A method and system for manufacturing a circuit board

Abstract

The present application relates to the field of circuit board production technology, and in particular to a circuit board production method and system. The method includes: obtaining a full-area scanning image; identifying the full-area scanning image, determining the abnormal area to be detected and the coordinates of the abnormal area; obtaining a circuit board design file; analyzing the circuit board design file, determining the circuit topology; generating a dynamic detection path according to the circuit topology, and controlling the micro-probe array to perform a conduction test on the abnormal area to be detected according to the dynamic detection path, and determining the breakpoint coordinates; obtaining historical production batch data; determining the cause of the process defect according to the breakpoint coordinates, historical production batch data and the coordinates of the abnormal area. Avoid temporary conduction failure areas caused by process fluctuations, reduce false breakpoints caused by false touches, improve the defect pattern matching efficiency of continuous batches of the same equipment, break through the ambiguity of qualitative analysis, and reduce the recurrence rate of similar defects caused by experimental verification.

Description

Technical Field

[0001] The present application relates to the technical field of circuit board production, and in particular to a circuit board production method and system. Background Art

[0002] In the manufacturing of printed circuit boards (PCBs) and high-density interconnect boards (HDIs), circuit continuity defects are a key factor leading to product failure. Traditional methods for detecting breakpoints rely primarily on manual visual inspection, point-by-point testing with contact probes, or infrared thermal imaging. With the rapid development of high-density interconnect technology and flexible electronic devices, modern PCB manufacturing is showing a trend toward denser microhole arrays, nanometer-sized conductor line widths, and multilayered substrates. Traditional manual visual inspection methods based on optical microscopes are no longer sufficient to detect submicron defects.

[0003] Existing technologies generally use multimodal inspection methods such as X-ray tomography and infrared thermal imaging. However, in the case of microvia fracture detection in high-density interconnect boards, existing technologies have a single defect recognition dimension, making some hidden process deviations impossible to detect, seriously restricting the closed-loop optimization of the production process. Summary of the Invention

[0004] The present application provides a circuit board production method and system to solve the above problems.

[0005] In a first aspect, the present application provides a method for producing a circuit board, the method comprising:

[0006] Acquire a full-area scan image; identify the full-area scan image, and determine the abnormal area to be detected and the coordinates of the abnormal area;

[0007] Obtaining a circuit board design file; analyzing the circuit board design file to determine a circuit topology;

[0008] Generate a dynamic detection path according to the circuit topology, and control the micro-probe array to perform a continuity test on the abnormal area to be detected according to the dynamic detection path to determine the breakpoint coordinates;

[0009] Obtain historical production batch data; and determine the cause of the process defect based on the breakpoint coordinates, the historical production batch data, and the abnormal area coordinates.

[0010] This solution captures full-area scan images, eliminating the boundary blind spots of traditional linear scanning. It identifies full-area scan images, identifies the abnormal areas to be detected, and their coordinates, reducing reliance on manual experience and lowering the false alarm rate for abnormal area marking. It also captures PCB design files, eliminating the disconnect between the detection path and electrical characteristics. It analyzes PCB design files to determine the circuit topology, adapting to topology updates required during complex circuit redesigns. Based on the circuit topology, it generates a dynamic detection path. Based on this dynamic detection path, it controls the microprobe array to perform continuity testing on the abnormal areas to be detected, determining breakpoint coordinates, avoiding temporary continuity failures caused by process fluctuations, and reducing false breakpoints caused by accidental touches. It also captures historical production batch data to improve the efficiency of defect pattern matching for consecutive batches of the same equipment. Based on breakpoint coordinates, historical production batch data, and abnormal area coordinates, it determines the cause of process defects, breaking through the ambiguity of qualitative analysis and reducing the recurrence rate of similar defects that can be verified through experimental verification.

[0011] Optionally, determining the cause of the process defect based on the breakpoint coordinates, the historical production batch data, and the abnormal area coordinates includes:

[0012] Coupling the breakpoint coordinates and the abnormal area coordinates to determine the microscopic morphological features of the breakpoint position;

[0013] Acquiring historical production batch data, and determining a process characteristic database based on the historical production batch data;

[0014] The microscopic morphological features are compared with the process feature database, and the cause of the process defect is determined based on the comparison result.

[0015] This solution couples the coordinates of the breakpoints and abnormal regions to determine the microscopic topographical features of the breakpoint locations. This ensures spatial consistency in inspection path planning, improves inspection coverage, and enables the computational conversion of microscopic defect features into process parameter analysis. Historical production batch data is obtained and used to establish a process feature database, shortening the response time for historical data retrieval. Microscopic topographical features are compared with the process feature database, and based on the comparison results, the cause of the process defect is determined, improving the accuracy of identifying process parameter anomalies.

[0016] Optionally, generating a dynamic detection path according to the circuit topology structure includes:

[0017] Analyze the circuit topology and determine the impedance characteristics and node connections between topological nodes;

[0018] Dynamically determining edge weights of the circuit topology according to the impedance characteristics;

[0019] Constructing a weighted network model according to the node connection status and the edge weights;

[0020] A dynamic detection path is constructed according to the weighted network model and the edge weights.

[0021] This solution analyzes the circuit topology and determines the impedance characteristics and node connectivity between topological nodes. This overcomes the probe array's inability to sense changes in electrical states and eliminates the problem of missed breakpoints due to dielectric layer obstruction during visual inspection. Based on the impedance characteristics, the edge weights of the circuit topology are dynamically determined to improve defect detection efficiency. Based on the node connectivity and edge weights, a weighted network model is constructed to dynamically couple the circuit functional topology with the detection logic, providing a spatial navigation reference for the probe array. Based on the weighted network model and edge weights, a dynamic detection path is constructed to avoid local over-detection caused by a single high-weight path.

[0022] Optionally, the historical production batch data includes a spatiotemporal identifier, a process parameter set, and historical breakpoint micromorphological features; and determining a process feature database based on the historical production batch data includes:

[0023] Analyzing the process parameter set, and determining the correlation feature weights between the process parameters and the micromorphological features of any breakpoint according to the historical breakpoint micromorphological features and the analysis results;

[0024] Constructing a correlation matrix including process parameters and breakpoint micromorphological features according to the spatiotemporal identifier, the process parameter set, and the correlation feature weights;

[0025] The correlation matrix is determined as the process feature database.

[0026] This solution analyzes the process parameter set and, based on the historical breakpoint micromorphological characteristics and the analysis results, determines the correlation feature weights between the process parameters and the micromorphological features at any breakpoint. This supports quantitative attribution of defect causes and provides a data basis for closed-loop optimization of process parameters. Based on spatiotemporal identifiers, process parameter sets, and correlation feature weights, a correlation matrix is constructed that includes process parameters and breakpoint micromorphological features. This strengthens the influence weights of highly correlated parameters and reflects the causal relationship between process and defect. This correlation matrix is defined as a process feature database, overcoming the limitation of ineffective utilization of spatiotemporal information in historical data and ensuring the timeliness and accuracy of the correlation matrix.

