CN-122023308-A - Copper foil production surface process defect detection method and system based on image analysis

CN122023308ACN 122023308 ACN122023308 ACN 122023308ACN-122023308-A

Abstract

The invention discloses a copper foil production surface process defect detection method and system based on image analysis, and relates to the technical field of image recognition. The method and the system for detecting the defects of the copper foil production surface process based on image analysis comprise the following steps of S1, collecting a multi-source copper foil production data set cooperatively by a plurality of modules, carrying out standardized normalization processing, S2, preprocessing and multi-path detection to generate a defect candidate area, extracting defect characteristics and binding space coordinates to form a defect event set, S3, carrying out cycle, aggregation and morphological analysis on similar defects to generate space-time pattern description vectors, S4, establishing process and equipment fingerprints to carry out online matching to generate root cause diagnosis and production linkage, S5, summarizing diagnosis information to construct health indexes, generating a linkage strategy and carrying out closed-loop optimization. The invention effectively improves the defect root cause distinguishing precision and the linkage efficiency, and solves the problems of difficult process and equipment distinguishing and high shutdown checking cost and risk in the prior art.

Inventors

LUO ZHIHONG
LI JIANWEI
YANG YUPING
GU MINGHUANG
LI XIUJUAN
LI ZHENJIE

Assignees

广东嘉元科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260123

Claims (10)

