Lens optimization and matching assembly method based on part error database

Abstract

This invention belongs to the field of precision optical manufacturing technology, and particularly relates to a lens optimization and assembly method based on a component error database. First, key parameters of each lens are collected and assigned unique identifiers to build a component-level error coding database. Then, a system response model mapping the overall assembly error vector and the overall performance index vector is constructed, and a comprehensive evaluation objective function is built, with basic constraints on the binary decision variables for component allocation. Under these constraints, an optimization algorithm is used to solve for the optimal solution. Based on the optimized system response model, the optimal overall performance vector is obtained, and a picking list, assembly instructions, and assembly suggestions are generated and sent to the workstations to complete the assembly. After assembly, the measured performance vector of the entire machine is collected and written back to the database. A residual loss function is constructed by comparing the measured performance with the predicted performance. The model parameters are iteratively updated until the loss converges, forming a complete closed loop of detection, selection, assembly, and feedback iteration. This offsets component processing errors and improves the imaging quality and batch consistency of the finished product.

Description

Technical Field

[0001] This invention belongs to the field of precision optical manufacturing technology, and in particular relates to a lens optimization and assembly method based on a component error database. Background Technology

[0002] A complete imaging optical system typically consists of multiple lens elements, spacers, lens barrels, and mounting structures. During the design phase, tolerance analysis is generally used to determine the allowable machining error range for each lens element, such as front and rear surface curvature radius error, center thickness error, surface shape deviation, surface eccentricity, surface tilt, element center deviation, and edge thickness difference, to ensure that the system can still meet the target imaging quality after assembly.

[0003] However, in actual mass production, parts with high curvature, high asphericity, high refractive index, or high assembly sensitivity are difficult to process, and are prone to situations where measured parameters approach tolerance boundaries or even locally exceed tolerances. If the assembly stage still adopts the traditional method of random part selection or assembly based solely on a single size, multiple errors may accumulate in the same direction, leading to a significant deterioration in system wave aberrations, MTF, focus shift, distortion, image plane tilt, or assembly compensation, ultimately resulting in a decrease in yield and an increase in scrap costs.

[0004] Existing fitting technologies mostly focus on mechanical dimensional fits, such as hole-shaft fits, thickness grouping, and clearance control. For imaging optical systems, these methods do not establish a unified evaluation framework and cannot consider the impact of measured part errors transmitted through the optical system on the final image quality. Therefore, it is difficult to formulate batch assembly fitting schemes based on optimal optical performance and maximum system yield.

[0005] Therefore, there is an urgent need for a selection and assembly method that can integrate part-level measured errors, system-level optical system response models, and batch combination optimization solutions, so that the assembly process can be transformed from random selection based on experience to optimized assembly based on measured data, predictable models, and calculable decisions. Summary of the Invention

[0006] In view of this, the present invention aims to provide a lens optimization and assembly method based on a part error database. This addresses the shortcomings of existing technologies where assembly decisions are not directly based on measured part errors and overall imaging quality predictions. This leads to the easy accumulation of system errors under random assembly, unnecessary scrapping of potentially usable out-of-tolerance parts, and limitations on inventory utilization, overall yield, and product consistency. The present invention eliminates reliance on random part selection during assembly. Instead, it automatically solves for the optimal or near-optimal part matching scheme based on the measured machining errors of each part and their predicted impact on system imaging quality. This method can improve the imaging quality and batch yield of qualified assemblies without compromising system performance requirements, reduce overall scrapping caused by localized out-of-tolerance errors in individual parts, lower manufacturing costs, and form a data closed loop for subsequent process feedback and model updates.

[0007] To achieve the above objectives, the technical solution created by this invention is implemented as follows: A lens optimization and assembly method based on a parts error database includes the following steps: S1: Obtain the key parameters of each lens element in the whole machine, assign a unique identifier to each lens element, and build a part-level error coding database; S2: Construct the overall machine combination error vector and overall machine performance index vector based on the component-level error coding database; S3: Establish a system response model between the overall machine combination error vector and the overall machine performance index vector; construct a comprehensive evaluation objective function and set three types of basic constraints for the allocation decision variables of binary parts; S4: Using three types of basic constraints as conditions and minimizing the comprehensive evaluation objective function as the objective, the system response model is optimized using an optimization algorithm, and the optimal vector of overall system performance is obtained using the optimized system response model. S5: Based on the optimal performance vector of the whole machine, generate the assembly instruction sheet, parts picking list and assembly suggestion sheet of the whole machine, and distribute them to the assembly station according to the whole machine number to complete the assembly of the whole machine. S6: After completing the assembly of the whole machine, collect the measured performance vector of the whole machine and write it back to the part-level error coding database; construct the residual loss function based on the measured performance vector and the predicted performance index vector, and use the residual loss function to iteratively update the system response model parameters and optimization algorithm parameters until convergence. Repeat steps S4-S5 using the final system response model and the final optimization algorithm to complete the assembly of the whole machine.

[0008] Furthermore, in step S1, the key parameters of each lens element include at least: front surface radius of curvature, rear surface radius of curvature, center thickness, surface off-axis amount, surface tilt angle, lens center deviation, edge thickness difference, and surface shape parameters; The database storage fields of the part-level error coding database shall include at least the following: unique identifier of each lens element, lens position category, material grade, testing equipment, testing time and process level.

[0009] Furthermore, in step S2, the expression for the overall assembly error vector is: ; in, Let be the overall assembly error vector of the u-th unit. For the first When lens element # is installed in the first lens position of the u-th complete machine, the transpose of the machining error vector introduced by the lens element itself. For the first When lens element # is installed in the second lens position of the uth complete machine, the transpose of the machining error vector introduced by the lens element itself. For the first When lens element # is installed at the Nth mirror position on the uth complete machine, the transpose of the machining error vector introduced by the lens element itself. To be assigned to the The first complete machine The lens element number of the mirror position; The expression for the overall performance index vector is: ; in, Let be the vector of overall performance indicators for the u-th machine. This represents the change in RMS wave aberration. This represents the change in the fourth-order Zernike coefficient from the ideal design value. This represents the change in the 5th-order Zernike coefficient from the ideal design value. This represents the change in the m-th order Zernike coefficient from the ideal design value. For spatial frequency place Change For spatial frequency place Change This represents the change in the position of the focal plane. This represents the amount of distortion change. This represents the change in image plane tilt.

