until the full-band errors meet the target. This application achieves full-band error collaborative control through a spatiotemporal co-processing method, accurately correcting low-frequency errors and suppressing mid- and high-frequency errors, while improving processing efficiency and accuracy, avoiding additional material removal, and improving processing convenience.

Description

Technical Field

[0001] This application relates to the field of ultra-precision optical processing, and in particular to a spatiotemporal fusion polishing method and equipment for full-band error collaborative control. Background Technology

[0002] As the performance of high-end optical systems continues to upgrade, terminal optical components face the challenge of coordinated control of errors across the entire frequency band: low-frequency height errors in the spatial domain directly affect the beam transmission path, while low-frequency gradient errors determine the system's imaging and energy focusing capabilities; mid-frequency ripple errors in the frequency domain easily induce light scattering, and directional textures of high-frequency surface roughness can reduce system contrast and even induce laser-induced damage.

[0003] Current mainstream sub-aperture polishing techniques have the following limitations: Traditional magnetorheological polishing relies on a time-invariant tool influence function (TIF) to correct low-frequency errors, but the convolution effect of regular scanning paths inevitably introduces mid-frequency ripple errors, creating an inherent contradiction between spatial domain and frequency domain error correction; high-frequency surface roughness produces directional textures due to the fixed abrasive scratch direction, which are difficult to eliminate using conventional methods; low-frequency gradient error convergence depends on repeated iterations and is limited by the dynamic performance of the machine tool, with abrupt changes in error gradients between adjacent processing points easily leading to discontinuous material removal; low-frequency error correction requires the introduction of an additional material removal layer, which reduces processing efficiency and may exacerbate mid-frequency errors. Furthermore, existing optimization algorithms suffer from high optimization dimensionality and low convergence efficiency, failing to achieve integrated error suppression. Summary of the Invention

[0004] The purpose of this application is to provide a spatiotemporal fusion polishing method and equipment for full-band error collaborative control. It can achieve full-band error collaborative control through spatiotemporal fusion processing, accurately correct low-frequency errors and suppress mid- and high-frequency errors, while improving processing efficiency and accuracy, avoiding additional material removal, and improving processing convenience.

[0005] To achieve the above objectives, this application provides the following solution.

[0006] Firstly, this application provides a spatiotemporal co-fusion polishing method for full-band error collaborative control, which is applied to a controllable spiral magnetorheological polishing system. The method includes: acquiring the full-band error of the sample to be processed; the full-band error includes: low-frequency error, mid-frequency error, and high-frequency error; dividing the mid-frequency and high-frequency errors into a time domain and the low-frequency error into a spatial domain; and constructing a time-space dual-dimensional control mechanism for the time and spatial domains based on the spatiotemporal co-fusion sub-aperture processing theory. The time-space dual-dimensional control mechanism includes: in the time domain, correcting the low-frequency error by optimizing the residence time distribution of the residence points; and in the spatial domain, determining the surface control points through a B-spline surface dimensionality reduction optimization strategy, using the control points to back-calculate the spiral angle distribution of the residence point tool influence function (TIF), and simultaneously suppressing the helical angle distribution by hybridizing the scratch trajectory through the helical angle distribution. The B-spline surface dimensionality reduction optimization strategy specifically includes: converting high-dimensional angle variables into a coefficient control point grid by representing the helical angle distribution using B-spline surfaces, and optimizing the helical angle of the stationary points; establishing a multi-objective fitness function with the goal of minimizing the full-band error, and constructing a multi-objective adaptive genetic algorithm optimization model based on the multi-objective fitness function and a time-space dual-dimensional control mechanism; inputting the full-band error into the multi-objective adaptive genetic algorithm optimization model for multi-objective solution to obtain the processing optimization parameters; inputting the processing optimization parameters into a controllable spiral magnetorheological polishing system to perform full-band error collaborative control processing on the sample to be processed; obtaining the residual error of the sample to be processed after processing, and if the residual error is still greater than the design standard, iteratively optimizing the model parameters of the multi-objective adaptive genetic algorithm optimization model until the residual error is lower than the design standard.

