Core board stacking sequence mistake proofing method based on visual recognition

Abstract

The present application relates to printed circuit board manufacturing technology field, propose based on visual identification's core board laminated sequence mistake proofing method, this method includes: the mapping baseline library corresponding between each layer and respective thermal-mechanical response characteristic parameter group is established; the real-time thermal-mechanical response characteristic parameter group of current core board is extracted;Real-time thermal-mechanical response characteristic parameter group and the thermal-mechanical response characteristic parameter group of each layer in mapping baseline library are matched in multidimensional mode;The actual layer information of current core board is compared with the layer information of target laminated sequence that should be pressed at present;The actual laminated record is compared with target laminated sequence, and the laminated integrity check result is obtained.Fundamentally, the present application avoids the cost increase and process complication caused by additional marking manufacturing process, and eliminates the risk of identification failure caused by marker wear, pollution or obstruction.

Description

Technical Field

[0001] This invention relates to the field of printed circuit board manufacturing technology, and in particular to a method for preventing errors in the stacking sequence of core boards based on visual recognition. Background Technology

[0002] In the manufacturing process of multilayer printed circuit boards (typically 8 layers or more), the lamination process is a critical step that determines product quality and reliability. The lamination process requires precisely stacking multiple inner core boards, prepreg (PP), and copper foil according to a predetermined layering order, and then pressing them together under high temperature and pressure. During this process, the correctness of the stacking order of the core boards directly affects the electrical performance and structural integrity of the final product—a single lamination error can lead to the scrapping of an entire batch of products, resulting in significant cost losses.

[0003] Currently, the main method for preventing errors in the core board stacking sequence during the lamination process involves setting identifiable marks on the core board and using external devices to read these marks for layer verification. For example, some solutions design layer barcodes or QR codes on the edge of the core board, using a scanning device to read the identification code and compare it with a preset sequence; others use a camera to photograph the layer markings on the core board, and computer software automatically compares them with standard images to determine consistency; still others set staggered copper blocks as layer markers on the edge of the core board, using capacitive sensors to detect the position of the copper blocks to identify the layers. Additionally, some solutions set retained conductive layer areas on the edge of the core board, using different positions of these areas to differentiate between layers.

[0004] While the above solutions achieve error prevention regarding the core board stacking order to some extent, they all share a common characteristic: they require additional markings (barcodes, numerical identifiers, copper blocks, conductive layers, etc.) to be made on the core board, and then these markings are read using specialized external reading devices (barcode scanners, CCD cameras, capacitive sensors, etc.). This "external marking—external reading—information comparison" error prevention mode has the following shortcomings: First, it requires additional processes and costs. Adding barcodes, digital labels, or copper blocks to the core board requires additional processing steps to the original core board manufacturing process, increasing material costs and process complexity.

[0005] Secondly, additional dedicated reading equipment is required. Barcode scanners, CCD cameras, capacitive sensors, and other such devices are not inherent components of the laminating equipment and must be installed and configured separately, increasing equipment investment costs and maintenance burden.

[0006] Third, there is a risk of labeling failure. Barcodes or digital labels may be worn, contaminated, or obscured during the handling and stacking of the core board, leading to reading failure or misreading; the manufacturing precision of physical markers such as copper blocks also directly affects the reliability of detection.

[0007] Fourth, the testing and lamination operations are separated. The layer verification in the above scheme is usually carried out during or after the core board is stacked, which is independent of the lamination operation itself and fails to organically integrate the error prevention function with the lamination process. Summary of the Invention

[0008] In view of the above, the main objective of this invention is to propose a method for preventing errors in the core board stacking sequence based on visual recognition, so as to solve the above-mentioned technical problems.

[0009] This invention proposes a method for preventing errors in the core board stacking sequence based on visual recognition, the method comprising the following steps: Step 1: Perform standard pressing procedures on the core boards of each layer. Simultaneously collect temperature dynamic response data and pressure dynamic response data of each core board during the entire pressing process through temperature and pressure sensors of the pressing equipment. Extract the thermal-mechanical synergistic response characteristic parameter group of each core board from the collected data to establish a mapping baseline library between each layer and its respective thermal-mechanical synergistic response characteristic parameter group. Step 2: During the actual layer-by-layer lamination process, the temperature sensor and pressure sensor synchronously collect the real-time temperature dynamic response data and real-time pressure dynamic response data of the current core board, and extract the real-time thermo-mechanical co-response characteristic parameter group of the current core board from the real-time temperature dynamic response data and real-time pressure dynamic response data. Step 3: Perform multi-dimensional pattern matching between the real-time thermal-mechanical co-response characteristic parameter group and the thermal-mechanical co-response characteristic parameter groups of each layer in the mapping baseline library to generate matching results, and identify the actual layer information of the current core board based on the matching results; Step 4: Compare the actual layer information of the current core board with the layer information that should be pressed in the target stacking sequence. If the comparison is consistent, the next layer can be pressed. If the comparison is inconsistent, an alarm is triggered and the pressing process is stopped. Step 5: Summarize the actual layer information of all core boards to generate actual stacking records, and perform a final verification between the actual stacking records and the target stacking order to obtain the stacking integrity verification result.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention utilizes the temperature-pressure coordinated dynamic response spontaneously generated by the inherent physical properties of the core board during the lamination process as the source of layer identification information. There is no need to set any external markings such as barcodes, digital labels, copper blocks or conductive layers on the core board, nor is there any need to install additional reading equipment such as barcode scanners, CCD cameras or capacitive sensors. Layer identification can be completed by simply reusing the temperature and pressure sensors of the lamination equipment. This fundamentally avoids the increased cost and process complexity caused by the external marking process, and at the same time eliminates the risk of identification failure caused by mark wear, contamination or obstruction.

[0011] 2. This invention meticulously divides the entire pressing process into preheating, resin flow, gelation, curing, and cooling stages. At each stage, it extracts independent temperature response features, independent pressure response features, and synergistic response features between temperature and pressure (including start-up time difference, rate ratio, and peak time difference). Furthermore, it plots temperature-pressure coupled response curves to extract waveform morphology features such as waveform symmetry and area, constructing a multidimensional thermo-mechanical synergistic response feature parameter set containing four types of characteristic parameters. Simultaneously, during the real-time pressing stage, it predicts the complete pressing process feature parameter set using early-stage data and verifies the deviations by comparing each parameter with measured values. This achieves a data closed loop between the calibration and real-time application stages, significantly improving the accuracy and reliability of layer identification.

[0012] 3. This invention constructs a multi-level tiered matching architecture. First, it quickly filters candidate layers based on the overall similarity of all feature parameters and performs admission judgment based on a first preset threshold. Then, it performs a detailed comparison of each type of feature parameter in the candidate layers and calculates the comprehensive confidence score. Finally, based on the tiered judgment of the second and third preset thresholds, the matching results are divided into three levels: high confidence automatically passes, medium confidence triggers a second-level review signal to pause the process and wait for manual confirmation, and low confidence triggers a third-level warning signal to stop the process. This realizes a differentiated hierarchical error prevention response mechanism and avoids the high false alarm rate or high false negative rate caused by a single threshold "one-size-fits-all" judgment.

[0013] 4. This invention, based on real-time comparison and control during the preceding layer-by-layer lamination process, further performs a dual final integrity verification after all core boards are laminated. The first step involves summarizing the actual layer information of all core boards into an actual stacking record and comparing it layer by layer with the target stacking sequence. The second step involves retrieving the standard thermo-mechanical synergistic response characteristic parameter set corresponding to each layer from the mapping baseline library and comparing it layer by layer with the real-time thermo-mechanical synergistic response characteristic parameter set actually collected during the lamination process. Four types of deviation values ​​are calculated: temperature independent response characteristic deviation value, pressure independent response characteristic deviation value, synergistic response characteristic deviation value, and waveform morphology characteristic deviation value. Then, a comprehensive judgment is made based on the single-layer characteristic consistency and overall characteristic consistency of each layer. This ensures that errors that may be missed in the layer-by-layer verification stage, such as incorrect layer order, abnormal characteristic parameters, and physical displacement, are fully intercepted before entering the next process after lamination, forming a complete error-proofing closed loop of "layer-by-layer process control + overall final confirmation".

[0014] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by means of embodiments of the invention. Attached Figure Description

[0015] Figure 1 This is a flowchart of the core board stacking sequence error prevention method based on visual recognition proposed in this invention. Detailed Implementation

[0016] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0017] These and other aspects of the embodiments of the present invention will become clear from the following description and accompanying drawings. In these descriptions and drawings, some specific embodiments of the present invention are specifically disclosed to illustrate some ways of implementing the principles of the embodiments of the present invention; however, it should be understood that the scope of the embodiments of the present invention is not limited thereto.

[0018] Please see Figure 1 This embodiment provides a method for preventing errors in the core board stacking sequence based on visual recognition. The method includes the following steps: Step 1: Perform standard pressing procedures on each layer of core board. Simultaneously collect temperature dynamic response data and pressure dynamic response data of each core board during the entire pressing process using temperature and pressure sensors of the pressing equipment. Extract the thermo-mechanical synergistic response characteristic parameter group of each core board from the collected data to establish a mapping baseline library between each layer and its respective thermo-mechanical synergistic response characteristic parameter group.

