Design method and lens of N-in-one free-form surface street lamp lens for easy operation and maintenance

Abstract

The present invention relates to the technical field of optical lenses, and discloses an N-in-one free-form surface street lamp lens design method and lens that are easy to operate and maintain. The method comprises determining the lighting demand of a target lighting scene and generating first lighting demand data; acquiring historical lens design parameters and historical lighting effect data of multiple LED street lamp lighting projects and generating second historical optimization data; based on the second historical optimization data, establishing a mapping model between free-form surface parameters and lighting effects through a machine learning algorithm; generating an initial free-form surface and performing regional division and optimization, and finally obtaining an optimized lens solution; the lens comprises three sub-areas: a central area, a transition area and an edge area; generating a final design solution through simulation verification and test optimization; the present invention realizes efficient and accurate LED street lamp lens design, improves lighting uniformity and visual comfort, adopts planar light output, and is convenient for subsequent operation and maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of optical lenses, and more specifically, to a design method and a lens for an N-in-one free-form surface street lamp lens that is easy to operate and maintain. Background Art

[0002] As the core optical component of LED street lamps, lenses directly affect the lighting performance of street lamps. How to design efficient, uniform, and comfortable LED street lamp lenses that are convenient for subsequent operation and maintenance is an urgent problem to be solved.

[0003] At present, there are many methods for designing lenses for LED street lamps. The Chinese patent application with publication number CN113552717A provides a lens design method, a lens and a laser radar, which can obtain lens parameters that optimize the optical performance of aspheric lenses. However, this method is mainly aimed at the field of laser radar, and does not consider the special needs of LED street lamp lighting, such as illumination uniformity, glare limitation, etc., and does not involve the maintainability design of the lens. The Chinese patent application with publication number CN117369122A proposes a meta-lens and a design method thereof. By performing near-field coupling analysis and parameter optimization on the meta-unit, a meta-lens with large size, high focusing efficiency and high focal spot quality can be designed. However, this method mainly focuses on the focusing performance of the meta-lens, does not consider the light distribution curve requirements of LED street lamp lighting, and does not involve the maintainability design of the lens structure. The Chinese patent application with publication number CN105276521A proposes a design method for a polarized lens, a polarized lens, and an LED street lamp, which directly form an ideal light intensity distribution through refraction of the incident surface, reduce the total reflection loss of the exit surface, and improve the utilization efficiency of the LED light source. However, the optimization goal of the lens is mainly focused on reducing the internal total reflection and improving the light output efficiency, and insufficient consideration is given to the lighting uniformity, visual comfort, etc., and the maintainability design of the lens is not involved.

[0004] In summary, the existing LED street light lens design methods mostly focus on a single optical performance indicator, such as light output efficiency, focusing effect, etc., lack of comprehensive consideration of key performance such as lighting uniformity and visual comfort, and ignore the maintainability design of the lens structure. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides an N-in-one free-form surface street lamp lens design method and lens that are easy to operate and maintain. A mapping model between free-form surface parameters and lighting effects is established through a machine learning algorithm, and the lens surface is divided and optimized for area division. While improving the lighting uniformity and visual comfort, a planar light-emitting structural design is adopted, which greatly simplifies the operation and maintenance operations such as cleaning and replacement of the lens, and reduces the difficulty and cost of operation and maintenance.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The N-in-one free-form surface street lamp lens design method that is easy to operate and maintain includes:

[0008] Determine the lighting demand of the target lighting scene and generate first lighting demand data; obtain historical lens design parameters and historical lighting effect data of multiple LED street lamp lighting projects and generate second historical optimization data; based on the second historical optimization data, establish a second mapping model M2 between free-form surface parameters and lighting effects through a machine learning algorithm; generate an initial free-form surface based on the second mapping model and the first lighting demand data;

[0009] The initial free-form surface is divided into three sub-areas: a central area, a transition area, and an edge area. The actual light-emitting angle distribution function of each sub-area is calculated, and the actual light-emitting angle distribution function is compared with the preset ideal light-emitting angle distribution function to obtain the angle deviation data of each sub-area. According to the angle deviation data, the coordinates and weights of the free-form surface control points of each sub-area are adjusted to continuously update the shape of the free-form surface to obtain an updated free-form surface, and the second angle deviation data of each sub-area of ​​the updated free-form surface is calculated until the second angle deviation data of all sub-areas are lower than the set angle deviation convergence threshold , obtaining a second optimized free-form surface; based on the second optimized free-form surface, introducing material screening, obtaining a third optimized lens solution;

[0010] The third optimized lens solution is simulated, verified and tested to generate the final N-in-one planar light-emitting LED street lamp lens design solution.

[0011] Further, the generating the first lighting requirement data includes:

[0012] Determine the illumination uniformity requirement U0, the glare limitation level requirement G0, the brightness longitudinal uniformity requirement UI and the brightness uniformity requirement L0 of the target lighting scene; use U0, G0, UI and L0 as lighting constraint parameters to generate corresponding first lighting requirement data D1.

[0013] Further, the generating the second historical optimization data includes:

[0014] Get the historical lens design parameters of LED street lighting projects;

[0015] Acquire historical lighting effect data of the LED street lamp lighting project, wherein the historical lighting effect data includes historical illumination uniformity, historical glare limit level, historical brightness longitudinal uniformity and historical brightness uniformity; form a data pair with historical lens design parameters and historical lighting effect data;

[0016] The first similarity between the historical lighting effect data and the first lighting demand data D1 in the data pair is calculated, and the data pairs whose first similarity is higher than a preset similarity threshold are extracted to generate the second historical optimization data D2.

[0017] Further, generating an initial free-form surface comprises:

[0018] Input the first lighting requirement data D1 into the second mapping model M2 to generate a first free-form surface parameter set C1, wherein the first free-form surface parameter set C1 includes the control point coordinates and weights of the free-form surface;

[0019] Generate a first B-spline free-form surface S1 based on the control point coordinates and weights of the free-form surface in the first free-form surface parameter set C1;

[0020] The first B-spline free-form surface S1 is used as the initial free-form surface.

[0021] Furthermore, dividing the initial free-form surface into three sub-areas, namely, a central area, a transition area and an edge area, comprises:

[0022] Taking the center of the lens as the origin, a three-dimensional rectangular coordinate system is established, which is marked as the free-form surface coordinate system. The function expression of the initial free-form surface in the free-form surface coordinate system is: ; is the radius of the sphere, is the zenith angle, is the azimuth angle; z' is the function expression of the initial free-form surface in the free-form surface coordinate system;

[0023] use represents the boundary of the central area, which will satisfy The surface is divided into the central area; ρ1 represents the spherical radius of the central area boundary, θ1 represents the zenith angle of the central area boundary, and φ1 represents the azimuth angle of the central area boundary;

[0024] use represents the boundary of the transition zone, which will satisfy The surface is divided into a transition zone; ρ2 represents the spherical radius of the transition zone boundary, θ2 represents the zenith angle of the transition zone boundary, and φ2 represents the azimuth angle of the transition zone boundary;

[0025] use represents the boundary of the marginal area, which will satisfy The surface is divided into edge areas; ρ3 represents the spherical radius of the edge area boundary, θ3 represents the zenith angle of the edge area boundary, and φ3 represents the azimuth angle of the edge area boundary.

[0026] Further, obtaining the angle deviation data of each sub-area includes:

[0027] Randomly generate N2 rays, marked as N2 incident rays, and inject the incident rays into the initial free-form surface; each incident ray is represented by the parametric equation of a spatial line, with its starting point being a random sampling point on the initial free-form surface and its direction being a randomly sampled unit vector on the unit sphere, marked as the direction vector of the incident ray; N2 is a positive integer;

[0028] For each incident light ray, determine the incident light ray I j Is the incident angle less than the critical angle of total reflection? j The incident angle is less than the critical angle of total reflection, then according to the incident light I j The direction vector of the incident light ray I is solved using Snell's law. j The direction vector of the corresponding refracted light; I j represents the jth incident light ray; 1≤j≤N2;

[0029] According to the direction vector of the refracted light, determine whether the refracted light can be emitted through the lens; if it can be emitted, record the emission angle ψ of the refracted light;

[0030] The emission angle ψ of all refracted light in each sub-region is counted, and the probability distribution histogram of the emission angle ψ is drawn. The probability distribution histogram is normalized to obtain the actual light emission angle distribution function g(ψ) of each sub-region. i ; g(ψ) i represents the actual light output angle distribution function of sub-region i, i is the index of the sub-region, 1≤i≤3;

[0031] The actual light output angle distribution function g(ψ) of sub-region i is i The ideal light output angle distribution function f(ψ) corresponding to sub-area i i Compare and calculate the angle deviation data of sub-area i .

[0032] Further, obtaining the second optimized free-form surface includes:

[0033] Step S2310, set the number of initialization iterations , set the angle deviation convergence threshold , let the second optimized free-form surface ,in is the initial free-form surface, represents the free surface obtained at the 0th iteration;

[0034] Step S2320, make , according to The free-form surface obtained by iteration , combined with the angle deviation data of each sub-region, calculate the coordinate offset and weight change of the control points in each sub-region;

[0035] Step S2330: using the coordinate offset and weight change to calculate the free-form surface The coordinates and weights of each control point are corrected to obtain the updated free-form surface ;

[0036] Step S2340, calculating the updated free-form surface The second angle deviation data of each sub-area , Represents the second angle deviation data of sub-area i; determines whether ; If satisfied, let the second optimized free-form surface , the iteration ends; otherwise, return to step S2320 and continue the iterative optimization.

