Chemical polishing method, device and equipment based on deep learning and storage medium

Abstract

The invention discloses a chemical polishing method, device and equipment based on deep learning and a storage medium. The chemical polishing method based on deep learning comprises the following steps: determining difference data of process parameters of an anodized sample to be polished on chemical polishing glossiness and / or chemical polishing time, wherein the difference data comprises influence result data and importance degree data; and inputting the process parameters and the difference data of the to-be-polished anodic oxidation sample wafer with the importance degree greater than a preset degree into a deep learning neural network with an autonomous learning ability, and predicting to obtain the polishing time of the to-be-polished anodic oxidation sample wafer. According to the technical scheme, the automation degree of the chemical polishing method for the anodic oxidation sample wafer to be polished is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of chemical polishing, and particularly to a chemical polishing method, device, equipment and storage medium based on deep learning. Background Art

[0002] In the manufacturing of flat metal casings, chemical polishing is a key surface treatment process. By using a chemical solution to remove minute protrusions on the metal surface, the surface becomes smooth, flat, and has a mirror-like effect. This process is applicable to metal casings with complex shapes, ensuring appearance consistency and uniformity, and enhancing the visual effect and corrosion resistance of the product.

[0003] Chemical polishing not only improves the wear resistance of the metal casing but also extends its service life. By monitoring parameters such as solution concentration, raw materials, temperature, and pH to precisely control the reaction time, consistency in the surface quality of the metal casing can be achieved. However, there are numerous parameters in the process, and any inaccurate operation will lead to a large error in manually estimating the chemical polishing time, ultimately resulting in fluctuations in surface quality. The existing chemical polishing solutions in the prior art adopt manual chemical polishing, with low automation, leading to high requirements for production personnel, a high learning threshold for novices, and a shortage of human resources. Summary of the Invention

[0004] The present invention provides a chemical polishing method, device, equipment and storage medium based on deep learning to improve the automation of the chemical polishing method for an anodized sample to be polished.

[0005] According to one aspect of the present invention, there is provided a chemical polishing method based on deep learning, including:

[0006] Determining difference data of the process parameters of the anodized sample to be polished with respect to chemical polishing glossiness and / or chemical polishing time, where the difference data includes influence result data and importance degree data;

[0007] Inputting the process parameters of the anodized sample to be polished and the difference data with an importance degree greater than a preset degree into a deep learning neural network with autonomous learning ability to predict the polishing time of the anodized sample to be polished.

[0008] According to another aspect of the present invention, there is provided a chemical polishing device based on deep learning, including:

[0009] A training data determination module for determining difference data of the process parameters of the anodized sample to be polished with respect to chemical polishing glossiness and / or chemical polishing time, where the difference data includes influence result data and importance degree data;

[0010] A preset module is configured to input the process parameters and the difference data of the anodized sample to be polished, whose importance level is greater than a preset level, into a deep learning neural network with self-learning ability, and predict the polishing time of the anodized sample to be polished.

[0011] According to another aspect of the present invention, there is provided an electronic device, which includes:

[0012] At least one processor; and

[0013] A memory communicatively connected to the at least one processor; wherein,

[0014] The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the chemical polishing method based on deep learning according to any embodiment of the present invention.

[0015] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for implementing the chemical polishing method based on deep learning according to any embodiment of the present invention when executed by a processor.

[0016] The chemical polishing method, device, equipment and storage medium based on deep learning provided by the embodiments of the present invention determine the difference data of the process parameters of the anodized sample to be polished with respect to the chemical polishing glossiness and / or the chemical polishing time. The difference data includes influence result data and importance level data, and inputs the process parameters and the difference data of the anodized sample to be polished, whose importance level is greater than a preset level, into a deep learning neural network with self-learning ability, and predicts the polishing time of the anodized sample to be polished. The prediction of the above polishing time combines an artificial intelligence (AI) algorithm to stabilize the chemical polishing process. Through an intelligent control system, numerous process parameters are monitored, the chemical polishing duration is precisely adjusted, human operation errors are avoided, and the automation level of the production line is greatly improved. It not only improves production efficiency but also reduces the scrap rate caused by process fluctuations, achieving a higher level of precision manufacturing.

[0017] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. Description of the Drawings

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0019] Figure 1 It is a flowchart of a chemical polishing method based on deep learning provided by an embodiment of the present invention;

[0020] Figure 2 It is a flowchart of another chemical polishing method based on deep learning provided by an embodiment of the present invention;

[0021] Figure 3 It is a schematic structural diagram of a chemical polishing device based on deep learning provided by an embodiment of the present invention;

[0022] Figure 4 It is a flowchart of yet another chemical polishing method based on deep learning provided by an embodiment of the present invention;

[0023] Figure 5 For Figure 4 a flowchart included in S350 in;

[0024] Figure 6 It is a schematic diagram of the statistical results of the glossiness of the sides of rework products and normal products in a first group of polished anodized samples and the statistical results of the glossiness of the large surfaces of rework products and normal products in a second group of polished anodized samples provided by an embodiment of the present invention;

[0025] Figure 7 It is a schematic diagram of the statistical results of the median glossiness of the sides of rework products and normal products in a first group of polished anodized samples and the statistical results of the median glossiness of the large surfaces of rework products and normal products in a second group of polished anodized samples provided by an embodiment of the present invention;

[0026] Figure 8 It is a schematic diagram of the result that the glossiness of the sides of rework products and normal products in a first group of polished anodized samples provided by an embodiment of the present invention conforms to a normal distribution;

[0027] Figure 9 It shows a schematic diagram of the result that the glossiness of the large surfaces of rework products and normal products in a second group of polished anodized samples conforms to a normal distribution;

[0028] Figure 10Schematic diagram of the Pearson coefficient of the increase in the number of defective products and the shortening of chemical polishing time in the first group of polished anodized samples provided by the embodiments of the present invention, and the schematic diagram of the Pearson coefficient of the decrease in the number of defective products and the shortening of chemical polishing time in the second group of polished anodized samples;

[0029] Figure 11 Schematic diagram of the Spearman coefficient of the increase in the number of defective products and the shortening of chemical polishing time in the first group of polished anodized samples provided by the embodiments of the present invention, and the decrease in the number of defective products and the shortening of chemical polishing time in the second group of polished anodized samples;

[0030] Figure 12 Schematic diagram of the statistical results of the glossiness of the sides of the first group of polished anodized samples from different suppliers and the statistical results of the glossiness of the large surfaces of the samples in the second group of polished anodized samples provided by the embodiments of the present invention;

[0031] Figure 13 Schematic diagram of the statistical results of the compliance ratio of the glossiness of the sides of the first group of polished anodized samples from different suppliers and the compliance ratio of the glossiness of the large surfaces of the second group of polished anodized samples provided by the embodiments of the present invention;

[0032] Figure 14 Schematic diagram of the statistical results of the proportion of the glossiness of the sides and large surfaces of the samples in the first group of polished anodized samples and the second group of polished anodized samples from different suppliers within the compliance glossiness range provided by the embodiments of the present invention;