[0019] 2. In step 1, the standard pressing procedure is performed on each layer of core board. The temperature dynamic response data and pressure dynamic response data of each core board are collected synchronously by the temperature sensor and pressure sensor of the pressing equipment throughout the pressing process. The thermo-mechanical co-response characteristic parameter group of each core board is extracted from the collected data to establish a mapping baseline library between each layer and its respective thermo-mechanical co-response characteristic parameter group. The specific steps include the following: S101. Obtain the core boards of each layer to obtain an unmarked set of core boards; S102. Randomly select a core board of a certain layer from the unmarked core board set and place it in the pressing equipment. Execute the standard pressing procedure so that the current core board sequentially goes through the preheating stage, resin flow stage, gel stage, curing stage, and cooling stage during the entire pressing process. At each stage, corresponding temperature dynamic response and pressure dynamic response are generated. The temperature dynamic response data and pressure dynamic response data of the current core board are synchronously collected by the temperature sensor and pressure sensor of the pressing equipment to obtain the temperature-pressure synchronous response data set of the current core board. S103. Extract the thermal-mechanical coordinated response characteristic parameter group of the current core board from the temperature-pressure synchronous response data group of the current core board, and store the thermal-mechanical coordinated response characteristic parameter group of the current core board in correspondence with the layer information of the current core board; wherein, the thermal-mechanical coordinated response characteristic parameter group includes temperature independent response characteristic parameters, pressure independent response characteristic parameters, and coordinated response characteristic parameters between temperature and pressure; S104. Repeat steps S101 to S103 for the other layers of core boards in the unmarked core board set that have not been calibrated, until all layers of core boards have been stored in order to establish a mapping baseline library between each layer and its respective thermal-mechanical co-response characteristic parameter group.

[0020] In this embodiment of the invention, taking the lamination production of a certain model of eight-layer high-multilayer board as an example, firstly, prepare one inner core board for each of the eight layers required for the product—that is, one each of layers L1 to L8. These core boards have all completed the inner layer circuit fabrication and browning treatment, but none of them have any barcodes, numerical markings, copper blocks, or conductive layer areas set on the edge of the board, thus obtaining a set of unmarked core boards. Then, take one core board from this set, for example, take the L3 core board first, place it on the hot plate worktable of the lamination equipment, and perform the lamination operation according to the standard lamination program corresponding to this model of product.

[0021] During the lamination process, the core board sequentially undergoes the following stages: a preheating stage (temperature gradually increases from room temperature to approximately 80-120°C, with low pressure of approximately 5-15 psi applied), a flow stage (temperature continues to rise to approximately 150-180°C, with pressure increased to approximately 50-150 psi to soften and flow the prepreg resin and fill the gaps between circuits), a gel stage (the resin begins to undergo a cross-linking reaction, and the viscosity increases sharply), a curing stage (temperature rises to approximately 180-220°C, with high pressure of approximately 300-500 psi applied to completely cure the resin), and a cooling stage (temperature is gradually reduced to below 80°C while maintaining pressure). Throughout the entire lamination process, temperature and pressure dynamic response data of the L3 layer core board are synchronously collected at the same sampling frequency using the lamination equipment's built-in temperature and pressure sensors—for example, recording a temperature value and its corresponding pressure value every second—thus obtaining a set of synchronous temperature-pressure response data for the L3 layer core board throughout the entire lamination process. This dataset is actually a sequence of temperature-pressure data pairs precisely aligned on the time axis, recording the temperature and pressure values ​​at every moment from the start to the end of the compression process.

[0022] It should be noted that the "visual recognition" referred to in this invention does not refer to optical image acquisition and recognition using cameras, image sensors or image processing algorithms, but rather to extracting the layer characteristics of the core board from the synchronously acquired temperature-pressure dynamic response data through the temperature and pressure sensors built into the pressing equipment, and "perceiving" and "identifying" the layer identity of the core board by means of feature parameter analysis—that is, achieving layer determination through sensor signals.

[0023] Next, the thermo-mechanical synergistic response characteristic parameter set of the L3 layer core board is extracted from the temperature-pressure synchronous response data set—the specific extraction method will be described in detail in subsequent embodiments—and the extracted characteristic parameter set is stored in the database along with the layer information of the L3 layer (i.e., "L3"). After the calibration of the L3 layer core board is completed, the L1 layer core board is taken out from the unmarked core board set, and the above steps are repeated for it to undergo the same standard pressing, synchronous data acquisition and feature extraction, and the characteristic parameter set of the L1 layer is stored along with "L1". In this way, the above complete calibration process is performed on the L2, L4, L5, L6, L7 and L8 layer core boards respectively, until all eight layers of core boards have been stored accordingly. Thus, a mapping baseline library is established in the database that corresponds one-to-one between each layer from L1 to L8 and its respective set of thermo-mechanical co-response characteristic parameters. For example, layer L1 corresponds to a specific set of heating rate values ​​and temperature peak arrival time values, layer L2 corresponds to another set of values, and layer L3 corresponds to a third set of values. Each layer is different from the others and has a unique correspondence.

[0024] In a preferred embodiment of the present invention, extracting the thermal-mechanical coordinated response characteristic parameter set of the current core board from the temperature-pressure synchronous response data set of the current core board specifically includes the following steps: From the current core board's temperature-pressure synchronous response data set, extract the heating rate and temperature peak arrival time to obtain a set of temperature-independent response characteristic parameters; From the current temperature-pressure synchronous response data set of the core board, extract the pressure build-up rate and pressure peak arrival time to obtain the pressure independent response characteristic parameter set; From the current core board's temperature-pressure synchronous response data set, extract the coordinated response characteristic parameters between the temperature dynamic response and the pressure dynamic response to obtain a temperature-pressure coordinated response characteristic parameter set; wherein, the coordinated response characteristic parameters include the start-up time difference between the heating start-up time and the pressurization start-up time, the ratio of the temperature change rate to the pressure change rate, and the arrival time difference between the temperature peak and the pressure peak. Using temperature data as the abscissa and pressure data as the ordinate in the temperature-pressure synchronous response data set, a temperature-pressure coupled response curve is plotted. The waveform symmetry and waveform area of ​​the temperature-pressure coupled response curve are extracted to obtain a set of waveform morphology characteristic parameters of the coupled response curve. The temperature-independent response characteristic parameter group, the pressure-independent response characteristic parameter group, the temperature-pressure co-response characteristic parameter group, and the coupled response curve waveform morphology characteristic parameter group are merged to obtain the current core board's thermo-mechanical co-response characteristic parameter group.

[0025] In this embodiment of the invention, taking the calibration process of the L3 core board in the aforementioned eight-layer board as an example, after obtaining the temperature-pressure synchronous response data set of the L3 core board through synchronous acquisition, the heating rate of the L3 core board in the preheating stage (for example, the actual heating rate of the L3 core board in the preheating range of 80~120℃ is 2.3℃ / min) and the time to reach the temperature peak (for example, the time taken from the start of pressing to reaching the highest curing temperature of 185℃ is 47min) are first extracted from the data set. Thus, the temperature independent response characteristic parameter set of the L3 core board is obtained - the set of parameters reflects the thermal response characteristics of the L3 core board in the pressing process.

[0026] Meanwhile, the pressure build-up rate of the L3 core plate during the pressurization stage (e.g., the slope of the pressure change over time is 12 psi / min during the process of building up from the initial low pressure of 5 psi to the main pressure of 350 psi) and the time to reach the pressure peak (e.g., the time taken from the start of pressing to reaching the maximum pressure of 380 psi is 52 min) were extracted from this data set, thereby obtaining a set of pressure-independent response characteristic parameters of the L3 core plate—this set of parameters reflects the mechanical response characteristics of the L3 core plate during the pressing process.

[0027] It should be noted that, due to the different stacking positions of the core boards of different layers during the lamination process, the heat conduction path and pressure transmission path they are subjected to also differ. The heating rate of the core board located in the middle of the stacking structure (such as L4 and L5 layers) is usually slower than that of the outer core boards (such as L1 and L8 layers) which are closer to the heat plate. Furthermore, the specifications of the prepreg sheets (such as type 1080, type 2116, type 7628, etc.) used with the core boards of different layers are different, and their resin content and flow characteristics are also different. These factors together lead to systematic differences in the heating rate, temperature peak arrival time, pressure build-up rate, and pressure peak arrival time of each core board layer.

[0028] Furthermore, the synergistic response characteristic parameters between the temperature dynamic response and pressure dynamic response of the L3 layer core board are extracted from this temperature-pressure synchronous response data set. Specifically, the start-up time difference between the heating start-up time and the pressurization start-up time is calculated—for example, after the L3 layer core board starts pressing, the temperature rises first, and the pressure only begins to increase from the initial low pressure to the main pressure when the material temperature reaches about 55~60℃; the time difference between the two is the start-up time difference; the ratio of the temperature change rate to the pressure change rate is calculated—for example, comparing the aforementioned heating rate of 2.3℃ / min with the pressure build-up rate of 12psi / min yields a ratio of approximately 0.19 (℃ / psi); the arrival time difference between the temperature peak and the pressure peak is calculated—for example, the temperature peak arrival time is 47 min while the pressure peak arrival time is 52 min, a difference of 5 min.

[0029] Summarizing these synergistic response characteristic parameters yields the temperature-pressure synergistic response characteristic parameter set for the L3-layer core board—this set of parameters reflects the coupling evolution relationship between the temperature and pressure fields during the lamination process. Furthermore, using the temperature data as the x-axis and the pressure data as the y-axis in this temperature-pressure synchronous response data set, a temperature-pressure coupling response curve (i.e., a Lissajous figure) is plotted, corresponding to the temperature and pressure values ​​collected synchronously throughout the lamination process. The waveform symmetry and area of ​​this curve are then extracted as waveform morphology characteristic parameters, thus obtaining the waveform morphology characteristic parameter set for the coupling response curve of the L3-layer core board. The shape of this curve is uniquely determined by the synergistic evolution relationship between temperature and pressure throughout the lamination process—the symmetry and area of ​​the coupling response curve vary for different core layers due to differences in material composition and stacking positions.

