Semiconductor punching production equipment detection and diagnosis system and method

Abstract

The present invention belongs to the field of equipment detection technology, and specifically relates to a semiconductor punching production equipment detection and diagnosis system and method. It realizes real-time collection and stable transmission of physical parameters by integrating high-precision monitoring devices to ensure data integrity and timeliness; uses dynamic data analysis to focus on the changes in variables that affect punching quality, and makes instant fine-tuning to reduce problems caused by parameter instability; combines with a visual system to perform high-precision edge detection, accurately evaluates the edge smoothness of each product and automatically records abnormal conditions; through rapid tracing and detailed review of abnormal conditions, determines the specific factors that cause burrs, and formulates targeted improvement measures. This method significantly improves the controllability of the punching process and product quality, and enhances production stability and product consistency.

Description

Technical Field

[0001] The invention belongs to the technical field of equipment detection, and in particular relates to a detection and diagnosis system and method for semiconductor punching production equipment. Background Art

[0002] In the semiconductor manufacturing industry, the punching process is a key step in cutting wafers into individual chips. Existing technologies usually use fixed parameter settings for punching operations, relying on pre-set tool paths, speeds, pressures and other parameters. In order to ensure product quality, some monitoring equipment is generally installed on the production line to collect physical parameters such as force, vibration and temperature during the punching process, and process these data through offline analysis. However, this traditional approach has certain limitations:

[0003] Data collection and transmission: Although a large amount of real-time data can be collected, existing systems often lack efficient data stream transmission mechanisms, resulting in data delays or losses, affecting the accuracy and timeliness of subsequent analysis.

[0004] Data analysis and adjustment: Traditional methods are mostly post-analysis, which cannot achieve real-time dynamic adjustment. Even if the analysis is carried out, it is difficult to quickly feedback to the production equipment for immediate fine-tuning, which may cause fluctuations in product quality.

[0005] Image capture and edge detection: Although some production lines are equipped with vision systems to check the edge quality of products, these systems have limited accuracy and automation and cannot effectively identify subtle burr problems.

[0006] Abnormal tracing and improvement: When quality problems are discovered, the process of tracing related production batches and reviewing parameter records is cumbersome and time-consuming, making it difficult to quickly locate the root cause of the problem and take effective measures.

[0007] The main problem of the above existing technologies is that they cannot achieve closed-loop management of the entire process from data collection to problem solving, especially in dealing with the fine burrs generated during punching and cutting. Specifically, due to the lack of efficient real-time data processing capabilities and accurate edge detection methods, the existing detection and diagnosis methods are difficult to timely discover and correct the specific factors that cause burrs, which in turn affects the quality and consistency of the final product. Summary of the invention

[0008] The purpose of the present invention is to provide a semiconductor punching production equipment detection and diagnosis system and method. By integrating high-precision monitoring devices, real-time data stream transmission paths, dynamic data analysis, visual systems and detailed retrospective analysis, this method can significantly improve the controllability of the punching process and product quality, so as to solve the problems in the prior art raised in the above background technology.

[0009] To achieve the above-mentioned object, the present invention proposes a semiconductor punching production equipment detection and diagnosis method, comprising the following steps:

[0010] Multiple monitoring devices are configured on the punching equipment to collect physical parameters during the punching process, and a real-time data stream transmission path is established based on the data obtained by the monitoring devices;

[0011] Use the data flow transmission path to implement dynamic data analysis, focus on the changes in variables that affect the punching quality, and fine-tune the punching operation based on the analysis results;

[0012] Combined with the adjustment results, the image of the punched product is captured by the visual system, and edge detection is performed. Based on the obtained product image information, the edge smoothness of each product is evaluated and abnormal conditions are recorded;

[0013] For recorded abnormal situations, trace the relevant production batches, review all parameter records within the batches for retrospective analysis, determine the specific factors causing the burrs based on the retrospective analysis, and formulate improvement measures accordingly.

[0014] Preferably, the punching equipment is provided with a plurality of monitoring devices, including:

[0015] Select multiple locations on the punching equipment to install monitoring devices, each device is responsible for collecting a set of physical parameters;

[0016] Using the installed monitoring device, the force F, vibration V and temperature T data during the punching process are recorded at fixed time intervals ΔT;

[0017] Based on the collected data, the average force of each punching operation is calculated as FAvg = (F1 + F2 + ... + Fn) / n, where n represents the number of data points recorded in the time interval ΔT;

[0018] Based on the obtained average force FAvg and the punching speed S, the energy consumption E of each punching is evaluated using the formula E=FAvg*S.

[0019] Preferably, establishing a real-time data stream transmission path based on the data acquired by the monitoring device includes:

[0020] Receive raw physical parameter data from the monitoring device, the data is represented in the form of time series, denoted as D(t), where t is the timestamp;

[0021] Using data D(t), a buffer is constructed. The buffer stores all data points in the most recent time period according to the first-in-first-out principle.

[0022] Based on the buffer, calculate the time difference of each data point Δτ=τ(n)-τ(n-1), and ensure that Δτ does not exceed the preset maximum allowable value MaxΔτ;

[0023] According to the obtained time difference Δτ, the quality index QoS of the data flow is evaluated by the formula QoS=Min(1,MaxΔτ / Δτ).

[0024] Preferably, the dynamic data analysis is performed using the data stream transmission path, focusing on the changes in variables that affect the punching quality, including:

[0025] Receive the constructed data flow quality index QoS value and physical parameter data D(t), and convert them into a standardized format;

[0026] Based on the standardized data, calculate the rate of change of each variable ΔV=(V(n)-V(n-1)) / Δτ, where V represents any physical parameter, including: force F, vibration V or temperature T;

[0027] Using the obtained change rate ΔV, determine the key variable set KV that affects the punching quality, select variables that meet the condition |ΔV|>Threshold and add them to KV, where Threshold is the preset threshold;

[0028] According to the determined key variable set KV, the impact score ImpactScore of each punching operation is calculated by the formula ImpactScore=Σ(|ΔV(i)|*Weight(i)), where Weight(i) is the weight assigned to each variable.

[0029] Preferably, the step of fine-tuning the punching operation according to the analysis results comprises:

[0030] According to the impact score ImpactScore, determine the punching parameter set P that needs to be adjusted, where each parameter p(j) corresponds to an impact score;

[0031] Based on the parameter set P, an adjustment factor AdjFactor(j) is set for each parameter p(j), and the coefficient is predetermined based on historical data and experimental results;

[0032] Using the adjustment factor AdjFactor(j), through the formula:

[0033] NewValue(j)=p(j)+AdjFactor(j)*ImpactScore calculates the new value NewValue(j) of each parameter;

[0034] According to the obtained new value NewValue(j), fine-tuning operation is performed and the changes before and after the adjustment are recorded in ChangeLog.

