PCB back drilling adaptive depth control system and method

Abstract

The application discloses a PCB back-drilling adaptive depth control system and method, and relates to the technical field of PCB machining.The system comprises a multi-source data acquisition module, an interface recognition algorithm module, a drill tip state dynamic inversion module, a feed compensation amount solving module, a variable impedance segmented control module and a processing quality online checking module.The application identifies the material layer interface by fusing the spindle current and axial vibration double characteristics, calculates the interface recognition confidence in real time, and locates the actual interface position.Combining the cumulative cutting parameters and the spindle temperature, the drill tip wear and thermal elongation are dynamically inverted, the Z-axis feed control instruction of multi-factor compensation is solved, the processing stage is divided according to the interface recognition confidence, the PID parameters and the feed speed are dynamically switched to realize segmented control, the residual stub length is estimated online after single-hole processing, the model self-learning is triggered by the batch CPK evaluation, and the depth control closed loop is formed.The application can effectively improve the back-drilling depth control precision and batch processing consistency.

Description

Technical Field

[0001] This invention relates to the field of machining technology, specifically to a PCB back-drilling adaptive depth control system and method. Background Technology

[0002] As the electronics and information industry develops towards higher speeds and higher densities, the signal transmission rate of multilayer PCBs continues to increase. Signal reflection, impedance discontinuities, and crosstalk issues caused by via residue can severely affect the integrity of high-speed signals. Back-drilling, which involves secondary drilling of through-holes to remove excess copper layers outside the target layer, is a core process for controlling residue length and ensuring the electrical performance of high-speed PCBs.

[0003] Currently, most back drilling depths in the industry are set using a fixed depth method, calculating the drilling depth based on the theoretical layer thickness in the PCB design file, and only using board height detection for basic compensation. Some solutions use a single current or vibration signal to assist in determining the contact position on the board surface, but due to factors such as the thickness tolerance of the board lamination layer, spindle load fluctuations, and material uniformity differences, the misjudgment rate of the interface between thin layers and high-frequency dielectric layers is relatively high, and the actual residual pile length fluctuates greatly, making it difficult to adapt to the processing requirements of high-end multilayer boards and special boards.

[0004] Furthermore, existing solutions mostly rely on fixed empirical coefficients and periodic manual calibration to address drill tip wear and spindle thermal expansion, which cannot provide real-time dynamic correction. During batch processing, depth drift is prone to accumulate, resulting in poor processing consistency. Feed control generally uses fixed PID parameters and a single feed rate, which cannot adapt to the load characteristics of different processing stages. It is difficult to balance efficiency and hole wall quality, and defects such as burrs at the hole opening and scratches in the medium layer are easily generated. Quality verification relies on offline sampling inspection, which cannot achieve online evaluation and closed-loop correction for each hole. Abnormal working conditions are difficult to detect in a timely manner, which can easily lead to batch defects. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a PCB back-drilling adaptive depth control system and method, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: a PCB back drilling adaptive depth control system, including a controller electrically connected to the Z-axis servo motor of the back drilling machine and a sensor group for collecting processing parameters, characterized in that it further includes a multi-source data acquisition module, an interface recognition algorithm module, a drill tip state dynamic inversion module, a feed compensation calculation module, a variable impedance segmented control module, and a processing quality online verification module; The multi-source data acquisition module is electrically connected to the sensor group and is configured to perform hardware synchronization triggering and timestamp alignment on the multi-source data acquired by the sensor group, and to cache the time-synchronized processing data. The interface recognition algorithm module is electrically connected to the multi-source data acquisition module and is configured to construct a heterogeneous feature mapping model of spindle load current and drill bit axial vibration, extract current mutation rate and high-frequency vibration energy entropy in real time, calculate interface recognition confidence, and determine the material layer interface crossing time and actual interface position. The drill tip state dynamic inversion module is electrically connected to the multi-source data acquisition module and is configured to combine the cumulative drilling cutting parameters and spindle temperature change data to establish an inversion model of the effective cutting length of the drill tip and calculate the drill tip wear and thermal elongation in real time. The feed compensation calculation module is electrically connected to the interface recognition algorithm module and the drill tip state dynamic inversion module, respectively. It is configured to calculate the Z-axis feed control command based on the theoretical back drilling target depth, drill tip wear, thermal elongation, and interface position deviation, and send it to the controller. The variable impedance segmented control module is electrically connected to the interface recognition algorithm module and the controller, respectively, and is configured to divide the back drilling process into multiple control stages according to the interface recognition confidence level, and dynamically switch the PID control gain and feed rate parameters of each stage. The online processing quality verification module is electrically connected to the multi-source data acquisition module and the feed compensation calculation module, respectively. It is configured to calculate the estimated value of the actual residual pile length and the process capability index CPK after the single hole processing is completed, and feed the quality deviation to the feed compensation calculation module to achieve closed-loop correction.

[0007] Preferably, the interface recognition algorithm module calculates the interface recognition confidence level using the following formula. : ; By integrating the characteristics of sudden changes in spindle load current and the characteristics of high-frequency vibration energy of drill bit, the reliability of drill bit crossing material interface is quantified by weighted summation. The value ranges from 0 to 1, with higher values ​​indicating higher reliability of interface crossing.

[0008] In the formula: For current characteristic weighting coefficients, The vibration characteristic weighting coefficient satisfies The weighting coefficients are preset according to the PCB board type; This represents the sudden change rate of the spindle load current within the current 10ms sliding time window, reflecting the instantaneous rate of change of the spindle load. The average current mutation rate within the current material layer is taken as the reference current mutation rate during the period of no interface change in the last 50ms, and used as the load fluctuation reference within the same material layer. The preset threshold for determining sudden current change is the typical current change during material layer interface crossing. It represents the energy entropy of the 10-20kHz high-frequency vibration along the drill bit axis within the current sliding time window, reflecting the degree of disorder in the drill bit cutting vibration; The reference vibration energy entropy within the current material layer is given by the reference value of vibration energy entropy under stable cutting conditions of the same material. The preset vibration energy threshold is the typical vibration energy change during material layer interface crossing; when When determining that the drill bit has penetrated the material layer interface, among which The confidence threshold for interface recognition is set between 0.85 and 0.95; the coordinates of the grating encoder corresponding to the moment of interface crossing are determined as the actual interface position of that layer. .

[0009] Preferably, the drill tip state dynamic inversion module calculates the drill tip wear in real time using the following two formulas. With drill tip thermal elongation : ; ; The wear length of the drill tip is estimated based on the cumulative cutting work of the drill bit. The greater the cumulative cutting work, the higher the wear of the drill tip.

[0010] In the formula: This is the drill tip wear rate coefficient, which is positively correlated with the drill bit material and PCB board type. The higher the material hardness, the larger the corresponding coefficient. This represents the total number of holes drilled by the current drill bit. For the first The average axial cutting force of the spindle during the machining of a hole is calculated from the average spindle load current of that hole through a preset current-cutting force mapping table; For the first The actual cutting time of each hole, that is, the time from when the drill bit contacts the plate surface to when the feed stops; By combining the steady-state temperature difference of the spindle with the rate of temperature change, the axial elongation of the spindle and drill tip due to thermal expansion is estimated, including both steady-state thermal elongation and dynamic temperature change hysteresis correction.

[0011] In the formula: is the linear coefficient of thermal expansion of the main spindle, and is the axial elongation corresponding to a unit temperature change of the main spindle. This is the temperature change rate correction coefficient, used to compensate for the dynamic thermal deformation hysteresis effect when the temperature changes rapidly. for Real-time temperature of the front bearing housing of the spindle; The spindle reference temperature when the machine tool is started from a cold state; for The rate of change of the principal axis temperature at any given time was calculated by linear fitting of 10 consecutive temperature sampling points.

[0012] Preferably, the feed compensation calculation module calculates the Z-axis feed control command using the following formula. : ; Based on the target depth, drill tip wear compensation, thermal elongation compensation and interface position correction are superimposed in sequence to obtain the final Z-axis feed target coordinates, realizing multi-factor adaptive compensation of back drill depth.

