A Smart Tolerance Allocation Method for Optical Systems Based on Measured Surface Profile Reference

By establishing a tolerance model based on measured surface datum, and combining wave aberration analysis and intelligent optimization algorithms, the problems of large computational load and poor adaptability in tolerance allocation of off-axis freeform surface optical systems in existing technologies are solved, achieving efficient and accurate tolerance allocation and cost control.

CN123111641APending Publication Date: 2026-11-17BEIJING INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611202523.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies for tolerance analysis of off-axis freeform surface optical systems suffer from problems such as large computational load, poor adaptability, and cost waste, making it difficult to achieve efficient and accurate tolerance allocation in complex systems.

Method used

By combining wavefront aberration analysis, sensitivity analysis, and intelligent optimization algorithms, a tolerance model based on the measured surface datum is established, a unified datum configuration is performed, an analytical mapping model is constructed, tolerance allocation is optimized, wavefront error is reduced, and manufacturing costs are controlled.

Benefits of technology

It achieves efficient and precise tolerance allocation in complex optical systems, improves optimization efficiency and accuracy, reduces production costs, and is suitable for various optical systems, including freeform surface lenses and off-axis optical systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111641A_ABST
    Figure CN123111641A_ABST
Patent Text Reader

Abstract

This invention discloses an intelligent tolerance allocation method for optical systems based on measured surface profile references. By combining wavefront aberration analysis, sensitivity analysis, and intelligent optimization algorithms, it integrates high-precision and high-efficiency calculations to establish a tolerance model that conforms to actual processing and inspection, effectively overcoming the shortcomings of existing methods. This method not only optimizes tolerance allocation and reduces wavefront errors but also controls manufacturing costs. Through intelligent optimization algorithms, it finds the optimal solution among multiple design variables, ensuring a balance between overall system performance and manufacturing efficiency. The proposed solution possesses strong physical interpretability and adaptability, capable of handling complex nonlinear relationships. Sensitivity analysis and error propagation models enable rapid identification and priority control of key optical components affecting wavefront errors, improving optimization efficiency and accuracy. Furthermore, this technology is widely applicable to various optical systems, including freeform lenses, off-axis optical systems, and precision instruments, ensuring efficient and accurate tolerance allocation in optical design and manufacturing processes across different fields.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the fields of tolerance analysis and allocation in optical systems, computer-aided optical design, and computer-aided optical manufacturing. Specifically, it relates to an intelligent tolerance allocation method and system for off-axis freeform surfaces and reflective optical systems, which performs small-sample tolerance analysis under unified reference tolerance constraints and achieves joint optimization of cost and performance based on wavefront aberration indices. Background Technology

[0002] Off-axis freeform surface optical systems possess advantages such as high image quality and small size, and have been widely used in fields such as space remote sensing, astronomical observation, and high-precision imaging. These systems have asymmetrical structures and multiple degrees of freedom, resulting in significant coupling between surface shape errors, assembly and adjustment errors, and manufacturing errors. Tolerance analysis has become a crucial link connecting optical design and optical manufacturing.

[0003] In publicly available literature such as the *Journal of the Optical Society of America A*, methods for analyzing perturbation models and tolerance indices in optical systems are relatively mature. Commercial optical design software (such as CODEV and Zemax OpticStudio) commonly employs the tracking index method or polynomial coefficient method to establish tolerance sensitivity, and combines this with Monte Carlo sampling statistical performance distribution to complete tolerance budget allocation. Typical analysis methods include: 1. Tolerance sensitivity analysis: Apply positive and negative perturbations to each tolerance, evaluate the changes in performance indicators such as wavefront RMS and MTF, and form a sensitivity matrix.

[0004] 2. Monte Carlo tolerance analysis: A large number of random samples are taken within the tolerance range, and the statistical performance distribution and cumulative probability quantiles are used for yield evaluation.

[0005] 3. Tolerance Budget Allocation: Under the constraint of the total performance budget, the tolerance is allocated to each component. Common methods include equal allocation and allocation inversely proportional to sensitivity.

[0006] The current state of research both domestically and internationally is as follows:

[0007] In the tolerance calculation and allocation of freeform surface optical systems, there are three main technical approaches: The first is to derive the influence of each system element tolerance on wave aberration and MTF using vector wave aberration theory. This method provides high-precision theoretical support by accurately calculating the impact of each optical element tolerance on system performance. However, in practical applications, due to the increased complexity of optical systems, especially with multiple elements and nonlinear effects, the derivation and calculation process becomes cumbersome and computationally intensive, making it difficult to adapt to large-scale production. The second approach uses deep learning methods to construct a mapping network between tolerances and image quality information for tolerance allocation. This method has good adaptability and can automatically learn and handle complex nonlinear relationships, but it relies on a large amount of experimental data and computational resources, and the model's interpretability is poor, potentially leading to an inability to fully grasp the physical characteristics of the optical system in actual production. Finally, the sensitivity matrix method can be used to analyze error sources and optimize tolerance allocation by establishing a mathematical model between geometric aberrations and assembly errors. This method is relatively simple to calculate and suitable for efficient computation, but it assumes that the error sources are linear or approximately linear, which may not hold true in some complex systems, and it has poor adaptability to high-dimensional systems. Summary of the Invention

