A crimping quality real-time detection method and system

Abstract

The present application relates to the technical field of semiconductor packaging, and particularly relates to a crimping quality real-time detection method and system, comprising: collecting pressure sensor data and displacement encoder data in a crimping process in real time, and generating a pressure-displacement curve; extracting feature parameters of the pressure-displacement curve, including a pressure fluctuation amplitude in a contact stage, a local slope change value in a main pressure interval, and a high-frequency oscillation frequency in a pressure maintaining stage; synchronously normalizing the pressure fluctuation amplitude, the local slope change value, and the high-frequency oscillation frequency into dimensionless risk values; weighting and fusing each risk value through a respective weight coefficient, calculating a comprehensive quality score, and judging the crimping quality according to the comprehensive quality score.In the present application, three-stage feature parameters are extracted through a pressure-displacement curve, replacing a traditional single-point threshold method, accurately capturing transient anomalies, and enabling rapid response, thereby breaking through the hysteresis problem of offline analysis of crimping quality.

Description

Technical Field

[0001] This invention relates to the field of semiconductor packaging technology, and in particular to a method and system for real-time detection of crimping quality. Background Technology

[0002] In the crimping process between connectors and PCBs, traditional quality inspection mainly relies on manual sampling or offline analysis, which suffers from low efficiency and poor real-time performance. Existing online monitoring systems mostly rely on single static thresholds for judgment, such as maximum pressure value or displacement endpoint value, which cannot effectively capture the complex dynamic behavior characteristics during the crimping process.

[0003] In actual production environments, sensor data is prone to drift due to fluctuations in ambient temperature and humidity, as well as mechanical vibrations of equipment, leading to an increased false judgment rate. Currently, there is a lack of quality assessment solutions capable of real-time integration of multi-dimensional dynamic parameters and adaptive adaptation to environmental changes, which hinders the improvement of production line intelligence. Summary of the Invention

[0004] This invention provides a method and system for real-time detection of crimping quality, which can effectively solve the problems pointed out in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A real-time crimping quality detection method, applied to a servo crimping system for connectors and PCBs, includes:

[0007] Real-time acquisition of pressure sensor data and displacement encoder data during the crimping process to generate pressure-displacement curves;

[0008] Extract the characteristic parameters of the pressure-displacement curve, including the pressure fluctuation amplitude during the contact stage, the local slope change value of the main pressure range, and the high-frequency oscillation frequency during the pressure holding stage.

[0009] The pressure fluctuation amplitude, local slope change value, and high-frequency oscillation frequency are synchronously normalized into dimensionless risk values ​​A, S, and F;

[0010] A, S, and F are weighted and integrated using weighting coefficients W1, W2, and W3 respectively, and the overall quality score Q is calculated as follows: Q = W1 × (1-A) + W2 × (1-S) + W3 × (1-F);

[0011] The quality of the crimping is determined based on the comprehensive quality score.

[0012] Furthermore, it also includes a parameter sensitivity enhancement process, including:

[0013] Apply zero-mean Gaussian noise perturbation to the original pressure-displacement curve to generate N sets of perturbation curves, where N≥100;

[0014] Calculate the coefficient of variation for each of the aforementioned characteristic parameters;

[0015] The weighting coefficients corresponding to the sensitivity of each of the aforementioned feature parameters are strengthened.

[0016] Furthermore, the weighting coefficients W1, W2, and W3 are dynamically adjusted, including:

[0017] Monitor the defect type distribution of multiple consecutive PCB boards;

[0018] When the proportion of the same type of defect exceeds the set proportion, the weight coefficient of the corresponding parameter of the defect is increased.

[0019] Furthermore, the step of determining the pressure fluctuation amplitude during the contact phase includes:

[0020] The point of displacement where the pressure continuously exceeds the initial contact pressure threshold is identified as the contact start point;

[0021] The standard deviation of pressure within a preset displacement interval after the contact point is calculated as the pressure fluctuation amplitude.

[0022] Furthermore, the steps for determining the local slope variation value of the main pressure interval include:

[0023] When the displacement is within the main pressure stroke range, the pressure change rate is calculated in real time using a sliding time window or a displacement window;

[0024] The absolute peak value of the pressure change rate between adjacent windows is recorded as the local slope change value.