[0010] Furthermore, in step S3, the system response model includes at least one of a first-order linear response model, a second-order coupled response model, or an extended model with adjustment compensation variables, wherein: The expression for the first-order linear response model is: ; ; ; ; in, This is the overall system sensitivity matrix. For the first The sensitivity submatrix of the mirror position error parameters to the overall system performance vector. This is the general form of the overall system performance index vector. For the first The local error subvector corresponding to the mirror position Let the nominal system performance vector be... Let be the overall assembly error vector of the u-th unit. Let be the vector of overall performance indicators for the u-th machine. For the first When lens element # is installed in the first lens position of the u-th complete machine, the machining error vector introduced by the lens element itself. For the first When lens element # is installed in the second lens position of the u-th complete machine, the machining error vector introduced by the lens element itself. For the first When lens element # is installed at the Nth mirror position of the uth complete machine, the processing error vector caused by the lens element itself; The expression for the second-order coupled response model is: ; in, For the first The sensitivity submatrix of the mirror position error parameters to the overall system performance vector. For the first The quadratic nonlinear term of the mirror position's inherent error. For the first Mirror position and the first The quadratic term of cross-coupling of mirror positions, where N is the total number of mirror positions. For the first The lens element is installed on the first unit of the u-th complete machine. The machining error vector introduced by the lens element itself during mirror positioning; The expression for the extended model with adjustment compensation variables is: ; in, For the first The assembly and adjustment compensation variable vector of the entire machine. This is the influence matrix of each assembly and adjustment compensation variable on the overall machine performance index vector.

[0011] Furthermore, in step S3, the comprehensive evaluation objective function includes a single-machine objective function and a batch objective function, wherein: The expression for the single-machine objective function is: ; in, For the first The objective function of the entire machine. For the first The quality cost of the entire machine For the first Compensation for the entire machine For the first Inventory cost of the complete machine For the first The soft penalty item for the entire machine. , , and All are weighting coefficients; Batch objective function The expression is: ; ; in, For batch consistency items, The total losses incurred from unused lens elements, lens elements requiring rework, or lens elements ultimately scrapped. and All are weighting coefficients. Indicates the first The overall performance indicators of the machine are The variance among the complete systems, where M is the total number of performance indicators included in the consistency evaluation. This represents the k-th overall performance indicator of the first complete machine. This represents the k-th overall performance indicator of the second complete machine. For the first The kth overall performance indicator of the complete machine. is the consistency weight for the k-th overall system performance index.

[0012] Furthermore, the first The expression for the quality cost of the entire machine is: ; in, Let be the vector of overall performance indicators for the u-th machine. For the target performance vector, This is the performance weight matrix; No. The expression for the compensation cost term of the entire machine is: ; in, The engineering costs for different compensation actions, For the first The compensation variable vector of the entire machine; No. The expression for the inventory cost item of a complete machine is: ; in, For the uth complete machine in the first The first mirror position was selected. The inventory cost corresponding to lens element number N, where N is the total number of lens positions; No. The expression for the soft penalty term of the entire machine is: ; in, For the first Each engineering constraint function For the first The penalty weight for each project constraint. For the first The threshold corresponding to each engineering constraint function, where R is the total number of engineering constraints.

[0013] Furthermore, in step S3, three types of basic constraints are set by introducing binary part allocation decision variables: ; ; ; j= ; in, Assign decision variables to binary components. 'i' is the overall serial number, 'i' is the lens position number, and 'j' is the lens element number.

[0014] Furthermore, the optimization algorithm includes at least one of the following: genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm, tabu search algorithm, greedy swap algorithm, and Hungarian algorithm.

[0015] Furthermore, in step S6, the residual loss function for: ; ; in, For the first The residual vector of the entire machine. For residual vectors The corresponding weighting matrix, The regularization coefficient is . For the prior parameter set, For parameter set, For the first The measured vector of the overall performance of the complete machine. For the first The predictive performance index vector of the entire machine.

[0016] Furthermore, the expression used to iteratively update the system response model parameters and optimization algorithm parameters using the residual loss function is as follows: ; Where t represents the current iteration round, To update the step size.

[0017] Compared with the prior art, the present invention can achieve the following beneficial effects: (1) The lens optimization and matching assembly method based on the component error database created by the present invention upgrades the traditional method of judging whether the individual tolerance of a component is qualified to a system-level method based on the final imaging quality of the whole machine.

[0018] (2) The lens optimization and matching assembly method based on the component error database created by the present invention realizes the joint utilization and complementary pairing of multi-dimensional errors of components, and reduces the probability of system errors superimposed in the same direction.

[0019] (3) The lens optimization and assembly method based on the part error database described in this invention can still form an acceptable whole machine through intelligent selection and assembly when some parts are out of tolerance, thereby improving the inventory utilization rate.

[0020] (4) The lens optimization and assembly method based on the part error database described in this invention effectively improves the batch assembly yield, reduces the production costs of rework and scrap, and ensures the consistency of imaging performance of batch products.

[0021] (5) The lens optimization and assembly method based on the part error database described in this invention is compatible with existing tolerance analysis software, measuring equipment and manufacturing execution system, and has excellent engineering application value.

[0022] (6) The lens optimization and matching assembly method based on the part error database described in this invention can continuously iterate and correct the model based on the actual measured data of the whole machine after assembly, forming a closed-loop optimization system covering design, tolerance simulation, part manufacturing and whole machine inspection. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic flowchart of the lens optimization and assembly method based on a part error database as described in an embodiment of the present invention; Figure 2 The following are schematic diagrams comparing the imaging quality and yield of random assembly and optimized matching in the embodiments of the present invention: (a) is a schematic diagram comparing the imaging quality of random assembly and optimized matching, and (b) is a schematic diagram comparing the yield of random assembly and optimized matching. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0026] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0027] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0028] The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0029] like Figure 1 As shown, this invention proposes a lens optimization and assembly method based on a parts error database, which specifically includes the following steps: S1: Obtain the key parameters of each lens element in the whole machine, assign a unique identifier to each lens element, and build a part-level error coding database; S2: Construct the overall machine combination error vector and overall machine performance index vector based on the component-level error coding database; S3: Establish a system response model between the overall machine combination error vector and the overall machine performance index vector; construct a comprehensive evaluation objective function and set three types of basic constraints for the allocation decision variables of binary parts; S4: Using three types of basic constraints as conditions and minimizing the comprehensive evaluation objective function as the objective, the system response model is optimized using an optimization algorithm, and the optimal vector of overall system performance is obtained using the optimized system response model. S5: Based on the optimal performance vector of the whole machine, generate the assembly instruction sheet, parts picking list and assembly suggestion sheet of the whole machine, and distribute them to the assembly station according to the whole machine number to complete the assembly of the whole machine. S6: After completing the assembly of the whole machine, collect the measured performance vector of the whole machine and write it back to the part-level error coding database; construct the residual loss function based on the measured performance vector and the predicted performance index vector, and use the residual loss function to iteratively update the system response model parameters and optimization algorithm parameters until convergence. Repeat steps S4-S5 using the final system response model and the final optimization algorithm to complete the assembly of the whole machine.

