PCB expansion and shrinkage compensation value prediction method based on time weight and related equipment

By using a time weight-based method in the prediction of the inner layer of PCB board, the time weight and weighted least squares method are used to establish a linear regression model, the problem of low prediction accuracy in the existing technology is solved and the sensitivity of the model is improved and robustness reduces the risk of adverse rise and contraction.

CN116610946BActive Publication Date: 2024-12-06江西景旺精密电路有限公司
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202310307145.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-03-27
Publication Date
2024-12-06
Estimated Expiration
2043-03-27

AI Technical Summary

Technical Problem

The existing linear regression model has a low sensitivity in the prediction of the inner layer of PCB board, resulting in a decrease in prediction accuracy. Especially when the increase and contraction change is large during the production process, there is a risk of bad increase and contraction increase.

Method used

Using the PCB increase and shrinkage compensation value prediction method based on time weight, the time weight of the training data is calculated by obtaining sample data and preprocessing it, and the regression coefficient of the initial linear regression model is calculated using the weighted least squares method, and the target linear regression model is established to improve the Model sensitivity and robustness.

Benefits of technology

The prediction accuracy and sensitivity of the linear regression model are improved, the robustness of the model is enhanced, and the risk of poor rise and fall is effectively reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116610946B_ABST
    Figure CN116610946B_ABST
Patent Text Reader

Abstract

The present invention discloses a PCB expansion and shrinkage compensation value prediction method based on time weight and related equipment. The method includes acquiring sample data, preprocessing the sample data, and obtaining training data, wherein the sample data at least includes production information, feature vectors, and corresponding label vectors of PCB boards; calculating the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data; calculating the regression coefficient of the initial linear regression model based on the weight index value, the training label value, and the feature vector using a preset weighted least squares method to obtain a target linear regression model; inputting the feature vector of the PCB board to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value, and obtaining the expansion and shrinkage compensation prediction value. The method improves the prediction accuracy of the PCB expansion and shrinkage compensation value prediction model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of PCB pressing technology, and in particular to a PCB expansion and contraction compensation value prediction method based on time weight and related equipment. Background Art

[0002] In the current production process of PCB boards, linear regression models are usually used to predict the expansion and shrinkage compensation coefficients of the inner layers of PCB boards. Historical data is collected to establish a linear regression model, and the modeling data has a certain time span. The currently established linear regression model does not take into account the timeliness of historical data. When the expansion and shrinkage of the inner layer of the PCB is greatly changed, the sensitivity of the linear regression model is low, and the prediction accuracy of the linear regression model decreases, resulting in the risk of increased expansion and shrinkage. Summary of the invention

[0003] The embodiment of the present invention provides a PCB expansion and contraction compensation value prediction method and related equipment based on time weight, aiming to solve the problem of low prediction accuracy of the expansion and contraction compensation coefficient of the inner layer of the PCB board using a linear regression model in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides a method for predicting a PCB expansion and contraction compensation value based on time weight, which includes:

[0005] Acquire sample data, and preprocess the sample data to obtain training data, wherein the sample data at least includes production information of the PCB board, a feature vector, and a corresponding label vector, wherein the production information at least includes the production time of the PCB board;

[0006] Calculating the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data, wherein the farther the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is;

[0007] Calculate the training label value of the training data based on the label vector;

[0008] Based on the weight index value, the training label value and the feature vector, the regression coefficient of the initial linear regression model is calculated using a preset weighted least squares method to obtain a target linear regression model;

[0009] The characteristic vector of the PCB board to be predicted is input into the target linear regression model to predict the expansion and shrinkage compensation value, so as to obtain the expansion and shrinkage compensation prediction value.

[0010] In a second aspect, an embodiment of the present invention provides a PCB expansion and contraction compensation value prediction device based on time weight, which includes:

[0011] An acquisition module, used to acquire sample data and preprocess the sample data to obtain training data, wherein the sample data at least includes production information of the PCB board, a feature vector and a corresponding label vector, wherein the production information at least includes the production time of the PCB board;

[0012] A weight calculation module, used to calculate the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data, wherein the longer the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is;

[0013] A label value calculation module, used to calculate the training label value of the training data based on the label vector;

[0014] A regression coefficient calculation module, used to calculate the regression coefficient of the initial linear regression model using a preset weighted least squares method based on the weight index value, the training label value and the feature vector, to obtain a target linear regression model;

[0015] The prediction module is used to input the feature vector of the PCB board to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value, so as to obtain the expansion and shrinkage compensation prediction value.