[0025] Furthermore, the step of determining the high-frequency oscillation frequency during the pressure holding stage includes:

[0026] Frequency domain transformation analysis was performed on the pressure data in the later stage of the pressure holding phase.

[0027] The ratio of the energy in the preset risk frequency band to the total energy in the entire frequency band is calculated and used as the quantized value of the high-frequency oscillation frequency;

[0028] The preset risk frequency band is set based on the resonance characteristics of the PCB substrate.

[0029] Furthermore, the frequency band distribution of high-frequency oscillation during the pressure holding phase is monitored in real time;

[0030] If a shift of high-frequency energy to low-frequency energy is detected, the weighting coefficient W1 of the pressure fluctuation amplitude is increased, and the weighting coefficient W3 of the high-frequency oscillation frequency is decreased.

[0031] If a migration of low-frequency energy to high-frequency energy is detected, the weighting coefficient W2 of the local slope change value is increased, and the weighting coefficient W1 of the pressure fluctuation amplitude is decreased.

[0032] Furthermore, environmental interference decoupling processing is performed before extracting the feature parameters, including:

[0033] Environmental data of the pressing area is collected in real time using temperature and humidity sensors;

[0034] The environmental data is input into a pre-trained LSTM network, which outputs the pressure compensation amount of the pressure-displacement curve.

[0035] A new pressure-displacement curve is generated based on the pressure compensation amount for the extraction of the characteristic parameters.

[0036] Furthermore, it also includes dynamic compensation credibility monitoring, including:

[0037] The normalized ratio of the absolute value of the pressure compensation amount to the preset compensation upper limit is used as the compensation confidence level;

[0038] When the compensation confidence level is lower than the threshold, perform any of the following operations:

[0039] If the displacement difference between the curves before and after compensation at the contact point exceeds the limit, the original curve is used to extract the characteristic parameters.

[0040] Otherwise, the overall quality score of the curves before and after the fusion compensation is weighted according to the confidence level.

[0041] A real-time crimping quality detection system, applied to a servo crimping system for connectors and PCBs, includes:

[0042] The data acquisition module collects pressure sensor data and displacement encoder data in real time and generates pressure-displacement curves.

[0043] The feature extraction module extracts the feature parameters of the pressure-displacement curve, including the pressure fluctuation amplitude during the contact stage, the local slope change value of the main pressure range, and the high-frequency oscillation frequency during the pressure holding stage.

[0044] The quality assessment module is used for:

[0045] The pressure fluctuation amplitude, local slope change value, and high-frequency oscillation frequency are synchronously normalized into dimensionless risk values ​​A, S, and F;

[0046] A, S, and F are weighted and integrated using weighting coefficients W1, W2, and W3 respectively, and the overall quality score Q is calculated as follows: Q = W1 × (1-A) + W2 × (1-S) + W3 × (1-F);

[0047] The quality of the crimping is determined based on the comprehensive quality score.

[0048] The technical solution of this invention can achieve the following technical effects:

[0049] In this invention, three-stage characteristic parameters are extracted through pressure-displacement curves to replace the traditional single-point threshold method, accurately capturing transient anomalies and achieving rapid response, thus overcoming the lag problem in offline analysis of crimping quality. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 A flowchart for a real-time detection method for crimping quality;

[0052] Figure 2 This is a schematic diagram of the servo crimping system;

[0053] Figure 3 This is an exploded view of the overall structure consisting of a servo press, a pressure sensor, and a modular pressure plate, along with the positioning end.

[0054] Reference numerals: 01, servo press; 02, pressure sensor; 03, positioning end; 04, module pressure plate. Detailed Implementation

[0055] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0056] Example 1

[0057] like Figure 1 As shown, a real-time crimping quality detection method is applied to a servo crimping system between a connector and a PCB board, comprising:

[0058] A1: Real-time acquisition of pressure sensor data and displacement encoder data during the crimping process to generate pressure-displacement curves;

[0059] A2: Extract the characteristic parameters of the pressure-displacement curve, including the pressure fluctuation amplitude during the contact stage, the local slope change value of the main pressure range, and the high-frequency oscillation frequency during the pressure holding stage.