[0013] In the formula: The theoretical back-drilling target depth is generated based on the PCGBerber design file, with the PCB board surface as the zero point of the Z-axis coordinate. To calculate the real-time drill tip wear, which causes the effective length of the drill tip to shorten due to wear, the wear is subtracted from the target depth to ensure that the actual cutting depth meets the target. To calculate the real-time thermal elongation of the drill tip, since thermal elongation causes the effective length of the drill tip to increase, the thermal elongation is superimposed on the target depth to avoid exceeding the drilling depth limit. This is the interface position correction amount, calculated by weighting the actual interface position and the theoretical interface position based on the interface recognition confidence level. It is used to correct the depth error caused by the manufacturing deviation of the sheet thickness.

[0014] Preferably, the variable impedance segmented control module divides the back drilling process into four control stages, and the control logic for each stage is as follows: Rapid feed phase: Interface recognition confidence level when the drill bit is not in contact with the PCB board surface. The first set of PID parameters is used, namely high proportional gain. Low integral gain Low differential gain At the preset maximum feed rate Quickly move to a safe height of 0.5mm above the board surface; Soft contact phase: when When the drill bit is close to the PCB board surface, it automatically switches to the second set of PID parameters, where the proportional gain... for 0.3-0.5 times the integral gain for 0.2 times, differential gain for Twice as fast, feed rate reduced to ,Right now 0.1-0.2 times the normal value, control the drill bit to slowly contact the board surface, and avoid burrs at the hole opening and copper layer flipping caused by impact; Layered cutting stage: when When the drill bit enters the corresponding material layer, the PID parameters and feed rate are switched according to the material type: when cutting copper layers, the third set of PID parameters, i.e., medium gain, is used. Medium integral gain Medium differential gain The feed rate is ,Right now 0.4-0.6 times; when cutting the FR-4 / PTFE dielectric layer, the fourth group of PID parameters is used, in which the proportional gain is 0.4-0.6 times; for 0.6 times, integral gain for 0.5 times, differential gain for 1.5 times that, feed rate is ,Right now 0.3-0.4 times, reducing hole wall scratches and nail head defects in the medium layer; Final Drilling Stop Phase: When the grating encoder detects the actual position of the drill bit Enter When the spindle is within the specified range, a feed stop command is triggered. After maintaining spindle rotation for 0.1-0.3 seconds for finishing, the Z-axis is controlled to move at a speed of... Quickly raise the drill to a safe height to complete the single-hole machining.

[0015] Preferably, the multi-source data acquisition module includes a hardware synchronization triggering unit, a sensor group, and a data buffer unit; The sensor group includes: a spindle load current sensor, which is a Hall current sensor installed in the three-phase drive circuit of the spindle motor, with a sampling frequency ≥20kHz, used to collect the real-time operating current of the spindle motor; a piezoelectric axial vibration sensor, installed on the end face of the spindle head near the drill chuck, with a sampling frequency ≥50kHz and a range of ±50g, used to collect the axial vibration acceleration signal of the drill bit; a PT100 platinum resistance temperature sensor, attached to the outer wall of the front bearing housing of the spindle, with a sampling frequency ≥10Hz and an accuracy of ±0.1℃, used to collect the temperature of the spindle bearing; and a high-precision grating position encoder, installed at the output end of the Z-axis servo motor, with a resolution of 0.05μm, used to detect the absolute Z-axis position coordinates of the drill bit in real time. The hardware synchronization triggering unit uses the Z-phase zero-position pulse of the grating position encoder as the global reference trigger signal, and synchronously triggers all sensors to perform synchronous sampling through hardware wiring to ensure that the time stamp phase difference of each sensor data is no greater than 1μs. The data caching unit uses a FIFO sliding time window to cache multi-source data. The time window length is 100ms and the sliding step is 10ms, providing continuous time-synchronized data segments for interface recognition and state inversion.

[0016] Preferably, the quality verification process of the online processing quality verification module is as follows: After single-hole machining stops, obtain the actual stopping position of the drill bit. The actual interface location of the target layer of the hole. Current drill tip compensation amount and By combining the deviation between the theoretical target depth and the theoretical interface position, the estimated value of the actual residual pile length is calculated. ; Will Target residual pile length as required by design Compare and calculate the deviation in the length of the remaining pile. ; like Then Feedback is sent to the feed compensation calculation module to fine-tune the interface correction weight of the next hole in the same network and at the same level. The adjustment step is 0.01, and the single adjustment range does not exceed 0.05. After completing the machining of 100 consecutive holes of the same type, the process capability index CPK of the residual pile length of the batch of holes is calculated. When CPK < 1.33, the drill tip model self-learning process is triggered; when CPK ≥ 1.67, the feed rate is automatically increased slightly to improve machining efficiency.

[0017] Preferably, the PCB back-drill adaptive depth control system further includes a drill tip model self-learning module, and the self-learning process of the drill tip model self-learning module is as follows: After receiving the self-learning trigger signal from the online processing quality verification module, the residual pile length deviation data of the most recent 30 consecutive processing holes are extracted, and the average residual pile deviation is calculated after removing abnormal hole data. Extract the cumulative cutting work parameters of the effective holes in this batch. Based on the ratio of the average residual pile deviation to the cumulative cutting work, and in conjunction with the preset learning rate, adjust the drill tip wear rate coefficient. Perform online corrections; The corrected drill tip wear rate coefficient is updated to the drill tip condition dynamic inversion module, replacing the original coefficient for subsequent borehole wear calculation. Real-time cumulative total number of boreholes ,when When the drill bit reaches 80% of its preset lifespan threshold for the corresponding model, a drill bit replacement warning will pop up on the human-machine interface; when When the lifespan threshold is reached, the machining function is locked until a new drill bit is replaced and the parameters are reset. The PCB back-drilling adaptive depth control system also includes a board parameter library module. This module pre-stores processing parameter sets for various commonly used PCB boards, including interface recognition weight coefficients for FR-4 epoxy board, PTFE high-frequency board, hydrocarbon ceramic board, and PI flexible board. and Initial values ​​of PID parameters and initial coefficients of drill tip wear rate for each control stage Thermal elongation coefficient and Default feed rate; Before processing begins, the user selects the type of PCB to be processed and the target residual length via a touchscreen human-machine interface. The system then automatically loads all initial parameters for the corresponding PCB and automatically calculates the theoretical target depth. There is no need to manually adjust parameters one by one; at the same time, the system supports users to import more than 50 trial processing sample data of new boards, and automatically generate parameter groups of new boards through the built-in least squares training algorithm and save them to the parameter library to achieve rapid adaptation of new boards.

[0018] Preferably, the PCB back-drilling adaptive depth control system further includes an abnormal working condition handling module, and the processing flow of the abnormal working condition handling module is as follows: Drill tip chipping / excessive wear anomaly identification: Real-time monitoring of high-frequency vibration energy entropy With current mutation rate ,when and When the drill tip chipping is detected, it is determined to be an abnormality; when 10 consecutive holes... When the calculated value exceeds 90% of the maximum allowable wear of the drill bit, it is determined that the drill bit is worn and has failed. Both types of abnormalities immediately send an emergency retraction command to the controller, control the Z-axis to lift the drill bit to a safe height at the maximum speed, stop the spindle rotation, mark the current hole position as an abnormal hole on the human-machine interface, and trigger an audible and visual alarm. Spindle overheating anomaly detection: Real-time monitoring of spindle temperature ,when At this time, the system automatically sends parameter adjustment commands to the variable impedance segmented control module, reducing the current feed rate to 50% of the normal value for the corresponding stage, and simultaneously activating the spindle cooling system's strong cooling mode, waiting for the temperature to drop back to normal. Then, restore the original feed rate and cooling mode; Handling internal anomalies in the board material: Confidence level of interface recognition over three consecutive sliding time windows. If the fluctuation range exceeds 0.5 and there is no corresponding design interface, it is determined to be an abnormality such as internal delamination, impurities, or cavities in the material. The current Z-axis position is recorded, and the feed speed is automatically reduced. 50% slow feed through the abnormal area; if the abnormality lasts for more than 1 second or the vibration value exceeds 80% of the range, immediately pull up the drill and mark the hole as abnormal.

