A method for predicting cv of each formed voltage of corroded aluminum foil based on machine learning

Abstract

The application discloses a CV prediction method for each formation voltage of corroded aluminum foil based on machine learning, acquires a SEM image of the corroded aluminum foil; pretreats the SEM image of the corroded aluminum foil to obtain a pretreated image; secondarily corrects the pretreated image under different formation voltages to remove existing blocked holes to obtain a corrected image under the corresponding voltage; and calculates the CV value of the aluminum foil per unit area under different formation voltages according to the corrected image. The application secondarily corrects the image to remove small blocked holes in the formation process under different formation voltages. The application can solve the problem of complicated and random artificial measurement and is more efficient and convenient. The application can improve the measurement accuracy by taking multiple photos of different areas, and the measurement result is more accurate compared with artificial measurement. The CV data obtained by the application can be used for comparison between different corroded foils to determine the applicable formation voltage interval of different foils.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of image processing and aluminum foil surface analysis, and particularly relates to a CV prediction method for etched aluminum foil under each forming voltage based on machine learning. BACKGROUND

[0002] Capacitors are one of the three major passive electronic components (resistors, capacitors and inductors) in the world, and play an important role in the electronic component industry. They are essential basic electronic components in electronic circuits. Aluminum electrolytic capacitors have the advantages of low cost, excellent performance, mature technology and convenient application, and have a very large demand in electronic components and complete machines.

[0003] With the continuous development of the electronic industry, the performance requirements for aluminum electrolytic capacitors are becoming higher and higher, prompting aluminum electrolytic capacitors to develop towards miniaturization and long service life. Anodized foil is a key material for aluminum electrolytic capacitors, and how to improve the performance of aluminum electrolytic capacitor anode foil is a core problem that needs to be solved for circuit system miniaturization.

[0004] Etched aluminum foil is a basic material for aluminum electrolytic capacitors, and its quality directly affects the performance of aluminum electrolytic capacitor anode foil. At present, there is less research on the microstructure and morphology characterization of etched foil, and there is a lack of systematic measurement method to characterize the performance of etched aluminum foil. Most of the micro-characterization of etched aluminum foil has large subjective factors, and the observation is only concentrated on one or a few holes, which cannot reflect the overall performance, the test results have randomness and are not persuasive. At the same time, the measurement of CV performance of etched foil still needs manual experiment, which consumes a lot of manpower and material resources. SUMMARY

[0005] In view of the above problems existing in the micro-analysis technology of aluminum foil for aluminum electrolytic capacitors in the prior art, the purpose of the present application is to provide a CV prediction method for etched aluminum foil under each forming voltage based on machine learning, which accurately predicts the CV value of etched foil under different forming voltages by taking SEM photos of the surface of etched foil, saving the consumed manpower and material resources of manual experiment.

[0006] To achieve the above purpose, the technical solution adopted by the present application is as follows:

[0007] A CV prediction method for etched aluminum foil under each forming voltage based on machine learning, comprising the following steps:

[0008] 1) Obtain the SEM image of etched aluminum foil;

[0009] 2) Preprocess the SEM image of etched aluminum foil to obtain a preprocessed image;

[0010] 3) The pre-processed image is modified again at different forming voltages to remove the holes with blockage, and the modified image at the corresponding voltage is obtained.

[0011] 4) The CV value of the aluminum foil per unit area at different forming voltages is calculated according to the modified image.

[0012] Further, the magnification of the SEM image of the etched aluminum foil is 2-10K.

[0013] Further, the specific steps of pre-processing the SEM image of the etched aluminum foil are as follows: using the imglab tool in the dlib library to label the sample holes in the SEM image, and using the data augmentation method to expand the sample.

[0014] Further, the data augmentation method includes rotation and cropping of the sample.

[0015] Further, the CV of the aluminum foil per unit area at different forming voltages U is calculated by the following formula:

[0016] C=ε0·εS / a

[0017] In the formula, C is the capacity of the aluminum foil in the area, ε0 is the dielectric constant of vacuum, ε is the dielectric constant of aluminum oxide, S is the total surface area of aluminum oxide formed per unit area of aluminum foil, and a is the formation constant.

[0018] Further, the total surface area of aluminum oxide formed per unit area of aluminum foil is the sum of the inner surface area of the aluminum oxide hole, the outer surface area of the aluminum foil, and the side surface area.

[0019] Further, the inner surface area of the aluminum oxide hole is the product of the number of aluminum oxide holes and the average inner diameter of the aluminum oxide hole.

[0020] Further, the average inner diameter R of the aluminum oxide hole is calculated by the following formula: R=0.5·L-0.39d, where L is the average side length and d is the thickness of the oxide film; the inner surface area of the aluminum oxide hole is 50·2·π·R·n, where n is the number of holes; and π is the circular constant.