[0037] Furthermore, the calculation of the coordinate offset and weight change of the control point in each sub-area includes:

[0038] Divide all sub-regions into n grid units, and take the area of ​​sub-region i through n grid units As the weight, calculate the weighted average of the angle deviation of each sub-area ; Indicates The weighted average of the angular deviations of the sub-areas;

[0039] according to Calculate the The shape adjustment coefficient of the sub-region ;

[0040] Adjust coefficients according to shape , calculate the control points The coordinate offset of and weight change ; is the first Line The control points of the column.

[0041] A lens is designed using the above-mentioned N-in-one free-form surface street lamp lens design method that is easy to operate and maintain; the lens includes three sub-areas: a central area, a transition area, and an edge area; and the light-emitting surface of the lens is a plane.

[0042] An electronic device includes a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the above-mentioned N-in-one free-form surface street lamp lens design method that is easy to operate and maintain is implemented.

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

[0044] The present invention realizes planar light emission through free-form surface design, which is different from the traditional "peanut-grain" beaded free-form surface lens. Planar light emission reduces the impact of pollutants on the optical performance of the lens, reduces light loss caused by pollution, and facilitates operation and maintenance. The surface of traditional curved lenses is prone to accumulation of pollutants such as dust and oil, which is difficult to clean and seriously affects the transmittance. The planar design of the present invention makes it difficult for dust to be adsorbed, and even if there are pollutants, it is easy to clean, which greatly reduces the cost and frequency of operation and maintenance.

[0045] The present invention uses historical optimization data and machine learning algorithms to construct a mapping model between free-form surface parameters and lighting effects, which greatly reduces the number of design iterations and shortens the design cycle. Machine learning-assisted design can quickly obtain high-quality initial solutions that meet lighting requirements, laying a good foundation for subsequent optimization. The present invention subdivides the free-form surface into three sub-areas: the central area, the transition area, and the edge area for independent optimization, and can perform more precise control of the light output in different fields of view. By comparing the deviation between the actual light output angle distribution function of each sub-area and the ideal distribution function, the deviation data is used to guide the adjustment of the control point coordinates and weights, and the ideal light output effect is continuously approached. This regional optimization strategy ensures the lighting quality of the entire field of view. After obtaining the optimized free-form surface, the present invention also introduces a material screening link. By selecting suitable lens materials (such as refractive index, transmittance, etc.), the optical performance of the lens can be further improved, and the light output efficiency, dispersion control, etc. can be improved. The introduction of material factors makes the design results more comprehensive and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 This is a principle flow chart of the N-in-one free-form surface street lamp lens design method that is easy to operate and maintain in the present invention;

[0048] Figure 2 A flow chart of a method for generating first lighting demand data in the N-in-one free-form surface street lamp lens design method for facilitating operation and maintenance of the present invention;

[0049] Figure 3 A flow chart of a method for generating second historical optimization data in the N-in-one free-form surface street lamp lens design method for facilitating operation and maintenance of the present invention;

[0050] Figure 4 A flow chart of a method for establishing a second mapping model M2 in the N-in-one free-form surface street lamp lens design method for facilitating operation and maintenance of the present invention;

[0051] Figure 5 A flow chart of a method for generating an initial free-form surface in the N-in-one free-form surface street lamp lens design method for facilitating operation and maintenance of the present invention;

[0052] Figure 6 A flow chart of a method for obtaining angle deviation data of each sub-area in the N-in-one free-form surface street lamp lens design method for facilitating operation and maintenance of the present invention;

[0053] Figure 7 The present invention is a flow chart of a method for obtaining a third optimized lens solution in the N-in-one free-form surface street lamp lens design method that is easy to operate and maintain. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Example 1

[0056] See also Figure 1 As shown, this embodiment provides an N-in-one free-form surface street lamp lens design method that is easy to operate and maintain, including:

[0057] Step S1000, determining the lighting demand of the target lighting scene, generating first lighting demand data; obtaining historical lens design parameters and historical lighting effect data of multiple LED street lamp lighting projects, generating second historical optimization data; based on the second historical optimization data, establishing a second mapping model M2 between free-form surface parameters and lighting effects through a machine learning algorithm; generating an initial free-form surface based on the second mapping model and the first lighting demand data;

[0058] Furthermore, step S1000 includes:

[0059] Step S1100, determining the lighting requirement of the target lighting scene and generating first lighting requirement data;

[0060] Furthermore, if Figure 2 As shown, step S1100 includes:

[0061] Step S1110, determining the illumination uniformity requirement U0, the glare limitation level requirement G0, the brightness longitudinal uniformity requirement UI and the brightness uniformity requirement L0 of the target lighting scene;

[0062] Step S1120, using U0, G0, UI and L0 as lighting constraint parameters to generate corresponding first lighting requirement data D1;

[0063] Step S1130 , performing parameter association on the first lighting demand data D1 , and defining U0 , G0 , UI and L0 as modifiable parameterized variables.

[0064] Specifically, step S1100 is to extract key lighting demand indicators by analyzing the target application scenario. The illumination uniformity requirement U0 reflects the uniformity of the road surface brightness distribution. The higher the U0, the more uniform the illumination, and the uneven light and dark areas can be avoided. The glare limit level requirement G0 reflects the control requirements for street lamp glare. The higher the G0 level, the smaller the allowed glare, and the better the visual comfort of drivers and pedestrians. The brightness longitudinal uniformity requirement UI is the ratio of the minimum brightness to the maximum brightness on the center line of each lane on the road surface, which measures the continuity of the road surface lighting in the longitudinal direction. Too low UI will cause stroboscopic effect; the brightness uniformity requirement L0 refers to the ratio of the minimum brightness to the average brightness on the road surface. L0 depends on the reflective characteristics of the road surface material. Too low L0 will cause uneven light and dark on the road surface. By inputting these lighting constraint parameters into professional lighting design software, using the optical calculation module therein and running the optical calculation engine, the lighting data set D1 that meets the constraints can be quickly calculated. D1 contains data such as light distribution curves and efficiency under different parameter combinations. In order to facilitate the subsequent optimization design, it is necessary to establish parameter associations and set U0, G0, UI and L0 as adjustable parameterized variables. The mapping relationship between constraint parameters and lighting data is realized through parameter tables. When the design requirements change, the corresponding lighting requirement data can be automatically generated by changing the parameters without recalculation, which greatly improves the design efficiency.

[0065] For example, a municipal road is to be retrofitted with LED street lamps. Through on-site surveys and communication, designers have determined the following lighting requirements: illumination uniformity requirement U0 ≥ 0.4, glare limit level requirement G0 is G3, brightness longitudinal uniformity requirement UI ≥ 0.6, and road surface brightness uniformity requirement L0 ≥ 0.4. U0, G0, UI and L0 are used as constraint parameters and input into the DIALux software. A set of lamp configuration parameters (such as lamp power, lamp spacing, installation height, etc.) that meet the requirements are obtained through optical calculations to generate the first lighting requirement data D1. U0, G0, UI and L0 in D1 are defined as adjustable parameterized variables. When the constraint indicators change, the corresponding configuration parameters can be quickly generated.

[0066] Step S1100 provides a clear optimization goal and evaluation basis for subsequent design by quantitatively describing the lighting requirements and generating a parameterized data set. The extraction of lighting constraint parameters requires the designer to have an in-depth understanding of the design specifications and user needs of the lighting project, while the parameterized processing relies on the computational analysis capabilities of professional lighting software. The combination of the two converts qualitative lighting requirements into quantitative design constraints, which is the basis for scientific lamp design. The generated first lighting demand data D1 covers the comprehensive performance of different working conditions and is a concentrated embodiment of multi-objective optimization. The parameter association gives D1 the ability to flexibly adjust, which is a strong guarantee for rapid response to changes in demand. This is fully prepared for subsequent historical data mining and machine learning modeling.

[0067] Step S1200, obtaining historical lens design parameters and historical lighting effect data of multiple LED street lamp lighting projects to generate second historical optimization data;

[0068] Furthermore, if Figure 3 As shown, step S1200 includes:

[0069] Step S1210, obtaining historical lens design parameters of the LED street lamp lighting project, wherein the historical lens design parameters include lens material parameters, lens outline dimensions, free-form surface control point coordinates, free-form surface control point coordinate weights, and light-emitting surface grid density;

[0070] Step S1220, obtaining historical lighting effect data of the LED street lamp lighting project, wherein the historical lighting effect data includes historical illumination uniformity, historical glare limit level, historical brightness longitudinal uniformity and historical brightness uniformity; historical lens design parameters and historical lighting effect data are combined into a data pair;

[0071] Step S1230, calculating the first similarity between the historical lighting effect data and the first lighting requirement data D1 in the data pair, extracting the data pairs whose first similarity is higher than a preset similarity threshold, and generating the second historical optimization data D2.

[0072] The calculating of the first similarity between the historical lighting effect data and the first lighting demand data D1 in the data pair includes:

[0073]

[0074] in:

[0075] : The first similarity, the value range is , the larger the value, the higher the similarity.

[0076] : Historical illumination uniformity in historical lighting effect data.

[0077] : Historical glare limit levels in historical lighting effect data.

[0078] : Longitudinal uniformity of historical brightness in historical lighting effect data.

[0079] : Historical brightness uniformity in historical lighting effect data.

[0080] :First lighting demand data The illumination uniformity requirement in

[0081] :First lighting demand data The glare limit level requirement in .

[0082] :First lighting demand data The requirement for longitudinal uniformity of brightness in the

[0083] :First lighting demand data The brightness uniformity requirement in

[0084] : The weight coefficient of historical illumination uniformity, indicating the influence of illumination uniformity on the first similarity.

[0085] : The weight coefficient of the historical glare restriction level, indicating the influence of the glare restriction level on the first similarity.

[0086] : The weight coefficient of the historical brightness longitudinal uniformity, which indicates the influence of the brightness longitudinal uniformity on the first similarity.

