Method for improving abnormal stress of carrier plate extruded by needle plate

Abstract

The invention relates to the technical field of I C carrier plate testing, in particular to a method for improving abnormal stress of a carrier plate extruded by a needle plate, which comprises the needle plate and a motor. The needle plate comprises an upper needle plate and a lower needle plate; the motor comprises an upper motor and a lower motor; the method further comprises the following steps: S1, acquiring needle plate information; s2, calculating the total elastic force of a needle and the stress of a needle plate; s3, driving a motor to enable the needle plate to extrude the I C carrier plate; s4, acquiring a current value and a position coordinate of the motor; s5, judging whether the current value and the relative position are in a preset interval range or not; and S6, determining the state of the needle plate according to a judgment result and adopting a corresponding adjustment strategy. The current value and the position coordinates of the motor are detected in real time, whether the current value and the relative position are within the preset interval range or not can be judged, the state of the needle plate is determined according to the judgment result, a corresponding adjustment strategy is adopted, and the abnormal stress condition that the needle plate extrudes the carrier plate can be improved.

Description

Technical Field

[0001] The invention relates to the technical field of IC carrier board testing, and in particular to a method for improving the abnormal force of a carrier board squeezed by a needle plate. Background Art

[0002] The chip is mainly composed of wafer, IC carrier, printed circuit board and other parts. The Pin pins of the wafer are small and dense, while the Pin pins of the printed circuit board are large and sparse, so the wafer cannot be directly connected to the printed circuit board; the Pin pins on the carrier are small and dense, while the Pin pins on the bottom are large and sparse, which plays a role in transferring the wafer and the printed circuit.

[0003] In order to ensure the quality of the chip, the connectivity test of the IC carrier must be carried out. Usually, the pin board is used to closely contact the Pin pin of the IC carrier to test the relevant electrical signals. However, too close contact will also cause the pin board to exert too much extrusion force on the IC carrier. The reaction force of the extrusion force is the bearing force of the pin board. Excessive bearing force will cause damage to the pin board. When the pin board produces debris due to damage, these debris may crush the Pin pin of the IC carrier, thereby affecting the test effect of the relevant electrical signals.

[0004] At present, the bearing capacity of the needle board in the market is mainly controlled by a fixed value. If it is equal to the fixed value, it is considered that the needle board is normally squeezing the substrate, and if it is greater or less than the fixed value, it is considered that the needle board is abnormally squeezing the substrate. However, the actual bearing capacity of the needle board varies depending on the material of the needle board and the number of needles. The disadvantage of the fixed value control of bearing capacity is obvious. Normal squeezing of the IC substrate by the needle board may be considered abnormal squeezing, and abnormal squeezing of the IC substrate by the needle board may be considered normal squeezing. Summary of the invention

[0005] The purpose of the present invention is to provide a method for improving the abnormal force of the needle plate squeezing the carrier plate in view of the above-mentioned deficiencies in the prior art.

[0006] The object of the present invention is achieved by the following technical scheme: A method for improving the abnormal force of a needle plate squeezing a carrier plate, characterized in that: comprising a needle plate and a motor; the needle plate comprises an upper needle plate and a lower needle plate; the motor comprises an upper motor and a lower motor; the upper motor is used to drive the upper needle plate to approach or move away from the lower needle plate; the lower motor is used to drive the lower needle plate to approach or move away from the upper needle plate;

[0007] The following steps are also included:

[0008] S1, obtain needle plate information;

[0009] S2. Calculate the total elastic force of the needle and the force on the needle plate;

[0010] S3, driving the motor to make the needle plate squeeze the IC carrier board;

[0011] S4, obtaining the motor current value and position coordinates;

[0012] S5, determining whether the current value and the relative position are within a preset range;

[0013] S6. Determine the needle plate state based on the judgment result and adopt corresponding adjustment strategies.

[0014] Furthermore, the calculation of the total elastic force of the needle and the force on the needle plate includes:

[0015] Obtain the number of needles of the needle plate; obtain the elastic coefficient of the needle plate material; obtain the type of needle and the corresponding elastic force value; calculate the total elastic force of the needle; calculate the force of the needle plate according to the elastic deformation coefficient of the needle plate material.

[0016] Further, the driving motor causes the needle plate to squeeze the IC carrier plate, including:

[0017] Compare and calculate the total elastic force of the needle and the force on the needle plate; select the smaller value as the input condition to the motor; calculate the torque output by the motor; drive the motor to make the needle plate squeeze the IC carrier.

[0018] Further, the obtaining of the motor current value and position coordinates includes:

[0019] The motor current value and position coordinate data are collected by the high-precision Hall current sensor and photoelectric encoder installed on the needle plate, and the original data is denoised and smoothed by wavelet transform filtering and Kalman filtering algorithm to obtain the cleaned feature data set;

[0020] The sliding window method is used to extract the key features of current peak value, current fluctuation rate and position deviation from the preprocessed data. The principal component analysis method is used to reduce the dimension of the features and screen out the most representative feature combination.

[0021] The selected feature data were trained using the support vector machine algorithm to establish a multi-category classification model for needle plate status. The model parameters were optimized through grid search and cross-validation to improve the classification accuracy and generalization ability.

[0022] The discrimination thresholds of various states are set, and the needle plate state evaluation rule base is constructed in combination with the Mamdan i fuzzy inference system for real-time state judgment and classification.

[0023] Further, the determining whether the current value and the relative position are within a preset range includes:

[0024] If the peak value of the motor current collected in real time exceeds the preset threshold and the position coordinate deviation is greater than the set value, it is judged to be an eccentric state. If the current fluctuation rate is abnormally high but the position coordinate is normal, it is judged to be an overvoltage state.

[0025] Furthermore, determining the needle plate state according to the judgment result and taking corresponding adjustment strategies includes:

[0026] By comparing the status classification results with the preset threshold, a needle plate adjustment suggestion is generated, including specific parameters of the adjustment direction and adjustment range, providing data support for precise adjustment;

[0027] The long short-term memory network is used to predict the state change trend of the needle plate, and the multi-sensor data fusion technology is combined to improve the accuracy of state classification and the foresight of prediction;

[0028] The adaptive fuzzy neural network algorithm is used to conduct online learning and optimization of the needle plate status classification system, dynamically adjust the classification threshold and rule base, and improve the system's ability to recognize new abnormal patterns.

[0029] Furthermore, the motor current value and position coordinate data are collected according to the high-precision Hall current sensor and photoelectric encoder installed on the needle plate, and the original data is denoised and smoothed by wavelet transform filtering and Kalman filtering algorithm to obtain the cleaned feature data set, including:

[0030] A1. Collect the original data of motor current value and position coordinates based on the high-precision Hall current sensor and photoelectric encoder installed on the needle plate;

[0031] A2. Use wavelet transform filtering algorithm to denoise the original data of motor current value to obtain current characteristic data;

[0032] A3. The original data of the position coordinates are smoothed by the Kalman filter algorithm to obtain the position characteristic data; if the current characteristic data and the position characteristic data are both within the preset interval, it is judged that the force of the needle plate squeezing the IC carrier is normal, and the needle plate and the IC carrier are coaxial; if the current characteristic data is greater than the upper limit of the preset interval and the position characteristic data is within the preset interval, it is judged that the force of the needle plate squeezing the IC carrier is too large, and the needle plate and the IC carrier are not coaxial; if the current characteristic data is less than the lower limit of the preset interval and the position characteristic data is within the preset interval, it is judged that the force of the needle plate squeezing the IC carrier is too small, and the needle plate and the IC carrier are not coaxial; if both the current characteristic data and the position characteristic data are not within the preset interval, it is judged that the needle plate and the IC carrier are not in place;

[0033] A4. According to the judgment result of whether the needle plate and the IC carrier are coaxial and the force of the needle plate squeezing the IC carrier, the corresponding needle plate force adjustment strategy is obtained through database query;

[0034] A5. Use PID control algorithm, combined with needle plate force adjustment strategy, to control the motor output torque, dynamically adjust the force of the needle plate squeezing the IC carrier until the optimal force state is reached, ensuring that the needle plate and IC carrier are coaxial and the force is reasonable.