[0033] Figure 15 Schematic diagram of the statistical results of the glossiness of the sides of the first group of polished anodized samples with different alloy codes and the statistical results of the glossiness of the large surfaces of the second group of polished anodized samples provided by the embodiments of the present invention;

[0034] Figure 16 Schematic diagram of the statistical results of the proportion of the glossiness of the sides of the first group of polished anodized samples with different alloy codes within the compliance glossiness range and the statistical results of the proportion of the glossiness of the large surfaces of the second group of polished anodized samples within the compliance glossiness range provided by the embodiments of the present invention;

[0035] Figure 17 Schematic diagram of the statistical results of the proportion of the glossiness of the first group of polished anodized samples and the second group of polished anodized samples with different production furnace numbers within the compliance glossiness range provided by the embodiments of the present invention;

[0036] Figure 18Schematic diagram of the glossiness statistical results of the first group of polished anodized samples at different positions in a slab provided by an embodiment of the present invention;

[0037] Figure 19 Schematic diagram of the glossiness statistical results of the second group of polished anodized samples at different positions of a billet in a slab provided by an embodiment of the present invention;

[0038] Figure 20 Schematic diagram of the statistical results of the proportion of the glossiness of the samples within the compliance glossiness range in the first group of polished anodized samples and the second group of polished anodized samples at different positions of a billet in a slab provided by an embodiment of the present invention;

[0039] Figure 21 For Figure 4 Another flowchart included in S350 and S360 in;

[0040] Figure 22 Bar chart of the importance degree of 8 influencing factors screened by superimposing the proportional factors of the process parameters of the to-be-polished anodized samples according to a preset ratio using the LightGBM and RandomForest algorithms provided by an embodiment of the present invention;

[0041] Figure 23 Pie chart of the importance degree of 8 influencing factors screened by superimposing the proportional factors of the process parameters of the to-be-polished anodized samples according to a preset ratio using the LightGBM and RandomForest algorithms provided by an embodiment of the present invention;

[0042] Figure 24 Flowchart of another chemical polishing method based on deep learning provided by an embodiment of the present invention;

[0043] Figure 25 For Figure 24 A flowchart included in S430 in;

[0044] Figure 26 Schematic diagram of the structure of a chemical polishing device based on deep learning provided by an embodiment of the present invention;

[0045] Figure 27 Schematic diagram of the structure of an electronic device that can be used to implement the embodiments of the present invention provided by an embodiment of the present invention. Detailed implementation manners

[0046] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0047] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0048] In order to improve the automation degree of the chemical polishing method for the to-be-polished anodized sample, an embodiment of the present invention provides a chemical polishing method based on deep learning. Figure 1 The following is a flowchart of a chemical polishing method based on deep learning provided by an embodiment of the present invention. This method can be executed by a chemical polishing device based on deep learning. The chemical polishing device based on deep learning can be implemented in the form of hardware and / or software, and the chemical polishing device based on deep learning can be configured in a chemical polishing electronic device based on deep learning. As Figure 1 shown, the method includes:

[0049] S110. Determine the difference data of the process parameters of the to-be-polished anodized sample on the chemical polishing glossiness and / or on the chemical polishing time. The difference data includes influence result data and importance degree data.

[0050] Among them, the to-be-polished anodized sample is a to-be-polished sample prepared by an anodizing process.

[0051] Optionally, the process parameters of the anodized sample to be polished include at least one of whether the anodized sample to be polished is a rework product, the area of the chemically polished surface of the anodized sample to be polished, the position information of the chemically polished surface of the anodized sample to be polished, the production batch of the anodized sample to be polished, the production supply source of the anodized sample to be polished, the material type of the anodized sample to be polished, the production device number of the anodized sample to be polished, the position of the anodized sample to be polished in the production device, and the primary chemical polishing parameters of the anodized sample to be polished. Among them, the primary chemical polishing parameters of the anodized sample to be polished include at least one of the chemical polishing tank number, chemical polishing time, chemical polishing temperature, chemical polishing corrosion liquid type, sampling time, chemical polishing specific gravity, and chemical polishing corrosion liquid concentration.

[0052] Specifically, the anodized sample to be polished includes rework products and non-rework products. A rework product refers to a sample for secondary chemical polishing of the anodized sample to be polished. A non-rework product refers to a sample for primary chemical polishing of the anodized sample to be polished. In the embodiments of the present invention, non-rework products can also be referred to as normal products.

[0053] According to the difference in the area of the chemically polished surface of the anodized sample to be polished, the chemically polished surface of the anodized sample to be polished can be divided into a side and a large surface. The side is located on one side of the large surface, and the area of the side is smaller than that of the large surface. Among them, the side is located at the edge position of the chemically polished surface. Correspondingly, the position of the chemically polished surface of the anodized sample to be polished can include the side and the large surface.

[0054] The process processes, sample properties, and surface morphologies of the anodized samples to be polished in the same production batch are relatively similar.

[0055] The production supply source of the anodized sample to be polished refers to the production supplier of the anodized sample to be polished. There may be differences in the process processes, sample properties, and surface morphologies of the samples from different production suppliers.

[0056] The material of the anodized sample to be polished refers to the original alloy of the anodized sample to be polished.

[0057] The production device number of the anodized sample to be polished includes the production furnace number of the anodized sample to be polished. There may be differences in the process processes, sample properties, and surface morphologies of the anodized samples to be polished produced by different furnace numbers.

[0058] The position of the anodized sample to be polished in the production device refers to the position of the steel corresponding to the original alloy in the billet.

[0059] The primary chemical polishing parameters of the anodized sample to be polished include at least one of the chemical polishing tank number, chemical polishing time, chemical polishing temperature, type of chemical polishing corrosion liquid, sampling time, chemical polishing specific gravity, and concentration of chemical polishing corrosion liquid. Exemplarily, the chemical polishing corrosion liquid comprises phosphoric acid, sulfuric acid, polishing additive (or nitric acid), moisture, aluminum salt, and so on. The concentration of the chemical polishing corrosion liquid includes at least one of the concentrations of phosphoric acid, sulfuric acid, polishing additive, and aluminum ions.

[0060] Among them, a certain number of anodized samples to be polished can be used as new feedstock for chemical polishing, and then the process parameters of the anodized samples to be polished are collected by the data collection module for the difference data of chemical polishing glossiness and / or chemical polishing time. The difference data includes impact result data and importance degree data.

[0061] S120. Input the process parameters and difference data of the anodized sample to be polished with an importance degree greater than the preset degree into a deep learning neural network with self-learning ability to predict the polishing time of the anodized sample to be polished.

[0062] In the process of predicting the polishing time of the anodized sample to be polished through a deep learning neural network with self-learning ability, for the process parameters that have a relatively large impact on glossiness and / or chemical polishing time.