[0035] Preferably, the combining the adjustment result, capturing the image of the punched product through a visual system, and performing edge detection, comprises:

[0036] After the punching operation is completed, the vision system is immediately started to capture the image of each product I, ensuring that the image covers the entire edge of the product;

[0037] Based on the acquired image I, the grayscale conversion formula is applied:

[0038] GrayLevel(x,y)=0.299*R(x,y)+0.587*G(x,y)+0.114*B(x,y), converts the color image to a grayscale image, where R, G, and B are the red, green, and blue channel values, respectively, and x and y are the pixel coordinates;

[0039] Using the obtained grayscale image, the edge detection formula is:

[0040] Edge(x,y)=|Gx(x,y)|+|Gy(x,y)|, calculates the edge strength of each pixel, where Gx and Gy represent the gradients in the horizontal and vertical directions respectively;

[0041] According to the edge strength Edge(x,y), determine the edge contour of the product, compare it with the standard template, and record the deviation Deviation.

[0042] Preferably, the step of evaluating the edge smoothness of each product based on the obtained product image information and recording abnormalities includes:

[0043] Based on the determined product edge contour, calculate the curvature K(p,q) of each pixel along the edge path using the formula:

[0044] K(p,q)=(Gx(p,q)*Gyy(p,q)-2*Gy(p,q)*Gxy(p,q)+Gx(p,q)*Gxx(p,q)) / (Gx(p,q)^2+Gy(p,q)^2)^(3 / 2), where Gxx, Gyy and Gxy are the second-order partial derivatives of the grayscale image in the p and q directions respectively;

[0045] Using the curvature K(p,q), we can calculate the average curvature of all pixels on the edge path, KAverage=ΣK(p,q) / N, where N is the total number of pixels on the edge path.

[0046] According to the obtained average curvature KAverage, a smoothness threshold SmoothnessThreshold is set. When KAverage exceeds this threshold, the product is marked as a potential abnormal product.

[0047] For the marked potential abnormal products, their position coordinates, curvature distribution and deviation from the standard template are recorded in detail (DeviationLog).

[0048] Preferably, for the recorded abnormal situation, tracing back the relevant production batches, reviewing all parameter records in the batches for retrospective analysis, including:

[0049] For products marked as potential abnormal products, extract their production batch identifier BatchID and create a batch product list ProductList;

[0050] Based on BatchID, access the production database to retrieve all parameter records of the batch to form a parameter record set ParamsSet, where each record contains a timestamp, punching parameters and environmental conditions;

[0051] Using ParamsSet, calculate the coefficient of variation CV = σ / μ for each parameter within the batch, where σ is the standard deviation of the parameter and μ is the mean value;

[0052] According to the obtained coefficient of variation CV, the unstable parameters whose coefficient of variation exceeds the preset threshold CVThreshold are identified, and a review report ReviewReport is generated, which lists the time period and specific values ​​of the parameters in detail.

[0053] Preferably, the specific factors causing the burrs are determined based on the retrospective analysis, and improvement measures are formulated accordingly, including:

[0054] Based on the generated review report ReviewReport, all unstable parameters whose coefficient of variation CV exceeds the preset threshold CVThreshold are identified and classified as potential influencing factors SetFactors;

[0055] SetFactors is used in combination with the product edge smoothness evaluation results to calculate the correlation factor CorrelationFactor between each unstable parameter and burr generation through the formula CorrelationFactor=Σ(|KAverage(k)-KTarget|*CV(k)), where KAverage(k) is the average curvature of the k-th product, KTarget is the target smoothness, and CV(k) is the coefficient of variation of the k-th parameter;

[0056] According to the obtained correlation factor CorrelationFactor, determine the specific factors CriticalFactors that cause the burr, select the factors that meet the condition CorrelationFactor>CorrThreshold and add them to the CriticalFactors set, where CorrThreshold is the preset correlation threshold;

[0057] According to the identified Critical Factors, formulate corresponding improvement measures Action Plan, which includes adjustment suggestions, expected results and implementation schedule.

[0058] On the other hand, the present invention provides a semiconductor punching production equipment detection and diagnosis system, comprising:

[0059] A data acquisition and transmission module is used to configure multiple monitoring devices on the punching equipment to collect physical parameters during the punching process and establish a real-time data stream transmission path based on the data obtained by the monitoring devices;

[0060] Dynamic data analysis and fine-tuning module, which is used to implement dynamic data analysis using the data flow transmission path, focusing on the changes in variables that affect the punching quality, and fine-tuning the punching operation based on the analysis results;

[0061] An image capture and edge smoothness evaluation module is used to combine the adjustment results, capture the image of the punched product through the visual system, and perform edge detection. Based on the obtained product image information, the edge smoothness of each product is evaluated and abnormal conditions are recorded;

[0062] The module of retrospective analysis and improvement measures formulation is used to trace the relevant production batches for the recorded abnormal situations, and to review all parameter records within the batches for retrospective analysis. Based on the retrospective analysis, the specific factors causing the burrs are determined, and improvement measures are formulated accordingly.

[0063] Technical effects and advantages of the present invention: The semiconductor punching production equipment detection and diagnosis system and method proposed by the present invention have the following advantages compared with the prior art:

[0064] The present invention realizes real-time acquisition and stable transmission of physical parameters by integrating high-precision monitoring devices to ensure data integrity and timeliness; uses dynamic data analysis to focus on the changes in variables that affect punching quality, and makes instant fine-tuning to reduce problems caused by unstable parameters; combines with the visual system to perform high-precision edge detection, accurately evaluates the edge smoothness of each product and automatically records abnormalities; through rapid tracing and detailed review of abnormalities, determines the specific factors that cause burrs, and formulates targeted improvement measures. This method significantly improves the controllability of the punching process and product quality, and enhances production stability and product consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flow chart of the semiconductor punching production equipment detection and diagnosis method of the present invention;

[0066] Figure 2 It is a block diagram of the semiconductor punching production equipment detection and diagnosis system of the present invention. DETAILED DESCRIPTION

[0067] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0068] The present invention provides Figure 1 The semiconductor punching production equipment detection and diagnosis method shown includes the following steps:

[0069] Step 1: arranging a plurality of monitoring devices on the punching equipment to collect physical parameters during the punching process; further comprising:

[0070] Monitoring devices are installed at multiple locations on the punching equipment, and each device is responsible for collecting a set of physical parameters. By installing monitoring devices at multiple key locations, comprehensive coverage of various physical parameters during the punching process is ensured. This helps to more accurately reflect changes in punching conditions and provide detailed basic data for subsequent data analysis.

[0071] The installed monitoring device records the force F, vibration V and temperature T data during the punching process at a fixed time interval ΔT; the data recording method with a fixed time interval ΔT ensures the continuity and consistency of the data, making the data analysis more reliable. At the same time, high-frequency data acquisition can capture instantaneous changes and improve the sensitivity of detection.