[0019] This invention also proposes a PCB back-drilling adaptive depth control method, comprising the following steps: S1. Initialization Preparation Phase: The system performs a power-on self-test. The user selects the type of PCB board to be processed and loads the Gerber processing file. The system automatically loads the corresponding processing parameters, performs zero-point calibration of all sensors, controls the spindle to move to the initial safe height above the first hole position, and resets the current drill bit's cumulative drilling count. Initialize drill tip wear Thermal elongation ; S2, Multi-source data synchronous acquisition stage: Start the spindle to rotate to the set speed, and synchronously acquire spindle load current, drill bit axial vibration, spindle temperature and Z-axis position data through the hardware synchronous trigger unit, and cache the synchronous data in real time through the sliding time window; S3. Real-time Interface Recognition Stage: Calculate the real-time current mutation rate based on current and vibration data within the sliding time window. With high-frequency vibration energy entropy The confidence level is identified by combining the current material layer reference value with the calculation interface. ,when Determine the interface between material layers and record the actual interface position. And update the current material layer status; S4. Drill tip condition inversion stage: based on cumulative borehole count The cutting force and cutting time of the machined holes, the current spindle temperature and temperature change rate are updated in real time to calculate the current drill tip wear. With thermal elongation ; S5. Feed compensation command calculation stage: combined with the theoretical target depth of the current hole. Drill tip wear, thermal elongation, correction for residual pile deviation from the previous hole, and correction for interface position. Finally, the Z-axis feed control command is generated. ; S6. Variable impedance segmented feed stage: Confidence level is identified based on the real-time interface. Switch the PID parameters and feed rate of the corresponding control stage to control the Z-axis servo motor to drive the drill bit to complete the feed cutting; judge in real time whether the actual position of the drill bit has reached the target range, stop the feed and perform a short finishing after reaching the target range, and control the drill bit to quickly lift the drill bit to a safe height. S7. Single Hole Quality Verification Stage: After a single hole is drilled, the estimated value of the actual residual pile length and the deviation of the residual pile are calculated. If the deviation exceeds the standard, the compensation coefficient is finely adjusted. After accumulating 100 holes, the batch CPK value is calculated. If the CPK does not meet the standard, the drill tip model self-learning process is triggered to update the wear rate coefficient. S8. Real-time anomaly judgment stage: Vibration, current and temperature data are monitored in real time throughout the entire processing process. If an abnormal condition is triggered, the corresponding abnormal handling process is executed immediately, and processing is resumed after troubleshooting. S9. Cyclic Machining Stage: After completing the entire process for a single hole, locate the next machining hole position according to the Gerber file, and accumulate the number of drilled holes. Return to step S2 to process the next hole until all holes are processed, and output a processing quality statistics report.

[0020] This invention provides an adaptive depth control system and method for PCB back drilling, which has the following advantages: This invention integrates multi-source features of spindle current and axial vibration to identify the board interface, and simultaneously establishes a dynamic inversion model of the drill tip state. The multi-feature fusion algorithm significantly improves the layer interface identification accuracy, compensates for board layer thickness tolerances, and stabilizes residual pile dimensions; the model calculates and dynamically compensates for drill tip wear and spindle thermal expansion in real time, suppressing batch processing depth drift and improving processing consistency and process capability.

[0021] This invention employs segmented variable parameter PID feed control and incorporates an online quality closed-loop verification mechanism. Speed ​​and control parameters are switched in stages based on interface confidence levels, balancing processing efficiency and hole wall quality to reduce defects such as burrs and media scratches. Single-hole residual piles are calculated online, and compensation parameters are fine-tuned in real time. For batch processing, the wear model is automatically optimized based on CPK, continuously improving control accuracy and reducing manual debugging work.

[0022] This invention integrates real-time handling logic for multiple types of anomalies with a self-adaptive library of sheet metal parameters. It can identify tool breakage, spindle overheating, and internal defects in the sheet metal in real time and automatically execute protective and alarm operations, reducing scrap and equipment wear. It also has built-in standard process parameters for various sheet metals and supports automatic training to generate parameters for new sheet metal samples, shortening the debugging cycle for new products and reducing operational difficulty. Attached Figure Description

[0023] Figure 1 Block diagram of PCB back-drilling adaptive depth control system; Figure 2 Flowchart of PCB back-drilling adaptive depth control method; Figure 3 Characteristic diagram of the segmented control stage for back drilling; Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0025] The PCB back drilling adaptive depth control system of this embodiment is mounted on a CNC PCB high-speed back drilling machine. The back drilling machine is equipped with a Z-axis servo feed unit and a high-speed electric spindle drilling unit. The Z-axis is driven by an absolute value servo motor and controlled by a grating ruler in a closed loop.

[0026] like Figure 1 As shown, this system includes a controller electrically connected to the Z-axis servo motor of the back drill, a sensor group for collecting machining parameters, and six core modules: a multi-source data acquisition module, an interface recognition algorithm module, a drill tip state dynamic inversion module, a feed compensation calculation module, a variable impedance segmented control module, and a machining quality online verification module. It also includes three auxiliary modules: a drill tip model self-learning module, a plate parameter library module, and an abnormal working condition handling module.

[0027] The controller adopts a high-performance embedded motion controller (such as a dedicated motion control card based on ARM+FPGA architecture) to execute servo commands, logic scheduling and peripheral interaction; each algorithm module runs in the controller's built-in processor as a software functional unit, and the modules interact with each other through an internal bus.

[0028] The multi-source data acquisition module is electrically connected to the sensor group and is configured to perform hardware synchronization triggering and timestamp alignment on the multi-source data acquired by the sensor group, and to cache the time-synchronized processing data. The module consists of three parts: a hardware synchronization triggering unit, a sensor group, and a data caching unit.

[0029] Sensor group configuration and installation: Spindle load current sensor: A high-precision Hall current sensor is used and installed in the U-phase circuit of the three-phase drive circuit of the spindle motor. The sampling frequency is ≥20kHz, and in this embodiment it is set to 25kHz with an accuracy of ±0.5%FS. The current signal is converted into a 0-5V standard voltage signal by the signal conditioning circuit and then input to the acquisition unit to collect the working current of the spindle motor in real time, which indirectly reflects the size of the cutting load.

[0030] Piezoelectric axial vibration sensor: An IEPE type piezoelectric accelerometer is used, which is rigidly installed on the front face of the spindle head near the drill chuck via a thread. The mounting surface is ground to ensure flatness ≤0.02mm; the sampling frequency is ≥50kHz, which is set to 50kHz in this embodiment; the range is ±50g; and the sensitivity is 100mV / g. It is used to collect axial vibration acceleration signals during the drilling process of the drill bit to reflect the cutting state and tool state.

[0031] PT100 Platinum Resistance Temperature Sensor: A-grade PT100 platinum resistance thermometer is used, which is mounted in a special mounting groove on the outer wall of the spindle front bearing housing with high-temperature thermally conductive adhesive. The sampling frequency is ≥10Hz, and in this embodiment it is set to 20Hz. The measurement accuracy is ±0.1℃. It is used to collect the real-time temperature of the spindle front bearing and provide data for thermal expansion compensation.

[0032] High-precision grating position encoder: It adopts a closed incremental grating ruler, which is installed next to the Z-axis transmission pair to form a fully closed-loop position detection with a resolution of 0.05μm. It is used to detect the absolute position coordinates of the drill bit's Z-axis in real time; at the same time, it outputs the Z-phase zero position pulse as a global synchronization reference signal.

[0033] The hardware synchronization triggering unit uses the Z-phase zero-position pulse of the grating position encoder as the global reference trigger signal. It outputs a synchronous sampling clock to all sensors through the hardware synchronization triggering circuit, ensuring that the sampling times of the four types of sensors—current, vibration, temperature, and position—are strictly aligned. The phase difference of the timestamps of the data from each sensor is no greater than 1μs, thus eliminating the time offset error of multi-source data from the root and providing a time-consistent data source for subsequent multi-feature fusion algorithms.