[0063] The technical solution provided by the embodiment of the present invention determines the difference data of the process parameters of the anodized sample to be polished for chemical polishing glossiness and / or chemical polishing time. The difference data includes impact result data and importance degree data, and inputs the process parameters and difference data of the anodized sample to be polished with an importance degree greater than the preset degree into a deep learning neural network with self-learning ability to predict the polishing time of the anodized sample to be polished. The prediction of the above polishing time combines the artificial intelligence (AI) algorithm to stabilize the chemical polishing process. Through an intelligent control system, numerous process parameters are monitored, and the chemical polishing duration is precisely adjusted, avoiding human operation errors and greatly improving the automation level of the production line. It not only improves production efficiency but also reduces the scrap rate caused by process fluctuations, achieving a higher level of precision manufacturing.

[0064] Figure 2 It is a flowchart of another chemical polishing method based on deep learning provided by the embodiment of the present invention. On the basis of the above embodiment, S120 is further refined. As Figure 2 shown, the chemical polishing method based on deep learning includes the following steps:

[0065] S210. Determine the difference data of the process parameters of the anodized sample to be polished with respect to the chemical polishing glossiness and / or the chemical polishing time. The difference data includes the influence result data and the importance data.

[0066] S220. Input the process parameters and the difference data of the anodized sample to be polished with an importance greater than the preset level into a deep learning neural network with self-learning ability.

[0067] S230. When the process parameters and the difference data of the anodized sample to be polished with an importance greater than the preset level change, the self-learning module of the deep learning neural network with self-learning ability updates the training model of the deep learning neural network according to the updated process parameters and the difference data of the anodized sample to be polished with an importance greater than the preset level, and predicts the polishing time of the anodized sample to be polished.

[0068] Figure 3 is a schematic structural diagram of a chemical polishing device based on deep learning provided by an embodiment of the present invention. As Figure 3 shown, on the basis of the above technical solution, in this embodiment, a certain number of anodized samples to be polished can be used as new inputs for chemical polishing, and then the data collection module collects the difference data of the process parameters of the anodized samples to be polished with respect to the chemical polishing glossiness and / or the chemical polishing time. The difference data includes the influence result data and the importance data. Based on the self-learning framework, the self-learning module can track the changes in the process parameters of the production line in real time, incorporate the changed parameters into the model in a timely manner, and keep the model in an optimal state all the time. The inference module predicts the polishing time of the anodized sample to be polished according to the process parameters and the difference data of the anodized sample to be polished with an importance greater than the preset level and the updated training model of the deep learning neural network.

[0069] Figure 4 is a flowchart of another chemical polishing method based on deep learning provided by an embodiment of the present invention. As Figure 4 shown, the chemical polishing method based on deep learning includes the following steps:

[0070] S310. Collect a first quantity of anodized samples to be polished.

[0071] S320. Perform chemical polishing on the first quantity of anodized samples to be polished to obtain polished anodized samples.

[0072] S330. Use the polished anodized samples with a glossiness within a preset range as a second quantity of polished anodized samples, and the second quantity is less than or equal to the first quantity.

[0073] S340. Divide the second quantity of polished anodized samples into a first group of polished anodized samples and a second group of polished anodized samples.

[0074] S350. Determine the difference data of the process parameters of the anodized sample to be polished with respect to chemical polishing glossiness and / or chemical polishing time. The difference data includes impact result data and importance degree data.

[0075] Based on the process parameters and chemical polishing glossiness of the first group of polished anodized samples and the second group of polished anodized samples, determine the difference data of the process parameters of the anodized sample to be polished with respect to chemical polishing glossiness and / or chemical polishing time. The difference data includes impact result data and importance degree data.

[0076] S360. Input the process parameters and difference data of the anodized sample to be polished with an importance degree greater than a preset degree into a deep learning neural network with self-learning ability, and predict the polishing time of the anodized sample to be polished.

[0077] Exemplarily, in the embodiment of the present invention, 283,884 (the first quantity) anodized samples to be polished of 2,199 rods can be collected. Then, the 283,884 anodized samples to be polished are chemically polished through a chemical polishing process to obtain polished anodized samples. The polished anodized samples with a glossiness of 5 - 30 GU are used as the second quantity of polished anodized samples, and the corresponding value of the second quantity is 282,841 polished anodized samples of 2,009 rods. Divide the 282,841 polished anodized samples into a first group of polished anodized samples and a second group of polished anodized samples. Based on the process parameters and chemical polishing glossiness of the first group of polished anodized samples and the second group of polished anodized samples, determine the difference data of the process parameters of the anodized sample to be polished with respect to chemical polishing glossiness and / or chemical polishing time. The difference data includes impact result data and importance degree data. And input the process parameters and difference data of the anodized sample to be polished with an importance degree greater than a preset degree into a deep learning neural network with self-learning ability, and predict the polishing time of the anodized sample to be polished.

[0078] Figure 5 For Figure 4 a flowchart included in S350. It should be noted that non-rework products can also be referred to as normal products in the embodiment of the present invention. Figure 6 It is a schematic diagram of the glossiness statistics results of the sides of rework products and normal products in a first group of polished anodized samples and the glossiness statistics results of the large surfaces of rework products and normal products in a second group of polished anodized samples provided by the embodiment of the present invention. Figure 7Schematic diagram of the statistical results of the median glossiness of the sides of rework products and normal products in a first group of polished anodized samples and the statistical results of the median glossiness of the large surfaces of rework products and normal products in a second group of polished anodized samples provided by the embodiments of the present invention. Figure 8 Schematic diagram of the result that the glossiness of the sides of rework products and normal products in a first group of polished anodized samples provided by the embodiments of the present invention conforms to a normal distribution. Figure 9 Schematic diagram showing the result that the glossiness of the large surfaces of rework products and normal products in a second group of polished anodized samples conforms to a normal distribution. Figure 10 Schematic diagram of the Pearson coefficient of the increase in the number of rework products and the shortening of the chemical polishing time in a first group of polished anodized samples and the Pearson coefficient of the decrease in the number of rework products and the shortening of the chemical polishing time in a second group of polished anodized samples provided by the embodiments of the present invention. Figure 11 Schematic diagram of the Spearman coefficient of the increase in the number of rework products and the shortening of the chemical polishing time in a first group of polished anodized samples and the decrease in the number of rework products and the shortening of the chemical polishing time in a second group of polished anodized samples provided by the embodiments of the present invention.

[0079] As Figure 5 As shown, the chemical polishing method based on deep learning includes the following steps: In the study of the influence of rework products on chemical polishing, the differential data of S350 for determining whether the anodized sample to be polished is a rework product, the area of the chemical polishing surface of the anodized sample to be polished, and the position information of the chemical polishing surface of the anodized sample to be polished on the chemical polishing glossiness and / or on the chemical polishing time includes:

[0080] S3501. Statistically calculate at least one of the average glossiness of the sides of rework products in a first group of polished anodized samples, the average glossiness of the sides of non-rework products in a first group of polished anodized samples, the overall average glossiness of the sides of rework products and non-rework products in a first group of polished anodized samples, the median glossiness of the sides of rework products in a first group of polished anodized samples, the median glossiness of the sides of non-rework products in a first group of polished anodized samples, and the overall median glossiness of the sides of rework products and non-rework products in a first group of polished anodized samples, and determine the influence of rework products and non-rework products on the chemical polishing glossiness of the sides of the anodized sample to be polished.

