Optical glass data prediction method and device, electronic device and storage medium

Abstract

The optical glass data prediction method and device, electronic device and storage medium provided in the present application relate to the technical field of optical glass data prediction. In the present application, first, sample formula data and transmittance labels are obtained; secondly, transmittance prediction is performed on the sample formula data through a candidate prediction network, and transmittance prediction data is output, and the sample formula data including the superposition effect between each formula component is predicted through the candidate prediction network, and the superposition effect characterization value corresponding to the sample formula data is output; then, based on the transmittance label, transmittance prediction data and superposition effect characterization value, the parameters of the candidate prediction network are optimized to form a target prediction network; finally, the target optical glass data is predicted using the target prediction network, and the predicted optical glass data is output. Based on the above content, the problem of high cost of determining optical glass data existing in the prior art can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of optical glass data prediction, and in particular to an optical glass data prediction method and device, electronic equipment, and storage medium. Background Art

[0002] Functional glass boasts high optical transmittance, high chemical and physical uniformity (structure and performance), and specific and precise optical constants. It has crucial applications in various optical instruments, information communications, and luminous displays, forming the foundation and a crucial component of the optoelectronic information technology industry. Functional glass, such as thermal insulation glass, laser glass, and photochromic glass, possesses unique optical properties. Transmittance is a key indicator of its optical performance, making research on the relationship between its composition and transmittance crucial.

[0003] In existing technology, glass formulations can be explored through trial-and-error experiments, with statistical analysis of glass performance (such as transmittance) to find the optimal component combination. However, the number of experimental factors and factor level values ​​in a formula affects the number of experimental formulas, leading to an explosion in the amount of glass performance data for each formula component combination. This results in long formula development cycles and high costs. Summary of the Invention

[0004] In view of this, the purpose of the present application is to provide an optical glass data prediction method and device, an electronic device and a storage medium, so as to improve the problem of high cost of determining optical glass data in the prior art.

[0005] To achieve the above objectives, this application adopts the following technical solutions:

[0006] An optical glass data prediction method, comprising:

[0007] Obtaining sample formula data corresponding to the optical glass and a transmittance label corresponding to the optical glass, wherein the sample formula data includes each formula component and the component content of each formula component;

[0008] Using a pre-constructed candidate prediction network, transmittance prediction is performed on the sample formula data, and transmittance prediction data corresponding to the sample formula data is output. The sample formula data, including the superposition effect between each formula component, is predicted using the candidate prediction network, and a superposition effect representation value corresponding to the sample formula data is output, wherein the superposition effect representation value is used to reflect the degree of matching between each formula component and a target linear constraint relationship, and the target linear constraint relationship is at least used to reflect that the sum of the molar percentages of the component contents corresponding to each formula component is 100%;

[0009] Optimizing parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data, and the superposition effect representation value to form a target prediction network;

[0010] The target prediction network is used to predict target optical glass data, and predicted optical glass data corresponding to the target optical glass data is output, wherein when the target optical glass data includes formula data, the predicted optical glass data includes transmittance data; when the target optical glass data includes transmittance data, the predicted optical glass data includes formula data.

[0011] In a preferred embodiment of the present application, in the above-mentioned optical glass data prediction method, the steps of predicting transmittance of the sample formula data using a pre-constructed candidate prediction network, outputting transmittance prediction data corresponding to the sample formula data, and predicting the sample formula data including the superposition effect between each formula component using the candidate prediction network, and outputting a superposition effect representation value corresponding to the sample formula data, include:

[0012] Performing feature mining on the sample recipe data through a feature mining unit included in a pre-built candidate prediction network, and outputting sample recipe features corresponding to the sample recipe data;

[0013] The first output unit of the candidate prediction network predicts and outputs the sample formula feature to obtain transmittance prediction data corresponding to the sample formula data;

[0014] matrixing each recipe component included in the sample recipe data by a projection unit included in the candidate prediction network to form at least one projection matrix corresponding to the sample recipe data, wherein each row of each projection matrix corresponds to a linear constraint relationship between each recipe component;

[0015] The at least one projection matrix is ​​processed by a second output unit included in the candidate prediction network to output a superposition effect representation value corresponding to the sample recipe data.

[0016] In a preferred embodiment of the present application, in the optical glass data prediction method, the step of processing the at least one projection matrix through the second output unit included in the candidate prediction network and outputting the superposition effect representation value corresponding to the sample formula data includes:

[0017] A weighted summation calculation is performed on the at least one projection matrix using the weight coefficients of the projection matrices in the second output unit included in the candidate prediction network, and a superposition effect representation value corresponding to the sample recipe data is output, wherein the weighted summation calculation formula includes:

[0018] ;

[0019] in, is the representation value of the superposition effect, represents the first linear constraint relationship in the target linear constraint relationship, represents the second linear constraint relationship in the target linear constraint relationship, are the weight coefficients corresponding to the projection matrix, 、 The projection matrix composed of each recipe component, 、 They are the configured constants respectively.