[0007] In a second aspect, this application also provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the spatiotemporal co-fusion polishing method for full-band error collaborative control as described in the first aspect.

[0008] Based on the specific embodiments provided in this application, the following technical effects are disclosed.

[0009] This application introduces the spatiotemporal fusion theory into sub-aperture polishing technology, resolving the coupling problem between spatial and frequency domain errors through co-optimization of dwell time and helix angle. A B-spline surface dimensionality reduction optimization strategy is used to determine surface control points, and the helix angle distribution of the tool influence function at the dwell point is inferred. By hybridizing the scratch trajectory through helix angle distribution, mid-to-high frequency errors are suppressed, solving the high-dimensional optimization problem and improving algorithm convergence efficiency. A multi-objective genetic algorithm is used to achieve flexible adaptation to processing requirements. This breakthrough overcomes the bottleneck of traditional processing's inability to coordinately control errors across the entire frequency band, achieving precise correction of low-frequency errors and coordinated suppression of mid-to-high frequency errors, while avoiding dependence on additional material removal layers during processing, thus improving processing efficiency and accuracy. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram illustrating the solution of dwell time based on B-spline curve dimensionality reduction, as provided in an embodiment of this application.

[0012] Figure 2 This is a schematic diagram of the intermediate frequency error on the surface of a component after controllable helical magnetorheological polishing, provided in an embodiment of this application; wherein, Figure 2 In Experiment 1, 'a' represents the mid-frequency error diagram of the component surface after conventional controllable spiral magnetorheological polishing with a line break distance of 1 mm and a spiral angle range of 0°. Figure 2 In Experiment 2, b represents the mid-frequency error diagram of the component surface after conventional controllable spiral magnetorheological polishing with a line break distance of 0.5 mm and a spiral angle range of 0°. Figure 2 In Experiment 3, 'c' represents the mid-frequency error diagram of the component surface after conventional controllable spiral magnetorheological polishing with a line break distance of 0.5 mm and a spiral angle range of 5°. Figure 2 In Experiment 4, d represents the mid-frequency error diagram of the component surface after controllable spiral magnetorheological polishing, where the line break distance is randomly selected between 0.5-1mm and the spiral angle range is 0°. Figure 2 In Experiment 5, 'e' represents the mid-frequency error diagram of the component surface after controllable helical magnetorheological polishing, where the line break distance is randomly selected between 0.5-1mm and the helical angle range is 5°. Figure 2 In Experiment 6, f represents the mid-frequency error diagram of the component surface after controllable spiral magnetorheological polishing, where the line break distance is randomly selected between 0.5-1mm and the spiral angle range is 20°. Figure 2In Experiment 7, g represents the mid-frequency error diagram of the component surface after optimized controllable helical magnetorheological polishing, where the line break distance is randomly selected between 0.5-1mm and the helical angle range is 20°. Figure 2 In Experiment 8, h represents the mid-frequency error diagram of the component surface after optimized controllable helical magnetorheological polishing, where the line break distance is randomly selected between 0.5-1mm and the helical angle range is 20°.

[0013] Figure 3 This is a schematic diagram illustrating the interaction between magnetorheological fluid ribbon forming and polishing particles and the surface of the sample to be processed under the action of a magnetic field, as provided in the embodiments of this application.

[0014] Figure 4 A schematic diagram illustrating the distribution of the optimal solution set at the Pareto front provided in an embodiment of this application; wherein... Figure 3 In this context, 'a' represents the curve showing the change in the average crossover probability per generation. Figure 3 In this context, b represents the curve showing the change in the average probability of mutation per generation. Figure 3 In this context, c represents the distribution of the optimal solution set at the Pareto front.

[0015] Figure 5 The experimental results curves for improving high-frequency surface roughness through controllable helical magnetorheological polishing are provided in the embodiments of this application; wherein, Figure 5 In the figure, 'a' represents the simulation and experimental surface roughness curves as a function of the helix angle. Figure 5 In the figure, b represents the power spectral density (PSD) curve analysis of the experimental results.

