A circuit board whole-process data traceability management method

Abstract

The present application relates to the field of data mining, and specifically discloses a circuit board whole-process data traceability management method, comprising the following steps: S1. State diagnosis step: through high-frequency acquisition of the spindle motor current bias signal sequence of a numerical control drilling machine during micro-hole drilling, and inputting the spindle motor current bias signal sequence and the current drill physical down-hole frequency into an adaptive time domain filter for smooth fundamental wave extraction, to obtain a net load current sequence after filtering out transient mechanical oscillation noise; S2. Copper deposition regulation step; S3. Reverse self-correction step. The technical scheme of the present application can predict and isolate potential broken blades and tool aging trends in the mechanical micro-ultra-deep hole drilling process to avoid mechanical oscillation interference, assist in reconstructing the uniformity of hole wall thickness, and establish a correlation between the end electrical network defects and the front physical / chemical / mechanical process disturbances.

Description

Technical Field

[0001] This invention relates to the field of data mining, and in particular to a method for tracing and controlling data throughout the entire circuit board manufacturing process. Background Technology

[0002] As electronic products evolve towards thinner, lighter, more multifunctional, and higher-frequency, higher-speed designs, the application of multilayer high-density interconnect (HDI) printed circuit boards and high-density packaging substrates in precision electronics manufacturing is becoming increasingly widespread. In the printing process of multilayer HDI circuit boards, mechanical microvia drilling, chemical copper plating of the via walls, pattern electroplating, and terminal electrical performance testing are the four core steps that determine the quality of interlayer interconnects and high-frequency impedance matching. With the rapid increase in circuit board integration, the widespread application of ultra-deep microvias with aspect ratios (AR) greater than 10:1 and apertures less than 0.15 mm, as well as high-frequency, high-speed electrical traces, places extremely stringent requirements on process data consistency and control methods.

[0003] However, traditional circuit board process control and data traceability solutions suffer from severe data chain breaks in actual production processes. Existing Manufacturing Execution Systems (MES) and quality control solutions tend to rely on "static logistics tracking," treating physical processes such as design, drilling, copper plating, electroplating, and testing as isolated islands. Testing equipment (such as flying probe microscopy and ICT) is only at the end of the production line, passively intercepting defective boards due to "open circuits / short circuits." This "post-event interception" mechanism not only fails to pinpoint the cause of faults in real time and accurately, but also cannot dynamically intervene in the process parameters of the front-end actuators. This results in significant delays or even failures in closed-loop defect regulation, hindering effective process self-healing.

[0004] Secondly, in ultra-deep micro-hole drilling, due to chip removal resistance and high frictional heat, the drill bit faces the risk of sudden breakage and rapid aging due to asymmetrical stress. Existing drill bit monitoring methods only focus on current amplitude, making it difficult to effectively filter out the mechanical vibration noise caused by abrupt chip removal. More seriously, worn drill bits not only lead to breakage themselves but also directly cause epoxy resin melting redshift (resin smear) and severe glass fiber tearing burrs on the hole wall. These deteriorations in microstructure significantly increase the local convection resistance within the hole for subsequent chemical copper plating, directly hindering the mass transfer and diffusion of the chemical solution, thereby accelerating the "dog-bone" (thick at both ends, thin in the middle, or even broken) uneven copper plating defects. This hidden "physical causal chain" is completely open-loop in actual machining operations, making cross-assisted diagnosis and proactive prevention impossible.

[0005] Furthermore, when flying probe testing detects impedance distortion or abnormal conduction resistance, due to numerous influencing factors (such as polarization potential drift, lamination thermal stress, and alignment slip), traditional analysis methods often rely on human experience, making it difficult to achieve high-confidence reverse decoupling in complex environments with multivariate coupling. This results in the inability to accurately reverse-adjust the equipment to target specific physical failure sources.

[0006] Therefore, there is an urgent need for a method that can predict and isolate potential tool breakage and tool aging trends during the mechanical micro-deep hole drilling process to avoid mechanical oscillation interference, assist in reconstructing hole wall thickness uniformity, and establish a full-process data traceability and control method for circuit boards that establishes the correlation between end-point electrical network defects and front-end physical / chemical / mechanical process disturbances. Summary of the Invention

[0007] This invention provides a method for full-process data traceability and control of circuit boards, which is capable of...

[0008] To solve the above-mentioned technical problems, this application provides the following technical solution: A method for full-process data traceability and control of circuit boards includes the following steps: S1. State Diagnosis Steps: The spindle motor current bias signal sequence of the CNC drilling machine during micro-hole drilling is acquired via high-frequency acquisition. This sequence, along with the current number of physical drilling operations, is input into an adaptive time-domain filter for smoothing and fundamental wave extraction, resulting in a net load current sequence after filtering out instantaneous mechanical oscillation noise. Based on this net load current sequence and a pre-calculated baseline health resistance current constant, the current deviation index variable resistance factor is calculated, and the first-order slope abrupt gradient of the net load current sequence is calculated. A time-division causal state determination logic is used to match and determine the first-order slope abrupt gradient and the current deviation index variable resistance factor, outputting the drill bit's working state. A corresponding first control command is then issued to the CNC drilling machine based on the drill bit's working state. S2. Copper Plating Control Steps: The cumulative alignment deviation of the optical target in the multilayer panel under mechanical alignment is read. A spatial alignment differential degradation operator is calculated based on the cumulative alignment deviation, and the spatial alignment differential degradation operator is used to self-correct the drilling coordinate axis feed command of the CNC drilling machine. The multilayer panel after correction drilling is delivered to the chemical copper plating tank. Based on the chemical reaction-physical liquid phase combined transport dynamic boundary model in high aspect ratio holes, the instantaneous deposition rate of the through-holes of the multilayer panel at different axial depths is calculated. The excitation force turbulence boundary gain and the forced jet circulation velocity gain are used as denominator correction constants applied to the exponential decay term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate. When the non-uniformity index is lower than a preset non-uniformity threshold, the corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate in the chemical copper plating tank. S3. Reverse Self-Correction Step: During the finished product electrical performance testing stage, the electrical deviation vector of the tested circuit is obtained through electrical testing equipment, and the electrical deviation vector is converted into a dimensionless normalized electrical deviation vector; a physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process disturbance relative change vector; based on the physical-electrical coupling transcription matrix, Tikhonov inverse matrix mapping analysis is performed on the normalized electrical deviation vector to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability; according to the dominant strength of each dimension of the physical component in the dominant feature relative vector of physical failure probability, the corresponding online physical equipment correction and reverse adjustment control behavior is automatically generated and executed.

[0009] The basic scheme principle and beneficial effects are as follows: In S1, the spindle motor current bias signal sequence of the CNC drilling machine during micro-hole drilling is acquired by high frequency, and the spindle motor current bias signal sequence and the current number of physical holes drilled by the drill bit are input into an adaptive time-domain filter for smoothing fundamental wave extraction to obtain the net load current sequence after filtering out instantaneous mechanical oscillation noise; based on the net load current sequence and the pre-calculated benchmark health resistance current constant, the current deviation index variable resistance factor is calculated, and the first-order slope abrupt gradient of the net load current sequence is calculated; the first-order slope abrupt gradient and the current deviation index variable resistance factor are matched and judged using time-division causal state determination logic, the drill bit working state is output, and the corresponding first control command is issued to the CNC drilling machine according to the drill bit working state.

[0010] The spindle motor current bias signal can reflect the changes in load resistance during drilling in real time, but it is highly susceptible to interference from instantaneous mechanical oscillation noise such as sudden chip removal. By introducing the current number of physical drilling operations (characterizing the tool wear stage) and combining it with an adaptive time-domain filter, the filtering characteristics can be dynamically adjusted according to the tool state. This effectively filters out high-frequency transient noise while preserving the fundamental characteristics, extracting the true net load current. Based on this, by calculating the current deviation index variable resistance factor (reflecting the static deviation degree) and the first-order slope abrupt change gradient (reflecting the dynamic abrupt change rate), and using time-division causal state judgment logic for multi-dimensional matching, different physical failure modes such as blade breakage (instantaneous weight loss, sudden slope drop) and wear aging (gradual slope, cumulative deviation) can be accurately isolated and diagnosed in the time domain.

[0011] It solves the problem of difficult-to-predict chip breakage and aging caused by chip removal resistance and high frictional heat in ultra-deep microhole drilling, eliminates the interference of high-frequency mechanical collisions and chip removal noise, realizes online high-precision diagnosis and real-time control of drill bit working status, and avoids the scrapping of circuit boards due to broken needles and the deterioration of hole wall micromorphology due to blunt cutting.

[0012] In S2, the cumulative alignment deviation of the optical target under mechanical alignment of the multilayer panel is read. Based on the cumulative alignment deviation, a spatial alignment differential degradation operator is calculated, and the spatial alignment differential degradation operator is used to self-correct the drilling coordinate axis feed command of the CNC drilling machine. The multilayer panel after correction drilling is delivered to the chemical copper plating tank. Based on the chemical reaction-physical liquid phase combined transport dynamic boundary model in the high aspect ratio hole, the instantaneous deposition rate of the through-hole of the multilayer panel at different axial depths is calculated. The excitation force turbulent boundary gain and the forced jet circulation velocity gain are used as denominator correction constants to apply to the exponential decay term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate. When the non-uniformity index is lower than the preset non-uniformity threshold, the corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate in the chemical copper plating tank.

