Multi-mode sensing circuit board pyrolysis parameter adaptive control system and method

Abstract

The invention belongs to the technical field of pyrolysis parameter self-adaptive control, and discloses a multi-mode sensing circuit board pyrolysis parameter self-adaptive control system and a multi-mode sensing circuit board pyrolysis parameter self-adaptive control method. The method comprises the following steps: carrying out feature extraction on spectral signals of R circuit boards to obtain a fluorescence spectrum feature vector set; performing feature extraction on the reflection spectrum data of the R circuit boards to obtain a near-infrared reflection spectrum feature vector set; performing feature extraction on the surface images of the R circuit boards to obtain a circuit board image feature vector set; evaluating the R circuit boards based on the fluorescence spectrum feature vector set, the near-infrared reflection spectrum feature vector set and the circuit board image feature vector set to obtain predicted manufacturing times and aging evaluation grades of the R circuit boards; performing intelligent self-adaptive control on pyrolysis parameters of the circuit board; according to the method, the circuit board pyrolysis parameters can be accurately matched with circuit board decomposition requirements under different manufacturing backgrounds and aging states, and the precision of circuit board pyrolysis parameter regulation and control is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of circuit board pyrolysis parameter adaptive control, more particularly, the present application relates to a multi-modal sensing circuit board pyrolysis parameter adaptive control system and method. BACKGROUND

[0002] With the rapid development of electronic information industry, the number of waste circuit boards is increasing, and pyrolysis as a recycling path can effectively realize circular economy and environmental protection because it can efficiently convert organic matter into combustible gas and simultaneously realize metal component enrichment. In order to improve the energy utilization rate and processing quality of the pyrolysis process, the existing technology usually controls based on preset temperature curve, heating rate, residence time and atmosphere ratio, etc. Process parameters, some schemes adjust the furnace temperature and air volume through infrared temperature measurement or tail gas concentration feedback to realize preliminary process closed-loop control.

[0003] However, in the actual processing process, the types of circuit board raw materials are various, and the metals, resins, fillers and even welding processes used in different manufacturing years have great differences, and the aging characteristics caused by the service life further exacerbate the composition fluctuation, making it difficult for traditional uniform process parameters to consider all types of raw materials, resulting in problems such as inaccurate parameter adaptation, incomplete reaction, and high energy consumption. In addition, most of the existing technologies rely on operators to set parameters based on experience, lack of intelligent identification and parameter adaptation capability for the actual properties of raw materials, and it is difficult to realize the precision and real-time of circuit board parameter regulation. Especially in the process of mixing batches or switching raw materials, the incorrect setting often causes coking, unstable gas composition or dioxin emission exceeding the standard, etc. Process risks restrict the popularization and application of pyrolysis processing devices.

[0004] Therefore, there is an urgent need for a new method of circuit board pyrolysis control that integrates raw material attribute sensing and intelligent parameter setting to address the resource utilization challenges of highly heterogeneous circuit board raw materials. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned purposes, the present application provides the following technical scheme: a multi-modal sensing circuit board pyrolysis parameter adaptive control method, comprising:

[0006] analyzing and extracting features of the R-block circuit board spectral signal obtained by the fluorescence spectrum acquisition region to obtain a set of fluorescence spectrum feature vectors;

[0007] extracting features and structuring the R-block circuit board reflectance spectrum data collected by the near-infrared reflectance spectrum acquisition region to obtain a set of near-infrared reflectance spectrum feature vectors;

[0008] extracting features from the R-block circuit board surface image obtained by the image acquisition device to obtain a set of circuit board image feature vectors;

[0009] Based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set and the circuit board image feature vector set, the R-block circuit board is evaluated to obtain the predicted manufacturing year and the aging evaluation grade of the R-block circuit board;

[0010] Based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, the circuit board image feature vector set, the predicted manufacturing year and the aging evaluation grade of the R-block circuit board, a pyrolysis parameter adjustment instruction is generated, and the pyrolysis parameter of the circuit board is intelligently and adaptively controlled.

[0011] Further, the method for intelligently and adaptively controlling the pyrolysis parameter of the circuit board comprises:

[0012] Based on the predicted manufacturing year and the aging evaluation grade of the R-block circuit board, a predicted manufacturing year binary tuple set and an aging evaluation grade binary tuple set are constructed;

[0013] Based on the predicted manufacturing year binary tuple set and the aging evaluation grade binary tuple set, a dominant component is determined and a dominant label value is constructed;

[0014] The dominant label value, the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set and the circuit board image feature vector set are input into a pyrolysis parameter setting model to obtain a corresponding pyrolysis parameter combination, wherein the pyrolysis parameter combination comprises a target pyrolysis temperature, a heating rate, a pyrolysis residence time, an oxygen and inert gas ratio and a pyrolysis reaction chamber initial pressure;

[0015] The system control platform issues the pyrolysis parameter combination to the corresponding control device and executes it.

[0016] Further, the method for obtaining the predicted manufacturing year binary tuple set and the aging evaluation grade binary tuple set comprises:

[0017] S400: Let the initial value of r be 1, and the value range of r be 1 to R;

[0018] S401: Obtain the predicted manufacturing year and the aging evaluation grade of the rth circuit board;

[0019] Determine whether the predicted manufacturing year exists in the predicted manufacturing year binary tuple set; if it exists, add 1 to the circuit board quantity count value of the predicted manufacturing year binary tuple of the predicted manufacturing year; if it does not exist, construct the predicted manufacturing year and 1 into a predicted manufacturing year binary tuple, and add it to the predicted manufacturing year binary tuple set; an example of the predicted manufacturing year binary tuple is (the predicted manufacturing year of the rth circuit board, the circuit board quantity count value);

[0020] determining whether the aging evaluation grade exists in the aging evaluation grade tuple set; if the aging evaluation grade exists, adding 1 to the board quantity count value of the aging evaluation grade tuple of the aging evaluation grade; if the aging evaluation grade does not exist, constructing the aging evaluation grade and 1 of the rth board into an aging evaluation grade tuple, and adding the aging evaluation grade tuple to the aging evaluation grade tuple set; an example of the aging evaluation grade tuple is (aging evaluation grade of the rth board, board quantity count value);

[0021] S402: setting r = r + 1, if r is less than or equal to R, returning to S401 to continue execution; if r is greater than R, ending the current process.

[0022] Further, the method for determining the dominant component based on the predicted manufacturing year tuple set and the aging evaluation grade tuple set and constructing the dominant label value comprises:

[0023] initializing the dominant identification of the predicted manufacturing year and the dominant identification of the aging evaluation grade as no;

[0024] respectively performing the modulo operation of the board quantity count value of each predicted manufacturing year tuple in the predicted manufacturing year tuple set by R to obtain the corresponding board quantity proportion, and constructing the board quantity proportions into a first quantity proportion set;

[0025] respectively performing the modulo operation of the board quantity count value of each aging evaluation grade tuple in the aging evaluation grade tuple set by R to obtain the corresponding board quantity proportion, and constructing the board quantity proportions into a second quantity proportion set;

[0026] selecting the maximum first quantity proportion from the first quantity proportion set, determining whether the first quantity proportion is greater than a preset first quantity proportion threshold, if the determination result is yes, setting the dominant identification of the predicted manufacturing year as yes;

[0027] selecting the maximum second quantity proportion from the second quantity proportion set, determining whether the second quantity proportion is greater than a preset second quantity proportion threshold, if the determination result is yes, setting the dominant identification of the aging evaluation grade as yes;

[0028] matching the dominant identification of the predicted manufacturing year and the dominant identification of the aging evaluation grade with the pre-constructed dominant label value mapping table to obtain the dominant label value.