[0016] Figure 6 This is a schematic flowchart of a spatiotemporal fusion polishing method for frequency band error collaborative control provided in an embodiment of this application.

[0017] Figure 7 This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The existing ultra-precision machining of optical components mainly suffers from the following problems: traditional sub-aperture polishing cannot coordinate the control of spatial and frequency domain errors; mid-frequency ripple error and low-frequency height error are mutually restrictive and difficult to optimize simultaneously; high-frequency surface roughness has directional textures, affecting the light transmission and laser damage resistance of optical components; the gradient error convergence process is random and is significantly limited by the dynamic performance of the machine tool, requiring multiple iterations to reach the target; low-frequency error correction relies on additional material removal layers, which not only reduces processing efficiency but also easily introduces new errors; existing optimization algorithms are difficult to handle multi-objective coupled full-band error control, and the optimization variables such as the dwell point helix angle have extremely high dimensionality, resulting in slow algorithm convergence and poor stability.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1, as Figure 6 As shown, this embodiment provides a spatiotemporal co-fusion polishing method with full-band error collaborative control. The spatiotemporal co-fusion polishing method with full-band error collaborative control is applied to a controllable spiral magnetorheological polishing system. The method includes the following steps.

[0022] S1. Obtain the full-frequency error of the sample to be processed; the full-frequency error includes: low-frequency error (height error, gradient error), mid-frequency error (0.5mm). -1 -5mm -1 (Frequency band) and high-frequency error.

[0023] Furthermore, step S1 specifically includes the following steps.

[0024] S11. Obtain the low-frequency error of the sample to be processed using a Zygo GPI XP laser interferometer.

[0025] S12. Use a Zygo Newview 9000 white light interferometer to measure the high-frequency surface roughness of the sample to be processed and collect the high-frequency error of the sample to be processed.

[0026] S13. Perform bandpass filtering on the high-frequency error to obtain the mid-frequency error of the sample to be processed.

[0027] Furthermore, the sample to be processed includes at least: fused silica element, sapphire element and silicon carbide element.

[0028] S2. For example Figure 1As shown, the mid-frequency error and high-frequency error are divided into the time domain, and the low-frequency error is divided into the spatial domain. Based on the spatiotemporal coexistence sub-aperture processing theory, a time-space dual-dimensional control mechanism is constructed for the time and spatial domains. The time-space dual-dimensional control mechanism includes: in the time domain dimension, correcting the low-frequency error by optimizing the residence time distribution of the residence points; in the spatial domain dimension, dynamically adjusting the helix angle of the tool influence function TIF through a B-spline surface dimensionality reduction optimization strategy to hybridize the scratch trajectory and simultaneously suppress the mid-frequency error and high-frequency error. The B-spline surface dimensionality reduction optimization strategy specifically includes: converting the high-dimensional angle variable into a coefficient control point grid by representing the helix angle distribution through a B-spline surface, thereby optimizing the helix angle of the residence points.

[0029] Furthermore, in the time domain dimension, low-frequency errors are corrected by optimizing the residence time distribution of residence points. Specifically, in the time domain dimension, the low-frequency error correction term in the spatial domain is solved by residence time. Based on the linear equation system model, the workpiece surface error and the material removal amount of TIF are correlated. Under the premise of ensuring accurate convergence of low-frequency errors, the residence time of each residence point is solved.

[0030] In practical applications, the purpose of solving for the optimal residence time and helix angle distribution (B-spline control point parameters) is to ensure that TIF is sharpened in the high error gradient region to enhance the removal capability, and widened in the flat region to ensure a smooth transition.

[0031] Furthermore, in the spatial dimension, in the spatial domain dimension, the surface control points are determined by the B-spline surface dimensionality reduction optimization strategy. The control points are used to back-calculate the dwell point tool to influence the helical angle distribution of the TIF function. The scrubbing trajectory is then hybridized by the helical angle distribution to simultaneously suppress mid-frequency and high-frequency errors. Specifically, in the spatial dimension, the surface control points are determined by the rotation axis of the controllable helical magnetorheological polishing system using the B-spline surface dimensionality reduction optimization strategy. After obtaining the control points, the helical angle distribution of the TIF function is influenced by the control points to back-calculate the dwell point tool. Based on the helical angle distribution, the high-frequency error is eliminated by scrambling the abrasive scrubbing trajectory, while the mid-frequency error is eliminated by erasing the periodic superposition features.