[0030] Finally, the temperature-independent response characteristic parameter group, pressure-independent response characteristic parameter group, temperature-pressure coordinated response characteristic parameter group, and coupled response curve waveform morphology characteristic parameter group of the L3 layer core board are merged to obtain the complete thermo-mechanical coordinated response characteristic parameter group of the L3 layer core board. This characteristic parameter group constitutes the unique "physical fingerprint" of the L3 layer core board from four dimensions: temperature response, pressure response, thermo-mechanical coordinated response, and waveform morphology. The corresponding thermo-mechanical coordinated response characteristic parameter groups are extracted in the same way during the calibration of other layer core boards. In existing technologies, error prevention methods for the stacking sequence of core boards in the lamination process generally adopt a technical path of "external marking - external reading - information comparison"—for example, setting barcodes or QR codes on the edge of the core board and reading them with a scanning device, or setting misaligned copper blocks on the edge of the core board and detecting them with a capacitive sensor, or setting a conductive layer area on the edge of the core board for layer differentiation. These solutions all require additional marking on the core board and rely on specialized reading equipment. Although the "multilayer circuit board hot pressing monitoring method" involves layer monitoring during the hot pressing process, it obtains external marking information such as "layer model" or "layer identification area image," and still fails to break away from the framework of external marking.

[0031] In contrast, this invention utilizes the spontaneous temperature-pressure dynamic response generated by the core board's inherent physical properties (material composition, thickness, resin content, copper foil thickness, etc.) during the lamination process as the information source for layer identification. Different layers of core boards exhibit systematic differences in their heating rate, pressure build-up rate, temperature peak arrival time, pressure peak arrival time, and the start-up time difference, rate ratio, and peak time difference between temperature and pressure during the five stages of preheating, flow, gelation, curing, and cooling. This embodiment executes a standard lamination procedure on each layer of core board and simultaneously collects temperature-pressure dynamic response data. It extracts four types of feature parameters, including independent temperature response features, independent pressure response features, temperature-pressure co-response features, and coupled response curve waveform morphology features. A one-to-one mapping baseline library is established between each layer and its respective feature parameter group, thereby creating a unique "physical identity fingerprint" for each layer of core board without relying on any external markings or additional reading devices.

[0032] Step 2: During the actual layer-by-layer lamination process, the temperature sensor and pressure sensor synchronously collect the real-time temperature dynamic response data and real-time pressure dynamic response data of the current core board, and extract the real-time thermo-mechanical synergistic response characteristic parameter group of the current core board from the real-time temperature dynamic response data and real-time pressure dynamic response data.

[0033] In step 2, during the actual layer-by-layer lamination process, real-time temperature dynamic response data and real-time pressure dynamic response data of the current core board are simultaneously collected by the temperature sensor and pressure sensor, and the real-time thermo-mechanical co-response characteristic parameter set of the current core board is extracted from the real-time temperature dynamic response data and real-time pressure dynamic response data. Specifically, the steps are as follows: In the actual layer-by-layer lamination process, temperature and pressure dynamic response data of the current core plate are collected in the preheating and early flow stages using temperature and pressure sensors to obtain early temperature-pressure response data sets. Early thermo-mechanical co-response characteristic parameter set is extracted from the early temperature-pressure response data set to obtain the early thermo-mechanical co-response characteristic parameter set; wherein, the early thermo-mechanical co-response characteristic parameter set includes the heating rate in the preheating stage, the pressure build-up rate in the early flow stage, and the start-up time difference between heating start-up and pressurization start-up; Using the aforementioned mapping baseline library as a mapping benchmark, the early thermo-mechanical co-response characteristic parameter set is mapped to the current core plate's predicted thermo-mechanical co-response characteristic parameter set during the pressing process; Based on the early temperature-pressure response data set, we continue to collect real-time temperature dynamic response data and real-time pressure dynamic response data of the current core plate in the remaining stages of the entire pressing process, and after pressing is completed, we extract the measured thermo-mechanical co-response characteristic parameter set of the current core plate from all the collected data. Calculate the deviation value of each characteristic parameter between the predicted thermo-mechanical coordinated response characteristic parameter set and the measured thermo-mechanical coordinated response characteristic parameter set to obtain the prediction-measurement deviation result; If the deviation values ​​of each characteristic parameter in the predicted-measured deviation result are all less than the preset tolerance threshold, the measured thermo-mechanical synergistic response characteristic parameter group will be output as the real-time thermo-mechanical synergistic response characteristic parameter group of the current core board; if the deviation value of any characteristic parameter in the predicted-measured deviation result is greater than or equal to the preset tolerance threshold, an early warning signal will be triggered and the current pressing process will be paused for manual confirmation.

[0034] In this embodiment of the invention, taking the actual lamination production of the aforementioned eight-layer high-multilayer board as an example, it is assumed that the layer-by-layer lamination operation of the L5 layer core board is currently underway. During the actual lamination process, after the L5 layer core board is placed in the lamination equipment, the temperature dynamic response data and pressure dynamic response data of the L5 layer core board are collected in the early stage of lamination—that is, the preheating stage and the early flow stage (usually the first 15 to 20 minutes after the start of lamination, when the material temperature rises from room temperature to about 120 to 150°C and the pressure gradually builds up from the initial low pressure to about 50 to 100 psi) by the temperature and pressure sensors of the lamination equipment. This yields the early temperature-pressure response data set of the L5 layer core board. This early data set records the heating rate of the L5 core plate during the initial pressing stage (e.g., the actual heating rate from room temperature to 120°C is 2.5°C / min), the pressure build-up rate in the early flow stage (e.g., the slope of the pressure change over time from the initial low pressure of 5 psi to 80 psi is 8 psi / min), and the start-up time difference between heating start-up and pressurization start-up (e.g., the pressure only begins to increase from the initial low pressure to the main pressure when the material temperature reaches about 60°C, and the time difference between the two is about 5 min). From this, the early thermo-mechanical synergistic response characteristic parameter set of the L5 core plate was extracted.

[0035] It should be noted that, due to the different positions of the core plates in the laminated structure, the heat conduction path and pressure transmission path they receive in the early stage of pressing are also different. The core plate located in the middle of the laminate usually heats up more slowly than the outer core plate near the heat plate. Therefore, the early thermal-mechanical synergistic response characteristic parameter sets of different layers are different from each other.

[0036] Subsequently, using the established mapping baseline library as the mapping benchmark—which pre-stores the complete thermo-mechanical synergistic response characteristic parameter sets corresponding to each of the L1 to L8 layer standard core boards, and whose correspondence between the early thermo-mechanical synergistic response characteristic parameter sets and the complete characteristic parameter sets has been pre-established through the calibration experiment in step 1—the early thermo-mechanical synergistic response characteristic parameter sets of the L5 layer core board are mapped to the predicted thermo-mechanical synergistic response characteristic parameter sets of the L5 layer core board throughout the entire pressing process. In other words, based solely on the early data from approximately 15-20 minutes in the initial pressing stage (approximately 15%-20% of the entire pressing process), the complete thermo-mechanical synergistic response characteristic parameter sets that the L5 layer core board should exhibit throughout the entire pressing process can be predicted. These parameters include all four categories of characteristic parameters: predicted heating rate, temperature peak arrival time, pressure build-up rate, pressure peak arrival time, temperature-pressure start-up time difference, rate ratio, peak time difference, and the predicted temperature-pressure coupling response curve waveform symmetry and waveform area.

[0037] After obtaining the predicted thermo-mechanical synergistic response characteristic parameter set, the real-time temperature dynamic response data and real-time pressure dynamic response data of the L5 core plate are continuously collected through temperature and pressure sensors during the remaining stages of the entire pressing process—namely, the middle and late stages of the flow stage, the gel stage, the curing stage, and the cooling stage. After the entire pressing process (usually lasting about 90 to 120 minutes) is completed, the measured thermo-mechanical synergistic response characteristic parameter set of the L5 core plate is extracted from all the collected temperature-pressure synchronous data including the early and remaining stages—that is, the complete four types of characteristic parameters actually exhibited by the L5 core plate during the entire pressing process. Subsequently, the predicted thermo-mechanical synergistic response characteristic parameter set is compared with the measured thermo-mechanical synergistic response characteristic parameter set item by item. The deviation value of each of the four types of characteristic parameters, namely temperature independent response characteristics, pressure independent response characteristics, synergistic response characteristics, and waveform morphology characteristics, is calculated. For example, if the predicted heating rate is 2.4℃ / min and the measured heating rate is 2.5℃ / min, then the deviation value of this item is 0.1℃ / min; if the predicted temperature peak arrival time is 48min and the measured time is 47min, then the deviation value of this item is 1min, and so on. The deviation value of all characteristic parameters is summarized to obtain the prediction-measurement deviation result of the L5 layer core board.

[0038] If the deviation values ​​of each characteristic parameter in the prediction-measured deviation result are all less than the preset tolerance threshold (e.g., the heating rate deviation does not exceed 0.3℃ / min, the peak arrival time deviation does not exceed 3min, etc.), it indicates that the actual pressing behavior of the L5 layer core board is highly consistent with the prediction result based on early data. At this time, the measured thermo-mechanical synergistic response characteristic parameter group is used as the real-time thermo-mechanical synergistic response characteristic parameter group output of the L5 layer core board for subsequent step 3 to perform multi-dimensional pattern matching to identify the actual layer of the core board.

[0039] Conversely, if any characteristic parameter in the predicted-measured deviation result has a deviation value greater than or equal to the preset tolerance threshold—for example, the deviation between the measured heating rate of 2.8℃ / min and the predicted value of 2.4℃ / min reaches 0.4℃ / min, exceeding the preset tolerance of 0.3℃ / min—it indicates that the actual pressing behavior of the L5 layer core board deviates significantly from the expectation, which may mean that an abnormality has occurred in the current pressing process (such as batch differences in core board materials, pressing parameter drift, sensor signal abnormality, etc.). At this time, an early warning signal is triggered to remind the operator to pay attention to the current pressing status of the core board.

[0040] In existing technologies, sensor arrays are deployed in laminating equipment to collect temperature and pressure data and make predictions. However, the object of these predictions is "layer offset" (the physical offset after lamination), and the predictions are based on data from the entire lamination process—meaning that predictive analysis can only be performed after the entire lamination process is completed, which falls under the category of post-processing analysis. While the "multilayer circuit board hot-pressing monitoring method" involves layer monitoring during the hot-pressing process, it acquires "hot-melt lamination layer parameters" (layer type or layer identification area image), compares them with preset layer parameters to prevent confusion, and then sends an early warning signal—this still belongs to the technical path of directly comparing after reading external markers, without involving any prediction based on early data. Furthermore, although existing technologies also have schemes that generate evaluation coefficients by collecting temperature and fill rate data to achieve intelligent judgment of lamination timing, their purpose is to determine whether the resin has fully filled the gaps between each layer to control the lamination timing, rather than to identify the core board layer identity.