[0035] Furthermore, the sliding window method is used to extract the key features of current peak value, current fluctuation rate, and position deviation from the preprocessed data. The principal component analysis method is used to reduce the dimension of the features and screen out the most representative feature combinations, including:

[0036] B1. According to the real-time value of the motor current, the current data is preprocessed using the sliding window method to obtain characteristic data such as current peak value and current fluctuation rate;

[0037] B2. Obtain the position coordinates of the needle plate relative to the IC carrier board, and determine the position deviation characteristics by calculating the deviation of the position coordinates;

[0038] B3. Combine the current peak value, current fluctuation rate and position deviation into a feature vector, use principal component analysis to reduce the dimension of the feature vector, and select the most representative feature combination;

[0039] B4. If the current value and the relative position are both within the preset range, it is judged that the upper and lower needle plates are coaxial and the force of the needle plates squeezing the IC carrier is normal; if the current value is greater than the preset range and the relative position is within the preset range, it is judged that the force of the needle plate squeezing the IC carrier is too large and the needle plate squeezing force needs to be adjusted to the normal range; if the current value is less than the preset range and the relative position is within the preset range, it is judged that the force of the needle plate squeezing the IC carrier is too small and the needle plate squeezing force needs to be increased to the normal range; if the current value and the relative position are both within the preset range, it is judged that the upper and lower needle plates are not in place and the motor needs to be driven to move until the needle plates are in place;

[0040] B5. Based on the calculation results of the needle plate bearing force and the analysis results of the motor current and position deviation, the fuzzy control algorithm is used to dynamically adjust the motor output torque to achieve precise control of the needle plate extrusion force;

[0041] B6. Through real-time monitoring and analysis of parameters such as motor current and needle plate position, the working status of the needle plate is judged, and the needle plate extrusion pressure control strategy is dynamically optimized according to the analysis results to achieve dynamic control of the needle plate bearing capacity.

[0042] Furthermore, the support vector machine algorithm is used to train the selected feature data to establish a multi-category classification model for needle plate status. The model parameters are optimized through grid search and cross-validation to improve the classification accuracy and generalization ability, including:

[0043] C1. Obtain characteristic data such as the number of needles of the needle plate, the elastic coefficient of the needle plate material, the type of needles, and the corresponding elastic force value;

[0044] C2. Calculate the total elastic force Fneedletotal of the needle and the force Fpressure of the needle plate based on the acquired characteristic data;

[0045] C3, taking the calculated total elastic force F needle total and the needle plate bearing force F pressure as the input features of the support vector machine algorithm;

[0046] C4. Optimize the model parameters of the support vector machine algorithm through grid search and cross validation to improve the accuracy and generalization ability of needle plate status classification;

[0047] C5. Use the optimized support vector machine algorithm to train the input feature data and establish a multi-category classification model for needle plate status;

[0048] C6. In the process of the needle plate squeezing the IC carrier, the current value of the motor and the position coordinates of the needle plate relative to the IC carrier are obtained in real time;

[0049] C7, input the acquired current value and relative position coordinates into the established needle plate state classification model to determine the stress state and coaxial state of the needle plate;

[0050] C8. If the needle plate is under normal stress and coaxial state, the needle plate extrusion force is moderate and the needle plate and IC carrier are in good contact. If the needle plate is under too much or too little stress, adjust the needle plate extrusion force to restore it to the normal range. If the needle plate is not coaxial, adjust the needle plate position in time to ensure that the needle plate and IC carrier are coaxial.

[0051] C9. According to the judgment results of the needle board status classification model, the needle board extrusion force and needle board position are dynamically adjusted to achieve real-time control of the needle board bearing capacity, avoid damage to the needle board, and improve the stability and reliability of IC substrate testing.

[0052] Furthermore, based on historical data and expert experience, the discrimination thresholds of various states are set, and the needle board state evaluation rule base is constructed in combination with the Mamdani fuzzy inference system for real-time state judgment and classification, including:

[0053] D1. Based on historical data and expert experience, the motor current range and the position range of the needle plate relative to the IC carrier when the needle plate is working normally are obtained as the threshold for determining the needle plate state;

[0054] D2. Using Mamdani fuzzy inference system, we build a needle plate state evaluation rule base through two input variables: needle plate motor current interval and needle plate relative position interval;

[0055] D3, obtain the motor current value when the needle plate is working in real time and the position coordinates of the needle plate relative to the IC carrier;

[0056] D4, taking the real-time acquired motor current value and relative position coordinates of the needle plate as input variables, and obtaining the fuzzy evaluation result of the needle plate state by querying the needle plate state evaluation rule library;

[0057] D5. Defuzzify the fuzzy evaluation results of the needle plate state using the centroid method to determine the specific classification of the needle plate state;

[0058] D6. If the needle plate state is judged to be that the force of the needle plate squeezing the IC carrier is normal and the needle plate is coaxial, the needle plate state is normal and the process ends; if the needle plate state is judged to be that the force of the needle plate squeezing the IC carrier is too large or too small, the motor output torque is adjusted to restore the force of the needle plate squeezing the IC carrier to normal; if the needle plate state is judged to be that the needle plate is not coaxial, the motor is controlled to adjust the position of the needle plate to restore the coaxial state of the needle plate; after the needle plate state adjustment is completed, the real-time working parameters of the needle plate are re-acquired to perform the next round of needle plate state evaluation and classification judgment.

[0059] Furthermore, if the peak value of the motor current collected in real time exceeds the preset threshold and the position coordinate deviation is greater than the set value, it is judged as an eccentric state; if the current fluctuation rate is abnormally high but the position coordinate is normal, it is judged as an overvoltage state, including:

[0060] E1. Obtain the number of needles and the elastic coefficient of the needle plate material according to the needle plate information, and calculate the total elastic force FneedleTotal of the needles and the force Fpressure of the needle plate;

[0061] E2, by comparing the total elastic force Fneedle and the needle plate bearing force Fpressure, select the minimum value as the input condition to the motor, and the motor calculates the output torque M through the given force;

[0062] E3, drive the motor to move, squeeze the IC carrier, and obtain the real-time value of the motor current and the position coordinates of the motor at the same time;

[0063] E4. Obtain the position of the needle board relative to the IC carrier board from the database and determine the current value and relative position;

[0064] E5. If the current value and the relative position are within the interval, it is determined that the upper and lower needle plates are coaxial and the force of the needle plates squeezing the IC carrier is normal; if the peak value of the motor current collected in real time exceeds the preset threshold and the position coordinate deviation is greater than the set value, it is determined to be an eccentric state, and the needle plate coaxiality and the force of squeezing the IC carrier need to be adjusted to be normal; if the current fluctuation rate is abnormally high but the position coordinate is normal, it is determined to be an overvoltage state, and the force of the needle plate squeezing the IC carrier needs to be adjusted to be normal; if the relative position is not within the interval and the current value is not within the interval, it is determined that the upper and lower needle plates are not in place, and return to E3 for action;

[0065] E6. Through comprehensive analysis of the motor current range and the position distance of the needle plate relative to the product, the bearing capacity of the needle plate, the coaxiality of the needle plate and the non-coaxiality of the needle plate and other working conditions are determined, and then different adjustment strategies are adopted for different working conditions to achieve dynamic control of the bearing capacity of the needle plate.