[0032] Furthermore, the B-spline surface has an order of 3.

[0033] In practical applications, the B-spline surface dimensionality reduction optimization strategy transforms high-dimensional angle variables into a sparse control point mesh (e.g., 10 control points each in the x and y directions, for a total of 100 control points), replacing the traditional independent optimization with 240×240 dwell points. Third-order B-spline surfaces can achieve smoothing of the second derivative of the helix angle, ensuring stable movement of the machine tool's rotary axis, avoiding abrupt changes in TIF morphology, and significantly reducing the optimization dimensionality, thus improving the algorithm's convergence efficiency.

[0034] Furthermore, the accuracy of the coefficient control point grid is 0.1 mm.

[0035] In practical applications, intermediate frequency error detection employs a 0.1mm fine grid and bandpass filtering, randomly selecting multiple 10mm×10mm regions to extract error features, ensuring detection accuracy as described above. Figure 2 As shown.

[0036] S3. With the goal of minimizing the error across the entire frequency band, a multi-objective fitness function is established, and based on the multi-objective fitness function and the time-space dual-dimensional control mechanism, a multi-objective adaptive genetic algorithm optimization model is constructed.

[0037] Step S3 specifically includes the following steps.

[0038] S31. Establish a multi-objective fitness function with the goal of minimizing the error across the entire frequency band.

[0039] S32. An adaptive genetic algorithm is used to dynamically adjust the crossover and mutation probabilities, and a time-space dual-dimensional control mechanism is used to collaboratively optimize the spiral angle distribution and residence time of the B-spline control points.

[0040] S33. Pareto front is obtained by non-dominated sorting and crowding calculation, and a multi-objective adaptive genetic algorithm optimization model is constructed accordingly.

[0041] Furthermore, the expression for the multi-objective fitness function is as follows.

[0042] .

[0043] In the formula, The value of the multi-objective fitness function; , , , Weighting coefficients are divided into low-frequency height, low-frequency gradient, mid-frequency, and high-frequency error. The root mean square of the low-frequency error is expressed in nm. The root mean square of the gradient error; The root mean square of the intermediate frequency error; The root mean square of the high-frequency error.

[0044] S4. Input the full-band error into the multi-objective adaptive genetic algorithm optimization model for multi-objective solution to obtain the processing optimization parameters.

[0045] S5. Input the processing optimization parameters into the controllable spiral magnetorheological polishing system to perform full-frequency band error collaborative control processing on the sample to be processed.

[0046] S6. After processing, obtain the residual error of the sample to be processed. If the residual error is still greater than the design standard, iteratively optimize the model parameters of the multi-objective adaptive genetic algorithm optimization model until the residual error is lower than the design standard.

[0047] Optionally, step S6 mainly involves starting the controllable spiral magnetorheological polishing system according to optimized parameters, and monitoring the TIF stability in real time during the processing (such as the magnetorheological fluid ribbon morphology, shear stress, etc.). Figure 3 As shown in the figure, after processing, repeat the previous detection, and iteratively optimize the parameters according to the residual error until the error of the entire frequency band meets the standard.

[0048] In practical applications, due to the complex coupling of height and gradient error correction processes, traditional sub-aperture polishing methods achieve gradient error convergence by correcting low-frequency surface shape errors to the limit of accuracy. This results in a near-random convergence process for gradient errors, often relying solely on repeated experiments and empirical adjustments to achieve indirect constraints. Furthermore, traditional sub-aperture processing cannot control mid-frequency errors due to the convolutional material removal mechanism, which inevitably introduces a large amount of convolutional mid-frequency ripple error. Therefore, frequency domain error control remains a bottleneck in traditional sub-aperture processing. The spatiotemporal co-fusion sub-aperture processing method can further improve surface processing accuracy and control the introduction of mid-frequency ripple error, building upon traditional time-invariant removal function processing. However, selecting different removal functions at various dwell points will inevitably yield completely different processing results. Therefore, it is essential to rationally determine the combination of removal functions at each processing position to synergistically optimize the error across the entire frequency band.