[0030] It should be noted that, for batches of parts to be assembled, this invention measures key processing error parameters for each part individually. In this invention, "parts" refers to lens elements. A uniquely identified part-level error coding database is established. Based on the tolerance analysis results from the design phase of the imaging optical system, a system response model is established between part errors and overall system performance indicators. This system response model includes at least one of a first-order system response model, a second-order coupled response model, and an extended model with assembly compensation variables. The measured errors of the batch of parts are substituted into the system response model to predict the imaging quality, focus shift compensation, and pass / fail probability corresponding to different combinations. Furthermore, by optimizing the selection algorithm, the assembly scheme with the optimal overall quality, highest yield, or lowest cost is solved under constraints of multiple lens positions, multiple complete systems, and multiple batches of parts. Finally, assembly is performed according to the determined scheme, and the actual test results are written back to the database to continuously revise the system response model used for prediction.

[0031] In some embodiments, in step S1, the key parameters of each lens element include at least: front surface radius of curvature, rear surface radius of curvature, center thickness, surface off-axis amount, surface tilt angle, lens center deviation, edge thickness difference, and surface shape parameters. The database storage fields of the part-level error coding database shall include at least the following: unique identifier of each lens element, lens position category, material grade, testing equipment, testing time and process level.

[0032] It should be noted that key parameters are measured for each part to be assembled, and a part error database is established. Specifically, for each lens element, one or more key parameters are measured, including the front surface radius of curvature, rear surface radius of curvature, center thickness, surface translation error (surface off-axis amount), surface tilt error (surface tilt angle), lens element center deviation (lens center deviation), edge thickness difference, surface shape parameters, and refractive index batch difference. Each part is assigned a unique identifier, and the measured error information of each part, along with the part number, lens position category, material grade, testing equipment, testing time, and process level, are written into the database to form a part-level error coding database.

[0033] In some embodiments, in step S2, the expression for the overall assembly error vector is: ; in, Let be the overall assembly error vector of the u-th unit. For the first When lens element # is installed in the first lens position of the u-th complete machine, the transpose of the machining error vector introduced by the lens element itself. For the first When lens element # is installed in the second lens position of the uth complete machine, the transpose of the machining error vector introduced by the lens element itself. For the first When lens element # is installed at the Nth mirror position on the uth complete machine, the transpose of the machining error vector introduced by the lens element itself. To be assigned to the The first complete machine The lens element number of the mirror position; The expression for the overall performance index vector is: ; in, Let be the vector of overall performance indicators for the u-th machine. This represents the change in RMS wave aberration. This represents the change in the fourth-order Zernike coefficient from the ideal design value. This represents the change in the 5th-order Zernike coefficient from the ideal design value. This represents the change in the m-th order Zernike coefficient from the ideal design value. For spatial frequency place Change For spatial frequency place Change This represents the change in the position of the focal plane. This represents the amount of distortion change. This represents the change in image plane tilt.

[0034] It should be noted that, for the j-th lens element at position i, the measured error vector of a single part is calculated using the following formula: ; in, Let j be the j-th lens element at position i. This represents the relative design value deviation of the front surface radius of curvature. This represents the relative design value deviation of the radius of curvature of the rear surface. For the center thickness deviation, This represents the translational offset of the current vertex of the lens element's surface in the X-axis direction. This represents the translational offset of the current vertex of the lens element's surface along the Y-axis. This represents the tilt angle deviation of the current lens element surface around the X-axis. This represents the tilt angle deviation of the current lens element surface around the Y-axis. This is due to lens center deviation. The vector dimension can be expanded according to the actual measurement capability to account for the edge thickness difference.

[0035] Each actual part's unique serial number, part category, material grade, batch number, measured error vector, inspection time, inspection equipment, and pass / fail grade are written into a database to form a "part-level error coding database." This database can be a relational database, a file database, or an embedded database.

[0036]

[0037] The overall assembly error vector is concatenated. For the first... The machine is awaiting assembly. The selection results for each lens position are as follows: Then the overall assembly error vector can be written as: ; In the formula, Indicates assignment to the first The actual part number of the mirror position i of the complete machine. This reflects the error status of all parts of the machine under the current configuration scheme.

[0038] Overall system performance metric vector definition. For ease of unified prediction and evaluation, overall system performance can be represented as a joint performance vector: ; in, This represents the change in RMS wave aberration of the system. For several key Zernike coefficient changes, For a given spatial frequency place Change This represents the change in the position of the focal plane. This represents the amount of distortion change. This represents the change in image plane tilt.

[0039] The overall performance indicators include at least one or more of the following: RMS wave aberration, one or more Zernike coefficient deviations, one or more MTF changes at a specified spatial frequency, focal plane position changes, distortion changes, and image plane tilt changes; in other embodiments, field curvature, exit pupil position, illuminance uniformity, or transmittance may also be included.

[0040] It is the nominal system performance vector, which is... Being in the same performance metric space, for example: ; in, For the first The assembly and adjustment compensation variable vector of the entire camera, such as focusing amount, spacer ring thickness correction amount, axial compensation amount of a certain lens group, eccentricity compensation amount, tilt compensation amount, etc. This is the influence matrix of each assembly and adjustment compensation variable on the overall machine performance index vector. Therefore, This is a 5-dimensional performance correction vector, and... , and They have the same dimensions.

[0041] For example, suppose the overall performance index vector includes RMS wavelet aberration, fourth-order Zernike coefficient, fifth-order Zernike coefficient, MTF at a spatial frequency of 50 lp / mm, and focal plane position change, then it can be denoted as: ; If the first The error parameter subvector of the mirror position includes the deviation of the front surface radius of curvature, the deviation of the center thickness, and the eccentricity in the X direction, which can be denoted as: ; in, The deviation of the front surface radius of curvature corresponding to the i-th mirror position. The center thickness deviation corresponding to the i-th mirror position. Let X be the X-axis eccentricity corresponding to the i-th mirror position.

[0042] When the system includes 3 mirror positions, the first The overall assembly error vector of the entire machine can be written as:

[0043] No. The local sensitivity matrix of the mirror position is:

[0044] For example, when It is a 5-dimensional vector and When it is a 3-dimensional vector, It is a 5×3 matrix; the total system sensitivity matrix It is a 5×9 matrix.

