Circuit board production quality evaluation system based on artificial intelligence

Abstract

The invention, which relates to the technical field of circuit board quality evaluation, discloses an artificial intelligence-based circuit board production quality evaluation system comprising a cloud platform, a data acquisition module, a feature extraction module, an artificial intelligence evaluation module and an early warning execution module. The data acquisition module acquires image data and process parameters in the circuit board production process, respective production-related data sets are obtained after data preprocessing, and the feature extraction module performs feature extraction on the production-related data sets to obtain feature point inspection items for representing the quality of the circuit board in the circuit board production process. And the artificial intelligence evaluation module constructs a circuit board quality evaluation model, obtains all the feature point inspection items after ciphertext transmission, carries out ciphertext decoding, and inputs the decoded feature point inspection items into the circuit board quality evaluation model to obtain the quality grade of the corresponding circuit board. And the early warning execution module carries out early warning on the circuit boards with substandard quality grades and notifies related operators to carry out screening.

Description

Technical Field

[0001] The invention relates to the technical field of circuit board quality assessment, and in particular to an artificial intelligence-based circuit board production quality assessment system. Background Art

[0002] In the existing circuit board production process, the quality assessment of circuit boards mainly relies on manual inspection. This method has certain limitations. First, manual inspection is inefficient and difficult to meet the needs of large-scale production. Second, manual inspection is easily affected by subjective factors, resulting in inaccurate inspection results.

[0003] In addition, the existing quality assessment system has security risks in the data transmission process, which can easily lead to data leakage. Therefore, it is necessary to study an artificial intelligence-based circuit board production quality assessment system to improve production efficiency, ensure the accuracy of the assessment results, and protect data security. Summary of the invention

[0004] In order to solve the above problems, the purpose of the present invention is to provide a circuit board production quality assessment system based on artificial intelligence.

[0005] The object of the present invention can be achieved by the following technical solutions: A circuit board production quality assessment system based on artificial intelligence, comprising a cloud platform, wherein the cloud platform is communicatively connected with a data acquisition module, a feature extraction module, an artificial intelligence assessment module and an early warning execution module;

[0006] The data acquisition module is used to collect image data and process parameters in the circuit board production process in real time, and perform data preprocessing on the image data and process parameters respectively, so as to obtain the corresponding production-related data sets;

[0007] The feature extraction module is used to extract features from production-related data sets, thereby obtaining feature inspection items used to characterize the quality of circuit boards during the production process of the circuit boards, and encrypting all feature inspection items;

[0008] The artificial intelligence evaluation module constructs a circuit board quality evaluation model according to an artificial intelligence algorithm, obtains all feature inspection items after ciphertext transmission, and inputs them into the circuit board quality evaluation model after deciphering the ciphertext to obtain the quality grade of the corresponding circuit board;

[0009] The early warning execution module is used to issue an early warning to the circuit boards whose quality levels do not meet the standards, and notify the relevant operators to screen out the circuit boards whose quality levels do not meet the standards.

[0010] Furthermore, the process of collecting image data and process parameters in the circuit board production process in real time, and preprocessing the image data and process parameters respectively, and then obtaining the corresponding production-related data sets includes:

[0011] Select several equipment deployment points on the circuit board processing production line, deploy sensor modules and cameras at each equipment deployment point, and determine whether the sensor modules and cameras are in normal working state. If so, the sensor module collects process parameters during circuit board processing and production, and the camera collects image data of each circuit board;

[0012] The process parameters include the working environment temperature, the working environment humidity, the imprinting pressure, the vibration frequency, the vibration amplitude and the conveying speed, and the image data includes the surface image of the circuit board corresponding to each production and processing stage of the circuit board;

[0013] If not, debug the sensor module and camera until they are in normal working condition;

[0014] Clean the process parameters in the circuit board production process, remove the incomplete, erroneous and redundant process parameters, binarize the circuit board surface image, and convert the circuit board surface image at each production and processing stage into a corresponding binary image;

[0015] All process parameters after data cleaning in the circuit board production process are integrated as the production-related data set corresponding to the production operation, and the binary images of the circuit board at each production and processing stage in the circuit board production process are integrated as the production-related data set corresponding to the target processing object.