[0020] In a preferred embodiment of the present application, in the optical glass data prediction method, the step of optimizing the parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data, and the superposition effect characterization value to form a target prediction network includes:

[0021] determining a first error index of the candidate prediction network based on a difference between the transmittance label and the transmittance prediction data;

[0022] Determining a second error indicator of the candidate prediction network based on a difference between the superposition effect representation value and a predetermined superposition effect target value;

[0023] Based on the first error indicator and the second error indicator, the parameters of the candidate prediction network are optimized so that the first error indicator converges and the second error indicator converges, and when converged, the corresponding target prediction network is determined based on the candidate prediction network.

[0024] In a preferred embodiment of the present application, in the above optical glass data prediction method, the step of using the target prediction network to predict target optical glass data and outputting predicted optical glass data corresponding to the target optical glass data includes:

[0025] Performing feature mining on the target optical glass data by a feature mining unit included in the target prediction network, and outputting target formula features corresponding to the target optical glass data, wherein the target optical glass data belongs to formula data;

[0026] The target formula feature is predicted and outputted by a first output unit included in the target prediction network to obtain transmittance prediction data corresponding to the target optical glass data, wherein the transmittance prediction data serves as predicted optical glass data corresponding to the target optical glass data;

[0027] Matrixing the recipe components included in the target optical glass data by a projection unit included in the target prediction network to form at least one projection matrix corresponding to the target optical glass data;

[0028] The target prediction network includes a second output unit, which processes at least one projection matrix corresponding to the target optical glass data and outputs a superposition effect representation value corresponding to the target optical glass data. The superposition effect representation value is used to reflect the degree of matching between each formula component in the target optical glass data and the target linear constraint relationship, and the degree of matching is positively correlated with the reliability of the predicted optical glass data corresponding to the target optical glass data.

[0029] In a preferred embodiment of the present application, in the above optical glass data prediction method, the step of using the target prediction network to predict target optical glass data and outputting predicted optical glass data corresponding to the target optical glass data includes:

[0030] For each formula component included in the sample formula data, determining the order of magnitude of the component content corresponding to the formula component, and determining a value interval corresponding to the formula component based on the order of magnitude, and performing random value processing based on the value interval to obtain multiple random component contents corresponding to the formula component, wherein each of the random component contents belongs to a corresponding value interval;

[0031] Combining multiple random component contents corresponding to each recipe component included in the sample recipe data to form multiple random recipe data, wherein each random recipe data includes each recipe component and the random component content of each recipe component;

[0032] The target prediction network is used to determine a random formula data that best matches the target optical glass data among the multiple random formula data, and the random formula data is determined as the predicted optical glass data corresponding to the target optical glass data, wherein the target optical glass data is transmittance data.

[0033] In a preferred embodiment of the present application, in the optical glass data prediction method, the step of using the target prediction network to determine, from the plurality of random formula data, a random formula data that best matches the target optical glass data, and determining the random formula data as the predicted optical glass data corresponding to the target optical glass data includes:

[0034] For each of the plurality of random recipe data, feature mining is performed on the random recipe data by a feature mining unit included in the target prediction network, and a random recipe feature corresponding to the random recipe data is outputted. The random recipe feature is predicted and outputted by a first output unit included in the target prediction network to obtain transmittance prediction data corresponding to the random recipe data. Furthermore, each recipe component included in the random recipe data is matrixed by a projection unit included in the target prediction network to form at least one projection matrix corresponding to the random recipe data. The at least one projection matrix corresponding to the random recipe data is processed by a second output unit included in the target prediction network to output a superposition effect representation value corresponding to the random recipe data, wherein the superposition effect representation value is used to reflect the degree of matching between each recipe component in the random recipe data and the target linear constraint relationship, and the degree of matching is positively correlated with the reliability of the transmittance prediction data corresponding to the random recipe data.

[0035] Based on the transmittance prediction data and the corresponding superposition effect characterization value corresponding to each of the random formula data, a random formula data that best matches the target optical glass data is determined among the multiple random formula data, and the random formula data is determined as the predicted optical glass data corresponding to the target optical glass data.

[0036] The present application also provides an optical glass data prediction device, comprising:

[0037] A sample label acquisition module is used to acquire sample formula data corresponding to the optical glass and a transmittance label corresponding to the optical glass, wherein the sample formula data includes each formula component and the component content of each formula component;

[0038] a sample transmittance prediction module, configured to predict transmittance of the sample formula data using a pre-constructed candidate prediction network, output transmittance prediction data corresponding to the sample formula data, and predict the sample formula data, including the superposition effect between each formula component, using the candidate prediction network, and output a superposition effect representation value corresponding to the sample formula data, wherein the superposition effect representation value is used to reflect the degree of match between each formula component and a target linear constraint relationship, wherein the target linear constraint relationship is at least used to reflect that the sum of the molar percentages of the component contents corresponding to each formula component is 100%;

[0039] A prediction network optimization module, configured to optimize the parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data, and the superposition effect characterization value to form a target prediction network;

[0040] An optical glass data prediction module is used to use the target prediction network to predict target optical glass data and output predicted optical glass data corresponding to the target optical glass data, wherein when the target optical glass data includes formula data, the predicted optical glass data includes transmittance data; when the target optical glass data includes transmittance data, the predicted optical glass data includes formula data.