[0041] In contrast, this invention is the first to utilize temperature-pressure dynamic response data from the early stages of pressing (the preheating and early flow stages, accounting for approximately 15% to 20% of the entire pressing process) to predict the complete thermo-mechanical synergistic response characteristic parameter set for the entire pressing process during the pressing process (rather than after pressing). The predicted value is then compared and verified item by item with the measured complete characteristic parameter set after pressing. This dual mechanism of "early prediction + real-time verification" eliminates the reliance on reading external markings on the core board for layer identification, and eliminates the need to wait until the entire pressing process is complete to determine layer correctness—predictive results can be obtained in the early stages of pressing and continuously verified during the pressing process. If the deviation between the predicted and measured values ​​exceeds the tolerance range, an early warning signal can be triggered and intervention measures taken before pressing is completed.

[0042] Step 3: Perform multi-dimensional pattern matching between the real-time thermal-mechanical co-response characteristic parameter group and the thermal-mechanical co-response characteristic parameter groups of each layer in the mapping baseline library to generate matching results, and identify the actual layer information of the current core board based on the matching results.

[0043] In step 3, the real-time thermo-mechanical co-response characteristic parameter set is matched with the thermo-mechanical co-response characteristic parameter sets of each layer in the mapping baseline library in a multi-dimensional pattern to generate a matching result. Based on the matching result, the actual layer information of the current core board is identified. Specifically, this includes the following steps: The overall similarity is calculated between all the feature parameters contained in the real-time thermo-mechanical co-response feature parameter group and all the feature parameters contained in the thermo-mechanical co-response feature parameter group of each layer in the mapping baseline library, and the overall similarity score corresponding to each layer is obtained. Sort the overall similarity scores of each layer in descending order, select the layer with the highest overall similarity score as the candidate layer, and obtain the overall similarity score of the candidate layer. Determine whether the overall similarity score corresponding to the candidate layer is greater than or equal to the first preset threshold. If the overall similarity score corresponding to the candidate layer is greater than or equal to the first preset threshold, output the candidate layer as the target layer. If the overall similarity score corresponding to the candidate layer is less than the first preset threshold, trigger a first-level warning signal and stop the current core board pressing process. After outputting the target layer, each type of feature parameter contained in the real-time thermo-mechanical co-response feature parameter group is compared with the corresponding type of feature parameter in the mapping baseline library of the target layer, and the class deviation value of each type of feature parameter is calculated to obtain the class deviation results of each type of feature. The comprehensive confidence score is calculated based on the class deviation results of each type of feature. The system determines whether the overall confidence score is greater than or equal to a second preset threshold. If the overall confidence score is greater than or equal to the second preset threshold, the target layer is output as a layer to be confirmed. If the overall confidence score is greater than or equal to a third preset threshold but less than the second preset threshold, a second-level review signal is triggered, the target layer is marked as pending review, and the lamination process is paused for manual confirmation. If the overall confidence score is less than the third preset threshold, a third-level warning signal is triggered, and the lamination process of the current core board is terminated. The third preset threshold is less than the second preset threshold. After outputting the layers to be confirmed, the layers to be confirmed are finally confirmed, and the final confirmation result is generated. If the final confirmation result is passed, the layers to be confirmed are output as the actual layer information of the current core board; if the final confirmation result is failed, a level three warning signal is triggered, and the pressing process of the current core board is stopped.

[0044] In this embodiment of the invention, taking the actual lamination production of the aforementioned eight-layer high-multilayer board as an example, assuming that the lamination of the L5 layer core board has been completed and the real-time thermo-mechanical synergistic response characteristic parameter set of the L5 layer core board has been obtained through step 2, that is, it includes four types of characteristic parameters: temperature-independent response characteristics (heating rate 2.5℃ / min, peak temperature arrival time 47min), pressure-independent response characteristics (pressure build-up rate 11psi / min, peak pressure arrival time 52min), temperature-pressure synergistic response characteristics (start-up time difference 5min, rate ratio approximately 0.21℃ / psi, peak time difference 5min), and coupled response curve waveform morphology characteristics (waveform symmetry 0.92, waveform area 1850℃·psi). At this time, it is necessary to match this real-time characteristic parameter set with the standard characteristic parameter sets of each layer (L1 to L8) in the mapping baseline library established in step 1 to identify the actual layer of the core board.

[0045] First, a first-level matching—overall similarity coarse screening—is performed. All feature parameters contained in the real-time thermo-mechanical co-response feature parameter set of the L5 core panel—that is, all specific parameter values ​​in the four categories of features mentioned above: temperature independence, pressure independence, co-response, and waveform morphology—are treated as a whole. This is then compared one by one with all feature parameters contained in the standard thermo-mechanical co-response feature parameter sets of each layer from L1 to L8 in the mapping baseline library. (For example, methods such as Euclidean distance or Pearson correlation coefficient are used to calculate the overall similarity between the two sets of multidimensional feature vectors.) This yields the overall similarity score between the L5 core panel and each layer from L1 to L8. Assume the calculated overall similarity scores for each layer are as follows: 0.45 with layer L1, 0.51 with layer L2, 0.92 with layer L3, 0.38 with layer L4, 0.96 with layer L5, 0.33 with layer L6, 0.29 with layer L7, and 0.41 with layer L8—with the highest similarity score of 0.96 with layer L5, followed by 0.92 with layer L3. After sorting the overall similarity scores for each layer in descending order, the layer with the highest overall similarity score—layer L5—is selected as the candidate layer, resulting in candidate layer L5 and its corresponding overall similarity score of 0.96.

[0046] The second level of matching—admission determination—is then performed. It checks whether the overall similarity score (0.96) corresponding to candidate layer L5 is greater than or equal to a first preset threshold (e.g., the first preset threshold can be set to 0.85). Since 0.96 ≥ 0.85, the admission condition is met, and candidate layer L5 is output as the target layer for subsequent refined verification. If, assuming, the highest overall similarity score calculated in a certain matching is only 0.72, which is lower than the first preset threshold of 0.85, it indicates that the real-time feature parameter set of the current core board does not match the standard feature parameter set of all layers in the mapping baseline library—this may mean that the current core board is not the standard core board corresponding to this product model (e.g., incorrect incoming materials), or that a serious abnormality occurred during the pressing process (e.g., temperature or pressure sensor malfunction causing data distortion). In this case, a first-level warning signal is triggered, immediately stopping the pressing process of the current core board to prevent the incorrect layer from being misjudged as a layer and production from continuing.

[0047] After outputting the target layer L5, the third level of matching—class-by-class refined verification—is performed. The four types of feature parameters contained in the real-time thermo-mechanical co-response feature parameter group of the L5 core board—namely, the temperature-independent response feature parameter group, the pressure-independent response feature parameter group, the temperature-pressure co-response feature parameter group, and the coupled response curve waveform morphology feature parameter group—are compared with the standard feature parameters of the corresponding categories of the target layer L5 in the mapping baseline library.

[0048] For example, the real-time independent temperature response characteristic parameters (heating rate 2.5℃ / min, peak temperature arrival time 47min) are compared with the L5 standard independent temperature response characteristic parameters (heating rate 2.4℃ / min, peak temperature arrival time 48min), and the deviation value of this type of characteristic is calculated. The real-time independent pressure response characteristic parameters are compared with the L5 standard independent pressure response characteristic parameters; the real-time temperature-pressure co-response characteristic parameters are compared with the L5 standard co-response characteristic parameters; and the real-time coupled response curve waveform morphology characteristic parameters are compared with the L5 standard waveform morphology characteristic parameters. After calculating the class deviation values ​​of each type of characteristic parameter—for example, the class deviation value for independent temperature response characteristics is 3%, for independent pressure response characteristics is 2%, for co-response characteristics is 4%, and for waveform morphology characteristics is 1%—a comprehensive confidence score is calculated based on these four types of deviation values ​​using a weighted average or other comprehensive scoring method. Assume the calculated comprehensive confidence score is 96.5 points (out of 100).

[0049] The process then proceeds to the fourth level of matching—a three-threshold grading determination. This involves assessing the relationship between the overall confidence score of 96.5 and the second (e.g., 90) and third (e.g., 75) preset thresholds. Since 96.5 ≥ 90, the second preset threshold condition is met, and the target layer L5 is output as the layer to be confirmed, entering the final confirmation stage. If, in a certain matching, the overall confidence score is 82—greater than or equal to the third preset threshold of 75 but less than the second preset threshold of 90—it indicates a certain degree of uncertainty in the matching result, possibly due to slight fluctuations in characteristic parameters caused by batch differences in the core board. In this case, a second-level review signal is triggered, marking the target layer as pending review and pausing the pressing process for manual confirmation. Operators can confirm the actual layer of the core board through visual inspection or auxiliary means before deciding whether to continue. If the overall confidence score is only 68—less than the third preset threshold of 75—it indicates an unreliable matching result, possibly due to abnormal incoming core board materials, severe drift in pressing parameters, or sensor malfunction. In this case, a third-level warning signal is triggered, and the current core board pressing process is immediately stopped.

[0050] Finally, the process proceeds to the fifth level of matching—final confirmation. After outputting the layer to be confirmed, L5, a final confirmation is performed—for example, by automatically checking whether the layer to be confirmed matches the layer expected to be pressed in the current lamination process, or by manual verification—generating a final confirmation result. If the final confirmation result is successful, the layer to be confirmed, L5, is output as the actual layer information of the current core board (i.e., the currently pressed core board is identified as layer L5), for comparison and control in subsequent step 4. If the final confirmation result is unsuccessful—for example, the system finds that the layer to be confirmed, L5, does not match the layer expected to be pressed in the current lamination process (assuming that layer L5 should be pressed but there is an actual anomaly)—a level three warning signal is triggered, and the current core board lamination process is immediately stopped.