[0013] Multi-layer panels exhibit cumulative deviations during mechanical alignment. By computationally calculating a spatial alignment differential degradation operator and performing self-correction on feed commands, hole ring offset caused by shearing can be suppressed at its source, preventing internal hole wall cracking. During the chemical copper plating stage, mass transfer of the chemical solution within high aspect ratio holes is limited, resulting in significant spatial diffusion attenuation. This solution constructs a chemical reaction-physical liquid phase combined transport dynamic boundary model to quantitatively describe the instantaneous deposition rate at different axial depths. It creatively incorporates the turbulent boundary gain from external excitation forces (such as mechanical or ultrasonic vibration) and the forced jet circulation velocity gain as correction constants, applying them to the denominator of the exponential decay term of the spatial diffusion attenuation operator. This enhances local convection and reduces laminar boundary layer resistance, adaptively generating a compensation rate at hole depth, mitigating the exponential decay trend with increasing depth. By calculating the non-uniformity index in real time and comparing it with a threshold, when the non-uniformity index is below the threshold, the uniform reconstruction of the pore wall thickness is achieved by adjusting the mechanical convection scale (enhancing drug mass transfer) and the chemical reaction rate (such as reducing the reactivity at the orifice by cooling and providing a molecular concentration recovery cycle for the depth of the pore).

[0014] It solves the quality defects of "thick at both ends and thin in the middle" (dog bone-like) or hollow copper deposition thickness in ultra-fine high aspect ratio through-holes, and achieves high uniformity control of hole wall deposition thickness, ensuring the continuity and uniformity of copper deposition on the surface of the via structure.

[0015] In S3, during the finished product electrical performance testing phase, the electrical deviation vector of the tested circuit is acquired through electrical testing equipment and converted into a dimensionless normalized electrical deviation vector. A physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process disturbance relative change vector. Based on the physical-electrical coupling transcription matrix, a Tikhonov inverse matrix mapping analysis is performed on the normalized electrical deviation vector to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability. According to the dominant strength of each dimension of the physical component in the dominant feature relative vector of physical failure probability, the corresponding online physical equipment correction and anti-adjustment control behavior is automatically generated and executed.

[0016] Electrical deviations in the finished product stage (such as impedance, inductance, and capacitance deviations) are the result of the coupling effects of physical disturbances (such as polarization potential drift, thermal stress, and alignment slip) from multiple upstream processes. By constructing a physical-electrical coupling transcription matrix, a linear mapping relationship is established between the dimensionless normalized electrical deviation vector and the normalized relative change vector of process disturbances, transforming the complex physical-electrical coupling process into mathematical matrix equations. Since the inverse solution process is usually ill-posed (with multiple solutions, degradation, or non-convergence), this scheme introduces Tikhonov inverse matrix mapping analysis. By introducing a regularization term, the inverse solution process achieves numerical convergence and does not suffer from dimensional degradation, thereby quantitatively decoupling the dominant characteristic relative vector of physical failure probability. Based on the dominant strength of each decoupled physical component, the underlying physical failure source causing electrical performance distortion can be accurately located, and the corresponding online physical equipment reverse control behavior can be automatically generated, achieving closed-loop self-healing of process parameters.

[0017] It solves the problem of unclear causal relationship between electrical network defects and front-end physical / chemical / mechanical process disturbances in the whole process, making it difficult to accurately locate and control them in a closed loop. It achieves high-confidence reverse decoupling from end-of-line electrical network defects to front-end process disturbances and precise reverse adjustment and correction of equipment, significantly reducing the product impedance consistency tolerance.

[0018] By employing adaptive time-domain filtering and time-division causal state determination in the state diagnosis step, the control system can respond to sudden faults such as broken blades within milliseconds and issue self-locking shaft lifting commands, effectively avoiding cascading equipment damage and large-scale scrapping.

[0019] In the reverse self-calibration step, a dimensionless physical-electrical coupling transcription matrix and Tikhonov inverse matrix mapping analysis are adopted, and a regularization coefficient is introduced to effectively overcome the multi-solution degradation problem under multivariable coupling environment, which significantly improves the convergence speed of reverse decoupling calculation and diagnostic positioning accuracy.

[0020] This solution is based on existing data from CNC drilling rig current acquisition, optical target detection, and electrical measurement equipment. It achieves adaptive control by constructing a physical-chemical-mechanical mechanism model and mathematical inversion algorithm. It eliminates the need for additional expensive online chemical composition analyzers or complex physical sensors, thus reducing the overall deployment and maintenance costs of the system.

[0021] Furthermore, in the state diagnosis step, the spindle motor current bias signal sequence and the current number of physical borehole runs are input together into an adaptive time-domain filter for smoothing fundamental wave extraction, including: A Hall current transformer is used to collect the spindle AC motor current bias signal sequence of the CNC drilling machine in real time during ultra-deep micro-hole drilling at a frequency greater than or equal to a preset sampling rate. The spindle motor current bias signal sequence and the current number of physical drilling operations are input together into a Savitzky-Golay adaptive time-domain filter based on second-order polynomial fitting. The Savitzky-Golay adaptive time-domain filter uses an asymmetric sliding window and uses the sliding quadratic term to minimize the preset smoothing benchmark to extract the fundamental wave of the spindle motor current bias signal sequence with high signal-to-noise ratio smoothing, thereby filtering out the mechanical oscillation noise caused by the instantaneous chip removal abrupt movement, and obtaining the net load current sequence. The semi-symmetrical window length of the sliding window of the asymmetric sliding axis window is dynamically adjusted according to the number of times the drill bit physically drills, so as to adapt to the fundamental characteristics of the load current under different tool wear stages.

[0022] Furthermore, in the state diagnosis step, the time-division causal state determination logic is used to match and determine the first-order slope abrupt gradient and the current deviation index variable resistance factor, outputting the drill bit working state, and issuing a corresponding first control command to the CNC drilling machine based on the drill bit working state, including: If the first-order slope abrupt gradient is lower than the negative boundary limit and the current deviation index variable resistance factor meets the preset first deviation threshold, then the working state of the drill bit is determined to be the broken edge state. The control system outputs the highest priority emergency stop command to the programmable logic controller PLC as the first control command, so as to feed the self-locking axial motor and force the shaft to lift. At the same time, the circuit board coordinates when the broken edge occurs are output. If the envelope of the first-order slope abrupt gradient in the time domain is in a monotonically rising state, and when the number of physical boreholes of the drill bit is close to the preset life limit, the cumulative current deviation index variable resistance factor meets the preset second deviation threshold, and the temperature rise rate of the drill shaft exceeds the preset temperature rise threshold, then the working state of the drill bit is determined to be a wear and aging state. The control system outputs an automatic tool change command to the CNC drilling machine as the first control command to switch to the spare tool holder to perform automatic tool change, and stores the current tool failure characteristics and process coordinate set into the traceability database.

[0023] Furthermore, in the copper plating control step, a spatial alignment differential degradation operator is calculated based on the cumulative alignment deviation of the optical target, and the spatial alignment differential degradation operator is used to perform self-correction on the drilling coordinate axis feed command of the CNC drilling machine, including: The optical inspection camera built into the CNC drilling rig is used to read the cumulative alignment deviation of the optical target under mechanical alignment of the multi-layer panel. A spatial alignment differential degradation operator is constructed. The spatial alignment differential degradation operator uses the cumulative alignment deviation of the optical target as the independent variable and calculates the spatial alignment degradation amount through the product of a sign function and a cosine function. The independent variable of the cosine function is determined by multiplying the physical displacement of the alignment deviation in a specific direction by a decimal dimension conversion constant of 1000 and then taking the ratio of the result with a preset step size parameter. The calculated spatial alignment degradation amount is used to perform self-correction on the initial coordinate axis feed command of the CNC drilling machine, and the corrected CNC drilling machine coordinate axis feed command is recalculated and output, thereby suppressing the internal cracking of the hole wall caused by the hole ring offset due to shearing from the source.

[0024] Furthermore, in the copper deposition control step, based on the chemical reaction-physical liquid phase combined transport dynamic boundary model within the high aspect ratio holes, the instantaneous deposition rate of the through-holes of the multilayer panel at different axial depths is calculated, including: A chemical reaction-physical-liquid phase combined transport dynamic boundary model is constructed within a high aspect ratio through-blind hole. This model characterizes the instantaneous deposition rate of the through-blind hole at its axial depth as the product of the initial base deposition rate and the spatial diffusion attenuation operator. The model is characterized by the excitation force turbulence boundary gain and the forced jet circulation velocity gain acting as denominator correction constants on the exponential attenuation term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The basic deposition rate is determined by the copper ion concentration, pH value, and tank temperature in the chemical copper plating bath. The spatial diffusion attenuation operator is the exponential decay term of the logarithmic and linear combination function of the surface damping matching constant of the hole wall and the depth-to-diameter ratio of the through-hole. The excitation force turbulent boundary gain is caused by the self-compliant vertical vibration excitation force of the external mechanical or ultrasonic fixture and is proportional to the equipment excitation frequency. The forced jet circulation velocity gain is generated by the bidirectional forced jet circulation velocity dynamics of the perforated plate and is proportional to the jet physical valve flow velocity at the bottom of the tank.