[0034] The data caching unit adopts an FPGA+RAM architecture to implement the FIFO sliding time window caching function. The time window length is set to 100ms and the sliding step is 10ms, that is, the window data is updated every 10ms, and all synchronous sampling data within the most recent 100ms is retained each time. The cached data is stored by timestamp index, which supports the upper-layer algorithm module to read the corresponding data on demand according to the time window, providing continuous time synchronization data segments for interface recognition and state inversion.

[0035] The interface recognition algorithm module is electrically connected to the multi-source data acquisition module and is configured to construct a heterogeneous feature mapping model of spindle load current and drill bit axial vibration, extract current mutation rate and high-frequency vibration energy entropy in real time, calculate interface recognition confidence, and determine the material layer interface crossing time and actual interface position.

[0036] The core of this module is a current-vibration dual-feature fusion interface recognition algorithm, and the specific implementation process is as follows: Real-time feature extraction: Current mutation rate calculation: Extract the spindle load current data within the current 10ms window from the sliding time window, and obtain the rate of change of current over time, i.e., the current current mutation rate, through least squares linear fitting. (Unit: A / ms), reflecting the instantaneous rate of change of spindle load. Simultaneously, current data from the most recent 50ms period without interface change within the current material layer is extracted, and the average current abrupt change rate during this period is calculated as the baseline current abrupt change rate. This serves as a reference for normal load fluctuations within the same material layer.

[0037] High-frequency vibration energy entropy calculation: Bandpass filtering is applied to the axial vibration acceleration signal within the current 10ms sliding time window to extract the high-frequency vibration component in the 10-20kHz frequency band; the signal in this frequency band is then subjected to 3-level wavelet packet decomposition to calculate the energy distribution of each sub-band, and finally, the energy entropy of the vibration signal is calculated. This reflects the degree of disorder in drill bit cutting vibration. Simultaneously, the average vibration energy entropy under the current stable cutting state of the material layer is taken as the benchmark vibration energy entropy. .

[0038] Interface recognition confidence calculation: The confidence level for interface identification is calculated by weighted fusion of current abrupt change characteristics and vibration energy entropy characteristics. The calculation formula is: ; In the formula: For current characteristic weighting coefficients, The vibration characteristic weighting coefficient satisfies The weighting coefficients are retrieved from the PCB material parameter library based on the PCB material type; in this embodiment, when processing FR-4 epoxy board, the weighting coefficients are... , When processing PTFE high-frequency boards, take , ; The preset threshold for determining sudden current change is the typical current change during material layer interface crossing, and the threshold is the copper-dielectric interface of the FR-4 substrate. The value is 0.8A; The preset vibration energy threshold represents the typical vibration energy change during material layer interface crossing, corresponding to the FR-4 plate. The value is 0.6.

[0039] Confidence The value ranges from 0 to 1, with higher values ​​indicating higher reliability in determining whether the drill bit has crossed the material interface. This formula normalizes two heterogeneous features with different dimensions—current abrupt changes and vibration energy changes—to map both types of features to a comparable range around 0-1. Then, through weighted fusion using weighting coefficients, it comprehensively quantifies the reliability of the drill bit crossing the material interface. This effectively suppresses misjudgments caused by electromagnetic interference and cutting fluctuations from a single signal, improving the robustness of interface recognition.

[0040] Taking the FR-4 epoxy board processing scenario in this embodiment as an example, the specific calculation process is as follows: Given the spindle load current mutation rate within the current sliding time window. The rate of change of reference current within the same material layer Threshold for determining sudden current changes Current high-frequency vibration energy entropy Reference vibration energy entropy Vibration energy determination threshold Current characteristic weights Vibration characteristic weights .

[0041] The first step is to calculate the normalized value of the current characteristic: ; The second step is to calculate the normalized values ​​of the vibration characteristics: ; The third step is to calculate the confidence level of interface recognition: ; Calculated Greater than the preset confidence threshold Therefore, it is determined that the drill bit is currently crossing the material layer interface.

[0042] Interface determination and position recording: Preset interface recognition confidence threshold In this embodiment (Value range: 0.85-0.95). The confidence level is calculated in real time. When the drill bit is determined to be crossing the material layer interface at that moment, the Z-axis coordinate value collected by the grating encoder at that moment is immediately latched and recorded as the actual interface position of the material layer. .

[0043] The drill tip dynamic inversion module is electrically connected to the multi-source data acquisition module and is configured to combine cumulative drilling cutting parameters and spindle temperature change data to establish an inversion model of the effective cutting length of the drill tip, and calculate the drill tip wear and thermal elongation in real time. The module corrects the effective length of the drill tip in real time from two dimensions: wear loss and thermal deformation elongation.

[0044] Calculation of drill tip wear: The wear length of the drill tip is estimated based on the cumulative cutting work of the drill bit. The greater the cumulative cutting work, the higher the wear of the drill tip. (Drill tip wear amount) The calculation formula is: ; In the formula: This is the drill tip wear rate coefficient, in units of... The coefficient is positively correlated with the drill bit material and the PCB board type; the higher the material hardness, the larger the coefficient. In this embodiment, when machining FR-4 board with a carbide drill bit... Initial value ; This represents the total number of holes drilled by the current drill bit. For the first The average axial cutting force of the spindle during the machining of each hole (unit: N) is calculated from the average spindle load current of the hole through a preset current-cutting force mapping table, which is established through previous calibration experiments. For the first The actual cutting time (in seconds) of each hole, that is, the time from when the drill bit contacts the plate surface to when the feed stops, is calculated from the triggering time of the grating position signal.

[0045] Based on the fundamental laws of abrasive wear, this formula characterizes the single-hole cutting load by multiplying the single-hole cutting force by the cutting time. The total cumulative cutting work of the drill bit is obtained by accumulating the cutting loads of all machined holes. The wear of the drill tip is linearly positively correlated with the cumulative cutting work, which can reflect the wear of the drill bit in real time as the machining process progresses.

[0046] Taking the processing scenario in this embodiment as an example, the specific calculation process is as follows: Known drill tip wear coefficient The current total number of holes drilled by the drill bit The average axial cutting force of a single hole was statistically obtained. Average actual cutting time per hole .

[0047] The first step is to calculate the total cumulative cutting work: ; The second step is to calculate the cumulative wear of the drill tip: ; The current cumulative wear of the drill tip is calculated to be: .

[0048] Calculation of thermal elongation of drill tip: By combining the steady-state temperature difference of the spindle and the rate of temperature change, the axial elongation of the spindle and drill tip due to thermal expansion is estimated, including both steady-state thermal elongation and dynamic temperature change hysteresis correction. Drill tip thermal elongation. The calculation formula is: ; In the formula: The linear coefficient for main shaft thermal elongation, in units of , is the axial elongation corresponding to a unit temperature change of the spindle, which is taken as a value in this embodiment. ; This is the correction factor for the rate of temperature change, in units of... This is used to compensate for the dynamic thermal deformation hysteresis effect when the temperature changes rapidly. In this embodiment, the value is taken as... ; for Real-time temperature of the front bearing housing of the spindle (unit: ); The spindle reference temperature when the machine tool is cold-started (unit: The system automatically collects and records data during initialization. for Spindle temperature change rate at time (unit: The result was obtained through linear fitting of 10 consecutive temperature sampling points.

[0049] The first term in the formula is the steady-state thermal elongation component of the spindle, which is linearly related to the spindle temperature rise and reflects the axial elongation of the spindle and drill tip under thermal equilibrium conditions. The second term is the dynamic temperature change correction component. The faster the temperature change rate, the larger the correction amount. It is used to compensate for the estimation error caused by the lag in temperature sensor measurement and the delay in the transmission of spindle thermal deformation, and improve the accuracy of thermal elongation prediction during the dynamic temperature rise process.