[0066] Furthermore, by comparing the status classification results with the preset threshold, needle plate adjustment suggestions are generated, including specific parameters such as adjustment direction and adjustment range, to provide data support for precise adjustment, including:

[0067] G1. Obtain the number of needles and the elastic coefficient of the needle plate material according to the needle plate information, and calculate the total elastic force FneedleTotal of the needles and the force Fpressure of the needle plate;

[0068] G2, by comparing the total elastic force Fneedle and the needle plate bearing force Fpressure, select a minimum value as the input condition to give to the motor. The motor calculates the torque M output by the program through the given force;

[0069] G3, drive the motor to move, squeeze the IC carrier, and obtain the real-time value of the motor current and the position coordinates of the motor at the same time;

[0070] G4. Obtain the position of the needle board relative to the IC carrier board from the database and determine the current value and relative position;

[0071] G5. If the current value and relative position are within the interval, it is determined that the upper and lower needle plates are coaxial and the force of the needle plates squeezing the IC carrier is normal;

[0072] G6. If the relative position is within the interval range and the current value is greater than the interval range, it is determined that the force of the needle plate squeezing the IC carrier is large, and the force of the needle plate squeezing the IC carrier needs to be adjusted to normal;

[0073] G7. If the current value is less than the interval range, it is determined that the force of the needle plate squeezing the IC carrier is small, and the needle plate coaxiality and the force squeezing the IC carrier need to be adjusted to normal;

[0074] G8, if the relative position is not within the interval range, and the current value is not within the interval range, it is determined that the upper and lower needle plates are not in place, and returns to G3 to drive the motor to continue to move;

[0075] G9. Compare the judgment result with the preset threshold value to determine whether the needle plate state is normal, the needle plate force is too large or too small, and the needle plate is not aligned with the same axis;

[0076] G10. Using the status classification results and combining them with the preset adjustment strategy, generate needle plate adjustment suggestions including parameters such as adjustment direction and adjustment range, providing data support for subsequent precise adjustment of the needle plate.

[0077] Furthermore, the long short-term memory network is used to predict the state change trend of the needle plate, and combined with multi-sensor data fusion technology, the accuracy of state classification and the foresight of prediction are improved, including:

[0078] H1. Obtain the historical status data of the needle board, including the material of the needle board, the number of needles, the type of needles, the elastic coefficient of the needle board, the force of the needle board, etc., and input these data into the long short-term memory network for training;

[0079] H2. Use the long short-term memory network to predict the state change trend of the needle board and obtain the state change of the needle board in the future;

[0080] H3. Use multiple sensors to monitor the real-time status of the needle plate and obtain information such as the real-time force and real-time position of the needle plate;

[0081] H4, integrate the needle board state changes predicted by the long short-term memory network with the real-time state of the needle board monitored by multiple sensors to comprehensively analyze the overall state of the needle board;

[0082] H5. According to the overall state of the needle plate, determine whether the needle plate is in normal working condition. If the needle plate is in normal working condition, continue to maintain the current test plan;

[0083] H6. If the needle plate is in an abnormal working state, further determine whether the needle plate is coaxial and whether the force of the needle plate squeezing the carrier plate is normal;

[0084] H7. If the needle plate is not coaxial and the force of the needle plate squeezing the carrier plate is abnormal, then according to the overall state of the needle plate, adopt the corresponding adjustment strategy to dynamically adjust the parameters such as the force on the needle plate and the position of the needle plate to restore the needle plate to a normal working state;

[0085] H8. After the adjustment is completed, the multi-sensor is used again to monitor the real-time status of the needle plate, and the monitoring data is input into the long short-term memory network to re-predict the state change trend of the needle plate;

[0086] H9. Repeat steps H3-H8 continuously to achieve real-time monitoring and dynamic adjustment of the needle plate status to ensure that the needle plate is always in the best working condition and avoid test failure caused by needle plate damage.

[0087] Furthermore, the adaptive fuzzy neural network algorithm is used to conduct online learning and optimization of the needle plate status classification system, dynamically adjust the classification threshold and rule base, and improve the system's ability to identify new abnormal patterns, including:

[0088] J1. According to the needle plate information and the number of needles, the elastic coefficient of the needle plate material and the elastic force value of the needle are obtained, and the total elastic force FneedleTotal of the needle and the force Fpressure of the needle plate are calculated;

[0089] J2. Adopting adaptive fuzzy neural network algorithm, taking total elastic force of needle FneedleTotal, force of needle plate Fpressure, real-time value of motor current and position of needle plate relative to IC carrier as input, the needle plate state is classified;

[0090] J3. Through online learning and optimization, dynamically adjust the classification threshold and rule base to improve the system's ability to identify new abnormal patterns;

[0091] J4. If the needle plate state classification result is that the needle plate extrusion pressure is normal, the needle plate is judged to be coaxial and the current motor output torque is maintained unchanged; if the needle plate state classification result is that the needle plate extrusion pressure is too large, the needle plate is judged to be not coaxial, the motor output torque is adjusted, and the needle plate extrusion pressure is reduced until the needle plate extrusion pressure is normal; if the needle plate state classification result is that the needle plate extrusion pressure is too small, the needle plate is judged to be not coaxial, the motor output torque is adjusted, and the needle plate extrusion pressure is increased until the needle plate extrusion pressure is normal; if the needle plate state classification result is that the needle plate is not in place, the drive motor continues to move until the needle plate is in place;

[0092] J5. According to the classification results of the needle plate state and the adjustment of the motor output torque, the dynamic control strategy of the needle plate bearing force is obtained;

[0093] J6. Use dynamic control strategy to adjust the motor output torque in real time to keep the needle plate extrusion pressure within a reasonable range, avoid needle plate damage, and ensure IC substrate test results.

[0094] Beneficial effects of the present invention: The present invention detects the motor current value and position coordinates in real time, and can judge whether the current value and the relative position are within a preset range, thereby determining the state of the needle plate according to the judgment result and taking corresponding adjustment strategies to improve the abnormal stress condition of the needle plate squeezing the carrier plate. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] The invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0096] Figure 1 It is a schematic diagram of the structure of the present invention;

[0097] Figure 2 is a system flow chart of the present invention;

[0098] Figure 3 is a system flow chart of the adjustment strategy of the present invention;

[0099] Among them: 1. Upper needle plate; 2. Lower needle plate. DETAILED DESCRIPTION

[0100] The present invention is further described in conjunction with the following examples.