[0087] : The weight coefficient of historical brightness uniformity, indicating the influence of brightness uniformity on the first similarity.

[0088] : Historical illumination uniformity The standard deviation is used to reflect the degree of change in the illumination uniformity in the historical lighting effect data. It indicates the dispersion of the illumination uniformity value. The smaller the standard deviation, the smaller the change in illumination uniformity in the historical data. On the contrary, the change is larger.

[0089] : Historical glare restriction level The standard deviation of represents the degree of change in the glare limitation level in the historical lighting effect data. , which reflects the distribution of glare levels in historical data and can affect the sensitivity to glare limitation.

[0090] : Historical brightness longitudinal uniformity The standard deviation of .

[0091] : Historical brightness uniformity The standard deviation of .

[0092] , , , These four standard deviation parameters play a role in adjusting the sensitivity in the similarity calculation, so that the formula can more accurately reflect the degree of match between historical lighting effect data and current lighting needs. , , , Make similarity assessment more targeted and set by experts based on actual needs, which may be derived through historical data analysis or domain knowledge.

[0093] This formula can comprehensively consider multiple parameters such as illumination uniformity, glare limitation, longitudinal brightness uniformity, brightness uniformity, etc., and provide a more comprehensive similarity evaluation. By setting the standard deviation and weight coefficient, the model can dynamically adapt to the requirements of different lighting scenarios. The use of exponential function form can better capture the nonlinear relationship between parameters and avoid the limitations that may be brought by linear models.

[0094] Specifically, step S1200 is a key step in the intelligent optimization design using historical data. The historical data here mainly includes two categories: one is the design parameters of the lens, and the other is the lighting effect achieved by the lens solution in actual application. The lens design parameters cover many factors such as the selection of lens materials, external dimensions, surface shape, processing accuracy, etc. Material parameters such as refractive index and transmittance determine the optical performance of the lens. Outer dimensions such as diameter and thickness determine the integrated compatibility of the lens. The shape of the free-form surface directly affects the emission direction of light and is the key to achieving the light distribution curve. Parameters such as control point coordinates, coordinate weights, and grid density control the generation method and fineness of the surface. The shape of the free-form surface is usually described by a set of control point coordinates and weights. The coordinates of the control points determine the approximate shape of the surface, while the weights affect the local shape changes. By reasonably setting the coordinates and weights of the control points, the shape of the free-form surface can be flexibly controlled, thereby affecting the light distribution of the lens.

[0095] The lighting effect data intuitively reflects the actual performance of the lens design, including indicators such as illumination uniformity, glare limit level, longitudinal brightness uniformity, and brightness uniformity. Illuminance uniformity quantifies the uniformity of road lighting and reflects the basic performance of street lamps. Glare level reflects the ability to control light pollution, which is related to driving and residents' comfort. Longitudinal brightness uniformity reflects the changes in road brightness observed by the driver along the driving direction, which is crucial to ensure that the driver can identify obstacles ahead in time. Brightness uniformity represents the consistency of the overall brightness distribution of the road surface, which directly affects the visual perception and driving safety of road users. These effect data are derived from field tests and long-term use of the design scheme and have authoritative reference value.

[0096] Putting the design parameters and effect data into complete data pairs constitutes valuable historical optimization data. The data pairs contain both influencing factors and objective functions, which describe the input-output relationship of the lighting design. However, not all historical data are of reference value and need further screening. The concept of similarity is introduced here, that is, to quantitatively compare the historical lighting effects with the current design requirements to see how close the two are. This comparison is based on the same physical quantity and is comparable. By extracting data pairs with high similarity, high-quality second historical optimization data D2 are obtained. D2 is small in scale but highly targeted, and can provide high-quality training samples for machine learning modeling. Data pairs whose first similarity is less than or equal to the preset similarity threshold are not included in the second historical optimization data D2.

[0097] For example, following the example in step S1100, the designer selected 100 lens design solutions from previous LED street light design cases. For each solution, the following historical data was extracted:

[0098] 1. Lens material parameters: Polycarbonate (PC) is selected, with a refractive index of 1.59 and a transmittance of 88%;

[0099] 2. Lens dimensions: outer diameter 50mm, center thickness 5mm, edge thickness 2mm;

[0100] 3. Free-form surface control point coordinates: 8x8=64 control points are selected, and the coordinate data is stored as a 64x3 matrix;

[0101] 4. Light-emitting surface grid density: For a light-emitting surface with a diameter of 50 mm, the grid is evenly divided, and the number of grids is 100x100;

[0102] 5. Illumination uniformity: ranges from 0.35 to 0.5;

[0103] 6. Glare limit level: Referring to the CIE glare evaluation standard, there are 6 levels from G1 to G6. The glare level of this batch of solutions is distributed between G2 and G4;

[0104] 7. Brightness longitudinal uniformity: ranges from 0.55 to 0.65;

[0105] 8. Brightness uniformity: ranges from 0.4 to 0.55.

[0106] The above 8 types of parameters are combined to form data pairs from historical lens design parameters to historical lighting effect data. Using the first lighting requirement data D1 (U0≥0.4, G0=G3, UI≥0.6, L0≥0.4) obtained in step S1100 as a benchmark, calculate the first similarity between the historical lighting effect data in each data pair and D1. Set the similarity threshold to 0.9, filter out data pairs with a first similarity greater than or equal to 0.9, and form the second historical optimization data D2. Assume that the final D2 contains 20 data pairs, in the following format:

[0107] Data pair 1: {PC,1.59,88%,50mm,5mm,2mm,[(0.2,1.5,3.3),..., (10.1,9.8,4.2)],100x100,0.42,G3,0.55,0.4}; lens material: PC (polycarbonate); lens material refractive index: 1.59; lens material transmittance: 88%; lens outer diameter: 50mm; lens center thickness: 5mm; lens edge thickness: 2mm; control point coordinates of lens free-form surface: [(0.2,1.5,3.3),..., (10.1,9.8,4.2)], ellipsis indicates that there are other control point coordinates in the middle; LED light array size: 100mm x 100mm; illumination uniformity: 0.42; glare limitation level: G3; brightness longitudinal uniformity: 0.55; brightness uniformity: 0.4;

[0108] Data pair 2: {PC, 1.59, 88%, 50mm, 5mm, 2mm, [(0.3, 1.2, 3.5), ..., (10.3, 9.5, 4.5)], 100x100, 0.45, G2, 0.60, 0.45};

[0109] …

[0110] Data pair 20: {PC,1.59,88%,50mm,5mm,2mm,[(0.5,1.7,3.1),..., (10.6,9.1,4.7)],100x100,0.40,G3,0.65,0.50}.

[0111] In this way, we get a highly targeted second historical optimization data D2, which is small in size but highly representative and condenses the essence of previous successful designs. It overcomes the problems of redundancy and noise in the original historical data, allowing machine learning to focus on the most relevant design rules.

[0112] Step S1300, based on the second historical optimization data, a second mapping model M2 between free-form surface parameters and lighting effects is established by a machine learning algorithm;

[0113] Furthermore, if Figure 4 As shown, step S1300 includes:

[0114] Step S1310, using the historical lens design parameters in the second historical optimization data D2 as sample features X' and the corresponding historical lighting effect data as sample output Y' to construct a machine learning sample set (X', Y');

[0115] Step S1320, selecting a neural network algorithm as a machine learning model, and building a deep neural network structure including n4 hidden layers; n4 is a positive integer;

[0116] Step S1330, randomly dividing the machine learning sample set into a training set and a test set, and using the training set data to train the neural network to obtain a first mapping model M1;

[0117] Step S1340: Use the test set data to perform performance evaluation on the first mapping model M1, and optimize the first mapping model M1 based on the evaluation result to obtain the second mapping model M2.

[0118] Specifically, step S1300 uses machine learning technology to automatically extract design rules from the second historical optimization data. The sample feature X' contains the key design parameters of the lens, mainly considering the shape characteristics of the free surface, such as control point coordinates, coordinate weights, B-spline surface order, etc. At the same time, attribute parameters such as lens material and thickness can also be added. The sample output Y' selects quantitative indicators of lighting effects, such as illumination uniformity, glare value, etc. The two together constitute a sample set (X', Y') for machine learning. In terms of algorithm selection, it is more appropriate to use a neural network algorithm for the complex nonlinear characteristics of optical design. By setting multiple hidden layers and reasonably configuring the number of neurons and activation functions, more complex mapping relationships can be fitted. The sample data is divided into a training set and a test set according to a certain ratio, and the network is trained and learned using the training set to establish a first mapping model M1. However, in order to prevent overfitting and improve the generalization performance of the model, it is also necessary to use a test set to evaluate the performance of M1. Based on the evaluation results, the model can be optimized by adjusting the number of layers, neurons, regularization parameters, etc. of the network, and finally a better performing second mapping model M2 is obtained.

[0119] Exemplarily, D2 obtained in step S1200 is used as a sample set for machine learning. Each data pair is as follows:

[0120] {x'1,x'2,x'3,x'4,x'5,x'6,x'7,y'1,y'2,y'3,y'4};

[0121] Among them, x'1~x'7 represent the lens design parameters, specifically:

[0122] x'1: material (PC);

[0123] x'2: refractive index (1.59);

[0124] x'3: transmittance (88%);

[0125] x'4: outer diameter (50mm);

[0126] x'5: center thickness (5mm);

[0127] x'6: edge thickness (2mm);

[0128] x'7: Control point coordinate matrix (64x3).

[0129] y'1~y'4 represent lighting effect parameters, specifically:

[0130] y'1: illumination uniformity (e.g. 0.42);

[0131] y'2: glare limit level (such as G3);

[0132] y'3: longitudinal uniformity of brightness (e.g. 0.65);

[0133] y'4: Brightness uniformity (eg 0.50).