[0050] Taking the processing scenario in this embodiment as an example, the specific calculation process is as follows: Given the linear coefficient of thermal expansion of the spindle Correction factor for rate of temperature change Machine tool cold reference temperature Current real-time temperature of the front bearing housing of the spindle Current spindle temperature change rate .

[0051] The first step is to calculate the steady-state thermal elongation: ; The second step is to calculate the dynamic correction amount: ; The third step is to calculate the total thermal elongation: ; The calculated total thermal elongation of the drill tip is as follows: .

[0052] The feed compensation calculation module is electrically connected to the interface recognition algorithm module and the drill tip state dynamic inversion module, respectively. It is configured to calculate the Z-axis feed control command based on the theoretical back-drilling target depth, drill tip wear, thermal elongation, and interface position deviation, and send it to the controller.

[0053] This module uses the theoretical depth generated from the PCGBerber design file as a basis, and sequentially overlays drill tip wear compensation, thermal expansion compensation, and interface position correction to obtain the final Z-axis feed target coordinates, achieving multi-factor adaptive compensation for back drilling depth. Z-axis feed control commands. The calculation formula is: ; In the formula: To determine the theoretical target depth for back drilling based on the PCBGerber design file, with the PCB board surface as the zero point of the Z-axis coordinate, the unit is... ; For real-time calculation of drill tip wear: Wear will shorten the effective length of the drill tip, so the wear amount is subtracted from the target depth to ensure that the actual cutting depth meets the target. The real-time calculation of drill tip thermal elongation: thermal elongation will increase the effective length of the drill tip, so the thermal elongation is added to the target depth to avoid exceeding the drilling depth limit; This is the adjustment amount for the interface position, based on the actual interface position. Location of the theoretical interface The deviation, combined with the confidence level of interface recognition The weighted calculation yields the depth error used to correct for manufacturing tolerances in the sheet layer thickness. The calculation formula is as follows: ; The above feed control command formula uses the theoretical target depth as a benchmark and achieves adaptive depth correction through the superposition of positive and negative compensation quantities of three types: drill tip wear shortens the effective cutting length, so the wear amount is subtracted to increase the feed depth and ensure that the cutting depth meets the target; drill tip thermal elongation increases the effective cutting length, so the thermal elongation amount is superimposed to raise the feed endpoint and avoid exceeding the drilling depth limit; the interface position correction amount is used to offset the deviation between the actual layer thickness of the plate and the theoretical value, ensuring that the residual pile length meets the design requirements.

[0054] Based on the aforementioned calculation results, the specific calculation process is as follows: Known theoretical back-drilling target depth Real-time drill tip wear Real-time thermal elongation of drill tip ; Actual interface location of the target layer Corresponding to the theoretical interface position Current interface recognition confidence level .

[0055] The first step is to calculate the interface position correction amount: ; The second step is to calculate the Z-axis feed control command: The final Z-axis feed target coordinates are calculated as follows: .

[0056] Meanwhile, this module receives residual pile deviation signals from the online processing quality verification module, fine-tunes the interface correction weights, and achieves closed-loop correction of depth control.

[0057] The variable impedance segmented control module is electrically connected to the interface recognition algorithm module and the controller, respectively. It is configured to divide the back drilling process into multiple control stages based on the interface recognition confidence level, and dynamically switch the PID control gain and feed rate parameters of each stage.

[0058] like Figure 3 As shown in the diagram, the PCB surface is taken as the zero point of the Z-axis coordinate, and the drilling feed direction (downward) is taken as the positive Z-axis direction. It fully presents the stage division and depth compensation logic of the entire stroke of single-hole back drilling. The diagram is divided into four working zones along the Z-axis: the idle rapid feed zone, the board surface soft contact transition zone, the material layer cutting zone, and the final drilling finishing stop zone. The actual positions of the surface copper layer interface, FR-4 / PTFE dielectric layer, and target copper layer interface are marked inside the board. At the same time, the spatial correspondence between the theoretical back drilling target depth, drill tip wear compensation, thermal expansion compensation, interface position correction, and the final actual residual pile length is marked, which intuitively reflects the control effect of multi-factor compensation on the residual pile accuracy.

[0059] This module divides the single-hole back-drilling process into four control stages and identifies the confidence level based on the real-time interface. The system automatically switches control parameters to match the load characteristics at each stage, balancing processing efficiency and hole wall quality. The control logic for each stage is as follows: Rapid feed phase: Interface recognition confidence level when the drill bit does not contact the PCB board surface This is determined to be the rapid feed phase.

[0060] This stage uses the first set of PID parameters: high proportional gain. Low integral gain Low differential gain In this embodiment , , The Z-axis feeds at a preset maximum speed. Rapid movement, in this embodiment The drill bit quickly moves to a safe height of 0.5mm above the PCB board surface to ensure efficient idle feed.

[0061] Soft contact phase: when At that time, it is determined that the drill bit is close to the PCB board surface and enters the soft contact stage.

[0062] This stage automatically switches to the second set of PID parameters: proportional gain. for 0.3-0.5 times, in this embodiment we take 0.4 times, that is... Integral gain for 0.2 times that is Differential gain for 2 times The feed rate decreased to ,for 0.1-0.2 times, in this embodiment we take 0.15 times, that is... By reducing the proportional gain and increasing the differential damping, the drill bit is controlled to slowly contact the board surface, avoiding burrs at the hole opening and copper layer flanging defects caused by impact.

[0063] Layered cutting stage: when When the drill bit traverses the interface and enters the corresponding material layer, it enters the layered cutting stage. The system switches the corresponding PID parameters and feed rate according to the identified material type. When cutting the copper layer: use the third set of PID parameters, i.e., medium gain. Medium integral gain Medium differential gain In this embodiment , , Feed rate for 0.4-0.6 times, in this embodiment we take 0.5 times, that is... It is adapted to the high cutting load characteristics of copper layers.

[0064] When cutting FR-4 / PTFE dielectric layers: the fourth set of PID parameters is used, where the proportional gain... for 0.6 times that is Integral gain for 0.5 times that is Differential gain for 1.5 times that is Feed rate for 0.3-0.4 times, in this embodiment we take 0.35 times, that is... By reducing the feed rate and increasing the differential damping, the hole wall scratches and nail head defects in the medium layer are reduced.

[0065] Final drilling cessation phase: When the grating encoder detects the actual position of the drill bit Enter When the target range is reached, a feed stop command is triggered, and the spindle rotation is maintained for 0.1-0.3s (0.2s in this embodiment) for finishing to smooth the end face of the residual pile at the bottom of the hole; then the Z-axis is controlled to move at a speed Quickly raise the drill to a safe height to complete the single-hole machining.

[0066] The online processing quality verification module is electrically connected to the multi-source data acquisition module and the feed compensation calculation module, respectively. It is configured to calculate the estimated value of the actual residual pile length and the process capability index CPK after the single hole processing is completed, and feed the quality deviation to the feed compensation calculation module to achieve closed-loop correction.

[0067] The quality verification process for this module is as follows: Single-hole residual pile length estimation: After single-hole machining stops, obtain the actual stopping position of the drill bit. The actual interface location of the target layer of the hole. Current drill tip wear With thermal elongation By combining the deviation between the theoretical target depth and the theoretical interface position, the estimated value of the actual residual pile length is calculated. The calculation formula is: ; This formula uses the actual interface position of the target layer as a benchmark, deducts the actual feed depth of the drill bit and adds the influence of the drill tip state compensation, to obtain the final estimated value of the remaining pile length, which can realize online estimation of single hole quality without damaging the PCB.

[0068] Based on the aforementioned parameters, the specific calculation process is as follows: Known actual interface location of the target layer The actual stopping position of the drill bit Current drill tip wear Thermal elongation of drill tip .

[0069] The first step is to calculate the equivalent value of the actual depth of cut: ; The second step is to calculate the estimated actual residual pile length: ; The estimated actual residual pile length of the hole is calculated to be: .

[0070] Calculation of residual pile deviation: The estimated value of the actual residual pile length is calculated. Target residual pile length as required by design Compare and calculate the deviation in the length of the remaining pile: ; Following the example above, if the design requires a target residual pile length... The deviation in the length of the remaining pile is: ; If the absolute value of the deviation is less than 5μm, there is no need to trigger compensation weight fine-tuning.

