Method for monitoring wafer polishing and electronic device

Abstract

The present application provides a method and an electronic device for monitoring wafer polishing, which relates to the field of wafer polishing detection. The method obtains the pressure observation value of the wafer at the current moment and the pressure observation value at the previous moment during the formal polishing process of the wafer; predicts the first pressure prediction value at the current moment according to the pressure observation value at the previous moment, and updates the first pressure prediction value at the current moment according to the pressure observation value at the current moment to obtain the second pressure prediction value at the current moment; the target pressure prediction value at the current moment is calculated based on the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment; based on the observation value at the current moment and the target pressure prediction value at the current moment, determine the pressure offset at the current moment, and if the pressure offset exceeds the dynamic range, control the polishing equipment to stop running. This method can capture the pressure change situation, give early warnings in time and take measures to improve the stability of the wafer polishing process.

Description

Technical Field

[0001] The present application relates to the field of wafer polishing detection, and particularly to a method for monitoring wafer polishing and an electronic device. Background Art

[0002] In the semiconductor manufacturing process, the polishing process of wafers is one of the key steps to ensure the flatness and quality of the wafer surface. Traditional polishing processes usually rely on the control of fixed parameters, such as the rotation speed of the polishing pad, the flow rate of the polishing liquid, and the applied pressure, etc. However, this fixed-parameter control method is difficult to adapt to the possible dynamic changes during the polishing process, such as the surface characteristics of the wafer, the uniformity of the polishing liquid, and equipment wear, etc. These factors may cause pressure fluctuations during the polishing process, thereby affecting the surface quality and processing accuracy of the wafer.

[0003] In addition, most of the polishing monitoring systems in the related art are based on manual experience or simple sensor data monitoring, lacking the ability of real-time dynamic analysis and prediction of pressure changes. This makes it difficult to detect abnormal situations in a timely manner during the polishing process, thus increasing the risk of wafer damage and reducing production efficiency and the yield rate. Summary of the Invention

[0004] Based on this, it is necessary to provide a method for monitoring wafer polishing and an electronic device for the above technical problems. This method can capture the pressure change situation, give early warnings in a timely manner and take measures to improve the stability of the wafer polishing process.

[0005] In a first aspect, the present application provides a method for monitoring wafer polishing. This method is applied to a wafer polishing system, which includes a pressure sensor. The pressure sensor is installed on the polishing equipment and is used to obtain the pressure value exerted by the polishing equipment on the wafer. The method is executed by a controller, and the method includes:

[0006] Obtain the pressure observation value z of the polishing equipment at the current moment when the wafer is in the formal polishing process k and the pressure observation value z of the previous moment k-1 ;

[0007] According to the pressure observation value z of the previous moment k-1 predict the first pressure prediction value at the current moment, and update the first pressure prediction value at the current moment according to the pressure observation value z of the current moment k to obtain the second pressure prediction value at the current moment;

[0008] Obtain the target pressure prediction value at the current moment. The target pressure prediction value at the current moment is calculated based on the second pressure prediction value at the current moment and the target pressure prediction value of the previous moment;

[0009] Based on the observation value z at the current momentk Based on the predicted value of the target pressure at the current moment, determine the pressure offset at the current moment. If the pressure offset exceeds the dynamic range, control the polishing equipment to stop running; the dynamic range can be dynamically adjusted based on the pressure offset within a unit time period.

[0010] In one embodiment, obtaining the predicted value of the target pressure at the current moment based on the predicted value of the second pressure at the current moment and the predicted value of the target pressure at the previous moment includes:

[0011] Using the moving weighted average algorithm, based on the predicted value of the second pressure at the current moment associated with the first coefficient and the predicted value of the target pressure at the previous moment associated with the second coefficient, obtain the predicted value of the target pressure at the current moment; wherein, the first coefficient is a dynamically adjustable coefficient, and the larger the first coefficient, the greater the influence on the pressure observation value z at the current moment k and the second coefficient is adjusted following the adjustment of the first coefficient.

[0012] In one embodiment, the moving weighted average algorithm is used to obtain the predicted value of the target pressure at the current moment, and the moving weighted average calculation formula is:

[0013] ,

[0014] wherein, EWMA k is the predicted value of the target pressure at the current moment, EWMA k-1 is the predicted value of the target pressure at the previous moment, is the predicted value of the second pressure at the current moment, is the first coefficient, is the second coefficient.

[0015] In one embodiment, dynamically adjusting the first coefficient includes:

[0016] Input the pressure observation value z at the current moment k into the pre-trained Gaussian regression model to obtain the predicted variance , and dynamically update the first coefficient according to the predicted variance ; wherein, the calculation formula for updating the first coefficient is:

[0017] ,

[0018] is the updated first coefficient, is the set pressure, is the first coefficient before update.

[0019] In one embodiment, according to the pressure observation value z at the current moment kUpdate the first pressure prediction value at the current moment to obtain the second pressure prediction value at the current moment, including:

[0020] Obtain the process noise covariance Q and the observation noise covariance R, and determine the error covariance at the current moment according to the observation noise covariance R and the error covariance at the previous moment;

[0021] Substitute the error covariance at the current moment and the observation noise covariance R into the gain equation to obtain the Kalman gain;

[0022] According to the Kalman gain, the pressure observation value z at the current moment k and the first pressure prediction value at the current moment, substitute them into the update equation to obtain the second pressure prediction value at the current moment.