[0101] Depend on Figures 1 to 3It can be seen that in this embodiment, a method for improving the abnormal force of the needle plate squeezing the carrier plate is characterized in that it includes: it includes a needle plate and a motor; the needle plate includes an upper needle plate 1 and a lower needle plate 2; the motor includes an upper motor and a lower motor; the upper motor and the lower motor are not drawn in the figure; the upper motor and the lower motor respectively drive the upper needle plate 1 and the lower needle plate 2 to move closer to or away from each other through the screw slide structure; the upper motor is used to drive the upper needle plate 1 to move closer to or away from the lower needle plate 2; the lower motor is used to drive the lower needle plate 2 to move closer to or away from the upper needle plate 1; it also includes the following steps: S1, obtaining needle plate information; S2, calculating the total elastic force of the needle and the force on the needle plate; S3, driving the motor to make the needle plate squeeze the IC carrier; S4, obtaining the motor current value and position coordinates; S5, judging whether the current value and the relative position are within the preset range; S6, determining the needle plate state according to the judgment result and taking corresponding adjustment strategies. Calculating the total elastic force of the needle and the force on the needle plate includes: obtaining the number of needles on the needle plate; obtaining the elastic coefficient of the needle plate material; obtaining the type of needle and the corresponding elastic force value; calculating the total elastic force of the needle; calculating the force on the needle plate according to the elastic deformation coefficient of the needle plate material. Driving the motor to make the needle plate squeeze the IC carrier includes: comparing and calculating the total elastic force of the needle and the force on the needle plate; selecting the smaller value as an input condition to give to the motor; calculating the torque output by the motor; driving the motor to make the needle plate squeeze the IC carrier.

[0102] This embodiment detects the motor current value and position coordinates in real time, and can determine whether the current value and the relative position are within a preset range, thereby determining the state of the needle plate based on the judgment result and taking corresponding adjustment strategies to improve the abnormal stress condition of the needle plate squeezing the carrier plate.

[0103] In this embodiment, the acquisition of the motor current value and position coordinates includes: collecting the motor current value and position coordinate data according to the high-precision Hall current sensor and photoelectric encoder installed on the needle board, denoising and smoothing the original data through wavelet transform filtering and Kalman filtering algorithm to obtain a cleaned feature data set; using the sliding window method to extract the current peak value, current fluctuation rate and position deviation key features from the preprocessed data, reducing the dimension of the features through the principal component analysis method, and screening out the most representative feature combination; using the support vector machine algorithm to train the screened feature data, establish a multi-category classification model for the needle board state, optimize the model parameters through grid search and cross-validation, and improve the classification accuracy and generalization ability; setting the discrimination threshold of each state, combining the Mamdani fuzzy inference system to build a needle board state evaluation rule library for real-time state judgment and classification. The judgment of whether the current value and relative position are within the preset range includes: if the real-time collected motor current peak value exceeds the preset threshold and the position coordinate deviation is greater than the set value, it is judged as an eccentric state, and if the current fluctuation rate is abnormally high but the position coordinate is normal, it is judged as an overvoltage state. The method of determining the needle plate state according to the judgment result and taking corresponding adjustment strategies includes: generating needle plate adjustment suggestions by comparing the state classification result with the preset threshold, including specific parameters of adjustment direction and adjustment range, to provide data support for precise adjustment; using long short-term memory network to predict the needle plate state change trend, combining multi-sensor data fusion technology to improve the accuracy of state classification and the foresight of prediction; using adaptive fuzzy neural network algorithm to conduct online learning and optimization of the needle plate state classification system, dynamically adjusting the classification threshold and rule base, and improving the system's recognition ability for newly emerging abnormal patterns.

[0104] In this embodiment, the motor current value and position coordinate data are collected according to the high-precision Hall current sensor and photoelectric encoder installed on the needle plate, and the original data are denoised and smoothed by wavelet transform filtering and Kalman filtering algorithm to obtain a cleaned feature data set, including: collecting the motor current value and position coordinate original data according to the high-precision Hall current sensor and photoelectric encoder installed on the needle plate; denoising the motor current value original data by wavelet transform filtering algorithm to obtain current feature data; smoothing the position coordinate original data by Kalman filtering algorithm to obtain position feature data; if the current feature data and the position feature data are both within the preset range, it is judged that the force of the needle plate squeezing the IC carrier is normal and the needle plate is coaxial with the IC carrier; if the current feature data is greater than the preset If the current characteristic data is less than the lower limit of the preset interval and the position characteristic data is within the preset interval, it is judged that the force of the needle plate squeezing the IC substrate is too large, and the needle plate and the IC substrate are not coaxial; if the current characteristic data is less than the lower limit of the preset interval and the position characteristic data is within the preset interval, it is judged that the force of the needle plate squeezing the IC substrate is too small, and the needle plate and the IC substrate are not coaxial; if both the current characteristic data and the position characteristic data are not within the preset interval, it is judged that the needle plate and the IC substrate are not in place; according to the judgment results of whether the needle plate and the IC substrate are coaxial and the force of the needle plate squeezing the IC substrate, the corresponding needle plate force adjustment strategy is obtained through database query; the PID control algorithm is adopted, combined with the needle plate force adjustment strategy, to control the motor output torque, and dynamically adjust the force of the needle plate squeezing the IC substrate until the optimal force state is reached to ensure that the needle plate and the IC substrate are coaxial and the force is reasonable.

[0105] Exemplarily, a high-precision Hall current sensor and a photoelectric encoder are installed on the needle board, wherein the measurement range of the Hall current sensor is 0-10A, the accuracy reaches 1%, and the resolution of the photoelectric encoder is 1000 pulses / turn. The system collects the original data of the motor current value and the position coordinate, performs wavelet transform filtering on the original data of the motor current value, selects the db4 wavelet basis function to perform a 5-layer decomposition of the signal, removes high-frequency noise, and obtains the current characteristic data. At the same time, the Kalman filter algorithm is used to smooth the original data of the position coordinate, sets the process noise covariance to 01, and the measurement noise covariance to 1 to obtain the position characteristic data. The system determines whether the current characteristic data and the position characteristic data are both within the preset range, wherein the preset range of the current characteristic data is 3-7A, and the preset range of the position characteristic data is 9-11mm. If both are within the range, the force of the needle board squeezing the IC carrier is normal, and the needle board is coaxial with the IC carrier. If the current characteristic data is greater than 7A and the position characteristic data is within 9-11mm, the force of the needle board squeezing the IC carrier is too large, and the needle board is not coaxial with the IC carrier. If the current characteristic data is less than 3A and the position characteristic data is within 9-11mm, the force of the needle plate squeezing the IC substrate is too small, and the needle plate and the IC substrate are not coaxial. If both the current characteristic data and the position characteristic data are not within the preset range, such as the current characteristic data is 1A and the position characteristic data is 5mm, the needle plate and the IC substrate are not in place. Based on the above judgment results, the system queries the corresponding needle plate force adjustment strategy from the database, adopts the PID control algorithm, sets the proportional coefficient to 5, the integral coefficient to 8, and the differential coefficient to 2, and combines the needle plate force adjustment strategy to control the motor output torque and dynamically adjust the force of the needle plate squeezing the IC substrate. Through continuous iterative optimization, the force of the needle plate squeezing the IC substrate reaches the optimal force state of 5-5N, ensuring that the needle plate and the IC substrate are coaxial and the force is reasonable, thereby ensuring the stability and reliability of the IC substrate test.