[0029] Further, the method for obtaining the predicted manufacturing year and the aging evaluation grade of the Rth board comprises:

[0030] S300: setting the initial value of r as 1, and the value range of r as 1 to R;

[0031] S301: Obtain the fluorescence spectrum feature vector of the rth block of circuit board from the set of fluorescence spectrum feature vectors, obtain the near-infrared reflectance spectrum feature vector of the rth block of circuit board from the set of near-infrared reflectance spectrum feature vectors, and obtain the circuit board image feature vector of the rth block of circuit board from the set of circuit board image feature vectors;

[0032] S302: Input the fluorescence spectrum feature vector, the near-infrared reflectance spectrum feature vector and the circuit board image feature vector of the rth block of circuit board into the comprehensive evaluation model to obtain the predicted manufacturing year and the aging evaluation grade of the rth block of circuit board;

[0033] S303: Let r=r+1, if r is less than or equal to R, return to S301 to continue execution, if r is greater than R, end the current process.

[0034] Further, the method for obtaining the set of fluorescence spectrum feature vectors comprises:

[0035] The X-ray fluorescence spectrum acquisition device emits a micro-focus X-ray beam with a focal spot size less than 100 microns to the surface of the R block of circuit board; under X-ray bombardment, the inner electrons of the surface layer elements of the circuit board jump and instantaneously release X-ray fluorescence with characteristic energy;

[0036] After entering the spectrometer detector, each fluorescence photon of the X-ray fluorescence forms an independent incident photon event, and the spectrometer detector collects and converts the electronic signal released by the incident photon event to generate a pulse voltage signal with a linear relationship with the fluorescence energy;

[0037] After preamplification, shaping and analog-to-digital conversion, the pulse voltage signal enters the multichannel analyzer for energy channel classification and statistical accumulation, and the spectral analysis software automatically extracts the peak energy, peak height and relative ratio of each element in the R block of circuit board;

[0038] The peak energy, peak intensity and relative ratio of the R block of circuit board are normalized and vectorized, and R sets of fluorescence spectrum feature vectors are constructed to form a set of fluorescence spectrum feature vectors.

[0039] Further, the method for obtaining the set of near-infrared reflectance spectrum feature vectors comprises:

[0040] The near-infrared reflectance spectrometer respectively emits near-infrared light with a wavelength range of 900 to 1700 nanometers to the surface of the R block of circuit board, and the near-infrared detector receives the near-infrared light signal reflected by the circuit board surface and samples according to the number Q of wavelength points, that is, the reflectivity of each block of circuit board is collected at Q wavelength points to form a Q-dimensional reflectivity vector;

[0041] The principal component analysis is performed on the Q-dimensional reflectivity vector corresponding to each circuit board to obtain a near-infrared reflectance spectrum characteristic vector corresponding to R circuit boards; and the near-infrared reflectance spectrum characteristic vectors corresponding to the R circuit boards are constructed into a near-infrared reflectance spectrum characteristic vector set.

[0042] Further, the method for obtaining the near-infrared reflectance spectrum characteristic vector corresponding to the R circuit boards comprises:

[0043] The Q-dimensional reflectivity vector corresponding to each circuit board is constructed into a reflectivity matrix;

[0044] The reflectivity data in each column of the reflectivity matrix is subtracted by the reflectivity mean value of the corresponding column to obtain a zero mean matrix, and a covariance matrix is calculated based on the zero mean matrix, which is used to measure the linear correlation between the reflectivity of each wave band;

[0045] Eigenvalues are calculated from the covariance matrix to obtain W eigenvalues, and the corresponding eigenvectors are calculated for each eigenvalue, and the W eigenvalues sorted in descending order are constructed into an eigenvalue set, and the eigenvectors corresponding to the eigenvalue set are constructed into an eigenvector set;

[0046] The number of principal components ZCF is determined according to the eigenvalue set and a preset cumulative contribution rate threshold;

[0047] The first ZCF eigenvectors are selected from the eigenvector set to construct a principal component vector set;

[0048] The Q-dimensional reflectivity vector of the R circuit boards is subjected to linear projection operation with the principal component vector set to obtain a principal component projection value sequence corresponding to the R circuit boards, and the principal component projection value sequence is the near-infrared reflectance spectrum characteristic vector of the circuit board.

[0049] Further, the method for determining the number of principal components ZCF according to the eigenvalue set and a preset cumulative contribution rate threshold comprises:

[0050] Step one: sum the eigenvalues in the eigenvalue set to obtain an eigenvalue sum; and let the initial value of the number of principal components ZCF be 1;

[0051] Step two: sum the first ZCF eigenvalues selected from the eigenvalue set to obtain a cumulative eigenvalue, and divide the cumulative eigenvalue by the eigenvalue sum to obtain a cumulative contribution rate;

[0052] Step three: if the cumulative contribution rate is greater than the cumulative contribution rate threshold, the number of principal components ZCF is obtained, and the current process is ended; if the cumulative contribution rate is less than or equal to the cumulative contribution rate threshold, ZCF=ZCF+1, and step two is returned to continue execution.

[0053] Further, the method for obtaining the principal component projection value sequence corresponding to the R-block circuit board comprises:

[0054] S100: construct the principal component vector set into a principal component matrix; the principal component matrix has Q rows and ZCF columns; each column in the principal component matrix corresponds to a principal component vector; let the initial value of r be 1, and the value range of r be 1 to R;

[0055] S101: obtain the r-th row matrix element from the zero-mean matrix, denoted as the centralized reflectivity vector;

[0056] S102: multiply the centralized reflectivity vector by the principal component matrix to obtain the principal component projection value sequence;

[0057] S103: let r = r + 1, if r is less than or equal to R, return to S101 for continuous execution, if r is greater than R, end the current process.

[0058] Further, the method for obtaining the principal component projection value sequence corresponding to the R-block circuit board comprises:

[0059] S200: let the initial value of r be 1, and the value range of r be 1 to R;

[0060] S201: pre-process the surface image of the r-th block circuit board to obtain the pre-processed surface image of the r-th block circuit board; the pre-processing includes brightness normalization, contrast enhancement, edge smoothing and artifact elimination;

[0061] S202: divide the pre-processed surface image of the r-th block circuit board into HxH equal-area image grids, extract the corresponding number of circuit board solder joints in each image grid, and calculate the solder joint distribution uniformity factor of the r-th block circuit board; mark the image grid corresponding to the number of circuit board solder joints greater than the preset number of circuit board solder joints threshold as a soldering area, and count the number of soldering areas as the number of soldering areas; extract the solder joint area corresponding to each image grid and the image grid area, and calculate the average solder joint density of the r-th block circuit board;

[0062] S203: construct the solder joint distribution uniformity factor, the number of soldering areas and the average solder joint density of the r-th block circuit board into the circuit board image feature vector of the r-th block circuit board, and add it to the circuit board image feature vector set;

[0063] S204: let r = r + 1, if r is less than or equal to R, return to S201 for continuous execution, if r is greater than R, end the current process.

[0064] A multi-modal perception circuit board pyrolysis parameter adaptive control system for implementing the multi-modal perception circuit board pyrolysis parameter adaptive control method, comprising:

[0065] The first processing module is configured to analyze and extract features of the R-block circuit board spectrum signal obtained by the fluorescence spectrum acquisition region, and obtain a fluorescence spectrum feature vector set;

[0066] The second processing module is configured to extract features and structure the R-block circuit board reflectance spectrum data collected by the near-infrared reflectance spectrum acquisition region, and obtain a near-infrared reflectance spectrum feature vector set;

[0067] The third processing module is configured to extract features of the R-block circuit board surface image obtained by the image acquisition device, and obtain a circuit board image feature vector set;

[0068] The comprehensive evaluation module is configured to evaluate the R-block circuit board based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set and the circuit board image feature vector set, and obtain a predicted manufacturing year and an aging evaluation grade of the R-block circuit board.

[0069] The parameter control module is configured to generate a pyrolysis parameter adjustment instruction based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, the circuit board image feature vector set, the predicted manufacturing year and the aging evaluation grade of the R-block circuit board, and intelligently and adaptively control the circuit board pyrolysis parameters.