[0041] Based on the above, the present application further provides an electronic device, including:

[0042] memory for storing computer programs;

[0043] The processor connected to the memory is used to execute the computer program stored in the memory to implement the above-mentioned optical glass data prediction method.

[0044] On the basis of the above, the present application further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is run, each step of the above-mentioned optical glass data prediction method is executed.

[0045] The optical glass data prediction method and device, electronic device and storage medium provided by the present application first obtain sample formula data and transmittance labels; secondly, transmittance prediction is performed on the sample formula data through a candidate prediction network, transmittance prediction data is output, and the sample formula data including the superposition effect between the various formula components is predicted through the candidate prediction network, and the superposition effect characterization value corresponding to the sample formula data is output; then, based on the transmittance label, transmittance prediction data and superposition effect characterization value, the parameters of the candidate prediction network are optimized to form a target prediction network; finally, the target prediction network is used to predict the target optical glass data and output the predicted optical glass data. Based on the above content, since the candidate prediction network can be trained using samples, the corresponding prediction can be directly made based on the target prediction network formed by training. Compared with the conventional technical solutions in the prior art that require experimental trial and error, the trial and error cost can be effectively reduced, and the problem of high cost of determining optical glass data in the prior art can be improved. In addition, after the target prediction network is formed through training, prediction can be made directly, and the efficiency can be greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

[0047] Figure 1 This is a structural block diagram of the electronic device provided in an embodiment of the present application.

[0048] Figure 2A schematic flow chart of the optical glass data prediction method provided in an embodiment of the present application.

[0049] Figure 3 This is a network architecture diagram of the candidate prediction network provided in an embodiment of the present application.

[0050] Figure 4 A block diagram of an optical glass data prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0052] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in the present application without creative work are within the scope of protection of the present application.

[0053] like Figure 1 As shown, an embodiment of the present application provides an electronic device, wherein the electronic device may include a memory, a processor, and an optical glass data prediction device.

[0054] Specifically, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, the memory and the processor can be electrically connected via one or more communication buses or signal lines. The optical glass data prediction device includes at least one software function module stored in the memory in the form of software or firmware. The processor is used to execute an executable computer program stored in the memory, such as the software function module and computer program included in the optical glass data prediction device, to implement the optical glass data prediction method provided in the embodiments of the present application.

[0055] Optionally, the memory may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0056] Furthermore, the processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0057] I understand. Figure 1 The structure shown is only for illustration, and the electronic device may also include Figure 1 More or fewer components than shown, or with Figure 1 The different configurations shown may, for example, further include a communication unit for exchanging information with other devices.

[0058] Combine Figure 2 The present application also provides an optical glass data prediction method applicable to the electronic device described above, wherein the steps defined in the process related to the optical glass data prediction method can be implemented by the electronic device.

[0059] The following will Figure 2 The specific process shown is explained in detail.

[0060] Step S110 , obtaining sample formula data corresponding to the optical glass and a transmittance label corresponding to the optical glass.

[0061] In an embodiment of the present application, the electronic device can obtain sample formula data corresponding to the optical glass and a transmittance label corresponding to the optical glass (i.e., the actual transmittance). The sample formula data includes each formula component (such as iron oxide, titanium oxide, lithium oxide, potassium oxide, sodium oxide, and silicon oxide) and the component content of each formula component (such as 0.5%, 1%, etc., in molar percentage).

[0062] In step S120, transmittance prediction is performed on the sample formula data through a pre-constructed candidate prediction network, and transmittance prediction data corresponding to the sample formula data is output. The sample formula data including the superposition effect between each formula component is predicted through the candidate prediction network, and a superposition effect characterization value corresponding to the sample formula data is output.

[0063] In an embodiment of the present application, after obtaining the sample formula data and the transmittance label, the electronic device can predict the transmittance of the sample formula data using a pre-built candidate prediction network (which can be a convolutional neural network), output transmittance prediction data corresponding to the sample formula data (i.e., the predicted transmittance), and use the candidate prediction network to predict the sample formula data, including the additive effect between each formula component, and output a additive effect representation value corresponding to the sample formula data. The additive effect representation value is used to reflect the degree of match between each formula component and a target linear constraint relationship. The target linear constraint relationship is at least used to reflect that the sum of the molar percentages of the component contents corresponding to each formula component is 100%. Exemplarily, it can include only one linear constraint relationship or multiple linear constraints. For example, it can also be used to reflect the relationship between the component contents of two components. For example, for iron oxide and titanium oxide with synergistic effects, the sum of the molar percentages of the component contents can be less than 2%. It should be noted that the transmittance can refer to the transmittance at the H line 404.7nm, or, in other embodiments, it can also refer to the transmittance under other indicator parameters.

[0064] Step S130 : Optimizing the parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data, and the superposition effect representation value to form a target prediction network.