[0027] Optionally, determining the correlation feature weight between the process parameters and the micromorphological features of any breakpoint based on the historical breakpoint micromorphological features and the analysis results includes:

[0028] Determining historical process parameters according to the analysis results;

[0029] Determining a breakpoint timestamp based on the microscopic morphological features of the historical breakpoint;

[0030] determining associated process parameters according to the breakpoint timestamp;

[0031] Determining a breakpoint grayscale gradient histogram according to the microscopic morphological characteristics of the historical breakpoints;

[0032] Determining the crack propagation angle according to the microscopic morphological characteristics of the historical breakpoints;

[0033] An associated feature weight is determined according to the breakpoint grayscale gradient histogram, the crack propagation angle, and the associated process parameters.

[0034] Through this solution, historical process parameters are determined based on the analysis results, eliminating the problem of ineffective extraction of spatiotemporal dimension information and ensuring the spatiotemporal consistency of the analysis. Based on the microscopic morphology characteristics of historical breakpoints, the breakpoint timestamp is determined to eliminate interference from non-related batches. Based on the breakpoint timestamp, the associated process parameters are determined to avoid mis-association between parameters and defects. Based on the microscopic morphology characteristics of historical breakpoints, the breakpoint grayscale gradient histogram is determined to capture the anisotropic characteristics during crack propagation and provide quantifiable spatial distribution characteristics for process parameter association. Based on the microscopic morphology characteristics of historical breakpoints, the crack propagation angle is determined to distinguish between stress concentration cracks and randomly distributed cracks. Based on the breakpoint grayscale gradient histogram, crack propagation angle and associated process parameters, the associated feature weights are determined to ensure that the association matrix only contains explainable causal relationships between process parameters and breakpoint features.

[0035] Optionally, constructing a correlation matrix including process parameters and breakpoint micromorphological features according to the spatiotemporal identifier, the process parameter set, and the correlation feature weights includes:

[0036] Performing spatiotemporal clustering on the process parameter set according to the spatiotemporal identifier to generate a spatiotemporal cluster;

[0037] Analyze the process parameter sets within each spatiotemporal cluster to determine the process parameter stability;

[0038] Comparing the process parameter stability with a preset threshold, and when the process parameter stability is higher than the preset threshold, expressing the production batch of the corresponding spatiotemporal cluster as a matrix row vector;

[0039] Expressing the associated feature weights as matrix column vectors;

[0040] A correlation matrix including process parameters and breakpoint micromorphology features is constructed through the matrix row vectors and the matrix column vectors.

[0041] Through this solution, the process parameter set is spatiotemporally clustered according to the spatiotemporal identifier to generate spatiotemporal clusters, isolate the interference of process differences between different equipment, and capture the short-term drift characteristics of the process parameters. The process parameter set within each spatiotemporal cluster is analyzed to determine the stability of the process parameters, quantify the degree of fluctuation of a single parameter, reflect the level of collaborative fluctuation of multiple parameters within the spatiotemporal cluster, and reduce the false alarm rate. The stability of the process parameters is compared with the preset threshold. When the stability of the process parameters is higher than the preset threshold, the production batch of the corresponding spatiotemporal cluster is expressed as a matrix row vector to eliminate the influence of random noise and ensure the comparability of cross-batch data. The associated feature weights are expressed as matrix column vectors to maintain the consistency of the feature space dimension. Through the matrix row vectors and matrix column vectors, an associated matrix containing process parameters and breakpoint micromorphological features is constructed so that the matrix interpretability meets the requirements of data-driven decision support.

[0042] Optionally, after determining the cause of the process defect according to the breakpoint coordinates, the historical production batch data, and the abnormal area coordinates, the method further includes:

[0043] Retrieve historical defect feature sets;

[0044] determining a set of associated process parameters according to the process feature database;

[0045] Calculating the feature similarity of the microscopic morphological features at the breakpoint position according to the historical defect feature set and the associated process parameter set;

[0046] Based on the feature similarity, a defect cause report is generated.

[0047] This solution retrieves historical defect feature sets, eliminating the issue of ineffective extraction of historical data's spatiotemporal dimensions. Based on the process feature database, a set of associated process parameters is determined, overcoming the disconnected nature of parameter-defect analysis. The feature similarity of micromorphological features at breakpoint locations is calculated based on the historical defect feature set and associated process parameter set, enabling quantitative correlation between microfeatures and macroparameters. This aligns with feature similarity calculation requirements and eliminates the problem of defect causes relying on empirical evidence. Based on feature similarity, a defect cause report is generated, enabling closed-loop optimization of defect causes.

[0048] Optionally, constructing a dynamic detection path according to the weighted network model and the edge weights includes:

[0049] obtaining a set of physical parameters of the microprobe array;

[0050] performing constrained optimization on the weighted network model according to the physical parameter set to obtain an optimized model;

[0051] A dynamic detection path is constructed according to the edge weights and the optimized model.

[0052] This solution captures the physical parameter set of the microprobe array, providing hardware capability constraints for path optimization and eliminating the difficulty of adapting fixed-path detection models to complex circuit topologies. Based on the physical parameter set, the weighted network model is constrained and optimized to achieve dynamic adaptation of detection logic to hardware capabilities. Based on edge weights and the optimized model, a dynamic detection path is constructed, eliminating the disconnect between abnormal region identification and circuit functional topology.

[0053] Optionally, determining the crack propagation angle according to the microscopic morphological characteristics of the historical breakpoints includes:

[0054] Extracting crack skeleton features based on the microscopic morphological features of the historical breakpoints;

[0055] Analyzing the crack skeleton characteristics and calculating the main crack direction angle;

[0056] Determine the crack branching angle based on fractal theory;

[0057] The crack propagation angle is determined according to the main crack direction angle and the crack branching angle.

[0058] This solution extracts crack skeleton features based on the microscopic morphology of historical breakpoints, avoiding interference from redundant details. The crack skeleton features are analyzed, the main crack orientation angle is calculated, and a numerical benchmark for the main crack direction is established, providing a directional indicator for defect cause analysis. Based on fractal theory, the crack branching angle is determined, and the geometric characteristics of the crack branches are quantified, revealing the fractal behavior during crack propagation. Based on the main crack orientation angle and crack branching angles, the crack propagation angle is determined, generating a unified crack propagation direction indicator.

[0059] In a second aspect, the present application provides a circuit board production system, the system comprising:

[0060] An image recognition module is used to obtain a global scan image; identify the global scan image, and determine the abnormal area to be detected and the coordinates of the abnormal area;

[0061] A structure determination module is used to obtain a circuit board design file; analyze the circuit board design file to determine a circuit topology;

[0062] a breakpoint determination module, configured to generate a dynamic detection path according to the circuit topology, and control a micro-probe array to perform a continuity test on the abnormal area to be detected according to the dynamic detection path to determine the breakpoint coordinates;

[0063] The cause determination module is used to obtain historical production batch data; and determine the cause of the process defect according to the breakpoint coordinates, the historical production batch data and the abnormal area coordinates.

