A method and device for detecting the slot width of a hot ultrasonic wire bonder wedge
By setting multiple standard slots in the thermo-ultrasonic wire bonding machine, performing forward and reverse measurements and establishing a mapping model, the problem of large errors in traditional testing instruments is solved, and high-precision slot width detection is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU SHENCUANG TECH CO LTD
- Filing Date
- 2026-02-13
- Publication Date
- 2026-05-15
AI Technical Summary
Traditional slot width measuring instruments have poor accuracy because the sensor contact movement is small, resulting in insignificant current changes.
By setting up multiple standard slots and performing forward and reverse measurements, a mapping model between sensor readings and actual width is established. Symmetrical and asymmetric errors are separated using forward and reverse measurements, and the final slot width and its reliability are calculated by combining error correction.
High-precision groove width measurement was achieved, and the detection accuracy was improved through error compensation, ensuring the accuracy and reliability of the fitting model.
Smart Images
Figure CN121702260B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of testing technology, and in particular to a method and equipment for detecting the width of a wedge groove in a thermo-ultrasonic wire bonding machine. Background Technology
[0002] Thermo-ultrasonic wire bonding machines are core devices that utilize the synergistic effects of heat, ultrasound, and pressure to bond metal wires between semiconductor chips and external circuits with extremely high precision and speed. The wedge is one of the core components of this equipment; it is a precision tool that directly contacts and manipulates the metal wire to form solder joints on the chip and substrate. Inside the wedge is a through-hole, extremely fine channel. With increasing bonding cycles, friction with the metal wire causes the channel opening to gradually widen. The channel width directly affects the shape and size of the solder joint, requiring regular and rigorous testing and control using a channel width measuring instrument. Traditional channel width measuring instruments detect changes in the conductivity of electrical components within the instrument by moving a sensor contact, which in turn affects the current within the instrument. This current change is then converted into the contact's movement length, thus measuring the channel width. However, because the wedge's channel opening is very small, the sensor contact can only move a small amount, resulting in insignificant changes in the electrical components. Consequently, the current change within the instrument and the contact movement cannot establish an accurate correlation, leading to poor accuracy in channel width measurement. Summary of the Invention
[0003] This invention provides a method for detecting the width of the slot in a cleaver used in a thermo-ultrasonic wire bonding machine, which can effectively solve the problems in the background art.
[0004] This invention provides a method for detecting the slot width of a wedge used in a thermo-ultrasonic wire bonding machine, comprising the following steps:
[0005] Set up J standard slots with different widths, the width of the J standard slots covering the expected range of the slot to be tested;
[0006] For each standard groove, perform multiple forward and reverse measurements and record the sensor readings respectively. Establish a mapping model between the sensor readings and the actual width, and calculate the model uncertainty parameters.
[0007] Perform preliminary forward and reverse measurements on the tank under test and record the sensor readings respectively, then select reliable readings.
[0008] The estimated width of the groove to be measured is calculated from the readings using a mapping model.
[0009] Error correction is applied to the estimated slot width, and the final slot width and its reliability are calculated.
[0010] Specifically, forward measurement and reverse measurement are as follows:
[0011] The two sidewalls of the test tank are designated as the first sidewall and the second sidewall, respectively.
[0012] Stop the sensor contact on the first side wall, clear the sensor reading to zero, move the sensor contact to the second side wall, and record the sensor reading at this time. This is recorded as a positive measurement.
[0013] After a forward measurement, the sensor reading is reset to zero. Then, the sensor contact is moved from the second sidewall to the first sidewall, and the sensor reading is recorded. This is recorded as a reverse measurement.
[0014] Furthermore, the width difference between any two adjacent standard slots is the same.
[0015] Furthermore, the mapping model between sensor readings and actual width is established as follows:
[0016] Let the width of the j-th standard groove be Bwid(j), and the average values of the sensor readings for forward and reverse measurements be Df(j) and Db(j), respectively.
[0017] Calculate the estimated reading of the j-th standard tank, D(j) = (Df(j) - Db(j)) / 2, and the hysteresis value is h(j) = (Df(j) + Db(j)) / 2;
[0018] The standard deviation of all reading estimates for the j-th standard tank is sD(j), and the standard deviation of all hysteresis values is sh(j).