[0051] In existing technologies, the "multilayer circuit board hot-pressing monitoring method" obtains the hot-melt pressing layer parameters (layer type or layer identification area image) of the layer board, performs "anti-mixing processing" on them with preset layer parameters to obtain the hot-pressing mixing amount, and then sends an early warning signal based on the mixing amount. The matching logic of this scheme is a direct comparison of a single time, a single dimension, and a single threshold—obtaining a layer parameter, calculating the difference with a preset value, and judging whether it passes based on a single threshold. Although existing technologies involve "extracting coupling features that are significantly related to layer offset response and constructing stage coupling state fingerprints," the purpose of extracting these coupling features is to predict layer offset and generate compensation instructions, rather than to identify the core board layer identity; and its matching method is end-to-end prediction of a neural network model, which does not involve multi-level threshold judgment and hierarchical response mechanisms.

[0052] This invention constructs a four-layer tiered matching architecture: "overall similarity coarse screening - category-by-category feature refinement verification - three-threshold hierarchical judgment - final confirmation". The first layer, overall similarity coarse screening, calculates and sorts the overall similarity between the real-time feature parameter group and each layer in the baseline library, and quickly eliminates obviously mismatched layers based on a first preset threshold. The second layer, category-by-category feature refinement verification, compares the four categories of feature parameters of the candidate layers one by one and calculates the comprehensive confidence score. The third layer, three-threshold hierarchical judgment, divides the matching results into three levels according to the tiered relationship between the second and third preset thresholds: high confidence automatically passes, medium confidence triggers a second-level review and pauses the process for manual confirmation, and low confidence triggers a third-level warning and stops the process. The fourth layer, final confirmation, performs final confirmation on the layers to be confirmed before outputting the actual layer information. This multi-dimensional, tiered verification architecture is groundbreaking in the field of PCB layer identification. Existing technologies' single-dimensional, single-threshold comparison mode cannot achieve differentiated hierarchical error prevention responses, nor can it intervene through manual review when the matching uncertainty is high. Step 3, however, ensures the reliability of the output results through four progressive verification layers, while avoiding the high false alarm rate or high false negative rate caused by a single threshold "one-size-fits-all" judgment.

[0053] Step 4: Compare the actual layer information of the current core board with the layer information that should be pressed in the target stacking sequence. If the comparison is consistent, the next layer can be pressed. If the comparison is inconsistent, an alarm is triggered and the pressing process is stopped.

[0054] In step 4, the actual layer information of the current core board is compared with the layer information to be laminated in the target stacking sequence. If the comparison matches, the next layer lamination is allowed to continue. If the comparison does not match, an alarm is triggered and the lamination process is stopped. The specific steps include the following: The actual layer information of the current core board is compared with the layer information that should be pressed, and it is determined whether the actual layer information and the layer information that should be pressed are consistent, so as to generate a real-time comparison result. If the real-time comparison result shows that the actual layer information is consistent with the current layer information to be pressed, the pressing operation of the next core board is allowed to continue, and the current layer pressing completion information is recorded to obtain the current layer pressing completion record; If the real-time comparison results show that the actual layer information is inconsistent with the layer information that should be pressed, an alarm signal is triggered and the pressing process of the current core board is stopped. At the same time, the error information of the current layer is recorded, the current layer error record is obtained, and the current layer pressing completion record or the current layer error record is stored in the production log.

[0055] In this embodiment of the invention, taking the actual lamination production of the aforementioned eight-layer high-multilayer board as an example, assuming that the layer-by-layer lamination operation of this product is currently underway, and the target stacking order (i.e., the correct stacking order of each core board layer from top to bottom as required by the product design) is L1→L2→L3→L4→L5→L6→L7→L8, the fifth layer is currently being laminated, that is, the layer that should be laminated is layer L5. Through the multi-dimensional pattern matching recognition in step 3, the actual layer information of the current core board has been identified as layer L5. At this time, it is necessary to compare the actual layer information L5 of the current core board with the layer information L5 that should be laminated.

[0056] First, obtain the actual layer information of the current core board (i.e., layer L5 output in step 3) and the layer information to be laminated in the target stacking sequence (i.e., layer L5 corresponding to the fifth layer), thus obtaining the actual layer information L5 of the current core board and the layer information to be laminated L5. Then, compare the two to determine whether the actual layer information is consistent with the layer information to be laminated—if L5 is the same, a "consistent" real-time comparison result is generated.

[0057] If the real-time comparison results are consistent (for example, L5 and L5 are consistent in this embodiment), it indicates that the current core board layer is correct. The system allows the pressing operation of the next core board (i.e., L6 layer) to continue, and automatically records the pressing completion information of the current layer (L5 layer) - including the actual layer information L5, pressing completion time, key process parameters during the pressing process (such as maximum temperature, maximum pressure, holding time, etc.) and the real-time thermo-mechanical synergistic response characteristic parameter group of the layer extracted in step 2 - to obtain the pressing completion record of the current layer (L5 layer) for subsequent quality traceability and final integrity verification in step 5.

[0058] If the real-time comparison results are inconsistent—for example, assuming the actual layer information of the current core board is identified as L4, while the layer information to be pressed is L5, the two are inconsistent—it indicates that the current core board layer being pressed is incorrect (this could be due to the operator taking the wrong core board from the material rack, or a deviation in the identification in step 3). In this case, the system immediately triggers an alarm signal (such as an audible and visual alarm sounding, a red warning window popping up on the operating interface) and automatically stops the current core board pressing process (e.g., by controlling the pressing equipment to stop urgently, depressurize, and open the pressing machine via PLC) to prevent the incorrectly layered core board from being pressed further, thus scrapping the entire batch of products. Simultaneously, the system automatically records the current layer error information—including the actual layer information L4, the layer information to be pressed L5, the specific differences between the two, the timestamp of the error occurrence, and relevant data collected during the pressing process—to obtain the current layer error record.

[0059] Finally, regardless of whether the comparison results are consistent or inconsistent, the system stores either the current layer lamination completion record or the current layer error record in the production log. If the comparison is consistent, the system stores the L5 layer lamination completion record—recording the fact that the L5 layer has been correctly laminated and related process data; if the comparison is inconsistent, the system stores the current layer error record—recording the abnormal event of the L4 layer being incorrectly placed in the L5 layer position. This production log data can be used for subsequent quality traceability and process analysis (e.g., analyzing whether specific layers are more prone to misplacement, whether the error rate is higher in specific time periods, etc.), and can also be retrieved for final integrity verification in subsequent steps.

[0060] In existing technologies, the "multilayer circuit board hot-pressing monitoring method" involves sending a "hot-pressing prohibition warning signal" based on the amount of hot-pressed mixed layers. However, the transmission of this warning signal is based on the calculation of the difference between a single layer parameter and a preset parameter—a simple binary judgment mode of "single comparison + immediate alarm." This solution only performs two actions: judgment and alarm, without involving status recording, log storage, or data linkage with upstream and downstream processes. Other existing solutions, such as the "circuit board fusion sleeve monitoring method," also only involve obtaining the sleeve status parameters, performing sequence processing with preset parameters to obtain the difference component, and then sending a signal. None of these solutions form a complete "judgment-response-recording" closed loop.

[0061] In contrast, this embodiment generates detailed records after the comparison is completed, regardless of whether the results are consistent or inconsistent. When consistent, a completion record is generated containing actual layer information, pressing completion time, key process parameters, and real-time characteristic parameter groups. When inconsistent, an error record is generated containing actual layer information, the layer to be pressed, difference details, error timestamp, and related data, and these records are stored in the production log. This integrated processing mode of "judgment + status recording + log storage" ensures that the pressing status of each layer is traceable, providing a complete data foundation for subsequent processing. Simultaneously, the accumulation of production logs provides data support for subsequent quality traceability and process analysis (such as analyzing the frequency of misplacement of specific layers and identifying systemic process deviations)—something that the existing binary mode of "single comparison + immediate alarm" cannot achieve. Furthermore, this embodiment achieves a three-way linkage of "alarm + stop + record" when the comparison is inconsistent; the alarm reminds operators, the stop prevents further loss, and the record preserves the error scene—forming a complete error response closed loop, effectively preventing the incorrect layer from being pressed further and causing the entire batch of products to be scrapped.

[0062] Step 5: Summarize the actual layer information of all core boards to generate actual stacking records, and perform a final verification between the actual stacking records and the target stacking order to obtain the stacking integrity verification result.

[0063] In step 5, the actual layer information of all core boards is summarized to generate an actual stack-up record, and the actual stack-up record is finally verified against the target stack-up order to obtain the stack-up integrity verification result. The specific steps include the following: Summarize the actual layer information of all core boards in the current product, arrange them according to the actual pressing order of each core board, and generate an actual stacking record; The actual stacking record is compared with the target stacking order layer by layer to determine whether the actual layer information of each layer is consistent with the layer information of the corresponding layer in the target stacking order, so as to obtain the layer order comparison result. If the layer order comparison results show that the actual layer information of each layer is consistent with the layer information of each corresponding layer in the target stacking order, the standard thermo-mechanical co-response characteristic parameter set corresponding to each layer in the actual stacking record is retrieved from the mapping baseline library, and compared with the real-time thermo-mechanical co-response characteristic parameter set actually collected during the pressing process of each layer, and then the characteristic deviation value of each layer is calculated to obtain the characteristic backtracking comparison result of each layer; If the layer order comparison result shows that the actual layer information of each layer is inconsistent with the layer information of the corresponding layers in the target stacking order, the final verification alarm is triggered and the process is suspended, pending manual confirmation; if the inconsistency is confirmed by manual confirmation, the stacking integrity verification is directly determined to fail, and a stacking integrity verification result containing erroneous layer information is generated. The overall feature matching degree is calculated based on the feature backtracking comparison results of each layer. If the overall feature matching degree is greater than or equal to the preset matching degree threshold, the layer integrity verification is deemed to have passed, and a layer integrity verification result that has passed the verification is generated. If the overall feature matching degree is less than the preset matching degree threshold, the layer integrity verification is deemed to have failed, and a layer integrity verification result containing feature deviation layer information is generated.