[0072] Based on the collected data, the average force of each punching operation is calculated as FAvg = (F1 + F2 + ... + Fn) / n, where n represents the number of data points recorded in the time interval ΔT; calculating the average force FAvg can smooth out the impact of short-term fluctuations and obtain an indicator that better reflects the long-term trend. This helps to identify the key factors that truly affect the quality of punching and reduce the interference of outliers.

[0073] Based on the average force FAvg, combined with the punching speed S, the energy consumption E of each punching is evaluated by the formula E=FAvg*S. The evaluation of energy consumption E can quantify the efficiency of the punching process, help optimize the punching parameter settings, improve energy utilization efficiency, and reduce production costs. In addition, energy consumption is also an important reference indicator for judging the degree of tool wear.

[0074] Assume that there is a semiconductor punching equipment, on which three key positions are selected to install monitoring devices, which are responsible for collecting data of force F, vibration V and temperature T. The fixed time interval ΔT is set to 0.1 seconds, that is, data is recorded every 0.1 seconds.

[0075] In one complete punching operation, a total of 10 data points (n=10) were recorded, corresponding to a time interval of 1 second.

[0076] The recorded force data are: F1=5N, F2=6N, F3=5.5N, F4=5.8N, F5=6.2N, F6=5.9N, F7=6.1N, F8=5.7N, F9=6.0N, F10=5.6N.

[0077] From these data, calculate the average force FAvg:

[0078] FAvg=(F1+F2+...+Fn) / n

[0079] =(5+6+5.5+5.8+6.2+5.9+6.1+5.7+6.0+5.6) / 10

[0080] =58.8 / 10

[0081] =5.88N

[0082] Assuming the punching speed S is 0.5m / s, the energy consumption E for each punching can be calculated by the following formula:

[0083] E=FAvg*S

[0084] =5.88*0.5

[0085] =2.94J.

[0086] Therefore, the energy consumption of this punching operation is 2.94 joules. In this way, the energy consumption of each punching operation can be continuously monitored, and the punching parameters can be adjusted in time to ensure that the equipment operates in the optimal state, thereby improving product quality and saving energy.

[0087] Step 2: Establishing a real-time data stream transmission path based on the data acquired by the monitoring device; further comprising:

[0088] Receive the original physical parameter data from the monitoring device, and the data is expressed in the form of time series, recorded as D(t), where t is the timestamp; receive the original physical parameter data from the monitoring device and record it in the form of time series. This method ensures the time continuity and traceability of the data, providing a solid foundation for subsequent real-time analysis.

[0089] Using data D(t), a buffer is constructed. The buffer stores all data points in the most recent time period according to the first-in-first-out principle. By constructing a buffer, a large amount of real-time data can be effectively managed to ensure the real-time and efficient data processing. The first-in-first-out (FIFO) principle ensures that new data replaces old data in a timely manner, avoids data backlog, and improves the response speed of the system.

[0090] Based on the buffer, the time difference Δτ=τ(n)-τ(n-1) of each data point is calculated, and Δτ is ensured not to exceed the preset maximum allowable value MaxΔτ; the time difference Δτ between adjacent data points is calculated to help evaluate the stability of data collection and transmission. Setting the maximum allowable value MaxΔτ can prevent data delay or loss and ensure the reliability and consistency of data flow.

[0091] According to the obtained time difference Δτ, the quality index QoS of the data stream is evaluated by the formula QoS=Min(1,MaxΔτ / Δτ). The quality index QoS is used to evaluate the stability and reliability of the data stream. This index reflects the effectiveness of the data transmission path and helps identify potential problem points, so that corresponding optimization measures can be taken to ensure the high quality of the entire data stream transmission process.

[0092] Assume that there is a semiconductor punching equipment with multiple monitoring devices installed on it to collect physical parameters such as force, vibration and temperature. These data are expressed in the form of time series, recorded as D(t). The following are the specific implementation steps:

[0093] Receive raw physical parameter data:

[0094] For example, in a complete punching operation, the monitoring device records data every 0.1 seconds, generating the following time series data:

[0095] D(t)={(t1,F1),(t2,F2),...,(tn,Fn)}

[0096] Constructing the buffer Buffer:

[0097] Suppose a buffer with a capacity of 50 data points is set up, and the latest 50 data points are stored according to the first-in-first-out principle. When a new data point arrives, the oldest data point will be removed to ensure that the buffer always contains the latest data.

[0098] Calculate the time difference Δτ and check MaxΔτ:

[0099] For each pair of adjacent data points, calculate the time difference Δτ between them. Assume that the maximum allowed time difference MaxΔτ is set to 0.15 seconds. If the Δτ calculated in a certain time exceeds this value, an alarm is triggered, indicating that there may be a data transmission problem.

[0100] Δτ=τ(n)-τ(n-1)

[0101] Evaluate the quality index QoS of data flow:

[0102] The formula QoS = Min (1, Max Δτ / Δτ) is used to evaluate the quality index of the data flow. For example, for a time difference Δτ = 0.12 seconds, the calculation is:

[0103] QoS=Min(1,0.15 / 0.12)

[0104] =Min(1,1.25)

[0105] =1.

[0106] If Δτ is small, close to 0.1 seconds, the QoS is close to 1, indicating that the data flow is very stable; if Δτ is large, close to 0.15 seconds, the QoS is reduced, indicating that the data flow may have unstable factors.

[0107] In this way, the quality of the data stream can be continuously monitored to ensure the stability and reliability of real-time data transmission, thereby supporting subsequent dynamic data analysis and precise adjustment of the punching operation.

[0108] Step 3: Implement dynamic data analysis using the data flow transmission path, focusing on the changes in variables that affect the punching quality; further including:

[0109] Receive the constructed data flow quality index QoS value and physical parameter data D(t) and convert them into a standardized format; receive and process high-quality data flows (QoS values) and physical parameter data D(t) from the buffer to ensure data consistency and reliability. Processing these data in a standardized format can eliminate dimensional differences between different variables, making subsequent analysis more accurate and intuitive.

[0110] Based on the standardized data, the change rate of each variable is calculated as ΔV = (V(n) - V(n-1)) / Δτ, where V represents any physical parameter, including force F, vibration V or temperature T; calculating the change rate of each physical parameter ΔV can capture the instantaneous change of the variable over time and help identify the rapid change points that may cause quality problems. This step is crucial to discovering potential problems because it can provide early warning of abnormal situations.

[0111] Using the obtained change rate ΔV, determine the key variable set KV that affects the punching quality, select variables that meet the condition |ΔV|>Threshold to add to KV, Threshold is the preset threshold; by setting the threshold Threshold, filter out the variables with significant changes to form the key variable set KV. This helps to focus on the factors that really affect the punching quality, reduce unnecessary interference information, and make subsequent analysis more targeted and efficient.

[0112] According to the determined key variable set KV, the impact score ImpactScore of each punching operation is calculated by the formula ImpactScore=Σ(|ΔV(i)|*Weight(i)), where Weight(i) is the weight assigned to each variable. The calculation of the impact score ImpactScore comprehensively considers the change range of each key variable and its importance to the punching quality. This quantitative evaluation method can provide an intuitive quality indicator to help decision makers quickly understand the quality status of each punching operation and take appropriate adjustment measures accordingly.