[0065] In an embodiment of the present application, after obtaining the transmittance prediction data and the superposition effect characterization value, the electronic device can optimize the parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data and the superposition effect characterization value to form a target prediction network. For example, the candidate prediction network after parameter optimization can be used as the target prediction network, or a target prediction network can be constructed based on the optimized parameters.

[0066] Step S140 , using the target prediction network to predict target optical glass data, and outputting predicted optical glass data corresponding to the target optical glass data.

[0067] In an embodiment of the present application, after forming the target prediction network, the electronic device can use the target prediction network to predict target optical glass data and output predicted optical glass data corresponding to the target optical glass data. When the target optical glass data includes recipe data, the predicted optical glass data includes transmittance data; and when the target optical glass data includes transmittance data, the predicted optical glass data includes recipe data. In other words, the target prediction network can be used not only to predict the corresponding transmittance based on the recipe, but also to predict the corresponding recipe based on the transmittance. This significantly improves the utilization of the target prediction network and maximizes resource utilization.

[0068] Based on the above, since samples can be used to train candidate prediction networks, predictions can be made directly based on the target prediction network formed through training. Compared to conventional technical solutions in the prior art that require trial and error, this can effectively reduce trial and error costs and address the high cost of determining optical glass data that exists in the prior art. Furthermore, after the target prediction network is trained, predictions can be made directly, significantly improving efficiency.

[0069] First, regarding step S110, it should be noted that multiple sample formula data sets may be obtained. To ensure data reliability, these multiple sample formula data sets can be preprocessed, such as by checking for missing values ​​and removing outliers using the 3σ (three times the standard deviation) principle. Furthermore, correlation analysis can be performed on the various formula components in the sample formula data to identify those with a close relationship to transmittance. Furthermore, the various components can be standardized to minimize dimensional effects between them.

[0070] Secondly, it should be noted that for step S120 , the specific method of performing corresponding predictions through the candidate prediction network is not limited and can be selected according to actual needs.

[0071] For example, in an alternative embodiment, in order to enable effective prediction through the candidate prediction network so that the reliability of the candidate optimization process can be guaranteed, the above-mentioned step S120 can further include step S121, step S122, step S123 and step S124, and the specific content of each step is described below.

[0072] Step S121 : performing feature mining on the sample recipe data through a feature mining unit included in a pre-built candidate prediction network, and outputting sample recipe features corresponding to the sample recipe data.

[0073] In the embodiment of the present application, the feature mining unit included in the pre-built candidate prediction network can be used to perform feature mining on the sample recipe data and output the sample recipe features corresponding to the sample recipe data. For example, the feature mining unit can include an input layer, a convolution layer, an activation layer, and a fully connected layer, such as Figure 3 shown.

[0074] Step S122 : predicting and outputting the sample recipe feature through the first output unit included in the candidate prediction network to obtain transmittance prediction data corresponding to the sample recipe data.

[0075] In an embodiment of the present application, after the sample formula features are mined, the sample formula features can be predicted and output through the first output unit included in the candidate prediction network to obtain transmittance prediction data corresponding to the sample formula data. For example, the first output unit can include a linear regression function for outputting transmittance prediction data.

[0076] Step S123 : matrixing each recipe component included in the sample recipe data by using the projection unit included in the candidate prediction network to form at least one projection matrix corresponding to the sample recipe data.

[0077] In an embodiment of the present application, the projection unit included in the candidate prediction network can also be used to matrix the recipe components included in the sample recipe data to form at least one projection matrix corresponding to the sample recipe data. Each row of each projection matrix corresponds to a linear constraint relationship between the recipe components. Specifically, a linear constrained optimization projection method (Linear Constrained Optimization Projection Method) can be used as an example. For example, if there are n recipe components, the relationship between the recipe components may be affected by linear constraints, such as , G is an m*n matrix, representing the coefficients of the constraint conditions, x is an n-dimensional column vector, representing the content of each formula component, b is an m-dimensional column vector, representing the right-hand constant of the constraint, where G and b can be configured according to actual needs, so there is , further, there is =0, where, according to the analysis In a certain glass system, it is fixed. For at least one linear constraint relationship, at least one projection matrix can be formed.

[0078] Step S124 : Processing the at least one projection matrix through the second output unit included in the candidate prediction network, and outputting a superposition effect representation value corresponding to the sample recipe data.

[0079] In this embodiment of the present application, after forming the at least one projection matrix, the at least one projection matrix can be processed by a second output unit included in the candidate prediction network to output a representation of the additive effect corresponding to the sample recipe data. In other words, the linear constraints represented by each projection matrix are fused to determine the matching of various content relationships between recipe components, thereby obtaining the corresponding representation of the additive effect.