[0064] Optionally, when the cause determination module determines the cause of the process defect based on the breakpoint coordinates, the historical production batch data, and the abnormal area coordinates, it is configured to:

[0065] Coupling the breakpoint coordinates and the abnormal area coordinates to determine the microscopic morphological features of the breakpoint position;

[0066] Acquiring historical production batch data, and determining a process characteristic database based on the historical production batch data;

[0067] The microscopic morphological features are compared with the process feature database, and the cause of the process defect is determined based on the comparison result.

[0068] Optionally, when the breakpoint determination module generates a dynamic detection path according to the circuit topology, it is configured to:

[0069] Analyze the circuit topology and determine the impedance characteristics and node connections between topological nodes;

[0070] Dynamically determining edge weights of the circuit topology according to the impedance characteristics;

[0071] Constructing a weighted network model according to the node connection status and the edge weights;

[0072] A dynamic detection path is constructed according to the weighted network model and the edge weights.

[0073] Optionally, the historical production batch data includes a spatiotemporal identifier, a process parameter set, and historical breakpoint micromorphological features; and when the cause determination module determines the process feature database based on the historical production batch data, it is configured to:

[0074] Analyzing the process parameter set, and determining the correlation feature weights between the process parameters and the micromorphological features of any breakpoint according to the historical breakpoint micromorphological features and the analysis results;

[0075] Constructing a correlation matrix including process parameters and breakpoint micromorphological features according to the spatiotemporal identifier, the process parameter set, and the correlation feature weights;

[0076] The correlation matrix is determined as the process feature database.

[0077] Optionally, when the cause determination module determines the correlation feature weight between the process parameters and the micromorphological features of any breakpoint based on the historical breakpoint micromorphological features and the analysis results, it is used to:

[0078] Determining historical process parameters according to the analysis results;

[0079] Determining a breakpoint timestamp based on the microscopic morphological features of the historical breakpoint;

[0080] determining associated process parameters according to the breakpoint timestamp;

[0081] Determining a breakpoint grayscale gradient histogram according to the microscopic morphological characteristics of the historical breakpoints;

[0082] Determining the crack propagation angle according to the microscopic morphological characteristics of the historical breakpoints;

[0083] An associated feature weight is determined according to the breakpoint grayscale gradient histogram, the crack propagation angle, and the associated process parameters.

[0084] Optionally, when the cause determination module constructs a correlation matrix including process parameters and breakpoint micromorphological features according to the spatiotemporal identifier, the process parameter set, and the correlation feature weights, it is configured to:

[0085] Performing spatiotemporal clustering on the process parameter set according to the spatiotemporal identifier to generate a spatiotemporal cluster;

[0086] Analyze the process parameter sets within each spatiotemporal cluster to determine the process parameter stability;

[0087] Comparing the process parameter stability with a preset threshold, and when the process parameter stability is higher than the preset threshold, expressing the production batch of the corresponding spatiotemporal cluster as a matrix row vector;

[0088] Expressing the associated feature weights as matrix column vectors;

[0089] A correlation matrix including process parameters and breakpoint micromorphology features is constructed through the matrix row vectors and the matrix column vectors.

[0090] Optionally, the circuit board production system further includes a report generation module, configured to:

[0091] Retrieve historical defect feature sets;

[0092] determining a set of associated process parameters according to the process feature database;

[0093] Calculating the feature similarity of the microscopic morphological features at the breakpoint position according to the historical defect feature set and the associated process parameter set;

[0094] Based on the feature similarity, a defect cause report is generated.

[0095] Optionally, when constructing a dynamic detection path according to the weighted network model and the edge weights, the breakpoint determination module is configured to:

[0096] obtaining a set of physical parameters of the microprobe array;

[0097] performing constrained optimization on the weighted network model according to the physical parameter set to obtain an optimized model;

[0098] A dynamic detection path is constructed according to the edge weights and the optimized model.

[0099] Optionally, when determining the crack propagation angle based on the microscopic morphological features of the historical breakpoints, the cause determination module is used to:

[0100] Extracting crack skeleton features based on the microscopic morphological features of the historical breakpoints;

[0101] Analyzing the crack skeleton characteristics and calculating the main crack direction angle;

[0102] Determine the crack branching angle based on fractal theory;

[0103] The crack propagation angle is determined according to the main crack direction angle and the crack branching angle. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0105] Figure 1 A schematic diagram of an application scenario provided in one embodiment of the present application;

[0106] Figure 2 A flowchart of a circuit board production method provided in one embodiment of the present application;

[0107] Figure 3 A schematic structural diagram of a circuit board production system provided in one embodiment of the present application. DETAILED DESCRIPTION

[0108] To make the purpose, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0109] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document, unless otherwise specified, generally indicates an "or" relationship between the related objects.

[0110] The embodiments of the present application are described in further detail below with reference to the accompanying drawings.

[0111] Modern circuit board manufacturing is trending toward denser microvia arrays, nanometer-sized conductor linewidths, and multilayered substrates. Traditional visual inspection methods based on optical microscopes are no longer sufficient to detect submicron defects. Existing technologies generally employ multimodal inspection methods such as X-ray tomography and infrared thermal imaging. However, in the context of microvia fracture detection in high-density interconnect boards, existing technologies rely on a single defect recognition dimension, making some hidden process deviations undetectable through inspection and severely restricting closed-loop optimization of the production process.

[0112] Based on this, the present application provides a circuit board production method and system to obtain a full-area scanning image and eliminate the boundary blind area of traditional linear scanning. Identify the full-area scanning image, determine the abnormal area to be detected and the coordinates of the abnormal area, reduce reliance on manual experience, and reduce the false alarm rate of abnormal area marking. Obtain the circuit board design file to eliminate the problem of disconnection between the detection path and the electrical characteristics. Analyze the circuit board design file, determine the circuit topology structure, and adapt to the topology update requirements when redesigning complex circuits. According to the circuit topology structure, generate a dynamic detection path, and according to the dynamic detection path, control the micro-probe array to perform a conduction test on the abnormal area to be detected, determine the breakpoint coordinates, avoid temporary conduction failure areas caused by process fluctuations, and reduce false breakpoints caused by false touches. Obtain historical production batch data to improve the defect pattern matching efficiency of continuous batches of the same equipment. According to the breakpoint coordinates, historical production batch data and abnormal area coordinates, determine the cause of the process defect, break through the ambiguity of qualitative analysis, and reduce the recurrence rate of similar defects that can be verified by experimental verification.

[0113] Figure 1 This is a schematic diagram of an application scenario provided by this application. The method provided by this application is applied when producing circuit boards.