[0070] Compared with the prior art, the technical effects and advantages of the multi-modal perception-based circuit board pyrolysis parameter adaptive control system and method are as follows:

[0071] The multi-modal perception-based circuit board pyrolysis parameter adaptive control system and method provided by the application fuses multi-source heterogeneous perception information such as fluorescence spectrum features, near-infrared reflectance spectrum features and image features, combines the predicted manufacturing year and the aging evaluation grade of the circuit board, constructs a dominant label value, and drives a pyrolysis parameter setting model based on the dominant label value to intelligently generate a pyrolysis parameter combination including a target pyrolysis temperature, a heating rate, a pyrolysis residence time, a ratio of oxygen and inert gas and an initial pressure of a pyrolysis reaction chamber.

[0072] Further, the application can dynamically adjust the multi-physical quantity control components of the pyrolysis device, so that the circuit board pyrolysis process accurately matches the decomposition requirements of circuit boards under different manufacturing backgrounds and aging states, significantly improves the reaction efficiency and control stability of the pyrolysis process. Compared with the uniform process parameters or manual experience setting method in the prior art, the application realizes an intelligent adaptive parameter setting mechanism based on data driving, effectively solves the problems of inaccurate parameter adaptation, high energy consumption and incomplete reaction in the diversified circuit board processing process, effectively improves the accuracy and real-time performance of the circuit board parameter regulation, and has strong technical advancement and industrial practical value. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 A multi-modal sensing circuit board pyrolysis parameter adaptive control system schematic diagram for embodiment 1 of the present application;

[0074] Figure 2 A multi-modal sensing circuit board pyrolysis parameter adaptive control method flow chart for embodiment 2 of the present application;

[0075] Figure 3 A R-block circuit board feature extraction method flow chart;

[0076] Figure 4 A dominant label value acquisition method flow chart;

[0077] Figure 5 A data-driven input and control linkage schematic diagram for circuit board pyrolysis parameters. DETAILED DESCRIPTION

[0078] The technical solutions in the embodiments of the present application will be described in detail, clearly and completely below with reference to the drawings in the embodiments of the present application. It should be particularly noted that the specific embodiments described below are only used to better illustrate and describe the technical solutions of the present application, and are intended to enable those skilled in the art to better understand and implement the present application, and should not be understood as limiting the scope of protection of the present application. Those skilled in the art can modify, adjust or equivalently replace the present application according to the content disclosed in the present application without departing from the spirit and essence of the present application, and these should be regarded as the protection scope of the present application.

[0079] Embodiment 1:

[0080] Please refer to Figure 1 As shown in the figure, the embodiment discloses a multi-modal sensing circuit board pyrolysis parameter adaptive control system, which comprises a first processing module, a second processing module, a third processing module, a comprehensive evaluation module and a parameter control module, each module is connected through wired and / or wireless connection to realize data transmission.

[0081] The first processing module is used for analyzing and extracting features of the R-block circuit board spectral signal obtained by the fluorescence spectrum acquisition region, to obtain a fluorescence spectrum feature vector set. R is the total number of circuit boards processed in the current batch.

[0082] The acquisition method of the fluorescence spectrum feature vector set comprises:

[0083] The X-ray fluorescence spectrum acquisition device emits a micro-focus X-ray beam with a focal spot size less than 100 microns to the surface of the R-block circuit board, to ensure that the excitation region is accurately positioned and the energy density is sufficient; under X-ray bombardment, the inner electrons of the surface layer elements of the circuit board jump and instantaneously release X-ray fluorescence with characteristic energy;

[0084] After entering the spectrometer detector, which is preferably a silicon drift detector (SDD) or a PIN diode detector, each fluorescent photon of the X-ray fluorescence forms an independent incident photon event. The spectrometer detector collects and converts the electronic signal released by the incident photon event to generate a pulse voltage signal with a linear relationship with the fluorescence energy;

[0085] After preamplification, shaping and analog-to-digital conversion, the pulse voltage signal enters the multichannel analyzer for energy channel classification and statistical accumulation. The spectral analysis software automatically extracts the peak energy (spectral peak center position), peak height (peak intensity) and relative ratio of each element in the R-block circuit board.

[0086] The peak energy, peak intensity and relative ratio of the R-block circuit board are normalized and vectorized, and an R-group fluorescent spectrum feature vector is constructed. The R-group fluorescent spectrum feature vector is constructed into a fluorescent spectrum feature vector set. The fluorescent spectrum feature vector set is used to comprehensively represent the metal components and their abundance of the circuit board, and provides high-credibility spectral input for manufacturing year determination, alloy type identification and metal load estimation.

[0087] It should be noted that the X-ray beam refers to the spatially concentrated radiation energy flow formed by continuous generation from the X-ray emitting tube and limited in direction by the collimator, with clear radiation directionality and energy density distribution. The micro-focus refers to the focal point diameter not greater than 100 microns formed by the electron beam bombarding the anode target inside the X-ray source, thereby forming an X-ray irradiation area with extremely small focal spot size on the sample surface, to realize micro-area excitation and high-resolution spectral analysis. By irradiating the target circuit board surface with a micro-focus X-ray beam, the characteristic fluorescence of specific material elements can be excited within a limited range, improving spectral collection accuracy and spatial resolution capability.

[0088] The X-ray fluorescence signal generated by exciting the surface of the circuit board is analyzed and the characteristic is extracted, and the obtained fluorescence spectrum characteristic vector can represent the types, contents and relative composition ratios of the metal elements contained in the circuit board, which has the following significant technical effects and comprehensive application value: first, the spectral peak energy in the fluorescence spectrum characteristic vector can accurately identify various metal elements in the surface layer of the circuit board, such as Sn (tin), PbPb (lead), Cu (copper) and Br (bromine), thereby realizing the consistency discrimination of manufacturing years and process systems (such as whether it is lead-free solder); second, the spectral peak intensity and relative ratio can reflect the distribution abundance of each metal element in the circuit board, which can be used to estimate the total alloy load level of the whole board and provide a basis for subsequent pyrolysis parameter control; third, the fluorescence spectrum characteristic vector as a highly structured and quantifiable material representation vector has the advantages of non-contact, rapidness, high resolution and the like, and can realize online classification and difference identification of mixed batch circuit board raw materials without interrupting the production rhythm, thereby improving the intelligent level, self-adaptive ability and resource utilization rate of the whole processing process line. Compared with the traditional raw material classification method relying on vision or manual identification, the fluorescence spectrum characteristic vector introduced in the present application has stronger anti-pollution interference ability and higher classification precision, and is particularly suitable for efficient intelligent identification of electronic waste circuit boards with complex structure and significant composition difference.

[0089] It should be noted that the peak energy, peak intensity and relative ratio corresponding to the elements are shown in Table 1 Element Data Table:

[0090] Table 1 Element Data Table

[0091]

[0092] The second processing module is configured to perform feature extraction and structural processing on the R-block circuit board reflectance spectrum data collected by the near-infrared reflectance spectrum acquisition region, and obtain a set of near-infrared reflectance spectrum characteristic vectors. The near-infrared reflectance spectrum characteristic vector is used to represent the material composition, aging state and organic residue of the circuit board.

[0093] The method for obtaining the set of near-infrared reflectance spectrum characteristic vectors comprises:

[0094] The near-infrared reflectance spectrometer emits near-infrared light with a wavelength range of 900 to 1700 nanometers to the surface of the R-block circuit board, and the near-infrared detector (preferably an InGaAs linear array detector) receives the reflected near-infrared light signal from the circuit board surface and samples according to the number Q of wavelength points (for example, sampling every 5 nanometers between 900 to 1700 nanometers, so the number of wavelength points is 161), that is, the reflectivity of each circuit board is collected at Q wavelength points to form a Q-dimensional reflectivity vector.

[0095] The Q-dimensional reflectivity vector corresponding to each circuit board is subjected to principal component analysis to obtain a near-infrared reflectance spectrum characteristic vector corresponding to the R circuit boards.

[0096] The method for obtaining the near-infrared reflectance spectrum characteristic vector corresponding to the R circuit boards comprises:

[0097] The Q-dimensional reflectivity vectors corresponding to the R circuit boards are constructed into a reflectivity matrix, which has R rows and Q columns. Each row of the reflectivity matrix corresponds to the reflectivity value of a circuit board at each wavelength point. The reflectivity data in each column of the reflectivity matrix is subtracted by the reflectivity mean value of the column to obtain a zero-mean matrix. A covariance matrix is calculated based on the zero-mean matrix, which is used to measure the linear correlation between the reflectivity of each wavelength band.