[0016] Furthermore, the process of extracting features from the production-related data set to obtain feature inspection items used to characterize the quality of the circuit board during the production process of the circuit board includes:

[0017] Feature extraction of production-related data sets, including first-category feature extraction and second-category feature extraction, by performing first-category feature extraction on production-related data sets of production operations, thereby obtaining several feature inspection items corresponding to the state of the production process when the circuit board processing production line performs production operations, and by performing second-category feature extraction on the production-related data sets of the target processing objects, thereby obtaining several feature inspection items representing the appearance state of the circuit board;

[0018] Among them, several characteristic inspection items that characterize the status of the production process and several characteristic inspection items that characterize the appearance status of the circuit board are used as characteristic inspection items for characterizing the quality of the circuit board during the production process of the circuit board.

[0019] Furthermore, the contents of the first type of feature extraction and the second type of feature extraction are as follows:

[0020] The first type of feature extraction includes: obtaining all the values ​​of the working environment temperature, working environment humidity, imprinting pressure, vibration frequency, vibration amplitude and transmission speed, and calculating their respective average values ​​as feature inspection items to characterize the corresponding production process status of the circuit board processing production line;

[0021] The content of the second type of feature extraction is: constructing standard reference images corresponding to the circuit board at each production and processing stage during the circuit board production process. The standard reference images are used to record the binary distribution of the circuit board at the corresponding production and processing stage when the circuit board is in a good appearance state. The binary images of the circuit board at each production and processing stage are overlapped and mapped with the standard reference images at the corresponding production and processing stage. Then, the pixel areas in the binary images of the circuit board at each production and processing stage that are different from the standard reference images are used as several feature inspection items to characterize the appearance status of the circuit board.

[0022] Furthermore, the process of ciphertext transmission of all feature inspection items includes:

[0023] Performing a first encryption on a number of characteristic inspection items representing the state of a production process, thereby converting the number of characteristic inspection items representing the state of the production process into first ciphertext data, and constructing a first decryption sequence for decrypting the first ciphertext data;

[0024] Performing a second encryption on a plurality of feature inspection items representing the appearance state, thereby converting the plurality of feature inspection items representing the appearance state into second ciphertext data, and constructing a second decryption sequence for decrypting the second ciphertext data;

[0025] Construct a ciphertext channel, and transmit the first ciphertext data carrying the first decryption sequence and the second ciphertext data carrying the second decryption sequence to the artificial intelligence evaluation module through the ciphertext channel.

[0026] Furthermore, the process of the artificial intelligence evaluation module constructing a circuit board quality evaluation model according to the artificial intelligence algorithm includes:

[0027] Through artificial intelligence algorithms, a convolutional neural network model in an initial state is constructed, and a model input layer and a model output layer are constructed respectively. A certain amount of image data of circuit boards whose appearance quality defects are known to exist are selected, and a certain amount of process parameters whose production process is known to exist during circuit board processing and production are selected, and the data are summarized as the model building data of the convolutional neural network model in the initial state.

[0028] The model building data is divided into a training set and a test set, and the test set is input into the convolutional neural network model in the initial state. The convolutional neural network model extracts feature parameter items corresponding to image data of circuit boards with different appearance quality defects, and extracts feature parameter items corresponding to process parameters with production process abnormalities. It also includes extracting feature parameter items corresponding to image data of circuit boards without appearance quality defects, and extracting feature parameter items corresponding to process parameters without production process abnormalities. The model output layer obtains the quality results corresponding to each circuit board, and the quality results include: the circuit board has quality problems and the circuit board does not have quality problems;

[0029] The quality result of the circuit board output by the model output layer is compared with the actual quality result of the circuit board, and then the model prediction rate of the convolutional neural network model in the initial state is obtained, and the critical threshold of the model prediction rate is set, which is recorded as Yz, and the model prediction rate is recorded as Sc;

[0030] If Sc≥Yz, the final circuit board quality assessment model is constructed;

[0031] If Sc<Yz, the convolutional neural network model in the initial state is trained through the training set, and the real-time model prediction rate is obtained through the test set until Sc≥Yz is satisfied. The model training is stopped and the final circuit board quality assessment model is constructed.