[0049] To address the problem of cross-band error control where each element is coupled and subject to conflicting effects, adaptive genetic algorithms, compared to traditional genetic algorithms, dynamically adjust crossover and mutation probabilities based on fitness functions, effectively improving computational efficiency and solution accuracy. Therefore, this embodiment employs a multi-objective adaptive genetic algorithm to optimize the helical angle combination of the removal function at each processing point. The helical angle distribution of the removal function is represented using B-spline surfaces, reducing the dimensionality of the optimization parameters and enabling faster and better convergence. To achieve coordinated optimization of spatial and frequency domain errors, a multi-objective genetic algorithm capable of co-optimizing low-frequency errors, gradient errors, and mid-frequency errors in specific frequency bands is required, along with a corresponding multi-objective fitness function. Due to the mutual constraints of optimization objectives, a Pareto front is introduced for multi-objective optimization. Optimizing one objective function may sacrifice the performance of others. Therefore, through multiple iterations and calculations, a relatively ideal solution is obtained, and a Pareto optimal solution set is introduced to represent the optimal helical angle combination of the removal function under different processing requirements.

[0050] The calculation process of the spatiotemporal symbiotic sub-aperture machining model based on controllable spiral magnetorheological polishing.

[0051] 1. Obtain the target machining surface shape and removal function (to meet the memory requirements of the calculation process, the calculation network is divided into 0.5mm).

[0052] 2. Given control points P ij The value of is obtained by using B-spline curves to obtain the distribution of all helical angles at the processing dwell point.

[0053] 3. Set control points P ij The values ​​are imported into the genetic algorithm, and the algorithm is initialized and optimized.

[0054] 4. Construct a system of linear equations and solve for the dwell time (using the algorithm in Section 2.3.2), and solve for the corresponding machining residual surface. H res Surface shape distribution.

[0055] 5. Obtain the machining residual surface Low-frequency error, gradient error, and mid-frequency error in a specific frequency band are used to determine the corresponding values ​​of the fitness function.

[0056] 6. Repeat the above operations for iterative optimization, and combine with the Pareto front to obtain the Pareto optimal solution set, thereby obtaining the corresponding optimal helix angle matching scheme under different processing requirements.

[0057] In this study, low-frequency surface shape errors in the fitness function were evaluated using surface RMS, while gradient errors were evaluated using GRMS. Mid-frequency errors in specific frequency bands required finer network partitioning; the computational network was divided into 0.1mm partitions (which, according to the sampling theorem, can be analyzed down to 5mm). -1 (Spatial frequency band). Ten 20mm×20mm local regions were randomly selected on the machined surface. Since the helix angle and dwell time were already obtained (the computational network was divided into 0.5mm sections), linear interpolation was used to interpolate the helix angle and dwell time distributions into local regions with a computational network size of 0.1mm (this closely matches the actual machining process, which is not discrete but continuous machining). The RMS error in a specific mid-frequency band was extracted by bandpass filtering of the obtained regions.

[0058] The fitness function is a multi-objective minimization problem. For this type of problem, , , recorded as Also known as solution θ 1. Strong Pareto Dominance Solution θ 2. If the solution θ 1. Superior to the solution for a certain objective θ 2, but inferior to the solution in another objective. θ 2, then it is called a solution. θ 1. Indistinguishable from solution θ2, that is θ 1, θ 2. The solutions are mutually non-dominant. In the decision space, if no solution has a superior objective function value compared to any other solution, then there is no other solution that can dominate the other solutions. θ 1. Solution θ 1 represents a Pareto optimal solution. The set of all Pareto optimal solutions is called the Pareto optimal solution set, and their corresponding function values ​​in the objective space constitute the Pareto front. The multi-objective genetic algorithm introduces a non-dominated sorting mechanism to stratify individuals in the population based on their dominance relationships. Individuals with higher dominance levels are closer to the Pareto front in the objective space, thus having a higher selection probability during evolution. Within the same level, a "crowding" index is further introduced to quantify the distribution density of individuals in the objective space, prioritizing the retention of sparsely distributed individuals that can maintain solution set diversity. During the selection phase, parent individuals are determined based on the non-dominated level and crowding, and the next generation of offspring is generated through crossover and mutation operations. Subsequently, the parent and offspring are merged to form a new population, and the above evolutionary steps are repeated until the termination condition is met. Finally, a Pareto optimal solution set is obtained for the three sets of objectives, and selection is made from the Pareto front based on the actual processing requirements for specific objectives.