[0060] A3: Synchronously normalize the pressure fluctuation amplitude, local slope change value, and high-frequency oscillation frequency into dimensionless risk values ​​A, S, and F;

[0061] A4: Weight A, S, and F are weighted and fused using weighting coefficients W1, W2, and W3 respectively, to calculate the overall quality score Q.

[0062] Q=W1×(1-A)+W2×(1-S)+W3×(1-F);

[0063] A5: Judge the crimping quality based on the overall quality score.

[0064] See Figure 2 and 3 This paper demonstrates the structure of a specific servo crimping system for a connector and a PCB board, including: a positioning end 03 for positioning the connector and the PCB board to be crimped; a servo press 01, a pressure sensor 02, and a module pressing plate 04, all located on top of the positioning end 03; the servo press 01 and the module pressing plate 04 are connected through the pressure sensor 02; a displacement encoder is integrated into the drive mechanism of the servo press 01, such as a servo motor, for real-time monitoring of the downward displacement of the module pressing plate 04; and the pressure sensor 02 is used to read the pressure data applied to the connector and the PCB.

[0065] In this embodiment, the pressure fluctuation amplitude during the contact stage reflects the degree of misalignment between the terminal and the PCB hole at the initial alignment, such as pin misalignment; the local slope change value in the main pressure range characterizes the abnormal resistance during the plastic deformation stage of the material, such as foreign objects in the hole or uneven plating; the high-frequency oscillation frequency during the pressure holding stage exposes stress release defects, such as microcracks or loose connections; the above three respectively cover the physical stages of contact, forming and stabilization in the pressing process, forming a relatively comprehensive defect perception closed loop, which can more comprehensively identify typical process failure modes than single parameter detection.

[0066] During implementation, pressure fluctuation amplitude, slope change value, and oscillation frequency with different dimensions are converted into a unified dimensionless risk value. As a specific implementation method, the risk value ranges from 0 to 1 to eliminate the interference of numerical scale differences between parameters on the fusion calculation.

[0067] Compared to traditional methods, in this embodiment, the Q-value is dynamically weighted and fused with three-parameter risk values ​​to form a more robust comprehensive evaluation index, including:

[0068] The quantification of defect coupling effects, such as the significant decrease in Q value when the contact stage offset A increases and the main pressure resistance anomaly S increases, is achieved; adaptive sensitivity is obtained by adjusting the weighting coefficients to specifically enhance the detection sensitivity of currently prevalent defects; and graded interception is made more efficient. Specifically, in this embodiment, dynamic graded judgment and interception include:

[0069] When Q ≥ the high-quality threshold, it is judged as high-quality;

[0070] When Q is in the re-inspection range, the optical re-inspection system is triggered in real time.

[0071] When Q < failure threshold, it is determined to be a failure.

[0072] The high-quality threshold, re-inspection interval, and failure threshold mentioned above can all be preset according to actual working conditions. In this invention, three-stage characteristic parameters are extracted through pressure-displacement curves, replacing the traditional single-point threshold method, accurately capturing transient anomalies, achieving rapid response, and overcoming the bottleneck of lag in offline analysis.

[0073] As a preferred embodiment of the above embodiments, the real-time detection method for crimping quality further includes a parameter sensitivity enhancement process, comprising:

[0074] B1: Apply zero-mean Gaussian noise perturbation to the original pressure-displacement curve to generate N sets of perturbation curves, N≥100. This step aims to simulate random disturbances in real working conditions, such as mechanical vibration and sensor drift, by actively introducing zero-mean Gaussian noise perturbation, thereby generating the required perturbation curve samples.

[0075] B2: Calculate the coefficient of variation of each characteristic parameter to quantify the sensitivity of each parameter to noise;

[0076] B3: The corresponding weight coefficients are strengthened according to the sensitivity of each feature parameter, so that the high-sensitivity parameters receive higher weight compensation in the Q value calculation, while the weight of the low-sensitivity parameters is relatively reduced. This adaptively suppresses abnormal parameter fluctuations caused by noise and improves the stability and reliability of the overall quality score Q in complex interference environments.