[0016] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for predicting PCB expansion and shrinkage compensation value based on time weight as described in the first aspect above is implemented.

[0017] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the PCB expansion and shrinkage compensation value prediction method based on time weight described in the first aspect above.

[0018] The embodiment of the present invention provides a PCB expansion and shrinkage compensation value prediction method based on time weight and related equipment. The method includes acquiring sample data and preprocessing the sample data to obtain training data, wherein the sample data at least includes production information, feature vectors and corresponding label vectors of PCB boards; calculating the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data, wherein the longer the production time, the smaller the weight index value, and the closer the production time, the larger the weight index value; calculating the training label value of the training data based on the label vector; calculating the regression coefficient of the initial linear regression model based on the weight index value, the training label value and the feature vector using a preset weighted least squares method to obtain a target linear regression model; inputting the feature vector of the PCB board to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value, and obtaining the expansion and shrinkage compensation prediction value. The method determines the weight index value of the training data by using a weight function according to the production information of the training data, fully considers the influence of data at different times on the prediction of the PCB expansion and shrinkage compensation value, and uses the training data with a relatively recent production time to train the linear regression model based on the fact that the farther the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is. The prediction accuracy of the linear regression model is improved. Based on the weight index value, the training label value and the feature vector, the regression coefficient of the initial linear regression model is calculated according to the weighted least squares method, so as to improve the sensitivity of the linear regression model and enhance the robustness of the linear regression model. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0020] Figure 1 A schematic flow chart of an embodiment of a method for predicting PCB expansion and contraction compensation values ​​based on time weight provided in an embodiment of the present invention;

[0021] Figure 2 A sub-flow diagram of step S110 of another embodiment of the method for predicting PCB expansion and contraction compensation value based on time weight provided by an embodiment of the present invention;

[0022] Figure 3 A schematic block diagram of a PCB expansion and contraction compensation value prediction device based on time weight provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] 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 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.

[0024] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0025] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0026] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0027] See also Figure 1 A schematic flow chart of a method for predicting a PCB expansion and contraction compensation value based on time weight provided in an embodiment of the present invention, the method comprising steps S110 to S150.

[0028] Step S110, acquiring sample data, and preprocessing the sample data to obtain training data, wherein the sample data at least includes production information of the PCB board, a feature vector, and a corresponding label vector, wherein the production information at least includes the production time of the PCB board;

[0029] In this embodiment, the data of PCB sheets produced within a period are collected as sample data. The sample data includes the production information, feature vectors and corresponding label vectors of the PCB sheets. The label vector includes the expansion and contraction compensation value of the PCB sheet production application and the actual expansion and contraction change value of the PCB sheet after pressing; the feature vector includes the Tg value, thickness, copper thickness, pressing thickness, pp number of sheets and residual copper rate of the PCB sheet; the production information includes the product code, layer number, production time and target distance of the PCB sheet. The collected sample data is preprocessed to obtain training data.

[0030] like Figure 2 As shown, in one embodiment, step S110 includes:

[0031] Step S210: sort the sample data in descending order according to production time, perform deduplication processing on the sample data, retain the most recent batch of sample data, and obtain deduplication data;

[0032] Step S220, deleting records with missing features and out-of-range label vectors in the deduplicated data to obtain retained data;

[0033] Step S230: Perform a Box-Cox transformation on the feature vector of the retained data to obtain the training data.

[0034] In this embodiment, the sample data is preprocessed, specifically including: first, the sample data is sorted in descending order according to the production time, and the sample data is deduplicated according to the product code, and the sample data of the most recent batch is retained to obtain deduplicated data; then, records with missing features in the deduplicated data are removed, and records with training label values ​​out of range are removed to obtain retained data; and the feature vector of the retained data is box-cox transformed to obtain training data.

[0035] Among them, the box-cox transformation of the feature vector of the retained data is performed according to the following formula:

[0036]

[0037] Where Z represents the eigenvector, Z is a positive value, λ represents the transformation parameter, which is estimated by the maximum likelihood method; ln( ) represents the logarithmic function;

[0038] Or, perform a Box-Cox transformation on the feature vector of the retained data as follows:

[0039]

[0040] Among them, Z is a negative value, μ is a constant, and μ>|Z|.

[0041] Step S120, calculating the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data, wherein the longer the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is;

[0042] In this embodiment, the weight index value in the training data is calculated according to the production time in the production information. The further the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is. For example, taking the farthest production time as the benchmark, t=1; the most recent production data as the end, t=30. The weight index value W is calculated according to the following formula t :

[0043] Wt =2 arctan(t) / π,

[0044] Where t represents the production time.