[0113] Assume that there is a semiconductor punching equipment with multiple monitoring devices installed on it to collect physical parameters such as force F, vibration V and temperature T. The following are the specific implementation steps:

[0114] Receive and normalize data:

[0115] Receive the constructed data flow quality index QoS value and physical parameter data D(t), for example:

[0116] QoS=0.95

[0117] D(t)={(t1,F1),(t2,F2),...,(tn,Fn);(t1,V1),(t2,V2),...,(tn,Vn);(t1,T1),(t2,T2),...,(tn,Tn)}.

[0118] All physical parameter data are converted into a standardized format, and the standardized data are assumed to be StdData(t).

[0119] Calculate the rate of change ΔV:

[0120] For each physical parameter, the time difference Δτ between adjacent data points is calculated, and the rate of change ΔV is calculated. For example, for force F:

[0121] Δτ=τ(n)-τ(n-1)

[0122] ΔF=(F(n)-F(n-1)) / Δτ

[0123] Assume that the following rate of change data is obtained:

[0124] ΔF=[0.2,0.1,-0.3,0.4,...]

[0125] ΔV=[0.15,-0.1,0.2,-0.25,...]

[0126] ΔT=[0.05,0.1,-0.05,0.15,...]

[0127] Determine the key variable set KV:

[0128] Set the threshold Threshold to 0.2, filter out the variables whose absolute value of change rate is greater than the threshold, and form the key variable set KV. For example:

[0129] KV={ΔF,ΔV}#Because some values ​​of |ΔF| and |ΔV| are greater than 0.2

[0130] Calculate the impact score ImpactScore:

[0131] Assign weights to each variable, for example, force F has a weight of 0.6, vibration V has a weight of 0.3, and temperature T has a weight of 0.1. Then calculate the impact score ImpactScore:

[0132] ImpactScore=Σ(|ΔV(i)|*Weight(i))

[0133] =|ΔF(1)|*0.6+|ΔV(1)|*0.3+|ΔT(1)|*0.1

[0134] =0.2*0.6+0.15*0.3+0.05*0.1

[0135] =0.12+0.045+0.005

[0136] =0.17.

[0137] The above calculation is repeated for all data points in the entire punching operation cycle to obtain the final impact score ImpactScore.

[0138] In this way, the changes in physical parameters during the punching process can be continuously monitored and analyzed, the key factors affecting quality can be identified in a timely manner, and the optimization adjustment of actual production operations can be guided by quantitative impact scores. This method not only improves the stability and consistency of production, but also effectively prevents the occurrence of quality problems.

[0139] Step 4: Fine-tune the punching operation based on the analysis results; further including:

[0140] According to the impact score ImpactScore, the punching parameter set P that needs to be adjusted is determined, where each parameter p(j) corresponds to an impact score; the parameters that have a significant impact on the punching quality are identified through the impact score ImpactScore to form the parameter set P. This step ensures the pertinence of the adjustment measures, avoids blind adjustments, and improves the efficiency and accuracy of the adjustments.

[0141] Based on the parameter set P, an adjustment coefficient AdjFactor(j) is set for each parameter p(j), and the coefficient is predetermined based on historical data and experimental results; the adjustment coefficient AdjFactor(j) is optimized based on historical data and experimental results, ensuring the rationality of the adjustment direction and amplitude. This method can minimize the cost of trial and error and quickly achieve the best adjustment effect.

[0142] Using the adjustment factor AdjFactor(j), through the formula:

[0143] NewValue(j)=p(j)+AdjFactor(j)*ImpactScore calculates the new value NewValue(j) of each parameter; by introducing the impact score ImpactScore and the adjustment coefficient AdjFactor(j), the new value NewValue(j) of each parameter can be scientifically calculated. This dynamic adjustment method can flexibly adjust parameters according to actual conditions, improving the adaptability and stability of punching and shearing operations.

[0144] According to the obtained new value NewValue(j), fine-tune the operation and record the changes before and after the adjustment in ChangeLog. After fine-tuning, record the changes before and after the adjustment in detail to facilitate subsequent evaluation and optimization. This method not only helps to provide immediate feedback on the adjustment effect, but also accumulates valuable historical data to provide a basis for further improvement.

[0145] Assume that there is a semiconductor punching equipment. After dynamic data analysis, the following impact score is obtained: ImpactScore = 0.17. The following are the specific implementation steps:

[0146] Determine the punching parameter set P that needs to be adjusted:

[0147] Assume that through analysis it is determined that three key parameters need to be adjusted: force F, vibration V and temperature T. The corresponding original parameter values ​​are:

[0148] p(1)=F=5.88N

[0149] p(2)=V=0.1m / s²

[0150] p(3)=T=25°C

[0151] Set the adjustment factor AdjFactor(j):

[0152] According to historical data and experimental results, set adjustment coefficients for each parameter:

[0153] AdjFactor(1)=0.05 (for force F)

[0154] AdjFactor(2)=0.02(for vibration V)

[0155] AdjFactor(3)=0.01(for temperature T)

[0156] Calculate the new value NewValue(j):

[0157] Use the formula NewValue(j)=p(j)+AdjFactor(j)*ImpactScore to calculate the new value of each parameter. For example:

[0158] NewValue(1)=p(1)+AdjFactor(1)*ImpactScore

[0159] =5.88+0.05*0.17

[0160] =5.88+0.0085

[0161] =5.8885N

[0162] NewValue(2)=p(2)+AdjFactor(2)*ImpactScore

[0163] =0.1+0.02*0.17

[0164] =0.1+0.0034

[0165] =0.1034m / s²

[0166] NewValue(3)=p(3)+AdjFactor(3)*ImpactScore

[0167] =25+0.01*0.17

[0168] =25+0.0017

[0169] =25.0017°C

[0170] Implement minor tweaks and document changes:

[0171] Apply the calculated new value to the actual punching operation and record the changes before and after the adjustment in detail in ChangeLog. For example:

[0172] ChangeLog:

[0173] BeforeAdjustment:

[0174] Force:5.88N,Vibration:0.1m / s²,Temperature:25°C

[0175] AfterAdjustment:

[0176] Force:5.8885N,Vibration:0.1034m / s²,Temperature:25.0017°C

[0177] In this way, the punching parameters can be adjusted scientifically and rationally, ensuring that each adjustment is based on precise data analysis and preset adjustment strategies. This not only improves the stability and consistency of production, but also effectively reduces the occurrence of quality problems such as burrs. At the same time, detailed adjustment records provide valuable reference for subsequent continuous improvement.

[0178] Step 5: Based on the adjustment results, the image of the punched product is captured by the visual system, and edge detection is performed; further comprising:

[0179] Immediately after the punching operation is completed, the visual system is started to capture the image of each product, ensuring that the image covers the entire edge of the product; Immediately start the visual system to capture the image of the product after punching, ensuring the real-time and accuracy of the data. Images covering the entire edge of the product can provide a comprehensive quality assessment basis and help to detect potential problems in a timely manner.