[0008] In view of this, the purpose of this invention is to provide an intelligent tolerance allocation method for optical systems based on measured surface profile references, offering a comprehensive, flexible, and efficient solution: by combining wavefront aberration analysis, sensitivity analysis, and intelligent optimization algorithms, high-precision calculation is combined with efficient calculation to establish a tolerance model that conforms to actual processing and inspection, effectively overcoming the shortcomings of existing methods. This method not only optimizes tolerance allocation and reduces wavefront errors but also controls manufacturing costs. Through intelligent optimization algorithms, it finds the optimal solution among multiple design variables, ensuring a balance between overall system performance and manufacturing efficiency. This solution possesses strong physical interpretability and adaptability, capable of handling complex nonlinear relationships. Through sensitivity analysis and error propagation models, it can quickly identify and prioritize the control of key optical components affecting wavefront errors, improving optimization efficiency and accuracy. Furthermore, this technology is widely applicable to various optical systems, including freeform lenses, off-axis optical systems, and precision instruments, ensuring efficient and accurate tolerance allocation in optical design and manufacturing processes across different fields.

[0009] A method for intelligent tolerance allocation in an optical system based on a measured surface profile datum includes the following steps: S1: Perform geometric correction on the measured surface data detected by the interferometer to align the measured surface data with the optical design coordinate system and generate reference surface data; perform reference injection on the optical system model based on the reference surface data, and filter the tolerance set to form a unified reference configuration; S2: Traverse each tolerance in the tolerance set, apply positive and negative perturbations to each tolerance under the unified reference configuration, perform tolerance analysis and extract the secondary degradation coefficient, and generate a sensitivity matrix; S3: Based on the sensitivity matrix and the distribution type of each tolerance, establish an analytical mapping model from the individual tolerance to the expected value of the system wavefront aberration degradation, and obtain the statistical expected wavefront aberration and the RSS spatial degradation budget. S4: Define the manufacturing cost model for each tolerance and determine the upper and lower limits of the half-width of each tolerance; S5: Based on the statistical expected wavefront difference, the RSS spatial degradation budget, and the manufacturing cost model, an optimization strategy is used to solve for the optimal half-width of each tolerance; S6: Discretize the optimal half-width by rounding to obtain a discrete tolerance value usable in engineering, and perform statistical verification by Monte Carlo sampling.

[0010] Preferably, the geometric correction in S1 includes performing elliptical or rectangular ratio correction and rotation processing on the measured surface data of the off-axis mirror to make the aperture shape, ratio and orientation of the measured surface data consistent with the optical design coordinate system; and performing interpolation repair on the measurement missing points within the aperture range.

[0011] Preferably, the reference injection in S1 includes: identifying the reflective mirror in the optical system model according to priority rules, establishing a matching mapping between the mirror and the reference surface data, and using the reference surface data as a unified reference injection analysis model for the reflective mirror; the tolerance screening is to remove parameterized surface tolerances and retain assembly tolerances.

[0012] Preferably, the quadratic degradation coefficient in S2 includes a constant term C, a linear term coefficient B, and a quadratic term coefficient A; the constant term C is determined by the wavefront difference under the nominal state, and the linear term coefficient B and the quadratic term coefficient A are obtained by solving the wavefront difference under positive and negative perturbations simultaneously with the constant term C.

[0013] Preferably, the analytical mapping model in S3 is as follows: for each tolerance, its wavefront error satisfies a quadratic polynomial degradation relationship near the nominal operating point; a statistical expectation wavefront error model is constructed by introducing a distribution factor, and a large-sample Monte Carlo random stacking analysis is replaced by a small-sample perturbation experiment plus statistical inference.

[0014] Preferably, the manufacturing cost model in S4 is described by a negative power-law model: [Form follows] The subscript i indicates the i-th type of tolerance. Indicates the half-width of the i-th tolerance class. , The calibration parameters represent the base cost level and the rate of cost increase as tolerances tighten, respectively. The calibration parameters are the fixed processing costs corresponding to the tolerance items; they are obtained by fitting the process standard data through three reference points.

[0015] Preferably, the optimization strategy in S5 is a constrained optimality conditional KKT strategy, with the goal of minimizing the total manufacturing cost of the system and the wavefront aberration degradation budget as a hard constraint. The closed-form optimal solution is obtained by constructing a Lagrange function. The optimal half-width is related to the cost parameter, sensitivity coefficient and Lagrange multiplier, and the Lagrange multiplier is obtained by bisection iterative solution.