[0019] Establish the design matrix X, weight matrix W, and response vector Y:
[0020] ;
[0021] ;
[0022] Y=[Bwid(1),Bwid(2),…,Bwid(J)] T ;
[0023] Where w(j) is the accuracy weight of the j-th standard slot, w(j) = 1 / (sD(j)). 2 +sh(j) 2 )
[0024] Let the parameter vector be θ = [α, β, γ]. T ;
[0025] Where α, β, and γ are the coefficients of the mapping model;
[0026] Solve the equation (X) T WX)=X T WY, obtain the values of α, β, and γ;
[0027] Construct a mapping model wid(x) = α·Df(x) + β·Db(x) + γ;
[0028] Where wid(x) is the estimated width of the groove x to be measured;
[0029] Df(x) is the reading of the positive measurement of the x-axis of the tank under test;
[0030] Db(x) is the reading of the reverse measurement of the groove x to be measured.
[0031] Furthermore, after constructing the mapping model, the residuals of each standard slot are calculated, and the residual of the j-th standard slot is denoted as δ(j).
[0032] δ(j)=Bwid(j)-[α·Df(j)+β·Db(j)+γ];
[0033] If the standard deviation of all residuals is equal to or greater than the set threshold, the standard slot corresponding to the residual with the largest absolute value is removed, and the mapping model is reconstructed using the readings corresponding to the remaining J-1 standard slots until the standard deviation of the residuals is less than the set threshold.
[0034] Furthermore, the uncertainty parameters of the calculation model are specifically as follows:
[0035] Calculate the standard deviation of the residuals in the model, ρmod.
[0036] ;
[0037] Calculate the parametric covariance matrix M = ρmod 2 (X T WX) -1 ;
[0038] Obtain the model uncertainty parameters: u(α)=sqrt(M(1,1)), u(β)=sqrt(M(2,2)), u(γ)=sqrt(M(3,3)), cov(α,β)=M(1,2), cov(α,γ)=M(1,3) and cov(β,γ)=M(2,3);
[0039] Where M(a,b) is the value corresponding to the a-th row and b-th column in matrix M.
[0040] Furthermore, the final slot width is calculated as follows:
[0041] Record the average forward measurement readings of the reliable readings of the test tank after screening as Dfavg, and the average reverse measurement readings as Dbavg; calculate the hysteresis value of each reliable reading and record the average of all hysteresis values as havg;
[0042] The estimated slot width is calculated using the mapping model as wid = α·Dfavg + β·Dbavg + γ;
[0043] Calculate the quantization error correction Q = havg - △·round(havg / △);
[0044] Where △ represents the minimum resolution of the sensor;
[0045] round() rounds the integer to the nearest integer.
[0046] The final slot width is widfin = wid - Q.
[0047] Furthermore, the confidence level for calculating the final slot width is specifically as follows:
[0048] Calculate the model uncertainty umod=Dfavg 2 ·u(α) 2 +Dbavg 2 ·u(β) 2 +u(γ) 2 +2·Dfavg·Dbavg·cov(α,β)+2·Dfavg·cov(α,γ)+2·Dbavg·cov(β,γ);
[0049] Calculate the quantization uncertainty uqua=△ / sqrt(12);
[0050] The reliability of calculating the final slot width is: uall = sqrt(umod) 2 +uqua 2 ).
[0051] Furthermore, the specific steps for selecting reliable readings for the test cell are as follows:
[0052] A total of M preliminary forward and reverse measurements were performed on the test tank.
[0053] Let the readings of the m-th forward and reverse measurements be Df(m) and Db(m), respectively.
[0054] Calculate the width estimate wid(m) for the m-th iteration as wid(m) = α·Df(m) + β·Db(m) + γ; and the hysteresis value h(m) = (Df(m) + Db(m)) / 2.
[0055] Set the screening width L, and select the top L readings with the smallest combined absolute values of the width estimate deviation and hysteresis value as reliable readings of the test cell.
[0056] Furthermore, when testing the standard slot and the slot under test, the movement parameters of the sensor contact remain the same.
[0057] The present invention also provides a wedge groove width detection device for a thermo-ultrasonic wire bonding machine, comprising a memory and a processor. The memory is used to store one or more program instructions; the processor is used to run one or more program instructions to perform the steps of the above-described thermo-ultrasonic wire bonding machine wedge groove width detection method.