[0134] The lens design parameters and lighting effect parameters are used as sample features and output responses to form the (X', Y') sample set. Considering the complex nonlinear relationship between design parameters and performance, a multi-layer neural network is selected as the machine learning model. The network structure is designed as follows:

[0135] Input layer: 7 nodes (corresponding to x'1~x'7);

[0136] Hidden layer: 4 layers, each layer has 100, 80, 60, and 40 nodes respectively, and the activation function is ReLU;

[0137] Output layer: 4 nodes (corresponding to y'1~y'4), the activation function is the identity function;

[0138] Loss function: mean square error (MSE), optimization algorithm: Adam, learning rate: 0.01.

[0139] The 20 sample data were randomly divided into training sets (14) and test sets (6) in a ratio of 7:3. The neural network was built using the PyTorch framework, and the network was trained with the training set. The network was iterated for 1000 rounds, and the performance on the test set was evaluated every 100 rounds. When the decrease in the loss function value for 5 consecutive evaluations was less than 0.1%, the early stopping mechanism was triggered, the training was stopped, and the current model was saved. On this basis, K-fold cross validation and grid search were further used to fine-tune the model and optimize the hyperparameters. For example, the number of hidden layers was adjusted from 3 to 5, the number of nodes per layer was optimized between 30 and 120, different activation functions (such as Sigmoid, tanh) were tried, and the mean square error and R square were used as evaluation indicators to find the best performing model structure. Finally, the second mapping model M2 with the best fitting effect and the strongest generalization performance on D2 was obtained. M2 achieved accurate prediction from free surface parameters to lighting performance, and the average relative error was controlled within 5%.

[0140] Step S1300 is the core of the entire design method. Through machine learning, the design knowledge and experience hidden behind the historical data can be excavated, and the quantitative relationship between design parameters and lighting performance can be established. Traditional design methods mainly rely on the experience and intuition of designers, and are often difficult to deal with complex design problems with multiple parameters and strong coupling. Machine learning just makes up for the shortcomings of the human brain and can automatically find rules from massive, high-dimensional data. The powerful nonlinear fitting ability of neural networks enables them to model complex mapping relationships in optical design well. Through continuous training and optimization of the model, a mapping model with both accuracy and generalization can be obtained. The model can quickly reverse the design parameters from the design requirements, greatly simplifying the design process. At the same time, machine learning transforms individual and fragmented design experience into general design knowledge, which is also conducive to the precipitation and inheritance of knowledge. Design experience will no longer be overly dependent on senior experts. This data-driven, machine learning-assisted design model represents the development trend of future design.

[0141] Step S1400, generating an initial free-form surface based on the second mapping model and the first lighting requirement data;

[0142] Furthermore, if Figure 5 As shown, step S1400 includes:

[0143] Step S1410, inputting the first lighting requirement data D1 into the second mapping model M2 to generate a first free-form surface parameter set C1, wherein the first free-form surface parameter set C1 includes the control point coordinates and weights of the free-form surface;

[0144] Step S1420, generating a first B-spline free-form surface S1 based on the control point coordinates and weights of the free-form surface in the first free-form surface parameter set C1;

[0145] Step S1430: taking the first B-spline free-form surface S1 as the initial free-form surface.

[0146] Specifically, step S1400 uses the trained mapping model to quickly generate an initial design plan from the lighting requirements. The first lighting requirement data D1 obtained in step S1100 is input into the model M2, and a set of free surface parameter sets C1 that meet the requirements can be predicted. The key parameters here are the control point coordinates and weights of the free surface, which determine the shape of the surface and the local shape changes. Using parametric CAD software, the control point coordinates and weights are imported, and a complete first B-spline free surface S1 is generated through B-spline surface fitting. S1 not only meets the lighting performance requirements, but also facilitates subsequent adjustments and optimizations with the help of parametric expressions. Based on S1, designers can quickly build the geometric shape of the entire lens. Compared with designing from scratch, this machine learning-based initialization method can greatly improve design efficiency, freeing designers from tedious modeling work and devoting more energy to innovative design thinking.

[0147] For example, if the first lighting requirement data D1 is input into the second mapping model M2, M2 quickly predicts a set of free surface parameter sets C1 that meet D1, including the following key information:

[0148] 1. Control point coordinate matrix Poles:

[0149] [[-0.23, 1.32, 3.41]; ... [10.37, 9.28, 4.56]];

[0150] 2. Control point weight matrix Weights:

[0151] [[1.0, 0.8, 1.2]; ... [0.9, 1.1, 1.0]];

[0152] Using Python's OpenCascade library, the steps to generate a cubic B-spline surface based on C1 are as follows:

[0153] 1. Define the order of the surface in the u and v directions to be 3 and 3 respectively, and the node vector u knots and v knots They are [0,0,0,0,0.25,0.5,0.75,1,1,1,1] and [0,0,0,0,0.2,0.4,0.6,0.8,1,1,1,1] respectively, and the number of control points is 8x8.

[0154] 2. Create an initial B-spline surface object S0.

[0155] 3. Traverse the control point coordinate matrix Poles and assign the coordinates of the 64 control points to the poles attribute of S0 in turn.

[0156] 4. Traverse the control point weight matrix Weights and assign the weights of the 64 control points to the weights attribute of S0 in turn. Note that the control point weights and coordinates of the B-spline surface in OpenCascade are defined separately and are set separately through the weights attribute.

[0157] 5. Finally, a fully defined weighted first B-spline free-form surface S1 is obtained. S1 not only contains the coordinate information of the control points, but also integrates the weight information of each point, which can make more precise control of the local surface shape.

[0158] By introducing the weights of the control points, the local shape of the surface can be changed only by adjusting the weights while keeping the coordinates of the control points unchanged. This undoubtedly provides more freedom and flexibility for the design of free-form surfaces. For example, increasing the weights of certain control points will make the surface convex near the point; conversely, reducing the weights will make the surface concave at that point. By cleverly using the weight characteristics, the shape of the surface can be optimized in subtle places to meet different lighting design requirements. The optimization knowledge learned by the second mapping model M2 is not only contained in the spatial coordinates of the control points, but also implicit in the ratio of the weight coefficients. The coordinates and weights work together to determine a high-quality free-form surface shape. The first B-spline free-form surface S1 generated in this way integrates historical experience and machine learning wisdom, and can fully meet lighting needs. This method of constructing parameterized free-form surfaces lays a solid foundation for subsequent lens modeling and optimization.

[0159] The core idea of ​​step S1400 is "intelligent design initialization". The traditional design process often requires a lot of repeated links such as modeling, simulation, evaluation, and modification, which are long and costly. The machine learning model can directly "predict" the design parameters close to the optimal solution from the demand, making the initial design solution closer to the global optimal solution. This "warm-start" strategy allows the optimization design algorithm to converge to the optimal solution faster. At the same time, since the generated initial surface is trained based on real engineering data, it itself meets the engineering constraints, which can reduce the workload of subsequent design modifications and verifications. In addition, the generation of parametric surfaces also lays a good foundation for subsequent optimization. Using the idea of ​​parametric design, design variables and constraints can be flexibly adjusted, making the design process more flexible and efficient. From a higher level, the idea of ​​step S1400 is in line with the concept of "semantic design". It establishes a direct semantic association between design requirements and design solutions, making the design process more natural and intuitive. This is an important feature of future intelligent design systems.

[0160] Step S1000 starts from the lighting requirements and first establishes a parameterized target description mechanism. On this basis, through deep mining of historical data and using machine learning algorithms, a mapping model from free-form surface design parameters to lighting performance is constructed. This data-driven design method can significantly improve design efficiency and quality. Finally, the trained mapping model is used to quickly generate high-quality free-form surface designs from quantitative lighting requirements. The parametric design concept runs through the whole process, which greatly improves the flexibility and adaptability of the design process. Step S1000 has made useful explorations in design automation and intelligence, providing new ideas for LED lenses and even the entire lighting design field.

[0161] Step S2000, dividing the initial free-form surface into three sub-areas, namely, a central area, a transition area, and an edge area, calculating the actual light emission angle distribution function of each sub-area, and comparing the actual light emission angle distribution function with the preset ideal light emission angle distribution function to obtain the angle deviation data of each sub-area; according to the angle deviation data, by adjusting the coordinates and weights of the free-form surface control points of each sub-area, the shape of the free-form surface is continuously updated to obtain an updated free-form surface, and the second angle deviation data of each sub-area of ​​the updated free-form surface is calculated until the second angle deviation data of all sub-areas are lower than the set angle deviation convergence threshold , obtaining a second optimized free-form surface; based on the second optimized free-form surface, introducing material screening, obtaining a third optimized lens solution;

[0162] Furthermore, step S2000 includes:

[0163] Step S2100, dividing the initial free-form surface into three sub-areas: a central area, a transition area, and an edge area, each sub-area corresponding to a different light emission angle range;

[0164] Furthermore, step S2100 includes:

[0165] Step S2110, taking the center of the lens as the origin, a three-dimensional rectangular coordinate system is established, which is marked as a free-form surface coordinate system. The function expression of the initial free-form surface in the free-form surface coordinate system is: ; is the spherical radius, which is used to describe the radial position of a point on a free-form surface; is the zenith angle, which is used to describe the longitudinal position of a point on a free-form surface; is the azimuth, which is used to describe the lateral position of a point on the free-form surface; z' is the function expression of the initial free-form surface in the free-form surface coordinate system;

[0166] Step S2120, use represents the boundary of the central area, which will satisfy The surface is divided into the central area; ρ1 represents the spherical radius of the central area boundary, θ1 represents the zenith angle of the central area boundary, and φ1 represents the azimuth angle of the central area boundary;

[0167] Step S2130, use represents the boundary of the transition zone, which will satisfy The surface is divided into a transition zone; ρ2 represents the spherical radius of the transition zone boundary, θ2 represents the zenith angle of the transition zone boundary, and φ2 represents the azimuth angle of the transition zone boundary;

[0168] Step S2140, use represents the boundary of the marginal area, which will satisfy The surface is divided into edge areas; ρ3 represents the spherical radius of the edge area boundary, θ3 represents the zenith angle of the edge area boundary, and φ3 represents the azimuth angle of the edge area boundary.