[0025] Furthermore, in the copper plating control step, a non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate; when the non-uniformity index is lower than a preset non-uniformity threshold, a corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate within the chemical copper plating tank, including: The ratio of the instantaneous deposition rate at the middle section of the hole along the axial direction to the instantaneous deposition rate at both ends of the hole along the axial direction is defined as an index of the non-uniformity of the hole wall deposition thickness. The non-uniformity index is compared with a preset non-uniformity threshold. If the non-uniformity index is lower than the non-uniformity threshold, it is determined that there is a risk of hollow or non-uniform defects in the copper plating inside the hole, and the following physical self-correction response is triggered: First, adjust the mechanical convection scale by increasing the equipment excitation frequency of the external mechanical or ultrasonic clamp and increasing the flow velocity of the physical valve jet at the bottom of the tank to reduce laminar boundary damping. The increment of the equipment excitation frequency is proportional to the difference between the non-uniformity index and the non-uniformity threshold. Second, the chemical reaction rate is reduced by controlling the cooling heat exchanger to interlock and cool the chemical copper plating bath solution, thereby reducing the chemical reaction activity at the orifice and providing a molecular concentration recovery period for the middle section of the orifice, which is under the diffusion limit, to achieve uniform reconstruction of the orifice wall thickness.

[0026] Further, in the reverse self-calibration step, the electrical deviation vector of the tested circuit is obtained, and a physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process perturbation relative change vector, including: During the final product electrical performance testing phase, the electrical deviation vector, consisting of the in-circuit micro-return impedance deviation, parasitic self-inductance deviation, and parasitic coupled electromagnetic distributed capacitive reactance deviation, is obtained by high-frequency sensitive electrical characteristic testing using flying probes. Each component of impedance, inductance, and capacitance in the electrical deviation vector is divided by a preset reference design parameter, and then converted into a dimensionless normalized electrical deviation vector. A dimensionless normalized process disturbance relative change vector is constructed. The normalized process disturbance relative change vector includes a dimensionless polarization ratio potential drift relative factor divided by a preset initial physical reference value, a thermal stress residual tension field relative exponent, and an alignment axial slip relative scale. A linear mapping relationship is established between the normalized electrical deviation vector and the normalized process perturbation relative change vector to construct a dimensionless physical-electrical coupling transcription matrix. The partial derivative correlation kernels in the physical-electrical coupling transcription matrix correspond to the partial derivative values ​​of the normalized terms of each electrical parameter with respect to the polarization ratio potential drift relative factor, the thermal stress residual tension field relative exponent, and the alignment axial slip relative scale.

[0027] Furthermore, in the reverse self-calibration step, based on the physical-electrical coupling transcription matrix, an inverse matrix mapping analysis is performed on the normalized electrical deviation vector to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability, including: A Tikhonov inverse matrix mapping analysis method with regularization coefficient is introduced to solve the physical-electric coupling transcription matrix and the normalized electrical deviation vector by inverse mapping, and to quantitatively calculate the relative vector of the dominant feature of physical failure probability; wherein, the relative vector of the dominant feature of physical failure probability includes the relative factor of micro-region polarization ratio potential drift in electroplating bath, the relative index of thermal stress residual tension field of laminated structure, and the relative scale of shear physical axial slip. When performing inverse matrix mapping analysis, the dimensionless physical-electrical coupling transcription matrix is ​​transposed and multiplied by itself, and the product of the dimensionless regularization coefficient and the identity matrix is ​​added. The inverse of the square matrix after multiplication and summation is then obtained. The inverse square matrix is ​​then multiplied sequentially by the transpose of the physical-electrical coupling transcription matrix and the dimensionless normalized electrical deviation vector. Finally, the decoupled calculation yields the dominant feature relative vector that is numerically convergent and does not undergo dimension degradation.

[0028] Furthermore, in the reverse self-calibration step, based on the dominant strength of each dimension of the relative physical component in the dominant feature relative vector of the physical failure probability, the corresponding online physical device correction and anti-adjustment control behavior is automatically generated and executed, including: If the polarization ratio potential drift relative factor in the relative vector of the physical failure probability dominant feature exceeds the first preset limit value, it is determined to be a polarization dominant failure. The control system automatically increases the reverse current ratio of bidirectional reverse pulse electroplating, and before pattern electroplating, it automatically controls the addition valve to increase the flow rate of polyethylene glycol additive in the electroplating tank, so as to remove the excess copper thickness at the edge of the micro blind hole through the peeling effect of reverse ultra-pulse discharge electrolytic polarization on the micro area. If the relative index of the thermal stress residual tension field in the relative vector of the dominant physical failure probability exceeds the second preset limit value, it is determined to be a stress-dominated failure. The control system initiates phase change release and feeds back the temperature drop rate gradient from the fourth temperature zone to the rapid cooling period of the control laminating equipment. If the relative scale of axial slip in the relative vector of the dominant feature of the physical failure probability exceeds the third preset limit value, it is determined to be an alignment error-dominated failure. The control system issues a correction gain vector to adjust the image distortion scaling factor matrix of the laser direct imaging (LDI) camera, thereby compensating for the camera distortion amplification operator in subsequent imaging and preventing asymmetric torsional shear damage.

[0029] Furthermore, it also includes a cross-module collaborative promotion step based on the state diagnosis step, the copper plating control step, and the reverse self-calibration step, including: First, if the working state of the drill bit is determined to be a worn and aged state, the degree of drill bit wear characterized by the current deviation index variable resistance factor is fed forward as the initial value of hole wall roughness to the copper plating control step. Based on the degree of drill bit wear, the spatial diffusion attenuation operator in the dynamic boundary model is actively corrected, thereby increasing the equipment excitation frequency in advance and increasing the flow velocity of the physical valve jet at the bottom of the tank, eliminating the dead zone impedance of the laminar boundary layer caused by the rough burrs on the hole wall. Second, the multi-layer physical layer alignment target offset data measured in the preceding lamination and alignment processes is fed forward as a positioning constraint to the reverse self-calibration step; if the measured alignment deviation is lower than the preset alignment deviation threshold, the column vectors corresponding to the axial slip in the physical-electric coupling transcription matrix are set to zero during the inverse matrix mapping analysis to complete matrix dimensionality reduction and pruning, thereby improving the convergence speed of the reverse decoupling calculation and the diagnostic positioning accuracy; Third, in the copper plating control step, the non-uniformity index is stably controlled within a preset safety range by adaptively adjusting the mechanical convection scale and chemical reaction rate in the chemical copper plating tank, and the obtained product performance non-uniformity distribution is used to adjust the reference state of the next batch of processed boards. The reference health resistance current constant lookup table of the next batch in the state diagnosis step is adaptively fine-tuned to reduce the random polarization noise of the system and improve the service life of the drill bit. Fourth, in the reverse self-correction step, if the relative axial slip obtained by decoupling shows a monotonically increasing trend, the relative axial slip is fed back to the state diagnosis step. By generating an absolute origin coordinate deviation axis compensation amount and injecting it into the position servo control channel of the CNC drilling machine for coordinate compensation, the cutting origin alignment position correction is achieved, preventing the relative tilting and off-center loading of the drill bit axis in the next batch. If the relative value of the polarization factor obtained by decoupling exceeds the preset polarization threshold, the relative value of the polarization factor is fed back to the copper plating control step, and the first-order correction parameters of the initial reaction rate and spatial diffusion decay operator in the dynamic model are periodically re-injected. Attached Figure Description

[0030] Figure 1 This is a block diagram of the closed-loop causal physical network topology related to the data interaction, cross-promotion, and anti-correction of the three physical functional layers involved in the embodiment. Figure 2 The flowchart in this embodiment describes the second-order polynomial sliding SG filtering of the spindle motor current signal in micro-hole drilling, the calculation of the first-order slope abrupt gradient, and the determination of needle breakage / wear status. Figure 3 This is a control flowchart in the embodiment for calculating the instantaneous deposition rate, evaluating non-uniformity, and executing physical self-correction response based on the kinetic boundary model during the chemical copper plating stage; Figure 4 This is a control flowchart in the embodiment, which describes how, during the finished product testing stage, based on the electrical deviation vector of the flying probe test, failure factors are decoupled and correction actions are assigned to each physical device using the Tikhonov inverse inversion model. Detailed Implementation