[0180] Based on the acquired image I, the grayscale conversion formula is applied:

[0181] GrayLevel(x,y)=0.299*R(x,y)+0.587*G(x,y)+0.114*B(x,y), converts the color image to a grayscale image, where R, G, and B are the red, green, and blue channel values, respectively, and x and y are pixel coordinates; grayscale conversion reduces the complexity of color information and simplifies subsequent processing steps. Grayscale images can highlight edge features more clearly and improve the accuracy and efficiency of edge detection.

[0182] Using the obtained grayscale image, the edge detection formula is:

[0183] Edge(x,y)=|Gx(x,y)|+|Gy(x,y)|, calculates the edge strength of each pixel, where Gx and Gy represent the gradients in the horizontal and vertical directions respectively; the edge detection formula can accurately identify the edge position in the image and provide high-resolution edge information. This step is crucial for accurately evaluating the edge quality of the product and helps identify any subtle burrs or defects.

[0184] According to the edge strength Edge(x,y), the edge contour of the product is determined and compared with the standard template, and the deviation Deviation is recorded. By comparing with the standard template, the difference between the product edge and the ideal shape is quantified. This comparison method can intuitively display the product quality and facilitate subsequent analysis and formulation of improvement measures.

[0185] Suppose there is a semiconductor punching machine, on which a vision system is installed to capture the image of the product after punching. The following are the specific implementation steps:

[0186] Capturing Product Images I:

[0187] Once the punching operation is complete, the vision system is activated and captures an image covering the entire edge of the product I. The image resolution is, for example, 1024x768 pixels.

[0188] Apply the grayscale conversion formula:

[0189] Use the grayscale conversion formula to convert a color image into a grayscale image. Assume that the color value of a pixel is:

[0190] R(512,384)=255

[0191] G(512,384)=128

[0192] B(512,384)=64

[0193] Calculate the gray value of the pixel:

[0194] GrayLevel(512,384)=0.299*R(512,384)+0.587*G(512,384)+0.114*B(512,384)

[0195] =0.299*255+0.587*128+0.114*64

[0196] =76.245+75.296+7.296

[0197] =158.837≈159 (rounded)

[0198] Calculate the edge strength Edge(x,y):

[0199] Use the Sobel operator or other gradient calculation methods to calculate the gradients Gx and Gy in the horizontal and vertical directions. Assume that the gradient value of a pixel is:

[0200] Gx(512,384)=10

[0201] Gy(512,384)=5

[0202] Calculate the edge strength of the pixel:

[0203] Edge(512,384)=|Gx(512,384)|+|Gy(512,384)|

[0204] =|10|+|5|

[0205] =10+5

[0206] =15

[0207] Determine edge profiles and record deviations:

[0208] According to the edge strength Edge(x,y), the edge contour of the product is extracted and compared with the preset standard template. Assume that the edge strength threshold of the standard template is set to 10, and the actual measured edge point strength exceeds this value, for example:

[0209] Deviation:

[0210] Pixel(512,384):EdgeStrength=15>Threshold=10

[0211] These deviations are recorded in detail to form a DeviationLog for subsequent review and analysis.

[0212] In this way, the edge quality of the product after punching can be accurately captured and analyzed, and any abnormalities can be discovered and recorded in time. This method not only improves the accuracy of product quality control, but also provides a scientific basis for continuous improvement.

[0213] Step 6: Based on the obtained product image information, evaluate the edge smoothness of each product and record abnormalities; further comprising:

[0214] Based on the determined product edge contour, calculate the curvature K(p,q) of each pixel along the edge path using the formula:

[0215] K(p,q)=(Gx(p,q)*Gyy(p,q)-2*Gy(p,q)*Gxy(p,q)+Gx(p,q)*Gxx(p,q)) / (Gx(p,q)^2+Gy(p,q)^2)^(3 / 2), where Gxx, Gyy and Gxy are the second-order partial derivatives of the grayscale image in the p and q directions respectively; by calculating the curvature K(p,q) of each pixel, the geometric characteristics of the edge can be accurately described. Curvature analysis helps to identify the smoothness of the edge and whether there are sudden changes or irregular parts, which is crucial for detecting subtle defects such as burrs.

[0216] Using the curvature K(p,q), we can calculate the average curvature of all pixels on the edge path, KAverage=ΣK(p,q) / N, where N is the total number of pixels on the edge path. Calculating the average curvature KAverage can comprehensively evaluate the smoothness of the entire edge path. This step provides a quantitative indicator, which is convenient for subsequent threshold setting for classification and judgment.

[0217] According to the average curvature KAverage, a smoothness threshold SmoothnessThreshold is set. When KAverage exceeds this threshold, the product is marked as a potential abnormal product. By setting the smoothness threshold SmoothnessThreshold, products that do not meet the quality standards can be automatically screened out. This method improves the detection efficiency and ensures that only products that meet the requirements enter the next stage.

[0218] For the marked potential abnormal products, the DeviationLog records their position coordinates, curvature distribution and deviation from the standard template in detail. The DeviationLog records the information of the abnormal products in detail, including the position coordinates, curvature distribution and deviation, to provide a specific basis for subsequent review and improvement measures. This method not only helps to trace the problem, but also accumulates valuable historical data to support long-term quality improvement.

[0219] Assume that there is a semiconductor punching machine whose vision system has captured and processed the image of the punched product. Here are the specific implementation steps:

[0220] Calculate the curvature K(p,q) of each pixel:

[0221] Assume that the gradient and second-order partial derivative values ​​of a pixel are:

[0222] Gx(512,384)=10

[0223] Gy(512,384)=5

[0224] Gxx(512,384)=0.2

[0225] Gyy(512,384)=0.1

[0226] Gxy(512,384)=0.05

[0227] Use the formula to calculate the curvature of the pixel:

[0228] K(512,384)=(Gx(512,384)*Gyy(512,384)-2*Gy(512,384)*Gxy(512,384 )+Gx(512,384)*Gxx(512,384)) / (Gx(512,384)^2+Gy(512,384)^2)^(3 / 2)

[0229] =(10*0.1-2*5*0.05+10*0.2) / (10^2+5^2)^(3 / 2)

[0230] =(1-0.5+2) / (100+25)^(3 / 2)

[0231] =2.5 / 125^(3 / 2)

[0232] ≈2.5 / 1953.125

[0233] ≈0.00128

[0234] Calculate the average curvature KAverage of all pixels on the edge path:

[0235] Assume that there are 10 pixels on the edge path, and the curvature values ​​are:

[0236] K=[0.00128,0.0013,0.00125,0.00132,0.00127,0.00131,0.00126,0.00133,0.00129,0.0013]

[0237] Calculate the mean curvature KAverage:

[0238] KAverage=ΣK(p,q) / N

[0239] =(0.00128+0.0013+0.00125+0.00132+0.00127+0.00131+0.00126+0.00133+0.00129+0.0013) / 10

[0240] =0.01291 / 10

[0241] =0.001291

[0242] Set the smoothness threshold SmoothnessThreshold:

[0243] Set the smoothness threshold SmoothnessThreshold to 0.0013. If KAverage>SmoothnessThreshold, mark the product as a potential outlier. For example:

[0244] KAverage=0.001291 <SmoothnessThreshold=0.0013

[0245] Recording of potential abnormal items:

[0246] For products marked as potential outliers, their position coordinates, curvature distribution, and deviations from the standard template are recorded in detail. For example, if a product is marked as a potential outlier, the record is as follows:

[0247] DeviationLog:

[0248] ProductID:P12345

[0249] PositionCoordinates:(512,384),(513,385),...