[0106] In this embodiment, a sliding window method is used to extract key features such as current peak value, current fluctuation rate, and position deviation from the preprocessed data, and the features are reduced in dimension by the principal component analysis method to screen out the most representative feature combination, including: preprocessing the current data according to the real-time value of the motor current by the sliding window method to obtain feature data such as current peak value and current fluctuation rate; obtaining the position coordinates of the needle plate relative to the IC carrier, and determining the position deviation feature by calculating the deviation of the position coordinates; combining the current peak value, current fluctuation rate, and position deviation features into a feature vector, and reducing the dimension of the feature vector by the principal component analysis method to screen out the most representative feature combination; if the current value and the relative position are both within the preset range, then it is judged that the upper and lower needle plates are coaxial and the force of the needle plate squeezing the IC carrier is normal; if the current value is greater than the preset range, If the current value is less than the preset range and the relative position is within the preset range, it is judged that the force of the needle plate squeezing the IC carrier is too large, and the needle plate squeezing force needs to be adjusted to the normal range; if the current value is less than the preset range and the relative position is within the preset range, it is judged that the force of the needle plate squeezing the IC carrier is too small, and the needle plate squeezing force needs to be increased to the normal range; if both the current value and the relative position are not within the preset range, it is judged that the upper and lower needle plates are not in place, and the motor needs to continue to be driven until the needle plates are in place; according to the calculation results of the needle plate bearing force and the analysis results of the motor current and position deviation, the fuzzy control algorithm is used to dynamically adjust the motor output torque to achieve precise control of the needle plate squeezing force; through real-time monitoring and analysis of parameters such as motor current and needle plate position, the working state of the needle plate is judged, and the needle plate squeezing force control strategy is dynamically optimized according to the analysis results to achieve dynamic control of the needle plate bearing force.

[0107] For example, in the process of pressing the IC carrier with the needle plate, the motor current data is collected in real time, collected once every 10 milliseconds, and collected continuously for 100 times to obtain a data window containing 100 current values. The current data in the window is analyzed, and the current peak value is 5A and the current fluctuation rate is 5%. At the same time, the position coordinates of the needle plate relative to the IC carrier are obtained through the machine vision system, and the deviation of the center of the needle plate relative to the center of the IC carrier is calculated to be 0.2mm. The current peak value, current fluctuation rate, and position deviation are combined into a three-dimensional feature vector, and the feature vector is reduced in dimension by the principal component analysis method to obtain a one-dimensional comprehensive feature value of 85. According to the pre-set threshold range, the current peak threshold is 0A~0A, and the position deviation threshold is 0~0.5mm, and the needle plate pressing state is judged. When the comprehensive feature value falls within the range of 8~9, it can be determined that the needle plate pressing force is normal and the upper and lower needle plates are coaxial. If the comprehensive feature value is greater than 9, the needle plate pressing force is too large, and the motor output torque needs to be dynamically adjusted through the fuzzy control algorithm to reduce the needle plate pressing force. If the comprehensive characteristic value is less than 8, the needle plate pressing force is too small, and the motor output torque needs to be increased to improve the needle plate pressing force. If the comprehensive characteristic value exceeds the range of 7 to 0, it is determined that the needle plate is not in place, and the motor needs to continue to be driven until the needle plate is pressed in place. Through real-time monitoring and analysis of parameters such as motor current and needle plate position, a fuzzy control rule between the needle plate pressing force and the motor current and position deviation is established. According to the real-time feedback of the needle plate pressing state, the needle plate pressing force control strategy is dynamically optimized to achieve precise control of the needle plate bearing capacity and ensure reliable pressing of the IC substrate.

[0108] In this embodiment, a support vector machine algorithm is used to train the screened feature data, establish a multi-category classification model of needle plate status, and optimize model parameters through grid search and cross-validation to improve classification accuracy and generalization ability, including: obtaining feature data such as the number of needles of the needle plate, the elastic coefficient of the needle plate material, the type of needles and the corresponding elastic force value; calculating the total elastic force F of the needles and the force F of the needle plate based on the acquired feature data; using the calculated total elastic force F of the needles and the force F of the needle plate as input features of the support vector machine algorithm; optimizing the model parameters of the support vector machine algorithm through grid search and cross-validation to improve the accuracy and generalization ability of needle plate status classification; using the optimized support vector machine algorithm to train the input feature data, and establish a multi-category classification model of needle plate status. Category classification model; in the process of the needle plate squeezing the IC carrier, the current value of the motor and the position coordinates of the needle plate relative to the IC carrier are obtained in real time; the obtained current value and relative position coordinates are input into the established needle plate state classification model to judge the stress state and coaxial state of the needle plate; if the stress state and coaxial state of the needle plate are normal, the needle plate squeezing force is moderate, and the needle plate and the IC carrier achieve good contact; if the needle plate is too large or too small, adjust the needle plate squeezing force to restore it to the normal range; if the needle plate is not coaxial, adjust the needle plate position in time to ensure that the needle plate and the IC carrier are coaxial; according to the judgment result of the needle plate state classification model, dynamically adjust the needle plate squeezing force and needle plate position to realize real-time control of the needle plate bearing capacity, avoid damage to the needle plate, and improve the stability and reliability of IC carrier test.

[0109] For example, the number of needles of the needle plate is 1000, the needle plate material is stainless steel, its elastic coefficient is 200GPa, the type of needle is spring needle, and the corresponding elastic force value is 5N. According to these characteristic data, it can be calculated that the total elastic force FneedleTotal of the needle is 500N, and the force Fpressure of the needle plate is 600N. These two force values ​​are used as input features of the support vector machine algorithm. Through grid search and 5-fold cross validation, the optimal model parameters are obtained: the kernel function is RBF, the penalty coefficient C is 10, and the kernel function coefficient γ is 1. At this time, the accuracy of the needle plate state classification can reach more than 98%. The input features are trained using the optimized support vector machine algorithm to establish a multi-category classification model of the needle plate force state (normal, too large, too small) and coaxial state (coaxial, non-coaxial). In the process of the needle plate squeezing the IC carrier, the current value of the motor and the position coordinates of the needle plate relative to the IC carrier are collected by the current sensor and the displacement sensor respectively. If the current value is about 1A and the position deviation is within 1mm, it is judged that the force and coaxial state of the needle plate are normal, and the needle plate can achieve reliable contact with the IC carrier. If the current value exceeds 5A, the force on the needle plate is too large, and the motor needs to be controlled to rotate in the opposite direction to reduce the needle plate squeezing force to less than 1A. If the current value is lower than 5A, the force on the needle plate is insufficient, and the motor needs to be controlled to rotate forward to increase the needle plate squeezing force to about 1A. If the needle plate position deviation exceeds 1mm, the needle plate position is adjusted by the XY translation stage to make it coaxial with the IC carrier, and the deviation is controlled within 1mm. The needle plate working state is judged in real time by the needle plate state classification model, and the needle plate squeezing force and position are dynamically adjusted in combination with the PID control algorithm, so as to realize the closed-loop control of the needle plate bearing force, avoid the needle plate being damaged due to excessive pressure or poor contact due to too little pressure, thereby improving the stability and reliability of IC carrier test.