[0081] S3502. Statistically analyze the average glossiness of the major surfaces of the rework products in the second group of polished anodized samples, the average glossiness of the major surfaces of the non-rework products in the second group of polished anodized samples, and the overall average glossiness of the major surfaces of the rework products and non-rework products in the second group of polished anodized samples. Also, analyze the median glossiness of the major surfaces of the rework products in the second group of polished anodized samples, the median glossiness of the major surfaces of the non-rework products in the second group of polished anodized samples, and the overall median glossiness of the major surfaces of the rework products and non-rework products in the second group of polished anodized samples. Determine the influence of rework products and non-rework products on the chemical polishing glossiness of the major surfaces of the polished anodized samples when one of the words is missing.

[0082] As Figure 6 shown, conduct a statistical analysis of the glossiness of normal products and rework products. The average glossiness of the rework products on the sides of the first group of polished anodized samples (Gloss_p1) is lower than the overall average glossiness of the rework products and normal products on the sides of the first group of polished anodized samples (Gloss_p1). There is no significant difference between the average glossiness of the normal products on the major surfaces of the second group of polished anodized samples (Gloss_p3), the average glossiness of the rework products, and the overall average glossiness.

[0083] As Figure 7 shown, conduct a statistical analysis of the glossiness of normal products and rework products. The median glossiness of the rework products on the sides of the first group of polished anodized samples (Gloss_p1) is lower than the overall median glossiness. There is no significant difference between the median glossiness of the normal products and rework products on the major surfaces of the second group of polished anodized samples (Gloss_p3) and the overall median glossiness.

[0084] S3503. When the influence of rework products and non-rework products on the chemical polishing glossiness of the sides of the polished anodized samples conforms to a normal distribution, and the influence of rework products and non-rework products on the chemical polishing glossiness of the major surfaces of the polished anodized samples conforms to a normal distribution, determine the chemical polishing time for the mixed placement of rework products and non-rework products for chemical polishing through linear analysis and non-linear analysis, which is greater than the chemical polishing time for the separate placement of rework products and non-rework products for chemical polishing. Among them, when predicting the polishing time through a chemical polishing method based on deep learning, predict the polishing time of rework products and non-rework products separately.

[0085] As Figure 8 shown, the glossiness of the corresponding sides of the rework products and normal products in the first group of polished anodized samples (Gloss_p1) conforms to a normal distribution. As Figure 9 shown, the glossiness of the corresponding major surfaces of the rework products and normal products in the second group of polished anodized samples (Gloss_p3) conforms to a normal distribution.

[0086] As Figure 10 andFigure 11 As shown in Figure 11 , the linear / non-linear relationship analysis of rework products and normal products under mixed feeding is as follows. The conclusion is that mixed feeding is not conducive to the refined management of the chemical polishing duration of rework products and the prediction of the first-pass chemical polishing by artificial intelligence (AI). In actual production experience, the chemical polishing time of rework products is slightly longer than that of normal products. Therefore, we improved the previous process and fed rework products and non-rework products separately to ensure that the predicted chemical polishing time obtained through the deep learning neural network with autonomous learning ability is more accurate. Among them, Figure 10 It can be seen from Figure 10 that there is a negative correlation between the increase in the number of rework products in the first group of polished anodized samples (Gloss_p1) and the shortening of the chemical polishing time, and there is a positive correlation between the decrease in the number of rework products in the second group of polished anodized samples (Gloss_p3) and the shortening of the chemical polishing time, that is, as the number of rework products increases, the chemical polishing time will increase. Figure 11 It can be seen from Figure 11 that there is a negative correlation between the increase in the number of rework products in the first group of polished anodized samples (Gloss_p1) and the shortening of the chemical polishing time, and there is a positive correlation between the decrease in the number of rework products in the second group of polished anodized samples (Gloss_p3) and the shortening of the chemical polishing time, that is, as the number of rework products increases, the chemical polishing time will increase. Figure 10 The analysis result in Figure 10 is the result of the linear relationship analysis of rework products and normal products under mixed feeding, Figure 11 The analysis result in Figure 11 is the result of the linear relationship analysis of rework products and normal products under mixed feeding.

[0087] Figure 12 This is a schematic diagram of the statistical results of the glossiness of the sides of the first group of polished anodized samples from different suppliers and the statistical results of the glossiness of the large surfaces of the samples in the second group of polished anodized samples provided by the embodiments of the present invention. Figure 13 This is a schematic diagram of the statistical results of the compliance ratio of the glossiness of the sides of the first group of polished anodized samples from different suppliers and the compliance ratio of the glossiness of the large surfaces of the second group of polished anodized samples provided by the embodiments of the present invention. Figure 14 This is a schematic diagram of the statistical results of the proportion of the glossiness of the sides and large surfaces of the samples in the first group of polished anodized samples and the second group of polished anodized samples from different suppliers within the compliance glossiness range provided by the embodiments of the present invention.

[0088] Or, for the study of the influence of suppliers on chemical polishing in raw materials, the difference data of the production supply sources of the anodized samples to be polished on chemical polishing glossiness and / or on chemical polishing time determined by S350 include:

[0089] S3504. Based on at least one of the average glossiness, overall average glossiness, glossiness median, glossiness median, finished product glossiness distribution, glossiness compliance ratio, and the number of polished anodized samples from different production and supply sources on the sides of the first group of polished anodized samples, and at least one of the average glossiness, overall average glossiness, glossiness median, glossiness median, finished product glossiness distribution, glossiness compliance ratio, and the number of polished anodized samples from different production and supply sources on the large surfaces of the second group of polished anodized samples, determine the preferred production and supply source. The glossiness compliance ratio of the anodized samples to be polished from the preferred production and supply source is greater than a preset ratio, so as to input the production and supply source and difference data of the anodized samples to be polished into a deep learning neural network with self-learning ability to predict the polishing time of the anodized samples to be polished.

[0090] It is known that different suppliers, or different alloys from the same supplier, or even raw materials from different batches may affect the chemical polishing glossiness. For example Figure 12 、 Figure 13 and Figure 14 shown, the glossiness compliance ratios of the samples from suppliers IM and KJ for the corresponding sides and large surfaces are higher respectively. It should be noted that the quantity of IM accounts for a relatively large proportion compared with the other three suppliers, and the influence of the data quantity cannot be excluded.

[0091] Figure 15 FIG. is a schematic diagram of the glossiness statistical results of the sides of the first group of polished anodized samples with different alloy codes and the glossiness statistical results of the large surfaces of the second group of polished anodized samples provided by an embodiment of the present invention. Figure 16 FIG. is a schematic diagram of the ratio statistical results of the glossiness of the sides of the first group of polished anodized samples with different alloy codes within the compliance glossiness range and the ratio statistical results of the glossiness of the large surfaces of the second group of polished anodized samples with different alloy codes within the compliance glossiness range provided by an embodiment of the present invention.