[0098] Eigenvalues are calculated from the covariance matrix to obtain W eigenvalues. The corresponding eigenvectors are calculated for each eigenvalue. The W eigenvalues sorted in descending order are constructed into an eigenvalue set, and the eigenvectors corresponding to the eigenvalue set are constructed into an eigenvector set.

[0099] The number of principal components ZCF is determined based on the eigenvalue set and a preset cumulative contribution rate threshold.

[0100] The first ZCF eigenvectors are selected from the eigenvector set to construct a principal component eigenvector set.

[0101] The Q-dimensional reflectivity vectors of the R circuit boards are subjected to linear projection operation with the principal component eigenvector set to obtain a principal component projection value sequence corresponding to the R circuit boards, which is the near-infrared reflectance spectrum characteristic vector of the circuit board.

[0102] The method for determining the number of principal components ZCF based on the eigenvalue set and a preset cumulative contribution rate threshold comprises:

[0103] Step 1: The eigenvalues in the eigenvalue set are summed to obtain an eigenvalue sum. The initial value of the number of principal components ZCF is set to 1.

[0104] Step 2: The first ZCF eigenvalues in the eigenvalue set are summed to obtain a cumulative eigenvalue. The cumulative eigenvalue is divided by the eigenvalue sum to obtain a cumulative contribution rate.

[0105] Step 3: If the cumulative contribution rate is greater than the cumulative contribution rate threshold, the number of principal components ZCF is obtained, and the current process is ended. If the cumulative contribution rate is less than or equal to the cumulative contribution rate threshold, ZCF is set to ZCF+1, and step 2 is returned to continue execution. In this application, the cumulative contribution rate threshold can be set to 0.95, for example.

[0106] The method for obtaining the principal component projection value sequence corresponding to the R circuit boards includes:

[0107] S100: constructing a principal component matrix from a set of principal component vectors; the principal component matrix has Q rows and ZCF columns; each column in the principal component matrix corresponds to a principal component vector; setting the initial value of r to 1, and the value range of r to 1 to R;

[0108] S101: Obtain the r-th row matrix element from the zero-mean matrix, and record it as the centralized reflectivity vector;

[0109] S102: Multiplying the central reflectivity vector by the principal component matrix to obtain a principal component projection value sequence;

[0110] The calculation of the principal component projection value sequence is as follows:

[0111] ;

[0112] in, is the principal component projection value sequence corresponding to the rth circuit board, is the matrix element in the rth row of the zero-mean matrix, that is, the centralized reflectivity vector corresponding to the rth circuit board; is the principal component matrix; is the first column element in the principal component matrix, The ZCF column element in the principal component matrix.

[0113] S103: Let r = r + 1. If r is less than or equal to R, return to S101 to continue execution. If r is greater than R, end the current process.

[0114] It should be noted that the surface of a circuit board is usually composed of an epoxy resin matrix, a glass fiber layer, a tin-based alloy welding area, and residual flux or thermal adhesive during use. Each component has specific absorption characteristics for light of different wavelengths in the near-infrared band. By standardizing the near-infrared reflectivity data of multiple circuit boards in the 900 to 1700 nanometer band, performing principal component analysis and principal component projection processing, the set of near-infrared reflectance spectrum feature vectors obtained constitutes an important component of the multimodal raw material fingerprint recognition system in this application, which plays a core role in multiple links such as raw material state identification, material difference quantification, and process adaptive control.

[0115] Specifically, the set of near-infrared reflectance spectrum feature vectors can accurately characterize the optical absorption behavior of the circuit board surface material in different bands, reflecting the physical aging degree of the raw materials during service, the molecular degradation level of the organic resin, the oxidation coverage of the metal surface, and the enrichment of organic residues such as flux or thermal grease. It has the ability to express the service status of the material with high sensitivity.

[0116] The near-infrared reflectance spectrum feature vector set is compressed by principal component analysis, and the spectral principal component factors that can best explain the differences between samples are retained, forming a standard vector form that can be input into a machine learning model, significantly improving the input efficiency and stability of subsequent manufacturing year discrimination, aging grade classification, and cracking process adaptability judgment, and avoiding model overfitting, computational redundancy, and feature coupling problems caused by high-dimensional original data.

[0117] In the present application, the near-infrared reflectance spectrum feature vector set is obtained by principal component analysis and projection processing of the reflectivity vectors of multiple circuit boards, and each principal component variable in it represents the absorption structure change characteristics of different wave bands, with clear material response meaning. The following are some representative features related to the aging degree and manufacturing year of circuit boards:

[0118] Aging resin absorption characteristic factor, this principal component mainly responds to the 1350-1449 nm wave band, which belongs to the typical absorption band of C-H, C=O, etc. functional groups (functional groups refer to the atomic group structure of organic molecules that exhibit specific properties in chemical reactions) in the epoxy resin matrix in the near-infrared region. The reflectivity of circuit boards with long service time or severe thermal aging in this wave band will decrease significantly, forming a deep absorption valley. This characteristic factor can be used to characterize the thermal oxidative aging degree of the resin matrix.

[0119] Moisture content and moisture absorption residual response factor, this principal component is mainly loaded on the 1450-1550 nm wave band, reflecting the absorption behavior of -OH groups and physically adsorbed water on the surface of the circuit board to infrared light. The absorption intensity of this wave band is higher for raw boards manufactured early and exposed to a humid environment for a long time. This feature helps to evaluate the environmental exposure history of the board during service.

[0120] Soldering flux or silicone grease contamination identification factor, this principal component is concentrated in the 1600-1700 nm region, which typically corresponds to the composite absorption structure of organic silicon and fatty acid additives. Some early process circuit boards do not use low-residue soldering technology, and their reflectivity characteristics in this wave band show a clear multi-peak overlap phenomenon. Therefore, this feature can assist in judging the manufacturing process generation characteristics and supporting the manufacturing year identification.

[0121] Oxidized solder layer surface characteristic factor, this principal component responds to the 1050-1250 nm interval, reflecting the infrared absorption shoulder peak characteristics of metal surface oxide films (such as SnO2, CuO). The reflectivity of the wave band changes significantly for long-service-time circuit boards with oxidized solder layers, and this principal component variable can be used to infer the solder state and potential service life.

[0122] It should be noted that the representative feature in the present application corresponds to the wavelength and aging relationship between the characteristics, as shown in Table 2 wavelength-aging characteristic data table:

[0123] Table 2 wavelength-aging characteristic data table

[0124]

[0125] The third processing module is used for feature extraction processing of the R block circuit board surface image obtained by the image acquisition device, and a circuit board image feature vector set is obtained. The circuit board image feature vector is used to reflect the appearance structure state, welding process distribution and aging signs of the circuit board.

[0126] For example, the image acquisition device in the present application is preferably an industrial vision camera, which adopts a linear array or a surface array CMOS or CCD image sensor, cooperates with a standard light source device (such as a diffuse LED area light source or a bar-shaped backlight source) to perform image acquisition, and can obtain surface image data of the target circuit board in the visible light range (400 to 700 nanometers). Each circuit board image should cover the complete area, and the image format preferably uses an 8-bit grayscale image or an RGB three-channel image.

[0127] The method for obtaining the circuit board image feature vector set comprises:

[0128] S200: Let the initial value of r be 1, and the value range of r be 1 to R;

[0129] S201: Preprocessing the surface image of the rth circuit board to obtain the preprocessed surface image of the rth circuit board; the preprocessing includes brightness normalization, contrast enhancement, edge smoothing and artifact elimination;

[0130] It should be noted that the brightness normalization usually adopts a histogram equalization method to redistribute the brightness value of the image pixels, so that the overall gray level of the image tends to be uniform within the output range, thereby eliminating the local dark or overexposure phenomenon caused by uneven light intensity; the contrast enhancement adopts an adaptive Gamma correction method, which dynamically adjusts the Gamma value according to the brightness distribution of the image, so that the low-contrast area obtains a higher dynamic range, thereby enhancing the image details; in terms of edge smoothing processing, Gaussian filtering or bilateral filtering algorithm is usually used, wherein the Gaussian filtering smoothes the image by constructing a normal distribution convolution kernel, which is suitable for overall noise suppression; the bilateral filtering realizes smoothing processing while retaining the edge structure, and is especially suitable for fine texture feature retention scenarios; the image artifact elimination processing usually combines a morphological reconstruction method, which effectively eliminates the non-structural artifacts in the image caused by background interference, dust residue or overexposure, and maintains the integrity and continuity of the real structure boundary in the image.