[0016] Preferably, the optimization strategy in S5 is the marginal equilibrium iterative algorithm (MEI) strategy, which takes minimizing the degradation of wavefront aberrations of the system as the optimization objective and the total manufacturing cost budget as the hard constraint. The closed-form optimal solution is obtained by constructing a Lagrange function. The optimal half-width is related to the cost parameter, sensitivity coefficient and Lagrange multiplier. The Lagrange multiplier is obtained by iterative solution using the bisection method.

[0017] Preferably, the optimization strategy in S5 is the sum of squares and root of the statistical standard deviation (RSS) method; without considering manufacturing costs, the tolerance half-width is allocated according to the principle of equal RSS contribution, that is, each tolerance contributes equally to the total degradation budget.

[0018] Preferably, in step S6, after rounding and discretization, if the relative deviation between the rounded total cost and the cost budget exceeds a preset threshold, a dimensionless scaling factor is introduced to perform recalibration: the continuous optimal solution is multiplied by the scaling factor and then rounded, and the result is limited to the upper and lower limits of the tolerance; by performing a binary search on the scaling factor, the rounded total cost is made to approach the cost budget.

[0019] The present invention has the following beneficial effects: 1. Existing technologies typically use parametric surface tolerances in optical design software to approximate mirror manufacturing errors, and mix manufacturing tolerances and assembly tolerances in the same list for sensitivity calculations. However, inherent deviations exist between the interferometer detection coordinate system, the optical design coordinate system, and the mechanical assembly coordinate system. Existing methods do not systematically correct for this, resulting in sensitivity analysis, allocation optimization, and final verification being performed on different coordinate references. The accumulated systematic errors are sufficient to distort the sensitivity ranking of critical tolerances, causing the allocation scheme to deviate from actual assembly requirements. This invention establishes a unified reference processing flow in the first step. First, geometric correction is performed on the measured surface of the interferometer, including ellipse correction, rectangle scaling correction, and rotation, to accurately align the surface data with the design coordinates. Then, the reflecting mirror is identified in the optical model according to priority rules, and the corrected reference surface is injected into the model. Finally, parametric surface tolerances are actively eliminated and completely replaced by the measured surface, retaining only assembly tolerances for subsequent analysis. This closed-loop mechanism ensures that sensitivity extraction, optimization allocation, and Monte Carlo verification are all based on the same physical benchmark, eliminating analysis errors caused by coordinate confusion. This allows the final tolerance value to directly guide the assembly and adjustment operations without the need for additional calculations or empirical corrections.

[0020] 2. Existing tolerance analysis mainly follows two paths: one is large-sample Monte Carlo simulation, which requires thousands or even tens of thousands of ray tracing operations on the system. This is time-consuming in complex freeform surface systems and only provides statistical distributions, failing to establish an analytical relationship between individual tolerances and performance degradation. The other is the worst-case RSS method, which assumes all tolerances reach their limits simultaneously, resulting in overly stringent requirements that far exceed actual manufacturing capabilities, leading to unnecessary manufacturing cost waste. This invention applies only one positive and one negative perturbation to each tolerance, obtains its second-order degradation characteristics near the nominal operating point through tolerance analysis, and then introduces corresponding statistical weighting factors based on the actual distribution of different tolerance types to construct an analytical mapping between individual tolerances and the expected value of system wavefront aberration degradation. The computational cost of this method is only proportional to the number of tolerance terms, far less than large-sample sampling, while avoiding overly conservative worst-case assumptions. Furthermore, because a differentiable analytical model is established, it can directly provide gradient information for subsequent optimization, significantly improving engineering efficiency.

[0021] 3. Existing tolerance allocation methods generally employ proportional scaling or inverse sensitivity-based allocation. These methods fail to consider the significant differences in cost growth rates required to tighten different tolerance items to the same precision. Tightening the mirror tilt tolerance relies on high-precision turntable assembly and adjustment, leading to rapid cost increases, while tightening the mirror spacing tolerance may only require replacing the precision grinding process of the distance ring, resulting in a relatively gradual cost increase. Ignoring this difference inevitably leads to suboptimal overall cost. This invention first individually calibrates the cost variation law of each tolerance with increasing precision, and then provides two optional optimization strategies: one is to automatically allocate the budget to each tolerance item under a given wavefront aberration budget to minimize the total manufacturing cost; the other is to automatically allocate the budget under a given total cost budget to minimize the final wavefront aberration degradation. Both strategies guarantee global optimization under constraints, and the solution process is stable and fast. This invention also includes a rounding calibration step to ensure that the theoretically optimal solution, after being discretized into an engineering-markable tolerance value, still strictly meets budget constraints. This avoids the dilemma of optimization results that look good but are not practical, enabling optical designers to proactively choose optimization directions based on project budget or performance priorities, and scientifically allocate precision resources among different tolerance items to avoid precision waste or insufficient performance. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the off-axis reflective optical system to which this invention applies.