[0080] It is understood that, in the above step S124, the specific method of determining the superposition effect characterization value is not limited and can be selected according to actual needs. For example, in an alternative embodiment, in order to ensure that multiple linear constraint relationships can be fully characterized, the above step S124 can further include the following specific implementation process:

[0081] The at least one projection matrix may be weighted and summed using the weight coefficients of the projection matrices in the second output unit of the candidate prediction network to output a superposition effect representation value corresponding to the sample recipe data, wherein the weighted summation calculation formula includes:

[0082] ;

[0083] in, is the representation value of the superposition effect, represents the first linear constraint relationship in the target linear constraint relationship, represents the second linear constraint relationship in the target linear constraint relationship (in other implementations, more linear constraints may be included, such as the third linear constraint relationship and the fourth linear constraint relationship), are the weight coefficients corresponding to the projection matrix, 、 The projection matrix of each recipe component (for example, combined with the relevant description above, q = , 、 They are the configuration constants (i.e. In different linear constraint relationships). Based on this, in theory, =0, then, ideally =0, = 0. Therefore, The ideal value should also be 0.

[0084] Thirdly, it should be noted that for step S130 , the specific method for optimizing the parameters of the candidate prediction network is not limited and can be selected according to actual needs.

[0085] For example, in an alternative embodiment, in order to improve the reliability of the optimized target prediction network, the above step S130 may further include the following specific implementation process:

[0086] First, based on the difference between the transmittance label and the transmittance prediction data, a first error indicator of the candidate prediction network, such as mean square error (MSE), can be determined;

[0087] Secondly, based on the difference between the superposition effect representation value and the predetermined superposition effect target value (such as difference calculation), the second error index of the candidate prediction network can be determined; illustratively, The ideal value of should be 0, so the target value of the superposition effect can be 0;

[0088] Then, the parameters of the candidate prediction network can be optimized based on the first error indicator and the second error indicator, so that the first error indicator converges and the second error indicator converges. When convergence occurs, the corresponding target prediction network is determined based on the candidate prediction network. Exemplarily, the optimization can be performed based on the first error indicator and the second error indicator respectively. Specifically, the parameters of the feature mining unit can be optimized in the direction of reducing the first error indicator. In addition, the parameters in the projection unit and the second output unit, such as the weight coefficients mentioned above, can be optimized in the direction of reducing the second error indicator.

[0089] Fourthly, it should be noted that for step S140 , the specific method of predicting the target optical glass data is not limited and can be selected according to actual needs.

[0090] For example, in an alternative embodiment, in order to achieve reliable prediction of transmittance through the target prediction network, the above step S140 may further include the following:

[0091] First, the target optical glass data may be subjected to feature mining by a feature mining unit included in the target prediction network, and target formula features corresponding to the target optical glass data may be output, wherein the target optical glass data belongs to formula data;

[0092] Secondly, the target formula feature can be predicted and output by the first output unit included in the target prediction network to obtain transmittance prediction data corresponding to the target optical glass data, wherein the transmittance prediction data serves as the predicted optical glass data corresponding to the target optical glass data;

[0093] Then, the projection unit included in the target prediction network can be used to matrix the formulation components included in the target optical glass data to form at least one projection matrix corresponding to the target optical glass data;

[0094] Finally, the target prediction network processes at least one projection matrix corresponding to the target optical glass data through a second output unit, and outputs a superposition effect representation value corresponding to the target optical glass data, wherein the superposition effect representation value is used to reflect the degree of matching between each formula component in the target optical glass data and the target linear constraint relationship, and the matching degree is positively correlated with the reliability of the predicted optical glass data corresponding to the target optical glass data; that is, the closer the superposition effect representation value is to 0, the more the relationship between the content of each formula component matches the target linear constraint relationship, thus indicating that the reliability of the content of each formula component is higher and the predicted transmittance prediction data is more reliable. Conversely, the further the superposition effect representation value is from 0, the less the relationship between the content of each formula component matches the target linear constraint relationship, thus indicating that the reliability of the content of each formula component is lower and the predicted transmittance prediction data is less reliable.

[0095] For example, in an alternative embodiment, in order to achieve reliable prediction of the recipe through the target prediction network, the above-mentioned step S140 may further include step S141, step S142 and step S143, and the specific content of each step is described as follows.

[0096] Step S141, for each formula component included in the sample formula data, determine the order of magnitude of the component content corresponding to the formula component, and determine the value range corresponding to the formula component based on the order of magnitude, and perform random value processing based on the value range to obtain multiple random component contents corresponding to the formula component.

[0097] In the embodiment of the present application, for each formula component included in the sample formula data, the order of magnitude of the component content corresponding to the formula component can be determined (for a formula component, the corresponding order of magnitude can also be determined based on the component content of the formula component in multiple sample formula data), and the value range corresponding to the formula component can be determined based on the order of magnitude (such as [0.8 ~1.2 ] ,in, represents an order of magnitude), and performing random value processing based on the value interval to obtain multiple random component contents corresponding to the formula components, such as randomly selecting 200 random component contents. Each of the random component contents belongs to a corresponding value interval.

[0098] Step S142 : performing combination processing based on the contents of multiple random components corresponding to each formula component included in the sample formula data to form multiple random formula data.

[0099] In an embodiment of the present application, after obtaining the multiple random component contents corresponding to each formula component, a combination process can be performed based on the multiple random component contents corresponding to each formula component included in the sample formula data to form multiple random formula data. Each of the random formula data includes each formula component and the random component content of each formula component.