[0045] Step S130, calculating a training label value of the training data based on the label vector;

[0046] In this embodiment, the training label value of the training data is calculated based on the expansion and contraction compensation value and the actual expansion and contraction change value of the PCB board production application. Specifically, the training label value of the training data is calculated according to the following formula.

[0047]

[0048] Among them, α represents the expansion and contraction compensation value of the training data, c represents the actual expansion and contraction change value, and j represents the target distance.

[0049] Step S140, based on the weight index value, the training label value and the feature vector, the regression coefficient of the initial linear regression model is calculated using a preset weighted least squares method to obtain a target linear regression model;

[0050] Step S150: input the feature vector of the PCB board to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value, and obtain the expansion and shrinkage compensation prediction value.

[0051] In this embodiment, the weight index value W t As a weight variable, the weighted least squares method is used to calculate the regression coefficients of the initial linear regression model. Specifically, the following is the expression of the initial linear regression model:

[0052] y=β0+β1x1+β2x2+...+β k x k +ε,

[0053] Among them, k represents the number of eigenvectors, x1~x k represents the eigenvector, ε represents the residual term;

[0054] Matrix operations are performed according to the following least squares method:

[0055]

[0056]

[0057]

[0058] From AB=D, or (X′X)B=X′Y,

[0059] We get: B = (X′X) -1X′Y,

[0060] Among them, matrix B is the regression coefficient of the linear regression model obtained Q represents the residual sum of squares, X represents the matrix of eigenvectors, Y represents the matrix of label vectors, n represents the number of samples, and k represents the number of variables;

[0061] Next, perform matrix operations according to the following weighted minimum multiplication:

[0062]

[0063] W=[w1 w2…w n ],

[0064]

[0065] Among them, W represents the weight matrix, I represents the unit matrix, and the least squares matrix operation formula is combined to obtain the B matrix of the weighted least squares method:

[0066]

[0067] Finally, the regression coefficient is obtained Substituting into the initial linear regression model, the linear regression equation of the target linear regression model is obtained as follows:

[0068]

[0069] Among them, k represents the number of eigenvectors, x1~x k represents a feature vector.

[0070] It should be noted that the expression of the initial linear regression model includes the residual term, which makes y equal to the actual value; while the linear regression equation of the target linear regression model does not include the residual term because there is a residual between the predicted value and the actual value.

[0071] Finally, the characteristic vector of the PCB board to be predicted is input into the target linear regression model, and the linear regression equation of the target linear regression model is used to predict the expansion and shrinkage compensation value to obtain the expansion and shrinkage compensation prediction value.

[0072] Furthermore, the present application has finally determined through experiments that the process of using the weighted least squares method to solve the regression coefficient of the initial linear regression model is better than using the least squares method as follows:

[0073] Step 1: Collect the historical data of the PCB board production in the past 30 days, including compensation coefficient, Tg value, board thickness, board copper thickness, pressing thickness, pp number, and residual copper rate;

[0074] Step 2: Preprocess the collected data to remove duplicate records, records with missing fields, and records with abnormal values;

[0075] Step 3: Logarithmically transform the plate thickness, pressing thickness, and residual copper rate fields according to the box-cox method;

[0076] Step 4: According to the production time sequence, take the first production time as the benchmark, count as t = 1, and the last production data as the end, count as t = 30, and output the weight index value W according to the weight function t ;

[0077] Step 5: Apply the stepwise regression algorithm, take the compensation coefficient as the dependent variable and the other fields as the independent variables, and use the least square method to output the regression coefficient, adjusted R square, and standard estimation error σ of the linear regression model. The adjusted R square value is 0.698 and the standard estimation error σ is 0.410.

[0078] Step 6: Set the time index value W t As the weight variable, the weighted least square method is used to output the regression coefficient, adjusted R square, and standard estimation error σ of the linear regression model. Among them, the adjusted R square value is 0.742, and the standard estimation error σ is 0.383;

[0079] Step 7: Compare the adjusted R-squared value and the standard estimation error σ value output in step 5 and step 6. The larger the adjusted R-squared value, the better, and the smaller the standard estimation error σ, the better. The result output in step 6 is better than the result output in step 5. Based on the standard estimation error σ as the comparison benchmark, the result in step 6 is improved by 6% compared with the result in step 5. Therefore, the weighted least squares method is better than the least squares method for solving the regression coefficient of the initial linear regression model.