[0058] The technical solution of this invention can achieve the following technical effects:
[0059] This method establishes a mapping relationship between sensor readings and the actual groove width using a standard groove, and then utilizes this relationship to perform high-precision measurement of the groove under test. The scheme combines forward and reverse measurements to separate symmetrical and asymmetrical errors, ensuring the accuracy of the fitted mapping model. Finally, error compensation improves the accuracy of the final groove width measurement. Attached Figure Description
[0060] 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.
[0061] Figure 1 This is a flowchart illustrating the method for detecting the groove width of a wedge used in a thermo-ultrasonic wire bonding machine. Detailed Implementation
[0062] 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.
[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.
[0064] This invention relates to a method for detecting the slot width of a wedge used in a thermo-ultrasonic wire bonding machine, such as... Figure 1 As shown, the main steps include:
[0065] Set up J standard slots with different widths. The width of the J standard slots covers the expected range of the slot to be tested. For example, if the expected width of the slot to be tested is about 2mm, then the width of the J standard slots should be 0.5~3.5mm to cover the expected width of the slot to be tested.
[0066] Perform multiple forward and reverse measurements on each standard tank and record the sensor readings for each measurement. Record the sensor readings for each forward and reverse measurement. The unit of the forward measurement reading is mm and represents a positive value; the unit of the reverse measurement reading is mm and represents a negative value.
[0067] Establish a mapping model between sensor readings and actual width, and calculate the model uncertainty parameters. The uncertainty parameters include multiple parameters, which reflect the uncertainty of the established mapping model from multiple dimensions, so as to represent the difference between the slot width calculated by the model and the actual slot width.
[0068] Perform preliminary forward and reverse measurements on the tank under test and record the sensor readings respectively, then select reliable readings.
[0069] The estimated width of the groove to be measured is calculated from the readings using a mapping model.
[0070] Error correction is applied to the estimated slot width, and the final slot width and its reliability are calculated.
[0071] In the above steps, the forward measurement and reverse measurement are specifically as follows:
[0072] The two sidewalls of the test tank are designated as the first sidewall and the second sidewall, respectively.
[0073] Stop the sensor contact on the first side wall, clear the sensor reading to zero, move the sensor contact to the second side wall, and record the sensor reading at this time. This is recorded as a positive measurement.
[0074] After a forward measurement, the sensor reading is reset to zero. Then, the sensor contact is moved from the second sidewall to the first sidewall, and the sensor reading is recorded. This is recorded as a reverse measurement.
[0075] This forward and reverse measurement method utilizes two forward and reverse movements to cause the electrical components in the sensor to undergo two symmetrical changes. This effectively separates symmetry errors (i.e., width values, which are the same regardless of whether the measurement is forward or reverse) and asymmetry errors (i.e., hysteresis values, where the changes in electrical components during the forward and reverse movements of the sensor are not completely symmetrical). If a problem arises where the changes in electrical components are not obvious or the correspondence is inaccurate due to a small contact movement distance, the hysteresis value will reflect this degree of asymmetry. Then, through subsequent calculations, this asymmetry is corrected, thereby obtaining a more accurate slot width detection value.
[0076] Preferably, the width difference between any two adjacent standard slots is the same, that is, the slot widths of all standard slots will form an arithmetic sequence. This uniformly spaced slot width can best constrain the fitting of the entire model, avoid large oscillations in the fitting results in sparse data regions or excessive sensitivity to individual points, thereby improving the accuracy of the fitted mapping model.
[0077] The mapping model between sensor readings and actual width is established as follows:
[0078] After sequentially testing the first to Jth standard slots, multiple sets of sensor readings for forward and reverse measurements can be obtained;
[0079] Let Bwid(j) be the width of the j-th standard slot;
[0080] The j-th standard tank will undergo multiple forward and reverse measurements, generating multiple forward measurement sensor readings (Df(j,1), Df(j,2), Df(j,3)...) and multiple reverse measurement sensor readings (Db(j,1), Db(j,2), Db(j,3)...). The average value of the multiple forward measurement sensor readings is denoted as Df(j), and the average value of the multiple reverse measurement sensor readings is denoted as Db(j).