[0077] From the N sets of curves generated after applying noise perturbation to the original pressure-displacement curve, the characteristic parameter i is re-extracted. This data set consists of pressure fluctuation amplitude, local slope change value, or high-frequency oscillation frequency. As a specific implementation method, the coefficient of variation is used. The calculation formula is:

[0078] ;

[0079] in, Let i be the standard deviation of parameter i under disturbance. Let i be the mean value of parameter i under the disturbance;

[0080] Weighting based on sensitivity enhancement parameter i The formula is:

[0081] ;

[0082] Where α is a preset enhancement factor, 0<α≤1.

[0083] As a preferred embodiment of the above, the weighting coefficients W1, W2 and W3 are dynamically adjusted, including: monitoring the defect type distribution of multiple consecutive PCB boards; and increasing the weighting coefficient of the corresponding parameter when the proportion of the same type of defect exceeds a set proportion.

[0084] In this embodiment, the set ratio is an empirical value determined based on historical data, and can be manually adjusted during implementation.

[0085] During implementation, the accuracy of PCB pad holes, the uniformity of plating, and the quality of lamination are the dominant factors for crimping reliability. Connector terminals are usually pre-inspected by suppliers, resulting in a low incoming defect rate. However, PCBs are prone to micro-deformation during transportation and temperature / humidity changes, making them a core target for real-time quality control on the production line. In this preferred solution, PCB defect identification relies on the mapping of characteristic parameters of the pressure-displacement curve. For example, for hole misalignment defects, the manifestation is severe pressure oscillation during the contact stage, quantified as an increase in pressure fluctuation amplitude. For plating peeling or foreign matter defects, it will cause a sudden change in resistance in the main pressure range, manifested as an abnormal increase in the local slope change value. For microcracks or stress concentration defects, it will lead to insufficient stress release during the holding pressure stage, resulting in a significant decrease in the high-frequency oscillation frequency.

[0086] When continuously monitoring the defect type distribution of multiple PCBs, occasional interference can be eliminated. A single abnormality may be random noise, while multiple consecutive PCBs with the same type of defect indicate a systematic process deviation. When the proportion of a certain type of defect exceeds the limit, the weight of the corresponding parameter is automatically increased, making the overall quality score Q more sensitive to capture the current high-risk factors.

[0087] As a preferred embodiment of the above, the step of determining the pressure fluctuation amplitude during the contact phase includes: identifying a displacement point where the pressure is continuously greater than the initial contact pressure threshold as the contact start point; and calculating the pressure standard deviation within a preset displacement interval after the contact start point as the pressure fluctuation amplitude.

[0088] In this preferred scheme, the continuous pressure exceeding the threshold is used as the trigger condition instead of single-point sampling. This avoids misjudging the contact timing due to instantaneous noise and ensures accurate starting point positioning. Furthermore, by selecting the displacement interval to calculate the standard deviation, the influence of changes in crimping speed on the results can be eliminated. This allows the fluctuation amplitude to directly reflect the degree of offset between the terminal and the PCB hole during spatial alignment. The statistical characteristics of the standard deviation can capture the dispersion of the pressure value within the interval, which better characterizes the overall fluctuation trend than the maximum or minimum value method.

[0089] Preferably, in this embodiment, the pressure continuously exceeding the initial contact pressure threshold is defined as follows: the pressure sensor reading is continuously higher than the threshold for more than 5-7 sampling points. In this embodiment, this time is approximately 5-14 milliseconds, while the initial contact pressure threshold is a system-set value. The preset displacement range is set according to the connector size, for example, 20%-50% of the connector pin diameter. This range covers the critical friction section from the contact hole to full insertion of the pin, which is the most sensitive area for offset defects. It can be directly calculated by measuring the physical dimensions of the terminal with calipers. After exceeding 50% of the diameter, the pin has been corrected by the hole wall, and the fluctuation disappears.

[0090] As a preferred embodiment of the above, the step of determining the local slope change value of the main pressure range includes: calculating the pressure change rate in real time using a sliding time window or a displacement window when the displacement is within the main pressure stroke range; and recording the absolute peak value of the pressure change rate between adjacent windows as the local slope change value.