[0021] Further, the average side length is obtained by the following process: using the support vector machine svm model in the dlib library to train the data to obtain a trained model, using the trained model to divide the SEM image region of the etched aluminum foil into aluminum substrate and hole regions, and through target detection of the holes, the number n of holes is counted, and at the same time, the coordinates of the square holes and the length and width of the square holes are obtained. According to the length and width of the hole and the number of holes, the average side length is obtained.

[0022] Further, the specific process of the secondary correction of the pre-processed image is as follows: under different forming voltages, the thickness d of the oxide film generated under the voltage U is obtained according to the aluminum oxide hole growth model, the relationship between the aluminum oxide thickness d and the forming voltage U is: d = aU, a is a forming constant, and then the bwareaopen function is used to remove the area smaller than π·d 2 to obtain the corrected image.

[0023] Further, the aluminum oxide hole growth model is that the square aluminum foil hole is grown into a circular aluminum oxide hole under the action of the forming voltage.

[0024] Compared with the prior art, the beneficial effects of the present application are:

[0025] The machine learning algorithm is used to train the hole model of the image in the present application, and the number of holes and the hole length of the aluminum etching foil can be obtained more accurately by using the present application. The machine learning algorithm is used to judge the image area in the present application, which reduces the calculation cost and reduces the consumption of memory and time. The image is corrected twice in the present application, and the small holes blocked in the formation process under different forming voltages are removed. The present application can solve the problem of tedious and random manual measurement, and is more efficient and convenient. The present application can improve the measurement accuracy by taking multiple photos of different areas, and the measurement result is more accurate than manual measurement. The CV data obtained by the present application can be used for comparison between different etching foils to determine the applicable forming voltage interval of different foils.

[0026] Further, the present application establishes a simple aluminum oxide hole growth model for calculating the inner diameter of the aluminum oxide hole under different forming voltages.

[0027] Further, the present application uses software for prediction, without consuming formation materials and electrolyte, saving energy and protecting the environment. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The flowchart of the present application is a CV prediction method for each forming voltage of the etching aluminum foil based on machine learning;

[0029] Figure 2 The SEM photograph of the present application is shown in Figure 1;

[0030] Figure 3 The schematic diagram of the present application after pre-processing and hole region identification is shown in Figure 2, wherein the black area is the aluminum base, and the white area is the hole;

[0031] Figure 4The results of image correction in Embodiment 1 of the present invention are shown in (a) as the original SEM image, (b) as the original image after region recognition, and (c), (d), (e) and (f) as the image correction results at 100V, 250V, 400V and 600V respectively.

[0032] Figure 5 The comparison between the predicted CV results obtained in Example 1 of this invention and the actual experimental values ​​shows that the average error between the two is only 1.59%. Detailed Implementation

[0033] The present invention will now be described in detail with reference to the accompanying drawings.

[0034] like Figure 1 As shown, this invention provides a machine learning-based method for predicting the CV (copper oxide) of etched aluminum foil at various formation voltages, comprising the following steps:

[0035] 1) SEM image of the etched foil: The surface of the etched aluminum foil is photographed using a scanning electron microscope to obtain the required SEM image, which has a magnification of 2-10K.

[0036] 2) Initialization and side length calculation: The captured SEM images are preprocessed using Python. The steps are as follows: The imglab tool from the dlib library is used to annotate the sample holes, and data augmentation methods are used to augment the samples. Data augmentation operations include rotating and cropping the samples.

[0037] The Support Vector Machine (SVM) model from the dlib library was used to train the data samples. Using the trained model, the SEM image region was divided into two parts: the aluminum substrate and the holes. The number of holes, *n*, was counted through hole detection. Simultaneously, the coordinates of the square holes were obtained, with their length and width being (hole.right-hole.left) and (hole.top-hole.bottom), respectively. The average side length *L* can be calculated from the hole length, width, and number of holes.

[0038] 3) Secondary image correction to remove blocked pores: Under different formation voltages, based on the alumina pore growth model, the thickness d of the oxide film formed at voltage U is calculated. The relationship between the alumina thickness d and the formation voltage U is: d = aU, where a is the formation constant. Then, the bwareaopen function is used to remove areas smaller than π·d. 2 The region was used to obtain corrected images under different formation voltages.

[0039] Among them, the alumina pore growth model is that under the action of voltage, square aluminum foil pores grow into circular alumina pores in the electrolyte.

[0040] 4) Inner diameter calculation: According to the corrected image under different forming voltage and the alumina hole growth model, the average inner diameter R of the alumina hole under the forming voltage U is calculated as R = 0.5·L-0.39d, wherein L is the average side length. The inner surface area of the alumina hole under different forming voltage U is calculated as 50·2·π·R·n, n is the number of holes, 50 is the unit of microns, representing the height of the tunnel hole of the aluminum foil, π is the circular constant, and · represents multiplication.

[0041] 5) Surface area calculation: The inner surface area of the alumina hole is calculated, and then the outer surface area and the side surface area of the aluminum foil are added to obtain the total surface area S of the alumina formed on the aluminum foil per unit area.