[0169] Specifically, step S2100 divides the initial free-form surface into three sub-areas: the central area, the transition area, and the edge area, and each sub-area corresponds to a different light-emitting angle range. Among them, the free-form surface refers to a parameterized, aspherical optical surface. Compared with traditional spherical and aspherical surfaces, the free-form surface has more design freedom and can more flexibly control the emission direction and angle distribution of light. Dividing the free-form surface into three sub-areas can specifically optimize the light intensity distribution within different light-emitting angle ranges and improve the overall performance of the lens.

[0170] The function expression of the initial free-form surface in the free-form surface coordinate system is z'=f(ρ,θ,φ); where ρ is the spherical radius, θ is the zenith angle, and φ is the azimuth angle. Here, the spherical coordinate system is used to describe the free-form surface. The spherical radius ρ represents the distance from any point in space to the origin of the coordinate system; the zenith angle θ represents the angle between the spatial point and the positive direction of the Z axis, and the value range is [0,π]; the azimuth angle φ represents the angle between the projection of the spatial point on the XOY plane and the positive direction of the X axis, and the value range is [0,2π). Using the spherical coordinate system, the position of any point in three-dimensional space can be uniquely determined, which is convenient for the subsequent parameterization and optimization of the free-form surface shape.

[0171] The central area mainly plays the role of focusing and collimation, which can make the light emitted by the LED light source converge into a parallel beam and improve the utilization rate of light energy. Reasonable setting of the size and curvature of the central area can reduce the imaging of the LED light source and improve the uniformity of the light spot. The transition area plays the role of diffusion and homogenization, which can appropriately expand the light output angle of the LED light source, so that the light is evenly distributed in a larger range and improve the comfort of lighting. By optimizing the shape parameters of the transition zone, the light intensity distribution between the central area and the edge area can be smoothed to reduce the glare. The edge area mainly plays the role of diffusion and fill light, which can make the lighting range of the LED street lamp cover the surrounding environment as much as possible and improve the uniformity of lighting. At the same time, the edge area can reduce the sharp attenuation of light energy at a large light output angle and improve visual comfort. Reasonable design of the edge area surface can significantly improve the user experience of LED street lamps.

[0172] By partitioning the free-form surface, the overall optimization problem can be transformed into multiple local optimization problems. Different optimization strategies and objective functions can be used in each sub-region to control the local light output effect more specifically. This divide-and-conquer approach can reduce the complexity of the optimization problem and improve the convergence speed and accuracy. At the same time, smooth transitions are used between sub-regions to avoid performance mutations at the region boundaries. This partition optimization method has been widely used in free-form surface design and can significantly improve the imaging quality and light output efficiency of the lens.

[0173] Step S2200, calculating the actual light output angle distribution function of each sub-area, and comparing the actual light output angle distribution function with a preset ideal light output angle distribution function to obtain angle deviation data of each sub-area;

[0174] Furthermore, if Figure 6 As shown, step S2200 includes:

[0175] Step S2210, randomly generate N2 rays, marked as N2 incident rays, and incident the incident rays onto the initial free-form surface; each incident ray is represented by a parametric equation of a space line, with its starting point being a random sampling point on the initial free-form surface and its direction being a randomly sampled unit vector on the unit sphere, marked as the direction vector of the incident ray; N2 is a positive integer;

[0176] Step S2220: for each incident light ray, determine the incident light ray I j Is the incident angle less than the critical angle of total reflection? j The incident angle is less than the critical angle of total reflection, then according to the incident light I j The direction vector of the incident light ray I is solved using Snell's law. j The direction vector of the corresponding refracted light; I j represents the jth incident light ray; 1≤j≤N2;

[0177] Step S2230, judging whether the refracted light can be emitted through the lens according to the direction vector of the refracted light; if it can be emitted, recording the emission angle ψ of the refracted light;

[0178] Specifically, the actual light output angle distribution function of each sub-area is calculated using the Monte Carlo ray tracing method; Monte Carlo ray tracing is a numerical method that simulates the light propagation process by random sampling. Its basic idea is to approximate the luminous flux with a large number of random light rays, and to estimate the performance of the optical system by statistically analyzing the interaction between these light rays and the surface of the object. In step S2210, the free-form surface coordinate system is the two-dimensional plane rectangular coordinate system established in step S2110. In this coordinate system, the analytical equation of the initial free-form surface is known, and random sampling points on the initial free-form surface can be generated based on it. The unit sphere represents a spherical surface with a radius of 1, corresponding to all possible light directions. By uniformly random sampling on the unit sphere, the light distribution in all directions can be obtained.

[0179] For each generated light ray, first determine whether the incident light ray can pass through the free-form surface, that is, whether the incident angle is less than the critical angle of total reflection. If the penetration condition is met, the direction vector of the refracted light ray is solved by Snell's law according to the direction of the incident light ray, the initial free-form surface normal vector and the refractive index of the initial free-form surface material. Then further determine whether the refracted light ray can pass through the entire lens. To determine whether the refracted light ray can exit the lens, two conditions must be considered: 1. The intersection of the refracted light ray and the outer surface of the lens must be within the lens aperture; 2. The incident angle of the refracted light ray with the outer surface must be less than the critical angle of total reflection. This requires an accurate description of the three-dimensional shape of the lens and the intersection calculation of the light ray and the surface. If the refracted light ray can finally exit the lens smoothly, the angle ψ at the time of its exit is recorded. ψ is the angle between the exiting light ray and the axis of the lens, which describes the exit direction of the light ray.

[0180] The step of determining whether the refracted light can be emitted through the lens comprises:

[0181]

[0182] in:

[0183] : refractive index of air, usually 1;

[0184] : refractive index of lens material;

[0185] : thickness of the lens;

[0186] : inner diameter of the concave surface of the lens;

[0187] : angle of incidence, i.e. and The angle of : direction vector of the incident light;

[0188] : refraction angle, that is and The angle of is the direction vector of the refracted ray, is the normal vector of the initial free-form surface at the intersection point;

[0189] : Coordinates of the intersection of the refracted light and the concave surface; is the coordinate of the intersection point of the refracted light ray and the concave surface in the z-axis direction;

[0190] : Lens output angle, that is, the angle between the output light and the central axis of the lens;

[0191] : Indicates that the intersection of the refracted light and the concave surface is The projection radius of the plane;

[0192] is a binary function, when When , it means that the light can pass through the lens; otherwise, it will be totally reflected or prevented from emitting.

[0193] The judgment conditions include 4 inequalities:

[0194] inequality It is used to determine whether total reflection occurs at the initial free-form surface. If not, the light is totally reflected and cannot pass through the lens.

[0195] inequality Ensure the projection radius of the refracted light on the xy plane It falls within half of the inner diameter of the lens. This ensures that the light is within the effective area of ​​the lens and prevents the light from emitting outside the lens, thus ensuring imaging performance.

[0196] inequality Determine the relationship between the refracted light and the lens thickness d to ensure the intersection of the light Within the range of lens thickness. Ensure that light can pass through the entire thickness of the lens to avoid light being blocked inside the lens.

[0197] inequality Used to determine the refraction angle of refracted light in the lens Whether the conditions for total reflection are exceeded. Ensure that the light can be emitted smoothly from the concave surface of the lens to avoid light loss caused by total reflection.

[0198] With the direction vector of the incident light , the geometric parameters of the lens (such as thickness , inner diameter ) and other changes, the model The value changes accordingly.

[0199] Increase the thickness of the lens May limit the exit angle , thereby reducing the transmittance; and adjusting the inner diameter Can affect the projection radius of the light , thereby improving light transmission efficiency.

[0200] The model can accurately determine the propagation path and emission conditions of light in the lens, which helps to optimize the lens design. By adjusting the design parameters, the light transmittance of the lens can be increased and the imaging quality can be improved. The model is suitable for the analysis and design of complex lens systems and can reduce the cost of physical prototype production. Combined with actual design, the optical system can be optimized and the overall performance can be improved.

[0201] Step S2240: Count the emission angles ψ of all refracted light rays in each sub-region, draw a probability distribution histogram of the emission angle ψ, and normalize the probability distribution histogram to obtain the actual light emission angle distribution function g(ψ) of each sub-region. i ; g(ψ) i represents the actual light output angle distribution function of sub-region i, i is the index of the sub-region, 1≤i≤3; the sub-region includes a central region, a transition region and an edge region;

[0202] The drawing of the probability distribution histogram of the emission angle ψ comprises:

[0203] According to the value range of the emission angle ψ, the emission angle ψ is divided into n3 small intervals of equal width, and the n3 small intervals of equal width are used as the abscissa of the probability distribution histogram; the abscissa and ordinate of the probability distribution histogram are the number of emission light rays falling in each small interval.

[0204] Step S2250: Subtract the actual light output angle distribution function g(ψ) of sub-region i from the i The ideal light output angle distribution function f(ψ) corresponding to sub-area i i Compare and calculate the angle deviation data of sub-area i .