[0080] According to the production information of the training data, the method uses a weight function to determine the weight index value of the training data, fully considers the influence of data at different times on the prediction of the PCB expansion and shrinkage compensation value, and uses the training data with a relatively recent production time to train the linear regression model based on the fact that the farther the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is. Based on the weight index value, the training label value and the feature vector, the regression coefficient of the initial linear regression model is calculated according to the weighted least squares method to improve the sensitivity of the linear regression model. The larger the amount of data in the model training process, the more robust the prediction ability of the output model is.

[0081] The embodiment of the present invention further provides a PCB expansion and contraction compensation value prediction device based on time weight, and the PCB expansion and contraction compensation value prediction device based on time weight is used to execute any embodiment of the above-mentioned PCB expansion and contraction compensation value prediction method based on time weight. Figure 3 , Figure 31 is a schematic block diagram of a PCB expansion and contraction compensation value prediction device based on time weight provided by an embodiment of the present invention. The PCB expansion and contraction compensation value prediction device 100 based on time weight can be configured in a server.

[0082] like Figure 3 As shown, the PCB expansion and contraction compensation value prediction device 100 based on time weight includes an acquisition module 110, a weight calculation module 120, a label value calculation module 130, a regression coefficient calculation module 140, and a prediction module 150.

[0083] The acquisition module 110 is used to acquire sample data and preprocess the sample data to obtain training data, wherein the sample data at least includes production information of the PCB board, a feature vector and a corresponding label vector, wherein the production information at least includes the production time of the PCB board;

[0084] The weight calculation module 120 is used to calculate the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data, wherein the longer the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is;

[0085] A label value calculation module 130, configured to calculate a training label value of the training data based on the label vector;

[0086] A regression coefficient calculation module 140 is used to calculate the regression coefficient of the initial linear regression model using a preset weighted least square method based on the weight index value, the training label value and the feature vector to obtain a target linear regression model;

[0087] The prediction module 150 is used to input the feature vector of the PCB board to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value, so as to obtain the expansion and shrinkage compensation prediction value.

[0088] In one embodiment, the acquisition module 110 includes:

[0089] A deduplication unit, used to sort the sample data in descending order according to production time, and perform deduplication processing on the sample data, retaining the most recent batch of sample data to obtain deduplication data;

[0090] A deleting unit, used to delete records with missing features and out-of-range training label values ​​in the deduplicated data, to obtain retained data;

[0091] The conversion unit is used to perform a Box-Cox transformation on the feature vector of the retained data to obtain the training data.

[0092] In one embodiment, the conversion unit is used to perform a Box-Cox transformation on the feature vector of the retained data according to the following formula:

[0093]

[0094] Where Z represents the eigenvector, Z is a positive value, λ represents the transformation parameter, and ln() represents the logarithmic function;

[0095] Or, perform a Box-Cox transformation on the feature vector of the retained data according to the following formula:

[0096]

[0097] Among them, Z is a negative value, μ is a constant, and μ>|Z|.

[0098] In one embodiment, the weight calculation module 120 is used to calculate the weight index value W according to the following formula: t :

[0099] W t =2*arctan(t) / π,

[0100] Where t represents the production time.

[0101] In one embodiment, the label value calculation module 130 is used to calculate the training label value of the training data according to the following formula:

[0102]

[0103] Among them, α represents the expansion and contraction compensation value of the training data, c represents the actual expansion and contraction change value, and j represents the target distance.

[0104] In one embodiment, the regression coefficient calculation module 140 is used to calculate the regression coefficient according to the following formula:

[0105]

[0106] Wherein, W represents the weight matrix of the weight index value, I represents the identity matrix, X represents the eigenvector, and Y represents the label matrix of the training label value.

[0107] In one embodiment, the prediction module 150 is used to calculate the expansion and contraction compensation prediction value according to the following formula:

[0108] Among them, k represents the number of eigenvectors, x1~x k represents a feature vector.

[0109] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the PCB expansion and contraction compensation value prediction method based on time weight as described above when executing the computer program.

[0110] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium may be a non-volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to execute the PCB expansion and contraction compensation value prediction method based on time weight as described above.

[0111] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the above-described equipment, devices and units can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. Those of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0112] In the several embodiments provided by the present invention, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. Units with the same function may also be combined into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices or units, or may be an electrical, mechanical or other form of connection.

[0113] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0114] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0115] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), disk or optical disk and other media that can store program codes.