[0032] Furthermore, all feature inspection items after ciphertext transmission are obtained, and after ciphertext decryption, they are input into the circuit board quality assessment model to obtain the quality level of the circuit board. The process includes:

[0033] Obtaining first ciphertext data and second ciphertext data of each of several circuit boards that need to be quality evaluated after ciphertext transmission, setting a data decryption period, and decrypting the first ciphertext data and the second ciphertext data according to the first decryption sequence and the second decryption sequence during the data decryption period, and then restoring the first ciphertext data and the second ciphertext data to the original feature inspection items;

[0034] A data entry period is set. During the data entry period, the feature inspection items of all circuit boards after ciphertext decryption are input into the circuit board quality assessment model to obtain the quality grade of each circuit board. When the assessment shows that the circuit board has appearance quality defects, or the circuit board processing line has production process abnormalities, the corresponding quality grade of the circuit board is unqualified. When the assessment shows that the circuit board does not have appearance quality defects, and the circuit board processing line does not have production process abnormalities, the corresponding quality grade of the circuit board is qualified.

[0035] Furthermore, the process of issuing an early warning to the circuit boards that do not meet the quality standards and notifying the relevant operators to screen out the circuit boards that do not meet the quality standards includes:

[0036] When the circuit board quality assessment model concludes that the quality level of the circuit board is qualified, it means that the quality level of the current circuit board meets the standard and no operation is performed. When the circuit board quality assessment model concludes that the quality level of the circuit board is unqualified, it means that the quality level of the circuit board does not meet the standard, and an early warning is issued for the circuit board with the corresponding unqualified quality level, generating a quality early warning message;

[0037] All quality warning messages are pushed to relevant operators, who will screen out circuit boards that do not meet quality standards, locate the specific abnormal locations of circuit board processing lines with abnormal production processes, and inspect and repair the circuit board processing lines.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows: the image data and process parameters of the circuit board production process are collected in real time by the data acquisition module, and data preprocessing is performed, thereby realizing the automatic collection of production-related data; the feature extraction module extracts features from the production-related data set to obtain feature inspection items that characterize the quality of the circuit board, thereby providing a data basis for subsequent quality assessment of the circuit board; an artificial intelligence algorithm is used to construct a circuit board quality assessment model, thereby realizing an intelligent assessment of the quality of the circuit board, avoiding the subjective influence of manual assessment, and effectively improving the efficiency and accuracy of circuit board quality assessment; and the security of the feature inspection items during the transmission process is ensured by means of encrypted transmission, thereby effectively preventing data leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0040] like Figure 1 As shown, a circuit board production quality assessment system based on artificial intelligence includes a cloud platform, and the cloud platform is communicatively connected with a data acquisition module, a feature extraction module, an artificial intelligence assessment module and an early warning execution module;

[0041] The data acquisition module is used to collect image data and process parameters in the circuit board production process in real time, and perform data preprocessing on the image data and process parameters respectively, so as to obtain the corresponding production-related data sets;

[0042] The feature extraction module is used to extract features from production-related data sets, thereby obtaining feature inspection items used to characterize the quality of circuit boards during the production process of the circuit boards, and encrypting all feature inspection items;

[0043] The artificial intelligence evaluation module constructs a circuit board quality evaluation model according to an artificial intelligence algorithm, obtains all feature inspection items after ciphertext transmission, and inputs them into the circuit board quality evaluation model after deciphering the ciphertext to obtain the quality grade of the corresponding circuit board;

[0044] The early warning execution module is used to issue an early warning to the circuit boards whose quality levels do not meet the standards, and notify the relevant operators to screen out the circuit boards whose quality levels do not meet the standards.