[0114] Specifically, the method provided in the present application is applied to any server, and the server interacts with the multispectral imaging device to obtain a full-area scanning image through the multispectral imaging device. Identify the full-area scanning image and determine the abnormal area to be detected and the coordinates of the abnormal area. Obtain the circuit board design file. Analyze the circuit board design file and determine the circuit topology. According to the circuit topology, a dynamic detection path is generated, and according to the dynamic detection path, the micro-probe array is controlled to perform a conduction test on the abnormal area to be detected, determine the breakpoint coordinates, avoid temporary conduction failure areas caused by process fluctuations, and reduce false breakpoints caused by false touches. Obtain historical production batch data to improve the defect pattern matching efficiency of continuous batches of the same equipment. According to the breakpoint coordinates, historical production batch data and the coordinates of the abnormal area, determine the cause of the process defect, break through the ambiguity of qualitative analysis, and reduce the recurrence rate of similar defects that can be verified by experiments.

[0115] For specific implementation methods, please refer to the following embodiments.

[0116] Figure 2 This is a flow chart of a circuit board production method provided in one embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes:

[0117] S201, obtaining a full-area scan image; identifying the full-area scan image, and determining the abnormal area to be detected and the coordinates of the abnormal area;

[0118] The full-area scan image may be a two-dimensional image dataset including the surface and internal structure of the circuit board.

[0119] The abnormal area to be detected may be a potential defect area.

[0120] The abnormal area coordinates may be a set of spatial positions obtained by mapping the image pixel positions to the physical coordinate system of the circuit board through a coordinate transformation model.

[0121] Specifically, a multispectral imaging device performs a full-area scan of a circuit board (PCB) with completed outer-layer circuitry, generating a full-area scan image that includes near-infrared reflectance spectra and visible light topography. Grayscale gradient histogram analysis is then performed on the full-area scan image to calculate the standard deviation of the grayscale value change rate within each pixel's neighborhood. Subsequently, based on the normal distribution assumption for statistical process control, a threshold is determined using the standard deviation distribution. Regions exceeding the set threshold are then marked as abnormal regions to be detected. Finally, based on affine transformation theory, a coordinate conversion model is used to map the image pixel coordinates to the PCB's physical coordinate system, outputting the coordinates of the abnormal region.

[0122] S202, obtaining a circuit board design file; analyzing the circuit board design file to determine a circuit topology;

[0123] The circuit board design file may be an electronic file of conductive layer pattern data, drilling files, and stacking structure information.

[0124] The circuit topology may be a circuit connection relationship.

[0125] Specifically, the CAM's data bus interface is called to load the Gerber format data stream of the circuit board design file; then, the RS-274X standard parser is used to extract the vector graphics data of the conductive layer layer by layer; and then, geometric topology analysis is performed on the parsed vector graphics to determine the circuit topology structure.

[0126] S203, generating a dynamic detection path according to the circuit topology, and controlling the micro-probe array to perform a continuity test on the abnormal area to be detected according to the dynamic detection path to determine the breakpoint coordinates;

[0127] The dynamic detection path can be a probe movement sequence.

[0128] The microprobe array can be a multi-probe detection device.

[0129] The breakpoint coordinates may be locations of breakpoints in the conductive path.

[0130] Specifically, based on the real-time impedance characteristics of each side in the circuit topology, the degree centrality index of the network node where the abnormal area is located is calculated; then, a dynamic detection path is generated in descending order of the degree centrality index, and areas with a high impact on the circuit function are prioritized; subsequently, the micro-probe array is controlled to move along the path, and a continuity test is performed at the coordinates of the abnormal area; finally, the breakpoint coordinates are determined based on the continuity test results.

[0131] S204. Obtain historical production batch data; determine the cause of the process defect based on the breakpoint coordinates, the historical production batch data, and the abnormal area coordinates.

[0132] Historical production batch data can be a structured data set stored in MES that contains spatiotemporal identifiers, process parameter sets, and historical breakpoint micromorphological features.

[0133] The cause of process defects can be abnormal core process parameters.

[0134] Specifically, based on the spatiotemporal dimension indexing mechanism, the process parameter set with the same batch number and production timestamp in the historical production batch data is retrieved; then, a process-defect correlation matrix is constructed, and the similarity between the crack fractal dimension at the breakpoint coordinates and the temperature fluctuation standard deviation at the same position point in the historical production batch data is calculated; further, based on the similarity calculation results and the coordinates of the abnormal area, the cause of the process defect caused by the process parameter fluctuation is determined.

[0135] This solution captures full-area scan images, eliminating the boundary blind spots of traditional linear scanning. It identifies full-area scan images, identifies the abnormal areas to be detected, and their coordinates, reducing reliance on manual experience and lowering the false alarm rate for abnormal area marking. It also captures PCB design files, eliminating the disconnect between the detection path and electrical characteristics. It analyzes PCB design files to determine the circuit topology, adapting to topology updates required during complex circuit redesigns. Based on the circuit topology, it generates a dynamic detection path. Based on this dynamic detection path, it controls the microprobe array to perform continuity testing on the abnormal areas to be detected, determining breakpoint coordinates, avoiding temporary continuity failures caused by process fluctuations, and reducing false breakpoints caused by accidental touches. It also captures historical production batch data to improve the efficiency of defect pattern matching for consecutive batches of the same equipment. Based on breakpoint coordinates, historical production batch data, and abnormal area coordinates, it determines the cause of process defects, breaking through the ambiguity of qualitative analysis and reducing the recurrence rate of similar defects that can be verified through experimental verification.

[0136] In some embodiments, the breakpoint coordinates and the abnormal area coordinates are coupled to determine the microscopic morphological features of the breakpoint position; historical production batch data are obtained, and a process feature database is determined based on the historical production batch data; the microscopic morphological features are compared with the process feature database, and the cause of the process defect is determined based on the comparison results.

[0137] The breakpoint location can be the precise physical coordinates of the circuit board where electrical conduction fails due to a physical break.

[0138] The microscopic morphological feature can be a set of breakpoint surface structure parameters.

[0139] The process characteristic database can be a structured data set organized by production batch data.

[0140] The comparison result can be a similarity index between the micromorphological features and the process feature database.

[0141] Specifically, the coordinates of the breakpoint and the coordinates of the abnormal region are spatially geometrically superimposed and mapped to the same process coordinate system using a coordinate transformation matrix. A grayscale gradient histogram matching algorithm is then used to extract the micromorphological feature set within the abnormal region. A three-dimensional spatiotemporal index structure is established based on the batch number, equipment number, and production timestamp in historical production batch data. Time-series slicing is performed according to process parameter categories to generate a process feature database corresponding to the temperature zone distribution of each production equipment. Furthermore, a dynamic time warping algorithm is used to align the fractal dimension sequence of the micromorphological features with the standard deviation sequence of temperature fluctuations for the same equipment number and adjacent production periods in the process feature database, and a morphological matching index is calculated. When the morphological matching index exceeds a preset correlation threshold determined by ROC curve analysis, the coordinates of the abnormal region are mapped to the production equipment temperature zone distribution map to locate the corresponding temperature zone number. The deviation of the temperature gradient of the temperature zone from the standard process parameters is calculated according to the process window drift determination rule. If the deviation is too high, the process defect is determined to be caused by temperature parameter fluctuations.