[0092] Alternatively, for the study of the influence of material types in raw materials on chemical polishing, the difference data of the material types of the anodized samples to be polished determined by S350 on chemical polishing glossiness and / or on chemical polishing time include:

[0093] S3505. Determine the preferred material type based on at least one of the average gloss, overall average gloss, median gloss, median gloss, finished product gloss distribution, gloss compliance ratio, and the number of polished anodized samples of different material types on the sides of the polished anodized samples in the first group, and at least one of the average gloss, overall average gloss, median gloss, median gloss, finished product gloss distribution, gloss compliance ratio, and the number of polished anodized samples of different material types on the large surfaces of the polished anodized samples in the second group. The gloss compliance ratio of the polished anodized samples of the preferred material type is greater than the preset ratio. Then, input the material type and difference data of the polished anodized samples to be polished into a deep learning neural network with self-learning ability to predict the polishing time of the polished anodized samples to be polished.

[0094] As Figure 15 and Figure 16 shown, the raw material alloy is analyzed as follows. It can be seen that the ratio of the gloss on the sides of the first group of polished anodized samples (Gloss_p1) of alloy 6R02 type within the compliance gloss range is higher. And the ratio of the gloss on the large surfaces of the second group of polished anodized samples of alloy 6R02 type within the compliance gloss range is higher. Also, the gloss on the sides of the first group of polished anodized samples (Gloss_p1) of alloy 6R02 type is higher, and the gloss on the large surfaces of the second group of polished anodized samples of alloy 6R02 type is higher. In summary, the polished anodized samples of 6R01 alloy are superior to those of 6R02 alloy in terms of the finished product gloss performance. The raw material alloy will affect the finished product gloss. Consider adding the raw material alloy as an influencing factor to the chemical polishing model of the chemical polishing method based on deep learning.

[0095] Figure 17 This is a schematic diagram of the statistical results of the ratio of the gloss of the first group of polished anodized samples and the second group of polished anodized samples of different production furnace numbers within the compliance gloss range provided by the embodiment of the present invention.

[0096] Alternatively, for the study of the influence of the production furnace number in the raw material on chemical polishing, the difference data determined by S350 for the production furnace number of the polished anodized samples to be polished on chemical polishing gloss and / or on chemical polishing time includes:

[0097] S3506. Determine the preferred production furnace number based on at least one of the average gloss, overall average gloss, gloss median, gloss median, finished product gloss distribution, gloss compliance ratio, and the number of polished anodized samples with different production furnace numbers on the side of the polished anodized samples in the first group, and at least one of the average gloss, overall average gloss, gloss median, gloss median, finished product gloss distribution, gloss compliance ratio, and the number of polished anodized samples with different production furnace numbers on the large surface of the polished anodized samples in the second group. The gloss compliance ratio of the polished anodized samples with the preferred production furnace number is greater than the preset ratio. Then, input the production furnace number and difference data of the polished anodized samples to be polished into a deep learning neural network with autonomous learning ability to predict the polishing time of the polished anodized samples to be polished.

[0098] As Figure 17 shown, through the analysis of the qualified rate of gloss under different furnace numbers, it can be seen that the finished products with furnace numbers 6, 7, 2, 8, and 9 have the best gloss, followed by 5, 3, and 4. The furnace number will affect the finished product gloss. It is considered to add the furnace number as an influencing factor to the chemical polishing model of the chemical polishing method based on deep learning.

[0099] Figure 18 It is a schematic diagram of the gloss statistical results of the first group of polished anodized samples at different positions in a slab provided by an embodiment of the present invention. Figure 19 It is a schematic diagram of the gloss statistical results of the second group of polished anodized samples at different positions of a steel billet in a slab provided by an embodiment of the present invention. Figure 20 It is a schematic diagram of the statistical results of the ratio of the gloss of the samples within the compliance gloss range in the first group of polished anodized samples and the second group of polished anodized samples at different positions of a steel billet in a slab provided by an embodiment of the present invention.

[0100] Alternatively, for the study of the influence of the production device on chemical polishing in the raw material, the difference data determined by S350 for the position of the polished anodized sample to be polished in the production device on the chemical polishing gloss and / or on the chemical polishing time includes:

[0101] S3507. Determine the preferred position of the steel billet in the continuous casting billet based on at least one of the average glossiness value, overall average glossiness value, median glossiness value, median glossiness value, finished product glossiness distribution, glossiness compliance ratio, and the number of polished anodized samples at different positions of the steel billet in the continuous casting billet on the side edges of the polished anodized samples in the first group, and at least one of the average glossiness value, overall average glossiness value, median glossiness value, median glossiness value, finished product glossiness distribution, glossiness compliance ratio, and the number of polished anodized samples at different positions of the steel billet in the continuous casting billet on the large surfaces of the polished anodized samples in the second group. The glossiness compliance ratio of the polished anodized sample to be polished at the preferred position of the steel billet in the continuous casting billet is greater than the preset ratio. Input the different positions of the steel billet of the polished anodized sample to be polished in the continuous casting billet and the difference data into a deep learning neural network with self-learning ability to predict the polishing time of the polished anodized sample to be polished.

[0102] As Figures 18 - 20 shown, by analyzing the position of the steel billet in the continuous casting billet, it is found that BO1 is the preferred position of the steel billet in the continuous casting billet. However, since the number of BO1 is too small to be statistically significant, the position of the steel billet in the continuous casting billet has no obvious effect on the finished product glossiness. Therefore, it is not necessary to input the different positions of the steel billet of the polished anodized sample to be polished in the continuous casting billet and the difference data into a deep learning neural network with self-learning ability to predict the polishing time of the polished anodized sample to be polished.

[0103] Figure 21 For Figure 4 another flowchart included in S350 and S360, as Figure 21 shown, S350 determines whether the polished anodized sample is a rework product, the area of the chemically polished surface of the polished anodized sample to be polished, and the position information of the chemically polished surface of the polished anodized sample to be polished. The difference data for the chemically polished glossiness and / or the chemically polished time includes:

[0104] S3508. Use the preset weight superposition algorithm to superpose the importance of the proportional factors of the process parameters of the polished anodized sample to be polished according to the preset ratio, and screen out the process parameters of the polished anodized sample to be polished whose importance degree is greater than the preset degree; the preset weight superposition algorithm includes the LightGBM algorithm and the Random Forest algorithm.

[0105] Exemplarily, a total of 47 influencing factors provided by engineering and production personnel are used to superpose the factor importance according to the ratio of 3.2:6.8 by using the LightGBM and Random Forest algorithms, and the top 8 most important influencing factors are screened out and sorted as followsFigure 22 and Figure 23 as shown. Among them, Figure 22 is a bar chart showing the importance levels of 8 influencing factors selected by superimposing the proportional factor importance of the process parameters of the anodized sample to be polished using the LightGBM and RandomForest algorithms according to a preset ratio in an embodiment of the present invention. Figure 23 is a pie chart showing the importance levels of 8 influencing factors selected by superimposing the proportional factor importance of the process parameters of the anodized sample to be polished using the LightGBM and RandomForest algorithms according to a preset ratio in an embodiment of the present invention.