[0059] As an optional implementation method, this application also uses the Corning 7980 fused silica element as an example to elaborate in detail the spatiotemporal co-polishing method for full-band error collaborative control, as follows.

[0060] The experiment was conducted using a controllable spiral magnetorheological polishing machine tool (equipped with a 200mm diameter polishing wheel and a B-axis rotation mechanism) independently developed in the laboratory. Fused silica elements made of Corning 7980 were selected as the processing samples, with specifications including 50mm×50mm×10mm (initial surface roughness RMS 0.363nm, initial surface shape error RMS 22.8nm) and 100mm diameter circular elements (initial surface shape error RMS λ / 20, λ=632.8nm).

[0061] Experimental parameters were set as follows: the magnetorheological fluid consisted of carbonyl iron powder (average particle size 3.77 μm, volume fraction 40%), nanodiamond particles (size 54 nm), deionized water (volume fraction 55%), and additives (5%); the polishing wheel speed was 250 rpm, the magnetic field current was 8 A, the magnetorheological fluid viscosity was 155 Pa·s, the flow rate was 110 L / h, the immersion depth of the polishing ribbon and the workpiece was 0.2 mm, the scanning interval was 0.5-1 mm (adaptive), the helix angle range was 20°, and the processing removal depth was 1 μm.

[0062] The experiment consisted of three steps: the first step was preliminary detection, using a Zygo GPI XP laser interferometer to measure the initial surface shape error, and then applying a 0.1mm mesh and bandpass filtering (0.5mm). -1 -5mm -1 The intermediate frequency error (initial RMS 0.503 nm) was extracted; the high frequency surface roughness (initial RMS 0.363 nm, with directional texture) was measured using a Zygo Newview 9000 white light interferometer (20x lens, detection area 0.43 × 0.43 mm); the initial gradient error GRMS was calculated to be 48.98 nm / cm.

[0063] The second step is parameter optimization, which involves importing the detection data into a multi-objective adaptive genetic algorithm: setting the population size to 15, the maximum number of iterations to 30, and adaptively adjusting the crossover and mutation probabilities; using a 3rd-order B-spline surface (10 control points each in the x and y directions) to represent the helical angle distribution; the optimization objectives are to minimize the low-frequency error RMS, the gradient error GRMS, and the mid-frequency error RMS; the optimal solution is selected through the Pareto front; the dwell time is solved using the LSQR algorithm, with the termination condition being a residual error L2 norm of 0.02 nm or 500 iterations; the helical angle distribution is determined by the B-spline control point parameters to ensure that the angle change between adjacent dwell points is ≤2°, avoiding abrupt changes in the machine tool's rotation axis.

[0064] The third step is processing and verification. The controllable spiral magnetorheological polishing system is started according to optimized parameters. During processing, the shear stress is monitored in real time using an Anton Paar MCR302 magnetorheological instrument (maintained at 21 kPa) to ensure TIF stability. After processing, the first step is repeated. The results show that the low-frequency surface shape error RMS has decreased to 0.939 nm, the gradient error GRMS has decreased to 5.496 nm / cm, and the mid-frequency error RMS has decreased to 0.189 nm (PSD characteristic peak value decreased from 0.711 nm). 2 •mm reduced to 0.312nm 2 The surface roughness RMS of the high-frequency surface is reduced to 0.251 nm (the directional texture is completely eliminated); compared with traditional magnetorheological polishing, the error accuracy of the whole frequency band is improved by 10-20 times, the processing efficiency is improved by 35%, and no additional material removal layer is required.