[0045] It should be further explained that, in the specific implementation process, the data acquisition module collects image data and process parameters in the circuit board production process in real time, and performs data preprocessing on the image data and process parameters respectively, and then obtains the corresponding production-related data sets, including:

[0046] Select a number of equipment deployment points at equal intervals on the circuit board processing production line, and number the equipment deployment points, and record the number as i, then i = 1, 2, 3, ..., n, where n is a natural number greater than 0;

[0047] Deploy sensor modules and cameras at each equipment deployment point, and determine whether the sensor modules and cameras are in normal working condition. If so, the sensor modules collect the corresponding process parameters during the production of the circuit boards, and the cameras collect the corresponding image data of each circuit board;

[0048] If not, debug the sensor module and camera until they are in normal working condition;

[0049] The sensor module is composed of a temperature sensor, a humidity sensor, a pressure sensor, a vibration sensor and a speed sensor. The temperature sensor and the humidity sensor are used to collect the temperature and humidity of the working environment when processing the circuit board, respectively. The pressure sensor is used to collect the printing pressure when printing the circuit board. The vibration sensor is used to monitor the vibration frequency and vibration amplitude of the production equipment that processes the circuit board on the circuit board processing production line. The speed sensor is used to collect the transmission speed of the circuit board processing production line.

[0050] The image data of the circuit board collected by the camera in real time are: the surface image of the circuit board corresponding to each production and processing stage of the circuit board;

[0051] The data preprocessing of the process parameters in the circuit board production process is data cleaning, which removes incomplete, erroneous and redundant process parameters. The data preprocessing of the circuit board appearance surface image is binarization, which converts the circuit board appearance surface image at each production and processing stage into a corresponding binary image.

[0052] All process parameters after data cleaning in the circuit board production process are integrated as the production-related data set corresponding to the production operation, and the binary images of the circuit board at each production and processing stage in the circuit board production process are integrated as the production-related data set corresponding to the target processing object.

[0053] It should be further explained that, in a specific implementation process, the feature extraction module extracts features from a production-related data set, and then obtains feature inspection items for characterizing the quality of circuit boards during the production process of the circuit boards, including:

[0054] The feature extraction module performs feature extraction on the production-related data set, including first-class feature extraction and second-class feature extraction. By performing first-class feature extraction on the production-related data set of the production operation, several feature inspection items corresponding to the state of the production process when the circuit board processing production line performs the production operation are obtained. By performing second-class feature extraction on the production-related data set of the target processing object, several feature inspection items representing the appearance state of the circuit board are obtained.

[0055] Among them, several characteristic inspection items characterizing the state of the production process and several characteristic inspection items characterizing the appearance state of the circuit board are used as characteristic inspection items for characterizing the quality of the circuit board during the production process of the circuit board;

[0056] The first type of feature extraction is to obtain all the corresponding values ​​of the working environment temperature, working environment humidity, imprinting pressure, vibration frequency, vibration amplitude and transmission speed, and then calculate their respective average values ​​as feature inspection items to characterize the corresponding production process status of the circuit board processing production line;

[0057] The second type of feature extraction content is: constructing standard reference images corresponding to the circuit board at each production and processing stage during the circuit board production process. The standard reference images are used to record the binary distribution of the circuit board at the corresponding production and processing stage when the circuit board is in a good appearance state. The binary images of the circuit board at each production and processing stage are overlapped and mapped with the standard reference images at the corresponding production and processing stage. Then, the pixel areas in the binary images of the circuit board at each production and processing stage that are different from the standard reference images are used as several feature inspection items to characterize the corresponding appearance state of the circuit board.