1. The method for detecting the defects of the copper foil production surface process based on image analysis is characterized by comprising the following steps of: s1, collecting a multi-source copper foil production data set, and carrying out standardized normalization processing on the multi-source copper foil production data set through geometric mapping and parameter correction; S2, constructing a defect candidate region by adopting pretreatment enhancement and multi-path detection based on the processed multi-source copper foil production data set, constructing a correlation system of defect characteristics, space-time coordinates and event records, and constructing a defect event set; S3, based on classification organization and multidimensional distribution analysis of the defect event set, extracting periodic, aggregate and morphological mode characteristics of the similar defects, and packaging space-time mode description vectors; s4, according to the space-time pattern description vector, process and equipment fingerprint definition, offline construction and self-evolution updating are carried out, on-line space-time pattern matching is carried out, and root cause diagnosis and production linkage closed loop is carried out; S5, carrying out multisource diagnosis information collection, comprehensive health index construction and lateral classification on the basis of root cause diagnosis, linkage strategy generation, man-machine interaction execution and closed-loop optimization self-evolution.
2. The method for detecting the defects of the copper foil production surface process based on the image analysis of claim 1, wherein the specific process of collecting the multi-source copper foil production data set and carrying out the standardized normalization processing of the multi-source copper foil production data set through geometric mapping and parameter correction is as follows: The method comprises the steps of collecting a multi-source copper foil production data set, wherein the multi-source copper foil production data set comprises an image data stream, a motion parameter data set, a coordinate mapping data set and a metadata record set, adopting a collaborative collection mechanism of synchronous triggering and multi-dimensional parameter linkage of an encoder, carrying out structuring temporary storage on the collected multi-source copper foil production data set through an industrial level buffer area, recording real-time production process parameters and equipment running state data, and establishing a process log table and an operation and maintenance work list table; The method comprises the steps of constructing a reference coordinate and sample data set and a historical defect data set as analysis and training data sources, realizing standardized normalization processing of space and attribute through geometric mapping and parameter correction for a multi-source copper foil production data set, mapping image pixels to a unified physical coordinate system through polynomial fitting, correcting roll length precision through pulse counting and linear velocity combination for motion parameters, and synchronously executing all standardized operations and data acquisition.
3. The method for detecting defects on a copper foil production surface process based on image analysis according to claim 1, wherein the specific process of constructing a defect candidate region by preprocessing enhancement and multi-path detection based on the processed multi-source copper foil production data set is as follows: Inputting an image data stream of a multi-source copper foil production data set, constructing a defect candidate region through image preprocessing enhancement and multi-channel anomaly detection, adopting a steady-state enhancement and multi-mode anomaly capturing collaborative detection mechanism, performing layering preprocessing enhancement on an original gray level image, generating a candidate region through multi-channel detection, performing logic fusion and morphological operation on multi-channel response, removing isolated noise, marking connected domains, and outputting a preprocessed enhanced image and a primary candidate connected domain set.
4. The method for detecting the defects of the copper foil production surface process based on the image analysis of claim 1, wherein the method for constructing a correlation system of defect characteristics, space-time coordinates and event records is characterized by comprising the following specific processes of: Inputting a preliminary candidate connected domain set, a unified physical coordinate system and a multi-stage timestamp, constructing a ternary association system of defect features, space-time coordinates and event records, adopting depth recognition, carrying out multidimensional feature extraction and space-time anchoring association mechanism, constructing a defect comprehensive characterization function, multiplying a geometric feature weight coefficient by the L2 norm square of a geometric feature vector to obtain a geometric item, multiplying a position feature weight coefficient by the L2 norm square of a position feature vector to obtain a position item, multiplying a texture feature weight coefficient by the L2 norm square of a texture feature vector to obtain a texture item, adding the geometric item, the position item and the texture item to obtain a defect comprehensive characterization score value, associating the space-time coordinates with the event, and outputting a defect event set with ternary association and a manual review interface.
5. The method for detecting the defects of the copper foil production surface process based on the image analysis of claim 1, wherein the classification organization and the multidimensional distribution analysis based on the defect event set are used for extracting periodic, aggregation and morphological pattern characteristics of similar defects, and the specific process of packaging space-time pattern description vectors is as follows: Constructing a defect event layering classification organization mechanism, screening and sorting according to volume numbers and defect types, inputting a defect event set, establishing a defect event ordering organization rule, carrying out volume length direction periodic feature extraction on the defect event set, extracting breadth direction aggregation features, analyzing directionality and morphology modes, constructing a space-time mode description vector, outputting periodic analysis results of each volume and each type of defects, a current space-time mode description vector, historical defect root cause marking data, a manual rechecking confirmation record, operation and maintenance work order information, a breadth aggregation analysis result, directionality and morphology mode analysis result, a unified format space-time mode description vector and a time sequence mode stability evaluation report.