[0106] S360. When inputting the process parameters and difference data of the anodized sample to be polished with an importance level greater than a preset level into a deep learning neural network with self-learning ability to predict the polishing time of the anodized sample to be polished, it includes:

[0107] S3601. Analyze the stability of the process parameters of the anodized sample to be polished with an importance level greater than a preset level on the influencing result data.

[0108] S3602. Control the process parameters of the anodized sample to be polished with an importance level greater than a preset level within a stable fluctuation range, and predict the polishing time of the anodized sample to be polished.

[0109] Exemplarily, stabilize the above 8 influencing factors to control them within a stable fluctuation range to prevent large fluctuations from affecting the prediction of the chemical polishing time.

[0110] Figure 24 is a flowchart of another chemical polishing method based on deep learning provided in an embodiment of the present invention. On the basis of the above embodiment, S120 is further refined. As Figure 24 shown, this chemical polishing method based on deep learning includes the following steps:

[0111] S410. Determine the difference data of the process parameters of the anodized sample to be polished on the chemical polishing glossiness and / or on the chemical polishing time. The difference data includes influencing result data and importance level data.

[0112] S420. Input the process parameters and difference data of the anodized sample to be polished with an importance level greater than a preset level into a deep learning neural network with self-learning ability.

[0113] S430. Select a preset optimizer based on a gradient-balanced adaptive hybrid optimizer to train the neural network in deep learning, which is used to adjust the parameter weights of the neural network and predict the polishing time of the anodized sample to be polished.

[0114] In the prior art, the mixed use of the AdamW optimizer and the SGD optimizer is mainly based on switching optimizers at fixed training stages. For example, in the early training stage, the AdamW optimizer is used for training, and in the later stage, the SGD optimizer is used for fine-tuning. Or, different optimizers are selected at fixed levels.

[0115] The disadvantages of the prior art are as follows: The AdamW optimizer and the SGD optimizer do not perform consistently under different data distributions, gradient change rates, batch sizes, etc. Sometimes, the AdamW optimizer may over-adjust the gradient in the early stage of training, while the SGD optimizer may still oscillate when locally converging in the later stage.

[0116] The existing methods cannot adjust the optimization strategy in real time during the training process and still require manually setting the switching time points or levels of the AdamW optimizer or the SGD optimizer.

[0117] In the embodiments of the present invention, a deep neural network (DNN) algorithm is used and optimized, and an adaptive hybrid optimizer based on gradient equilibrium (Gradient Equilibrium Hybrid Optimizer, GEHO) is designed to select a preset optimizer to train a neural network in deep learning, which is used to adjust the parameter weights of the neural network and predict the polishing time of the anodized sample to be polished. This method dynamically measures the balance of gradient updates and adaptively selects the AdamW optimizer or the SGD optimizer based on gradient information to adjust the parameter weights in deep learning, so that the loss function is as small as possible, rather than presetting the switching time or level.

[0118] Figure 25 For Figure 24 a flowchart included in S430, as Figure 25 shown, S430 selects a preset optimizer to train a neural network in deep learning based on the adaptive hybrid optimizer of gradient equilibrium, and is used to adjust the parameter weights of the neural network, including:

[0119] S4301. Calculate the gradient balance index; wherein, the gradient balance index satisfies the relationship:

[0120]

[0121] wherein, ▽L t is the gradient of the current batch, ▽L t-1 is the gradient of the previous batch, and ε is a constant to prevent division by zero errors;

[0122] S4302. When the change degree of the current gradient is greater than the set gradient equilibrium threshold, select the AdamW optimizer to train the neural network in deep learning for adjusting the parameter weights of the neural network;

[0123] In the case where the degree of change of the current gradient is less than or equal to the gradient setting equilibrium threshold, the SGD optimizer is selected to train the neural network in deep learning for adjusting the parameter weights of the neural network.

[0124] In the embodiment of the present invention, the working principle of GEHO is as follows: During the training process, the gradient equilibrium metric (GEM) is used to monitor the gradient change, and the selection of the optimizer is automatically adjusted based on the GEM feedback. The threshold equilibrium (TEQ) is set. When the gradient changes violently (is unbalanced), the AdamW optimizer is used to accelerate convergence. When the gradient tends to be stable (is balanced), the SGD optimizer is used to improve the generalization ability. In this way, the optimization strategy can be dynamically adjusted throughout the training process, rather than statically switched in stages. Among them, the gradient equilibrium metric satisfies the relationship:

[0125]

[0126] Among them, is the gradient of the current batch, is the gradient of the previous batch, and ε is a constant to prevent division-by-zero errors.

[0127] If GEM > TEQ (the gradient changes violently): It indicates that the gradient fluctuates greatly, and the AdamW optimizer is used for training.

[0128] If GEM ≤ TEQ (the gradient is relatively stable): It indicates that the gradient is stable, and the SGD optimizer is used for optimization.

[0129] In the embodiment of the present invention, the chemical polishing method based on deep learning includes the following steps: During the initialization process, the learning rate η, the weight decay parameter λ, the gradient setting equilibrium threshold (TEQ), and the hyperparameters of the AdamW optimizer and the SGD optimizer can be set. Among them, the gradient setting equilibrium threshold (TEQ) can be set to 0.1. During the training process, by calculating the gradient of the current batch the gradient equilibrium metric is calculated. When the gradient changes violently (is unbalanced), the AdamW optimizer is used to accelerate convergence. When the gradient tends to be stable (is balanced), the SGD optimizer is used to improve the generalization ability. And in the above process, the weight parameters need to be updated and the gradient value ▽L of the previous batch needs to be updated t-1 ←▽L t , and the above steps are repeated until convergence. As Figure 3As shown, based on the design of the autonomous learning architecture, due to the continuous increase in production data and the continuous optimization of the manufacturing process, the situation of each rod material may continue to change. Therefore, the autonomous learning architecture is designed with the aims of: accelerating the model iteration speed; always maintaining the latest model and providing the accuracy of predicting the polishing time of the anodized sample to be polished.

[0130] It should be noted that the engineering architecture of the chemical polishing method based on deep learning in the embodiments of the present invention is as follows: including an interface layer, an algorithm layer, a logic layer, and a data layer. The functions of each layer are as follows: The function of the Data Layer is to read the mapping relationship between the model type and color in the configuration file or database and provide a standardized data preprocessing function. The function of the Logic Layer is to define logical rules, such as which algorithms need to be called for a specific model type, and implement the dynamic scheduling of algorithm selection. The function of the Algorithm Layer is that each algorithm module is independent and does not affect each other, supports dynamic expansion, and no other modules need to be modified when adding new algorithms. The function of the Interface Layer is to provide an external interface for users to call, such as a prediction function, a model update function, etc., accept user input, and distribute the input to the logic layer. The purpose of the above engineering architecture is to modularize the chemical polishing method based on deep learning, on the one hand, separating the model type, color, and algorithm logic for easy maintenance; on the other hand, achieving high scalability to support the addition of new model types, colors, and algorithms.