[0110] In this embodiment, the discrimination thresholds of various states are set according to historical data and expert experience, and the needle plate state evaluation rule base is constructed in combination with the Mamdani fuzzy inference system for real-time state judgment and classification, including: D1. According to historical data and expert experience, the motor current interval and the position interval of the needle plate relative to the IC carrier when the needle plate works normally are obtained as the threshold for discriminating the needle plate state; Mamdani fuzzy inference system is used to determine the state of the needle plate; i fuzzy reasoning system, through the two input variables of needle plate motor current interval and needle plate relative position interval, construct a needle plate state evaluation rule base; obtain the motor current value of the needle plate when it is working in real time and the position coordinates of the needle plate relative to the IC carrier; use the real-time motor current value and needle plate relative position coordinates as input variables, and obtain the fuzzy evaluation result of the needle plate state by querying the needle plate state evaluation rule base; use the center of gravity method to defuzzify the fuzzy evaluation result of the needle plate state to determine the specific classification of the needle plate state; if the needle plate state is judged to be that the force of the needle plate squeezing the IC carrier is normal and the needle plate is coaxial, the needle plate state is normal and the process ends; if the needle plate state is judged to be that the force of the needle plate squeezing the IC carrier is too large or too small, adjust the motor output torque to restore the force of the needle plate squeezing the IC carrier to normal; if the needle plate state is judged to be that the needle plate is not coaxial, control the motor to adjust the needle plate position to restore the needle plate to the coaxial state; after the needle plate state adjustment is completed, re-obtain the real-time working parameters of the needle plate, and perform the next round of needle plate state evaluation and classification judgment.

[0111] For example, through statistical analysis of historical data and summary of expert experience, it is determined that the normal working range of the needle plate motor current is 5-5A, and the position deviation of the needle plate relative to the IC carrier does not exceed 2mm. Using the Mamdani fuzzy inference system, the needle plate motor current and position deviation are used as input variables. For different current and position deviation combinations, 9 fuzzy rules are set to construct a needle plate state evaluation rule base. When the needle plate is working in real time, the sensor collects motor current and needle plate position data every 1 second, and the collected current value and position coordinates are used as inputs of the fuzzy inference system. By querying the evaluation rule base, the fuzzy evaluation results of the needle plate state are obtained, including the needle plate force state and coaxial state membership. The centroid method is used to defuzzify the fuzzy evaluation results, and the specific membership values ​​of the needle plate force state and coaxial state are calculated. If the needle plate force membership is above 6 and the coaxial membership is also above 6, it is determined that the needle plate is working normally. If the force membership of the needle plate is lower than 4, the motor output torque is adjusted through the PID control algorithm to restore the force of the needle plate squeezing the IC carrier to the normal range of 5-5N. If the coaxial membership of the needle plate is lower than 4, the motor position is adjusted through the PID control algorithm to control the position deviation between the needle plate and the IC carrier to within 2mm. After the needle plate state adjustment is completed, the system automatically returns to the data acquisition stage, re-acquires the real-time working parameters of the needle plate, and repeatedly performs state evaluation and adjustment control to keep the needle plate in the best working state, thereby ensuring the stability and reliability of IC carrier detection.

[0112] In this embodiment, if the peak value of the motor current collected in real time exceeds a preset threshold and the position coordinate deviation is greater than a set value, it is judged to be an eccentric state; if the current fluctuation rate is abnormally high but the position coordinate is normal, it is judged to be an overvoltage state, including: obtaining the number of needles and the elastic coefficient of the needle plate material according to the needle plate information, and calculating the total elastic force F of the needle and the force F of the needle plate; by comparing the total elastic force F of the needle and the force F of the needle plate, the minimum value is selected as the input condition to the motor, and the motor calculates the output torque M through the given force; the motor is driven to move, squeeze the IC carrier, and at the same time obtain the real-time value of the motor current and the position coordinates of the motor; obtain the position of the needle plate relative to the IC carrier from the database, and judge the current value and the relative position; if the current value and the relative position are in the interval If the motor current peak value collected in real time exceeds the preset threshold and the position coordinate deviation is greater than the set value, it is judged to be in an eccentric state, and the needle plate coaxiality and the force of squeezing the IC carrier need to be adjusted to be normal; if the current fluctuation rate is abnormally high but the position coordinate is normal, it is judged to be an overvoltage state, and the needle plate squeezing the IC carrier need to be adjusted to be normal; if the relative position is not within the interval and the current value is not within the interval, it is judged that the upper and lower needle plates are not in place, and return to E3 for action; by comprehensively analyzing the motor current range and the position distance of the needle plate relative to the product, the working conditions such as the bearing capacity of the needle plate, the coaxiality of the needle plate and the non-coaxiality of the needle plate are judged, and different adjustment strategies are adopted for different working conditions to realize dynamic control of the bearing capacity of the needle plate.

[0113] For example, according to the needle plate information, the number of needles is 1000, the needle plate material is stainless steel, and the elastic coefficient is 193GPa. The total elastic force of the needle is calculated by the formula F needle total = 1000 × 193 × 109 × π ​​× (0001) 2 / 4 = 1517N. The force F pressure of the needle plate is calculated to be 1600N by finite element analysis. Compare F needle total and F pressure, select the minimum value 1517N as the input condition to the motor, and the motor is calculated by the formula M = F × r, where r is the motor shaft radius 01m, and the output torque M = 118N·m. The drive motor moves to squeeze the IC carrier, and at the same time obtains the real-time value of the motor current through the Hall sensor, and obtains the position coordinates of the motor through the encoder. The position of the needle plate relative to the IC carrier is obtained from the database as 1mm±01mm, and the current value and relative position are determined. If the current value is within 5A±1A and the relative position is within 1mm±01mm, it is determined that the upper and lower needle plates are coaxial and the force of the needle plate squeezing the IC carrier is normal. If the peak value of the motor current collected in real time exceeds 7A and the position coordinate deviation is greater than 0.2mm, the needle plate position is adjusted through the fuzzy control algorithm to make the needle plate coaxial, and the motor output torque is adjusted through the PID control algorithm to make the force of squeezing the IC carrier plate 1517N. If the current fluctuation rate exceeds 10% but the position coordinate is within 1mm±0.1mm, the current value is filtered through the Kalman filter algorithm, and then the motor output torque is adjusted through the PID control algorithm to make the force of squeezing the IC carrier plate 1517N. If the relative position exceeds 1mm±0.1mm and the current value exceeds 5A±1A, it is determined that the upper and lower needle plates are not in place, and the drive motor action step is returned to readjust. Through comprehensive analysis of the motor current range and the position distance of the needle plate relative to the product, a mapping model of the needle plate bearing capacity, needle plate coaxiality, current value, and position coordinates is established. Fuzzy control, PID control, Kalman filtering and other strategies are adopted for different working conditions to achieve dynamic control of the needle plate bearing capacity and ensure the test accuracy and reliability of the IC carrier.