6. The method for detecting surface process defects of copper foil production based on image analysis according to claim 1, wherein the specific processes of process and equipment fingerprint definition, offline construction and self-evolution update according to the space-time pattern description vector are as follows: The method comprises the steps of establishing a standardized design of a multi-type fingerprint template, off-line calibration sub-library and a self-adaptive updating mechanism, inputting space-time pattern description vectors, historical defect root cause labeling data, manual rechecking confirmation records and operation and maintenance work order information, establishing fingerprint classification and template structuring rules, establishing a fingerprint library off-line construction, establishing a fingerprint library self-learning updating mechanism, extracting a corresponding space-time pattern vector, comparing the similarity with the existing fingerprint to obtain similarity, merging the similarity into a corresponding fingerprint sample set when the similarity is larger than or equal to a similarity threshold value, marking the similarity as a candidate new fingerprint when the similarity is smaller than the fingerprint threshold value, and outputting a structuring fingerprint library, a fingerprint vector set, a covariance matrix and threshold interval table, an increment sample labeling record table and a fingerprint library item example table.
7. The method for detecting the defects of the copper foil production surface process based on the image analysis of claim 1, wherein the specific processes of on-line space-time pattern matching, root cause diagnosis and production linkage closed loop are as follows: The method comprises the steps of establishing fingerprint library matching of a real-time space-time mode, establishing on-line matching and diagnosis, outputting root cause diagnosis, suggesting cross-dimension collaborative investigation of equipment and technology, structuring diagnosis records, carrying out production system linkage and closed-loop assessment, and outputting real-time root cause diagnosis reports, candidate fingerprint matching list, production linkage control quantity detail, diagnosis and treatment closed-loop assessment record list and fingerprint matching and diagnosis visual view.
8. The method for detecting the defects of the copper foil production surface process based on the image analysis of claim 1, wherein the specific processes of carrying out multisource diagnosis information collection, comprehensive health index construction and lateral classification on the basis of root cause diagnosis are as follows: The method comprises the steps of establishing multi-dimensional diagnosis information fusion, quantifying health index calculation and grading discrimination mechanism, inputting space-time mode description vector, fingerprint diagnosis table, defect event table, real-time production process parameter and equipment operation state data, carrying out standardized collection on multi-source diagnosis information, establishing comprehensive health index, obtaining density item by multiplying defect density weight coefficient by defect density normalization value of unit length, obtaining serious item by multiplying serious defect influence normalization value by severity weight coefficient, obtaining density item by multiplying mode anomaly weight coefficient by mode anomaly normalization value, and obtaining comprehensive health score value by summing three items; Synchronously constructing equipment side and process side risk scores to form a two-dimensional risk quantification basis, dividing diagnosis grades by combining the comprehensive health score value and the equipment side and process side risk scores, and outputting a multi-source diagnosis information collection report, a quality health score table, equipment and process side risk scores, comprehensive diagnosis results, fingerprint matching details, production process constraints, equipment capacity parameters and order delivery requirements.
9. The method for detecting the defects of the copper foil production surface process based on the image analysis of claim 1, wherein the specific processes of linkage strategy generation, man-machine interaction execution and closed-loop optimization self-evolution are as follows: The method comprises the steps of constructing a process and equipment linkage strategy library, a multi-constraint matching execution mechanism and an effect feedback optimization system, inputting comprehensive diagnosis results, fingerprint matching details, production process constraints, equipment capacity parameters and order delivery requirements, designing linkage strategy library structuring, carrying out strategy matching and accurate screening, carrying out man-machine interaction and production system linkage, evaluating and self-evolving a closed loop effect, obtaining a defect descending item by multiplying a defect density descending rate weight coefficient by a defect density descending rate, obtaining a fingerprint confidence item by multiplying a fingerprint confidence descending rate weight coefficient by a non-batch scrapping mark weight coefficient, obtaining a scrapping mark item by multiplying a non-batch scrapping mark, and summing the defect descending item, the fingerprint confidence item and the scrapping mark item to obtain a strategy effect grading value; The method comprises the steps of dynamically adjusting strategy recommendation priority based on strategy effect scoring values, improving recommendation sorting weights under similar diagnosis scenes and expanding applicable diagnosis grade ranges of strategies with the strategy effect scoring values larger than or equal to a primary effect threshold value, reducing recommendation priority or tightening applicable conditions for the strategies with the strategy effect scoring values smaller than the primary effect threshold value, marking to be optimized and triggering manual strategy revision, and outputting a recommendation linkage strategy list, a strategy execution record list, an effect evaluation report, a fingerprint library and feature extraction parameter updating suggestions.
10. A copper foil production surface process defect detection system based on image analysis, applying the copper foil production surface process defect detection method based on image analysis as claimed in any one of claims 1 to 9, comprising: the coil image acquisition and unified coordinate calibration module is used for acquiring a multi-source copper foil production data set and carrying out standardized normalization processing on the multi-source copper foil production data set through geometric mapping and parameter correction; the surface defect detection and basic feature extraction module is used for constructing a defect candidate area based on the processed multi-source copper foil production data set by adopting pretreatment enhancement and multi-path detection, constructing a correlation system of defect features, space-time coordinates and event records, and establishing a defect event set; the defect space-time mode construction and aggregation analysis module is used for carrying out periodic, aggregation and morphological mode feature extraction of similar defects based on classification organization and multidimensional distribution analysis of a defect event set and packaging space-time mode description vectors; the process fingerprint library construction and online matching module is used for defining process and equipment fingerprints, constructing offline and updating self-evolution according to the space-time pattern description vector, matching the space-time pattern online, diagnosing root cause and producing linkage closed loop; The fault lateral diagnosis and process linkage decision module is used for carrying out multisource diagnosis information collection, comprehensive health index construction and lateral classification on the basis of root cause diagnosis, linkage strategy generation, man-machine interaction execution and closed-loop optimization self-evolution.