[0091] In this embodiment, the main pressure stroke interval specifically refers to the period from when the connector pins are fully inserted into the PCB hole until the crimping depth reaches the standard. A sliding window mechanism is used to calculate the pressure change rate in real time. Specifically, the window width is the displacement increment or a fixed duration, such as 0.05mm or 5ms per window. The instantaneous slope of the data points within the window is calculated using linear fitting, specifically the ratio of pressure change to displacement change. After recording the slope differences between adjacent windows, the entire main pressure interval is traversed, and the absolute peak value is taken as the local slope change value. In the above method, the sliding window mechanism effectively avoids single-point fluctuation interference, ensuring that the slope reflects the true deformation trend, and the absolute peak value captures resistance change events, making it more suitable for transient anomaly detection than the mean method.

[0092] As a preferred embodiment of the above, the step of determining the high-frequency oscillation frequency during the pressure holding stage includes: performing frequency domain transformation analysis on the pressure data in the later stage of the pressure holding stage; calculating the ratio of the energy of the preset risk frequency band to the energy of the entire frequency band, as the quantization value of the high-frequency oscillation frequency; wherein, the preset risk frequency band is set based on the resonance characteristics of the PCB substrate.

[0093] In this embodiment, the later stage of the pressure holding phase refers to the last 30% of the time after the pressure has stabilized. For example, if the total pressure holding time is 100ms, the last 30ms is taken, at which point the stress is fully released and mechanical interference is minimal. The determination of the high-frequency oscillation frequency specifically includes the following steps:

[0094] The pressure data is first detrended to eliminate the steady-state baseline, and then a Hanning window is added to suppress spectral leakage. The time-domain signal is further converted into frequency-domain energy distribution using fast Fourier transform. The preset risk frequency band is set according to the PCB substrate type. In this embodiment, it is 200-500Hz for aluminum substrate and 500-1000Hz for fiberglass board.

[0095] This preferred solution avoids mechanical inertial oscillation interference in the later stage of pressure holding, so that the spectrum truly reflects the microstructure of the weld joint. Microcracks or loose connections cause high-frequency energy attenuation, which significantly reduces the high-frequency oscillation frequency and is more resistant to noise than single-peak frequency detection.

[0096] As a preferred embodiment of the above, the frequency band distribution of the high-frequency oscillation during the pressure holding stage is monitored in real time;

[0097] If a shift from high-frequency energy to low-frequency energy is detected, the weighting factor W1 for pressure fluctuation amplitude is increased, and the weighting factor W3 for high-frequency oscillation frequency is decreased.

[0098] If a migration of low-frequency energy to high-frequency energy is detected, the weighting coefficient W2 of the local slope change value is increased, and the weighting coefficient W1 of the pressure fluctuation amplitude is decreased.

[0099] This preferred embodiment, based on the above embodiments, further monitors the dynamic characteristics of the frequency band distribution of the high-frequency oscillation during the pressure holding stage in real time, and dynamically adjusts the weighting coefficients according to the energy migration mode. As a specific implementation method, the frequency band distribution specifically includes:

[0100] In the later stage of the pressure holding phase, the frequency domain transformation analysis process is continuously executed with a displacement window that overlaps by 50%, such as every 10ms window with a step size of 5ms.

[0101] By comparing the preset risk frequency band proportion and low frequency band energy proportion of adjacent windows, in this embodiment, the low frequency band specifically refers to 200Hz, and the weight adjustment is triggered when the following conditions are met:

[0102] If the energy proportion of the preset risk frequency band decreases by more than 10% and the energy proportion of the low frequency band increases by more than 15%, it is determined that the high frequency is migrating to the low frequency. At this time, the weighting coefficient W1 of the pressure fluctuation amplitude is increased and the weighting coefficient W3 of the high frequency oscillation frequency is decreased. The specific physical defect associated with this is abnormal contact stress caused by PCB hole position displacement.

[0103] If the proportion of energy in the low-frequency band decreases by more than 10%, and the proportion of energy in the preset risk frequency band increases by more than 15%, it is determined that the low-frequency band is migrating to the high-frequency band. At this time, the weighting coefficient W2 of the local slope change value is increased, and the weighting coefficient W1 of the pressure fluctuation amplitude is decreased. The associated physical defects are terminal plating peeling and foreign object jamming causing sudden changes in the resistance of the main pressure section. In this embodiment, W1 + W2 + W3 = 1 is always maintained during the adjustment of the weighting coefficients.