[0114] In this embodiment, by comparing the state classification result with the preset threshold, a needle plate adjustment suggestion is generated, including specific parameters such as adjustment direction and adjustment amplitude, to provide data support for precise adjustment, including: obtaining the number of needles and the elastic coefficient of the needle plate material according to the needle plate information, and calculating the total elastic force F of the needles and the force F of the needle plate; by comparing the total elastic force F of the needles and the force F of the needle plate, a minimum value is selected as an input condition to be given to the motor, and the motor calculates the torque M output by the program through the given force; the motor is driven to move, squeeze the IC carrier, and at the same time obtain the real-time value of the motor current and the position coordinates of the motor; the position of the needle plate relative to the IC carrier is obtained from the database, and the current value and the relative position are determined; if the current value and the relative position are within the interval, it is determined that the upper and lower needle plates are in the same range. axis, the force of the needle plate squeezing the IC carrier is normal; if the relative position is within the interval range and the current value is greater than the interval range, it is determined that the force of the needle plate squeezing the IC carrier is large, and the force of the needle plate squeezing the IC carrier needs to be adjusted to be normal; if the current value is less than the interval range, it is determined that the force of the needle plate squeezing the IC carrier is small, and the needle plate coaxiality and the force of squeezing the IC carrier need to be adjusted to be normal; if the relative position is not within the interval range and the current value is not within the interval range, it is determined that the upper and lower needle plates are not in place, and the drive motor returns to continue to move; according to the judgment result, the preset threshold is compared to determine whether the needle plate state is normal, the needle plate force is too large or too small, and the needle plate is not in the same axis; the state classification result is used, combined with the preset adjustment strategy, to generate needle plate adjustment suggestions including parameters such as adjustment direction and adjustment amplitude, to provide data support for subsequent precise adjustment of the needle plate.

[0115] For example, first, according to the needle plate information, the number of needles is 1000, the needle plate material is stainless steel, and its elastic coefficient is 193GPa. Using the formula FneedleTotal=n×k×Δx, where n is the number of needles, k is the elastic coefficient of the needle, and Δx is the displacement of the needle, the total elastic force FneedleTotal of the needle is calculated to be 965N. At the same time, according to the force analysis of the needle plate, the force of the needle plate is calculated using the finite element analysis method, and it is concluded that the pressure Fpressure borne by the needle plate is 1000N. By comparing the total elastic force FneedleTotal of the needle and the force borne by the needle plate Fpressure, the smaller value 965N is selected as the input condition of the motor. According to the input condition, the motor uses the formula M=F×L to calculate the motor output torque M as 295N·m, where F is 965N and L is 0.3m. Then the control system drives the motor to run, driving the needle plate to squeeze the IC carrier, and at the same time collects the motor current value and position coordinates in real time. The motor collects data every 5ms and transmits it to the database. Extract the position data of the needle plate relative to the IC carrier from the database to determine whether the needle plate is coaxial with the IC carrier. Assume that the coaxial position range of the needle plate and the IC carrier is 95-05, and the normal current range is 5-5A. If the collected current value and position data are both within the normal range, it is determined that the needle plate is coaxial with the IC carrier and the extrusion force is appropriate. If the position data is within the range but the current value is higher than 5A, it is determined that the extrusion force is too large and the motor torque needs to be reduced. If the position data is within the range but the current value is lower than 5A, it is determined that the extrusion force is too small and the motor torque needs to be increased and the needle plate position needs to be fine-tuned. If both the position data and the current value are out of range, the needle plate is not aligned with the IC carrier, and it is necessary to return to the motor operation step and adjust the needle plate position until it is coaxial. Finally, compare the judgment result with the preset threshold to obtain the needle plate status. If the position and current values ​​are both within the range, the needle plate status is normal. If the current value is out of range, the needle plate force is abnormal. If the position is out of range, the needle plate is not coaxial. According to the status diagnosis results, the system automatically generates needle plate adjustment suggestions, including parameters such as adjustment direction and adjustment amplitude. If the needle plate is subjected to excessive force, it is recommended to reduce the motor torque by 2N·m. If the needle plate is not coaxial, it is recommended to move the needle plate 0.2mm in the positive direction of the X axis. The adjustment suggestions provide data reference for subsequent precise adjustment of the needle plate to ensure the stability and reliability of IC substrate testing.

[0116] In this embodiment, a long short-term memory network is used to predict the state change trend of the needle board, and combined with multi-sensor data fusion technology, the accuracy of state classification and the foresight of prediction are improved, including: obtaining historical state data of the needle board, including information such as the material of the needle board, the number of needles, the type of needles, the elastic coefficient of the needle board and the force of the needle board, and inputting these data into the long short-term memory network for training; predicting the state change trend of the needle board through the long short-term memory network to obtain the state change of the needle board in the future; using multiple sensors to monitor the real-time state of the needle board to obtain information such as the real-time force of the needle board and the real-time position of the needle board; fusing the state change of the needle board predicted by the long short-term memory network with the real-time state of the needle board monitored by the multi-sensor, and comprehensively analyzing the overall state of the needle board; and The state of the needle plate is judged to determine whether the needle plate is in a normal working state. If the needle plate is in a normal working state, the current test plan is maintained; if the needle plate is in an abnormal working state, it is further judged whether the needle plate is coaxial and whether the force of the needle plate squeezing the carrier plate is normal; if the needle plate is not coaxial and the force of the needle plate squeezing the carrier plate is abnormal, then according to the overall state of the needle plate, the corresponding adjustment strategy is adopted to dynamically adjust the parameters such as the force on the needle plate and the position of the needle plate to restore the needle plate to a normal working state; after the adjustment is completed, the real-time state of the needle plate is monitored again by multiple sensors, and the monitoring data is input into the long short-term memory network to re-predict the state change trend of the needle plate; it is repeated continuously to realize the real-time monitoring and dynamic adjustment of the state of the needle plate, to ensure that the needle plate is always in the best working state, and to avoid the problem of test failure caused by damage to the needle plate.

[0117] Exemplarily, the historical state data of the needle plate is first obtained, including information such as the needle plate is made of stainless steel, the number of needles is 1000, the type of needle is spring needle, the elastic coefficient of the needle plate is 200GPa, and the force of the needle plate is 10N. These data are input into an LSTM network consisting of 3 hidden layers for training. Each hidden layer contains 128 neurons, the amount of training data is 10,000 groups, and the number of training iterations is 500. The state change trend of the needle plate is predicted by the trained LSTM network, and the state change of the needle plate in the next hour is predicted. At the same time, 5 six-dimensional force sensors and 3 high-precision displacement sensors are used to monitor the real-time force and position of the needle plate, with a sampling frequency of 100Hz. The state change of the needle plate predicted by the LSTM network is fused with the real-time state monitored by the sensor through Kalman filtering, and the overall state of the needle plate is comprehensively analyzed. When the force change of the needle plate is less than 1N and the position offset of the needle plate is less than 1mm, the needle plate is judged to be in normal working state. When the needle plate is in an abnormal working state, further determine whether the coaxiality deviation of the needle plate is greater than 0.5mm and whether the force of the needle plate squeezing the carrier plate is greater than 20N. If the needle plate is not coaxial and the squeezing force is too large, the PID control algorithm is used according to the overall state of the needle plate to dynamically adjust the force and position of the needle plate so that the coaxiality deviation of the needle plate is less than 0.2mm and the squeezing force is controlled at about 15N. After the adjustment is completed, the sensor is used again to monitor the real-time state of the needle plate, and the monitoring data is input into the LSTM network to re-predict the state change trend of the needle plate, continuously iterate and optimize, realize real-time monitoring and dynamic adjustment of the needle plate state, ensure that the needle plate is always in the best working state, effectively avoid test failures caused by needle plate wear, deformation, etc., and improve test efficiency and accuracy.