[0100] Step S143 , using the target prediction network, determines a random formula data that best matches the target optical glass data from the multiple random formula data, and determines the random formula data as the predicted optical glass data corresponding to the target optical glass data.

[0101] In an embodiment of the present application, after obtaining multiple random formula data, the target prediction network can be used to determine, from the multiple random formula data, a random formula data that best matches the target optical glass data, and determine this random formula data as the predicted optical glass data corresponding to the target optical glass data. Where the target optical glass data is transmittance data, the random formula data corresponding to the transmittance that best matches the transmittance data is determined.

[0102] It is understood that, in the above step S143, the specific method of determining the random formula data that best matches the target optical glass data from the multiple random formula data is not limited and can be selected according to actual needs. For example, in an alternative embodiment, in order to ensure the reliability of the determined formula data, the above step S143 may further include:

[0103] First, for each random recipe data among the multiple random recipe data, feature mining is performed on the random recipe data by a feature mining unit included in the target prediction network, and a random recipe feature corresponding to the random recipe data is output. The random recipe feature is predicted and output by a first output unit included in the target prediction network to obtain transmittance prediction data corresponding to the random recipe data. Furthermore, each recipe component included in the random recipe data is matrixed by a projection unit included in the target prediction network to form at least one projection matrix corresponding to the random recipe data. The at least one projection matrix corresponding to the random recipe data is processed by a second output unit included in the target prediction network to output a superposition effect representation value corresponding to the random recipe data, wherein the superposition effect representation value is used to reflect the degree of matching between each recipe component in the random recipe data and the target linear constraint relationship, and the degree of matching is positively correlated with the reliability of the transmittance prediction data corresponding to the random recipe data.

[0104] Secondly, based on the transmittance prediction data and the corresponding superposition effect characterization value corresponding to each of the random formula data, a random formula data that best matches the target optical glass data can be determined among the multiple random formula data, and the random formula data can be determined as the predicted optical glass data corresponding to the target optical glass data; illustratively, some random formula data with relatively small superposition effect characterization values ​​(such as the difference between the superposition effect target value "0" does not exceed a threshold, etc.) can be selected first, and then, among these random formula data, a random formula data with the smallest difference from the target optical glass data can be selected.

[0105] Combine Figure 4 The present application also provides an optical glass data prediction device applicable to the aforementioned electronic device. The optical glass data prediction device may include a sample label acquisition module, a sample transmittance prediction module, a prediction network optimization module, and an optical glass data prediction module.

[0106] In detail, the sample label acquisition module can be used to obtain the sample formula data corresponding to the optical glass and the transmittance label corresponding to the optical glass, wherein the sample formula data includes each formula component and the component content of each formula component. In the embodiment of the present application, the sample label acquisition module can be used to execute Figure 2 Regarding step S110 shown, the relevant contents of the sample label acquisition module can refer to the above description of step S110.

[0107] In detail, the sample transmittance prediction module can be used to predict the transmittance of the sample formula data through a pre-built candidate prediction network, output the transmittance prediction data corresponding to the sample formula data, and predict the sample formula data including the superposition effect between each formula component through the candidate prediction network, and output the superposition effect characterization value corresponding to the sample formula data, wherein the superposition effect characterization value is used to reflect the degree of matching between each formula component and the target linear constraint relationship, and the target linear constraint relationship is at least used to reflect that the sum of the molar percentages of the component contents corresponding to each formula component is 100%. In an embodiment of the present application, the sample transmittance prediction module can be used to perform Figure 2 Regarding step S120 shown, the relevant contents of the sample transmittance prediction module can refer to the above description of step S120.

[0108] In detail, the prediction network optimization module can be used to optimize the parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data and the superposition effect characterization value to form a target prediction network. In the embodiment of the present application, the prediction network optimization module can be used to perform Figure 2 As shown in step S130, for the relevant content of the prediction network optimization module, reference can be made to the above description of step S130.

[0109] In detail, the optical glass data prediction module can be used to predict the target optical glass data using the target prediction network, and output the predicted optical glass data corresponding to the target optical glass data, wherein when the target optical glass data includes formula data, the predicted optical glass data includes transmittance data, and when the target optical glass data includes transmittance data, the predicted optical glass data includes formula data. In the embodiment of the present application, the optical glass data prediction module can be used to perform Figure 2 Regarding step S140 shown, the relevant contents of the optical glass data prediction module can refer to the above description of step S140.

[0110] In an embodiment of the present application, corresponding to the optical glass data prediction method applied to the electronic device, a computer-readable storage medium is also provided. The computer-readable storage medium stores a computer program that, when executed, executes the various steps of the optical glass data prediction method. The steps executed by the aforementioned computer program are not described in detail here; reference is made to the aforementioned explanation of the optical glass data prediction method.