[0131] The image preprocessing means of brightness normalization, contrast enhancement, edge smoothing and artifact elimination are introduced in the image feature extraction process, aiming to improve the image input quality and the accuracy of subsequent feature recognition, which is a reasonable optimization selection for realizing the feasibility and adaptability of the technical scheme of the application, and does not constitute the core innovation point of the technical scheme of the application.

[0132] S202: The preprocessed surface image of the rth circuit board is divided into HxH equal-area image grids (such as 4x4), the number of corresponding solder joints of each image grid is extracted, and the solder joint distribution uniformity factor of the rth circuit board is calculated; the image grid corresponding to the number of solder joints greater than the preset number of solder joints threshold of the circuit board is marked as a soldering area, and the number of soldering areas is counted as the number of soldering areas; the solder joint area corresponding to each image grid and the image grid area are extracted, and the average solder joint density of the rth circuit board is calculated; in this application, the number of solder joints threshold of the circuit board can be set to 10, for example.

[0133] S203: The solder joint distribution uniformity factor, the number of soldering areas and the average solder joint density of the rth circuit board are constructed into the circuit board image feature vector of the rth circuit board, and are added to the circuit board image feature vector set;

[0134] S204: Let r=r+1, if r is less than or equal to R, return to S201 to continue execution, if r is greater than R, end the current process.

[0135] The calculation method of the solder joint distribution uniformity factor includes:

[0136] ;

[0137] Wherein, is the solder joint distribution uniformity factor, represents the number of corresponding solder joints of the hth image grid in the HxH equal-area image grid, represents the mean of the number of solder joints, represents the standard deviation of the number of solder joints, and the quotient of the standard deviation of the number of solder joints and the mean of the number of solder joints is the solder joint distribution uniformity factor of the circuit board. When the solder joints are uniformly distributed in each area, the standard deviation of the number of solder joints is small, and the solder joint distribution uniformity factor tends to 0; if there is a concentrated dense or large area of blank area, the standard deviation increases, and the solder joint distribution uniformity factor tends to 1. The solder joint distribution uniformity factor is used to measure whether the layout of the circuit board solder joints is concentrated and dense or sparse and has no solder joints, so as to reflect the process complexity and alloy distribution strategy.

[0138] The calculation method of the average solder joint density of the rth circuit board includes:

[0139] ;

[0140] wherein, is the average solder density of the r-th board, h is an index variable in the summation formula, is the solder area of the h-th image grid corresponding to the HxH equal-area image grid, is the image grid area of the h-th image grid in the HxH equal-area image grid.

[0141] It should be noted that in the embodiments of the present application, the constructed set of board image feature vectors realizes indirect modeling and quantifiable evaluation of the manufacturing year and the service aging state of the board by extracting observable physical features of the board solder structure at the image level. The board image feature vectors include the solder distribution uniformity factor, the solder area number and the average solder density, have statistical behavior rules highly related to the process iteration level, the manufacturing batch characteristics and the long-term service state, and can be used as auxiliary basis for manufacturing year and aging degree identification.

[0142] Specifically, boards of different manufacturing years usually adopt different generations of soldering process standards and layout templates. For example, early boards often have features such as uneven solder distribution, local over-soldering or different solder pad sizes due to a higher proportion of manual or semi-automatic soldering. However, new generation boards mostly use precision patching and full-automatic soldering, and the solder distribution tends to be more regular and the arrangement density is higher. Therefore, by analyzing the solder distribution uniformity factor and the solder area number, the possible manufacturing process year of the board can be identified in a passive state.

[0143] On the other hand, after a long time of operation, the solder shape of the board may also be slightly degraded due to factors such as thermal expansion and contraction, current thermal stress and environmental pollution. For example, phenomena such as blurring of solder reflow shape, corrosion distortion of solder pad boundary or unstable growth of solder area, etc. These degradation phenomena can be manifested as solder size abnormalities, edge blurring or solder area number fluctuations in image solder density statistics and region connectivity, thereby providing quantitative clues at the image level for aging state identification.

[0144] In summary, by constructing a unified image feature data set from multiple board image feature parameters and combining it with spectral data and working condition parameters for fusion modeling, not only the recognition ability of the system for manufacturing year and aging state is enhanced, but also key visual structure input is provided for subsequent smelting and pyrolysis parameter adaptive adjustment model, improving the adaptability and intelligent level of the system.

[0145] It should be noted that the R-block board feature extraction method flowchart is as shown in Figure 3 Figure 3 ​The acquisition method of the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set and the circuit board image feature vector set, i.e., the final construction of the feature data set.

[0146] The comprehensive evaluation module evaluates the R-block circuit board based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set and the circuit board image feature vector set, and obtains the predicted manufacturing year and the aging evaluation grade of the R-block circuit board. The predicted manufacturing year refers to a preset time interval to which the manufacturing time of the circuit board belongs, for example, it can be divided into "before 2010", "from 2010 to 2015", "from 2016 to 2020" and "after 2021".

[0147] The acquisition method of the predicted manufacturing year and the aging evaluation grade of the R-block circuit board comprises:

[0148] S300: Let the initial value of r be 1, and the value range of r be 1 to R;

[0149] S301: Obtain the fluorescence spectrum feature vector of the rth block circuit board from the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector of the rth block circuit board from the near-infrared reflectance spectrum feature vector set, and the circuit board image feature vector of the rth block circuit board from the circuit board image feature vector set;

[0150] S302: Input the fluorescence spectrum feature vector, the near-infrared reflectance spectrum feature vector and the circuit board image feature vector of the rth block circuit board into the comprehensive evaluation model, and obtain the predicted manufacturing year and the aging evaluation grade of the rth block circuit board;

[0151] S303: Let r = r + 1, if r is less than or equal to R, return to S301 to continue execution, if r is greater than R, end the current process.

[0152] The training method of the comprehensive evaluation model comprises:

[0153] Pre-construct a comprehensive evaluation data set, the comprehensive evaluation data set comprises ZH groups of comprehensive evaluation data and the predicted manufacturing year and the aging evaluation grade corresponding to the ZH groups of comprehensive evaluation data, ZH is a positive integer; the comprehensive evaluation data comprises fluorescence spectrum feature vectors, near-infrared reflectance spectrum feature vectors and circuit board image feature vectors; divide the comprehensive evaluation data set into a training set and a validation set, the training set is used for comprehensive evaluation model parameter learning, and the validation set is used for real-time monitoring of the generalization performance and the overfitting degree of the comprehensive evaluation model;

[0154] A deep neural network with a multi-layer perceptron as its core is used as the comprehensive evaluation model. The comprehensive evaluation data is standardized and vectorized and then input into the deep neural network. The deep neural network consists of an input layer, a hidden layer, and an output layer. Each hidden layer uses a nonlinear activation function to extract high-order features, and the output layer uses a softmax activation function to obtain the probability distribution corresponding to each predicted manufacturing year and aging assessment level. Finally, the predicted manufacturing year and aging assessment level corresponding to the maximum probability are taken as the prediction results of the comprehensive evaluation model. During the training process, the cross-entropy loss function is used as the optimization target, and a gradient descent optimization algorithm is used to update the network weights. An early stopping strategy is set: when the prediction accuracy on the validation set reaches or exceeds the preset threshold, the comprehensive evaluation model is determined to have converged and training is terminated.