[0045] In some embodiments, in step S3, the system response model includes at least one of a first-order linear response model, a second-order coupled response model, or an extended model with adjustment compensation variables, wherein: The expression for the first-order linear response model is: ; ; ; ; in, This is the overall system sensitivity matrix. For the first The sensitivity submatrix of the mirror position error parameters to the overall system performance vector. This is the general form of the overall system performance index vector. For the first The local error subvector corresponding to the mirror position Let the nominal system performance vector be... Let be the overall assembly error vector of the u-th unit. Let be the vector of overall performance indicators for the u-th machine. For the first When lens element # is installed in the first lens position of the u-th complete machine, the machining error vector introduced by the lens element itself. For the first When lens element # is installed in the second lens position of the u-th complete machine, the machining error vector introduced by the lens element itself. For the first When lens element # is installed at the Nth mirror position of the uth complete machine, the processing error vector caused by the lens element itself; The expression for the second-order coupled response model is: ; in, For the first The sensitivity submatrix of the mirror position error parameters to the overall system performance vector. For the first The quadratic nonlinear term of the mirror position's inherent error. For the first Mirror position and the first The quadratic term of cross-coupling of mirror positions, where N is the total number of mirror positions. For the first The lens element is installed on the first unit of the u-th complete machine. The machining error vector introduced by the lens element itself during mirror positioning; The expression for the extended model with adjustment compensation variables is: ; in, For the first The assembly and adjustment compensation variable vector of the entire machine. This is the influence matrix of each assembly and adjustment compensation variable on the overall machine performance index vector.

[0046] It should be noted that, based on the tolerance analysis results during the optical system design phase, a system response model of the system performance indicators to the error parameters of each component can be established.

[0047] When the error amplitude is small, the coupling between mirror positions is weak, and the system operating point is relatively stable, a first-order linear response model can be used: ; Alternatively, it can be written in blocks according to the camera position: ; in, Let the nominal system performance vector be... This is the overall system sensitivity matrix. For mirror position The corresponding local system response model. The key error parameters of mirror position i can be obtained by perturbing each item and then regressing them using design software.

[0048] When the optical system of the entire instrument is a high-NA, high-magnification, and high-sensitivity structure, or when there is significant coupling between multiple mirror position errors, a second-order coupled response model can be used: ; In the formula, The nonlinear quadratic effect characterizing the error at the same mirror position i. Characterization mirror position With mirror position The coupling effects between them. For example, the curvature deviation of the front group and the thickness deviation of the rear group may jointly affect spherical aberration and focal shift. Such effects cannot be fully described by a pure first-order model.

[0049] In actual production, a certain range of focusing, spacer ring compensation, or local axial position compensation is usually allowed after assembly. An extended model with assembly adjustment compensation variables can be used: ; in, For the first The assembly and adjustment compensation variable vector of the entire machine, such as focusing amount, spacer ring thickness correction amount, axial compensation amount of a certain lens group, etc. This is the influence matrix of assembly and adjustment compensation variables on the overall machine performance. This allows for a direct evaluation of the overall machine result after the combined effect of "part error + allowable compensation".

[0050] The system response model can be obtained through any of the following methods or a combination thereof: First, establish a first-order matrix based on the tolerance and sensitivity analysis results of design software such as CodeV and Zemax; Second, perform DOE scanning on key parameters and fit them using multinomial regression, response surface methodology, or Kriging surrogate model; Third, perform supervised learning regression based on historical assembly-inspection data; Fourth, use a first-order model for rapid initial screening, and then use a higher-order model for fine calculation on the shortlisted candidate combinations.

[0051] To ensure the usability of the project, the model needs to be validated using retained samples or historical assembly samples. The mean square error, maximum deviation, and consistency rate between the predicted and measured values ​​should be calculated. When the model error exceeds the set threshold, feature parameters should be reselected, coupling terms should be added, or weights should be corrected.

[0052] In some embodiments, in step S3, the comprehensive evaluation objective function includes a single-machine objective function and a batch objective function, wherein: The expression for the single-machine objective function is: ; in, For the first The objective function of the entire machine. For the first The quality cost of the entire machine For the first Compensation for the entire machine For the first Inventory cost of the complete machine For the first The soft penalty item for the entire machine. , , and All are weighting coefficients; Batch objective function The expression is: ; ; in, For batch consistency items, The total losses incurred from unused lens elements, lens elements requiring rework, or lens elements ultimately scrapped. and All are weighting coefficients. Indicates the first The overall performance indicators of the machine are The variance among the complete systems, where M is the total number of performance indicators included in the consistency evaluation. This represents the k-th overall performance indicator of the first complete machine. This represents the k-th overall performance indicator of the second complete machine. For the first The kth overall performance indicator of the complete machine. is the consistency weight for the k-th overall system performance index.

[0053] In some embodiments, the first The expression for the quality cost of the entire machine is: ; in, Let be the vector of overall performance indicators for the u-th machine. For the target performance vector, This is the performance weight matrix; No. The expression for the compensation cost term of the entire machine is: ; in, The engineering costs for different compensation actions, For the first The compensation variable vector of the entire machine; No. The expression for the inventory cost item of a complete machine is: ; in, For the uth complete machine in the first The first mirror position was selected. The inventory cost corresponding to lens element number N, where N is the total number of lens positions; No. The expression for the soft penalty term of the entire machine is: ; in, For the first Each engineering constraint function For the first The penalty weight for each project constraint. For the first The threshold corresponding to each engineering constraint function, where R is the total number of engineering constraints.

[0054] It should be noted that a unified evaluation of different combinations should be applied, unifying "better optics," "more economical manufacturing," and "more stable batch production" under a single evaluation criterion. The engineering objective function should not only focus on minimizing single-wavelength aberrations but also consider imaging quality, compensability, inventory utilization, process cycle time, and scrap risk. A comprehensive objective function can be constructed. Taking into account imaging quality, yield, inventory utilization, and assembly costs, it can be specifically subdivided into the following categories: (1) For the first For a complete machine, given the selected configuration and compensation variables, the quality cost term can be defined as: ; in, For the target performance vector, This is the performance weight matrix. If we are more concerned with lower-order aberrations or specific spatial frequencies... , can It is given higher weight.

[0055] (2) For schemes that require focusing or interval ring correction, the compensation cost term can be defined as: ; in, This reflects the engineering costs of different compensation actions. For example, the cost of simply adjusting the focus is relatively low, while the cost of replacing a spacer ring of a specific thickness or repeatedly disassembling and reassembling lens assemblies is relatively high.

[0056] (3) To avoid excessive and concentrated consumption of high-quality parts, or to achieve batch balance and first-in-first-out, an inventory cost item can be defined: ; in, It can represent the inventory priority, batch weight, near-expiration risk, cross-station handling cost, or key part number retention cost of parts.

[0057] (4) To avoid unstable selection schemes that are computationally feasible but whose indicators are close to the threshold, a soft penalty term is introduced: ; in, For the first Engineering constraint functions, such as RMS wave aberration, focus shift, and a certain frequency. Lower limit, image plane tilt, upper limit of compensation, etc.; The threshold value can transform hard constraint overflow into a large penalty, thereby enhancing solution stability.