[0071] Single-hole deviation fine adjustment: If Then Feedback is sent to the feed compensation calculation module to fine-tune the interface correction weight of the next hole in the same network and at the same level. The adjustment step is 0.01, and the single adjustment range does not exceed 0.05. The accumulation of depth deviation is suppressed by fine-tuning hole by hole.

[0072] Batch quality assessment: After machining 100 consecutive holes of the same type, the residual length data of the holes in that batch is collected, and the process capability index CPK is calculated. When CPK < 1.33, the process capability is deemed insufficient, and a self-learning trigger signal is sent to the drill tip model self-learning module; when CPK ≥ 1.67, the process capability is deemed sufficient, and the feed rate at each stage is automatically increased slightly (by 5%) to improve machining efficiency while ensuring quality.

[0073] This system also includes a drill tip model self-learning module, used for online optimization of the drill tip wear rate coefficient, improving wear inversion accuracy, and enabling full lifecycle management of the drill bit. The self-learning process is as follows: Data extraction and cleaning: After receiving the self-learning trigger signal from the online processing quality verification module, the residual pile length deviation data of the most recent 30 consecutive processed holes are extracted. After removing abnormal hole data using the 3σ criterion, the average residual pile deviation of the effective holes is calculated.

[0074] Online coefficient correction: Extract the cumulative cutting work parameters of the effective holes in this batch, and adjust the drill tip wear rate coefficient based on the ratio of the average residual pile deviation to the cumulative cutting work, combined with a preset learning rate (0.1 in this embodiment). Perform online correction; the correction formula is as follows: ; in For learning rate, This represents the average residual pile deviation.

[0075] The correction formula uses the ratio of average residual pile deviation to cumulative cutting work as the deviation correction benchmark. It combines the learning rate to control the single correction amplitude and updates the wear rate coefficient online through small step iterations to avoid control oscillations caused by excessive single corrections. This allows the wear inversion model to gradually adapt to the actual wear characteristics of the drill bit and the plate.

[0076] Taking the processing scenario in this embodiment as an example, the specific calculation process is as follows: Given the current drill tip wear rate coefficient Preset learning rate After data cleaning, the average residual pile deviation of the most recent 30 effective boreholes The total cumulative cutting power of the effective holes in this batch .

[0077] The first step is to calculate the deviation correction increment: The second step is to calculate the corrected wear rate coefficient: The corrected wear rate coefficient is .

[0078] Parameter update: Adjust the drill tip wear rate coefficient. Update the drill tip state dynamic inversion module to replace the original coefficients for subsequent borehole wear calculations.

[0079] Drill bit life management: The system accumulates the total number of boreholes in real time. ,when When the drill bit reaches 80% of its preset lifespan threshold for the corresponding model, a drill bit replacement warning will pop up on the human-machine interface; when When the lifespan threshold is reached, the machining function is locked until the operator replaces the drill bit and resets the borehole count and wear parameters.

[0080] This system also includes a board material parameter library module, which pre-stores processing parameter groups for various commonly used PCB boards, including interface recognition weight coefficients for FR-4 epoxy boards, PTFE high-frequency boards, hydrocarbon ceramic boards, and PI flexible boards. and Initial values ​​of PID parameters and initial coefficients of drill tip wear rate for each control stage Thermal elongation coefficient and Default values ​​for feed rates at each stage, etc.

[0081] Before processing begins, the user selects the type of PCB to be processed and inputs the target residual length via the touchscreen human-machine interface. The system then automatically loads all initial parameters of the corresponding PCB from the parameter library and automatically calculates the theoretical target depth. There is no need to manually adjust the parameters one by one.

[0082] Meanwhile, the system supports rapid adaptation to new materials: after users import trial processing sample data of more than 50 new materials, the system automatically fits and generates a complete set of parameters for the new materials through the built-in least squares training algorithm and saves them to the parameter library, shortening the process debugging cycle of new products.

[0083] This system also includes an abnormal operating condition handling module, which is used to identify abnormal operating conditions in real time throughout the entire processing process and automatically execute protection strategies to reduce the risk of batch scrapping and equipment damage. The specific handling process is as follows: Identification of abnormal drill tip chipping / excessive wear: Drill tip chipping detection: Real-time monitoring of high-frequency vibration energy entropy With current mutation rate ,when and At that time, it was determined to be an abnormality of drill tip chipping; Excessive wear judgment: When 10 consecutive holes... When the calculated value exceeds 90% of the maximum allowable wear of the drill bit, it is determined that the drill bit has worn out.

[0084] When either of the above two types of abnormalities is triggered, the system immediately sends an emergency retraction command to the controller, controls the Z-axis to raise the drill bit to a safe height at maximum speed, stops the spindle rotation, marks the current hole position as an abnormal hole on the human-machine interface, and triggers an audible and visual alarm to prompt the operator to check and replace the drill bit.

[0085] Spindle overheating anomaly detection: Real-time monitoring of spindle temperature ,when If the system detects an overheating anomaly in the spindle, it automatically sends a parameter adjustment command to the variable impedance segmented control module, reducing the current feed rate to 50% of the normal value for the corresponding stage, and simultaneously activating the spindle cooling system's forced cooling mode; wait for the temperature to drop back to normal. Afterwards, it automatically restores the original feed rate and cooling mode to avoid accuracy drift and lifespan reduction caused by long-term high-temperature operation of the spindle.

[0086] Handling internal anomalies in the board material: Confidence level of interface recognition over three consecutive sliding time windows. If the fluctuation range exceeds 0.5 and there is no corresponding design interface in the Gerber file, it is determined to be an abnormality such as internal delamination, impurities, or cavities in the sheet metal. The system records the current Z-axis position and automatically reduces the feed rate. If the abnormal area is 50% of the range, feed slowly through it; if the abnormality lasts for more than 1 second or the vibration value exceeds 80% of the sensor range, immediately lift the drill and mark the hole as abnormal to avoid drill bit breakage and batch defects.

[0087] like Figure 2 As shown, an example of a PCB back-drilling adaptive depth control method is presented: Based on the above system, the PCB back-drilling adaptive depth control method of this embodiment includes the following steps: S1. Initialization Preparation Phase: After power-on, the system performs a full hardware self-test to confirm normal communication between all sensors, servo units, and spindle units. The user selects the type of PCB board to be processed (e.g., FR-4 epoxy board) through the human-machine interface and loads the Gerber processing file. The system automatically loads the complete set of processing parameters for the corresponding board from the board parameter library, including weighting coefficients, PID parameters, compensation coefficients, feed rates, etc.; performs zero-point calibration on all sensors to eliminate initial zero-point deviation; controls the spindle to move to the initial safe height above the first hole (10mm above the board surface); resets the current drill bit's cumulative drilling count. Initialize drill tip wear Thermal elongation Record the current spindle temperature as the reference temperature. .

[0088] S2. Multi-source data synchronous acquisition stage: Start the spindle to rotate to the set speed (12000rpm in this embodiment). The hardware synchronous triggering unit uses the grating Z-phase pulse as a reference to synchronously trigger all sensor sampling. Real-time acquisition of spindle load current, drill bit axial vibration, spindle temperature, and Z-axis position data. The data caching unit caches the synchronous data in real time with a 100ms sliding time window and a 10ms step size for use by the upper-level algorithm.

[0089] S3, Real-time Interface Recognition Stage: The interface recognition algorithm module extracts current and vibration data from the sliding time window and calculates the current mutation rate within the current 10ms window. Energy entropy of 10-20kHz high-frequency vibration Combined with the current reference current mutation rate of the current material layer Compared with the reference vibration energy entropy Substitute the values ​​into the confidence formula to calculate the confidence level of the real-time interface recognition. ;when When the drill bit traverses a material layer interface, the current grating coordinates are latched as the actual interface position. And update the current material layer state.