[0142] This solution couples the coordinates of the breakpoints and abnormal regions to determine the microscopic topographical features of the breakpoint locations. This ensures spatial consistency in inspection path planning, improves inspection coverage, and enables the computational conversion of microscopic defect features into process parameter analysis. Historical production batch data is obtained and used to establish a process feature database, shortening the response time for historical data retrieval. Microscopic topographical features are compared with the process feature database, and based on the comparison results, the cause of the process defect is determined, improving the accuracy of identifying process parameter anomalies.

[0143] In some embodiments, the circuit topology is analyzed to determine the impedance characteristics and node connection status between topological nodes; based on the impedance characteristics, the edge weights of the circuit topology are dynamically determined; based on the node connection status and edge weights, a weighted network model is constructed; based on the weighted network model and edge weights, a dynamic detection path is constructed.

[0144] Topological nodes can be physical structure endpoints with electrical connection functions.

[0145] The impedance characteristic may be a parameter that characterizes the conductive performance of a line between adjacent topological nodes.

[0146] The node connection status can be a set of binary relationships determined based on physical connectivity.

[0147] The edge weight can be a normalized indicator reflecting the conductivity reliability of the line.

[0148] A weighted network model can be a computable graph model consisting of nodes, edges, and edge weights.

[0149] Specifically, based on the circuit topology, the four-wire method is used to perform point-to-point impedance measurements of adjacent topological nodes to construct an impedance signature. The connection relationships of each layer of the circuit topology are then extracted to generate node connectivity information. The impedance signature is then normalized, a baseline impedance is calculated, and a normalized weight matrix is generated. A real-time impedance offset is introduced to dynamically determine the edge weights of the circuit topology. A weighted network model is then constructed, with nodes as vertices and node connectivity as edges. The calculated edge weights are then injected into the corresponding edge attribute sets. Furthermore, an improved Dijkstra algorithm is used to traverse the weighted network model, prioritizing edge weights as the objective function. High-risk paths are identified, thereby constructing a dynamic detection path.

[0150] This solution analyzes the circuit topology and determines the impedance characteristics and node connectivity between topological nodes. This overcomes the probe array's inability to sense changes in electrical states and eliminates the problem of missed breakpoints due to dielectric layer obstruction during visual inspection. Based on the impedance characteristics, the edge weights of the circuit topology are dynamically determined to improve defect detection efficiency. Based on the node connectivity and edge weights, a weighted network model is constructed to dynamically couple the circuit functional topology with the detection logic, providing a spatial navigation reference for the probe array. Based on the weighted network model and edge weights, a dynamic detection path is constructed to avoid local over-detection caused by a single high-weight path.

[0151] In some embodiments, the process parameter set is analyzed, and the correlation feature weights between the process parameters and the micromorphological features of any breakpoint are determined based on the historical breakpoint micromorphological features and the analysis results; based on the spatiotemporal identifier, the process parameter set and the correlation feature weights, an association matrix containing the process parameters and the breakpoint micromorphological features is constructed; and the association matrix is determined as a process feature database.

[0152] The process parameter set may be a set of process conditions recorded in historical production batches.

[0153] The historical breakpoint micromorphological characteristics can be quantitative characteristics of the microstructure of the breakpoint defect that has occurred.

[0154] The analytical result can be a numerical matrix after the process parameter set is standardized.

[0155] Process parameters can be physical quantities that affect the quality of circuit boards during the production process.

[0156] The microscopic morphological characteristics of any breakpoint can be a quantitative indicator of the morphology of a single breakpoint defect.

[0157] The correlation feature weight may be a weight value representing the correlation strength between the process parameter and the micromorphology feature.

[0158] The spatiotemporal identifier can be a composite identifier consisting of a production line number and a production date.

[0159] The breakpoint micromorphological features may be quantifiable morphological features of the breakpoint defect in a microscopic image.

[0160] The association matrix may be a structured data matrix with a spatiotemporal identifier as the first dimension, integrating a process parameter set, a breakpoint feature vector, and an association feature weight.

[0161] Specifically, the process parameter set from the historical production batch data is extracted and Z-score normalized. Then, based on the microscopic morphology characteristics of the historical breakpoints, the grayscale gradient histogram, crack propagation angle, and fractal dimension are extracted to generate the breakpoint feature vector. Then, the mutual information algorithm is used to calculate the nonlinear correlation between the process parameters and the microscopic characteristics of any breakpoint, and an initial weight matrix is generated. Subsequently, a significance threshold is preset based on the statistical significance test theory in the mutual information analysis to determine the existence of a valid association between the parameters and the features, retain the weight value, and generate the associated feature weight. Then, with the spatiotemporal identifier as the first dimension, a tensor product operation is performed on the normalized process parameter set, the breakpoint feature vector, and the associated feature weight to generate an association matrix. Finally, the proportion of non-zero elements in the association matrix and the coverage rate of each breakpoint feature are checked. Once the conditions are met, the association matrix is confirmed as a process feature database.

[0162] This solution analyzes the process parameter set and, based on the historical breakpoint micromorphological characteristics and the analysis results, determines the correlation feature weights between the process parameters and the micromorphological features at any breakpoint. This supports quantitative attribution of defect causes and provides a data basis for closed-loop optimization of process parameters. Based on spatiotemporal identifiers, process parameter sets, and correlation feature weights, a correlation matrix is constructed that includes process parameters and breakpoint micromorphological features. This strengthens the influence weights of highly correlated parameters and reflects the causal relationship between process and defect. This correlation matrix is defined as a process feature database, overcoming the limitation of ineffective utilization of spatiotemporal information in historical data and ensuring the timeliness and accuracy of the correlation matrix.

[0163] In some embodiments, historical process parameters are determined based on the analysis results; the breakpoint timestamp is determined based on the historical breakpoint micromorphological characteristics; the associated process parameters are determined based on the breakpoint timestamp; the breakpoint grayscale gradient histogram is determined based on the historical breakpoint micromorphological characteristics; the crack propagation angle is determined based on the historical breakpoint micromorphological characteristics; and the associated feature weight is determined based on the breakpoint grayscale gradient histogram, the crack propagation angle and the associated process parameters.

[0164] The breakpoint timestamp may be a time-space coordinate composed of a time-space identifier.

[0165] The associated process parameters may be obtained by indexing a time-space mapping table according to the breakpoint timestamp to obtain a subset of process parameters corresponding to the production batch.

[0166] The breakpoint grayscale gradient histogram can be a multidimensional histogram that characterizes the spatial distribution characteristics of the crack surface texture.

[0167] The crack propagation angle may be an angle that characterizes the direction of crack propagation in the substrate.