[0104] By implementing the above-mentioned preferred solutions, the root causes of defects can be targeted and suppressed, the energy migration mode can be directly associated with specific process defects, the weight adjustment can target and strengthen defect sensitivity indicators, and the bidirectional threshold for migration judgment during implementation can avoid false triggering by single parameter fluctuations.

[0105] As a preferred embodiment of the above, in order to eliminate feature distortion caused by environmental drift, environmental interference decoupling processing is performed before extracting feature parameters, including:

[0106] C1: Real-time environmental data of the pressing area is collected by temperature and humidity sensors; in specific implementation, a high-precision temperature and humidity sensor is integrated into the side wall of the pressing mold to collect the temperature T and relative humidity H of the pressing area in real time. The environmental data and the pressure-displacement curve use the same time scale to ensure spatiotemporal alignment. In this embodiment, the sampling rate is ≥10Hz.

[0107] C2: Input environmental data into a pre-trained LSTM network and output the pressure compensation amount of the pressure-displacement curve; the pre-trained LSTM network takes temperature, humidity and original pressure values ​​as input and outputs the pressure compensation amount. The training data of the LSTM network is specifically collected under operating conditions of -10℃~60℃ and 20%~90% humidity, collecting pressure-displacement curves of different defects and normal welds, labeled as the pressure compensation amount under constant temperature and humidity environment in the laboratory; in this preferred scheme, the pressure compensation amount essentially includes temperature drift compensation and humidity distortion compensation;

[0108] C3: Generate a new pressure-displacement curve based on the pressure compensation amount for the extraction of characteristic parameters.

[0109] In some embodiments of the present invention, dynamic compensation reliability monitoring is also included, including:

[0110] The normalized ratio of the absolute value of the pressure compensation to the preset compensation upper limit is used as the compensation confidence level; when the compensation confidence level is lower than the threshold, one of the following operations is performed:

[0111] If the displacement difference between the curves before and after compensation at the contact point exceeds the limit, the original curve is used to extract the characteristic parameters.

[0112] Otherwise, the composite quality score of the curves before and after compensation is weighted according to the confidence level. For example, if Q1 is the composite quality score generated by the compensated curve, Q0 is the composite quality score generated by the original curve, and C is the compensation confidence level, then the composite quality score Q2 after fusion is: .

[0113] In this preferred embodiment, both the preset compensation upper limit and the threshold can be set according to specific operating conditions; in this embodiment, they are system settings. This preferred embodiment can maintain quality continuity and prevent production line interruptions when the compensation component fails.

[0114] Example 2

[0115] A real-time crimping quality detection system, applied to a servo crimping system for connectors and PCBs, includes:

[0116] The data acquisition module collects pressure sensor data and displacement encoder data in real time and generates pressure-displacement curves.

[0117] The feature extraction module extracts the feature parameters of the pressure-displacement curve, including the pressure fluctuation amplitude during the contact stage, the local slope change value of the main pressure range, and the high-frequency oscillation frequency during the pressure holding stage.

[0118] The quality assessment module is used for:

[0119] The pressure fluctuation amplitude, local slope change value, and high-frequency oscillation frequency are synchronously normalized into dimensionless risk values ​​A, S, and F;

[0120] A, S, and F are weighted and integrated using weighting coefficients W1, W2, and W3 respectively, and the overall quality score Q is calculated as follows: Q = W1 × (1-A) + W2 × (1-S) + W3 × (1-F);

[0121] The quality of the crimping is judged based on the overall quality score.

[0122] The technical effects achieved in this embodiment are the same as in Embodiment 1, and will not be repeated here.