[0090] S4. Drill Tip Status Inversion Stage: The drill tip status dynamic inversion module reads the cumulative number of boreholes. Average cutting force of each machined hole With cutting time Substitute the values ​​into the wear formula to calculate the current drill tip wear. Simultaneously read the current spindle temperature. Calculate the rate of temperature change Substitute into the thermal elongation formula to calculate the current thermal elongation of the drill tip. The two compensation values ​​are updated in real time and transmitted to the feed compensation value calculation module.

[0091] S5. Feed Compensation Command Calculation Stage: The feed compensation calculation module reads the theoretical target depth of the current hole. The interface position correction is calculated by combining real-time drill tip wear, thermal elongation, and the deviation between the actual and theoretical interface positions. Simultaneously, the residual pile deviation feedback correction amount from the previous hole is superimposed to generate the Z-axis feed control command. And send it to the controller.

[0092] S6. Variable Impedance Segmented Feed Stage: The variable impedance segmented control module identifies the confidence level based on the real-time interface. The system determines the current machining stage and automatically switches the corresponding PID parameters and feed rate. Based on the control parameters, the controller drives the Z-axis servo motor to move the drill bit and complete the feed cutting. The actual position of the optical encoder is read in real time. ,when Enter When the spindle reaches the specified range, a feed stop command is triggered, maintaining spindle rotation for 0.2 seconds for finishing; subsequently, the Z-axis is controlled to... The drill bit is quickly raised to a safe height to complete the single-hole drilling feed.

[0093] S7. Single-hole quality verification stage: After a single hole is drilled, the online processing quality verification module calculates the estimated value of the actual residual pile length and the residual pile deviation of the hole based on the actual stop position, interface position, and drill tip compensation amount. If the absolute value of the deviation is >5μm, the interface correction weight of the next hole is finely adjusted. After processing 100 holes of the same type, the CPK value of the batch residual pile length is calculated. If CPK <1.33, the drill tip model self-learning process is triggered, and the wear rate coefficient is updated online. If CPK ≥1.67, the feed rate is slightly increased.

[0094] S8. Real-time anomaly judgment stage: Throughout the entire processing process, the abnormal working condition handling module monitors the vibration energy entropy, current mutation rate, and spindle temperature data in real time. If any abnormal condition such as drill tip breakage, spindle overheating, or internal abnormality of the plate is triggered, the corresponding abnormal handling process will be executed immediately. Processing will resume after the operator has investigated the fault and reset it.

[0095] S9. Cyclic Machining Stage: After completing the entire single-hole process, the system locates the next machining hole position based on the hole position coordinates in the Gerber file, accumulating the number of drilled holes. Then return to step S2 to perform the next hole machining cycle; until all holes are machined, the system outputs a machining quality statistics report, including information such as the total number of machined holes, the number of abnormal holes, the residual pile length CPK, and the drill bit wear status.

[0096] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made based on the content of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A PCB back-drilling adaptive depth control system, characterized in that, The system includes a controller electrically connected to the Z-axis servo motor of the back drilling rig and a sensor group for acquiring machining parameters. It is characterized by further including a multi-source data acquisition module, an interface recognition algorithm module, a drill tip state dynamic inversion module, a feed compensation calculation module, a variable impedance segmented control module, and a machining quality online verification module. The multi-source data acquisition module is electrically connected to the sensor group and is configured to perform hardware synchronization triggering and timestamp alignment on the multi-source data acquired by the sensor group, and to cache the time-synchronized processing data. The interface recognition algorithm module is electrically connected to the multi-source data acquisition module and is configured to construct a heterogeneous feature mapping model of spindle load current and drill bit axial vibration, extract current mutation rate and high-frequency vibration energy entropy in real time, calculate interface recognition confidence, and determine the material layer interface crossing time and actual interface position. The drill tip state dynamic inversion module is electrically connected to the multi-source data acquisition module and is configured to combine the cumulative drilling cutting parameters and spindle temperature change data to establish an inversion model of the effective cutting length of the drill tip and calculate the drill tip wear and thermal elongation in real time. The feed compensation calculation module is electrically connected to the interface recognition algorithm module and the drill tip state dynamic inversion module, respectively. It is configured to calculate the Z-axis feed control command based on the theoretical back drilling target depth, drill tip wear, thermal elongation, and interface position deviation, and send it to the controller. The variable impedance segmented control module is electrically connected to the interface recognition algorithm module and the controller, respectively, and is configured to divide the back drilling process into multiple control stages according to the interface recognition confidence level, and dynamically switch the PID control gain and feed rate parameters of each stage. The online processing quality verification module is electrically connected to the multi-source data acquisition module and the feed compensation calculation module, respectively. It is configured to calculate the estimated value of the actual residual pile length and the process capability index CPK after the single hole processing is completed, and feed the quality deviation to the feed compensation calculation module to achieve closed-loop correction.

2. The PCB back-drilling adaptive depth control system according to claim 1, characterized in that, The interface recognition algorithm module calculates the interface recognition confidence level using the following formula. : ； In the formula: For current characteristic weighting coefficients, The vibration characteristic weighting coefficient satisfies ; This represents the spindle load current mutation rate within the current 10ms sliding time window. This represents the reference current abrupt change rate within the current material layer; The preset threshold for determining sudden current changes; The entropy of the high-frequency vibration energy of the drill bit axial direction at 10-20kHz within the current sliding time window; This is the reference vibrational energy entropy within the current material layer; The preset threshold for determining vibration energy; when When determining that the drill bit has penetrated the material layer interface, among which The confidence threshold for interface recognition is set between 0.85 and 0.95; the coordinates of the grating encoder corresponding to the moment of interface crossing are determined as the actual interface position of that layer. .

3. The PCB back-drilling adaptive depth control system according to claim 2, characterized in that, The drill tip condition dynamic inversion module calculates the drill tip wear in real time using the following two formulas. With drill tip thermal elongation : ； ； In the formula: This is the drill tip wear rate coefficient; This represents the total number of holes drilled by the current drill bit. For the first Average axial cutting force of the spindle during the machining of each hole; For the first The actual cutting time for each hole; In the formula: The linear coefficient of thermal elongation of the main shaft; This is the correction factor for the rate of temperature change; for Real-time temperature of the front bearing housing of the spindle; The spindle reference temperature when the machine tool is started from a cold state; for The rate of change of principal axis temperature at any given time.

4. The PCB back-drilling adaptive depth control system according to claim 3, characterized in that, The feed compensation calculation module calculates the Z-axis feed control command using the following formula. : ； In the formula: The theoretical back-drilling target depth is generated based on the PCGBerber design file, with the PCB board surface as the zero point of the Z-axis coordinate. The calculated real-time drill tip wear amount; The calculated real-time thermal elongation of the drill tip; This is the amount of adjustment for the interface position.

5. A PCB back-drilling adaptive depth control system according to claim 4, characterized in that, The variable impedance segmented control module divides the back drilling process into four control stages, and the control logic for each stage is as follows: Rapid feed phase: Interface recognition confidence level when the drill bit is not in contact with the PCB board surface. The first set of PID parameters is used, namely high proportional gain. Low integral gain Low differential gain At the preset maximum feed rate Quickly move to a safe height of 0.5mm above the board surface; Soft contact phase: when When the drill bit is close to the PCB board surface, it automatically switches to the second set of PID parameters, where the proportional gain... for 0.3-0.5 times the integral gain for 0.2 times, differential gain for Twice as fast, feed rate reduced to ,Right now 0.1-0.2 times the normal value, control the drill bit to slowly contact the board surface, and avoid burrs at the hole opening and copper layer flipping caused by impact; Layered cutting stage: when When the drill bit enters the corresponding material layer, the PID parameters and feed rate are switched according to the material type: when cutting copper layers, the third set of PID parameters, i.e., medium gain, is used. Medium integral gain Medium differential gain The feed rate is ,Right now 0.4-0.6 times; when cutting the FR-4 / PTFE dielectric layer, the fourth group of PID parameters is used, in which the proportional gain is 0.4-0.6 times; for 0.6 times, integral gain for 0.5 times, differential gain for 1.5 times that, feed rate is ,Right now 0.3-0.4 times, reducing hole wall scratches and nail head defects in the medium layer; Final Drilling Stop Phase: When the grating encoder detects the actual position of the drill bit Enter When the spindle is within the specified range, a feed stop command is triggered. After maintaining spindle rotation for 0.1-0.3 seconds for finishing, the Z-axis is controlled to move at a speed of... Quickly raise the drill to a safe height to complete the single-hole machining.