[0168] Specifically, historical process parameters such as temperature gradient and electroplating current density are extracted from the analysis results. Then, based on the historical breakpoint micromorphological features and the production batch number in the breakpoint microscopic image metadata, the breakpoint timestamp is extracted. Furthermore, the spatiotemporal mapping table is indexed using the breakpoint timestamp to obtain the associated process parameters for the corresponding production batch. Subsequently, a restriction method is used to perform multi-directional Sobel operator convolution on the historical breakpoint micromorphological features, statistically analyzing the gradient amplitude distribution in each direction to generate a breakpoint grayscale gradient histogram. Based on the historical breakpoint micromorphological features, an image skeletonization algorithm is used to extract the main crack axis, and the crack propagation angle is detected using a Hough transform. Furthermore, the associated process parameters are strictly aligned with the breakpoint grayscale gradient histogram and crack propagation angle of the corresponding breakpoint according to the timestamp to construct a data pair set. The mutual information algorithm is then used to calculate the correlation strength between the associated process parameters, the breakpoint grayscale gradient histogram, and the crack propagation angle, and weight normalization is performed to generate the associated feature weights.

[0169] Through this solution, historical process parameters are determined based on the analysis results, eliminating the problem of ineffective extraction of spatiotemporal dimension information and ensuring the spatiotemporal consistency of the analysis. Based on the microscopic morphology characteristics of historical breakpoints, the breakpoint timestamp is determined to eliminate interference from non-related batches. Based on the breakpoint timestamp, the associated process parameters are determined to avoid mis-association between parameters and defects. Based on the microscopic morphology characteristics of historical breakpoints, the breakpoint grayscale gradient histogram is determined to capture the anisotropic characteristics during crack propagation and provide quantifiable spatial distribution characteristics for process parameter association. Based on the microscopic morphology characteristics of historical breakpoints, the crack propagation angle is determined to distinguish between stress concentration cracks and randomly distributed cracks. Based on the breakpoint grayscale gradient histogram, crack propagation angle and associated process parameters, the associated feature weights are determined to ensure that the association matrix only contains explainable causal relationships between process parameters and breakpoint features.

[0170] In some embodiments, the process parameter sets are spatiotemporally clustered according to spatiotemporal identifiers to generate spatiotemporal clusters; the process parameter sets within each spatiotemporal cluster are analyzed to determine the stability of the process parameters; the process parameter stability is compared with a preset threshold, and when the process parameter stability is higher than the preset threshold, the production batches of the corresponding spatiotemporal cluster are expressed as matrix row vectors; the associated feature weights are expressed as matrix column vectors; and an association matrix containing process parameters and breakpoint micromorphological features is constructed through the matrix row vectors and matrix column vectors.

[0171] A spatiotemporal cluster can be a production unit with spatiotemporal consistency.

[0172] The process parameter stability can be the arithmetic mean of the coefficient of variation of the parameters within the spatiotemporal cluster.

[0173] The preset threshold may be a critical value for determining the stability of the process parameters.

[0174] The corresponding space-time cluster may be a space-time cluster verified by the stability of process parameters.

[0175] A production batch can be a data set of process parameters with a unique spatiotemporal identifier.

[0176] The row vectors of the matrix may be process parameter mean sequences extracted from stable spatiotemporal clusters.

[0177] The matrix column vector can be a multidimensional vector consisting of associated feature weights arranged in a fixed order.

[0178] Specifically, based on spatiotemporal identifiers, a density-based spatiotemporal clustering with noise is performed on the process parameter sets to generate spatiotemporal clusters. Then, based on the process parameter sets within each spatiotemporal cluster, the coefficient of variation is calculated by parameter category. The arithmetic mean of the coefficient of variation is taken as the cluster stability indicator for the spatiotemporal cluster. The process parameter stability is then determined based on the cluster stability indicator. A threshold is then set based on empirical research findings from authoritative literature on semiconductor electroplating process quality control. The process parameter stability is then compared with the threshold. When the process parameter stability exceeds the threshold, the corresponding spatiotemporal clusters are arranged according to the pre-set process parameter order, generating a matrix row vector. The associated feature weights are then arranged in a fixed order to form a matrix column vector. Finally, a correlation matrix containing the process parameters and breakpoint micromorphological features is constructed by multiplying the matrix row vector by the matrix column vector.

[0179] Through this solution, the process parameter set is spatiotemporally clustered according to the spatiotemporal identifier to generate spatiotemporal clusters, isolate the interference of process differences between different equipment, and capture the short-term drift characteristics of the process parameters. The process parameter set within each spatiotemporal cluster is analyzed to determine the stability of the process parameters, quantify the degree of fluctuation of a single parameter, reflect the level of collaborative fluctuation of multiple parameters within the spatiotemporal cluster, and reduce the false alarm rate. The stability of the process parameters is compared with the preset threshold. When the stability of the process parameters is higher than the preset threshold, the production batch of the corresponding spatiotemporal cluster is expressed as a matrix row vector to eliminate the influence of random noise and ensure the comparability of cross-batch data. The associated feature weights are expressed as matrix column vectors to maintain the consistency of the feature space dimension. Through the matrix row vectors and matrix column vectors, an associated matrix containing process parameters and breakpoint micromorphological features is constructed so that the matrix interpretability meets the requirements of data-driven decision support.

[0180] In some embodiments, a historical defect feature set is retrieved; an associated process parameter set is determined based on a process feature database; the feature similarity of the microscopic morphological features of the breakpoint position is calculated based on the historical defect feature set and the associated process parameter set; and a defect cause report is generated based on the feature similarity.

[0181] The historical defect feature set may be a historical data set having the same abnormal region coordinates as the current breakpoint position.

[0182] The associated process parameter set may be a set of process parameters that matches the current production batch.

[0183] The feature similarity may be a quantitative indicator calculated using a cosine similarity algorithm.

[0184] The defect cause report may be an output attribution conclusion file containing an associated process parameter set and a corresponding defect type label.

[0185] Specifically, the historical defect feature set with the same abnormal region coordinates as the current breakpoint coordinates is extracted from the association matrix. Then, based on the spatiotemporal identifier of the current production batch, the associated process parameter set of the corresponding spatiotemporal cluster is retrieved from the process feature database. Furthermore, the microscopic morphological features of the breakpoint position are expressed as feature vectors; then, they are compared with each feature vector in the historical defect feature set, and the similarity score is calculated using a vector distance metric. Subsequently, based on the associated process parameter set, a weighted calculation is performed to prioritize matching cases with similar process parameters, outputting the highest feature similarity. Furthermore, based on the feature similarity, historical defect cases are selected; finally, the process parameter data of the historical defect cases is extracted from the associated process parameter set, and a defect cause report is automatically generated by combining it with quantitative attribution logic.

[0186] This solution retrieves historical defect feature sets, eliminating the issue of ineffective extraction of historical data's spatiotemporal dimensions. Based on the process feature database, a set of associated process parameters is determined, overcoming the disconnected nature of parameter-defect analysis. The feature similarity of micromorphological features at breakpoint locations is calculated based on the historical defect feature set and associated process parameter set, enabling quantitative correlation between microfeatures and macroparameters. This aligns with feature similarity calculation requirements and eliminates the problem of defect causes relying on empirical evidence. Based on feature similarity, a defect cause report is generated, enabling closed-loop optimization of defect causes.