[0250] CurvatureDistribution:[0.00132,0.00131,...]

[0251] DeviationfromStandardTemplate:EdgeStrengthat(512,384)=15>Threshold=10

[0252] In this way, the edge smoothness of each product can be accurately evaluated, and any abnormalities can be discovered and recorded in a timely manner. This method not only improves the accuracy of product quality control, but also provides a scientific basis for continuous improvement, ensuring the stability and consistency of the production process.

[0253] Step 7: For recorded abnormalities, trace the relevant production batches and review all parameter records within the batch for retrospective analysis; further including:

[0254] For products marked as potential abnormal products, extract their production batch ID BatchID and create a batch product list ProductList; by extracting the production batch ID BatchID and creating a product list ProductList, all products related to the abnormal products can be quickly located to ensure the comprehensiveness and accuracy of subsequent analysis. This step provides a clear direction for tracing the root cause of the problem.

[0255] Based on BatchID, access the production database to retrieve all parameter records of the batch, forming a parameter record set ParamsSet, where each record contains a timestamp, punching parameters and environmental conditions; by accessing the production database and retrieving relevant parameter records, detailed historical data can be obtained. These data include not only punching parameters, but also influencing factors such as environmental conditions, which help to comprehensively evaluate various variables in the production process.

[0256] Using ParamsSet, the coefficient of variation CV = σ / μ of each parameter within the batch is calculated, where σ is the standard deviation of the parameter and μ is the mean value; calculating the coefficient of variation CV can quantify the stability of each parameter. The smaller the coefficient of variation, the more stable the parameter is during the production process; otherwise, there may be unstable factors. This method helps identify parameters that need to be focused on.

[0257] According to the obtained coefficient of variation CV, the unstable parameters whose coefficient of variation exceeds the preset threshold CVThreshold are identified, and a review report ReviewReport is generated, which lists the time period and specific values ​​of the parameters in detail. By setting the coefficient of variation threshold CVThreshold to filter out unstable parameters, the possible source of the problem can be quickly identified. A detailed review report ReviewReport is generated to facilitate subsequent in-depth analysis and the formulation of improvement measures.

[0258] Assume there is a semiconductor punching machine that has already flagged some potential anomalies. Here are the specific steps:

[0259] Extract the production batch identifier BatchID and create the product list ProductList:

[0260] For products marked as potential abnormal products, extract their production batch identifier BatchID "B20241224". Create a product list ProductList for this batch, for example:

[0261] ProductList=[P12345,P12346,...,P12360]

[0262] Access the production database to retrieve the parameter record set ParamsSet:

[0263] Based on BatchID "B20241224", access the production database, retrieve the parameter records of all products in the batch, and form the parameter record set ParamsSet. Assume that the acquired data is as follows:

[0264] ParamsSet=[

[0265] {Timestamp:"2024-12-2408:00",Force:5.88N,Vibration:0.1m / s²,Temperature:25°C},

[0266] {Timestamp:"2024-12-2408:01",Force:5.90N,Vibration:0.103m / s²,Temperature:25.1°C}, ... ]

[0268] Calculate the coefficient of variation CV for each parameter:

[0269] For each parameter (force F, vibration V, temperature T), calculate its standard deviation σ and mean μ, and then calculate the coefficient of variation CV. For example, suppose the data of the internal force F of a batch is:

[0270] F=[5.88,5.90,5.89,5.92,...]

[0271] Calculate the mean μ and standard deviation σ:

[0272] μ_F=(5.88+5.90+5.89+5.92+...) / n

[0273] =5.90N

[0274] σ_F=sqrt(Σ(F(i)-μ_F)^2 / n)

[0275] ≈0.015N

[0276] Calculate the coefficient of variation CV:

[0277] CV_F=σ_F / μ_F

[0278] =0.015 / 5.90

[0279] ≈0.00254

[0280] Identify unstable parameters and generate a review report ReviewReport:

[0281] Set the coefficient of variation threshold CVThreshold to 0.005. If the CV of a parameter exceeds this threshold, the parameter is considered unstable. For example, suppose the CV calculation result of vibration V is 0.006, which exceeds the threshold:

[0282] CV_V=0.006>CVThreshold=0.005

[0283] Generate a review report ReviewReport, which lists in detail the time period and specific values ​​of unstable parameters. For example:

[0284] ReviewReport:

[0285] BatchID:B20241224

[0286] UnstableParameter:Vibration

[0287] TimePeriod:2024-12-2408:00to2024-12-2408:10

[0288] SpecificValues:

[0289] -Timestamp:2024-12-2408:00,Vibration:0.1m / s²

[0290] -Timestamp:2024-12-2408:01,Vibration:0.103m / s²

[0291] In this way, all parameter records within the production batch can be systematically traced and reviewed, the specific causes of abnormal product quality can be identified, and a detailed review report can be generated. This method not only improves the efficiency of problem diagnosis, but also provides a scientific basis for continuous improvement to ensure the stability and consistency of production.

[0292] Step 8: Based on the review and analysis, determine the specific factors that cause the burrs and formulate improvement measures accordingly; further including:

[0293] Based on the generated review report ReviewReport, all unstable parameters whose coefficient of variation CV exceeds the preset threshold CVThreshold are identified and classified as potential influencing factors SetFactors; through the detailed data in the review report, those parameters that are unstable during the production process are identified. This step can focus on the key factors that may affect product quality, reduce unnecessary interference information, and improve the accuracy of problem location.

[0294] By using SetFactors combined with the product edge smoothness evaluation results, the correlation factor CorrelationFactor between each unstable parameter and burr generation is calculated through the formula CorrelationFactor=Σ(|KAverage(k)-KTarget|*CV(k)), where KAverage(k) is the average curvature of the kth product, KTarget is the target smoothness, and CV(k) is the coefficient of variation of the kth parameter; by calculating the correlation factor CorrelationFactor, the influence of each unstable parameter on burr generation can be quantified. This method helps to clarify which parameter changes are most likely to cause quality problems, thereby providing a scientific basis for subsequent improvements.