[0058] It should be further explained that, in the specific implementation process, the process of ciphertext transmission of all feature inspection items includes:

[0059] Performing a first encryption on a number of characteristic inspection items representing the state of a production process, thereby converting the number of characteristic inspection items representing the state of the production process into first ciphertext data, and constructing a first decryption sequence for decrypting the first ciphertext data;

[0060] Setting the data encryption points of the first encryption, each data encryption point has a corresponding encryption frequency, encryption operation code and encryption strength, and then converting a number of characteristic inspection items representing the state of the production process into the first ciphertext data;

[0061] Locate a number of data encryption points corresponding to a number of characteristic inspection items representing the state of the production process, and set the decryption frequency, decryption resource computing power and decryption operation code corresponding to each data encryption point, thereby constructing a first decryption sequence corresponding to the first ciphertext data;

[0062] Performing a second encryption on a plurality of feature inspection items representing the appearance state, thereby converting the plurality of feature inspection items representing the appearance state into second ciphertext data, and constructing a second decryption sequence for decrypting the second ciphertext data;

[0063] The second encrypted content is as follows:

[0064] Converting several feature inspection items representing the appearance state into a binary string, and dividing the first 1 / 3 of the binary string into the first data to be operated, the middle 1 / 3 into the second data to be operated, and the last 1 / 3 into the third data to be operated;

[0065] Performing sixteen-bit encryption on the first data to be operated to generate a sixteen-bit ciphertext data segment;

[0066] Performing thirty-two-bit encryption on the second data to be operated to generate a thirty-two-bit ciphertext data segment;

[0067] Performing sixty-four-bit encryption on the third data to be operated to generate a sixty-four-bit ciphertext data segment;

[0068] By performing reverse decryption of corresponding digits on the 16-bit ciphertext data segment, the 32-bit ciphertext data segment, and the 64-bit ciphertext data segment, respectively, the corresponding decryption subkeys are obtained, wherein the digits of the decryption subkeys of the 16-bit ciphertext data segment, the 32-bit ciphertext data segment, and the 64-bit ciphertext data segment are 16 digits, 32 digits, and 64 digits, respectively;

[0069] Integrate the 16-bit ciphertext data segment, the 32-bit ciphertext data segment, and the 64-bit ciphertext data segment to generate second ciphertext data, and integrate decryption subkeys of different digits to generate a second decryption sequence;

[0070] Construct a ciphertext channel, and transmit the first ciphertext data carrying the first decryption sequence and the second ciphertext data carrying the second decryption sequence to the data temporary storage area corresponding to the artificial intelligence evaluation module through the ciphertext channel.

[0071] It should be further explained that, in the specific implementation process, the process of the artificial intelligence evaluation module constructing a circuit board quality evaluation model according to the artificial intelligence algorithm includes:

[0072] A convolutional neural network model in an initial state is constructed by using a deep learning algorithm in an artificial intelligence algorithm. A model input layer and a model output layer are constructed for the convolutional neural network model in the initial state. Image data of a certain amount of circuit boards whose appearance quality defects are known to exist are selected, and process parameters whose production process is known to exist with certain amount of data when the circuit boards are processed and produced are selected, and the data are summarized as the model building data of the convolutional neural network model in the initial state.

[0073] Setting a data partition ratio for model building data, and then dividing the model building data into a training set and a test set, inputting the test set into the convolutional neural network model in an initial state, and then extracting feature parameter items corresponding to image data of circuit boards with different appearance quality defects, and extracting feature parameter items corresponding to process parameters with production process abnormalities, and also extracting feature parameter items corresponding to image data of circuit boards without appearance quality defects, and extracting feature parameter items corresponding to process parameters without production process abnormalities, and obtaining the quality results corresponding to each circuit board from the model output layer;

[0074] The quality results include: the circuit board has quality problems and the circuit board does not have quality problems;

[0075] The quality result of the circuit board output by the model output layer is compared with the actual quality result of the circuit board, and then the model prediction rate of the convolutional neural network model in the initial state is obtained, and the critical threshold of the model prediction rate is set, which is recorded as Yz;

[0076] The model prediction rate is denoted as Sc;

[0077] Wherein, Sc=Zq / Zg, Zq is the number of quality results output by the model output layer of the circuit board that are consistent with their actual quality results, and Zg is the total number of quality results output by the model output layer of the circuit board;

[0078] If Sc≥Yz, the final circuit board quality assessment model is constructed;

[0079] If Sc<Yz, the convolutional neural network model in the initial state is trained through the training set, and the real-time model prediction rate is obtained through the test set until Sc≥Yz is satisfied. The model training is stopped and the final circuit board quality assessment model is constructed.