[0023] Figure 2 This is a flowchart illustrating the overall process of the intelligent tolerance allocation method for optical systems based on measured surface profile references of the present invention.

[0024] Figure 3 This is a schematic diagram of the original interferometer-detected surface data before surface preprocessing in this invention.

[0025] Figure 4 This is a schematic diagram of the reference surface data after geometric correction (ellipse to circle / rectangle ratio correction, rotation) and hole repair of the present invention.

[0026] Figure 5 This is the sensitivity matrix of the single-tolerance wavefront difference of the present invention as a function of the disturbance amplitude, and the corresponding MTF and RMS yields.

[0027] Figure 6 This represents the RSS degradation contribution corresponding to each tolerance after allocation under the KKT optimization strategy of this invention.

[0028] Figure 1 In the middle, 1-aperture plane, 2~4-reflecting mirror plane, 5-image plane. Detailed Implementation

[0029] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0030] To realize the intelligent tolerance allocation technology proposed in this invention, we will mainly conduct theoretical research and numerical simulation, and complete the research work in three major parts: the construction of unified benchmark tolerance constraints, small sample tolerance analysis, and research on intelligent tolerance allocation technology based on wave aberration.

[0031] 1. Research on the Construction Technology of Unified Datum Tolerance Constraints To implement the intelligent tolerance allocation technology proposed in this solution, it is first necessary to study the technology for constructing unified datum tolerance constraints. In off-axis freeform surface optical systems, tolerances are divided into manufacturing tolerances and assembly tolerances. Manufacturing tolerances include surface shape errors, tilt and eccentricity tolerances of off-axis mirror surfaces. Assembly errors include inter-component distances, component tilt, and eccentricity tolerances. In off-axis three-mirror optical systems, optical components are typically fixed and adjusted through mechanical structures. The tolerances of optical components are not only constrained by optical design but also influenced by the design and manufacturing of mechanical structures. Mechanical tolerances related to optical components include dimensional tolerances, form and position tolerances, and assembly accuracy of the mechanical structure. These mechanical tolerances directly affect the actual position and orientation of the optical components, thereby affecting the imaging quality of the system. Therefore, it is necessary to establish a mathematical relationship between the positional tolerances of optical components and the tolerances of the mechanical structure. Through coordinate transformation and error propagation models, optical tolerances are transformed into mechanical tolerances. Considering stress analysis during mechanical assembly, the relationship between mechanical stress and optical surface shape errors is incorporated into the tolerance analysis model. In the tolerance modeling stage, in addition to geometric parameters, material properties are introduced as input variables, and sensitivity analysis of multi-material components in the folding and reversing system is also supported.

[0032] A ray tracing model of the off-axis system was established using optical design software. Combined with the differential geometric parameters of the freeform surface, the influence of mechanical deviations on the optical path was quantified, serving as prediction samples for subsequent tolerance analysis models. For example, mirror position offset. It will cause a change in the angle of incidence of light. This leads to wavefront error. The lack of a reference point during the assembly and adjustment of freeform off-axis reflective systems makes assembly and adjustment extremely difficult. To address this issue, and considering ease of assembly and adjustment, the research team will use the measured surface shape of the primary mirror as a unified reference, and rationally construct tolerance constraints for the off-axis freeform optical system based on manufacturing feasibility. This will lay the technical foundation for subsequent intelligent tolerance allocation and ultimately build a tolerance constraint system based on a unified reference.

[0033] 2. Research on Small Sample Tolerance Analysis Techniques Based on a unified tolerance constraint system, the research team will use wavefront differential theory to study the impact of each tolerance term on the system's wavefront aberration. The tolerance term manifests as disturbances caused by mechanical structures, such as surface deformation, surface tilt, displacement, and curvature changes. These disturbances alter the propagation path of light, thus causing changes in wavefront aberration. By performing a Taylor expansion of the expression and retaining the second-order terms, the relationship between system performance and tolerance parameters can be approximated as a quadratic function, with coefficients determined through a mathematical model. Due to the high manufacturing cost and low batch size of off-axis reflectors, the number of actual samples available is small, necessitating small-sample tolerance analysis. Specific measurements will be performed on several actually manufactured off-axis freeform surface optical systems and catadioptric systems. Based on the small-sample method, the coefficients of the quadratic function will be calibrated through a finite number of disturbance experiments, and statistical inference will be performed by combining the tolerance distribution characteristics, achieving accurate tolerance analysis of the off-axis freeform surface optical system.