6. A PCB back-drilling adaptive depth control system according to claim 5, characterized in that, The multi-source data acquisition module includes a hardware synchronization triggering unit, a sensor group, and a data buffer unit; The sensor group includes: a spindle load current sensor, which is a Hall current sensor installed in the three-phase drive circuit of the spindle motor; a piezoelectric axial vibration sensor, installed on the end face of the spindle head near the drill chuck; a PT100 platinum resistance temperature sensor, which is attached to the outer wall of the front bearing housing of the spindle; and a high-precision grating position encoder, which is installed at the output end of the Z-axis servo motor. The hardware synchronous triggering unit uses the Z-phase zero-position pulse of the grating position encoder as the global reference trigger signal. The data caching unit uses a FIFO sliding time window to cache multi-source data.

7. A PCB back-drilling adaptive depth control system and method according to claim 6, characterized in that, The quality verification process of the online processing quality verification module is as follows: After single-hole machining stops, obtain the actual stopping position of the drill bit. The actual interface location of the target layer of the hole. Current drill tip compensation amount and By combining the deviation between the theoretical target depth and the theoretical interface position, the estimated value of the actual residual pile length is calculated. ; Will Target residual pile length as required by design Compare and calculate the deviation in the length of the remaining pile. ; like Then Feedback is sent to the feed compensation calculation module to fine-tune the interface correction weight of the next hole in the same network and at the same level. The adjustment step is 0.01, and the single adjustment range does not exceed 0.

05. After completing the machining of 100 consecutive holes of the same type, the process capability index CPK of the residual pile length of the batch of holes is calculated. When CPK < 1.33, the drill tip model self-learning process is triggered; when CPK ≥ 1.67, the feed rate is automatically increased slightly to improve machining efficiency.

8. A PCB back-drilling adaptive depth control system according to claim 7, characterized in that, The PCB back-drilling adaptive depth control system also includes a drill tip model self-learning module, and the self-learning process of the drill tip model self-learning module is as follows: After receiving the self-learning trigger signal from the online processing quality verification module, the residual pile length deviation data of the most recent 30 consecutive processing holes are extracted, and the average residual pile deviation is calculated after removing abnormal hole data. Extract the cumulative cutting work parameters of the effective holes in this batch. Based on the ratio of the average residual pile deviation to the cumulative cutting work, and in conjunction with the preset learning rate, adjust the drill tip wear rate coefficient. Perform online corrections; The corrected drill tip wear rate coefficient is updated to the drill tip condition dynamic inversion module, replacing the original coefficient for subsequent borehole wear calculation. Real-time cumulative total number of boreholes ,when When the drill bit reaches 80% of its preset lifespan threshold for the corresponding model, a drill bit replacement warning will pop up on the human-machine interface; when When the lifespan threshold is reached, the machining function is locked until a new drill bit is replaced and the parameters are reset. The PCB back-drilling adaptive depth control system also includes a board parameter library module. This module pre-stores processing parameter sets for various commonly used PCB boards, including interface recognition weight coefficients for FR-4 epoxy board, PTFE high-frequency board, hydrocarbon ceramic board, and PI flexible board. and Initial values ​​of PID parameters and initial coefficients of drill tip wear rate for each control stage Thermal elongation coefficient and Default feed rate.

9. A PCB back-drilling adaptive depth control system according to claim 8, characterized in that, The PCB back-drilling adaptive depth control system also includes an abnormal working condition handling module, the processing flow of which is as follows: Drill tip chipping / excessive wear anomaly identification: Real-time monitoring of high-frequency vibration energy entropy With current mutation rate ,when and At that time, it was determined to be an abnormality of drill tip chipping; When 10 holes in a row When the calculated value exceeds 90% of the maximum allowable wear of the drill bit, it is determined that the drill bit is worn and has failed. Both types of abnormalities immediately send an emergency retraction command to the controller, control the Z-axis to lift the drill bit to a safe height at the maximum speed, stop the spindle rotation, mark the current hole position as an abnormal hole on the human-machine interface, and trigger an audible and visual alarm. Spindle overheating anomaly detection: Real-time monitoring of spindle temperature ,when At this time, the system automatically sends parameter adjustment commands to the variable impedance segmented control module, reducing the current feed rate to 50% of the normal value for the corresponding stage, and simultaneously activating the spindle cooling system's strong cooling mode, waiting for the temperature to drop back to normal. Then, restore the original feed rate and cooling mode; Handling internal anomalies in the board material: Confidence level of interface recognition over three consecutive sliding time windows. If the fluctuation range exceeds 0.5 and there is no corresponding design interface, it is determined to be an abnormality such as internal delamination, impurities, or cavities in the material. The current Z-axis position is recorded, and the feed speed is automatically reduced. 50% slow feed through the abnormal area; if the abnormality lasts for more than 1 second or the vibration value exceeds 80% of the range, immediately pull up the drill and mark the hole as abnormal.

10. A PCB back-drilling adaptive depth control method, employing the PCB back-drilling adaptive depth control system as described in claim 9, characterized in that, Includes the following steps: S1. Initialization Preparation Phase: The system performs a power-on self-test. The user selects the type of PCB board to be processed and loads the Gerber processing file. The system automatically loads the corresponding processing parameters, performs zero-point calibration of all sensors, controls the spindle to move to the initial safe height above the first hole position, and resets the current drill bit's cumulative drilling count. Initialize drill tip wear Thermal elongation ; S2, Multi-source data synchronous acquisition stage: Start the spindle to rotate to the set speed, and synchronously acquire spindle load current, drill bit axial vibration, spindle temperature and Z-axis position data through the hardware synchronous trigger unit, and cache the synchronous data in real time through the sliding time window; S3. Real-time Interface Recognition Stage: Calculate the real-time current mutation rate based on current and vibration data within the sliding time window. With high-frequency vibration energy entropy The confidence level is identified by combining the current material layer reference value with the calculation interface. ,when Determine the interface between material layers and record the actual interface position. And update the current material layer status; S4. Drill tip condition inversion stage: based on cumulative borehole count The cutting force and cutting time of the machined holes, the current spindle temperature and temperature change rate are updated in real time to calculate the current drill tip wear. With thermal elongation ; S5. Feed compensation command calculation stage: combined with the theoretical target depth of the current hole. Drill tip wear, thermal elongation, correction for residual pile deviation from the previous hole, and correction for interface position. Finally, the Z-axis feed control command is generated. ; S6. Variable impedance segmented feed stage: Confidence level is identified based on the real-time interface. Switch the PID parameters and feed rate of the corresponding control stage to control the Z-axis servo motor to drive the drill bit to complete the feed cutting; judge in real time whether the actual position of the drill bit has reached the target range, stop the feed and perform a short finishing after reaching the target range, and control the drill bit to quickly lift the drill bit to a safe height. S7. Single Hole Quality Verification Stage: After a single hole is drilled, the estimated value of the actual residual pile length and the deviation of the residual pile are calculated. If the deviation exceeds the standard, the compensation coefficient is finely adjusted. After accumulating 100 holes, the batch CPK value is calculated. If the CPK does not meet the standard, the drill tip model self-learning process is triggered to update the wear rate coefficient. S8. Real-time anomaly judgment stage: Vibration, current and temperature data are monitored in real time throughout the entire processing process. If an abnormal condition is triggered, the corresponding abnormal handling process is executed immediately, and processing is resumed after troubleshooting. S9. Cyclic Machining Stage: After completing the entire process for a single hole, locate the next machining hole position according to the Gerber file, and accumulate the number of drilled holes. Return to step S2 to process the next hole until all holes are processed, and output a processing quality statistics report.

Images