[0081] Calculate the estimated value of the k-th group of readings for the j-th standard tank: D(j,k) = (Df(j,k) - Db(j,k)) / 2, and the hysteresis value: h(j,k) = (Df(j,k) + Db(j,k)) / 2
[0082] The standard deviation of all estimated readings (D(j,1), D(j,2), D(j,3)...) for the j-th standard groove is denoted as sD(j), and the standard deviation of all hysteresis values (h(j,1), h(j,2), h(j,3)...) is denoted as sh(j). sD(j) and sh(j) reflect the repeatability accuracy of width measurement and the repeatability accuracy of hysteresis measurement, respectively. The smaller the values, the more accurate the detection of the standard groove, and the higher the reliability of the data.
[0083] Then, based on the principle of weighted least squares algorithm, a linear model is established, and the optimal model parameters are solved based on the precision weights, thereby establishing a mapping model. The specific steps are as follows:
[0084] Firstly, based on the special method of forward and reverse measurement in this approach, the independent variables in the mapping model can be determined as the forward measurement reading Df and the reverse measurement reading Db. The final mapping model can be predicted as follows:
[0085] The slot width wid = α·Df + β·Db + γ;
[0086] α, β, and γ are the coefficients of the mapping model; the physical meaning of α·Df is the contribution of the forward measurement reading to the slot width; the physical meaning of β·Db is the contribution of the reverse measurement reading to the slot width; γ represents the contribution of the inherent deviation of the measurement system to the slot width, which may include: sensor electrical zero point deviation, software zero point setting error, and fixed deviations that the model cannot fully describe, etc., and these effects are quantified by γ.
[0087] Once the form of the mapping model is determined, the actual width of each standard groove and the forward and reverse measurement readings obtained from each standard groove can be substituted into the assumed mapping model to solve for the values of α, β, and γ in reverse. The solution process is as follows:
[0088] Establish the design matrix X, weight matrix W, and response vector Y:
[0089] ;
[0090] Each row in the design matrix X is an eigenvector (Df,Db,1) that describes the forward and reverse measurement readings of each standard slot. The influence of the system is the same for any number of measurements, so it can be denoted as 1.
[0091] ;
[0092] Where w(j) is the accuracy weight of the j-th standard slot, w(j) = 1 / (sD(j)). 2 +sh(j) 2 ), which represents the reliability of the data measured by the j-th standard cell. As mentioned above, the more accurate the detection of the standard cell, the smaller the values of sD(j) and sh(j) will be, and the larger w(j) will be overall, thereby amplifying the weight of the measurement reading of the standard cell in subsequent calculations.
[0093] Y=[Bwid(1),Bwid(2),…,Bwid(J)] T ;
[0094] The response vector Y is the actual width of each known standard slot.
[0095] Let the variables α, β, and γ that we need to find be represented as a parameter vector θ = [α, β, γ]. T ;
[0096] Establish equation (X) T WX)θ=X TWY, mathematically speaking, means that for vector θ, the weighted sum of squared deviations between the predicted slot width wid=α·Df+β·Db+γ and the true value Bwid is minimized in the calculations corresponding to all standard slots, that is, the difference between the predicted value and the true value is minimized.
[0097] Then only the equation (X) needs to be solved. T WX)θ=X T By performing WY, the numerical value of vector θ can be obtained. Then, the corresponding α, β, and γ in vector θ can also be calculated. After that, we only need to substitute the values of α, β, and γ into the previously assumed mapping model to obtain the final mapping model:
[0098] wid(x)=α·Df(x)+β·Db(x)+γ;
[0099] Where wid(x) is the estimated width of the groove x to be measured;
[0100] Df(x) is the reading of the positive measurement of the x-axis of the tank under test;
[0101] Db(x) is the reading of the reverse measurement of the groove x to be measured.
[0102] After constructing the mapping model, it is necessary to verify its usability. If the mapping model has a large error, it needs to be adjusted in a timely manner. This method provides a verification approach, specifically:
[0103] Calculate the residuals of each standard tank, and denote the residual of the j-th standard tank as δ(j);
[0104] δ(j)=Bwid(j)-[α·Df(j)+β·Db(j)+γ];
[0105] The residual δ(j) represents the difference between the predicted width of the j-th standard slot α·Df(j)+β·Db(j)+γ by the mapping model and the actual width of the j-th standard slot Bwid(j).