[0080] It should be further explained that, in the specific implementation process, all feature inspection items after ciphertext transmission are obtained, and after ciphertext is deciphered, they are input into the circuit board quality assessment model to obtain the quality level of the corresponding circuit board. The process includes:

[0081] Obtain all feature inspection items corresponding to each of the several circuit boards that need to be evaluated for quality after ciphertext transmission, that is, the first ciphertext data and the second ciphertext data, set a data decryption period, and record the data decryption period as Tpy, then Tpy = [t1, t2], where t1 is the start time of the data decryption period, and t2 is the end time of the data decryption period;

[0082] During the data decryption period, the first ciphertext data and the second ciphertext data are decrypted according to the first decryption sequence and the second decryption sequence, respectively, so as to restore the first ciphertext data and the second ciphertext data to the original feature inspection items;

[0083] Set a data entry period, record the data entry period as Trl, then Trl = [t3, t4], where t3 is the start time of the data entry period, t4 is the end time of the data entry period, and within the data entry period, input the feature inspection items of all circuit boards after ciphertext decryption into the circuit board quality assessment model, and then obtain the quality level of each circuit board;

[0084] If the evaluation shows that the circuit board has appearance quality defects, or the circuit board processing line has production process abnormalities, the corresponding quality grade of the circuit board is unqualified;

[0085] When the assessment concludes that the circuit board has no appearance quality defects and that there are no production process abnormalities in the circuit board processing production line, the corresponding quality grade of the circuit board is qualified.

[0086] It should be further explained that, in the specific implementation process, the early warning execution module issues an early warning to the circuit boards that do not meet the quality standards, and notifies the relevant operators to screen out the circuit boards that do not meet the quality standards, and the process includes:

[0087] When the circuit board quality assessment model concludes that the quality level of the circuit board is qualified, it means that the quality level of the current circuit board meets the standard and no operation is performed. When the circuit board quality assessment model concludes that the quality level of the circuit board is unqualified, it means that the quality level of the circuit board does not meet the standard, and an early warning is issued for the circuit board with the corresponding unqualified quality level, generating a quality early warning message;

[0088] The format of the quality warning message is: circuit board number + warning time + quality defect type + pusher device IP address, where the circuit board number is the unique identity of the circuit board, the warning time is the time corresponding to when the circuit board is found to be substandard in quality, the quality defect type includes abnormal production process of the circuit board processing production line where the current circuit board is located, and appearance quality defects of the circuit board, which are used by relevant operators to screen the circuit boards, and the relevant operators are equipped with mobile device terminals, which are used to receive quality warning messages, and the pusher device IP address is the device IP address of the mobile device terminal held by the relevant operators, and the relevant operators are the pushers of the quality warning messages;

[0089] All quality warning messages are pushed to relevant operators, who then screen out circuit boards that do not meet quality standards, locate the specific abnormal locations of circuit board processing lines where abnormalities exist in the production process, and inspect and repair the circuit board processing lines, thereby achieving efficient control of circuit board quality and ensuring the smooth and efficient operation of the circuit board processing production lines.