[0187] In some embodiments, a physical parameter set of a microprobe array is obtained; based on the physical parameter set, a weighted network model is constrained optimized to obtain an optimized model; and a dynamic detection path is constructed based on the edge weights and the optimized model.

[0188] The physical parameter set may be a set of intrinsic hardware properties of the microprobe array.

[0189] The optimized model can be a topological network model that adapts to the hardware capabilities of the probe.

[0190] Specifically, the device's communication interface directly reads and extracts a set of physical parameters, including the number of probes in the microprobe array, minimum movement step length, and maximum detection coverage. Constraints are imposed based on these physical parameters, and a linear programming algorithm is used to optimize the node connectivity of the weighted network model to generate an optimized model. Using edge weights as impedance eigenvalues and the topological structure of the optimized model, the Dijkstra algorithm is used to calculate a prioritized detection path sequence and output a dynamic detection path.

[0191] This solution captures the physical parameter set of the microprobe array, providing hardware capability constraints for path optimization and eliminating the difficulty of adapting fixed-path detection models to complex circuit topologies. Based on the physical parameter set, the weighted network model is constrained and optimized to achieve dynamic adaptation of detection logic to hardware capabilities. Based on edge weights and the optimized model, a dynamic detection path is constructed, eliminating the disconnect between abnormal region identification and circuit functional topology.

[0192] In some embodiments, crack skeleton features are extracted based on the microscopic morphological features of historical breakpoints; the crack skeleton features are analyzed to calculate the main crack direction angle; the crack branching angle is determined based on fractal theory; and the crack propagation angle is determined based on the main crack direction angle and the crack branching angle.

[0193] The crack skeleton feature can be a simplified geometric structure of the crack trunk path.

[0194] The main crack direction angle can be the angle between the main crack and the reference coordinate axis.

[0195] Fractal theory can be a theoretical method for analyzing the fractal characteristics of cracks.

[0196] The crack branching angle can be the angle between the branch crack and the main crack.

[0197] Specifically, a simplified geometric structure representing the main crack path is extracted from the microscopic morphology of historical breakpoints and identified as the crack skeleton feature. The crack skeleton feature is then analyzed to identify the linear orientation of the main crack. Furthermore, the main crack orientation angle is calculated based on the geometric coordinates of the main crack. Subsequently, fractal theory is applied to analyze the branching structure of the crack fractal feature. The geometric characteristics of the crack branching structure are calculated using the fractal dimension to determine the crack branching angle. Finally, the main crack orientation angle and crack branching angle are integrated and synthesized through geometric relationships to determine the crack propagation angle.

[0198] This solution extracts crack skeleton features based on the microscopic morphology of historical breakpoints, avoiding interference from redundant details. The crack skeleton features are analyzed, the main crack orientation angle is calculated, and a numerical benchmark for the main crack direction is established, providing a directional indicator for defect cause analysis. Based on fractal theory, the crack branching angle is determined, and the geometric characteristics of the crack branches are quantified, revealing the fractal behavior during crack propagation. Based on the main crack orientation angle and crack branching angles, the crack propagation angle is determined, generating a unified crack propagation direction indicator.

[0199] Figure 3 A schematic diagram of a circuit board production system according to an embodiment of the present application is shown in FIG. Figure 3 As shown, the circuit board production system 300 of this embodiment includes: an image recognition module 301 , a structure determination module 302 , a breakpoint determination module 303 , and a cause determination module 304 .

[0200] The image recognition module 301 is used to obtain a global scan image; identify the global scan image, and determine the abnormal area to be detected and the coordinates of the abnormal area;

[0201] The structure determination module 302 is used to obtain a circuit board design file; analyze the circuit board design file to determine the circuit topology;

[0202] A breakpoint determination module 303 is configured to generate a dynamic detection path according to the circuit topology, and control the micro-probe array to perform a continuity test on the abnormal area to be detected according to the dynamic detection path to determine the breakpoint coordinates;

[0203] The cause determination module 304 is configured to obtain historical production batch data and determine the cause of the process defect based on the breakpoint coordinates, the historical production batch data, and the abnormal region coordinates.

[0204] Optionally, when determining the cause of a process defect based on the breakpoint coordinates, the historical production batch data, and the abnormal region coordinates, the cause determination module 304 is configured to:

[0205] Coupling the breakpoint coordinates and the abnormal area coordinates to determine the microscopic morphological features of the breakpoint position;

[0206] Acquiring historical production batch data, and determining a process characteristic database based on the historical production batch data;

[0207] The microscopic morphological features are compared with the process feature database, and the cause of the process defect is determined based on the comparison result.

[0208] Optionally, when generating a dynamic detection path according to the circuit topology, the breakpoint determination module 303 is configured to:

[0209] Analyze the circuit topology and determine the impedance characteristics and node connections between topological nodes;

[0210] Dynamically determining edge weights of the circuit topology according to the impedance characteristics;

[0211] Constructing a weighted network model according to the node connection status and the edge weights;

[0212] A dynamic detection path is constructed according to the weighted network model and the edge weights.

[0213] Optionally, the historical production batch data includes a spatiotemporal identifier, a process parameter set, and historical breakpoint micromorphological features; when the cause determination module 304 determines the process feature database based on the historical production batch data, it is used to:

[0214] Analyzing the process parameter set, and determining the correlation feature weight between the process parameter and the micromorphological feature of any breakpoint according to the historical breakpoint micromorphological feature and the analysis result;

[0215] Constructing a correlation matrix including process parameters and breakpoint micromorphological features according to the spatiotemporal identifier, the process parameter set, and the correlation feature weights;

[0216] The correlation matrix is determined as the process feature database.

[0217] Optionally, when determining the correlation feature weights between process parameters and any breakpoint micromorphology features based on the historical breakpoint micromorphology features and the analysis results, the cause determination module 304 is configured to:

[0218] Determining historical process parameters according to the analysis results;

[0219] Determining a breakpoint timestamp based on the microscopic morphological features of the historical breakpoint;

[0220] determining associated process parameters according to the breakpoint timestamp;

[0221] Determining a breakpoint grayscale gradient histogram according to the microscopic morphological characteristics of the historical breakpoints;

[0222] Determining the crack propagation angle according to the microscopic morphological characteristics of the historical breakpoints;

[0223] An associated feature weight is determined according to the breakpoint grayscale gradient histogram, the crack propagation angle, and the associated process parameters.