[0058] (5) Construct the single-machine integrated objective function. The engineering evaluation objective of the u-th complete machine can be written as: ; in, , , , These are the weighting coefficients for each sub-item, and their values ​​can be jointly set by the process department, quality department, and production management department based on their needs.

[0059] (6) Construct the batch synthesis objective function, for The overall objective function for the batch selection and configuration of complete machines can be written as: ; in, This indicates a batch consistency item, used to suppress excessive performance variation between different complete machines; This refers to the combined losses resulting from unused parts, parts requiring rework, or parts that are ultimately scrapped.

[0060] (7) Preferred form of batch consistency term. To improve batch stability, it can be further defined as follows: ; in, Indicates the first Each performance metric The variance between complete units. This item applies to imaging systems requiring batch consistency, such as lens modules and inspection lenses.

[0061] In some embodiments, in step S3, three types of basic constraints are set by introducing binary part allocation decision variables: ; ; ; j= ; in, Assign decision variables to binary components. 'i' is the overall serial number, 'i' is the lens position number, and 'j' is the lens element number.

[0062] It should be noted that, for mass production scenarios, a binary part allocation decision variable can be introduced. When the first The first part was assigned to the... Lens position of the entire machine hour, Otherwise, it is 0. Therefore, the constraint can be written as: Type I constraint: ; Second type of constraint: ; Third type of constraint: ; The first type of constraint states that each mirror position in each machine is assigned only one part; the second type of constraint states that the same actual part cannot be reused; and the third type of constraint states that the variable is a binary variable. Depending on the specific form of the objective function, this problem can be constructed as a mixed-integer linear programming problem, a mixed-integer quadratic programming problem, or a heuristic combinatorial optimization problem.

[0063] In some embodiments, the optimization algorithm includes at least one of the following: genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm, tabu search algorithm, greedy swap algorithm, and Hungarian algorithm.

[0064] It should be noted that by substituting the batch part error parameters from the part-level error coding database into the aforementioned system response model and single-machine synthesis objective function, the batch assembly task can be expressed as a constrained combinatorial optimization problem. For the... For each mirror assembly unit, the candidate parts in the part-level error coding database are first pre-screened based on mirror type, part category, diameter, material grade, coating condition, batch compatibility, inventory status, structural interface, and process release conditions, forming a set of candidate parts for each mirror position. Parts that clearly do not meet assembly boundary conditions, pose assembly risks, or exceed the engineering allowable error envelope can be eliminated before solving. Parts whose local parameters exceed the single-part design tolerance, but whose system response model predicts can still meet the overall performance requirements by complementing other mirror parts, can be retained as controlled candidate parts for subsequent optimization.

[0065] When solving the model, the mirror positions of each machine can be determined. Should the first option be selected? Each candidate part is represented as a decision variable for assigning binary parts. And based on the aforementioned comprehensive objective function Single-machine objective function Or batch objective function As the solution objective, in addition to the unique allocation constraint given in the above formula, constraints such as batch assembly constraints, mirror assembly sequence constraints, inventory upper and lower limits constraints, parts traceability constraints, compensation mechanism stroke constraints, assembly tooling capability constraints, and overall system performance threshold constraints can be further superimposed. For systems with subsequent adjustable degrees of freedom, the image plane position compensation, mirror group interval adjustment, eccentricity compensation, and tilt compensation can also be solved as additional variables, thereby simultaneously outputting the joint result of "parts selection scheme + assembly and adjustment suggestion scheme".

[0066] When the problem to be solved is small in scale and the constraint structure is relatively clear, precise methods such as 0-1 integer programming, mixed integer linear programming, or mixed integer quadratic programming can be used to directly solve it and obtain the globally optimal result. When the number of mirror positions, candidate parts, or complete machines to be assembled is large, resulting in an exponential increase in the combinatorial space, genetic algorithms, particle swarm optimization, simulated annealing, tabu search, greedy swapping, Hungarian algorithm, or a combination of multiple algorithms can be used to solve the problem. Among them, a preferred implementation method is to first use rule filtering, cost sorting, or greedy matching to quickly generate an initial solution, and then further optimize it through genetic search, local swapping, neighborhood rearrangement, or simulated annealing to balance the solution speed and the quality of the solution.

[0067] In practical engineering execution, the optimization solution module can be linked with the production management system, barcode traceability system, or testing database. After the solution is completed, the system can output the part numbers that should be installed at each lens position for each complete machine, the key error parameters of the corresponding parts, the optimal performance vector of the whole machine, the expected compensation amount, the comprehensive evaluation score, and whether the solution meets the factory threshold. The optimal performance vector of the whole machine may include, but is not limited to, RMS wavelet aberration, Zernike coefficients of various orders, MTF, focal plane position, image plane tilt, distortion, transmittance, or other quality indicators related to the system's application.

[0068] Furthermore, assembly instructions, parts picking lists, and assembly suggestions can be automatically generated based on the optimal performance vector of the entire machine, and distributed to assembly stations according to the machine number or batch number. Assembly personnel or automated assembly equipment complete parts selection, mirror assembly, and initial compensation settings according to the assembly instructions; for combinations where the predicted results are close to the boundary, manual review or secondary simulation verification can be added before execution. Through the above methods, the assembly process can be transformed from random parts picking to a data-driven, predictable selection process, thereby reducing the failure risk caused by the accumulation of errors in the same direction, and improving inventory utilization, overall machine yield, and batch consistency.

[0069] In step S5, for example, for the first complete machine, the system outputs lens positions P1~P6 to assemble parts L1-03 (the third lens in a batch of lenses adapted to L1 lens), L2-11, L3-07, L4-22, L5-05, and L6-18 respectively, while providing the expected RMS wavelet aberration of 0.042λ, MTF50 decrease of 4.8%, and recommended focusing compensation of +0.03 mm. The assembly station completes parts picking, barcode scanning verification, and initial focusing settings according to this instruction. The assembly instruction sheet refers to the first complete machine, where the system output lens positions P1~P6 are respectively equipped with parts L1-03 (the third lens in a batch of lenses adapted to L1 lenses), L2-11, L3-07, L4-22, L5-05, and L6-18. The parts picking list is L1-03, L2-11, L3-07, L4-22, L5-05, and L6-18, and the assembly and adjustment suggestion table is a focus compensation amount of +0.03 mm.

[0070] In some embodiments, in step S6, the residual loss function for: ; ; in, For the first The residual vector of the entire machine. For residual vectors The corresponding weighting matrix, The regularization coefficient is . For the prior parameter set, For parameter set, For the first The measured vector of the overall performance of the complete machine. For the first The predictive performance index vector of the entire machine.