[0034] 3. Research on Intelligent Tolerance Allocation Technology Based on Wavelength Aberration The core of intelligent tolerance allocation technology lies in rationally allocating the manufacturing tolerances of each optical component in an optical system through a global optimization algorithm, thereby maximizing the overall performance of the system. Based on an established mapping model between error and wavefront aberration, dynamic optimization is performed by comprehensively considering the performance requirements of each component and its contribution to the overall system wavefront error. Through iterative optimization, this technology can find the optimal balance between system imaging quality and manufacturing cost, thus optimizing the production process while minimizing wavefront error. This technology, based on the specific design requirements of optical components, simultaneously adjusts the tolerances of multiple optical components through a global optimization algorithm. These optimization algorithms can find the optimal solution among multiple design variables, ensuring the minimization of system wavefront error and guaranteeing optimal final imaging quality. This intelligent optimization method not only significantly improves the performance of the optical system but also reduces unnecessary precision waste in the manufacturing process, effectively reducing production costs. This method provides a new and efficient solution for the design, manufacturing, and application of high-precision optical systems.

[0035] The tolerance allocation algorithm provided by this invention is applicable to refractive, reflective, and catadioptric optical systems, and is particularly suitable for high-precision systems containing mirrors that require interferometer-measured surface profiles to characterize mirror errors, such as off-axis reflective systems. Figure 1 This approach is also applicable to other optical engineering scenarios where a joint trade-off between wavefront aberration budget and manufacturing cost budget is required. Features in the following embodiments can be combined with each other unless otherwise specified.

[0036] The method of the present invention is executed in sequence according to six steps S1 to S6.

[0037] S1 is the unified configuration for surface preprocessing correction and tolerance datum. The input is the interferometer detection data and the optical system model, and the output is the datum surface data aligned with the design coordinate system and the selected tolerance set.

[0038] S2 is for tolerance set traversal and sensitivity matrix generation. The input is the filtered tolerance set and unified reference configuration, and the output is a sensitivity matrix containing the quadratic degradation coefficients A, B, C and the reference half-width.

[0039] S3 is used to establish a small-sample distribution perception degradation model for tolerance and wavefront aberration. The input is the sensitivity matrix and the tolerance distribution type, and the output is the statistical expectation wavefront aberration and the RSS spatial degradation budget.

[0040] S4 is used for calibrating the economic tolerance manufacturing cost model. The inputs are process reference data and accuracy level settings, and the outputs are the cost model and the upper and lower limits of the tolerance half-width. and .

[0041] S5 solves for the optimal tolerance half-width and allocation using three strategies: KKT, MEI, and RSS. The inputs are statistical sensitivity, cost model, and various budget constraints, and the output is the optimal half-width in the continuous domain.

[0042] S6 is for process allocation rounding, calibration, and statistical verification. Its inputs are the continuous optimal half-width t and a uniform baseline configuration. The output is the discrete tolerance half-width usable in engineering. The comparison results with Monte Carlo quantile performance are shown in the attached document. Figure 2 .

[0043] The following will elaborate on the detailed process of each stage.

[0044] Phase S1 focuses on establishing a unified benchmark. First, coordinate system analysis and parameter extraction are performed on the interferometer-detected surface data. Then, geometric correction algorithms are applied to the off-axis mirror detection data, encompassing ellipse correction, rectangle scaling, and rotation correction. Interpolation repair is performed on missing measurement points within the aperture range to generate benchmark surface data strictly aligned with the design coordinate system. In the optical system model, mirror surfaces are identified according to priority rules, a matching mapping between the mirror surface and the benchmark surface is established, and a unified benchmark is injected. Finally, tolerance screening is performed, discarding parametric surface tolerances by default and retaining assembly tolerances, forming a consistent analysis benchmark across the entire process.

[0045] The surface data preprocessing algorithm supports two types of input: for surface data output by the interferometer, integer surface values ​​are converted to wavelength units using a scaling factor; for 3D coordinate data in point cloud format, a 2D mesh is reconstructed based on planar coordinates, and the height values ​​are converted to wavelength units before being connected to a unified processing link. For off-axis detection of circular lenses, the ellipse correction algorithm automatically estimates the ellipse's center, major axis, minor axis, and principal axis tilt angle. After correcting the principal axis to align with the coordinate axes, interpolation and stretching are performed along the minor axis to transform the ellipse into a perfect circle, meeting the circular aperture index requirements. Areas outside the aperture are marked as invalid data. For rectangular lenses, the effective area detection and scale correction algorithm estimates the rectangle's center, width, height, and rotation angle. After correcting, interpolation and stretching are performed according to the design target aspect ratio to restore the design coordinate system's scale. The hole repair algorithm distinguishes between two types of invalid data: invalid areas outside the aperture are marked as invalid and not included in the calculation; holes within the aperture that failed measurement are repaired using nearest neighbor, linear, or bicubic interpolation to avoid error accumulation caused by multiple filling operations. The preprocessed datum surface data ensures coordinate system consistency throughout the entire process of inspection, design, and simulation, serving as the physical basis for subsequent tolerance analysis. Figure 3 and Figure 4 This is a schematic diagram showing the effect before and after surface preprocessing.