[0111] In summary, the optical glass data prediction method and device, electronic device and storage medium provided by the present application first obtain sample formula data and transmittance labels; secondly, transmittance prediction is performed on the sample formula data through a candidate prediction network, transmittance prediction data is output, and the sample formula data including the superposition effect between each formula component is predicted through the candidate prediction network, and the superposition effect characterization value corresponding to the sample formula data is output; then, based on the transmittance label, transmittance prediction data and superposition effect characterization value, the parameters of the candidate prediction network are optimized to form a target prediction network; finally, the target prediction network is used to predict the target optical glass data and output the predicted optical glass data. Based on the above content, since the candidate prediction network can be trained using samples, the corresponding prediction can be directly made based on the target prediction network formed by training. Compared with the conventional technical solutions in the prior art that require experimental trial and error, the trial and error cost can be effectively reduced, and the problem of high cost of determining optical glass data in the prior art can be improved. In addition, after the target prediction network is formed through training, prediction can be made directly, and the efficiency can be fully improved, which also improves the problem of long experimental trial and error cycle in the prior art.

[0112] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0113] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0114] If the functions are implemented in the form of software modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. It should be noted that, in this document, the terms "comprise," "include," or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. Without further constraints, an element defined by the phrase "comprises a..." does not preclude the existence of additional identical elements in the process, method, article or apparatus that includes the element.

[0115] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for predicting optical glass data, characterized in that: include: Acquire sample formula data corresponding to the optical glass and a transmittance label corresponding to the optical glass, wherein the sample formula data includes each formula component and the component content of each formula component; The transmittance of the sample formula data is predicted by a pre-constructed candidate prediction network, and transmittance prediction data corresponding to the sample formula data is outputted. The sample formula data including the superposition effect between the formula components is predicted by the candidate prediction network, and a superposition effect characterization value corresponding to the sample formula data is outputted, wherein the superposition effect characterization value is used to reflect the matching degree between the formula components and the target linear constraint relationship, and the target linear constraint relationship is at least used to reflect that the sum of the molar percentages of the component contents corresponding to the formula components is 100%; Based on the transmittance label, the transmittance prediction data and the superposition effect characterization value, optimizing the parameters of the candidate prediction network to form a target prediction network; Utilizing the target prediction network, predicting target optical glass data, and outputting predicted optical glass data corresponding to the target optical glass data, wherein when the target optical glass data includes formula data, the predicted optical glass data includes transmittance data, and when the target optical glass data includes transmittance data, the predicted optical glass data includes formula data; The steps of predicting transmittance of the sample formula data by using a pre-constructed candidate prediction network, outputting transmittance prediction data corresponding to the sample formula data, and predicting the sample formula data including the superposition effect between each formula component by using the candidate prediction network, and outputting the superposition effect characterization value corresponding to the sample formula data include: Performing feature mining on the sample recipe data through a feature mining unit included in a pre-built candidate prediction network, and outputting sample recipe features corresponding to the sample recipe data; The sample formula feature is predicted and outputted by a first output unit included in the candidate prediction network to obtain transmittance prediction data corresponding to the sample formula data; Matrixing the recipe components included in the sample recipe data by a projection unit included in the candidate prediction network to form at least one projection matrix corresponding to the sample recipe data, wherein each row of each projection matrix corresponds to a linear constraint relationship between the recipe components; The at least one projection matrix is ​​processed by a second output unit included in the candidate prediction network to output a superposition effect representation value corresponding to the sample recipe data.

2. The optical glass data prediction method according to claim 1, characterized in that: The step of processing the at least one projection matrix through the second output unit included in the candidate prediction network and outputting the superposition effect representation value corresponding to the sample formula data includes: The at least one projection matrix is ​​weighted and summed by the weight coefficients of each projection matrix in the second output unit included in the candidate prediction network, and the superposition effect characterization value corresponding to the sample formula data is output, wherein the weighted summation calculation formula includes: ； in, is the characterization value of the superposition effect, represents the first linear constraint relationship in the target linear constraint relationship, represents the second linear constraint relationship in the target linear constraint relationship, are the weight coefficients corresponding to the projection matrix, , are the projection matrices of each recipe component, , They are the configured constants respectively.

3. The optical glass data prediction method according to claim 1, characterized in that: The step of optimizing the parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data and the superposition effect characterization value to form a target prediction network includes: Determining a first error index of the candidate prediction network based on a difference between the transmittance label and the transmittance prediction data; Determining a second error indicator of the candidate prediction network based on a difference between the stacking effect characterization value and a predetermined stacking effect target value; Based on the first error indicator and the second error indicator, the parameters of the candidate prediction network are optimized so that the first error indicator converges and the second error indicator converges, and when converged, a corresponding target prediction network is determined based on the candidate prediction network.

4. The optical glass data prediction method according to any one of claims 1 to 3, characterized in that: The step of using the target prediction network to predict target optical glass data and outputting predicted optical glass data corresponding to the target optical glass data comprises: Performing feature mining on the target optical glass data through a feature mining unit included in the target prediction network, and outputting target formula features corresponding to the target optical glass data, wherein the target optical glass data belongs to formula data; The target formula feature is predicted and outputted by the first output unit included in the target prediction network to obtain transmittance prediction data corresponding to the target optical glass data, wherein the transmittance prediction data is used as the predicted optical glass data corresponding to the target optical glass data; Matrixing the formula components included in the target optical glass data through a projection unit included in the target prediction network to form at least one projection matrix corresponding to the target optical glass data; Through the second output unit included in the target prediction network, at least one projection matrix corresponding to the target optical glass data is processed, and a superposition effect characterization value corresponding to the target optical glass data is output, wherein the superposition effect characterization value is used to reflect the degree of matching between each formula component in the target optical glass data and the target linear constraint relationship, and the degree of matching is positively correlated with the reliability of the predicted optical glass data corresponding to the target optical glass data.