[0131] The present invention optimizes the optimizer of the DNN algorithm, uses automatic adjustment of training weights to improve the model training effect, and the model accuracy reaches 99.3%. The present invention designs an autonomous learning architecture that can track the changes in the manufacturing process of the production line in real time, incorporate the changed parameters into the model in a timely manner, and keep the model in an optimal state all the time. Moreover, the present invention improves the degree of automation of the production line; the CPK of the finished product gloss is increased by about 15.4%. Among them, CPK refers to the Process Capability Index for the Centered Process, the process capability index with offset. Based on CP, CPK takes into account the offset between the process output mean and the specification center, and its calculation comprehensively considers the position and dispersion degree of the process.

[0132] Figure 26 It is a schematic structural diagram of a chemical polishing device based on deep learning provided by an embodiment of the present invention. As Figure 26 shown, the chemical polishing device based on deep learning includes:

[0133] A training data determination module 100 is configured to determine difference data of process parameters of an anodized sample to be polished with respect to chemical polishing glossiness and / or chemical polishing time. The difference data includes influence result data and importance degree data.

[0134] A prediction module 200 is configured to input process parameters and difference data of an anodized sample to be polished with an importance degree greater than a preset degree into a deep learning neural network with an autonomous learning ability, and predict the polishing time of the anodized sample to be polished.

[0135] The chemical polishing device based on deep learning provided by an embodiment of the present invention can execute the chemical polishing method based on deep learning provided by any embodiment of the present invention, and has function modules and beneficial effects corresponding to the execution of the method.

[0136] Figure 27 It is a schematic structural diagram of an electronic device that can be used to implement an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device (such as a helmet, glasses, a watch, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described herein and / or claimed.

[0137] As Figure 27 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. Various programs and data required for the operation of the electronic device 10 can also be stored in the RAM 13. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0138] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0139] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the chemical polishing method based on deep learning.

[0140] In some embodiments, the chemical polishing method based on deep learning can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the chemical polishing method based on deep learning described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute the chemical polishing method based on deep learning by any other suitable means (e.g., by means of firmware).

[0141] The various embodiments of the systems and technologies described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), systems-on-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: implemented in one or more computer programs, the one or more computer programs can be executed and / or interpreted on a programmable system including at least one programmable processor, the programmable processor can be a dedicated or general-purpose programmable processor, can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0142] A computer program for implementing the method of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, a special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer program may be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0143] In the context of the present invention, a computer-readable storage medium may be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium may be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0144] In order to provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0145] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0146] The computing system can include clients and servers. The clients and servers are generally far from each other and typically interact through a communication network. The client-server relationship is created by computer programs that run on the respective computers and have a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0147] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is imposed herein.

[0148] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A chemical polishing method based on deep learning, characterized in that: include: Determine the difference data of the process parameters of the anodized sample to be polished on the chemical polishing glossiness and / or the chemical polishing time, wherein the difference data includes the impact result data and the importance degree data; The process parameters of the anodized sample to be polished whose importance is greater than a preset degree and the difference data are input into a deep learning neural network with autonomous learning capability to predict the polishing time of the anodized sample to be polished.

2. The chemical polishing method based on deep learning according to claim 1, characterized in that: Inputting the process parameters of the anodized sample to be polished whose importance is greater than a preset degree and the difference data into a deep learning neural network with autonomous learning capability, and predicting the polishing time of the anodized sample to be polished includes: Inputting the process parameters of the anodized sample to be polished and the difference data whose importance is greater than a preset degree into a deep learning neural network with autonomous learning capability; The autonomous learning module of the deep learning neural network with autonomous learning capability updates the training model of the deep learning neural network and predicts the polishing time of the anodized sample to be polished based on the updated process parameters of the anodized sample to be polished with a degree of importance greater than a preset degree and the difference data, when the process parameters of the anodized sample to be polished with a degree of importance greater than a preset degree and the difference data change.

3. The chemical polishing method based on deep learning according to claim 1, characterized in that: The process parameters of the anodized sample to be polished include: whether the anodized sample to be polished is a rework, the area of ​​the chemically polished surface of the anodized sample to be polished, the position information of the chemically polished surface of the anodized sample to be polished, the production batch of the anodized sample to be polished, the production supply source of the anodized sample to be polished, the material type of the anodized sample to be polished, the production device number of the anodized sample to be polished, the position of the anodized sample to be polished in the production device, and at least one of the primary chemical polishing parameters of the anodized sample to be polished, wherein the primary chemical polishing parameters of the anodized sample to be polished include: at least one of the chemical polishing tank number, chemical polishing time, chemical polishing temperature, chemical polishing etching solution type, sampling time, chemical polishing specific gravity, and chemical polishing etching solution concentration.