[0064] In this embodiment of the invention, taking the actual lamination production of the aforementioned eight-layer high-multilayer board as an example, it is assumed that the target stacking sequence of the product is L1→L2→L3→L4→L5→L6→L7→L8. In the actual production process, through the layer-by-layer identification and comparison control in steps 3 and 4, each core board layer is sequentially laminated. For example, layer L1 is identified as L1 and the comparison is consistent, layer L2 is identified as L2 and the comparison is consistent, layer L3 is identified as L3 and the comparison is consistent, layer L4 is identified as L4 and the comparison is consistent, layer L5 is identified as L5 and the comparison is consistent, layer L6 is identified as L6 and the comparison is consistent, layer L7 is identified as L7 and the comparison is consistent, and layer L8 is identified as L8 and the comparison is consistent—all eight layers have been laminated and the layer-by-layer comparison has passed. At this point, the system summarizes the actual layer information of all eight core boards (i.e., L1, L2, L3, L4, L5, L6, L7, L8), arranges them according to the actual pressing order of each core board, and generates the actual stacking record [L1, L2, L3, L4, L5, L6, L7, L8].

[0065] The actual stacking records [L1, L2, L3, L4, L5, L6, L7, L8] are then compared layer by layer with the target stacking order [L1, L2, L3, L4, L5, L6, L7, L8]—the first layer L1 to L1, the second layer L2 to L2, and so on, up to the eighth layer L8 to L8—to determine whether the actual layer information of each layer is consistent with the layer information of the corresponding layers in the target stacking order, thus obtaining the layer order comparison result. In this embodiment, all layers are consistent, therefore the layer order comparison result shows that all are consistent.

[0066] When the layer order comparison results are all consistent, the system retrieves the standard thermo-mechanical co-response characteristic parameter sets corresponding to each layer (L1 to L8) in the actual stacking record from the mapping baseline library established in step 1—that is, the standard characteristic parameter set of layer L1, the standard characteristic parameter set of layer L2, ..., the standard characteristic parameter set of layer L8—and performs a layer-by-layer backtracking comparison with the real-time thermo-mechanical co-response characteristic parameter sets actually collected during the pressing process. For example, the standard characteristic parameter set of layer L1 is compared with the real-time characteristic parameter set of layer L1 collected during pressing to calculate the characteristic deviation value of layer L1; the standard characteristic parameter set of layer L2 is compared with the real-time characteristic parameter set of layer L2 to calculate the characteristic deviation value of layer L2, and so on until layer L8, to obtain the characteristic deviation values ​​of each layer from L1 to L8, that is, the characteristic backtracking comparison results of each layer. These characteristic deviation values ​​reflect the degree of agreement between the actual pressing behavior of each core board layer and the standard "physical fingerprint" of that layer. The smaller the deviation value, the closer the pressing quality of that core board layer is to the standard state; the larger the deviation value, the more likely there is an abnormality.

[0067] Assuming that in the above layer-by-layer feature backtracking comparison, the feature deviation values ​​of each layer are relatively small, for example, the overall deviation value of layer L1 is 2%, layer L2 is 1.5%, layer L3 is 3%, layer L4 is 2.5%, layer L5 is 1.8%, layer L6 is 2.2%, layer L7 is 1.2%, and layer L8 is 2.8%—the overall feature matching degree is calculated based on these feature backtracking comparison results of each layer. For example, the average of the single-layer matching degrees of each layer (such as 98%, 98.5%, 97%, 97.5%, 98.2%, 97.8%, 98.8%, 97.2%) yields an overall feature matching degree of 97.9%. If the preset matching degree threshold is 95%, since 97.9% ≥ 95%, the stack integrity verification is deemed to have passed, a stack integrity verification result of passing verification is generated, and the product is allowed to proceed to the next process.

[0068] If, in a certain production run, although the layer order comparison results are all consistent (i.e., each layer is placed correctly), but a significant deviation value is found in the layer-by-layer feature backtracking comparison of a certain layer—for example, layer L5—(e.g., the single-layer conformity of layer L5 is only 82%, far lower than the above 95% of other layers), and the calculated overall feature conformity is 89.5%, which is less than the preset conformity threshold of 95%, then the stack integrity verification is determined to be unsuccessful. A stack integrity verification result containing the feature deviation layer information (i.e., layer L5) is generated, prompting the operator that although the layer order of the product is correct, the lamination quality of layer L5 is abnormal (possibly due to differences in the batch of core board materials for this layer, or local delamination of this layer during the lamination process, etc.), and further inspection and confirmation are required before proceeding to the next process.

[0069] If the layer order comparison results are inconsistent—for example, the actual layer record is [L1, L2, L3, L5, L4, L6, L7, L8] while the target layer order is [L1, L2, L3, L4, L5, L6, L7, L8], meaning the order of the fourth and fifth layers has been swapped—then there is no need to perform feature backtracking comparison. The layer integrity check is directly determined to have failed, and a layer integrity check result containing incorrect layer information (such as "the order of L4 and L5 layers is incorrect") is generated. The product is then marked as a defective product to prevent it from flowing into subsequent processes.

[0070] Existing final verification schemes generally adopt a path of "external marking - external reading - simple judgment". For example, "a circuit board for rapid detection and prevention of incorrect lamination layers" uses a first mark on the edge of the large panel process and a second mark on the edge of the delivered board process. After lamination, the mark lengths are arranged in ascending or descending order, and visual identification is performed using the copper foil with equal gradients on the board edge. "A circuit board structure for visually distinguishing board layers" sets a cutout area on each circuit board layer and marks it with a numerical layer number. After lamination, it is identified by the naked eye or visual scanning equipment. "A method for manufacturing PCB boards and preventing incorrect layer order from flowing into the milling process" designs exposed copper areas on the inner layer film. After lamination and milling, the exposed copper areas are exposed on the side of the PCB board, forming a diagonal line. The correctness of the layer order is judged by checking whether the connection of the exposed copper areas on the side is a diagonal line. In addition, "a method for preventing errors in multilayer board lamination" obtains the weight information of the circuit board and the preset lamination order by scanning the graphic code, and performs detection in combination with changes in vehicle load weight. These schemes are all single final verifications - the verification ends when the layer order is correct, and the verification basis is all external markings. Although the "multilayer circuit board hot-pressing monitoring method" involves layer monitoring, it obtains "hot-melt pressing layer parameters" (layer model or layer identification area image), and sends an early warning signal after anti-mixing processing with preset layer parameters. It is also a single comparison and does not involve retrospective verification of feature layers.

[0071] This invention employs a dual final verification architecture: "layer sequence verification + feature parameter backtracking verification." The first layer verifies whether the actual layer stacking record matches the target layer stacking order—addressing issues such as incorrect, missing, or excessive layer placement. The second layer retrieves the standard thermo-mechanical co-response feature parameter sets corresponding to each layer from the mapping baseline library and performs a layer-by-layer backtracking comparison with the real-time thermo-mechanical co-response feature parameter sets actually collected during the pressing process, calculating the feature deviation value for each layer—addressing the issue of abnormal pressing quality in a particular layer even if the layer order is correct. This dual verification of "sequence + feature" ensures that any errors that might be missed during the layer-by-layer verification stage—including incorrect layer order, abnormal single-layer feature parameters, and physical displacement of the core board—can be fully intercepted after pressing and before proceeding to the next process.

[0072] In a preferred embodiment of the present invention, the standard thermo-mechanical co-response characteristic parameter set corresponding to each layer in the actual stacking record is retrieved from the mapping baseline library, and compared layer by layer with the real-time thermo-mechanical co-response characteristic parameter set actually collected during the pressing process of each layer. Then, the characteristic deviation value of each layer is calculated to obtain the characteristic backtracking comparison result of each layer. The specific steps include the following: Obtain the standard thermo-mechanical co-response characteristic parameter set corresponding to each layer in the actual stacked record, and obtain the standard characteristic parameter set set of each layer; The real-time thermo-mechanical co-response characteristic parameter set of each layer is obtained by acquiring the actual collected real-time characteristic parameter set of each layer during the pressing process; The standard characteristic parameter set of the same layer is associated with each parameter in the real-time characteristic parameter set, and the temperature independent response characteristic deviation value, pressure independent response characteristic deviation value, cooperative response characteristic deviation value and waveform morphology characteristic deviation value of each layer are calculated to obtain the four types of characteristic deviation values ​​of each layer. The single-layer feature matching degree of each layer is calculated based on the four types of feature deviation values ​​of each layer. If the single-layer feature matching degree of any layer is less than the preset single-layer matching degree threshold, the layer corresponding to the single-layer matching degree threshold is marked as the feature deviation layer. The number of layers marked as feature deviations is counted, and the overall feature fit is calculated based on the ratio of the number of feature deviation layers to the total number of layers in the actual stacked records. The single-layer feature matching degree and the overall feature matching degree of each layer are used as the feature backtracking comparison results of each layer.

[0073] In this embodiment of the invention, taking the actual lamination production of the aforementioned eight-layer high-multilayer board as an example, assuming that the layer order comparison results are all consistent, the system enters the feature backtracking comparison stage. First, the standard thermo-mechanical synergistic response characteristic parameter sets corresponding to each layer in the actual stacking record [L1, L2, L3, L4, L5, L6, L7, L8] are obtained—that is, the standard characteristic parameter sets of layer L1, L2, L3, L4, L5, L6, L7, and L8 are retrieved from the mapping baseline library respectively—to obtain the standard characteristic parameter set set of each layer. At the same time, the real-time thermo-mechanical synergistic response characteristic parameter sets actually collected during the lamination process of each layer are obtained—that is, the real-time characteristic parameter sets of layer L1, L2, ..., L8 output layer by layer—to obtain the real-time characteristic parameter set set of each layer.