[0090] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A circuit board production quality assessment system based on artificial intelligence, including a cloud platform, characterized in that: The cloud platform is communicatively connected with a data acquisition module, a feature extraction module, an artificial intelligence evaluation module and an early warning execution module; The data acquisition module is used to collect image data and process parameters in the circuit board production process in real time, and perform data preprocessing on the image data and process parameters respectively, so as to obtain the corresponding production-related data sets; The feature extraction module is used to extract features from production-related data sets, thereby obtaining feature inspection items used to characterize the quality of circuit boards during the production process of the circuit boards, and encrypting all feature inspection items; The artificial intelligence evaluation module constructs a circuit board quality evaluation model according to an artificial intelligence algorithm, obtains all feature inspection items after ciphertext transmission, and inputs them into the circuit board quality evaluation model after deciphering the ciphertext to obtain the quality grade of the corresponding circuit board; The early warning execution module is used to issue an early warning to the circuit boards whose quality levels do not meet the standards, and notify the relevant operators to screen out the circuit boards whose quality levels do not meet the standards.

2. According to the artificial intelligence-based circuit board production quality assessment system of claim 1, it is characterized in that: The process of real-time acquisition of image data and process parameters in the circuit board production process, and data preprocessing of the image data and process parameters respectively, and then obtaining the corresponding production-related data sets includes: Select several equipment deployment points on the circuit board processing production line, deploy sensor modules and cameras at each equipment deployment point, and determine whether the sensor modules and cameras are in normal working state. If so, the sensor module collects process parameters during circuit board processing and production, and the camera collects image data of each circuit board; The process parameters include the working environment temperature, the working environment humidity, the imprinting pressure, the vibration frequency, the vibration amplitude and the conveying speed, and the image data includes the surface image of the circuit board corresponding to each production and processing stage of the circuit board; If not, debug the sensor module and camera until they are in normal working condition; Clean the process parameters in the circuit board production process, remove the incomplete, erroneous and redundant process parameters, binarize the circuit board surface image, and convert the circuit board surface image at each production and processing stage into a corresponding binary image; All process parameters after data cleaning in the circuit board production process are integrated as the production-related data set corresponding to the production operation, and the binary images of the circuit board at each production and processing stage in the circuit board production process are integrated as the production-related data set corresponding to the target processing object.

3. The circuit board production quality assessment system based on artificial intelligence according to claim 2 is characterized in that: The process of extracting features from production-related data sets and obtaining feature inspection items used to characterize the quality of circuit boards during the production process of circuit boards includes: Feature extraction of production-related data sets, including first-class feature extraction and second-class feature extraction, by performing first-class feature extraction on production-related data sets of production operations, thereby obtaining several feature inspection items corresponding to the state of the production process when the circuit board processing production line performs production operations, and by performing second-class feature extraction on the production-related data sets of the target processing objects, thereby obtaining several feature inspection items representing the appearance state of the circuit board; Among them, several characteristic inspection items that characterize the status of the production process and several characteristic inspection items that characterize the appearance status of the circuit board are used as characteristic inspection items for characterizing the quality of the circuit board during the production process of the circuit board.

4. The circuit board production quality assessment system based on artificial intelligence according to claim 3 is characterized in that: The contents of the first and second feature extractions are as follows: The first type of feature extraction includes: obtaining all the values ​​of the working environment temperature, working environment humidity, imprinting pressure, vibration frequency, vibration amplitude and transmission speed, and calculating their respective average values ​​as feature inspection items to characterize the corresponding production process status of the circuit board processing production line; The content of the second type of feature extraction is: constructing standard reference images corresponding to the circuit board at each production and processing stage during the circuit board production process. The standard reference images are used to record the binary distribution of the circuit board at the corresponding production and processing stage when the circuit board is in a good appearance state. The binary images of the circuit board at each production and processing stage are overlapped and mapped with the standard reference images at the corresponding production and processing stage. Then, the pixel areas in the binary images of the circuit board at each production and processing stage that are different from the standard reference images are used as several feature inspection items to characterize the appearance status of the circuit board.

5. The circuit board production quality assessment system based on artificial intelligence according to claim 4 is characterized in that: The process of ciphertext transmission of all feature inspection items includes: Performing a first encryption on a number of characteristic inspection items representing the state of a production process, thereby converting the number of characteristic inspection items representing the state of the production process into first ciphertext data, and constructing a first decryption sequence for decrypting the first ciphertext data; Performing a second encryption on a plurality of feature inspection items representing the appearance state, thereby converting the plurality of feature inspection items representing the appearance state into second ciphertext data, and constructing a second decryption sequence for decrypting the second ciphertext data; Construct a ciphertext channel, and transmit the first ciphertext data carrying the first decryption sequence and the second ciphertext data carrying the second decryption sequence to the artificial intelligence evaluation module through the ciphertext channel.