[0205] Specifically, the probability distribution histogram of the emission angle ψ is actually a finite number of discrete intervals that approximate the continuous angle distribution function. The finer the interval division, the closer the histogram is to the true distribution. By normalizing the histogram so that the total area is 1, the actual light emission angle distribution function g(ψ) of sub-area i is obtained. i . Converted from histogram to g(ψ) i The process can be realized by using piecewise linear interpolation and other methods. g(ψ) i It is a probability density function, which indicates the probability that the emission angle falls near ψ. It quantitatively describes the actual distribution characteristics of the light emission angle in sub-region i.

[0206] f(ψ) i It reflects the expected requirements for the light output angle distribution of sub-area i, and can be set according to the light distribution curve characteristics to be achieved. The calculation formula is:

[0207]

[0208] in:

[0209] Integration interval : This is a sub-area The light output angle range can be set according to the specific sub-area characteristics.

[0210] : Quantifies the degree of deviation between the actual light output angle distribution and the ideal light output angle distribution. The smaller it is, the closer the actual distribution is to the ideal distribution; This indicates that it fully complies with the ideal distribution.

[0211] Ideal light angle distribution function It is often determined according to the specific requirements of the light distribution design. For example, to achieve a light pattern with high brightness in the center and low brightness around the lens, Take a larger value in a small angle range and a smaller value in a large angle range. On the contrary, if you want to achieve wide-angle uniform light distribution, Should be as flat as possible. Angle deviation The calculation of uses the L1 norm in functional analysis, that is, Integrate over the light output angle range. This calculation method can better reflect the local difference between the two functions than the simple mean square error.

[0212] Introducing angular deviation As an optimization target, it can make the design of the initial free-form surface more quantitative and goal-oriented. Deviation from the ideal distribution In the subsequent steps, we can use The gradient information of the free-form surface guides the iterative correction of the free-form surface shape. This quantitative feedback mechanism avoids the drawbacks of manual judgment and trial and error in traditional optimization methods, and can greatly improve the efficiency and accuracy of the initial free-form surface design. At the same time, the sub-region division makes the optimization target more refined and targeted, and different light requirements can be flexibly applied to different regions.

[0213] Step S2200 uses numerical simulation and statistical analysis to establish a quantitative connection between the lens microstructure and macroscopic performance. It overcomes the limitations of the traditional forward design method and provides a clear evaluation criterion and optimization path for the initial free-form surface configuration optimization. It has good physical interpretability, high design freedom, and can give full play to the flexible control ability of the free-form surface, and has obvious advantages in the precise redirection of light energy. At the same time, this method also has good applicability and can flexibly handle different light output requirements and complex light source conditions.

[0214] In general, the method given in step S2200 has a solid physical and mathematical foundation, and the technical route is clear and reasonable. It introduces the ideas and methods of illumination optics on the basis of traditional imaging optical design, and realizes the quantitative control of light energy distribution. It not only makes full use of the design freedom of free-form surfaces, but also avoids blind search and manual experience judgment. It lays an important foundation for further optimization. This has certain enlightenment significance and promotion value for LED lenses and even the entire field of non-imaging optics.

[0215] Step S2300: According to the angle deviation data, the free-form surface shape is continuously updated by adjusting the coordinates and weights of the free-form surface control points of each sub-region to obtain an updated free-form surface. , calculate the updated free-form surface The second angle deviation data of each sub-area Until the second angle deviation data of all sub-areas are lower than the set angle deviation convergence threshold , and obtain the second optimized free-form surface;

[0216] Further, step S2300 includes:

[0217] Step S2310, set the number of initialization iterations , set the angle deviation convergence threshold , let the second optimized free-form surface ,in is the initial free-form surface obtained in step S1000; represents the free surface obtained at the 0th iteration;

[0218] Step S2320, make , according to The free-form surface obtained by iteration , combining the angle deviation data of each sub-region obtained in step S2200, calculating the coordinate offset and weight change of the control point in each sub-region;

[0219] Further, step S2320 includes:

[0220] Step S2321: Divide all sub-regions into n grid units, and take the area of ​​sub-region i through n grid units. As the weight, calculate the weighted average of the angle deviation of each sub-area ,Right now:

[0221]

[0222] in, Indicates The weighted average of the angle deviations of the sub-areas, Indicates In the sub-area The area of ​​a grid cell, Indicates In the sub-area is the angular deviation of each grid cell, and n is the total number of grid cells.

[0223] Step S2322, according to Calculate the The shape adjustment coefficient of the sub-region :

[0224]

[0225] in, represents the maximum value of the weighted average of the angle deviations of all sub-areas, is the shape adjustment rate factor, with a value of within the range.

[0226] Step S2323, adjusting coefficients according to shape , for Each control point in the sub-region , calculate its coordinate offset and weight change :

[0227]

[0228] in, is the first Line The control points of the columns, For control points The angular deviation of the grid cell, For the The arithmetic mean of the angle deviations of all grid cells in a sub-region, is the Z-axis unit vector.

[0229] Step S2330, using the coordinate offset and weight change , for free-form surfaces The coordinates and weights of each control point are corrected to obtain the updated free-form surface ;

[0230] make:

[0231]

[0232] in, :Indicates in In each sub-area, after After the iteration update Row and The coordinates of the control points. These are the positions of the control points adjusted during the optimization process and used to define the shape of the freeform surface.

[0233] :Indicates in In the sub-area, Row and The control points are listed in The coordinates of the control points at the previous iteration. This is the position of the control points in the previous iteration and is the basis for the update process.

[0234] :Indicates in In each sub-area, after After the iteration update Row and The weight of the control points. This parameter is used to adjust the influence of the control points on the free-form surface shape. The larger the weight, the more significant the influence of the control points on the surface shape.

[0235] :Indicates in In the sub-area, Row and The control points are listed in The weight at the iteration.

[0236] Combining the updated control point data of all sub-areas, we can get Free-form surface .

[0237] Step S2340, using the method of step S2200, calculate the updated free-form surface The second angle deviation data of each sub-area , Represents the second angle deviation data of sub-area i; determines whether ; If satisfied, let the second optimized free-form surface , the iteration ends; otherwise, return to step S2320 and continue the iterative optimization.

[0238] Specifically, the iterative optimization process of step S2300 adopts a heuristic search strategy, which continuously approaches the optimal solution by tentatively adjusting the shape of the free-form surface. Different from traditional numerical optimization algorithms such as gradient descent, this method makes full use of the heuristic information contained in the angle deviation data and applies different correction strategies to different areas. This targeted and adaptive optimization method can quickly lock in favorable directions in a complex solution space, greatly improving the convergence efficiency.

[0239] The core of iterative optimization is the construction of the shape update function. It calculates the coordinate offset and weight change of each control point based on the deviation between the current surface shape and the ideal light distribution. The offset direction is always consistent with the Z axis, reflecting the control freedom of the "convex and concave deformation" of the free-form surface. The size of the offset and weight is driven by the deviation data. The larger the deviation, the more radical the adjustment; only fine-tuning is performed in areas with small deviations. This is the intelligence of the algorithm. It can automatically identify the key areas of shaping and focus limited computing resources. It is worth mentioning that the shape update also takes into account the coupling effect between adjacent areas and introduces weighted averaging to smooth the shaping boundaries and ensure the smoothness of the surface.

[0240] Iterative optimization also introduces a shape adjustment rate factor λ to control the convergence speed and stability of the algorithm. The value of λ needs to strike a balance between efficiency and robustness, and can usually be selected between 0.1 and 0.5. If λ is too large, it may cause oscillation and miss the optimal solution; if it is too small, it will affect the convergence speed. λ can be dynamically adjusted through empirical formulas or adaptive strategies, such as using the idea of ​​simulated annealing to gradually reduce λ as the number of iterations increases. This helps to quickly approach the optimal area in the early stage, and then search in small steps in the later stage to jump out of the local minimum point.

[0241] The termination condition of the algorithm is based on the angle deviation convergence threshold δ. The setting of δ needs to comprehensively consider factors such as design requirements, processing technology, and computational cost. The smaller δ is, the higher the requirement for surface accuracy and the better the corresponding lens performance, but the greater the processing difficulty and computational complexity. The selection of δ can refer to industry standards and expert experience, such as estimating the allowable range of light angle deviation based on indicators such as lighting uniformity and glare level.

[0242] From a macroscopic perspective, it is to find the optimal surface shape in the solution space; from a microscopic perspective, it is to adjust the coordinates and weights of the control points in a specific parametric surface family. This idea gives full play to the flexibility of free-form surfaces and the physical characteristics of LED lighting, and broadens new ideas for intelligent design. Compared with traditional forward design, inverse optimization is more efficient and accurate, which is of great value in shortening the development cycle and improving product performance.

[0243] The innovation of this algorithm lies in the use of ray tracing to introduce geometric information and realize closed-loop optimization of the surface through a feedback mechanism. It cleverly combines the partitioning idea and the weight strategy to achieve fine control of complex surfaces. At the same time, the algorithm is universal and is not limited to a specific surface parameterization method. It can be easily extended to other types of free-form surfaces. This provides a powerful tool for the development of personalized and customized LED lenses.

[0244] Step S2400: Based on the second optimized free-form surface, material screening is introduced to obtain a third optimized lens solution.

[0245] Furthermore, if Figure 7 As shown, step S2400 includes:

[0246] Step S2410, selecting L candidate lens materials, and obtaining a refractive index value range and an extinction coefficient value range of each candidate lens material; L is a positive integer greater than 1;

[0247] Step S2420, for each candidate lens material, uniformly sample M parameter combinations within the value range of its refractive index and extinction coefficient, each parameter combination corresponds to a lens material instance; M is a positive integer;

[0248] Step S2430, calculating the transmittance T and light extraction efficiency E of each lens material instance in combination with the second optimized free-form surface, and constructing an objective function F in combination with the cost factor C of the lens material instance;

[0249] F = αT + βE + γC;

[0250] Among them, α is the weight coefficient of the transmittance T, β is the weight coefficient of the light extraction efficiency E, and γ is the weight coefficient of the cost factor C of the lens material instance.