[0155] In an embodiment of the present invention, a comprehensive assessment model is constructed and the multi-source feature vectors corresponding to each circuit board (including fluorescence spectrum feature vectors, near-infrared reflectance spectrum feature vectors, and circuit board image feature vectors) are input into the comprehensive assessment model for joint analysis and inference processing. This effectively determines the predicted manufacturing age and aging assessment level of each circuit board. This process not only achieves high-precision, non-destructive identification of the circuit board's service background and aging status, but also significantly improves the targeted and adaptable setting of subsequent melting and recycling parameters. The comprehensive assessment model possesses cross-domain feature fusion capabilities and strong generalized recognition capabilities, making it particularly suitable for circuit board reuse scenarios where the proportion of old samples is small and material variations are complex.

[0156] Specifically, fluorescence spectral eigenvectors reflect the composition, content distribution, and ratio changes of metal elements on the circuit board surface, indirectly mapping differences in material systems such as solder and metal coatings used across generations of circuit boards. Near-infrared reflectance spectral eigenvectors reveal the decay characteristics of molecular functional groups in the circuit board's dielectric layer, polymeric materials, or encapsulating coatings, providing a strong indicator of material aging. Circuit board image eigenvectors capture visual manufacturing and degradation information, such as the circuit board's structural layout, solder joint density, and distribution uniformity. These three types of eigenvectors collectively constitute a feature fusion space that is representative, complementary, and capable of dimensionality reduction and abstraction.

[0157] Compared with the existing technology, this application does not rely on traditional serial number analysis, RFID chip tags or physical batch documents. Instead, it uses data-driven soft inference to achieve accurate identification of the manufacturing background and service degradation status of unlabeled circuit boards, significantly improving the intelligent perception and adaptability of automated recycling and smelting pyrolysis processes, and effectively avoiding problems such as low processing efficiency, excessive energy consumption or increased pollution risks caused by misjudgment of manufacturing batches or misestimation of aging levels. It has outstanding practical value and industrial promotion prospects.

[0158] The parameter control module generates pyrolysis parameter adjustment instructions based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, the circuit board image feature vector set, the predicted manufacturing year of the R-block circuit board, and the aging assessment grade, and intelligently and adaptively controls the pyrolysis parameters of the circuit board.

[0159] The method for intelligently and adaptively controlling the pyrolysis parameters of the circuit board comprises the following steps.

[0160] The predicted manufacturing year binary tuple set and the aging assessment grade binary tuple set are constructed based on the predicted manufacturing year and the aging assessment grade of the R-block circuit board.

[0161] The dominant label value is determined based on the predicted manufacturing year binary tuple set and the aging assessment grade binary tuple set.

[0162] The dominant label value, the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, and the circuit board image feature vector set are input into a pyrolysis parameter setting model to obtain a corresponding pyrolysis parameter combination, wherein the pyrolysis parameter combination comprises a target pyrolysis temperature, a heating rate, a pyrolysis residence time, an oxygen and inert gas ratio, and an initial pressure of a pyrolysis reaction chamber.

[0163] The system control platform sends the pyrolysis parameter combination to the corresponding control device and executes it.

[0164] The method for obtaining the predicted manufacturing year binary tuple set and the aging assessment grade binary tuple set comprises the following steps.

[0165] S400: Let the initial value of r be 1, and the value range of r be 1 to R.

[0166] S401: Obtain the predicted manufacturing year and the aging assessment grade of the rth circuit board.

[0167] Determine whether the predicted manufacturing year exists in the predicted manufacturing year binary tuple set; if it exists, add 1 to the circuit board quantity count value of the predicted manufacturing year binary tuple of the predicted manufacturing year; if it does not exist, construct the predicted manufacturing year and 1 of the rth circuit board into a predicted manufacturing year binary tuple and add it to the predicted manufacturing year binary tuple set; an example of the predicted manufacturing year binary tuple is (predicted manufacturing year of the rth circuit board, circuit board quantity count value).

[0168] determining whether the aging evaluation grade exists in the aging evaluation grade pair set; if the aging evaluation grade exists, adding 1 to the board quantity count value of the aging evaluation grade pair; if the aging evaluation grade does not exist, constructing the aging evaluation grade of the rth board and 1 into an aging evaluation grade pair, and adding the aging evaluation grade pair to the aging evaluation grade pair set; an example of the aging evaluation grade pair is (aging evaluation grade of the rth board, board quantity count value);

[0169] S402: setting r = r + 1, if r is less than or equal to R, returning to S401 to continue execution; if r is greater than R, ending the current process.

[0170] As shown in Figure 4 the method for determining the dominant component and constructing the dominant label value based on the predicted manufacturing year pair set and the aging evaluation grade pair set comprises:

[0171] initializing the dominant identification of the predicted manufacturing year and the dominant identification of the aging evaluation grade as no;

[0172] respectively taking the board quantity count value of each predicted manufacturing year pair in the predicted manufacturing year pair set and R as quotient, obtaining the corresponding board quantity proportion, and constructing as a first quantity proportion set;

[0173] respectively taking the board quantity count value of each aging evaluation grade pair in the aging evaluation grade pair set and R as quotient, obtaining the corresponding board quantity proportion, and constructing as a second quantity proportion set;

[0174] selecting the maximum first quantity proportion from the first quantity proportion set, determining whether the first quantity proportion is greater than a preset first quantity proportion threshold, if the determination result is yes, setting the dominant identification of the predicted manufacturing year as yes;

[0175] selecting the maximum second quantity proportion from the second quantity proportion set, determining whether the second quantity proportion is greater than a preset second quantity proportion threshold, if the determination result is yes, setting the dominant identification of the aging evaluation grade as yes;

[0176] For example, the first quantity proportion threshold and the second quantity proportion threshold in the present application can be set to 60%.

[0177] matching the dominant identification of the predicted manufacturing year and the dominant identification of the aging evaluation grade with the pre-constructed dominant label value mapping table to obtain the dominant label value; the dominant label value mapping table comprises the dominant identification of the predicted manufacturing year, the dominant identification of the aging evaluation grade, and the corresponding dominant label value.

[0178] For example, the dominant label value mapping table is shown in Table 3:

[0179] Table 3 Dominant label value mapping table

[0180]

[0181] It should be noted that in this embodiment of the present invention, a statistical distribution constructed based on a set of predicted manufacturing year pairs and a set of aging assessment grade pairs is used to identify the dominant identifiers for the predicted manufacturing year and the dominant identifier for the aging assessment grade. This identifier is then matched against a dominant label value mapping table to ultimately obtain a dominant label value. This dominant label value serves as a unified control strategy index, simplifying the subsequent pyrolysis parameter setting logic and improving the efficiency and reusability of the decision-making process.

[0182] Specifically, the dominant tag value realizes the rapid indexing of the four combination states through numerical coding, which is convenient for calling and expanding in the control model, parameter configuration table or logical judgment node. By introducing the dominant tag value, it is possible to build a tag-driven parameter control strategy to achieve rapid parameter matching for batches of different types of circuit boards. For example, when the dominant tag value is 1, that is, the predicted manufacturing year and aging level are both dominant, high-confidence parameters can be enabled first; when the dominant tag value is 4, that is, neither type of attribute is dominant, the fusion regression model is triggered to dynamically calculate the parameter group, realizing refined adaptive control, thereby improving the system's classification response capability to batches with different feature distributions.

[0183] The training method of the pyrolysis parameter setting model includes:

[0184] A pyrolysis parameter setting data set is pre-constructed, the pyrolysis parameter setting data set including QH group pyrolysis parameter setting data and pyrolysis parameter combinations corresponding to the QH group pyrolysis parameter setting data, where QH is a positive integer; the pyrolysis parameter setting data includes a dominant label value, a fluorescence spectrum feature vector set, a near-infrared reflectance spectrum feature vector set, and a circuit board image feature vector set; the pyrolysis parameter setting data set is divided into a training set and a validation set, the training set is used for learning parameters of the pyrolysis parameter setting model, and the validation set is used for real-time monitoring of the generalization performance and overfitting degree of the pyrolysis parameter setting model;

[0185] The deep neural network with a multilayer perceptron as the core is used as the pyrolysis parameter setting model. The pyrolysis parameter setting data is input into the deep neural network after being standardized and vectorized. The deep neural network is composed of an input layer, hidden layers and an output layer. Each hidden layer uses a nonlinear activation function to extract high-order features, and the output layer uses a Softmax activation function to obtain the probability distribution corresponding to each pyrolysis parameter combination. Finally, the pyrolysis parameter combination corresponding to the maximum probability is taken as the prediction result of the pyrolysis parameter setting model. In the training process, the cross-entropy loss function is used as the optimization objective, the gradient descent type optimization algorithm is used to update the network weights, and the early stopping strategy is set: when the prediction accuracy on the validation set reaches or exceeds the preset threshold, it is determined that the pyrolysis parameter setting model has converged and the training is terminated.