[0224] Optionally, when the cause determination module 304 constructs a correlation matrix including process parameters and breakpoint micromorphological features based on the spatiotemporal identifier, the process parameter set, and the correlation feature weights, it is configured to:

[0225] Performing spatiotemporal clustering on the process parameter set according to the spatiotemporal identifier to generate a spatiotemporal cluster;

[0226] Analyze the process parameter sets within each spatiotemporal cluster to determine the process parameter stability;

[0227] Comparing the process parameter stability with a preset threshold, and when the process parameter stability is higher than the preset threshold, expressing the production batch of the corresponding spatiotemporal cluster as a matrix row vector;

[0228] Expressing the associated feature weights as matrix column vectors;

[0229] A correlation matrix including process parameters and breakpoint micromorphology features is constructed through the matrix row vectors and the matrix column vectors.

[0230] Optionally, the circuit board production system further includes a report generation module 305, which is used to:

[0231] Retrieve historical defect feature sets;

[0232] determining a set of associated process parameters according to the process feature database;

[0233] Calculating the feature similarity of the microscopic morphological features at the breakpoint position according to the historical defect feature set and the associated process parameter set;

[0234] Based on the feature similarity, a defect cause report is generated.

[0235] Optionally, when constructing a dynamic detection path according to the weighted network model and the edge weights, the breakpoint determination module 303 is configured to:

[0236] obtaining a set of physical parameters of the microprobe array;

[0237] performing constrained optimization on the weighted network model according to the physical parameter set to obtain an optimized model;

[0238] A dynamic detection path is constructed according to the edge weights and the optimized model.

[0239] Optionally, when determining the crack propagation angle based on the microscopic morphological features of the historical breakpoints, the cause determination module 304 is configured to:

[0240] Extracting crack skeleton features based on the microscopic morphological features of the historical breakpoints;

[0241] Analyzing the crack skeleton characteristics and calculating the main crack direction angle;

[0242] Determine the crack branching angle based on fractal theory;

[0243] The crack propagation angle is determined according to the main crack direction angle and the crack branching angle.

[0244] The system of this embodiment can be used to execute the method of any of the above embodiments. Its implementation principles and technical effects are similar and will not be described in detail here.

Claims

1. A method for producing a circuit board, characterized in that: include: Acquire a full-area scan image; identify the full-area scan image, and determine the abnormal area to be detected and the coordinates of the abnormal area; Get the circuit board design file; Analyzing the circuit board design file to determine the circuit topology; Generate a dynamic detection path according to the circuit topology, and control the micro-probe array to perform a continuity test on the abnormal area to be detected according to the dynamic detection path to determine the breakpoint coordinates; Obtaining historical production batch data; determining the cause of the process defect based on the breakpoint coordinates, the historical production batch data, and the coordinates of the abnormal area; Determining the cause of the process defect based on the breakpoint coordinates, the historical production batch data, and the abnormal area coordinates includes: Coupling the breakpoint coordinates and the abnormal area coordinates to determine the microscopic morphological features of the breakpoint position; Acquiring historical production batch data, and determining a process characteristic database based on the historical production batch data; Comparing the microscopic morphology features with the process feature database, and determining the cause of the process defect based on the comparison results; Generating a dynamic detection path according to the circuit topology structure includes: Analyze the circuit topology and determine the impedance characteristics and node connections between topological nodes; Dynamically determining edge weights of the circuit topology according to the impedance characteristics; Constructing a weighted network model according to the node connection status and the edge weights; A dynamic detection path is constructed according to the weighted network model and the edge weights.

2. The method according to claim 1, characterized in that The historical production batch data includes a spatiotemporal identifier, a process parameter set, and historical breakpoint micromorphological features; and determining a process feature database based on the historical production batch data includes: Analyzing the process parameter set, and determining the correlation feature weights between the process parameters and the micromorphological features of any breakpoint according to the historical breakpoint micromorphological features and the analysis results; Constructing a correlation matrix including process parameters and breakpoint micromorphological features according to the spatiotemporal identifier, the process parameter set, and the correlation feature weights; The correlation matrix is determined as the process feature database.

3. The method according to claim 2, characterized in that Determining the correlation feature weights between process parameters and any breakpoint micromorphology features based on the historical breakpoint micromorphology features and the analysis results includes: Determining historical process parameters according to the analysis results; Determining a breakpoint timestamp based on the microscopic morphological features of the historical breakpoint; determining associated process parameters according to the breakpoint timestamp; Determining a breakpoint grayscale gradient histogram according to the microscopic morphological characteristics of the historical breakpoints; Determining the crack propagation angle according to the microscopic morphological characteristics of the historical breakpoints; An associated feature weight is determined according to the breakpoint grayscale gradient histogram, the crack propagation angle, and the associated process parameters.

4. The method according to claim 2, characterized in that The step of constructing a correlation matrix including process parameters and breakpoint micromorphological features according to the spatiotemporal identifier, the process parameter set, and the correlation feature weights includes: Performing spatiotemporal clustering on the process parameter set according to the spatiotemporal identifier to generate a spatiotemporal cluster; Analyze the process parameter sets within each spatiotemporal cluster to determine the process parameter stability; Comparing the process parameter stability with a preset threshold, and when the process parameter stability is higher than the preset threshold, expressing the production batch of the corresponding spatiotemporal cluster as a matrix row vector; Expressing the associated feature weights as matrix column vectors; A correlation matrix including process parameters and breakpoint micromorphology features is constructed through the matrix row vectors and the matrix column vectors.

5. The method according to claim 1, wherein After determining the cause of the process defect according to the breakpoint coordinates, the historical production batch data, and the abnormal area coordinates, the method further includes: Retrieve historical defect feature sets; determining a set of associated process parameters according to the process feature database; Calculating the feature similarity of the microscopic morphological features at the breakpoint position according to the historical defect feature set and the associated process parameter set; Based on the feature similarity, a defect cause report is generated.

6. The method according to claim 1, characterized in that The constructing of a dynamic detection path according to the weighted network model and the edge weights includes: obtaining a set of physical parameters of the microprobe array; performing constrained optimization on the weighted network model according to the physical parameter set to obtain an optimized model; A dynamic detection path is constructed according to the edge weights and the optimized model.

7. The method according to claim 2, characterized in that Determining the crack propagation angle according to the microscopic morphological characteristics of the historical breakpoints includes: Extracting crack skeleton features based on the microscopic morphological features of the historical breakpoints; Analyzing the crack skeleton characteristics and calculating the main crack direction angle; Determine the crack branching angle based on fractal theory; The crack propagation angle is determined according to the main crack direction angle and the crack branching angle.

8. A circuit board production system, applied to the method according to any one of claims 1 to 7, characterized in that: include: An image recognition module is used to obtain a global scan image; identify the global scan image, and determine the abnormal area to be detected and the coordinates of the abnormal area; A structure determination module is used to obtain a circuit board design file; analyze the circuit board design file to determine a circuit topology; a breakpoint determination module, configured to generate a dynamic detection path according to the circuit topology, and control a micro-probe array to perform a continuity test on the abnormal area to be detected according to the dynamic detection path to determine the breakpoint coordinates; The cause determination module is used to obtain historical production batch data; and determine the cause of the process defect according to the breakpoint coordinates, the historical production batch data and the abnormal area coordinates.

Images