[0116] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. PCB expansion and shrinkage compensation value prediction method based on time weight, It is characterized in that include: Acquire sample data, and preprocess the sample data to obtain training data, wherein the sample data at least includes production information of the PCB board, a feature vector, and a corresponding label vector, wherein the production information at least includes the production time of the PCB board; Calculating the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data, wherein the farther the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is; Calculate the training label value of the training data based on the label vector; Based on the weight index value, the training label value and the feature vector, the regression coefficient of the initial linear regression model is calculated using a preset weighted least squares method to obtain a target linear regression model; Inputting the characteristic vector of the PCB board to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value, and obtaining the expansion and shrinkage compensation prediction value; The calculating the time weight of the training data according to the production information and the preset weight function to obtain the weight index value in the training data includes: The weight index value W is calculated according to the following formula t : W t =2·arctan(t) / π, Where, t represents the production time; The method of calculating the regression coefficient of the initial linear regression model based on the weight index value, the training label value and the feature vector by using a preset weighted least square method to obtain a target linear regression model includes: The regression coefficient is calculated according to the following formula Wherein, W represents the weight matrix of the weight index value, I represents the identity matrix, X represents the eigenvector, and Y represents the label matrix of the training label value.

2. According to the time weighted PCB expansion and contraction compensation value prediction method of claim 1, It is characterized in that Preprocessing the sample data to obtain training data includes: Sorting the sample data in descending order according to production time, performing deduplication processing on the sample data, retaining the most recent batch of sample data, and obtaining deduplication data; Deleting records with missing features and out-of-range training label values ​​in the deduplicated data to obtain retained data; The feature vector of the retained data is subjected to a Box-Cox transformation to obtain the training data.

3. According to the time weighted PCB expansion and contraction compensation value prediction method of claim 2, It is characterized in that The performing a Box-Cox transformation on the feature vector of the retained data comprises: The feature vector of the retained data is box-cox transformed according to the following formula: Where Z represents the eigenvector, Z is a positive value, λ represents the transformation parameter, and ln() represents the logarithmic function; Or, perform a Box-Cox transformation on the feature vector of the retained data according to the following formula: Among them, Z is a negative value, μ is a constant, and μ>|Z|.

4. According to the time weighted PCB expansion and contraction compensation value prediction method of claim 1, It is characterized in that The production information also includes the target distance of the PCB board, and the label vector includes the expansion and contraction compensation value and the actual expansion and contraction change value; and the training label value of the training data is calculated based on the label vector, including: The training label value of the training data is calculated according to the following formula: Among them, α represents the expansion and contraction compensation value of the training data, c represents the actual expansion and contraction change value, and j represents the target distance.

5. According to the time weighted PCB expansion and contraction compensation value prediction method of claim 1, It is characterized in that The step of inputting the characteristic vector of the PCB board material to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value to obtain the expansion and shrinkage compensation prediction value includes: The expansion and contraction compensation prediction value is calculated according to the following formula Among them, k represents the number of eigenvectors, x 1 ~x k represents a feature vector.

6. A PCB expansion and contraction compensation value prediction device based on time weight, It is characterized in that include: An acquisition module, used to acquire sample data and preprocess the sample data to obtain training data, wherein the sample data at least includes production information of the PCB board, a feature vector and a corresponding label vector, wherein the production information at least includes the production time of the PCB board; A weight calculation module, used to calculate the time weight of the training data according to the production information and a preset weight function to obtain a weight index value in the training data, wherein the longer the production time is, the smaller the weight index value is, and the closer the production time is, the larger the weight index value is; A label value calculation module, used to calculate the training label value of the training data based on the label vector; A regression coefficient calculation module, used to calculate the regression coefficient of the initial linear regression model using a preset weighted least squares method based on the weight index value, the training label value and the feature vector, to obtain a target linear regression model; A prediction module, used for inputting the feature vector of the PCB board to be predicted into the target linear regression model to predict the expansion and shrinkage compensation value, so as to obtain the expansion and shrinkage compensation prediction value; The weight calculation module is used to calculate the weight index value W according to the following formula t : W t =2*arctan(t) / π, Where, t represents the production time; The regression coefficient calculation module is used to calculate the regression coefficient according to the following formula Wherein, W represents the weight matrix of the weight index value, I represents the identity matrix, X represents the eigenvector, and Y represents the label matrix of the training label value.

7. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the method for predicting PCB expansion and contraction compensation value based on time weight as described in any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the PCB expansion and contraction compensation value prediction method based on time weight as described in any one of claims 1 to 5.

Citation Information

Patent Citations

  • Semiconductor batch process fault diagnosis method based on time sequence model coefficient

    CN108459579A

  • Abnormal behavior detection method and device based on regression training and electronic equipment

    CN111160485A