Description

Copper foil production surface process defect detection method and system based on image analysis Technical Field The invention relates to the technical field of image recognition, in particular to a copper foil production surface process defect detection method and system based on image analysis. Background Along with the explosive growth of the requirements of the fields of AI, new energy automobiles and the like on high-end products such as HVLP copper foil, ultrathin copper foil and the like and the mainstream trend of the copper foil industry towards ultrathin, intelligent and high-quality transformation, the prior art has realized real-time online detection and accurate positioning of the surface defects of the copper foil. By means of technical supports such as a high-resolution industrial camera and a deep learning algorithm, the existing detection system can efficiently identify micron-sized defects such as pinholes, scratches and chromatic aberration, and finish quantitative evaluation of defect sizes and distribution densities, and quality screening requirements of large-scale continuous production are met. But is limited by a technical framework of pure image analysis, the prior art can only realize binary judgment of whether defects exist, cannot deeply excavate the cause association behind the defects, cannot combine the information of morphological characteristics, roll length direction periodicity, width direction repeated positions and the like of the defects, can distinguish process-adjustable defects such as electrolytic parameter drift, improper additive proportion and the like, and also cannot define equipment fault defects such as scraper damage, roll surface scratch and the like. For example, the invention patent with publication number CN120543548A discloses a surface process defect detection method and system for copper foil production, comprising the steps of collecting copper foil images in the production process, obtaining each image block in the copper foil images, carrying out edge detection on the copper foil images to obtain each edge line in the copper foil images, determining texture characteristic coefficients of each image block, constructing illumination contrast coefficients of each image block, correcting standard deviation parameters in a Gaussian filter, carrying out image enhancement on each image block by utilizing the Gaussian filter after the correction parameters and a local Retinex algorithm, and carrying out surface defect detection on the copper foil by the copper foil images after the image enhancement. Thereby improving the precision of the detection of the technological defects on the surface of the copper foil. For example, the invention patent with publication number CN120107195A discloses a method, a system and equipment for detecting the surface defects of a copper foil based on machine learning, which comprises the steps of shooting the surface of the copper foil to be detected, moving the copper foil at a preset speed to obtain a plurality of images with repeated areas, carrying out defect detection on each image to obtain a first defect area and position information, based on the defect position and the moving speed of the first image, deducing the position of the defect in a subsequent image, calculating the deviation value of the actual position and the deduced position, counting the number of images with the deviation value smaller than a preset threshold, judging the defect area as a second defect area if the ratio of the number to the total number of images is larger than or equal to the preset threshold, splitting the small-range area where the second defect area is located from each image, splicing the images into one image, and further detecting and classifying the defects by using a machine learning technology. The invention can improve the efficiency and accuracy of copper foil surface defect detection. In the prior art, the existing system depends on pure image surface defect detection technology, and only can realize defect identification and positioning, but lacks depth analysis capability for defect causes. The defect analysis method cannot be based on key information such as morphological distribution characteristics of defects, periodicity rules in the roll length direction, repeated positions in the width direction and the like, and cannot judge whether defects are caused by technological problems such as electrolytic parameter drift, improper additive proportion and the like or not, and whether the defects are caused by equipment faults such as scraper damage, roll surface flaws, foreign matter entrainment and the like is difficult to define. The pure image detection can only feed back the result of 'defect', and can not give out the tendency judgment of 'more bias process problem or bias equipment problem', so that when the situation of frequent defect is faced on site, the basis of accurate treatment is lacking, the passive mode