[0046] In off-axis freeform surface optical systems, the coordinate systems for lens processing and inspection, assembly and adjustment, and optical design typically deviate from each other. If parametric surface tolerances or uncorrected raw interferometric data are directly used for tolerance analysis, the surface errors in sensitivity calculation, tolerance allocation, and verification stages will not match the actual system in terms of position, scale, and direction, introducing systematic analysis errors. To address this, this invention constructs a two-layer unified reference system: the first layer is a measured surface geometry correction layer, which corrects the raw surface data output by the interferometer to reference data consistent with the design coordinate system, ensuring coordinate system consistency at the data level; the second layer is an optical model injection layer, which uses the corrected reference surface as the unified reference for the reflector in the analysis model, eliminating parametric surface tolerances, retaining assembly and adjustment tolerances, and reusing them throughout the entire analysis chain, ensuring reference consistency at the simulation level.

[0047] In the datum injection and tolerance screening algorithm, the identification of reflective surfaces follows a predetermined priority: first, surfaces with tolerance definitions are identified; second, surfaces possessing both reflective and circular properties are identified; and finally, all reflective surfaces are covered. Surface shape data is fuzzily matched according to the mirror role identifier or surface number to establish a mapping relationship between surfaces and datum surface shapes, ensuring that each reflective surface corresponds to the correct measured surface shape data. Datum positioning parameters can be automatically calculated or obtained through model queries. After datum injection, only eccentric adjustment tolerances are retained by default; clearing other tolerances does not affect the datum data, ensuring that the datum remains effective throughout the entire analysis process. The unified datum configuration adopts a full-link reuse mechanism, bound to the sensitivity matrix for transmission. Subsequent calculations directly read the same configuration, avoiding analysis errors introduced by datum differences.

[0048] S2 iterates through the tolerances item by item to generate the sensitivity coefficient matrix. First, the unified reference configuration is read and injected into the optical model. Full tolerance superposition analysis can be run first, exporting nominal performance parameters and quantile references, where the nominal wavefront deviation is the constant term C of the quadratic degradation model. Then, the tolerance set is traversed, clearing the remaining tolerances for each tolerance while retaining the unified reference. Positive and negative perturbations and corresponding compensators are applied to the current tolerance term, and tolerance analysis is run to output sensitivity data. Finally, the sensitivity results are analyzed, and the quadratic degradation coefficients A, B, and C, as well as the reference half-width, are extracted and assembled to form a complete sensitivity matrix.

[0049] S3 establishes a distribution-aware degradation model for small samples of individual tolerances and wavelet aberrations. Based on distribution-aware modeling, it first matches the corresponding distribution type and distribution factor k value according to the tolerance type, and then reads the quadratic degradation coefficients of each tolerance from the sensitivity coefficient matrix. , , Compared with reference half width Based on this, degradation feature extraction can be completed by applying only a finite number of perturbations to each tolerance, which is what this invention refers to as small sample tolerance analysis. Figure 5 As shown. Since the wavefront error of a single tolerance satisfies a quadratic polynomial degradation relation near the nominal operating point, its three coefficients A, B, and C can be uniquely determined by three independent data points: the nominal state provides C, and the positive and negative disturbances each provide an independent equation, i.e. , , Solving the system of equations simultaneously yields the following results: , RMS² is the wavefront error, the mean of the squared wavefront errors, a core indicator of system performance. A is the quadratic degradation intensity coefficient, representing the nonlinear influence of tolerance disturbances on the wavefront error; B is the linear asymmetric term coefficient, representing the asymmetric degradation characteristics in the positive and negative disturbance directions; C is the nominal wavefront error, corresponding to the system's baseline performance without tolerance disturbances. This means that only three rounds of tolerance analysis are needed to calibrate the quadratic degradation characteristics of a single tolerance, eliminating the need for thousands of Monte Carlo simulations of the entire system.

[0050] Building upon this foundation, a statistical expected wavefront difference model is constructed by introducing a distribution factor. This model replaces large-sample random stacking analysis with small-sample perturbation experiments and statistical inference, while avoiding the overly conservative nature of traditional worst-case RSS methods. Given a target degradation amount, the RSS spatial degradation budget is calculated, and the statistical expected wavefront difference corresponding to any candidate half-width vector can be evaluated, providing a reference for subsequent tolerance allocation. This technical approach can directly interface with the tolerance analysis output of mainstream optical design software, possessing advantages such as analytical differentiability and high computational efficiency, making it more suitable for engineering-scale batch applications.