[0074] Then, the standard characteristic parameter groups of the same layer are correlated with the parameters in the real-time characteristic parameter groups. For example, the standard characteristic parameter group of layer L5 (standard heating rate 2.4℃ / min, standard peak temperature arrival time 48min, standard pressure build-up rate 11.5psi / min, standard peak pressure arrival time 52min, standard start-up time difference 4.8min, standard rate ratio 0.21℃ / psi, standard peak time difference 4min, standard waveform symmetry 0.93, standard waveform area 1860℃·psi) is correlated with the real-time characteristic parameter group of layer L5 (real-time heating rate 2.5℃ / min, real-time peak temperature arrival time 47min, real-time pressure build-up rate 11psi / min, real-time peak pressure arrival time 52min, real-time start-up time difference 5min, real-time rate ratio 0.21℃ / psi, real-time peak time difference 5min, real-time waveform symmetry 0.92, real-time waveform area 1850℃·psi) is further correlated. One-to-one correspondence is established—and four types of characteristic deviation values ​​are calculated for each layer: temperature independent response characteristic deviation value (for example, the temperature rise rate deviation of layer L5 is |2.5-2.4| / 2.4≈4.2%, and the temperature peak arrival time deviation is |47-48| / 48≈2.1%), pressure independent response characteristic deviation value (pressure build-up rate deviation is |11-11.5| / 11.5≈4.3%, and pressure peak arrival time deviation is |52-52| / 52=0%), and cooperative response. The characteristic deviation values ​​(start-up time difference deviation: |5-4.8| / 4.8≈4.2%, rate ratio deviation: |0.21-0.21| / 0.21=0%, peak time difference deviation: |5-4| / 4=25%) and waveform morphology characteristic deviation values ​​(waveform symmetry deviation: |0.92-0.93| / 0.93≈1.1%, waveform area deviation: |1850-1860| / 1860≈0.5%) are summarized to obtain the four types of characteristic deviation values ​​for each layer of L5. Similarly, the four types of characteristic deviation values ​​for each layer from L1 to L8 are calculated.

[0075] After obtaining the four types of characteristic deviation values ​​for each layer, the single-layer characteristic matching degree for each layer is calculated based on the four types of characteristic deviation values ​​for each layer. For example, the temperature independent response characteristic matching degree for layer L5 is (1-4.2%)×(1-2.1%)≈93.8%, the pressure independent response characteristic matching degree is (1-4.3%)×(1-0%)≈95.7%, the cooperative response characteristic matching degree is (1-4.2%)×(1-0%)×(1-25%)≈71.9%, and the waveform morphology characteristic matching degree is (1-1.1%)×(1-0.5%)≈98.4%. The average of the four types of matching degrees is taken to obtain the single-layer characteristic matching degree of layer L5, which is approximately 89.9%. If the preset single-layer matching threshold is 90%, since the single-layer matching degree of L5 layer is 89.9%, which is less than the preset single-layer matching threshold of 90%, L5 layer is marked as a feature deviation layer, indicating that although the layer order is correct, its collaborative response characteristics (especially the peak time difference) have a large deviation from the standard value.

[0076] The number of layers marked as feature deviations is counted—for example, in this embodiment only layer L5 is marked, with a feature deviation layer count of 1—and the overall feature fit is calculated based on the ratio of the number of feature deviation layers to the total number of layers (8 layers) in the actual stacked record. For example, the overall feature fit can be calculated by the average of the fit of each individual layer, or by (total number of layers - number of feature deviation layers) / total number of layers × 100%—for example, (8-1) / 8 × 100% = 87.5%. Finally, the single-layer feature matching degree of each layer (L1 layer 98.5%, L2 layer 97.8%, L3 layer 96.5%, L4 layer 97.2%, L5 layer 89.9%, L6 layer 97.5%, L7 layer 98.2%, L8 layer 96.8%) and the overall feature matching degree of 87.5% are output together as the feature backtracking comparison results of each layer, which are used in the previous stage of step 5 to compare and determine the overall feature matching degree with the preset matching degree threshold.

[0077] Existing technologies all employ single-step verification schemes—passing verification results in release, while failing verification results in scrapping, without involving independent feature deviation analysis for each layer. The "Multilayer Circuit Board Hot-Pressure Monitoring Method" obtains the hot-pressing layer mixing amount by comparing the hot-melt pressing layer parameters of the board with preset layer parameters for anti-mixing processing—this mixing amount is a single overall difference, unable to pinpoint which specific layer has a problem or distinguish which type of feature deviation occurred. The "Circuit Board Fusion Kit Monitoring Method" obtains the fusion kit status parameters of the core board with preset kit status parameters for order processing to obtain a kit order difference component—again, only outputting a single overall difference component. Both of these schemes are single-step, overall, and single-dimensional comparisons, unable to pinpoint the specific problematic layer or specific problem feature type. The "A Foolproof Design for Multilayer Circuit Board Layer Appearance Inspection Based on Capacitive Sensors" identifies layers by detecting the position of copper blocks after pressing using capacitive sensors; this is also a single-step overall inspection, without involving layer-by-layer fine-grained comparison of feature parameters for each layer.

[0078] In contrast, the feature backtracking comparison scheme in this embodiment implements a refined feature backtracking verification mechanism that performs layer-by-layer, independent calculation of four types of features, and dual evaluation of single-layer and overall consistency. This scheme does not simply output an overall pass / fail conclusion, but rather calculates the deviation values ​​of the four types of features (temperature independent response features, pressure independent response features, coordinated response features, and waveform morphology features) for each layer in the actual stacked record. It then calculates the single-layer feature consistency for each layer, precisely locates the feature deviation layers based on the comparison results between each layer's consistency and a preset single-layer consistency threshold, and finally calculates the overall feature consistency by counting the number of feature deviation layers. This layer-by-layer refined feature backtracking verification enables operators to accurately identify which layer has an anomaly and in which type of feature parameter the anomaly occurs. For example, in the above embodiment, the L5 layer has the largest collaborative response feature deviation value (peak time difference deviation reaches 25%). Based on this, operators can focus on checking whether the specifications of the L5 layer's prepreg are correct, whether the temperature-pressure timing of this layer is abnormal during the pressing process, etc., providing clear directional guidance for subsequent cause investigation and process improvement. This is a technical effect that cannot be achieved by a single overall difference in the existing technology.

[0079] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0080] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0081] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for preventing errors in the stacking sequence of core boards based on visual recognition, characterized in that, The method includes the following steps: Step 1: Perform standard pressing procedures on the core boards of each layer. Simultaneously collect temperature dynamic response data and pressure dynamic response data of each core board during the entire pressing process through temperature and pressure sensors of the pressing equipment. Extract the thermal-mechanical synergistic response characteristic parameter group of each core board from the collected data to establish a mapping baseline library between each layer and its respective thermal-mechanical synergistic response characteristic parameter group. Step 2: During the actual layer-by-layer lamination process, the temperature sensor and pressure sensor synchronously collect the real-time temperature dynamic response data and real-time pressure dynamic response data of the current core board, and extract the real-time thermo-mechanical co-response characteristic parameter group of the current core board from the real-time temperature dynamic response data and real-time pressure dynamic response data. Step 3: Perform multi-dimensional pattern matching between the real-time thermal-mechanical co-response characteristic parameter group and the thermal-mechanical co-response characteristic parameter groups of each layer in the mapping baseline library to generate matching results, and identify the actual layer information of the current core board based on the matching results; Step 4: Compare the actual layer information of the current core board with the layer information that should be pressed in the target stacking sequence. If the comparison is consistent, the next layer can be pressed. If the comparison is inconsistent, an alarm is triggered and the pressing process is stopped. Step 5: Summarize the actual layer information of all core boards to generate actual stacking records, and perform a final verification between the actual stacking records and the target stacking order to obtain the stacking integrity verification result.

2. The method for preventing errors in the core board stacking sequence based on visual recognition according to claim 1, characterized in that, In step 1, a standard pressing procedure is performed on each core board layer. Temperature and pressure dynamic response data of each core board are synchronously collected through temperature and pressure sensors of the pressing equipment throughout the pressing process. The thermo-mechanical co-response characteristic parameter set of each core board is extracted from the collected data to establish a mapping baseline library between each layer and its respective thermo-mechanical co-response characteristic parameter set. The specific steps include the following: S101. Obtain the core boards of each layer to obtain an unmarked set of core boards; S102. Randomly select a core board of a certain layer from the unmarked core board set and place it in the pressing equipment. Execute the standard pressing procedure so that the current core board sequentially goes through the preheating stage, resin flow stage, gel stage, curing stage, and cooling stage during the entire pressing process. At each stage, corresponding temperature dynamic response and pressure dynamic response are generated. The temperature dynamic response data and pressure dynamic response data of the current core board are synchronously collected by the temperature sensor and pressure sensor of the pressing equipment to obtain the temperature-pressure synchronous response data set of the current core board. S103. Extract the thermal-mechanical coordinated response characteristic parameter group of the current core board from the temperature-pressure synchronous response data group of the current core board, and store the thermal-mechanical coordinated response characteristic parameter group of the current core board in correspondence with the layer information of the current core board; wherein, the thermal-mechanical coordinated response characteristic parameter group includes temperature independent response characteristic parameters, pressure independent response characteristic parameters, and coordinated response characteristic parameters between temperature and pressure; S104. Repeat steps S101 to S103 for the other layers of core boards in the unmarked core board set that have not been calibrated, until all layers of core boards have been stored in order to establish a mapping baseline library between each layer and its respective thermal-mechanical co-response characteristic parameter group.