[0031] The following detailed description illustrates the specific implementation method: A method for full-process data traceability and control of circuit boards (e.g.) Figure 1 (As shown), including the following steps: S1. State Diagnosis Steps: The spindle motor current bias signal sequence of the CNC drilling machine during micro-hole drilling is acquired via high-frequency acquisition. This sequence, along with the current number of physical drilling operations, is input into an adaptive time-domain filter for smoothing and fundamental wave extraction, resulting in a net load current sequence after filtering out instantaneous mechanical oscillation noise. Based on this net load current sequence and a pre-calculated baseline health resistance current constant, the current deviation index variable resistance factor is calculated, and the first-order slope abrupt gradient of the net load current sequence is calculated. A time-division causal state determination logic is used to match and determine the first-order slope abrupt gradient and the current deviation index variable resistance factor, outputting the drill bit's working state. A corresponding first control command is then issued to the CNC drilling machine based on the drill bit's working state. S2. Copper Plating Control Steps: The cumulative alignment deviation of the optical target in the multilayer panel under mechanical alignment is read. A spatial alignment differential degradation operator is calculated based on the cumulative alignment deviation, and the spatial alignment differential degradation operator is used to self-correct the drilling coordinate axis feed command of the CNC drilling machine. The multilayer panel after correction drilling is delivered to the chemical copper plating tank. Based on the chemical reaction-physical liquid phase combined transport dynamic boundary model in high aspect ratio holes, the instantaneous deposition rate of the through-holes of the multilayer panel at different axial depths is calculated. The excitation force turbulence boundary gain and the forced jet circulation velocity gain are used as denominator correction constants applied to the exponential decay term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate. When the non-uniformity index is lower than a preset non-uniformity threshold, the corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate in the chemical copper plating tank. S3. Reverse Self-Correction Step: During the finished product electrical performance testing stage, the electrical deviation vector of the tested circuit is obtained through electrical testing equipment, and the electrical deviation vector is converted into a dimensionless normalized electrical deviation vector; a physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process disturbance relative change vector; based on the physical-electrical coupling transcription matrix, Tikhonov inverse matrix mapping analysis is performed on the normalized electrical deviation vector to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability; according to the dominant strength of each dimension of the physical component in the dominant feature relative vector of physical failure probability, the corresponding online physical equipment correction and reverse adjustment control behavior is automatically generated and executed.

[0032] In practical application, the spindle motor current bias signal sequence refers to the current fluctuation signal generated by the spindle drive motor due to load changes during CNC drilling. This signal reflects the resistance when the drill bit contacts and cuts the circuit board material (such as copper-clad laminate, fiberglass cloth, epoxy resin, etc.). Optional acquisition devices include Hall current sensors, electromagnetic transformers, or current detection modules built into the motor driver.

[0033] The physical number of drill holes refers to the cumulative number of times the drill bit has performed drilling operations. Since the drill bit inevitably wears down during the cutting process, the number of drill holes is an important cumulative physical quantity for assessing the tool wear stage.

[0034] An adaptive time-domain filter is a digital time-domain filter whose filtering parameters (such as filter window length and filter coefficients) can be dynamically adjusted according to external input variables (such as the number of times the drill bit physically drills). Available types include adaptive Kalman filters, adaptive moving average filters, and adaptive Savitzky-Golay filters.

[0035] The net load current sequence refers to the current sequence that has been filtered after the original current signal has been processed to eliminate transient random noise caused by sudden changes in cutting chip removal resistance, high-frequency mechanical vibration, etc., and only retains the current sequence that reflects the fundamental characteristics of the cutting resistance of the tool.

[0036] Time-sharing causal state determination logic refers to a logical determination method that combines dynamic change rate (first-order slope abrupt gradient) and static deviation (current deviation index variable resistance factor) to perform multi-condition causal matching at different time scales or tool life stages.

[0037] Construction and calibration process of adaptive time-domain filters: a. An adaptive time-domain filter is constructed based on the Savitzky-Golay (SG) filtering algorithm. This filter smooths the data by performing local polynomial least-squares fitting within a sliding window.

[0038] b. Set adaptive rules: adjust the half-window length of the sliding window. Established based on the number of times the drill bit physically runs through the hole. Related functions. In the initial stages of tool use ( Smaller size), sharper cutting edge, smoother chip removal, lower mechanical vibration noise, and a smaller window to preserve transient signal details; in the later stages of tool use ( When the tool is nearing its lifespan limit, it wears severely, chip removal resistance increases and is accompanied by severe chip removal abrupt changes, and mechanical vibration noise is significantly enhanced. At this point, increasing the window length can enhance the smoothing effect.

[0039] c. Calibrate the smooth approximation coefficients: Pre-calculate different window lengths using the least squares method. The second-order polynomial fitting coefficients This forms a coefficient lookup table, which is used for filtering based on the current... Direct call.

[0040] In this embodiment, the spindle speed of the CNC drilling machine Axial feed rate Reference health resistance current constant .

[0041] High-frequency sampling rate The adaptive time-domain filter employs a Savitzky-Golay filter fitted with a second-order polynomial, with a half-window length of... Based on the number of holes Dynamic adjustment: when hour, ;when hour, ;when hour, .

[0042] by For example, the corresponding smooth approximation coefficient vector is: This coefficient was used to analyze the acquired spindle motor current bias signal sequence. Perform convolution smoothing to obtain the net load current sequence. .

[0043] Calculate the current deviation index variable resistance factor: Calculate the first-order slope gradient abrupt change As attached Figure 2 As shown, this illustrates the comparison of current signals before and after filtering and the response to abrupt gradient changes.

[0044] In the state diagnosis step, the spindle motor current bias signal sequence and the current number of physical borehole runs are input together into an adaptive time-domain filter for smooth fundamental wave extraction, including: A Hall current transformer is used to collect the spindle AC motor current bias signal sequence of the CNC drilling machine in real time during ultra-deep micro-hole drilling at a frequency greater than or equal to a preset sampling rate. The spindle motor current bias signal sequence and the current number of physical drilling operations are input together into a Savitzky-Golay adaptive time-domain filter based on second-order polynomial fitting. The Savitzky-Golay adaptive time-domain filter uses an asymmetric sliding window and uses the sliding quadratic term to minimize the preset smoothing benchmark to extract the fundamental wave of the spindle motor current bias signal sequence with high signal-to-noise ratio smoothing, thereby filtering out the mechanical oscillation noise caused by the instantaneous chip removal abrupt movement, and obtaining the net load current sequence. The semi-symmetrical window length of the sliding window of the asymmetric sliding axis window is dynamically adjusted according to the number of times the drill bit physically drills, so as to adapt to the fundamental characteristics of the load current under different tool wear stages.

[0045] A Hall current transformer is a sensor that uses the Hall effect to convert a large current into a small voltage / current signal with the same frequency and phase. It features fast response and a wide measurement range. The Hall current transformer used in this embodiment is a closed-loop Hall current sensor with a sampling frequency of... Set as .

[0046] An asymmetric sliding window refers to a situation where, during time-domain filtering, the number of data points on the left and right sides of the current processing point is unequal, or the center of the window is biased towards one side of the current processing point. In this embodiment, to reduce the group delay caused by the filtering algorithm and ensure the real-time performance of state diagnosis, an asymmetric sliding window is used, with the left historical data window length being greater than the right look-ahead data window length. The specific window configuration is as follows: .

[0047] In drilling depth-to-diameter ratio When using micropores, as the pore depth increases, the chip removal path becomes longer, and the chip removal resistance increases non-linearly. The Savitzky-Golay adaptive time-domain filter dynamically adjusts the semi-symmetric window length. (as in) Time adjustment to By utilizing the sliding quadratic term to minimize the smoothing benchmark, it can dynamically filter out random oscillation noise caused by glass fiber entanglement and instantaneous chip accumulation, and accurately extract the fundamental current drift caused by tool wear.

[0048] In the state diagnosis step, time-division causal state determination logic is used to match and determine the first-order slope abrupt gradient and the current deviation index variable resistance factor, outputting the drill bit working state, and issuing a corresponding first control command to the CNC drilling machine based on the drill bit working state, including: If the first-order slope abrupt gradient is lower than the negative boundary limit and the current deviation index variable resistance factor meets the preset first deviation threshold, then the working state of the drill bit is determined to be the broken edge state. The control system outputs the highest priority emergency stop command to the programmable logic controller PLC as the first control command, so as to feed the self-locking axial motor and force the shaft to lift. At the same time, the circuit board coordinates when the broken edge occurs are output. If the envelope of the first-order slope abrupt gradient in the time domain is in a monotonically rising state, and when the number of physical boreholes of the drill bit is close to the preset life limit, the cumulative current deviation index variable resistance factor meets the preset second deviation threshold, and the temperature rise rate of the drill shaft exceeds the preset temperature rise threshold, then the working state of the drill bit is determined to be a wear and aging state. The control system outputs an automatic tool change command to the CNC drilling machine as the first control command to switch to the spare tool holder to perform automatic tool change, and stores the current tool failure characteristics and process coordinate set into the traceability database.

[0049] A first-order slope abrupt gradient below the negative boundary limit means that the rate of change of the current signal becomes extremely negative, indicating that the load disappears instantaneously.

[0050] The first and second deviation thresholds are used to distinguish between static current deviations from the limits of broken edges and wear and aging.

[0051] In this embodiment, the negative boundary limit is set as follows: The first deviation threshold is set to If during the drilling process, it is detected that... and If the condition is detected as broken, the control system immediately sends a high-priority emergency stop command to the PLC, with a response delay controlled within [timeframe missing]. Internally, a self-locking axial motor forces the shaft to lift, outputting the broken blade coordinates (e.g., ).