[0295] According to the obtained correlation factor CorrelationFactor, the specific factors CriticalFactors that cause burrs are determined, and the factors that meet the condition CorrelationFactor>CorrThreshold are selected to add to the CriticalFactors set. CorrThreshold is the preset correlation threshold. By setting the correlation threshold CorrThreshold, factors with high correlation are screened out, and the key factors that really affect the generation of burrs are locked. This method ensures the pertinence and effectiveness of improvement measures and avoids waste of resources.

[0296] According to the specific factors identified, the Action Plan for improvement is formulated, which includes adjustment suggestions, expected results and implementation schedule. Based on the specific factors identified, detailed improvement measures are formulated, including specific adjustment suggestions, expected results and implementation schedule. This step ensures the operability and timeliness of the improvement measures, helps to quickly solve problems and continuously optimize the production process.

[0297] Assume that there is a semiconductor punching equipment, a review report ReviewReport has been generated, and some unstable parameters have been identified. The following are the specific implementation steps:

[0298] Identify unstable parameters and classify them as potential influencing factors SetFactors:

[0299] Assume that the review report shows that the coefficient of variation CV of vibration V and temperature T exceeds the preset threshold CVThreshold (0.005). Classify these parameters as potential influencing factors SetFactors:

[0300] SetFactors=[Vibration,Temperature]

[0301] Calculate the correlation factor CorrelationFactor:

[0302] Assume that there are 10 products in a batch, and the average curvature KAverage(k) and target smoothness KTarget of each product are:

[0303] KAverage=[0.001291,0.0013,...,0.00132]

[0304] KTarget=0.0013

[0305] CV=[0.006,0.004]#Corresponding to the coefficient of variation of Vibration and Temperature

[0306] The correlation factor CorrelationFactor for each unstable parameter is calculated using the formula:

[0307] CorrelationFactor_V=Σ(|KAverage(k)-KTarget|*CV_V)

[0308] =|0.001291-0.0013|*0.006+|0.0013-0.0013|*0.006+...+|0.00132-0.0013|*0.006

[0309] ≈0.000006

[0310] CorrelationFactor_T=Σ(|KAverage(k)-KTarget|*CV_T)

[0311] =|0.001291-0.0013|*0.004+|0.0013-0.0013|*0.004+...+|0.00132-0.0013|*0.004

[0312] ≈0.000004

[0313] Determine the specific factors that cause the burr CriticalFactors:

[0314] Set the correlation factor threshold CorrThreshold to 0.000005. According to the calculation results, the correlation factor of vibration V exceeds the threshold:

[0315] CorrelationFactor_V=0.000006>CorrThreshold=0.000005

[0316] Therefore, the vibration V is added to the CriticalFactors collection of specific factors that cause glitches:

[0317] CriticalFactors=[Vibration]

[0318] Develop an Action Plan for improvement:

[0319] According to the specific factors (Vibration V), formulate a detailed improvement measure Action Plan:

[0320] ActionPlan:

[0321] -Adjustment suggestions: Check and calibrate the vibration sensor to ensure it works optimally; adjust the punching speed to reduce vibration.

[0322] -Expected effect: Reduce edge irregularities caused by vibration and improve product edge smoothness.

[0323] -Implementation timeline:

[0324] -Week 1: Complete inspection and calibration of vibration sensors.

[0325] -Week 2: Adjust punching speed and conduct test verification.

[0326] -Week 3: Fully implement improvement measures and monitor results.

[0327] In this way, the specific factors causing the burrs can be systematically determined and effective improvement measures can be developed.

[0328] On the other hand, the present invention provides a semiconductor punching production equipment detection and diagnosis system, such as Figure 2 As shown, including:

[0329] A data acquisition and transmission module is used to configure multiple monitoring devices on the punching equipment to collect physical parameters during the punching process and establish a real-time data stream transmission path based on the data obtained by the monitoring devices;

[0330] Dynamic data analysis and fine-tuning module, which is used to implement dynamic data analysis using the data flow transmission path, focusing on the changes in variables that affect the punching quality, and fine-tuning the punching operation based on the analysis results;

[0331] An image capture and edge smoothness evaluation module is used to combine the adjustment results, capture the image of the product after punching through the visual system, and perform edge detection. Based on the obtained product image information, the edge smoothness of each product is evaluated and abnormal conditions are recorded;

[0332] The module of retrospective analysis and improvement measures formulation is used to trace the relevant production batches for the recorded abnormal situations, and to review all parameter records within the batches for retrospective analysis. Based on the retrospective analysis, the specific factors causing the burrs are determined, and improvement measures are formulated accordingly.

[0333] In addition, the above-mentioned data acquisition and transmission module, dynamic data analysis and fine-tuning module, image capture and edge smoothness evaluation module, and review analysis and improvement measures formulation module are also used to implement the other steps of the above-mentioned semiconductor punching production equipment detection and diagnosis method when executed, which will not be described one by one here.

[0334] In summary, the back side integrates high-precision monitoring devices to achieve real-time acquisition and stable transmission of physical parameters, ensuring data integrity and timeliness; uses dynamic data analysis to focus on the changes in variables that affect punching quality, and makes instant fine-tuning to reduce problems caused by unstable parameters; combines with the visual system to perform high-precision edge detection, accurately evaluates the edge smoothness of each product and automatically records abnormalities; through rapid tracing and detailed review of abnormalities, determines the specific factors that cause burrs, and formulates targeted improvement measures. This method significantly improves the controllability of the punching process and product quality, and enhances production stability and product consistency.

[0335] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting and diagnosing semiconductor punching production equipment, characterized in that: The following steps are involved: Multiple monitoring devices are configured on the punching equipment to collect physical parameters during the punching process; Based on the data obtained by the monitoring device, a real-time data stream transmission path is established, specifically including: Receive raw physical parameter data from the monitoring device, the data is represented in the form of time series, denoted as D(t), where t is the timestamp; Using data D(t), a buffer is constructed, which stores all data points in the most recent time period according to the first-in-first-out principle; Based on the buffer, calculate the time difference of each data point Δτ=τ(n)-τ(n-1), and ensure that Δτ does not exceed the preset maximum allowable value MaxΔτ, where n represents the number of data points recorded in the time interval ΔT, and τ(n) is the time of the data point; According to the obtained time difference Δτ, the quality index QoS of the data flow is evaluated by the formula QoS=Min(1,MaxΔτ / Δτ); Use the data streaming path to perform dynamic data analysis, focusing on the changes in variables that affect punching quality, including: Receive the constructed data flow quality index QoS value and physical parameter data D(t), and convert them into a standardized format; Based on the standardized data, calculate the rate of change of each variable ΔV=(V(n)-V(n-1)) / Δτ, where V represents any physical parameter, including: force F, vibration V or temperature T; Using the obtained change rate ΔV, determine the key variable set KV that affects the punching quality, select variables that meet the condition |ΔV|>Threshold and add them to KV, where Threshold is the preset threshold; According to the determined key variable set KV, the impact score ImpactScore of each punching operation is calculated by the formula ImpactScore=Σ(|ΔV(i)|*Weight(i)), where Weight(i) is the weight assigned to each variable; Based on the results of the analysis, fine-tune the punching operation; Combined with the adjustment results, the image of the punched product is captured by the visual system, and edge detection is performed. Based on the obtained product image information, the edge smoothness of each product is evaluated and abnormal conditions are recorded; For recorded abnormal situations, trace the relevant production batches, review all parameter records within the batches for retrospective analysis, determine the specific factors causing the burrs based on the retrospective analysis, and formulate improvement measures accordingly.