[0065] Experimental results verification: By comparing the error distribution before and after processing, as shown below... Figure 5The experimental results curves of high-frequency surface roughness improvement by controllable spiral magnetorheological polishing and the distribution of various residual errors on the surface after CPP experimental processing clearly verify that the method of this application not only realizes the synchronous and coordinated control of low, medium and high frequency errors of optical components, but also improves the full-band error accuracy by 10-20 times compared with traditional magnetorheological polishing, and improves the processing efficiency by more than 35%. It effectively avoids the accumulation of error coupling and dependence on additional material removal layers, and fully meets the requirements of high-end fields such as inertial confinement fusion and extreme ultraviolet lithography for the full-band accuracy of optical components.

[0066] The technical effects of this application are as follows.

[0067] First, this application achieves coordinated control of the full-band error of optical components, breaking through the bottleneck of "mutual constraint between spatial and frequency domain errors" in traditional processing: through the "time-space" dual-dimensional control of TIF, low-frequency errors are precisely corrected by residence time, mid-frequency errors are suppressed by helix angle rotation to eliminate periodic ripples, and high-frequency errors are eliminated by disrupting the scratch trajectory to eliminate directional textures. Experiments verify that the peak value of mid-frequency error PSD characteristics is reduced by 99.938% (from 0.711nm). 2 •mm reduced to 0.312nm 2 The high-frequency surface roughness RMS was optimized from 0.363nm to 0.251nm, and the gradient error GRMS was reduced from 48.98nm / cm to 5.496nm / cm, achieving simultaneous compliance with the error across the entire frequency band.

[0068] Secondly, the dimensionality reduction optimization of B-spline surfaces in this application improves processing efficiency and stability: it transforms high-dimensional helix angle optimization into low-dimensional control point optimization, avoiding sudden changes in the machine tool rotation axis, while improving the optimization convergence efficiency by more than 3 times; it can achieve high-precision correction of low-frequency errors without introducing an additional material removal layer (traditional methods require an additional 200nm layer, which can be omitted in this application), improving processing efficiency by more than 30%, and avoiding new errors introduced by the additional layer.

[0069] Furthermore, the multi-objective adaptive genetic algorithm and Pareto front enhance adaptability: the optimization objective can be flexibly selected according to actual processing requirements, such as focusing on mid-frequency suppression of inertial confinement fusion elements or gradient convergence of extreme ultraviolet lithography elements; the algorithm avoids premature convergence by dynamically adjusting crossover and mutation probabilities, such as... Figure 4 As shown, the consistency of solutions obtained from multiple independent runs is over 90%.

[0070] Furthermore, this application is not only applicable to magnetorheological polishing, but can also be extended to other spatiotemporally controllable TIF processing technologies such as rotating water jets and ion beams, which has important promotional value for ultra-precision optical manufacturing.

[0071] Example 2: This application also provides a computer device, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 7 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores and processes data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements the methods described above.

[0072] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0073] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0074] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0075] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A spatiotemporal co-fusion polishing method with full-band error collaborative control, wherein the spatiotemporal co-fusion polishing method with full-band error collaborative control is applied to a controllable spiral magnetorheological polishing system, characterized in that, The method includes: Obtain the full-frequency error of the sample to be processed; the full-frequency error includes: low-frequency error, mid-frequency error and high-frequency error; The mid-frequency and high-frequency errors are divided into the time domain, and the low-frequency error is divided into the spatial domain. Based on the spatiotemporal coexistence sub-aperture processing theory, a time-space dual-dimensional control mechanism is constructed for both the time and spatial domains. This time-space dual-dimensional control mechanism includes: in the time domain, optimizing the residence time distribution of residence points to correct low-frequency errors; in the spatial domain, determining surface control points through a B-spline surface dimensionality reduction optimization strategy, using these control points to infer the helix angle distribution of the residence point influence function TIF, and simultaneously suppressing mid-frequency and high-frequency errors by hybridizing the scratch trajectory through the helix angle distribution. Specifically, the B-spline surface dimensionality reduction optimization strategy includes: converting high-dimensional angle variables into a coefficient control point grid by representing the helix angle distribution using a B-spline surface, thereby optimizing the helix angle of the residence points. With the goal of minimizing the error across the entire frequency band, a multi-objective fitness function is established, and based on the multi-objective fitness function and the time-space dual-dimensional control mechanism, a multi-objective adaptive genetic algorithm optimization model is constructed. The full-band error is input into a multi-objective adaptive genetic algorithm optimization model for multi-objective solution to obtain the processing optimization parameters; The processing optimization parameters are input into the controllable spiral magnetorheological polishing system to perform full-frequency band error collaborative control processing on the sample to be processed. After processing, the residual error of the sample to be processed is obtained. If the residual error is still greater than the design standard, the model parameters of the multi-objective adaptive genetic algorithm optimization model are iteratively optimized until the residual error is lower than the design standard.

2. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, Obtain the full-frequency error of the sample to be processed, specifically including: Low-frequency errors of the sample to be processed were obtained using a Zygo GPI XP laser interferometer. The high-frequency surface roughness of the sample to be processed was measured using a Zygo Newview 9000 white light interferometer, and the high-frequency error of the sample to be processed was collected. Bandpass filtering is applied to the high-frequency error to obtain the mid-frequency error of the sample to be processed.

3. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, In the time domain, low-frequency errors are corrected by optimizing the residence time distribution of residence points, specifically including: In the time domain, the low-frequency error correction term in the spatial domain is solved by dwell time. Based on the linear equation model, the workpiece surface error and the amount of material removed by TIF are correlated. Under the premise of ensuring accurate convergence of low-frequency error, the dwell time of each dwell point is solved.

4. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, In the spatial dimension, and in the spatial domain dimension, the control points of the surface are determined through a B-spline surface dimensionality reduction optimization strategy. The helix angle distribution of the influence function TIF is then used to infer the dwell point from the control points. Furthermore, the helix angle distribution is used to hybridize the scratch trajectory, simultaneously suppressing mid-frequency and high-frequency errors. Specifically, this includes: In the spatial dimension, the control points of the surface are determined by the rotation axis of the controllable helical magnetorheological polishing system using a B-spline surface dimensionality reduction optimization strategy. After obtaining the control points, the helical angle distribution of the TIF influence function is determined by using the control points to back-infer the dwell point tool. Based on the helical angle distribution, high-frequency errors are eliminated by scrambling the abrasive grain scratching trajectory, while mid-frequency errors are eliminated by erasing the periodic superposition features.

5. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, With the goal of minimizing the error across the entire frequency band, a multi-objective fitness function is established. Based on the multi-objective fitness function and a time-space dual-dimensional control mechanism, a multi-objective adaptive genetic algorithm optimization model is constructed, specifically including: A multi-objective fitness function is established with the goal of minimizing the error across the entire frequency band. An adaptive genetic algorithm is used to dynamically adjust the crossover and mutation probabilities, and a time-space dual-dimensional control mechanism is used to synergistically optimize the spiral angle distribution and residence time of B-spline control points. Pareto front is obtained by non-dominated sorting and crowding calculation, and a multi-objective adaptive genetic algorithm optimization model is constructed accordingly.

6. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, The expression for the multi-objective fitness function is as follows: ； In the formula, The value of the multi-objective fitness function; , , , Weighting coefficients are divided into low-frequency height, low-frequency gradient, mid-frequency, and high-frequency error. The root mean square of the low-frequency error is expressed in nm. The root mean square of the gradient error; The root mean square of the intermediate frequency error; The root mean square of the high-frequency error.

7. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, The B-spline surface is of order 3.

8. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, The accuracy of the coefficient control point grid is 0.1 mm.

9. The spatiotemporal co-fusion polishing method for full-band error collaborative control according to claim 1, characterized in that, The sample to be processed includes at least: fused silica elements, sapphire elements, and silicon carbide elements.

10. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the spatiotemporal co-fusion polishing method for full-band error collaborative control as described in any one of claims 1-9.

Images