[0051] S4 is the calibration manufacturing cost model. The manufacturing cost of a single tolerance is described using a negative power-law model, in the form of: In other words, the smaller the tolerance half-width, the higher the processing difficulty, and the manufacturing cost increases according to a negative power law, where the subscript i indicates the i-th type of tolerance. Indicates the half-width of the i-th tolerance class. , The calibration parameters represent the base cost level and the rate of cost increase as tolerances tighten, respectively. This represents the fixed processing cost corresponding to the tolerance item. Model parameters are calibrated using three reference points based on process standard data. The algorithm incorporates linked parameters for three precision levels: Base level is the basic precision level, with a relative cost coefficient of 1.00, a lower half-width limit of 0.10 times the reference value, and an upper half-width limit of 3.00 times, suitable for conventional precision systems; Precision level is the high precision level, with a relative cost coefficient of 1.35 times, a lower half-width limit of 0.05 times, and an upper half-width limit of 2.00 times, suitable for medium-high precision systems; High-Precision level is the high precision level, with a relative cost coefficient of 2.35 times, a lower half-width limit of 0.02 times, and an upper half-width limit of 1.50 times, suitable for ultra-high precision systems. The algorithm supports customization of cost parameters for each tolerance item, with user-defined parameters having higher priority than the built-in standard model to adapt to the process cost characteristics of different manufacturers.

[0052] In the S5 stage, three optimization strategies are used to solve for the optimal tolerance half-width, each adapting to different engineering constraint scenarios.

[0053] The first type is the constrained optimality condition (KKT) strategy, which solves for the minimum cost under given wavefront aberration degradation constraints, such as... Figure 6 This strategy aims to minimize the total system manufacturing cost, with wavefront aberration degradation budget as a hard constraint, making it a typical convex optimization problem. The optimization objective is to minimize the sum of the costs of all individual tolerances, with the constraint that the RSS degradation of all tolerances does not exceed a preset degradation budget, and that the half-width of each tolerance is within the allowable upper and lower limits. By constructing a Lagrangian function, taking the partial derivative with respect to each tolerance half-width and setting it to zero, a closed-form optimal solution can be derived. The optimal half-width is related to the cost parameters. , Sensitivity coefficient And related to the Lagrange multiplier λ, specifically in the form of Since both the optimization objective and constraints are convex functions, the KKT conditions are necessary and sufficient conditions for global optimum, and this closed-form solution corresponds to the global optimum. The Lagrange multiplier λ is solved iteratively using the bisection method; the RSS degeneration monotonically decreases as λ increases, and approximately 1 × 10⁻⁶ steps are needed to reach the optimal solution. -12The relative accuracy is on the order of magnitude, and both solution efficiency and accuracy meet engineering requirements. This strategy is suitable for scenarios with clearly defined system performance objectives, requiring the minimization of manufacturing costs without exceeding wavefront aberration degradation targets.

[0054] The second strategy is the Marginal Equilibrium Iterative Algorithm (MEI), which solves for the minimum wavefront aberration degradation under a given cost constraint. This strategy aims to minimize the system's wavefront aberration degradation, with the total manufacturing cost budget as a hard constraint, forming a dual optimization relationship with the KKT strategy. The optimization objective is to minimize the RSS degradation of all tolerance items, with the constraint that the sum of the costs of all individual tolerances does not exceed the preset cost budget, and that the half-width of each tolerance must be within the allowable upper and lower limits. The derived closed-form optimal solution is isomorphic to the KKT strategy, and the optimal half-width is related to the cost parameter. , Sensitivity coefficient And related to the Lagrange multiplier μ, specifically in the form of The optimal allocation satisfies the marginal equilibrium principle, meaning the ratio of marginal cost to marginal degradation is equal for all tolerance items. This is the origin of the name of the marginal equilibrium iterative algorithm. The Lagrange multiplier μ is solved iteratively using the bisection method. The total cost increases monotonically with increasing μ, and it can quickly converge to the budget constraint point. This strategy is suitable for scenarios where the manufacturing budget is fixed and the optimal wavefront RMS performance needs to be obtained within the budget constraint.

[0055] The third method is the Statistical Standard Deviation (RSS) method, which does not consider manufacturing costs and allocates the tolerance half-width according to the principle of equal RSS contribution. That is, each tolerance contributes equally to the total degradation budget. The formula for calculating the half-width is: , where n is the total number of tolerance items involved in tolerance allocation. The RSS space degradation budget, calculated from the target degradation amount, is the core constraint indicator for tolerance allocation optimization. This scheme mainly serves as a benchmark, used to compare the effects with the two cost optimization strategies and quantify the benefits of cost optimization.

[0056] S6 represents rounding, calibration, and statistical verification. The continuous domain optimal solution is discretized by rounding to two significant figures by default, yielding tolerance values ​​consistent with engineering representation conventions. For the MEI strategy scenario, if the relative deviation between the rounded total cost and the cost budget exceeds 0.5%, a dimensionless scaling factor s is introduced for recalibration: the continuous optimal solution is multiplied by the scaling factor and then rounded, limited to the upper and lower limits of the tolerance. A binary search is performed on the scaling factor to make the rounded total cost approach the preset cost budget, ensuring that the budget constraint still holds after discretization. The KKT strategy uses a degenerate budget as a hard constraint and typically does not perform scaling calibration to avoid violating performance constraints.