[0106] For a usable model, all residuals δ should be close to zero and randomly distributed without obvious trends or patterns. Therefore, if the standard deviation of all residuals is equal to or greater than the set threshold, it indicates that the model has a large bias. The set of data with the largest residual needs to be removed. That is, after removing the standard slot data corresponding to the residual with the largest absolute value, the mapping model is reconstructed using the readings corresponding to the remaining J-1 standard slots.
[0107] It is important to note that although the data removed has a large error for the obtained mapping model, assuming the removed data is from the j1-th standard slot, the j1-th standard slot data may not be the data with the largest error during measurement. Therefore, if the data with the largest error is re-involved in building the mapping model, the constructed mapping model is very likely to still be unusable. Another set of data with the largest residual will be removed. Assuming the data to be removed is from the j2-th standard slot, then the previously removed j1-th standard slot data needs to be reused to ensure that there are readings corresponding to J-1 standard slots to rebuild the mapping model, until the standard deviation of the residual is less than the set threshold.
[0108] The specific parameters for calculating the model uncertainty are as follows:
[0109] Calculate the standard deviation of the residuals in the model, ρmod.
[0110] ;
[0111] ρmod is the root mean square error of the model prediction after weighting and degree of freedom correction. It represents how much the mapping model itself has the error in predicting new data during detection.
[0112] Calculate the parametric covariance matrix M = ρmod 2 (X T WX) -1 ;
[0113] The covariance matrix M, after calculation, will be a 3×3 matrix, which represents the confidence level of the parameters themselves and the mutual influence of the parameters. The form of the covariance matrix M after solving is as follows:
[0114] ;
[0115] Obtain the model uncertainty parameters: u(α)=sqrt(M(1,1)), u(β)=sqrt(M(2,2)), u(γ)=sqrt(M(3,3)), cov(α,β)=M(1,2), cov(α,γ)=M(1,3) and cov(β,γ)=M(2,3);
[0116] The sqrt() function calculates the square root of the expression within the parentheses.
[0117] M(a,b) is the value corresponding to the a-th row and b-th column in matrix M.
[0118] u(α), u(β), and u(γ) represent the confidence levels of α, β, and γ themselves, respectively; cov(α,β), cov(α,γ), and cov(β,γ) represent the correlation between the estimation errors of two coefficients. For example, cov(α,β) represents the correlation between the estimation errors of α and β. This correlation is due to the correlation between the forward measurement reading Df and the reverse measurement reading Db (forward and reverse measurements from the same slot). Ignoring the covariance term will cause omissions in the subsequent uncertainty propagation, affecting the accuracy of the final confidence level calculation.
[0119] The final slot width is calculated as follows:
[0120] Record the average forward measurement readings of the reliable readings of the test tank after screening as Dfavg, and the average reverse measurement readings as Dbavg; calculate the hysteresis value of each reliable reading and record the average of all hysteresis values as havg;
[0121] The estimated slot width is calculated using the mapping model as wid = α·Dfavg + β·Dbavg + γ.
[0122] Calculate the quantization error correction Q = havg - △·round(havg / △);
[0123] Where △ represents the minimum resolution of the sensor. For example, if the sensor can only resolve to a minimum of 0.05mm, then △ = 0.05mm.
[0124] round() means rounding to the nearest integer. The principle of this algorithm is: under ideal noise-free conditions, for a stable system, the hysteresis value havg after averaging multiple measurements should stably fall on a certain quantization level of the sensor output discretization, that is, it should be an integer multiple of Δ.
[0125] The physical meaning of Q is the deviation between the observed average hysteresis value havg and the most recent ideal quantization level. This deviation is mainly due to the systematic offset caused by rounding errors in the quantization process. When calculating the final slot width, it needs to be subtracted from the slot width estimate calculated by the mapping model to obtain the final slot width widfin=wid-Q.
[0126] The reliability of calculating the final slot width is as follows:
[0127] Calculate the model uncertainty umod=Dfavg 2 ·u(α) 2 +Dbavg 2 ·u(β) 2 +u(γ) 2+2·Dfavg·Dbavg·cov(α,β)+2·Dfavg·cov(α,γ)+2·Dbavg·cov(β,γ);
[0128] Calculate the quantization uncertainty uqua=△ / sqrt(12);
[0129] The reliability of calculating the final slot width is: uall = sqrt(umod) 2 +uqua 2 ).