[0042] 6) CV prediction: The CV value of the aluminum foil per unit area under different forming voltage U can be calculated as ε0·εS / a from the formula C = ε0·εS / d = ε0·εS / aU, wherein C is the capacity of the aluminum foil in the area, ε0 is the dielectric constant of vacuum, and ε is the dielectric constant of alumina.

[0043] Example 1

[0044] 1) SEM of etched foil: The surface of the etched aluminum foil is photographed using a scanning electron microscope to obtain a SEM image as shown in FIG. 1, wherein the magnification of the image is 2.0K; Figure 2

[0045] 2) Initialization and length statistics: The SEM photo is enhanced, labeled, and trained by using python language, and the picture area is divided into aluminum substrate and hole two parts according to the trained model. Then the average side length of the hole is calculated by hole detection. Referring to FIG. 2, the average pore diameter calculated in this embodiment is 1.99 microns, and the number of holes is 403. Figure 3

[0046] 3) Image correction: Under different forming voltages, the thickness d of the generated oxide film under the voltage U is calculated according to the alumina growth model. Then the bwareaopen function is used to remove the areas with an area less than π·d2 to obtain the corrected image under the forming voltage. As shown in FIG. 3, the right side shows the hole image correction under voltages of 100V, 250V, 400V and 600V. Figure 4

[0047] 4) Inner diameter calculation: According to the alumina growth model, the inner diameter R of the alumina hole under the forming voltage U is calculated as R = 0.5·L-0.39d.

[0048] 5) Surface area calculation: The inner surface area of the alumina hole is calculated, and then the outer surface area and the side surface area of the aluminum foil are added to obtain the total surface area S of the alumina formed on the aluminum foil per unit area. ​​​

[0049] 6) CV prediction: The CV of the aluminum foil under different forming voltages U can be calculated by the formula C = C = ε0·εS / d = ε0·εS / aU, which is ε0·εS / a. The prediction result of the CV of the No. 1 foil of Factory A is shown in Table 2, which has only 1.59% error with the actual experimental value, and can completely replace the experimental test. Figure 5 Table 2

Claims

1. A method for predicting CV at various forming voltages of a machine learning based etched aluminum foil, characterized by, The method comprises the following steps: 1) obtaining a SEM image of the etched aluminum foil; 2) preprocessing the SEM image of the etched aluminum foil to obtain a preprocessed image; 3) performing secondary correction on the preprocessed image under different forming voltages to remove the existing blocked pores to obtain a corrected image under the corresponding voltage; 4) calculating the CV value of the aluminum foil per unit area under different forming voltages according to the corrected image; The CV of the aluminum foil per unit area under different forming voltages U is calculated by the following formula: C = ε0·εS / a In the formula, C is the capacity of the aluminum foil in the area, ε0 is the dielectric constant of vacuum, ε is the dielectric constant of aluminum oxide, S is the total surface area of aluminum oxide after the formation of the aluminum foil per unit area, and a is the formation constant; The total surface area of aluminum oxide after the formation of the aluminum foil per unit area is the sum of the inner surface area of the aluminum oxide pores, the outer surface area of the aluminum foil, and the side surface area; The inner surface area of the aluminum oxide pores is the product of the number of aluminum oxide pores and the average inner diameter of the aluminum oxide pores; The average inner diameter R of the aluminum oxide pores is calculated by the following formula: R = 0.5·L-0.39d, wherein L is the average side length, and d is the thickness of the oxide film; the inner surface area of the aluminum oxide pores is 50·2·π·R·n, wherein n is the number of pores; and π is the circular constant; The average side length is obtained by the following process: training the data using the support vector machine svm model in the dlib library to obtain a trained model, dividing the SEM image region of the etched aluminum foil using the trained model into aluminum substrate and pore regions, detecting the target of the pores, and counting the number of pores n, and simultaneously obtaining the coordinates and length and width of the square pores, and obtaining the average side length according to the length and width of the pores and the number of pores.

2. The method of claim 1, wherein the method is based on machine learning. The magnification of the SEM image of the etched aluminum foil is 2-10K.

3. The method of claim 1, wherein the method is based on machine learning. The specific steps of preprocessing the SEM image of the etched aluminum foil are as follows: using the imglab tool in the dlib library to label the sample pores in the SEM image, and using the data enhancement method to expand the sample.

4. The method of claim 3, wherein the method is based on machine learning. The data enhancement method includes rotating and cutting the sample.

5. The method of claim 1, wherein the method is based on machine learning. The specific process of the secondary modification of the pretreated image is as follows: under different forming voltages, the thickness d of the oxide film generated under the voltage U is obtained according to the aluminum oxide hole growth model, the relationship between the aluminum oxide thickness d and the forming voltage U is d=aU, a is a forming constant, and then the bwareaopen function is used to remove the area smaller than π·d 2 of the region, to obtain the modified image. ​

Images