[0071] In some embodiments, the expression used to iteratively update the system response model parameters and optimization algorithm parameters using the residual loss function is as follows: ; Where t represents the current iteration round.

[0072] It should be noted that after assembly, the entire machine should undergo MTF, wavefront, focal plane position, aberration balance, distortion, image plane tilt, or other performance tests. The test results, along with the corresponding part number, lens position allocation relationship, batch information, assembly compensation amount, testing environment, testing equipment, and testing time, should be written back to the part-level error coding database, forming a full-link traceability record of "part measured parameters - predicted assembly scheme - actual assembly result - overall machine measured performance". Based on this written-back data, a residual sample set between predicted and measured values ​​can be established, which can be used to correct the first-order linear response model, the second-order coupled response model, or the extended model with assembly compensation variables.

[0073] Preferably, the performance residual vector of the u-th complete machine can be defined. , in, The predicted overall machine performance vector is calculated by the system response model from the part error vector and compensation variables in the part-level error coding database. This is the overall performance vector obtained from actual measurements after assembly.

[0074] Furthermore, a residual loss function can be constructed. ,in This represents the set of parameters of the model to be identified. This represents a weighted matrix for different performance metrics. This refers to prior parameters obtained during the design phase or in historical batches. This is the regularization coefficient. By minimizing this residual loss function, the system response model can gradually approximate the error propagation law in the actual manufacturing and assembly process.

[0075] System response model updates can employ supervised learning for offline batch training or parameter identification for online or quasi-online updates. For production stages with a small sample size or relatively stable system structure, batch-wise parameter regression correction can be prioritized. For production stages with large batch sizes, significant changes in process conditions, or drifting part distributions, newly added assembly samples can be periodically used to adjust the local sensitivity matrix. Second-order coupling terms The compensation mapping matrix and bias terms are jointly updated. If necessary, sample confidence, outlier removal, cross-validation, or validation set retention mechanisms can be introduced to avoid excessive perturbation of the model by individual outliers.

[0076] ; in, Represents the set of model parameters. Represents the residual loss function. To update the step size, when using gradient descent, stochastic gradient descent, Adam, or other iterative optimization methods, the model parameters can be updated according to the above formula. During engineering implementation, update trigger conditions can also be set, such as initiating model correction when the cumulative number of assembly samples reaches a preset threshold, the mean prediction error exceeds the allowable range, or a key performance indicator shows systematic drift.

[0077] It should be noted that the aforementioned model update and database write-back mechanism is not the only necessary condition for the implementation of this invention; even without performing online model updates, this invention can still complete part selection and assembly decisions based solely on the system response model established during the design phase. However, in the preferred embodiment, by continuously writing back the assembly results and iteratively correcting the model parameters, the prediction accuracy, batch consistency, and process robustness of subsequent batches can be further improved, thereby forming a closed-loop optimization system from design prediction, manufacturing inspection, assembly execution to actual measurement feedback.

[0078] Example 1: Six-Mirror Batch Selection Based on First-Order Linear Response Model Suppose an imaging system contains 6 lens positions. The plan is to assemble 20 complete sets, with 30 candidate parts for each mirror position. Four key parameters will be measured for each part: front surface radius of curvature error, rear surface radius of curvature error, center thickness error, and component eccentricity.

[0079] Step 1: Establish a critical error parameter table. The process department confirms this based on the tolerance decomposition results. right , , , The four parameters are the most sensitive, so they are used as modeling features. All parameters are uniformly converted to units consistent with the design model, and their symbols are defined based on the principal optical axis coordinate system of the mirror position.

[0080] Step 2: Establish a part-level error coding database and candidate set. Number each of the 180 actual parts as follows: The batch, measured parameters, measurement uncertainty, and available lens positions are recorded. Since the parts for each lens position are not interchangeable, the candidate set for each lens position comes only from the corresponding lens position hopper.

[0081] Step 3: Establish a system response model. Design software to obtain the RMS wavelet aberrations and 50° for each mirror position offline. The first-order linear response model of the MTF on the axis is established, and then the first-order linear response model is established: .in, At least include and Two components.

[0082] Step 4: Construct the evaluation objective function. Prioritize the quality cost term, followed by the compensation cost term, and then the inventory cost term. Since a small range of focusing is allowed during the assembly and adjustment stage, the focusing amount is included. And limit its absolute value to not exceed a set threshold.

[0083] Step 5: Execute the optimization solution and output assembly instructions. Simultaneously solve the problem for 20 complete machines, with constraints that each mirror position of each complete machine uses exactly one part, each part is used at most once, and the predicted MTF is not lower than the lower limit. First, use a first-order linear response model to quickly find candidate combinations, and then verify the top few optimal solutions.

[0084] The system outputs a list of lens positions and parts for each complete unit. For example, the first complete unit uses... , , , , , It also provides the estimated RMS wave aberration, estimated MTF, and recommended focusing amount.

[0085] Step 6: Test and write back. After assembly, test the write-back of all 20 units. The focal plane position is detected, and the measured values ​​are compared with the predicted values. If a systematic deviation of the model for a certain lens position is found to be large, the system response model (i.e., the first-order linear response model) for that lens position is updated.

[0086] Compared with random assembly, the predicted pass rate of the present invention is increased from about 65% to about 88%, while the utilization rate of parts is significantly improved, indicating that the optimized selection based on the first-order linear response model can achieve significant benefits in medium-complexity systems.

[0087] Example 2: Fine-tuning based on second-order coupled response model and discrete compensation loop When the mirror-position coupling is significant and the compensation variables are discretized, suppose a high-sensitivity imaging system contains 8 mirror positions and plans to assemble 10 complete units. In addition to the conventional errors, axial compensation is also allowed by replacing the three-stage discrete interval ring.

[0088] "Three-stage discrete spacer rings" refers to spacer rings with three fixed thicknesses pre-machined for a specific lens assembly. By selecting one stage, the axial spacing between adjacent lenses or lens assemblies is changed, thereby compensating for assembly errors such as focus shift and spherical aberration. For example, assuming a nominal thickness of 5.000 mm and a stage spacing of 0.020 mm, the three stages could be: ; Here, "discrete" means that the compensation amount cannot be continuously and arbitrarily adjusted; it can only be selected from three preset thicknesses. "Three levels" does not refer to installing three spacer rings simultaneously, but rather to having three selectable specifications. The actual level interval should be determined based on the system's axial sensitivity, machining capabilities, and allowable residual aberrations. In actual use, the actual level interval should be determined based on the system's axial sensitivity, machining capabilities, and allowable residual aberrations.