[0130] The physical meaning of the reliability uall is as follows: Assuming that model error and quantization error are independent sources, the total standard uncertainty is the square root of the sum of squares of the independent terms. uall is the combined standard uncertainty of the final slot width widfin, which quantitatively represents the dispersion of the measurement results, i.e., the reliability. The smaller the value, the higher the reliability, meaning the more accurate the final slot width widfin value.
[0131] Preferably, the reliable readings of the test cell are as follows:
[0132] A total of M preliminary forward and reverse measurements were performed on the test tank.
[0133] Let the readings of the m-th forward and reverse measurements be Df(m) and Db(m), respectively.
[0134] Calculate the width estimate wid(m) for the m-th iteration as wid(m) = α·Df(m) + β·Db(m) + γ; and the hysteresis value h(m) = (Df(m) + Db(m)) / 2.
[0135] Calculate the absolute values of the deviation and hysteresis for each width estimate. The smaller the deviation, the closer the width estimate is to the average level, and the lower the probability of error. The smaller the hysteresis, the closer the current changes in the forward and reverse measurements are during detection, and the lower the probability of error.
[0136] Set a screening width L, and select the top L readings with the smallest combined deviation of width estimate (i.e., the difference between each width estimate and the average of all width estimates) and absolute value of hysteresis as reliable readings of the tank to be measured. For example, multiply the deviation of width estimate by the absolute value of hysteresis, and then arrange them from smallest to largest. The top L with smaller values are the data that are relatively accurate and less likely to have errors during measurement.
[0137] When testing the standard slot and the slot under test, the movement parameters of the sensor contact must be kept the same. For example, the movement speed and the trigger force of the electric shock signal. This ensures consistency at the physical level and avoids inaccuracies in the mapping model caused by changes in physical properties. For instance, if the speed is inconsistent, such as if the speed is too fast during a measurement, it may cause slight differences in the changes of electrical components when the contact moves the same distance, resulting in inconsistent changes in the current within the sensor. This, in turn, affects the accuracy of the readings and causes the mapping model to fail.
[0138] The present invention also relates to a wedge groove width detection device for a thermo-ultrasonic wire bonding machine, comprising a memory and a processor, wherein the memory is used to store one or more program instructions; and the processor is used to run one or more program instructions to perform the steps of the above-described method for detecting the wedge groove width of a thermo-ultrasonic wire bonding machine.
[0139] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.
Claims
1. A method for detecting the groove width of a wedge used in a thermo-ultrasonic wire bonding machine, characterized in that the steps include... include: Set up J standard slots with different widths, the width of the J standard slots covering the expected range of the slot to be tested; For each standard groove, perform multiple forward and reverse measurements and record the sensor readings respectively. Establish a mapping model between the sensor readings and the actual width, and calculate the model uncertainty parameters. Perform preliminary forward and reverse measurements on the tank under test and record the sensor readings respectively, then select reliable readings. The estimated width of the groove to be measured is calculated from the readings using a mapping model. Error correction is applied to the estimated slot width, and the final slot width and its reliability are calculated. Specifically, forward measurement and reverse measurement are as follows: The two sidewalls of the test tank are designated as the first sidewall and the second sidewall, respectively. Stop the sensor contact on the first side wall, clear the sensor reading to zero, move the sensor contact to the second side wall, and record the sensor reading at this time. This is recorded as a positive measurement. After a forward measurement, the sensor reading is cleared to zero. Then, the sensor contact is moved from the second sidewall to the first sidewall, and the sensor reading is recorded. This is recorded as a reverse measurement. The mapping model between sensor readings and actual width is established as follows: Let the width of the j-th standard groove be Bwid(j), and the average values of the sensor readings for forward and reverse measurements be Df(j) and Db(j), respectively. Calculate the estimated value of the k-th group of readings for the j-th standard tank: D(j,k) = (Df(j,k) - Db(j,k)) / 2, and the hysteresis value is h(j,k) = (Df(j,k) + Db(j,k)) / 2; The standard deviation of all reading estimates for the j-th standard tank is sD(j), and the standard deviation of all hysteresis values is sh(j). Establish the design matrix X, weight matrix W, and response vector Y: ； ； Y=[Bwid(1),Bwid(2),……,Bwid(J)] T ; Where w(j) is the accuracy weight of the j-th standard slot, w(j) = 1 / (sD(j)). 