2. The semiconductor punching production equipment detection and diagnosis method according to claim 1, characterized in that: The punching equipment is provided with a plurality of monitoring devices, including: Select multiple locations on the punching equipment to install monitoring devices, each device is responsible for collecting a set of physical parameters; Using the installed monitoring device, the force F, vibration V and temperature T data during the punching process are recorded at fixed time intervals ΔT; Based on the collected data, the average force of each punching operation is calculated as FAvg = (F1 + F2 + ... + Fn) / n, where n represents the number of data points recorded in the time interval ΔT; Based on the obtained average force FAvg and the punching speed S, the energy consumption E of each punching is evaluated using the formula E=FAvg*S.

3. The semiconductor punching production equipment detection and diagnosis method according to claim 2, characterized in that: According to the analysis results, the punching operation is fine-tuned, including: According to the impact score ImpactScore, determine the punching parameter set P that needs to be adjusted, where each parameter p(j) corresponds to an impact score; Based on the parameter set P, an adjustment coefficient AdjFactor(j) is set for each parameter p(j), and the adjustment coefficient AdjFactor(j) is predetermined according to historical data and experimental results; Using the adjustment factor AdjFactor(j), through the formula: NewValue(j)=p(j)+AdjFactor(j)*ImpactScore calculates the new value NewValue(j) of each parameter; According to the obtained new value NewValue(j), fine-tuning operation is performed and the changes before and after the adjustment are recorded.

4. The semiconductor punching production equipment detection and diagnosis method according to claim 3, characterized in that: The combined adjustment result is used to capture the image of the punched product through a visual system and perform edge detection, including: After the punching operation is completed, the vision system is immediately started to capture the image of each product I, ensuring that the image covers the entire edge of the product; Based on the acquired image I, the grayscale conversion formula is applied: GrayLevel(x,y)=0.299*R(x,y)+0.587*G(x,y)+0.114*B(x,y), converts the color image to a grayscale image, where R, G, and B are the red, green, and blue channel values, respectively, and x and y are the pixel coordinates; Using the obtained grayscale image, the edge detection formula is: Edge(x,y)=|Gx(x,y)|+|Gy(x,y)|, calculates the edge strength of each pixel, where Gx and Gy represent the gradients in the horizontal and vertical directions respectively; According to the edge strength Edge(x,y), determine the edge contour of the product, compare it with the standard template, and record the deviation.

5. The semiconductor punching production equipment detection and diagnosis method according to claim 4, characterized in that: Based on the product image information obtained, the edge smoothness of each product is evaluated and abnormal conditions are recorded, including: Based on the determined product edge contour, calculate the curvature K(p,q) of each pixel along the edge path using the formula: K(p,q)=(Gx(p,q)*Gyy(p,q)-2*Gy(p,q)*Gxy(p,q)+Gx(p,q)*Gxx(p,q)) / (Gx(p,q)^2+Gy(p,q)^2)^(3 / 2), where Gxx, Gyy and Gxy are the second-order partial derivatives of the grayscale image in the p and q directions respectively; Using the curvature K(p,q), we can calculate the average curvature of all pixels on the edge path, KAverage=ΣK(p,q) / N, where N is the total number of pixels on the edge path. According to the obtained average curvature KAverage, a smoothness threshold is set. When KAverage exceeds this threshold, the product is marked as a potential abnormal product. For the marked potential abnormal products, their position coordinates, curvature distribution and deviation from the standard template are recorded in detail.

6. The semiconductor punching production equipment detection and diagnosis method according to claim 5, characterized in that: For the abnormal situations recorded, trace the relevant production batches and review all parameter records within the batches for retrospective analysis, including: For products marked as potential abnormal products, extract their production batch identification and create a product list of the batch; Based on the production batch identification, access the production database to retrieve all parameter records of the batch to form a parameter record set, where each record contains a timestamp, punching parameters and environmental conditions; Using the parameter record set, calculate the coefficient of variation CV = σ / μ for each parameter within the batch, where σ is the standard deviation of the parameter and μ is the mean value; Based on the obtained coefficient of variation CV, unstable parameters whose coefficient of variation exceeds the preset threshold are identified, and a review report is generated, listing the time period and specific values ​​of the parameters in detail.

7. The semiconductor punching production equipment detection and diagnosis method according to claim 6, characterized in that: Based on the review and analysis, the specific factors causing the burrs are determined, and improvement measures are formulated accordingly, including: Based on the generated review report, all unstable parameters whose coefficient of variation (CV) exceeds the preset threshold are identified and classified as potential influencing factors; Using the potential influencing factors combined with the product edge smoothness evaluation results, the correlation factor CorrelationFactor between each unstable parameter and burr generation is calculated by the formula CorrelationFactor=Σ(|KAverage(k)-KTarget|*CV(k)), where KAverage(k) is the average curvature of the k-th product, KTarget is the target smoothness, and CV(k) is the coefficient of variation of the k-th parameter; According to the obtained correlation factor CorrelationFactor, determine the specific factors CriticalFactors that cause the burr, select the factors that meet the condition CorrelationFactor>CorrThreshold and add them to the CriticalFactors set, where CorrThreshold is the preset correlation threshold; According to the specific Critical Factors identified, corresponding improvement measures are formulated. The plan includes adjustment suggestions, expected results and implementation timetable.

8. A semiconductor punching production equipment detection and diagnosis system for executing the method according to any one of claims 1 to 7, characterized in that: include: A data acquisition and transmission module is used to configure multiple monitoring devices on the punching equipment to collect physical parameters during the punching process and establish a real-time data stream transmission path based on the data obtained by the monitoring devices; Dynamic data analysis and fine-tuning module, which is used to implement dynamic data analysis using the data flow transmission path, focusing on the changes in variables that affect the punching quality, and fine-tuning the punching operation based on the analysis results; An image capture and edge smoothness evaluation module is used to combine the adjustment results, capture the image of the product after punching through the visual system, and perform edge detection. Based on the obtained product image information, the edge smoothness of each product is evaluated and abnormal conditions are recorded; The module of retrospective analysis and improvement measures formulation is used to trace the relevant production batches for the recorded abnormal situations, and to review all parameter records within the batches for retrospective analysis. Based on the retrospective analysis, the specific factors causing the burrs are determined, and improvement measures are formulated accordingly.

Images