[0052] For the wear and aging condition (Worn), the preset life limit is... Next, the second deviation threshold is set to The preset temperature rise threshold is set to .like The envelope line rises monotonically and gently, and in hour, Furthermore, the infrared temperature sensor measured the drill shaft temperature rise rate to reach If the failure is detected, it is determined to be in a worn and aged state. The control system outputs an automatic tool change command and records the failure characteristics and process coordinates.

[0053] S2. Copper Plating Control Steps: The cumulative alignment deviation of the optical target in the multilayer panel under mechanical alignment is read. A spatial alignment differential degradation operator is calculated based on the cumulative alignment deviation, and the spatial alignment differential degradation operator is used to self-correct the drilling coordinate axis feed command of the CNC drilling machine. The multilayer panel after correction drilling is delivered to the chemical copper plating tank. Based on the chemical reaction-physical liquid phase combined transport dynamic boundary model in high aspect ratio holes, the instantaneous deposition rate of the through-holes of the multilayer panel at different axial depths is calculated. The excitation force turbulence boundary gain and the forced jet circulation velocity gain are used as denominator correction constants applied to the exponential decay term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate. When the non-uniformity index is lower than a preset non-uniformity threshold, the corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate in the chemical copper plating tank.

[0054] The spatial alignment differential degradation operator is a mathematical operator used to quantitatively describe the degradation law of spatial geometric alignment accuracy of multilayer panels as physical position changes under the cumulative effects of mechanical lamination and alignment tolerances.

[0055] The chemical reaction-physical liquid phase combined transport dynamic boundary model is a dynamic boundary model constructed by coupling the chemical reaction consumption kinetics (such as copper ion reduction and complexing agent dissociation) inside the micropore with the physical liquid phase mass transfer and diffusion (such as Fick diffusion and convective mass transfer).

[0056] The excitation force turbulent boundary gain and the forced jet circulation velocity gain refer to the micro-area turbulent convection promotion effect caused by external mechanical / ultrasonic vibration, and the in-hole liquid phase penetration enhancement effect caused by forced jet circulation, respectively.

[0057] Construction process of dynamic boundary model: a. Establish a one-dimensional axial diffusion-reaction equation: in, This represents the concentration of copper ions inside the pore. For the effective diffusion coefficient, The aperture is [not specified].

[0058] b. Instantaneous deposition rate It is characterized as the product of the reaction kinetics term and the spatial decay term, and a correction term for the excitation force and the forced jet is introduced.

[0059] c. By using different depth-to-diameter ratios ( A copper plating experiment was conducted inside the standard sample hole to determine the actual copper thickness at various depths of the hole wall. The surface damping matching constant of the hole wall was then calibrated using least squares fitting. .

[0060] In this embodiment, the final electrical insulation layer thickness of the multilayer board physical lamination Mechanical micro-hole drilling diameter Aspect Ratio .

[0061] Instantaneous deposition rate The calculation formula is: Among them, the spatial diffusion attenuation operator After calibration, , Without external gain, .

[0062] Excitation force turbulent boundary gain and forced jet circulation velocity gain The constant is used as a denominator correction constant in the exponential decay term. This is used to calculate the inhomogeneity index. .like Below the preset threshold If this is the case, a self-correcting response will be triggered.

[0063] In the copper plating control step, a spatial alignment differential degradation operator is calculated based on the accumulated alignment deviation of the optical target, and the spatial alignment differential degradation operator is used to perform self-correction on the drilling coordinate axis feed command of the CNC drilling machine, including: The optical inspection camera built into the CNC drilling rig is used to read the cumulative alignment deviation of the optical target under mechanical alignment of the multi-layer panel. A spatial alignment differential degradation operator is constructed. The spatial alignment differential degradation operator uses the cumulative alignment deviation of the optical target as the independent variable and calculates the spatial alignment degradation amount through the product of a sign function and a cosine function. The independent variable of the cosine function is determined by multiplying the physical displacement of the alignment deviation in a specific direction by a decimal dimension conversion constant of 1000 and then taking the ratio of the result with a preset step size parameter. The calculated spatial alignment degradation amount is used to perform self-correction on the initial coordinate axis feed command of the CNC drilling machine, and the corrected CNC drilling machine coordinate axis feed command is recalculated and output, thereby suppressing the internal cracking of the hole wall caused by the hole ring offset due to shearing from the source.

[0064] The decimal dimension conversion constant 1000 is used to convert the physical displacement of alignment deviations, which is measured in millimeters (mm), to micrometers (µm). The values ​​are in units of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 ...

[0065] In this embodiment, the alignment deviation read by the optical detection camera is in The physical displacement in the direction is Preset step size parameters The spatial alignment differential degradation operator is calculated as follows: If the actual deviation displacement is ,but: The space alignment degradation is used to compensate for the initial feed command, and the coordinate axis feed command is recalculated to achieve self-correction and eliminate hole ring offset.

[0066] In the copper deposition control step, based on the chemical reaction-physical liquid phase combined transport dynamic boundary model within the high aspect ratio vias, the instantaneous deposition rate of the through-holes in the multilayer panel at different axial depths is calculated, including: A chemical reaction-physical-liquid phase combined transport dynamic boundary model is constructed within a high aspect ratio through-blind hole. This model characterizes the instantaneous deposition rate of the through-blind hole at its axial depth as the product of the initial base deposition rate and the spatial diffusion attenuation operator. The model is characterized by the excitation force turbulence boundary gain and the forced jet circulation velocity gain acting as denominator correction constants on the exponential attenuation term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The basic deposition rate is determined by the copper ion concentration, pH value, and tank temperature in the chemical copper plating bath. The spatial diffusion attenuation operator is the exponential decay term of the logarithmic and linear combination function of the surface damping matching constant of the hole wall and the depth-to-diameter ratio of the through-hole. The excitation force turbulent boundary gain is caused by the self-compliant vertical vibration excitation force of the external mechanical or ultrasonic fixture and is proportional to the equipment excitation frequency. The forced jet circulation velocity gain is generated by the bidirectional forced jet circulation velocity dynamics of the perforated plate and is proportional to the jet physical valve flow velocity at the bottom of the tank.

[0067] The surface damping matching constant of the pore wall is a comprehensive influence coefficient characterizing the physical properties of the pore wall, such as micro-roughness and charge distribution, on the resistance to liquid phase diffusion.

[0068] In this embodiment, the initial base deposition rate copper ion concentration in the bath solution , ,temperature At that time, it was calibrated as .

[0069] Excitation force turbulent boundary gain ,in Reference frequency Equipment excitation frequency Calculated .

[0070] Forced jet circulation velocity gain ,in Reference flow rate Actual jet velocity Calculated .

[0071] Substituting the dynamic boundary model, the middle section of the hole along the axis ( The instantaneous deposition rate of ) is: Axial end of hole ( The instantaneous deposition rate of ) is: In the copper plating control step, a non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate; when the non-uniformity index is lower than a preset non-uniformity threshold, a corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate within the chemical copper plating tank, including: The ratio of the instantaneous deposition rate at the middle section of the hole along the axial direction to the instantaneous deposition rate at both ends of the hole along the axial direction is defined as an index of the non-uniformity of the hole wall deposition thickness. The non-uniformity index is compared with a preset non-uniformity threshold. If the non-uniformity index is lower than the non-uniformity threshold, it is determined that there is a risk of hollow or non-uniform defects in the copper plating inside the hole, and the following physical self-correction response is triggered: First, adjust the mechanical convection scale by increasing the equipment excitation frequency of the external mechanical or ultrasonic clamp and increasing the flow velocity of the physical valve jet at the bottom of the tank to reduce laminar boundary damping. The increment of the equipment excitation frequency is proportional to the difference between the non-uniformity index and the non-uniformity threshold. Second, the chemical reaction rate is reduced by controlling the cooling heat exchanger to interlock and cool the chemical copper plating bath solution, thereby reducing the chemical reaction activity at the orifice and providing a molecular concentration recovery period for the middle section of the orifice, which is under the diffusion limit, to achieve uniform reconstruction of the orifice wall thickness.

[0072] The non-uniformity index of pore wall deposition thickness is The closer this value is to 1, the more uniform the deposition thickness within the pores.

[0073] Based on the calculation in the previous step, the non-uniformity index (Note: If defined as the ratio of the middle section to both ends, this is the reciprocal of the ratio of the thickness at both ends to the thickness at the middle section, i.e.) ).

[0074] because If a defect risk is detected, a physical self-calibration response is triggered: First, adjust the scale of mechanical convection: the increment of the equipment's excitation frequency is proportional to the difference. Set the scaling factor. Assuming we get At the same time, the flow rate of the bottom jet physical valve will be... Depend on Increase to .

[0075] Second, reduce the chemical reaction rate: control the cooling heat exchanger to lower the bath temperature. Depend on Reduce to This reduces the reactivity at the orifice, allowing copper ions sufficient time to diffuse to the middle section of the orifice axis. (See attached image.) Figure 3 As shown, this illustrates the change in deposition rate in the middle of the borehole wall before and after compensation.