3. The core board stacking sequence error prevention method based on visual recognition according to claim 2, characterized in that, Extract the thermal-mechanical coordinated response characteristic parameter set of the current core board from the current temperature-pressure synchronous response data set, specifically including the following steps: From the current core board's temperature-pressure synchronous response data set, extract the heating rate and temperature peak arrival time to obtain a set of temperature-independent response characteristic parameters; From the current temperature-pressure synchronous response data set of the core board, extract the pressure build-up rate and pressure peak arrival time to obtain the pressure independent response characteristic parameter set; From the current core board's temperature-pressure synchronous response data set, extract the coordinated response characteristic parameters between the temperature dynamic response and the pressure dynamic response to obtain a temperature-pressure coordinated response characteristic parameter set; wherein, the coordinated response characteristic parameters include the start-up time difference between the heating start-up time and the pressurization start-up time, the ratio of the temperature change rate to the pressure change rate, and the arrival time difference between the temperature peak and the pressure peak. Using temperature data as the abscissa and pressure data as the ordinate in the temperature-pressure synchronous response data set, a temperature-pressure coupled response curve is plotted. The waveform symmetry and waveform area of ​​the temperature-pressure coupled response curve are extracted to obtain a set of waveform morphology characteristic parameters of the coupled response curve. The temperature-independent response characteristic parameter group, the pressure-independent response characteristic parameter group, the temperature-pressure co-response characteristic parameter group, and the coupled response curve waveform morphology characteristic parameter group are merged to obtain the current core board's thermo-mechanical co-response characteristic parameter group.

4. The core board stacking sequence error prevention method based on visual recognition according to claim 3, characterized in that, In step 2, during the actual layer-by-layer lamination process, the temperature sensor and pressure sensor simultaneously collect real-time temperature dynamic response data and real-time pressure dynamic response data of the current core board, and extract the real-time thermo-mechanical co-response characteristic parameter set of the current core board from the real-time temperature dynamic response data and real-time pressure dynamic response data, specifically including the following steps: In the actual layer-by-layer lamination process, temperature and pressure dynamic response data of the current core plate are collected in the preheating and early flow stages using temperature and pressure sensors to obtain early temperature-pressure response data sets. Early thermo-mechanical co-response characteristic parameter set is extracted from the early temperature-pressure response data set to obtain the early thermo-mechanical co-response characteristic parameter set; wherein, the early thermo-mechanical co-response characteristic parameter set includes the heating rate in the preheating stage, the pressure build-up rate in the early flow stage, and the start-up time difference between heating start-up and pressurization start-up; Using the aforementioned mapping baseline library as a mapping benchmark, the early thermo-mechanical co-response characteristic parameter set is mapped to the current core plate's predicted thermo-mechanical co-response characteristic parameter set during the pressing process; Based on the early temperature-pressure response data set, we continue to collect real-time temperature dynamic response data and real-time pressure dynamic response data of the current core plate in the remaining stages of the entire pressing process, and after pressing is completed, we extract the measured thermo-mechanical co-response characteristic parameter set of the current core plate from all the collected data. Calculate the deviation value of each characteristic parameter between the predicted thermo-mechanical coordinated response characteristic parameter set and the measured thermo-mechanical coordinated response characteristic parameter set to obtain the prediction-measurement deviation result; If the deviation values ​​of each characteristic parameter in the predicted-measured deviation result are all less than the preset tolerance threshold, the measured thermo-mechanical synergistic response characteristic parameter group will be output as the real-time thermo-mechanical synergistic response characteristic parameter group of the current core board; if the deviation value of any characteristic parameter in the predicted-measured deviation result is greater than or equal to the preset tolerance threshold, an early warning signal will be triggered and the current pressing process will be paused for manual confirmation.

5. The core board stacking sequence error prevention method based on visual recognition according to claim 4, characterized in that, In step 3, the real-time thermal-mechanical co-response characteristic parameter set is matched with the thermal-mechanical co-response characteristic parameter sets of each layer in the mapping baseline library in a multi-dimensional pattern to generate a matching result. Based on the matching result, the actual layer information of the current core board is identified. Specifically, this includes the following steps: The overall similarity is calculated between all the feature parameters contained in the real-time thermo-mechanical co-response feature parameter group and all the feature parameters contained in the thermo-mechanical co-response feature parameter group of each layer in the mapping baseline library, and the overall similarity score corresponding to each layer is obtained. Sort the overall similarity scores of each layer in descending order, select the layer with the highest overall similarity score as the candidate layer, and obtain the overall similarity score of the candidate layer. Determine whether the overall similarity score corresponding to the candidate layer is greater than or equal to the first preset threshold. If the overall similarity score corresponding to the candidate layer is greater than or equal to the first preset threshold, output the candidate layer as the target layer. If the overall similarity score corresponding to the candidate layer is less than the first preset threshold, a first-level warning signal is triggered, and the current core board pressing process is stopped. After outputting the target layer, each type of feature parameter contained in the real-time thermo-mechanical co-response feature parameter group is compared with the corresponding type of feature parameter in the mapping baseline library of the target layer, and the class deviation value of each type of feature parameter is calculated to obtain the class deviation results of each type of feature. The comprehensive confidence score is calculated based on the class deviation results of each type of feature. Determine whether the overall confidence score is greater than or equal to the second preset threshold. If the overall confidence score is greater than or equal to the second preset threshold, output the target layer as the layer to be confirmed. If the overall confidence score is greater than or equal to the third preset threshold and less than the second preset threshold, a second-level review signal is triggered, the target layer is marked as pending review, and the lamination process is paused for manual confirmation; if the overall confidence score is less than the third preset threshold, a third-level warning signal is triggered, and the lamination process of the current core board is terminated; wherein, the third preset threshold is less than the second preset threshold; After outputting the layers to be confirmed, the layers to be confirmed are finally confirmed, and the final confirmation result is generated. If the final confirmation result is passed, the layers to be confirmed are output as the actual layer information of the current core board; if the final confirmation result is failed, a level three warning signal is triggered, and the pressing process of the current core board is stopped.

6. The core board stacking sequence error prevention method based on visual recognition according to claim 5, characterized in that, In step 4, the actual layer information of the current core board is compared with the layer information to be laminated in the target stacking sequence. If the comparison matches, the next layer lamination is allowed to continue. If the comparison does not match, an alarm is triggered and the lamination process is stopped. Specifically, the steps are as follows: The actual layer information of the current core board is compared with the layer information that should be pressed, and it is determined whether the actual layer information and the layer information that should be pressed are consistent, so as to generate a real-time comparison result. If the real-time comparison result shows that the actual layer information is consistent with the current layer information to be pressed, the pressing operation of the next core board is allowed to continue, and the current layer pressing completion information is recorded to obtain the current layer pressing completion record; If the real-time comparison results show that the actual layer information is inconsistent with the layer information that should be pressed, an alarm signal is triggered and the pressing process of the current core board is stopped. At the same time, the error information of the current layer is recorded, the current layer error record is obtained, and the current layer pressing completion record or the current layer error record is stored in the production log.

7. The method for preventing errors in core board stacking sequence based on visual recognition according to claim 6, characterized in that, In step 5, the actual layer information of all core boards is summarized to generate an actual stacking record, and the actual stacking record is finally verified against the target stacking order to obtain the stacking integrity verification result. The specific steps include the following: Summarize the actual layer information of all core boards in the current product, arrange them according to the actual pressing order of each core board, and generate an actual stacking record; The actual stacking record is compared with the target stacking order layer by layer to determine whether the actual layer information of each layer is consistent with the layer information of the corresponding layer in the target stacking order, so as to obtain the layer order comparison result. If the layer order comparison results show that the actual layer information of each layer is consistent with the layer information of each corresponding layer in the target stacking order, the standard thermo-mechanical co-response characteristic parameter set corresponding to each layer in the actual stacking record is retrieved from the mapping baseline library, and compared with the real-time thermo-mechanical co-response characteristic parameter set actually collected during the pressing process of each layer, and then the characteristic deviation value of each layer is calculated to obtain the characteristic backtracking comparison result of each layer; If the layer order comparison result shows that the actual layer information of each layer is inconsistent with the layer information of the corresponding layers in the target stacking order, the final verification alarm is triggered and the process is suspended, pending manual confirmation. If inconsistencies are found upon manual verification, the stack integrity check is directly deemed to have failed, and a stack integrity check result containing information about the incorrect stack layer is generated. The overall feature matching degree is calculated based on the feature backtracking comparison results of each layer. If the overall feature matching degree is greater than or equal to the preset matching degree threshold, the layer integrity verification is deemed to have passed, and a layer integrity verification result that has passed the verification is generated. If the overall feature matching degree is less than the preset matching degree threshold, the layer integrity verification is deemed to have failed, and a layer integrity verification result containing feature deviation layer information is generated.

8. The method for preventing errors in core board stacking sequence based on visual recognition according to claim 7, characterized in that, The standard thermo-mechanical co-response characteristic parameter set corresponding to each layer in the actual stacking record is retrieved from the mapping baseline library, and compared layer by layer with the real-time thermo-mechanical co-response characteristic parameter set actually collected during the pressing process. Then, the characteristic deviation value of each layer is calculated to obtain the characteristic backtracking comparison result of each layer. The specific steps include the following: Obtain the standard thermo-mechanical co-response characteristic parameter set corresponding to each layer in the actual stacked record, and obtain the standard characteristic parameter set set of each layer; The real-time thermo-mechanical co-response characteristic parameter set of each layer is obtained by acquiring the actual collected real-time characteristic parameter set of each layer during the pressing process; The standard characteristic parameter set of the same layer is associated with each parameter in the real-time characteristic parameter set, and the temperature independent response characteristic deviation value, pressure independent response characteristic deviation value, cooperative response characteristic deviation value and waveform morphology characteristic deviation value of each layer are calculated to obtain the four types of characteristic deviation values ​​of each layer. The single-layer feature matching degree of each layer is calculated based on the four types of feature deviation values ​​of each layer. If the single-layer feature matching degree of any layer is less than the preset single-layer matching degree threshold, the layer corresponding to the single-layer matching degree threshold is marked as the feature deviation layer. The number of layers marked as feature deviations is counted, and the overall feature fit is calculated based on the ratio of the number of feature deviation layers to the total number of layers in the actual stacked records. The single-layer feature matching degree and the overall feature matching degree of each layer are used as the feature backtracking comparison results of each layer.

Images