[0251] Step S2440: Select the lens solution with the largest objective function F value from among the L×M lens solutions as the third optimized lens solution to be output.

[0252] Specifically, step S2400 further introduces material screening based on the optimization of free-form surfaces, in order to find the global optimal solution in the design space of "shape-material" coupling. The refractive index and extinction coefficient of optical materials are key parameters that affect the performance of lenses. The refractive index determines the propagation path of light inside the lens and is closely related to the focusing characteristics. The extinction coefficient reflects the degree of absorption and attenuation of light energy by the material, affecting the transmittance and light extraction efficiency. The parameter ranges of different materials are different, and they need to be reasonably selected according to the specific material properties in the design.

[0253] The candidate material library contains L kinds of commonly used lens materials, such as PMMA, PC, optical glass, etc. The initial selection of materials can be based on previous design experience and recommendations in the material manual. For each candidate material, discrete sampling is performed within the range of its parameter variation to cover as many parameter combinations as possible. The density of the sampling points is determined by the M value. The larger the M, the more detailed the sampling and the more accurate the optimal solution, but the greater the amount of calculation. When sampling parameters, special attention should be paid to the combined effect of the refractive index and the extinction coefficient, which influence each other. A high refractive index often leads to an increase in dispersion and absorption loss, and vice versa. Therefore, the selection of sampling points should be as balanced as possible to avoid overly extreme parameter combinations.

[0254] By substituting the sampled material instances into the second optimized free-form surface one by one, a series of complete lens solutions can be obtained. Using the Monte Carlo ray tracing in step S2200, the transmittance T and light extraction efficiency E of each solution can be simulated and analyzed. Compared with "ideal material + free-form surface", the optical performance after considering the material effect will be reduced. Reduced transmittance means more light energy loss, and reduced light extraction efficiency affects the uniformity of lighting. Therefore, it is necessary to focus on the matching characteristics of the material instance and the free-form surface, and select material parameters that can maximize the performance of the free-form surface.

[0255] The final material selection needs to comprehensively consider optical performance and economic cost. In order to balance the influence of the two, the weighted objective function F is introduced. The values ​​of α, β, and γ reflect the designer's relative emphasis on transmittance, light output efficiency, and cost. Generally speaking, optical performance is the primary consideration, and α and β should be given a larger weight; cost factors are second, and γ can take a moderate value. Through linear weighting, the three indicators are normalized to the same dimension, and the obtained objective function value F can directly reflect the pros and cons of the solution. The solution with the largest F value is the third optimized lens solution, which can maximize the performance requirements proposed by the first lighting demand data under the current constraints.

[0256] The third optimization scheme is a perfect combination of shape design and material selection. The free-form surface creates ideal optical performance, and the optimal material ensures full performance. Compared with the design that simply pursues extreme performance, the third optimization scheme takes into account material costs while achieving high-level lighting effects, and is more practical. In actual production, standard materials can be purchased according to the specifications of the third optimization scheme, and high-performance, low-cost LED lens products can be manufactured using processes such as CNC machining and ultra-precision turning, thereby promoting the large-scale application of LED lighting technology in public places such as roads, tunnels, and squares.

[0257] Step S3000: Perform simulation verification and test optimization on the third optimized lens solution to generate a final N-in-one planar light-emitting LED street lamp lens design solution.

[0258] Furthermore, step S3000 includes:

[0259] Step S3100, import the design parameters of the third optimized lens solution into the optical simulation software, build a simulation environment, set the light source parameters, material parameters, geometric parameters and mesh parameters, run the ray tracing algorithm to simulate and analyze the lens, and obtain the theoretical lighting performance parameters of the lens, wherein the theoretical lighting performance parameters include the illuminance distribution curve, uniformity and glare index.

[0260] Among them, the optical simulation software can be commercial software such as ASAP (Advanced Systems Analysis Program), LightTools, TracePro, etc., or open source software such as FRED (Full Reference Evaluation Device) can be used. The construction of the simulation environment includes setting the spectral distribution and output angle distribution parameters of the LED light source, setting the material refractive index, absorptivity, transmittance and other parameters of the free-form surface lens, and the geometric parameters such as the lens's outer dimensions and surface shape data. In order to ensure the accuracy of the simulation, it is usually necessary to generate a large number of grid cells on the surface and inside of the lens. The ray tracing algorithm simulates the propagation process of light in the lens through the interaction of a large number of light rays with the surface and inside of the lens (such as refraction, reflection, scattering, absorption, etc.), and then obtains the output light distribution of the lens.

[0261] Step S3200, judging whether the lighting requirements determined in step S1000 are met according to the theoretical lighting performance parameters obtained by simulation. If so, output the final N-in-one planar light-emitting LED street lamp lens design scheme; if not, return to step S2000, adjust the shape of the second optimized free-form surface, and regenerate the third optimized lens scheme until the lighting requirements are met.

[0262] Among them, the judgment criteria for lighting needs can refer to the following aspects:

[0263] (1) Whether the illumination uniformity meets the U0 requirement, which is usually required to be above 0.4;

[0264] (2) Whether the glare index meets the limit level of G0. G0 is usually required to be above G3 (G3, G4, G5, G6);

[0265] (3) Whether the luminous flux meets the standard requirements for road lighting. Generally, the standard for urban main roads is 10-15lx, for secondary roads is 8-12lx, and for branch roads is 6-10lx;

[0266] (4) Whether the longitudinal uniformity of brightness meets the UI requirements. Generally, the standard for urban main roads is above 0.6, and for secondary roads is above 0.4;

[0267] (5) Whether the brightness uniformity meets the L0 requirement. Generally, the standard for urban main roads is above 0.4, and for secondary roads it is above 0.35.

[0268] This iterative optimization process can further improve the lighting performance of the lens design so that it can better meet the needs of actual applications.

[0269] Step S3300: Make a lens sample and perform a lighting test in an actual scene.

[0270] Lens samples are made by 3D printing or injection molding, and lighting tests are carried out in actual scenes. The test items include illumination distribution, glare evaluation, luminous flux, longitudinal brightness uniformity, brightness uniformity, etc. The test results are compared and analyzed with the simulation results. If necessary, the lens design scheme is further optimized to finally obtain an N-in-one flat light-emitting LED street lamp lens product that meets the use requirements.

[0271] Among them, 3D printing technology can directly print out transparent samples based on the CAD model of the lens, which is suitable for small batch or customized production; injection molding is to mass produce lenses through injection molds, with mature technology and low cost, which is suitable for large-scale production. The two processing methods have their own characteristics and can be selected according to needs.

[0272] In actual lighting tests, a standard street light test platform can be built to simulate different road conditions (such as asphalt roads, cement roads, asphalt roads, etc.), and the lighting effects of street lights can be tested at different installation heights and spacings. You can also directly select the target road for on-site testing to more realistically evaluate the performance of street lights.

[0273] Through a series of verification and optimization processes such as simulation, processing, and testing, we can finally obtain a high-performance, low-cost, scalable N-in-one flat light-emitting LED street lamp lens. The lens can efficiently converge and shape the light emitted by the LED light source to form a uniform and soft rectangular light spot, significantly improving the lighting quality and energy efficiency of the street lamp. Compared with the traditional free-form surface optical design method, the present invention uses historical optimization data to train the machine learning model to assist in generating the initial free-form surface, and then refines and optimizes each area, which can greatly shorten the design cycle and reduce the requirements for the designer's experience and skills. At the same time, the lens material screening link is introduced to take into account the lens cost while ensuring the optical performance, further improving the cost-effectiveness and practicality of the design scheme.

[0274] In summary, the N-in-one free-form surface street lamp lens design method provided by the present invention, which is easy to operate and maintain, integrates multidisciplinary technologies such as machine learning, computational optimization, and materials science to form a complete and efficient design optimization process, which can significantly improve the research and development efficiency and cost-effectiveness of LED street lamp lenses, and is of great significance for promoting the large-scale application of semiconductor lighting technology in the field of urban infrastructure.

[0275] Example 2

[0276] This embodiment provides a lens based on Embodiment 1, wherein the lens includes three sub-regions: a central region, a transition region, and an edge region; and the light emitting surface of the lens is a plane.

[0277] The free-form surface of the lens has undergone multi-objective optimization design, which not only meets multiple lighting requirements such as illumination uniformity, glare limitation, longitudinal brightness uniformity, and brightness uniformity, but also takes into account the improvement of optical efficiency. Since the machine learning algorithm is used to establish the mapping model and the high-performance free-form surface parameters are obtained through iterative optimization, the designed lens can achieve the best comprehensive performance under complex lighting constraints.

[0278] The surface of the lens is divided into three sub-areas: the central area, the transition area, and the edge area. The curved shape of each area is independently optimized and controlled. The central area is mainly used to achieve precise light distribution, the transition area is responsible for uniform transition, and the edge area is used for wide-angle scattering. The three areas are linked to efficiently shape the ideal light distribution curve. The regional parametric design increases the degree of freedom of the curved surface, and can evaluate and improve different light output angles in a targeted manner to achieve more refined optical performance control.

[0279] The design of this lens is derived from the mining and learning of a large amount of historical data, and regular empirical knowledge is extracted from existing excellent design cases. The machine learning model can establish the intrinsic relationship between design parameters and lighting performance, providing intelligent guidance for the design. At the same time, it also integrates the latest materials and process frontiers, such as ultra-high transparent materials and nano-scale surface processing, further expanding the design space and continuously iterating better solutions. Therefore, the design of this lens inherits the successful experience of history.