[0123] Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A real-time crimping quality detection method, applied to a servo crimping system for connectors and PCB boards, characterized in that, include: Real-time acquisition of pressure sensor data and displacement encoder data during the crimping process to generate pressure-displacement curves; Extract the characteristic parameters of the pressure-displacement curve, including the pressure fluctuation amplitude during the contact stage, the local slope change value of the main pressure range, and the high-frequency oscillation frequency during the pressure holding stage. The pressure fluctuation amplitude, local slope change value, and high-frequency oscillation frequency are synchronously normalized into dimensionless risk values ​​A, S, and F; A, S, and F are weighted and integrated using weighting coefficients W1, W2, and W3 respectively, and the overall quality score Q is calculated as follows: Q = W1 × (1-A) + W2 × (1-S) + W3 × (1-F); The pressing quality is determined based on the comprehensive quality score. The steps for determining the pressure fluctuation amplitude during the contact phase include: identifying a displacement point where the pressure is continuously greater than the initial contact pressure threshold as the contact start point; and calculating the pressure standard deviation within a preset displacement interval after the contact start point as the pressure fluctuation amplitude. The steps for determining the local slope change value of the main pressure range include: calculating the pressure change rate in real time using a sliding time window or a displacement window when the displacement is within the main pressure stroke range; and recording the absolute peak value of the pressure change rate between adjacent windows as the local slope change value. The steps for determining the high-frequency oscillation frequency during the pressure holding stage include: performing frequency domain transformation analysis on the pressure data in the later stage of the pressure holding stage; calculating the ratio of the energy of the preset risk frequency band to the energy of the entire frequency band, as the quantized value of the high-frequency oscillation frequency; wherein, the preset risk frequency band is set based on the resonance characteristics of the PCB substrate.

2. The real-time detection method for crimping quality according to claim 1, characterized in that, It also includes parameter sensitivity enhancement processes, including: Apply zero-mean Gaussian noise perturbation to the original pressure-displacement curve to generate N sets of perturbation curves, where N≥100; Calculate the coefficient of variation for each of the aforementioned characteristic parameters; The weighting coefficients corresponding to the sensitivity of each of the aforementioned feature parameters are strengthened.

3. The real-time detection method for crimping quality according to claim 1, characterized in that, The weighting coefficients W1, W2, and W3 are dynamically adjusted, including: Monitor the defect type distribution of multiple consecutive PCB boards; When the proportion of the same type of defect exceeds the set proportion, the weight coefficient of the corresponding parameter of the defect is increased.

4. The real-time detection method for crimping quality according to claim 1, characterized in that, Real-time monitoring of the frequency band distribution of high-frequency oscillations during the pressure holding phase; If a shift of high-frequency energy to low-frequency energy is detected, the weighting coefficient W1 of the pressure fluctuation amplitude is increased, and the weighting coefficient W3 of the high-frequency oscillation frequency is decreased. If a migration of low-frequency energy to high-frequency energy is detected, the weighting coefficient W2 of the local slope change value is increased, and the weighting coefficient W1 of the pressure fluctuation amplitude is decreased.

5. The real-time detection method for crimping quality according to claim 1, characterized in that, Before extracting the feature parameters, environmental interference decoupling processing is performed, including: Environmental data of the pressing area is collected in real time using temperature and humidity sensors; The environmental data is input into a pre-trained LSTM network, which outputs the pressure compensation amount of the pressure-displacement curve. A new pressure-displacement curve is generated based on the pressure compensation amount for the extraction of the characteristic parameters.

6. The real-time detection method for crimping quality according to claim 5, characterized in that, It also includes dynamic compensation credibility monitoring, including: The normalized ratio of the absolute value of the pressure compensation amount to the preset compensation upper limit is used as the compensation confidence level; When the compensation confidence level is lower than the threshold, perform any of the following operations: If the displacement difference between the curves before and after compensation at the contact point exceeds the limit, the original curve is used to extract the characteristic parameters. Otherwise, the overall quality score of the curves before and after the fusion compensation is weighted according to the confidence level.

7. A real-time crimping quality detection system, employing the real-time crimping quality detection method as described in claim 1, applied to a servo crimping system between a connector and a PCB board, characterized in that, include: The data acquisition module collects pressure sensor data and displacement encoder data in real time and generates pressure-displacement curves. The feature extraction module extracts the feature parameters of the pressure-displacement curve, including the pressure fluctuation amplitude during the contact stage, the local slope change value of the main pressure range, and the high-frequency oscillation frequency during the pressure holding stage. The quality assessment module is used for: The pressure fluctuation amplitude, local slope change value, and high-frequency oscillation frequency are synchronously normalized into dimensionless risk values ​​A, S, and F; A, S, and F are weighted and integrated using weighting coefficients W1, W2, and W3 respectively, and the overall quality score Q is calculated as follows: Q = W1 × (1-A) + W2 × (1-S) + W3 × (1-F); The quality of the crimping is determined based on the comprehensive quality score.

Images