[0089] Step 1: In addition to measuring curvature, thickness, and eccentricity, the key aspheric coefficient deviation and refractive index batch variation are also measured. All data are temperature normalized, and parts with measurement repeatability below the threshold are discarded.

[0090] Step 2: In addition to the basic measured items, the part-level error coding database also records the range of interval ring gears that the part can match, the batch priority, and whether it is recommended to consume it first.

[0091] Step 3: Establish a second-order coupled response model using DOE scanning. Combine perturbations are applied to key mirror position errors, and the model is fitted to obtain... .in, The variable is a discrete compensation variable, representing the specific choice among the three-level interval loops.

[0092] Step 4: In addition to the quality cost term, the objective function is further enhanced by adding a compensation cost term and a soft penalty term, especially imposing a greater penalty on schemes that are "qualified but whose compensation amount is close to the upper limit". This can avoid generating unstable boundary assembly schemes in mass production.

[0093] Step 5: First, filter out obviously suboptimal parts at each mirror position, and then use mixed integer quadratic programming or heuristic algorithms to jointly solve the part allocation and spacing cycle selection.

[0094] In addition to the lens part number, the output results also include the spacer ring position to be used for each complete machine, the theoretical residual spherical aberration, and the focal shift margin.

[0095] Step 6: After actual assembly, perform wavefront testing and focal plane retesting, and then... The coefficients are fed back into the second-order coupled response model for correction. .

[0096] This indicates that when the system has significant coupling and discrete compensation structures, the present invention can not only complete the selection and matching, but also simultaneously optimize the component matching and compensation scheme, thereby further improving the assembly consistency of high-performance systems.

[0097] Example 3: Multi-objective matching and closed-loop update for mass production lines Suppose a certain lens testing production line needs to assemble 100 complete machines per batch. The factory is concerned not only with the first pass rate, but also with the balanced consumption of inventory, assembly cycle time and batch consistency.

[0098] Step 1: Production line selection based on historical data , , , Four types of high-frequency measurement parameters serve as basic features, while surface shape and refractive index fields are retained for subsequent model expansion.

[0099] Step 2: Establish FIFO (First-In, First-Out) tags, batch age tags, and key part number retention tags in the database, so that the inventory strategy can be written into the evaluation function.

[0100] Step 3: The response model employs a hierarchical screening structure. The first layer uses a first-order linear response model for rapid prediction, scoring, and initial screening of candidate solutions in a large-scale combination space. The second layer uses a second-order coupled response model for high-precision performance prediction of the small number of candidate solutions obtained from the first layer, and for final ranking of the candidate solutions based on the prediction results. The first-order linear response model and the second-order coupled response model are not weighted superpositions, but rather are the initial screening response model and the recalculated response model called separately at different calculation stages.

[0101] Step 4: Batch objective function adopted ,in Used to reduce performance discrepancies among 100 complete machines. Used to suppress low utilization and obsolescence.

[0102] Step 5: The solver first satisfies all performance thresholds, and then minimizes the overall batch objective within the feasible region that satisfies the constraints. For the very few parts that cannot meet the requirements, the system automatically marks them as requiring retesting, rework, or degraded use.

[0103] The assembly workstation automatically calls for materials based on the list generated by the system, and displays the risk level of the whole machine, suggested compensation actions, and whether enhanced final inspection is required.

[0104] Step 6: After each batch is completed, write the final inspection results of the whole machine back to the database and update the model weights using parameter identification. If a significant deviation is found in the label of a certain batch in multiple consecutive batches, reverse process correction is performed on the upstream processing steps.

[0105] This indicates that the present invention is not only applicable to optimal assembly of a single batch, but can also be embedded in mass production manufacturing execution systems to form a continuous optimization closed loop across batches.

[0106] Simulation results show that, under the same batch conditions for parts, compared with random assembly or single-parameter graded assembly, the present invention can significantly reduce the overall performance dispersion, increase the proportion of qualified assemblies, and reduce unnecessary scrap due to local deviations. In systems with focusing or interval compensation capabilities, the present invention can also jointly optimize part matching and compensation strategies, further improving mass production yield and consistency.

[0107] As shown in Figure 2, under the same batch of parts and the same overall performance evaluation conditions, the traditional random assembly method is compared with the optimized selection method of the present invention. Figure 2 (a) shows the comparison results of key performance indicators under the two assembly methods, including RMS wave aberration, MTF drop at a specified spatial frequency, and focal shift compensation. Figure 2 (b) shows a comparison of quality and disposal results under two assembly methods, including overall machine pass rate, rework rate, and scrap rate. Figure 2 As can be seen from (a) and (b) in the present invention, compared with random assembly, the present invention uses a part-level error database, a system response model and a combination optimization solution to complement and match the errors of parts at different mirror positions, and combines the allowable assembly and adjustment compensation amount for comprehensive evaluation. This can reduce the degree of degradation of the key performance indicators of the whole machine, improve the overall machine qualification rate, and at the same time reduce the rework rate and scrap rate.

[0108] In addition, performance indicators are not limited to wavefronts. or It can also be distortion, field curvature, focus shift, exit pupil position, image plane tilt, illumination uniformity, etc.

[0109] The system response model is not limited to linear matrix form; it can also be a higher-order polynomial, response surface, lookup table, machine learning model, or digital twin model.

[0110] The optimization algorithm is not limited to integer programming; it can also be solved using genetic algorithms, particle swarm optimization, simulated annealing, tabu search, ant colony optimization, greedy swapping, Hungarian algorithm, or a combination of multiple algorithms.

[0111] Part error parameters are not limited to curvature, thickness, eccentricity, and tilt; they can also be extended to refractive index batch differences, Abbe number deviations, surface parameters, aspheric coefficient deviations, coating characteristics, etc.

[0112] The part-level error coding database can be established according to single lenses, cemented assemblies, lens modules, or sub-assemblies with mechanical references.

[0113] In addition to a parts matching list, the output results may also include the thickness of the matching compensation shims, the focusing position, the assembly and adjustment direction, or suggestions for subsequent processing corrections.