[0076] S3. Reverse Self-Correction Step: During the finished product electrical performance testing stage, the electrical deviation vector of the tested circuit is obtained through electrical testing equipment, and the electrical deviation vector is converted into a dimensionless normalized electrical deviation vector; a physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process disturbance relative change vector; based on the physical-electrical coupling transcription matrix, Tikhonov inverse matrix mapping analysis is performed on the normalized electrical deviation vector to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability; according to the dominant strength of each dimension of the physical component in the dominant feature relative vector of physical failure probability, the corresponding online physical equipment correction and reverse adjustment control behavior is automatically generated and executed.

[0077] The electrical deviation vector is a multi-dimensional vector obtained from the electrical testing of the finished product, reflecting abnormalities in the electrical performance of traces and vias. It typically includes resistance deviation, inductance deviation, capacitance deviation, etc.

[0078] The physical-electrical coupling transcription matrix is ​​a Jacobian matrix that characterizes the relationship between the first-order partial derivatives of the front-end physical process perturbation (independent variable) and the end-end electrical deviation (dependent variable).

[0079] Tikhonov inverse matrix mapping analysis is a regularization method for solving ill-posed inverse problems. By introducing a penalty term (regularization matrix) into the objective function, it overcomes matrix ill-conditioning and obtains stable numerical solutions.

[0080] Construction process of physical-electrical coupling transcription matrix: a. Determine the state variable: electrical deviation vector Process disturbance vector .

[0081] b. Perform a Taylor expansion with small perturbations near the baseline operating point, ignoring higher-order terms, to obtain a linear mapping relationship: .

[0082] c. Using electromagnetic simulation software (such as ANSYS HFSS) or by designing orthogonal experiments, obtain electrical deviation response data under different process disturbances, calculate each partial derivative component in the matrix through numerical differentiation, and construct the transcription matrix.

[0083] In this embodiment, the electrical deviation vector measured by the flying probe test is: .

[0084] Normalized (divided by the baseline design parameters) ), thus obtaining the dimensionless normalized electrical bias vector: The constructed dimensionless physical-electric coupling transcription matrix is ​​as follows: Introducing regularization coefficients Using the Tikhonov inverse matrix mapping formula: The solution yields the dominant feature relative vector of physical failure probability: because If the value significantly exceeds the preset limit, it is determined to be a polarization-dominant failure, and the system automatically generates corresponding counter-modulation control behavior. The specific process is attached. Figure 4 As shown.

[0085] In the reverse self-calibration step, the electrical deviation vector of the circuit under test is obtained, and a physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process perturbation relative change vector, including: During the final product electrical performance testing phase, the electrical deviation vector, consisting of the in-circuit micro-return impedance deviation, parasitic self-inductance deviation, and parasitic coupled electromagnetic distributed capacitive reactance deviation, is obtained by high-frequency sensitive electrical characteristic testing using flying probes. Each component of impedance, inductance, and capacitance in the electrical deviation vector is divided by a preset reference design parameter, and then converted into a dimensionless normalized electrical deviation vector. A dimensionless normalized process disturbance relative change vector is constructed. The normalized process disturbance relative change vector includes a dimensionless polarization ratio potential drift relative factor divided by a preset initial physical reference value, a thermal stress residual tension field relative exponent, and an alignment axial slip relative scale. A linear mapping relationship is established between the normalized electrical deviation vector and the normalized process perturbation relative change vector to construct a dimensionless physical-electrical coupling transcription matrix. The partial derivative correlation kernels in the physical-electrical coupling transcription matrix correspond to the partial derivative values ​​of the normalized terms of each electrical parameter with respect to the polarization ratio potential drift relative factor, the thermal stress residual tension field relative exponent, and the alignment axial slip relative scale.

[0086] Flying probe high-frequency sensitive electrical characteristic measurement is a method that uses a mobile flying probe tester to inject a weak high-frequency excitation signal into the network under test, and measures the reflection coefficient or transmission parameters, thereby extracting parasitic parameters such as resistance, inductance, and capacitance with high precision.

[0087] In this embodiment, the excitation signal frequency of the flying needle tester is set to... .

[0088] The physical reference value in the normalized process disturbance relative change vector is set as: polarization potential reference. (vs. SCE), thermal stress reference Alignment sliding reference .

[0089] The partial derivative correlation kernel in the matrix is ​​calculated using the numerical perturbation method, for example: Finally, a dimensionless physical-electric coupling transcription matrix was constructed. .

[0090] In the reverse self-calibration step, an inverse matrix mapping analysis is performed on the normalized electrical deviation vector based on the physical-electrical coupling transcription matrix to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability, including: A Tikhonov inverse matrix mapping analysis method with regularization coefficient is introduced to solve the physical-electric coupling transcription matrix and the normalized electrical deviation vector by inverse mapping, and to quantitatively calculate the relative vector of the dominant feature of physical failure probability; wherein, the relative vector of the dominant feature of physical failure probability includes the relative factor of micro-region polarization ratio potential drift in electroplating bath, the relative index of thermal stress residual tension field of laminated structure, and the relative scale of shear physical axial slip. When performing inverse matrix mapping analysis, the dimensionless physical-electrical coupling transcription matrix is ​​transposed and multiplied by itself, and the product of the dimensionless regularization coefficient and the identity matrix is ​​added. The inverse of the square matrix after multiplication and summation is then obtained. The inverse square matrix is ​​then multiplied sequentially by the transpose of the physical-electrical coupling transcription matrix and the dimensionless normalized electrical deviation vector. Finally, the decoupled calculation yields the dominant feature relative vector that is numerically convergent and does not undergo dimension degradation.

[0091] The numerical convergence without dimensional degradation is due to the fact that all elements of the matrix have been dimensionless and a regularization penalty term has been introduced, which avoids numerical divergence and physical dimension confusion caused by singular values ​​approaching zero when solving the inverse matrix.

[0092] In this embodiment, the calculation process is as follows: a. Calculate the transpose multiplication term: .

[0093] b. Introduce regularization terms: .

[0094] c. Inverse operation: .

[0095] d. Final projection: .

[0096] By following the steps above, the problem of directly inverting the equation is avoided. Solution failures caused by excessively large condition numbers (ill-conditioned matrices) should be addressed to ensure the physical rationality of the decoupling results.

[0097] In the reverse self-calibration step, based on the dominant strength of each dimension of the relative physical component in the dominant feature relative vector of the physical failure probability, the corresponding online physical equipment correction and inverse adjustment control behavior is automatically generated and executed, including: If the polarization ratio potential drift relative factor in the relative vector of the physical failure probability dominant feature exceeds the first preset limit value, it is determined to be a polarization dominant failure. The control system automatically increases the reverse current ratio of bidirectional reverse pulse electroplating, and before pattern electroplating, it automatically controls the addition valve to increase the flow rate of polyethylene glycol additive in the electroplating tank, so as to remove the excess copper thickness at the edge of the micro blind hole through the peeling effect of reverse ultra-pulse discharge electrolytic polarization on the micro area. If the relative index of the thermal stress residual tension field in the relative vector of the dominant physical failure probability exceeds the second preset limit value, it is determined to be a stress-dominated failure. The control system initiates phase change release and feeds back the temperature drop rate gradient from the fourth temperature zone to the rapid cooling period of the control laminating equipment. If the relative scale of axial slip in the relative vector of the dominant feature of the physical failure probability exceeds the third preset limit value, it is determined to be an alignment error-dominated failure. The control system issues a correction gain vector to adjust the image distortion scaling factor matrix of the laser direct imaging (LDI) camera, thereby compensating for the camera distortion amplification operator in subsequent imaging and preventing asymmetric torsional shear damage.

[0098] Bidirectional reverse pulse plating (PPR) is an electroplating process that uses alternating forward plating current and reverse stripping current to significantly improve the uniformity of coating distribution in holes with high aspect ratios.

[0099] In this embodiment, the first preset limit value is set as follows: The second preset limit value is set to The third preset limit value is set as follows: .

[0100] Due to decoupling This was determined to be a polarization-dominated failure. The control system automatically adjusted the pulse parameters of the patterned electroplating tank: the ratio of the forward to reverse current was changed from the initial... Increase the reverse current ratio The reverse pulse time duty cycle is adjusted to [number] times. Simultaneously, the self-addition valve of the electroplating tank is controlled to adjust the flow rate of polyethylene glycol (PEG) additive. Increase to This is to suppress excessive deposition at the edges of blind holes.