[0280] Compared with the curved light-emitting surface, the flat light-emitting surface is more convenient to clean dust and stains. Curved surfaces are easy to hide dirt and are difficult to clean, but flat surfaces do not have this problem. Therefore, the use of a flat light-emitting surface is conducive to reducing the difficulty of operation and maintenance of the lens. The processing and manufacturing difficulty and cost of a flat light-emitting surface are lower than those of a curved surface, which is more conducive to the large-scale production and promotion and application of lenses. The efficiency and yield of lens production can be improved. The present invention proposes to establish a mapping model between free surface parameters and lighting effects through machine learning, and to optimize the internal curved surface of the lens by region, so that while ensuring the planarity of the light-emitting surface, excellent lighting uniformity and visual comfort can still be achieved. The assembly of the flat light-emitting surface and the lamp is simpler and more firm, and it is also very convenient to replace the lens in subsequent maintenance.

[0281] Example 3

[0282] This embodiment discloses an electronic device, which may include one or more processors and one or more memories. The memories store computer-readable codes, which, when executed by the one or more processors, may execute the N-in-one free-form surface street lamp lens design method that is convenient for operation and maintenance as described above.

[0283] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. A storage device in the electronic device, such as a ROM or a hard disk, can store the N-in-one free-form surface street lamp lens design method provided in the present application that is convenient for operation and maintenance. An N-in-one free-form surface street lamp lens design method that is easy to operate and maintain may, for example, include: determining the lighting requirements of the target lighting scene to generate first lighting requirement data; obtaining historical lens design parameters and historical lighting effect data of multiple LED street lamp lighting projects to generate second historical optimization data; based on the second historical optimization data, a second mapping model M2 between free-form surface parameters and lighting effects is established through a machine learning algorithm; based on the second mapping model and the first lighting requirement data, an initial free-form surface is generated; the initial free-form surface is divided into three sub-areas: a central area, a transition area, and an edge area, the actual light output angle distribution function of each sub-area is calculated, and the actual light output angle distribution function is compared with a preset ideal light output angle distribution function to obtain angle deviation data of each sub-area; according to the angle deviation data, the free-form surface shape is continuously updated by adjusting the coordinates and weights of the free-form surface control points of each sub-area until the second angle deviation data of all sub-areas are lower than the set angle deviation convergence threshold. , and obtain the second optimized free-form surface; based on the second optimized free-form surface, introduce material screening to obtain the third optimized lens solution; simulate and verify the third optimized lens solution and perform test optimization to generate the final N-in-one planar light-emitting LED street lamp lens design solution.

[0284] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0285] The specific implementation modes described above further describe the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A design method for an N-in-one free-form surface street lamp lens that is convenient for operation and maintenance, characterized in that: The method comprises: Determine the lighting demand of the target lighting scene and generate first lighting demand data; obtain historical lens design parameters and historical lighting effect data of multiple LED street lamp lighting projects and generate second historical optimization data; based on the second historical optimization data, establish a second mapping model M2 between free-form surface parameters and lighting effects through a machine learning algorithm; generate an initial free-form surface based on the second mapping model and the first lighting demand data; The initial free-form surface is divided into three sub-areas: a central area, a transition area, and an edge area. The actual light-emitting angle distribution function of each sub-area is calculated, and the actual light-emitting angle distribution function is compared with the preset ideal light-emitting angle distribution function to obtain the angle deviation data of each sub-area. According to the angle deviation data, the coordinates and weights of the free-form surface control points of each sub-area are adjusted to continuously update the shape of the free-form surface to obtain an updated free-form surface, and the second angle deviation data of each sub-area of ​​the updated free-form surface is calculated until the second angle deviation data of all sub-areas are lower than the set angle deviation convergence threshold , obtaining a second optimized free-form surface; based on the second optimized free-form surface, introducing material screening, obtaining a third optimized lens solution; The third optimized lens solution was simulated, verified and tested to generate the final N-in-1 planar light-emitting LED street lamp lens design solution; The generating of the initial free-form surface comprises: Input the first lighting requirement data D1 into the second mapping model M2 to generate a first free-form surface parameter set C1, wherein the first free-form surface parameter set C1 includes the control point coordinates and weights of the free-form surface; Generate a first B-spline free-form surface S1 based on the control point coordinates and weights of the free-form surface in the first free-form surface parameter set C1; The first B-spline free-form surface S1 is used as the initial free-form surface; The obtaining of the second optimized free-form surface comprises: Step S2310, set the number of initialization iterations , set the angle deviation convergence threshold , let the second optimized free-form surface ,in is the initial free-form surface, represents the free surface obtained at the 0th iteration; Step S2320, make , according to The free-form surface obtained by iteration , combined with the angle deviation data of each sub-region, calculate the coordinate offset and weight change of the control points in each sub-region; Step S2330: using the coordinate offset and weight change to align the free-form surface The coordinates and weights of each control point are corrected to obtain the updated free-form surface ; Step S2340, calculating the updated free-form surface The second angle deviation data of each sub-area , Represents the second angle deviation data of sub-area i; determines whether ; If satisfied, let the second optimized free-form surface , the iteration ends; otherwise, return to step S2320 and continue iterative optimization; The calculation of the coordinate offset and weight change of the control points in each sub-area includes: Divide all sub-regions into n grid units, and take the area of ​​sub-region i through n grid units As the weight, calculate the weighted average of the angle deviation of each sub-area ; Indicates The weighted average of the angular deviations of the sub-areas; according to Calculate the The shape adjustment coefficient of the sub-region ; Adjust coefficients according to shape , calculate the control points The coordinate offset of and weight change ; is the first Line The control points of the column.

2. The method for designing an N-in-one free-form surface street lamp lens that is easy to operate and maintain according to claim 1, characterized in that: The generating the first lighting requirement data comprises: Determine the illumination uniformity requirement U0, the glare limitation level requirement G0, the brightness longitudinal uniformity requirement UI and the brightness uniformity requirement L0 of the target lighting scene; use U0, G0, UI and L0 as lighting constraint parameters to generate corresponding first lighting requirement data D1.

3. The method for designing an N-in-one free-form surface street lamp lens that is easy to operate and maintain according to claim 1, characterized in that: The generating of the second historical optimization data comprises: Get the historical lens design parameters of LED street lighting projects; Acquire historical lighting effect data of the LED street lamp lighting project, wherein the historical lighting effect data includes historical illumination uniformity, historical glare limit level, historical brightness longitudinal uniformity and historical brightness uniformity; form a data pair with historical lens design parameters and historical lighting effect data; The first similarity between the historical lighting effect data and the first lighting demand data D1 in the data pair is calculated, and the data pairs whose first similarity is higher than a preset similarity threshold are extracted to generate the second historical optimization data D2.

4. The method for designing an N-in-one free-form surface street lamp lens that is easy to operate and maintain according to claim 1, characterized in that: The dividing of the initial free-form surface into three sub-areas, namely, the central area, the transition area and the edge area, comprises: Taking the center of the lens as the origin, a three-dimensional rectangular coordinate system is established, which is marked as the free-form surface coordinate system. The function expression of the initial free-form surface in the free-form surface coordinate system is z'=f ; is the radius of the sphere, is the zenith angle, is the azimuth angle; z' is the function expression of the initial free-form surface in the free-form surface coordinate system; use represents the boundary of the central area, which will satisfy The surface is divided into the central area; ρ1 represents the spherical radius of the central area boundary, θ1 represents the zenith angle of the central area boundary, and φ1 represents the azimuth angle of the central area boundary; use represents the boundary of the transition zone, which will satisfy The surface is divided into a transition zone; ρ2 represents the spherical radius of the transition zone boundary, θ2 represents the zenith angle of the transition zone boundary, and φ2 represents the azimuth angle of the transition zone boundary; use represents the boundary of the marginal area, which will satisfy The surface is divided into edge areas; ρ3 represents the spherical radius of the edge area boundary, θ3 represents the zenith angle of the edge area boundary, and φ3 represents the azimuth angle of the edge area boundary.

5. The method for designing an N-in-one free-form surface street lamp lens that is easy to operate and maintain according to claim 1, characterized in that: The obtaining of the angle deviation data of each sub-area comprises: Randomly generate N2 rays, marked as N2 incident rays, and inject the incident rays into the initial free-form surface; each incident ray is represented by the parametric equation of a spatial line, with its starting point being a random sampling point on the initial free-form surface and its direction being a randomly sampled unit vector on the unit sphere, marked as the direction vector of the incident ray; N2 is a positive integer; For each incident light ray, determine the incident light ray I j Is the incident angle less than the critical angle of total reflection? j The incident angle is less than the critical angle of total reflection, then according to the incident light I j The direction vector of the incident light ray I is solved using Snell's law. j The direction vector of the corresponding refracted light; I j represents the jth incident light ray; 1≤j≤N2; According to the direction vector of the refracted light, determine whether the refracted light can be emitted through the lens; if it can be emitted, record the emission angle ψ of the refracted light; The emission angle ψ of all refracted light in each sub-region is counted, and the probability distribution histogram of the emission angle ψ is drawn. The probability distribution histogram is normalized to obtain the actual light emission angle distribution function g(ψ) of each sub-region. i ; g(ψ) i represents the actual light output angle distribution function of sub-region i, i is the index of the sub-region, 1≤i≤3; The actual light output angle distribution function g(ψ) of sub-region i is i The ideal light output angle distribution function f(ψ) corresponding to sub-area i i Compare and calculate the angle deviation data of sub-area i .

6. A lens, characterized in that: The lens is designed using the N-in-one free-form surface street lamp lens design method that is easy to operate and maintain as described in any one of claims 1-5; the lens includes three sub-areas: a central area, a transition area, and an edge area; and the light-emitting surface of the lens is a plane.

Images