[0057] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent tolerance allocation in an optical system based on a measured surface profile datum, characterized in that, Includes the following steps: S1: Perform geometric correction on the measured surface data detected by the interferometer to align the measured surface data with the optical design coordinate system and generate reference surface data; Based on the aforementioned reference surface data, a reference injection is performed on the optical system model, and a tolerance set is selected to form a unified reference configuration; S2: Traverse each tolerance in the tolerance set, apply positive and negative perturbations to each tolerance under the unified reference configuration, perform tolerance analysis and extract the secondary degradation coefficient, and generate a sensitivity matrix; S3: Based on the sensitivity matrix and the distribution type of each tolerance, establish an analytical mapping model from the individual tolerance to the expected value of the system wavefront aberration degradation, and obtain the statistical expected wavefront aberration and the RSS spatial degradation budget. S4: Define the manufacturing cost model for each tolerance and determine the upper and lower limits of the half-width of each tolerance; S5: Based on the statistical expected wavefront difference, the RSS spatial degradation budget, and the manufacturing cost model, an optimization strategy is used to solve for the optimal half-width of each tolerance; S6: Discretize the optimal half-width by rounding to obtain a discrete tolerance value usable in engineering, and perform statistical verification by Monte Carlo sampling.

2. The method according to claim 1, characterized in that, The geometric correction in S1 includes performing elliptical or rectangular ratio correction and rotation processing on the measured surface data of the off-axis mirror to make the aperture shape, ratio and orientation of the measured surface data consistent with the optical design coordinate system; and performing interpolation repair on the measurement missing points within the aperture range.

3. The method according to claim 1, characterized in that, The reference injection in S1 includes: identifying the reflective mirror in the optical system model according to priority rules, establishing a matching mapping between the mirror and the reference surface data, and using the reference surface data as a unified reference injection analysis model for the reflective mirror; the tolerance screening is to remove parameterized surface tolerances and retain assembly tolerances.

4. The method according to claim 1, characterized in that, The quadratic degradation coefficient in S2 includes a constant term C, a linear term B, and a quadratic term A. The constant term C is determined by the wavefront difference under the nominal state, and the linear term B and the quadratic term A are obtained by solving the wavefront difference under positive and negative perturbations simultaneously with the constant term C.

5. The method according to claim 1, characterized in that, The analytical mapping model in S3 is as follows: for each tolerance, its wavefront error satisfies a quadratic polynomial degradation relationship near the nominal operating point; a statistical expectation wavefront error model is constructed by introducing a distribution factor, and a large-sample Monte Carlo random stacking analysis is replaced by a small-sample perturbation experiment plus statistical inference.

6. The method according to claim 1, characterized in that, The manufacturing cost model in S4 is described using a negative power-law model: [Form follows] The subscript i indicates the i-th type of tolerance. Indicates the half-width of the i-th tolerance class. , The calibration parameters represent the base cost level and the rate of cost increase as tolerances tighten, respectively. The calibration parameters are the fixed processing costs corresponding to the tolerance items; they are obtained by fitting the process standard data through three reference points.

7. The method according to claim 1, characterized in that, The optimization strategy in S5 is a constrained optimality conditional KKT strategy, with the goal of minimizing the total manufacturing cost of the system and wavefront aberration degradation budget as a hard constraint. The closed-form optimal solution is obtained by constructing a Lagrange function. The optimal half-width is related to the cost parameter, sensitivity coefficient and Lagrange multiplier. The Lagrange multiplier is obtained iteratively by the bisection method.

8. The method according to claim 1, characterized in that, The optimization strategy in S5 is the marginal equilibrium iterative algorithm (MEI) strategy, which aims to minimize the degradation of wavefront aberrations in the system and uses the total manufacturing cost budget as a hard constraint. It solves the closed-form optimal solution by constructing a Lagrange function. The optimal half-width is related to the cost parameter, sensitivity coefficient, and Lagrange multiplier, which is solved iteratively by the bisection method.

9. The method according to claim 1, characterized in that, The optimization strategy in S5 is the sum of squares and root of the statistical standard deviation (RSS) method; without considering manufacturing costs, the tolerance half-width is allocated according to the principle of equal RSS contribution, that is, each tolerance contributes equally to the total degradation budget.

10. The method according to claim 1, characterized in that, If the relative deviation between the rounded total cost and the cost budget exceeds a preset threshold after discretization in S6, a dimensionless scaling factor is introduced to perform recalibration: the continuous optimal solution is multiplied by the scaling factor and then rounded, and the result is limited to the upper and lower limits of the tolerance; by performing a binary search on the scaling factor, the rounded total cost is made to approach the cost budget.