[0118] In this embodiment, an adaptive fuzzy neural network algorithm is used to perform online learning and optimization on the needle plate state classification system, and the classification threshold and rule base are dynamically adjusted to improve the system's ability to recognize newly emerging abnormal patterns, including: according to the needle plate information and the number of needles, the elastic coefficient of the needle plate material and the elastic force value of the needle are obtained, and the total elastic force FneedleTotal of the needle and the force Fpressure of the needle plate are calculated; an adaptive fuzzy neural network algorithm is used to classify the needle plate state with the total elastic force FneedleTotal of the needle, the force Fpressure of the needle plate, the real-time value of the motor current and the position of the needle plate relative to the IC carrier as input; through online learning and optimization, the classification threshold and rule base are dynamically adjusted to improve the system's ability to recognize newly emerging abnormal patterns; if the needle plate state classification result is that the needle plate extrusion pressure is normal, it is judged that The needle plate is coaxial, and the current motor output torque is maintained unchanged; if the needle plate state classification result is that the needle plate extrusion pressure is too large, the needle plate is judged to be not coaxial, and the motor output torque is adjusted to reduce the needle plate extrusion pressure until the needle plate extrusion pressure is normal; if the needle plate state classification result is that the needle plate extrusion pressure is too small, the needle plate is judged to be not coaxial, and the motor output torque is adjusted to increase the needle plate extrusion pressure until the needle plate extrusion pressure is normal; if the needle plate state classification result is that the needle plate is not in place, the drive motor continues to move until the needle plate is in place; according to the needle plate state classification result and the motor output torque adjustment, the dynamic control strategy of the needle plate bearing capacity is obtained; the dynamic control strategy is adopted to adjust the motor output torque in real time to maintain the needle plate extrusion pressure within a reasonable range, avoid damage to the needle plate, and ensure the IC substrate test effect.

[0119] Exemplarily, the needle plate material is high-strength alloy steel with an elastic modulus of 210GPa. There are 1000 needles on the needle plate, and the elastic force value of each needle is 5N. According to Hooke's law, the total elastic force of the needle FneedleTotal = 1000×5N = 500N, and the extrusion force Fpressure on the needle plate = 500N. The needle plate state is classified using an adaptive fuzzy neural network algorithm, and the input parameters include the total elastic force of the needle 500N, the force on the needle plate 500N, the motor current 2A, and the position of the needle plate relative to the IC carrier 1mm. The input is mapped to the fuzzy membership through the Gaussian membership function, and the needle plate state membership vector is obtained by combining the I F-THEN rule base reasoning, and then the needle plate state classification result is output through neural network fitting. The normal range of the needle plate extrusion force is set to 480N~520N, and the needle plate coaxial deviation threshold is set to 0.5mm. If the classification result is that the needle plate extrusion force is 505N, which is within the normal range, and the needle plate deviation is 03mm<05mm, the needle plate is determined to be coaxial, and the motor output torque remains unchanged at 2N·m. If the classification result is that the needle plate extrusion force is 530N, which exceeds the normal range, the needle plate is determined to be not coaxial, and the motor output torque is adjusted to 8N·m, reducing the needle plate extrusion force to 505N. If the classification result is that the needle plate extrusion force is 470N, which is lower than the normal range, the needle plate is determined to be not coaxial, and the motor output torque is adjusted to 2N·m, increasing the needle plate extrusion force to 495N. If the classification result is that the needle plate is 2mm relative to the IC carrier board and is not in place, the motor is controlled to continue pressing down at a speed of 01mm / s until the needle plate is in place. Based on the analysis of the needle plate force and the adjustment of the motor, a dynamic control strategy for the needle plate bearing force is formulated, and the motor output torque is adjusted in real time to stabilize the needle plate extrusion force within the range of 500N±20N, avoid deformation and damage of the needle plate, and thus ensure the accuracy and reliability of IC carrier board testing.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. A method for improving the abnormal force of the needle plate squeezing the carrier plate, characterized by: It includes a needle plate and a motor; the needle plate includes an upper needle plate and a lower needle plate; the motor includes an upper motor and a lower motor; the upper motor is used to drive the upper needle plate to move closer to or away from the lower needle plate; the lower motor is used to drive the lower needle plate to move closer to or away from the upper needle plate; The following steps are also included: S1. Get needle plate information; S2. Calculate the total elastic force of the needle and the force on the needle plate; S3, driving the motor to make the needle plate squeeze the IC carrier board; S4, obtain the motor current value and position coordinates; S5, determining whether the current value and the relative position are within a preset range; S6. Determine the needle plate status based on the judgment result and adopt corresponding adjustment strategies.

2. The method for improving the abnormal force of the needle plate squeezing the carrier plate according to claim 1, characterized in that: Calculation of the total elastic force of the needle and the force on the needle plate includes: Obtain the number of needles on the needle plate; obtain the elastic coefficient of the needle plate material; obtain the type of needle and the corresponding elastic force value; calculate the total elastic force of the needle; calculate the force on the needle plate based on the elastic deformation coefficient of the needle plate material.

3. The method for improving the abnormal force of the needle plate squeezing the carrier plate according to claim 1, characterized in that: The drive motor causes the needle plate to squeeze the IC carrier plate including: Compare and calculate the total elastic force of the needle and the force on the needle plate; select the smaller value as the input condition to the motor; calculate the torque output by the motor; drive the motor to make the needle plate squeeze the IC carrier.

4. The method for improving the abnormal force of the needle plate squeezing the carrier plate according to claim 1, characterized in that: The obtaining of the motor current value and position coordinates includes: The motor current value and position coordinate data are collected using the high-precision Hall current sensor and photoelectric encoder installed on the needle board. The raw data is denoised and smoothed using wavelet transform filtering and Kalman filtering algorithms to obtain a cleaned feature data set. The sliding window method is used to extract the key features of current peak value, current fluctuation rate and position deviation from the preprocessed data. The principal component analysis method is used to reduce the dimension of the features and screen out the most representative feature combination. The support vector machine algorithm is used to train the filtered feature data to establish a multi-category classification model for needle plate status. The model parameters are optimized through grid search and cross-validation to improve classification accuracy and generalization ability. The discrimination thresholds of various states are set, and the needle plate state evaluation rule base is constructed in combination with the Mamdani fuzzy inference system for real-time state judgment and classification.

5. The method for improving the abnormal force of the needle plate squeezing the carrier plate according to claim 4, characterized in that: The determining whether the current value and the relative position are within a preset range includes: If the peak value of the motor current collected in real time exceeds the preset threshold and the position coordinate deviation is greater than the set value, it is judged to be an eccentric state. If the current fluctuation rate is abnormally high but the position coordinates are normal, it is judged to be an overvoltage state.

6. The method for improving the abnormal force of the needle plate squeezing the carrier plate according to claim 5, characterized in that: Determining the needle plate state according to the judgment result and taking corresponding adjustment strategies include: By comparing the status classification results with the preset threshold, a needle plate adjustment suggestion is generated, including specific parameters of the adjustment direction and adjustment range, providing data support for precise adjustment. Long-short-term memory network is used to predict the state change trend of the needle plate, combined with multi-sensor data fusion technology to improve the accuracy of state classification and the foresight of prediction; The adaptive fuzzy neural network algorithm is used to conduct online learning and optimization of the needle plate status classification system, dynamically adjust the classification threshold and rule base, and improve the system's ability to recognize newly emerging abnormal patterns.

Images