5. The optical glass data prediction method according to any one of claims 1 to 3, characterized in that: The step of using the target prediction network to predict target optical glass data and outputting predicted optical glass data corresponding to the target optical glass data comprises: For each formula component included in the sample formula data, determine the order of magnitude of the component content corresponding to the formula component, and determine the value interval corresponding to the formula component based on the order of magnitude, and perform random value processing based on the value interval to obtain multiple random component contents corresponding to the formula component, wherein each of the random component contents belongs to the corresponding value interval; Combining and processing multiple random component contents corresponding to each formula component included in the sample formula data to form multiple random formula data, wherein each of the random formula data includes each formula component and the random component content of each formula component; The target prediction network is used to determine a random formula data that best matches the target optical glass data among the multiple random formula data, and the random formula data is determined as the predicted optical glass data corresponding to the target optical glass data, wherein the target optical glass data belongs to transmittance data.

6. The optical glass data prediction method according to claim 5, characterized in that: The step of using the target prediction network to determine a random formula data that best matches the target optical glass data among the multiple random formula data, and determining the random formula data as the predicted optical glass data corresponding to the target optical glass data includes: For each of the multiple random formula data, feature mining is performed on the random formula data through the feature mining unit included in the target prediction network, and the random formula features corresponding to the random formula data are outputted, and the random formula features are predicted and outputted through the first output unit included in the target prediction network to obtain transmittance prediction data corresponding to the random formula data, and each formula component included in the random formula data is matrixed through the projection unit included in the target prediction network to form at least one projection matrix corresponding to the random formula data, and at least one projection matrix corresponding to the random formula data is processed through the second output unit included in the target prediction network to output the superposition effect characterization value corresponding to the random formula data, wherein the superposition effect characterization value is used to reflect the matching degree between each formula component in the random formula data and the target linear constraint relationship, and the matching degree is positively correlated with the reliability of the transmittance prediction data corresponding to the random formula data; Based on the transmittance prediction data and the corresponding superposition effect characterization value corresponding to each of the random formula data, a random formula data that best matches the target optical glass data is determined among the multiple random formula data, and the random formula data is determined as the predicted optical glass data corresponding to the target optical glass data.

7. An optical glass data prediction device, characterized in that: include: A sample label acquisition module, used to acquire sample formula data corresponding to the optical glass and a transmittance label corresponding to the optical glass, wherein the sample formula data includes each formula component and the component content of each formula component; A sample transmittance prediction module, used to predict the transmittance of the sample formula data through a pre-constructed candidate prediction network, output the transmittance prediction data corresponding to the sample formula data, and predict the sample formula data including the superposition effect between the formula components through the candidate prediction network, and output the superposition effect representation value corresponding to the sample formula data, wherein the superposition effect representation value is used to reflect the matching degree between the formula components and the target linear constraint relationship, and the target linear constraint relationship is at least used to reflect that the sum of the molar percentages of the component contents corresponding to the formula components is 100%; A prediction network optimization module, configured to optimize the parameters of the candidate prediction network based on the transmittance label, the transmittance prediction data and the superposition effect characterization value to form a target prediction network; an optical glass data prediction module, used to predict the target optical glass data using the target prediction network, and output predicted optical glass data corresponding to the target optical glass data, wherein when the target optical glass data includes formula data, the predicted optical glass data includes transmittance data, and when the target optical glass data includes transmittance data, the predicted optical glass data includes formula data; Among them, the transmittance prediction of the sample formula data is performed through the pre-constructed candidate prediction network, and the transmittance prediction data corresponding to the sample formula data is output, and the sample formula data including the superposition effect between each formula component is predicted through the candidate prediction network, and the superposition effect characterization value corresponding to the sample formula data is output, including: Performing feature mining on the sample recipe data through a feature mining unit included in a pre-built candidate prediction network, and outputting sample recipe features corresponding to the sample recipe data; The sample formula feature is predicted and outputted by a first output unit included in the candidate prediction network to obtain transmittance prediction data corresponding to the sample formula data; Matrixing the recipe components included in the sample recipe data by a projection unit included in the candidate prediction network to form at least one projection matrix corresponding to the sample recipe data, wherein each row of each projection matrix corresponds to a linear constraint relationship between the recipe components; The at least one projection matrix is ​​processed by a second output unit included in the candidate prediction network to output a superposition effect representation value corresponding to the sample recipe data.

8. An electronic device, characterized in that: include: Memory for storing computer programs; A processor connected to the memory is used to execute the computer program stored in the memory to implement the optical glass data prediction method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is run, the optical glass data prediction method according to any one of claims 1 to 6 is executed.

Images