[0101] It also includes a cross-module collaborative promotion step based on the state diagnosis step, the copper plating control step, and the reverse self-calibration step, including: First, if the working state of the drill bit is determined to be a worn and aged state, the degree of drill bit wear characterized by the current deviation index variable resistance factor is fed forward as the initial value of hole wall roughness to the copper plating control step. Based on the degree of drill bit wear, the spatial diffusion attenuation operator in the dynamic boundary model is actively corrected, thereby increasing the equipment excitation frequency in advance and increasing the flow velocity of the physical valve jet at the bottom of the tank, eliminating the dead zone impedance of the laminar boundary layer caused by the rough burrs on the hole wall. Second, the multi-layer physical layer alignment target offset data measured in the preceding lamination and alignment processes is fed forward as a positioning constraint to the reverse self-calibration step; if the measured alignment deviation is lower than the preset alignment deviation threshold, the column vectors corresponding to the axial slip in the physical-electric coupling transcription matrix are set to zero during the inverse matrix mapping analysis to complete matrix dimensionality reduction and pruning, thereby improving the convergence speed of the reverse decoupling calculation and the diagnostic positioning accuracy; Third, in the copper plating control step, the non-uniformity index is stably controlled within a preset safety range by adaptively adjusting the mechanical convection scale and chemical reaction rate in the chemical copper plating tank, and the obtained product performance non-uniformity distribution is used to adjust the reference state of the next batch of processed boards. The reference health resistance current constant lookup table of the next batch in the state diagnosis step is adaptively fine-tuned to reduce the random polarization noise of the system and improve the service life of the drill bit. Fourth, in the reverse self-correction step, if the relative axial slip obtained by decoupling shows a monotonically increasing trend, the relative axial slip is fed back to the state diagnosis step. By generating an absolute origin coordinate deviation axis compensation amount and injecting it into the position servo control channel of the CNC drilling machine for coordinate compensation, the cutting origin alignment position correction is achieved, preventing the relative tilting and off-center loading of the drill bit axis in the next batch. If the relative value of the polarization factor obtained by decoupling exceeds the preset polarization threshold, the relative value of the polarization factor is fed back to the copper plating control step, and the first-order correction parameters of the initial reaction rate and spatial diffusion decay operator in the dynamic model are periodically re-injected.

[0102] Cross-module collaborative promotion refers to the establishment of a closed-loop connection between three independent process modules—state diagnosis, copper plating control, and reverse self-correction—through data feedforward and feedback channels, thereby achieving cross-process process self-healing and error compensation.

[0103] Specific implementation: First coordination (S1 feedforward S2): When S1 determines that the drill bit is in a worn and aging state, the variable resistance factor will be adjusted. The degree of uneven wear is fed forward to S2 as an indicator of wear. S2 utilizes a corrected mapping. The diffusion attenuation operator is modified (where the rough convection damping factor is used). The corrected damping constant was calculated. This allows S2 to pre-adjust the equipment's excitation frequency to [a higher level]. The jet velocity was increased to Eliminate laminar flow dead zone impedance caused by burrs from blunt cutting tools.

[0104] Second coordination (lamination feedforward S3): The target offset measured in the preceding lamination process is... Below the preset alignment deviation threshold S3, when solving, will use the transcription matrix. The third column (corresponding to axial slip) is set to zero, reducing the matrix dimension to... This reduces the time required for Tikhonov inversion calculations and improves diagnostic positioning accuracy.

[0105] The third coordination (S2 feedback S1): S2 stabilizes and controls the non-uniformity index within... The above eliminates the scattered structural noise caused by uneven electroplating on the hole walls. This is then fed back to S1, setting the reference healthy resistance current constant for the next batch of spindle motors. The fluctuation range is narrowed, random polarization noise in the system is eliminated, and the service life of the drill bit is extended.

[0106] Fourth coordination (S3 feedback S1 and S2): If the axial slip relative to the scale of S3 decoupling If three consecutive plates show an upward trend, then coordinate axis compensation (e.g., ...) is injected into the CNC drilling rig servo channel corresponding to S1. This corrects eccentric cutting and prevents the drill bit from breaking due to tilting and uneven loading. If the relative value of the polarization factor exceeds the preset polarization threshold... This is then fed back to S2, where a first-order correction parameter re-injection is performed on the initial reaction rate in the kinetic model: .

[0107] The above are merely embodiments of the present invention. The invention is not limited to the fields covered by these embodiments. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are able to access all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for full-process data traceability and control of circuit boards, characterized in that, Includes the following steps: S1. State Diagnosis Steps: The spindle motor current bias signal sequence of the CNC drilling machine during micro-hole drilling is acquired via high-frequency acquisition. This sequence, along with the current number of physical drilling operations, is input into an adaptive time-domain filter for smoothing and fundamental wave extraction, resulting in a net load current sequence after filtering out instantaneous mechanical oscillation noise. Based on this net load current sequence and a pre-calculated baseline health resistance current constant, the current deviation index variable resistance factor is calculated, and the first-order slope abrupt gradient of the net load current sequence is calculated. A time-division causal state determination logic is used to match and determine the first-order slope abrupt gradient and the current deviation index variable resistance factor, outputting the drill bit's working state. A corresponding first control command is then issued to the CNC drilling machine based on the drill bit's working state. S2. Copper Plating Control Steps: The cumulative alignment deviation of the optical target in the multilayer panel under mechanical alignment is read. A spatial alignment differential degradation operator is calculated based on the cumulative alignment deviation, and the spatial alignment differential degradation operator is used to self-correct the drilling coordinate axis feed command of the CNC drilling machine. The multilayer panel after correction drilling is delivered to the chemical copper plating tank. Based on the chemical reaction-physical liquid phase combined transport dynamic boundary model in high aspect ratio holes, the instantaneous deposition rate of the through-holes of the multilayer panel at different axial depths is calculated. The excitation force turbulence boundary gain and the forced jet circulation velocity gain are used as denominator correction constants applied to the exponential decay term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate. When the non-uniformity index is lower than a preset non-uniformity threshold, the corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate in the chemical copper plating tank. S3. Reverse self-calibration step: During the finished product electrical performance testing stage, the electrical deviation vector of the tested circuit is obtained through electrical testing equipment, and the electrical deviation vector is converted into a dimensionless normalized electrical deviation vector; a physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process disturbance relative change vector, and Tikhonov inverse matrix mapping analysis is performed on the normalized electrical deviation vector based on the physical-electrical coupling transcription matrix to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability; Based on the dominant strength of each physical component in the relative vector of the dominant features of the physical failure probability, the corresponding online physical equipment correction and counter-adjustment control behavior is automatically generated and executed.

2. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the state diagnosis step, the spindle motor current bias signal sequence and the current number of physical borehole runs are input together into an adaptive time-domain filter for smooth fundamental wave extraction, including: A Hall current transformer is used to collect the spindle AC motor current bias signal sequence of the CNC drilling machine in real time during ultra-deep micro-hole drilling at a frequency greater than or equal to a preset sampling rate. The spindle motor current bias signal sequence and the current number of physical drilling operations are input together into a Savitzky-Golay adaptive time-domain filter based on second-order polynomial fitting. The Savitzky-Golay adaptive time-domain filter uses an asymmetric sliding window and uses the sliding quadratic term to minimize the preset smoothing benchmark to extract the fundamental wave of the spindle motor current bias signal sequence with high signal-to-noise ratio smoothing, thereby filtering out the mechanical oscillation noise caused by the instantaneous chip removal abrupt movement, and obtaining the net load current sequence. The semi-symmetrical window length of the sliding window of the asymmetric sliding axis window is dynamically adjusted according to the number of times the drill bit physically drills, so as to adapt to the fundamental characteristics of the load current under different tool wear stages.

3. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the state diagnosis step, time-division causal state determination logic is used to match and determine the first-order slope abrupt gradient and the current deviation index variable resistance factor, outputting the drill bit working state, and issuing a corresponding first control command to the CNC drilling machine based on the drill bit working state, including: If the first-order slope abrupt gradient is lower than the negative boundary limit and the current deviation index variable resistance factor meets the preset first deviation threshold, then the working state of the drill bit is determined to be the broken edge state. The control system outputs the highest priority emergency stop command to the programmable logic controller PLC as the first control command, so as to feed the self-locking axial motor and force the shaft to lift. At the same time, the circuit board coordinates when the broken edge occurs are output. If the envelope of the first-order slope abrupt gradient in the time domain is in a monotonically rising state, and when the number of physical boreholes of the drill bit is close to the preset life limit, the cumulative current deviation index variable resistance factor meets the preset second deviation threshold, and the temperature rise rate of the drill shaft exceeds the preset temperature rise threshold, then the working state of the drill bit is determined to be a wear and aging state. The control system outputs an automatic tool change command to the CNC drilling machine as the first control command to switch to the spare tool holder to perform automatic tool change, and stores the current tool failure characteristics and process coordinate set into the traceability database.

4. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the copper plating control step, a spatial alignment differential degradation operator is calculated based on the accumulated alignment deviation of the optical target, and the spatial alignment differential degradation operator is used to perform self-correction on the drilling coordinate axis feed command of the CNC drilling machine, including: The optical inspection camera built into the CNC drilling rig is used to read the cumulative alignment deviation of the optical target under mechanical alignment of the multi-layer panel. A spatial alignment differential degradation operator is constructed. The spatial alignment differential degradation operator uses the cumulative alignment deviation of the optical target as the independent variable and calculates the spatial alignment degradation amount through the product of a sign function and a cosine function. The independent variable of the cosine function is determined by multiplying the physical displacement of the alignment deviation in a specific direction by a decimal dimension conversion constant of 1000 and then taking the ratio of the result with a preset step size parameter. The calculated spatial alignment degradation amount is used to perform self-correction on the initial coordinate axis feed command of the CNC drilling machine, and the corrected CNC drilling machine coordinate axis feed command is recalculated and output, thereby suppressing the internal cracking of the hole wall caused by the hole ring offset due to shearing from the source.

5. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the copper deposition control step, based on the chemical reaction-physical liquid phase combined transport dynamic boundary model within the high aspect ratio vias, the instantaneous deposition rate of the through-holes in the multilayer panel at different axial depths is calculated, including: A chemical reaction-physical-liquid phase combined transport dynamic boundary model is constructed within a high aspect ratio through-blind hole. This model characterizes the instantaneous deposition rate of the through-blind hole at its axial depth as the product of the initial base deposition rate and the spatial diffusion attenuation operator. The model is characterized by the excitation force turbulence boundary gain and the forced jet circulation velocity gain acting as denominator correction constants on the exponential attenuation term of the spatial diffusion attenuation operator to adaptively generate different compensation rates at different hole depths. The basic deposition rate is determined by the copper ion concentration, pH value, and tank temperature in the chemical copper plating bath. The spatial diffusion attenuation operator is the exponential decay term of the logarithmic and linear combination function of the surface damping matching constant of the hole wall and the depth-to-diameter ratio of the through-hole. The excitation force turbulent boundary gain is caused by the self-compliant vertical vibration excitation force of the external mechanical or ultrasonic fixture and is proportional to the equipment excitation frequency. The forced jet circulation velocity gain is generated by the bidirectional forced jet circulation velocity dynamics of the perforated plate and is proportional to the jet physical valve flow velocity at the bottom of the tank.

6. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the copper plating control step, a non-uniformity index of the hole wall deposition thickness is calculated based on the instantaneous deposition rate; when the non-uniformity index is lower than a preset non-uniformity threshold, a corresponding physical self-correction response is executed to adaptively adjust the mechanical convection scale and chemical reaction rate within the chemical copper plating tank, including: The ratio of the instantaneous deposition rate at the middle section of the hole along the axial direction to the instantaneous deposition rate at both ends of the hole along the axial direction is defined as an index of the non-uniformity of the hole wall deposition thickness. The non-uniformity index is compared with a preset non-uniformity threshold. If the non-uniformity index is lower than the non-uniformity threshold, it is determined that there is a risk of hollow or non-uniform defects in the copper plating inside the hole, and the following physical self-correction response is triggered: First, adjust the mechanical convection scale by increasing the equipment excitation frequency of the external mechanical or ultrasonic clamp and increasing the flow velocity of the physical valve jet at the bottom of the tank to reduce laminar boundary damping. The increment of the equipment excitation frequency is proportional to the difference between the non-uniformity index and the non-uniformity threshold. Second, the chemical reaction rate is reduced by controlling the cooling heat exchanger to interlock and cool the chemical copper plating bath solution, thereby reducing the chemical reaction activity at the orifice and providing a molecular concentration recovery period for the middle section of the orifice, which is under the diffusion limit, to achieve uniform reconstruction of the orifice wall thickness.

7. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the reverse self-calibration step, the electrical deviation vector of the circuit under test is obtained, and a physical-electrical coupling transcription matrix is ​​constructed to characterize the linear mapping relationship between the normalized electrical deviation vector and the normalized process perturbation relative change vector, including: During the final product electrical performance testing phase, the electrical deviation vector, consisting of the in-circuit micro-return impedance deviation, parasitic self-inductance deviation, and parasitic coupled electromagnetic distributed capacitive reactance deviation, is obtained by high-frequency sensitive electrical characteristic testing using flying probes. Each component of impedance, inductance, and capacitance in the electrical deviation vector is divided by a preset reference design parameter, and then converted into a dimensionless normalized electrical deviation vector. A dimensionless normalized process disturbance relative change vector is constructed. The normalized process disturbance relative change vector includes a dimensionless polarization ratio potential drift relative factor divided by a preset initial physical reference value, a thermal stress residual tension field relative exponent, and an alignment axial slip relative scale. A linear mapping relationship is established between the normalized electrical deviation vector and the normalized process perturbation relative change vector to construct a dimensionless physical-electrical coupling transcription matrix. The partial derivative correlation kernels in the physical-electrical coupling transcription matrix correspond to the partial derivative values ​​of the normalized terms of each electrical parameter with respect to the polarization ratio potential drift relative factor, the thermal stress residual tension field relative exponent, and the alignment axial slip relative scale.

8. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the reverse self-calibration step, an inverse matrix mapping analysis is performed on the normalized electrical deviation vector based on the physical-electrical coupling transcription matrix to quantitatively decouple and obtain the dominant feature relative vector of physical failure probability, including: A Tikhonov inverse matrix mapping analysis method with regularization coefficient is introduced to solve the physical-electric coupling transcription matrix and the normalized electrical deviation vector by inverse mapping, and to quantitatively calculate the relative vector of the dominant feature of physical failure probability; wherein, the relative vector of the dominant feature of physical failure probability includes the relative factor of micro-region polarization ratio potential drift in electroplating bath, the relative index of thermal stress residual tension field of laminated structure, and the relative scale of shear physical axial slip. When performing inverse matrix mapping analysis, the dimensionless physical-electrical coupling transcription matrix is ​​transposed and multiplied by itself, and the product of the dimensionless regularization coefficient and the identity matrix is ​​added. The inverse of the square matrix after multiplication and summation is then obtained. The inverse square matrix is ​​then multiplied sequentially by the transpose of the physical-electrical coupling transcription matrix and the dimensionless normalized electrical deviation vector. Finally, the decoupled calculation yields the dominant feature relative vector that is numerically convergent and does not undergo dimension degradation.

9. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, In the reverse self-calibration step, based on the dominant strength of each dimension of the relative physical component in the dominant feature relative vector of the physical failure probability, the corresponding online physical equipment correction and inverse adjustment control behavior is automatically generated and executed, including: If the polarization ratio potential drift relative factor in the relative vector of the physical failure probability dominant feature exceeds the first preset limit value, it is determined to be a polarization dominant failure. The control system automatically increases the reverse current ratio of bidirectional reverse pulse electroplating, and before pattern electroplating, it automatically controls the addition valve to increase the flow rate of polyethylene glycol additive in the electroplating tank, so as to remove the excess copper thickness at the edge of the micro blind hole through the peeling effect of reverse ultra-pulse discharge electrolytic polarization on the micro area. If the relative index of the thermal stress residual tension field in the relative vector of the dominant physical failure probability exceeds the second preset limit value, it is determined to be a stress-dominated failure. The control system starts phase change release and feeds back the temperature drop rate gradient from the fourth temperature zone to the rapid cooling period of the control laminating equipment. If the relative scale of axial slip in the relative vector of the dominant feature of the physical failure probability exceeds the third preset limit value, it is determined to be an alignment error-dominated failure. The control system issues a correction gain vector to adjust the image distortion scaling factor matrix of the laser direct imaging (LDI) camera, thereby compensating for the camera distortion amplification operator in subsequent imaging and preventing asymmetric torsional shear damage.

10. The method for full-process data traceability and control of circuit boards according to claim 1, characterized in that, It also includes a cross-module collaborative promotion step based on the state diagnosis step, the copper plating control step, and the reverse self-calibration step, including: First, if the working state of the drill bit is determined to be a worn and aged state, the degree of drill bit wear characterized by the current deviation index variable resistance factor is fed forward as the initial value of hole wall roughness to the copper plating control step. Based on the degree of drill bit wear, the spatial diffusion attenuation operator in the dynamic boundary model is actively corrected, thereby increasing the equipment excitation frequency in advance and increasing the flow velocity of the physical valve jet at the bottom of the tank, eliminating the dead zone impedance of the laminar boundary layer caused by the rough burrs on the hole wall. Second, the multi-layer physical layer alignment target offset data measured in the preceding lamination and alignment processes is fed forward as a positioning constraint to the reverse self-calibration step; if the measured alignment deviation is lower than the preset alignment deviation threshold, the column vectors corresponding to the axial slip in the physical-electric coupling transcription matrix are set to zero during the inverse matrix mapping analysis to complete matrix dimensionality reduction and pruning, thereby improving the convergence speed of the reverse decoupling calculation and the diagnostic positioning accuracy; Third, in the copper plating control step, the non-uniformity index is stably controlled within a preset safety range by adaptively adjusting the mechanical convection scale and chemical reaction rate in the chemical copper plating tank, and the obtained product performance non-uniformity distribution is used to adjust the reference state of the next batch of processed boards. The reference health resistance current constant lookup table of the next batch in the state diagnosis step is adaptively fine-tuned to reduce the random polarization noise of the system and improve the service life of the drill bit. Fourth, in the reverse self-correction step, if the relative axial slip obtained by decoupling shows a monotonically increasing trend, the relative axial slip is fed back to the state diagnosis step. By generating an absolute origin coordinate deviation axis compensation amount and injecting it into the position servo control channel of the CNC drilling machine for coordinate compensation, the cutting origin alignment position correction is achieved, preventing the relative tilting and off-center loading of the drill bit axis in the next batch. If the relative value of the polarization factor obtained by decoupling exceeds the preset polarization threshold, the relative value of the polarization factor is fed back to the copper plating control step, and the first-order correction parameters of the initial reaction rate and spatial diffusion decay operator in the dynamic model are periodically re-injected.

Images