[0114] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0115] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A lens optimization and assembly method based on a parts error database, characterized in that: Specifically, the steps include the following: S1: Obtain the key parameters of each lens element in the whole machine, assign a unique identifier to each lens element, and build a part-level error coding database; S2: Construct the overall machine combination error vector and overall machine performance index vector based on the component-level error coding database; S3: Establish a system response model between the overall machine combination error vector and the overall machine performance index vector; construct a comprehensive evaluation objective function and set three types of basic constraints for the allocation decision variables of binary parts; S4: Using three types of basic constraints as conditions and minimizing the comprehensive evaluation objective function as the objective, the system response model is optimized using an optimization algorithm, and the optimal vector of overall system performance is obtained using the optimized system response model. S5: Based on the optimal performance vector of the whole machine, generate the assembly instruction sheet, parts picking list and assembly suggestion sheet of the whole machine, and distribute them to the assembly station according to the whole machine number to complete the assembly of the whole machine. S6: After completing the assembly of the whole machine, collect the measured performance vector of the whole machine and write it back to the part-level error coding database; construct the residual loss function based on the measured performance vector and the predicted performance index vector, and use the residual loss function to iteratively update the system response model parameters and optimization algorithm parameters until convergence. Repeat steps S4-S5 using the final system response model and the final optimization algorithm to complete the assembly of the whole machine.

2. The lens optimization and assembly method based on a part error database according to claim 1, characterized in that: In step S1, the key parameters of each lens element include at least: front surface radius of curvature, rear surface radius of curvature, center thickness, surface off-axis amount, surface tilt angle, lens center deviation, edge thickness difference, and surface shape parameters; The database storage fields of the part-level error coding database shall include at least the following: unique identifier of each lens element, lens position category, material grade, testing equipment, testing time and process level.

3. The lens optimization and assembly method based on a part error database according to claim 1, characterized in that: In step S2, the expression for the overall assembly error vector is: ； in, Let be the overall assembly error vector of the u-th unit. For the first When lens element # is installed in the first lens position of the u-th complete machine, the transpose of the machining error vector introduced by the lens element itself. For the first When lens element # is installed in the second lens position of the uth complete machine, the transpose of the machining error vector introduced by the lens element itself. For the first When lens element # is installed at the Nth mirror position on the uth complete machine, the transpose of the machining error vector introduced by the lens element itself. To be assigned to the The first complete machine The lens element number of the mirror position; The expression for the overall performance index vector is: ； in, Let be the vector of overall performance indicators for the u-th machine. This represents the change in RMS wave aberration. This represents the change in the fourth-order Zernike coefficient from the ideal design value. This represents the change in the 5th-order Zernike coefficient from the ideal design value. This represents the change in the m-th order Zernike coefficient from the ideal design value. For spatial frequency place Change For spatial frequency place Change This represents the change in the position of the focal plane. This represents the amount of distortion change. This represents the change in image plane tilt.

4. The lens optimization and assembly method based on a part error database according to claim 3, characterized in that: In step S3, the system response model includes at least one of a first-order linear response model, a second-order coupled response model, or an extended model with adjustment compensation variables, wherein: The expression for the first-order linear response model is: ； ； ； ； in, This is the overall system sensitivity matrix. For the first The sensitivity submatrix of the mirror position error parameters to the overall system performance vector. This is the general form of the overall system performance index vector. For the first The local error subvector corresponding to the mirror position Let the nominal system performance vector be... Let be the overall assembly error vector of the u-th unit. Let be the vector of overall performance indicators for the u-th machine. For the first When lens element # is installed in the first lens position of the u-th complete machine, the machining error vector introduced by the lens element itself. For the first When lens element # is installed in the second lens position of the u-th complete machine, the machining error vector introduced by the lens element itself. For the first When lens element # is installed at the Nth mirror position of the uth complete machine, the processing error vector caused by the lens element itself; The expression for the second-order coupled response model is: ； in, For the first The sensitivity submatrix of the mirror position error parameters to the overall system performance vector. For the first The quadratic nonlinear term of the mirror position's inherent error. For the first Mirror position and the first The quadratic term of cross-coupling of mirror positions, where N is the total number of mirror positions. For the first The lens element is installed on the first unit of the u-th complete machine. The machining error vector introduced by the lens element itself during mirror positioning; The expression for the extended model with adjustment compensation variables is: ； in, For the first The assembly and adjustment compensation variable vector of the entire machine. This is the influence matrix of each assembly and adjustment compensation variable on the overall machine performance index vector.

5. The lens optimization and assembly method based on a part error database according to claim 1, characterized in that: In step S3, the comprehensive evaluation objective function includes a single-machine objective function and a batch objective function, wherein: The expression for the single-machine objective function is: ； in, For the first The objective function of the entire machine. For the first The quality cost of the entire machine For the first Compensation for the entire machine For the first Inventory cost of the complete machine For the first The soft penalty item for the entire machine. , , and All are weighting coefficients; Batch objective function The expression is: ； ； in, For batch consistency items, The total losses incurred from unused lens elements, lens elements requiring rework, or lens elements ultimately scrapped. and All are weighting coefficients. Indicates the first The overall performance indicators of the machine are The variance among the complete systems, where M is the total number of performance indicators included in the consistency evaluation. This represents the k-th overall performance indicator of the first complete machine. This represents the k-th overall performance indicator of the second complete machine. For the first The kth overall performance indicator of the complete machine. is the consistency weight for the k-th overall system performance index.

6. The lens optimization and assembly method based on a part error database according to claim 5, characterized in that: No. The expression for the quality cost of the entire machine is: ； in, Let be the vector of overall performance indicators for the u-th machine. For the target performance vector, This is the performance weight matrix; No. The expression for the compensation cost term of the entire machine is: ； in, The engineering costs for different compensation actions, For the first The compensation variable vector of the entire machine; No. The expression for the inventory cost item of a complete machine is: ； in, For the uth complete machine in the first The first mirror position was selected. The inventory cost corresponding to lens element number N, where N is the total number of lens positions; No. The expression for the soft penalty term of the entire machine is: ； in, For the first Each engineering constraint function For the first The penalty weight for each project constraint. For the first The threshold corresponding to each engineering constraint function, where R is the total number of engineering constraints.

7. The lens optimization and assembly method based on a part error database according to claim 1, characterized in that: In step S3, three types of basic constraints are set by introducing binary part allocation decision variables: ； ； ； j= ； in, Assign decision variables to binary components. 'i' is the overall serial number, 'i' is the lens position number, and 'j' is the lens element number.

8. The lens optimization and assembly method based on a part error database according to claim 1, characterized in that: The optimization algorithm includes at least one of the following: genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm, tabu search algorithm, greedy swap algorithm, and Hungarian algorithm.

9. The lens optimization and assembly method based on a part error database according to claim 1, characterized in that: In step S6, the residual loss function for: ； ； in, For the first The residual vector of the entire machine. For residual vectors The corresponding weighting matrix, The regularization coefficient is . For the prior parameter set, For parameter set, For the first The measured vector of the overall performance of the complete machine. For the first The predictive performance index vector of the entire machine.

10. The lens optimization and assembly method based on a part error database according to claim 9, characterized in that: The expression used to iteratively update the system response model parameters and optimization algorithm parameters using the residual loss function is as follows: ； Where t represents the current iteration round, To update the step size.

Images