4. The chemical polishing method based on deep learning according to claim 3 is characterized in that: Determine the difference data of the process parameters of the anodized sample to be polished on the chemical polishing glossiness and / or the chemical polishing time, wherein the difference data includes the impact result data and the importance data and also includes: collecting a first number of anodized samples to be polished; Chemically polishing the first number of anodized samples to be polished to obtain polished anodized samples; taking the polished anodized samples whose glossiness is within a preset range as a second number of polished anodized samples, wherein the second number is less than or equal to the first number; dividing the second number of polished anodized samples into a first group of polished anodized samples and a second group of polished anodized samples; The difference data for determining whether the anodized sample to be polished is a reworked product, the area of ​​the chemically polished surface of the anodized sample to be polished, and the position information of the chemically polished surface of the anodized sample to be polished to the chemical polishing glossiness and / or the chemical polishing time include: Count at least one of the following: the average glossiness of the reworked product side in the first group of polished anodized samples, the average glossiness of the non-reworked product side in the first group of polished anodized samples, the overall average glossiness of the reworked product side and the non-reworked product side in the first group of polished anodized samples, the median glossiness of the reworked product side in the first group of polished anodized samples, the median glossiness of the non-reworked product side in the first group of polished anodized samples, and the overall median glossiness of the reworked product side and the non-reworked product side in the first group of polished anodized samples, and determine the influence of the reworked product and the non-reworked product on the chemical polishing glossiness of the side of the polished anodized sample; Count at least one of the following: the average glossiness of the large surface of the reworked products in the second group of polished anodized samples, the average glossiness of the large surface of the non-reworked products in the second group of polished anodized samples, and the overall average glossiness of the large surface of the reworked products and the large surface of the non-reworked products in the second group of polished anodized samples, the median glossiness of the large surface of the reworked products in the second group of polished anodized samples, the median glossiness of the large surface of the non-reworked products in the second group of polished anodized samples, and the overall median glossiness of the large surface of the reworked products and the large surface of the non-reworked products in the second group of polished anodized samples, and determine the influence of the reworked products and the non-reworked products on the chemical polishing glossiness of the large surface of the polished anodized samples; When the influence of the rework and non-rework on the chemical polishing glossiness of the side of the anodized sample to be polished conforms to the normal distribution, and when the influence of the rework and non-rework on the chemical polishing glossiness of the large surface of the anodized sample to be polished conforms to the normal distribution, it is determined through linear analysis and nonlinear analysis that the chemical polishing time of the mixed rework and non-rework is greater than the chemical polishing time of the separate rework and non-rework; wherein, when the polishing time is predicted by the chemical polishing method based on deep learning, the polishing time of the rework and non-rework is predicted separately; Alternatively, data to determine differences in chemical polishing gloss and / or chemical polishing time between production and supply sources of anodized samples to be polished include: According to at least one of the average glossiness of the sides of the polished anodized samples from different production and supply sources in the first group of polished anodized samples, the overall average glossiness, the median glossiness, the median glossiness, the gloss distribution of the finished product, the glossiness compliance ratio, and the number of polished anodized samples from different production and supply sources, and at least one of the average glossiness of the large surfaces of the polished anodized samples from different production and supply sources in the second group of polished anodized samples, the preferred production and supply source is determined, and the gloss compliance ratio of the anodized samples to be polished from the preferred production and supply source is greater than a preset ratio, so as to input the production and supply source of the anodized samples to be polished and the difference data into a deep learning neural network with autonomous learning ability, and predict the polishing time of the anodized samples to be polished; Alternatively, data for determining the effect of the type of material of the anodized sample to be polished on the chemical polishing gloss and / or on the chemical polishing time include: According to at least one of the average glossiness of the sides of the polished anodized samples of different material types in the first group of polished anodized samples, the overall average glossiness, the median glossiness, the median glossiness, the gloss distribution of the finished product, the glossiness compliance ratio, and the number of polished anodized samples of different material types, and at least one of the average glossiness of the large surfaces of the polished anodized samples of different material types in the second group of polished anodized samples, the preferred material type is determined, and the gloss compliance ratio of the anodized samples to be polished of the preferred material type is greater than a preset ratio, so as to input the material type of the anodized sample to be polished and the difference data into a deep learning neural network with autonomous learning ability, and predict the polishing time of the anodized sample to be polished; Alternatively, the data for determining the difference between the production heat number of the anodized sample to be polished and the chemical polishing gloss and / or the chemical polishing time include: According to at least one of the average glossiness of the sides of the polished anodized samples with different production furnace numbers in the first group of polished anodized samples, the overall average glossiness, the median glossiness, the median glossiness, the gloss distribution of the finished product, the glossiness compliance ratio, and the number of polished anodized samples with different production furnace numbers, and at least one of the average glossiness of the large surfaces of the polished anodized samples with different production furnace numbers in the second group of polished anodized samples, the preferred production furnace number is determined, and the gloss compliance ratio of the anodized samples to be polished of the preferred production furnace number is greater than a preset ratio, so as to input the production furnace number of the anodized samples to be polished and the difference data into a deep learning neural network with autonomous learning ability, and predict the polishing time of the anodized samples to be polished; Alternatively, the data for determining the difference in chemical polishing glossiness and / or chemical polishing time due to the position of the anodized sample to be polished in the production device includes: According to at least one of the average glossiness of the sides of the polished anodized samples at different positions of the steel billet in the cast billet in the first group of polished anodized samples, the overall average glossiness, the median glossiness, the median glossiness, the gloss distribution of the finished product, the gloss compliance ratio and the number of polished anodized samples at different positions of the steel billet in the cast billet, and at least one of the average glossiness of the large surface of the polished anodized samples at different positions of the steel billet in the cast billet in the second group of polished anodized samples, the preferred position of the steel billet in the cast billet is determined, and the gloss compliance ratio of the anodized samples to be polished at the preferred position of the steel billet in the cast billet is greater than a preset ratio, so that the different positions of the steel billet in the cast billet and the difference data of the anodized samples to be polished are input into a deep learning neural network with autonomous learning ability, and the polishing time of the anodized samples to be polished is predicted.

5. The chemical polishing method based on deep learning according to claim 3, characterized in that: The difference data for determining whether the polished anodized sample is a reworked product, the area of ​​the chemically polished surface of the anodized sample to be polished, and the position information of the chemically polished surface of the anodized sample to be polished to the chemical polishing glossiness and / or the chemical polishing time include: Using a preset weight superposition algorithm to superimpose the importance of the proportional factors of the process parameters of the anodized sample to be polished according to a preset proportion, and screening out the process parameters of the anodized sample to be polished whose importance is greater than the preset degree; the preset weight superposition algorithm includes the LightGBM algorithm and the Random Forest algorithm; Inputting the process parameters of the anodized sample to be polished and the difference data whose importance is greater than a preset degree into a deep learning neural network with autonomous learning capability, and predicting the polishing time of the anodized sample to be polished includes: Analyzing the influence of the process parameters of the anodized sample to be polished whose importance is greater than a preset degree on the stability of the result data; The process parameters of the anodized sample to be polished whose importance is greater than a preset degree are controlled within a stable fluctuation range, and the polishing time of the anodized sample to be polished is predicted.

6. The chemical polishing method based on deep learning according to claim 1 or 2, characterized in that: Inputting the process parameters of the anodized sample to be polished whose importance is greater than a preset degree and the difference data into a deep learning neural network with autonomous learning capability, and predicting the polishing time of the anodized sample to be polished includes: Inputting the process parameters of the anodized sample to be polished and the difference data whose importance is greater than a preset degree into a deep learning neural network with autonomous learning capability; The adaptive hybrid optimizer based on gradient balance selects a preset optimizer to train a neural network in deep learning, so as to adjust the parameter weights of the neural network and predict the polishing time of the anodized sample to be polished.

7. The chemical polishing method based on deep learning according to claim 6, characterized in that: The adaptive hybrid optimizer based on gradient balancing selects the preset optimizer to train the neural network in deep learning. The parameters weights used to adjust the neural network include: Calculate the gradient balance index; wherein the gradient balance index satisfies the relationship: Among them, ▽L t is the gradient of the current batch, ▽L t-1 is the gradient of the previous batch, and ε is a constant to prevent division by zero errors; When the current gradient change is greater than the gradient set equalization threshold, the AdamW optimizer is selected to train the neural network in deep learning to adjust the parameter weights of the neural network; When the current gradient change is less than or equal to the gradient set balance threshold, the SGD optimizer is selected to train the neural network in deep learning to adjust the parameter weights of the neural network.

8. A chemical polishing device based on deep learning, characterized in that: include: A training data determination module, used to determine difference data of process parameters of anodized samples to be polished on chemical polishing glossiness and / or chemical polishing time, wherein the difference data includes impact result data and importance data; The prediction module is used to input the process parameters of the anodized sample to be polished whose importance is greater than a preset degree and the difference data into a deep learning neural network with autonomous learning ability, so as to predict the polishing time of the anodized sample to be polished.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the deep learning-based chemical polishing method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the deep learning-based chemical polishing method described in any one of claims 1 to 7 when executed.

Images