[0186] It is further supplemented that, in order to more clearly illustrate the application of the pyrolysis parameter setting result in the actual pyrolysis process, the following will illustrate the parameter execution and device adaptive adjustment process driven by the dominant label value in combination with specific control examples. The data-driven input and control linkage diagram of the circuit board pyrolysis parameters is as shown in Figure 5

[0187] Suppose the current processing object is a batch of R=120 pieces of waste circuit boards. The system has obtained the fluorescence spectrum feature vector, near-infrared reflectance spectrum feature vector and image feature vector of each circuit board through the first processing module, the second processing module and the third processing module, and constructed the prediction manufacturing year two-tuple set and the aging evaluation level two-tuple set.

[0188] After proportional judgment, the system identifies that the number of circuit boards with the predicted manufacturing year of “2015-2017” in this batch is 78, accounting for about 65%, which exceeds the first number proportion threshold of 60%; the number of circuit boards with the aging evaluation level of “level 3” is 74, accounting for about 61.6%, which also exceeds the second number proportion threshold of 60%. Therefore, the system determines that the dominant predicted manufacturing year and the dominant aging evaluation level are both established, and determines the dominant label value as “1” according to the dominant label value mapping table.

[0189] Subsequently, the system inputs the dominant label value, the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set and the circuit board image feature vector set into the pyrolysis parameter setting model, and outputs the following pyrolysis parameter combination:

[0190] Target pyrolysis temperature = 820°C;

[0191] Heating rate = 6.5°C / min;

[0192] Pyrolysis residence time = 22min;

[0193] Oxygen to inert gas ratio = 30:70;​

[0194] Pyrolysis reaction chamber initial pressure = 0.85 atm.

[0195] The system automatically issues the above parameter instructions to the pyrolysis control device: sets the target pyrolysis temperature to the main heating device temperature control device; adjusts the temperature rise rate into the temperature rise speed controller; configures the pyrolysis residence time in the time controller; injects the oxygen and inert gas ratio into the atmosphere mixing valve group controller; and configures the pyrolysis reaction chamber initial pressure in the gas inlet pressure regulating unit. Ultimately, under a variety of mixed fluctuation conditions of circuit boards, the risk of coking and dioxin exceeding the standard can be greatly reduced, and the energy efficiency, safety and intelligent level of the circuit board resource pyrolysis process can be significantly improved.

[0196] Embodiment 2:

[0197] Please refer to Figure 2 The embodiment provides a multi-modal sensing circuit board pyrolysis parameter adaptive control method, which comprises:

[0198] The R-block circuit board spectral signal obtained by the fluorescence spectrum acquisition area is analyzed and feature extracted to obtain a fluorescence spectrum feature vector set;

[0199] The R-block circuit board reflectance spectrum data collected by the near-infrared reflectance spectrum acquisition area is feature extracted and structured to obtain a near-infrared reflectance spectrum feature vector set;

[0200] The R-block circuit board surface image obtained by the image acquisition device is feature extracted and processed to obtain a circuit board image feature vector set;

[0201] Based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set and the circuit board image feature vector set, the R-block circuit board is evaluated to obtain the predicted manufacturing year and aging evaluation grade of the R-block circuit board;

[0202] Based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, the circuit board image feature vector set, the predicted manufacturing year and the aging evaluation grade of the R-block circuit board, a pyrolysis parameter adjustment instruction is generated, and the circuit board pyrolysis parameter is intelligently and adaptively controlled.

[0203] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0204] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. A multi-modal sensing circuit board pyrolysis parameter adaptive control method, characterized in that: include: Analyze and extract features of the R-block circuit board spectral signals acquired in the fluorescence spectrum acquisition area to obtain a set of fluorescence spectrum feature vectors; Perform feature extraction and structured processing on the R-block circuit board reflectance spectrum data collected in the near-infrared reflectance spectrum collection area to obtain a set of near-infrared reflectance spectrum feature vectors; Performing feature extraction processing on the surface images of the R blocks of circuit boards acquired by the image acquisition device to obtain a set of circuit board image feature vectors; Evaluate R circuit boards based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, and the circuit board image feature vector set to obtain the predicted manufacturing age and aging assessment grade of the R circuit boards; Based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, the circuit board image feature vector set, and the predicted manufacturing age and aging assessment level of the R-block circuit board, pyrolysis parameter adjustment instructions are generated, and the circuit board pyrolysis parameters are intelligently adaptively controlled.

2. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 1, characterized in that: The method for intelligent adaptive control of circuit board pyrolysis parameters includes: Based on the predicted manufacturing years and aging assessment levels of R circuit boards, a set of predicted manufacturing years and a set of predicted aging assessment levels are constructed; Based on the predicted manufacturing year binary set and the aging assessment grade binary set, the dominant component is determined and the dominant label value is constructed; Inputting the dominant tag values, the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, and the circuit board image feature vector set into a pyrolysis parameter setting model to obtain a corresponding pyrolysis parameter combination, wherein the pyrolysis parameter combination includes a target pyrolysis temperature, a heating rate, a pyrolysis residence time, an oxygen to inert gas ratio, and an initial pressure of the pyrolysis reaction chamber; The system control platform sends the pyrolysis parameter combination to the corresponding control device for execution.

3. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 2, characterized in that: The method for obtaining the predicted manufacturing year two-tuple set and the aging assessment grade two-tuple set includes: S400: Let the initial value of r be 1, and the value range of r be 1 to R; S401: Obtain the predicted manufacturing year and aging assessment level of the rth circuit board; Determine whether the predicted manufacturing year already exists in the set of predicted manufacturing year two-tuples; if so, add 1 to the circuit board quantity count value of the predicted manufacturing year two-tuple for the predicted manufacturing year; if not, construct a predicted manufacturing year two-tuple by adding the predicted manufacturing year of the rth circuit board and 1, and add the two-tuple to the set of predicted manufacturing year two-tuples; Determine whether the aging assessment level already exists in the aging assessment level two-tuple set; if so, add 1 to the circuit board quantity count value of the aging assessment level two-tuple of the aging assessment level; if not, construct an aging assessment level two-tuple by adding the aging assessment level of the r-th circuit board and 1, and add the two-tuple to the aging assessment level two-tuple set; S402: Let r = r + 1. If r is less than or equal to R, return to S401 to continue execution; if r is greater than R, end the current process.

4. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 2, characterized in that: The method for determining the dominant component and constructing the dominant label value based on the predicted manufacturing year binary set and the aging assessment level binary set includes: Initialize the dominant flag of the predicted manufacturing age and the dominant flag of the aging assessment level to be negative; Taking the quotient of the circuit board quantity count value of each predicted manufacturing year binary in the predicted manufacturing year binary set and R as the quotient, the corresponding circuit board quantity ratio is obtained, and the first quantity ratio set is constructed; Taking the quotient of the circuit board quantity count value of each aging assessment level binary group in the aging assessment level binary group set and R as the quotient, the corresponding circuit board quantity ratio is obtained, and a second quantity ratio set is constructed; Selecting the largest first quantity proportion from the first quantity proportion set, determining whether the first quantity proportion is greater than a preset first quantity proportion threshold, and if the determination result is yes, setting the dominant flag of the predicted manufacturing year to yes; Selecting the largest second quantity proportion from the second quantity proportion set, determining whether the second quantity proportion is greater than a preset second quantity proportion threshold, and if the determination result is yes, setting the dominant flag of the aging assessment level to yes; The dominant identifier of the predicted manufacturing age and the dominant identifier of the aging assessment level are matched with a pre-built dominant label value mapping table to obtain the dominant label value.

5. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 1, characterized in that: The method for obtaining the predicted manufacturing age and aging assessment level of R circuit boards includes: Based on a set of fluorescence spectrum feature vectors, a set of near-infrared reflection spectrum feature vectors, and a set of circuit board image feature vectors, multimodal feature vectors of each circuit board are obtained respectively, and the multimodal feature vectors are respectively input into a comprehensive evaluation model to obtain a predicted manufacturing age and aging assessment grade of each circuit board; the multimodal feature vectors are fluorescence spectrum feature vectors, near-infrared reflection spectrum feature vectors, and circuit board image feature vectors.

6. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 1, characterized in that: The method for obtaining the fluorescence spectrum feature vector set includes: The X-ray fluorescence spectrum acquisition device emits a micro-focus X-ray beam with a focal spot size of less than 100 microns to the surface of the R-block circuit board. Under the bombardment of X-rays, the inner electrons of the surface elements of the circuit board undergo transitions and instantaneously release X-ray fluorescence with characteristic energy. After the X-ray fluorescence enters the spectrometer detector, each fluorescence photon of the X-ray fluorescence forms an independent incident photon event. The spectrometer detector collects and converts the electronic signal released by the incident photon event to generate a pulsed voltage signal with an amplitude linearly related to the fluorescence energy. After pre-amplification, shaping and analog-to-digital conversion, the pulsed voltage signal enters the multi-channel analyzer for energy channel classification and statistical accumulation. The spectrum analysis software automatically extracts the peak energy, peak height and relative ratio of each element in the R block circuit board; The peak energy, peak intensity and relative ratio of R circuit boards are normalized and vectorized, and R groups of fluorescence spectrum feature vectors are constructed, and the R groups of fluorescence spectrum feature vectors are constructed into a fluorescence spectrum feature vector set.

7. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 1, characterized in that: The method for obtaining the near-infrared reflectance spectrum feature vector set includes: The near-infrared reflectance spectrometer emits near-infrared light with a wavelength range of 900 to 1700 nanometers onto the surfaces of R circuit boards. The near-infrared detector receives the near-infrared light signal reflected back from the circuit board surface and samples it according to the number of wavelength points Q. That is, the reflectivity of each circuit board is collected at Q wavelength points to form a Q-dimensional reflectivity vector. The principal component analysis is performed on the Q-dimensional reflectivity vector corresponding to each circuit board to obtain the near-infrared reflectance spectrum feature vectors corresponding to the R circuit boards; and the near-infrared reflectance spectrum feature vectors corresponding to the R circuit boards are constructed into a near-infrared reflectance spectrum feature vector set.

8. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 7, characterized in that: The method for obtaining the near-infrared reflectance spectrum feature vector corresponding to the R circuit boards includes: Construct the Q-dimensional reflectivity vector corresponding to each circuit board into a reflectivity matrix; The reflectivity data of each column in the reflectivity matrix is ​​respectively subtracted from the reflectivity mean of the corresponding column to obtain a zero mean matrix, and the covariance matrix is ​​calculated based on the zero mean matrix. The covariance matrix is ​​used to measure the linear correlation between the reflectivity of each band; The covariance matrix is ​​evaluated for eigenvalues ​​to obtain W eigenvalues, and the corresponding eigenvector is obtained for each eigenvalue. The W eigenvalues ​​sorted in descending order are constructed into an eigenvalue set, and the eigenvectors corresponding to the eigenvalue set are constructed into an eigenvector set. Determine the number of principal components ZCF based on the eigenvalue set and the preset cumulative contribution rate threshold; Select the first ZCF eigenvectors from the eigenvector set to construct the principal component vector set; A linear projection operation is performed on the Q-dimensional reflectivity vectors of the R circuit boards and the principal component vector set to obtain the principal component projection value sequence corresponding to the R circuit boards. The principal component projection value sequence is the near-infrared reflectance spectrum feature vector of the circuit board.

9. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 8, characterized in that: Methods for determining the number of principal components ZCF based on the eigenvalue set and a preset cumulative contribution rate threshold include: Step 1: Sum the eigenvalues ​​in the eigenvalue set to obtain the sum of the eigenvalues; set the initial value of the principal component number ZCF to 1; Step 2: Select the first ZCF eigenvalues ​​from the eigenvalue set and sum them to get the cumulative eigenvalue. Divide the cumulative eigenvalue by the sum of the eigenvalues ​​to get the cumulative contribution rate. Step 3: If the cumulative contribution rate is greater than the cumulative contribution rate threshold, the number of principal components ZCF is obtained and the current process ends; if the cumulative contribution rate is less than or equal to the cumulative contribution rate threshold, ZCF is set to ZCF + 1 and the process returns to step 2 to continue.

10. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 8, characterized in that: The method for obtaining the principal component projection value sequence corresponding to the R circuit boards includes: S100: constructing a principal component matrix from a set of principal component vectors; the principal component matrix has Q rows and ZCF columns; each column in the principal component matrix corresponds to a principal component vector; setting the initial value of r to 1, and the value range of r to 1 to R; S101: Obtain the r-th row matrix element from the zero-mean matrix, and record it as the centralized reflectivity vector; S102: Multiplying the central reflectivity vector by the principal component matrix to obtain a principal component projection value sequence; S103: Let r = r + 1. If r is less than or equal to R, return to S101 to continue execution. If r is greater than R, end the current process.

11. The method for adaptively controlling circuit board thermal decomposition parameters based on multimodal sensing according to claim 1, characterized in that: The method for obtaining the circuit board image feature vector set includes: S200: Let the initial value of r be 1, and the value range of r be 1 to R; S201: Preprocessing the surface image of the rth circuit board to obtain a preprocessed surface image of the rth circuit board; the preprocessing includes brightness normalization, contrast enhancement, edge smoothing, and artifact removal; S202: Divide the preprocessed surface image of the rth circuit board into H×H image grids of equal area, extract the number of circuit board solder joints corresponding to each image grid, and calculate the solder joint distribution uniformity factor of the rth circuit board; mark the image grids corresponding to the number of circuit board solder joints greater than a preset circuit board solder joint number threshold as soldering areas, count the number of soldering areas and record it as the number of soldering areas; extract the solder joint area corresponding to each image grid and the image grid area, and calculate the average solder joint density of the rth circuit board; S203: constructing a circuit board image feature vector of the rth circuit board based on the solder joint distribution uniformity factor, the number of soldering areas, and the average solder joint density of the rth circuit board, and adding the vector to the circuit board image feature vector set; S204: Let r = r + 1. If r is less than or equal to R, return to S201 to continue execution. If r is greater than R, end the current process.

12. A multi-modal sensing circuit board pyrolysis parameter adaptive control system, used to implement the multi-modal sensing circuit board pyrolysis parameter adaptive control method according to any one of claims 1 to 11, characterized in that: include: The first processing module is used to analyze and extract features from the spectrum signals of the R circuit boards acquired in the fluorescence spectrum acquisition area to obtain a set of fluorescence spectrum feature vectors; The second processing module is used to extract and structure the R-block circuit board reflectance spectrum data collected in the near-infrared reflectance spectrum collection area to obtain a near-infrared reflectance spectrum feature vector set; A third processing module is used to perform feature extraction processing on the surface images of the R blocks of circuit boards acquired by the image acquisition device to obtain a set of circuit board image feature vectors; A comprehensive evaluation module evaluates R circuit boards based on a set of fluorescence spectrum feature vectors, a set of near-infrared reflectance spectrum feature vectors, and a set of circuit board image feature vectors to obtain a predicted manufacturing age and aging evaluation grade of the R circuit boards; The parameter control module generates pyrolysis parameter adjustment instructions based on the fluorescence spectrum feature vector set, the near-infrared reflectance spectrum feature vector set, the circuit board image feature vector set, the predicted manufacturing age and aging assessment level of the R block circuit board, and performs intelligent adaptive control of the circuit board pyrolysis parameters.

Images