2 +sh(j) 2 ); Let the parameter vector be θ = [α, β, γ]. T ; Where α, β, and γ are the coefficients of the mapping model; Solve the equation (X) T WX)θ=X T WY, obtain the values of α, β, and γ; The mapping model is obtained as: wid(x) = α·Df(x) + β·Db(x) + γ; Where wid(x) is the estimated width of the groove x to be measured; Df(x) is the reading of the positive measurement of the x-axis of the tank under test; Db(x) is the reading of the reverse measurement of the test tank x; After constructing the mapping model, calculate the residual of each standard slot, and denote the residual of the j-th standard slot as δ(j); δ(j)=Bwid(j)-[α·Df(j)+β·Db(j)+γ]; If the standard deviation of all residuals is equal to or greater than the set threshold, the standard slot corresponding to the residual with the largest absolute value is removed, and the mapping model is reconstructed using the readings corresponding to the remaining J-1 standard slots until the standard deviation of the residuals is less than the set threshold. The final slot width is calculated as follows: Record the average forward measurement readings of the reliable readings of the test tank after screening as Dfavg, and the average reverse measurement readings as Dbavg; calculate the hysteresis value of each reliable reading and record the average of all hysteresis values as havg; The estimated slot width is calculated using the mapping model as wid = α·Dfavg + β·Dbavg + γ; Calculate the quantization error correction Q = havg - △·round(havg / △); Where △ represents the minimum resolution of the sensor; round() rounds the integer to the nearest integer. The final slot width is widfin = wid - Q.
2. The method for detecting the wedge groove width of a thermo-ultrasonic wire bonding machine according to claim 1, characterized in that, The width difference between any two adjacent standard slots is the same.
3. The method for detecting the wedge groove width of a thermo-ultrasonic wire bonding machine according to claim 1, characterized in that, The specific parameters for calculating the model uncertainty are as follows: Calculate the standard deviation of the residuals in the model, ρmod. ； Calculate the parametric covariance matrix M = ρmod 2 (X T WX) -1 ; Obtain the model uncertainty parameters: u(α)=sqrt(M(1,1)), u(β)=sqrt(M(2,2)), u(γ)=sqrt(M(3,3)), cov(α,β)=M(1,2), cov(α,γ)=M(1,3) and cov(β,γ)=M(2,3); Where M(a,b) is the value corresponding to the a-th row and b-th column in matrix M.
4. The method for detecting the slot width of a wedge used in a thermo-ultrasonic wire bonding machine according to claim 3, characterized in that, The reliability of calculating the final slot width is as follows: Defaulting in the uppercase umod=Dfavg 2 ·u(α) 2 +Dbavg 2 ·u(β) 2 +u(γ) 2 +2·Dfavg·Dbavg·cov(α,β)+2·Dfavg·cov(α,γ)+2·Dbavg·cov(β,γ) Calculate the quantization uncertainty uqua=△ / sqrt(12); The reliability of calculating the final slot width is: uall = sqrt(umod) 2 +uqua 2 ).
5. The method for detecting the wedge groove width of a thermo-ultrasonic wire bonding machine according to claim 1, characterized in that, The specific reliable readings for screening the test tank are as follows: A total of M preliminary forward and reverse measurements were performed on the test tank. Let the readings of the m-th forward and reverse measurements be Df(m) and Db(m), respectively. Calculate the width estimate wid(m) for the m-th iteration as wid(m) = α·Df(m) + β·Db(m) + γ; and the hysteresis value h(m) = (Df(m) + Db(m)) / 2. Set the screening width L, and select the top L readings with the smallest combined absolute values of the width estimate deviation and hysteresis value as reliable readings of the test cell.
6. The method for detecting the wedge groove width of a thermo-ultrasonic wire bonding machine according to claim 1, characterized in that, When testing the standard slot and the slot under test, the movement parameters of the sensor contact remain the same.
7. A device for detecting the width of a wedge groove in a thermo-ultrasonic wire bonding machine, characterized in that, It includes a storage device and a processor, the storage device being used to store one or more program instructions; the processor being used to execute one or more program instructions for performing the steps of the method for detecting the slot width of a wedge for a thermo-ultrasonic wire bonding machine as described in any one of claims 1 to 6.