6. The circuit board production quality assessment system based on artificial intelligence according to claim 5, characterized in that: The process of building a circuit board quality assessment model based on the artificial intelligence algorithm by the artificial intelligence assessment module includes: Through artificial intelligence algorithms, a convolutional neural network model in an initial state is constructed, and a model input layer and a model output layer are constructed respectively. A certain amount of image data of circuit boards whose appearance quality defects are known to exist are selected, and a certain amount of process parameters whose production process is known to exist during circuit board processing and production are selected, and the data are summarized as the model building data of the convolutional neural network model in the initial state. The model building data is divided into a training set and a test set, and the test set is input into the convolutional neural network model in the initial state. The convolutional neural network model extracts feature parameter items corresponding to image data of circuit boards with different appearance quality defects, and extracts feature parameter items corresponding to process parameters with production process abnormalities. It also includes extracting feature parameter items corresponding to image data of circuit boards without appearance quality defects, and extracting feature parameter items corresponding to process parameters without production process abnormalities. The model output layer obtains the quality results corresponding to each circuit board, and the quality results include: the circuit board has quality problems and the circuit board does not have quality problems; The quality result of the circuit board output by the model output layer is compared with the actual quality result of the circuit board, and then the model prediction rate of the convolutional neural network model in the initial state is obtained, and the critical threshold of the model prediction rate is set, which is recorded as Yz, and the model prediction rate is recorded as Sc; If Sc≥Yz, the final circuit board quality assessment model is constructed; If Sc<Yz, the convolutional neural network model in the initial state is trained through the training set, and the real-time model prediction rate is obtained through the test set until Sc≥Yz is satisfied. The model training is stopped and the final circuit board quality assessment model is constructed.

7. The circuit board production quality assessment system based on artificial intelligence according to claim 6, characterized in that: The process of obtaining all feature inspection items after ciphertext transmission and inputting them into the circuit board quality assessment model after ciphertext decryption to obtain the quality level of the circuit board includes: Obtaining first ciphertext data and second ciphertext data of each of several circuit boards that need to be quality evaluated after ciphertext transmission, setting a data decryption period, and decrypting the first ciphertext data and the second ciphertext data according to the first decryption sequence and the second decryption sequence during the data decryption period, and then restoring the first ciphertext data and the second ciphertext data to the original feature inspection items; A data entry period is set. During the data entry period, the feature inspection items of all circuit boards after ciphertext decryption are input into the circuit board quality assessment model to obtain the quality grade of each circuit board. When the assessment shows that the circuit board has appearance quality defects, or the circuit board processing line has production process abnormalities, the corresponding quality grade of the circuit board is unqualified. When the assessment shows that the circuit board does not have appearance quality defects, and the circuit board processing line does not have production process abnormalities, the corresponding quality grade of the circuit board is qualified.

8. The circuit board production quality assessment system based on artificial intelligence according to claim 7, characterized in that: The process of issuing early warnings to circuit boards that do not meet the quality standards and notifying relevant operators to screen out circuit boards that do not meet the quality standards includes: When the circuit board quality assessment model concludes that the quality level of the circuit board is qualified, it means that the quality level of the current circuit board meets the standard and no operation is performed. When the circuit board quality assessment model concludes that the quality level of the circuit board is unqualified, it means that the quality level of the circuit board does not meet the standard, and an early warning is issued for the circuit board with the corresponding unqualified quality level, generating a quality early warning message; All quality warning messages are pushed to relevant operators, who will screen out circuit boards that do not meet quality standards, locate the specific abnormal locations of circuit board processing lines with abnormal production processes, and inspect and repair the circuit board processing lines.

Images