[0023] In one embodiment, the polishing equipment includes upper and lower polishing pads; determining that the wafer is in the formal polishing process includes:

[0024] Obtain the rotation speed range of the upper polishing pad and the rotation speed range of the lower polishing pad, wherein the rotation speed ranges of the upper and lower polishing pads are determined based on the rotation speeds of the upper polishing pad and the lower polishing pad in the historical formal polishing process;

[0025] Obtain the latest n1 pieces of data in the current wafer polishing process. If the rotation speeds of the upper polishing pad in the latest n1 pieces of data are all within the rotation speed range of the upper polishing pad and the rotation speeds of the lower polishing pad are all within the rotation speed range of the lower polishing pad, determine that the wafer is in the formal polishing process.

[0026] In one embodiment, the method further includes:

[0027] Obtain the latest pressure observation value sequence within a preset time period, and perform a sliding window operation on the pressure observation value sequence;

[0028] Determine the first-order difference value diff of each pressure observation value within the window. If the window size is m, the set P of the first-order difference values within the window diff =[diff1, diff2,... diff m-1 ;

[0029] If there are L consecutive first-order difference values within the window that exceed the differential dynamic threshold, control the polishing equipment to stop running; wherein, the differential dynamic threshold is determined according to the mean and standard deviation of multiple first-order difference values collected within the window.

[0030] In one embodiment, dynamically adjusting the dynamic range based on the pressure deviation within a unit time period includes:

[0031] Obtain multiple pressure observation values in the sliding window within a unit time period, and based on each pressure observation value and the corresponding target pressure prediction value within the sliding window, obtain the pressure offsets within the sliding window.

[0032] Calculate the offset mean and offset standard deviation of the sliding window based on the pressure offsets within the sliding window, and adjust the dynamic range based on the offset mean and offset standard deviation of the sliding window.

[0033] In a second aspect, the present application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program. It is characterized in that when the processor executes the computer program, it implements the method for monitoring wafer polishing in the first aspect.

[0034] In a third aspect, the present application also provides a wafer polishing system, which includes the electronic device in the second aspect. In addition, the wafer polishing system further includes: an upper polishing disc for polishing one side of the wafer; a lower polishing disc for polishing the other side of the wafer; a supporting device for fixing the wafer, and the supporting device applies a supporting force to the wafer to make the wafer contact with the upper and lower polishing discs; a polishing liquid pipeline for coating the polishing liquid on the wafer and / or the upper and lower polishing discs; a rotational speed sensor for monitoring the instantaneous rotational speed of the upper and lower polishing discs; a pressure sensor for monitoring the instantaneous pressure applied by the upper and lower polishing discs.

[0035] For the method of the above-mentioned wafer monitoring method, when it is confirmed that the wafer is in the formal polishing process, obtain the pressure observation value z of the polishing equipment at the current moment k and the pressure observation value z of the previous moment k-1 ; Based on z k-1 predict the first pressure prediction value at the current moment, and use z k to update it to obtain the second pressure prediction value; Combine the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment to calculate the target pressure prediction value at the current moment; Determine the pressure offset according to the observation value z k at the current moment and the target pressure prediction value. If the offset exceeds the dynamic range, control the polishing equipment to stop running. Among them, the dynamic range is dynamically adjusted based on the pressure offset data within the unit time period to adapt to the pressure changes in the actual polishing process. This method can effectively capture abnormal pressure changes through real-time monitoring and dynamic adjustment, can identify potential abnormal trends in time before the wafer is polished and broken, give early warnings and take measures, thereby improving the stability of the wafer polishing process. Description of the Drawings

[0036] Figure 1 It is a schematic diagram of a wafer polishing system in an embodiment;

[0037] Figure 2Flowchart of a method for monitoring wafer polishing in an embodiment;

[0038] Figure 3 Flowchart of obtaining the second pressure prediction value at the current moment in an embodiment;

[0039] Figure 4 Flowchart of determining that the wafer is in the formal polishing process in an embodiment;

[0040] Figure 5 Flowchart of monitoring wafer polishing in an embodiment;

[0041] Figure 6 Flowchart of dynamically adjusting the dynamic range based on the pressure offset within a unit time period in an embodiment;

[0042] Figure 7 Structural diagram of an electronic device in an embodiment;

[0043] Figure 8 Schematic diagram of a wafer polishing system in another embodiment. Detailed implementation manners

[0044] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0045] In one embodiment, as Figure 1 shown, a method for monitoring wafer polishing is provided. This method is applied to a wafer polishing system 100. The system includes a pressure sensor 10. The pressure sensor 10 is installed on a polishing device 20 and is used to obtain the pressure value exerted by the polishing device 20 on the wafer. This method is executed by a controller 30. The polishing device 20 includes an upper polishing disc 21 and a lower polishing disc 22. This method is as Figure 2 shown and includes the following steps:

[0046] Step 201: Obtain the pressure observation value z of the polishing device 20 at the current moment when the wafer is in the formal polishing process k and the pressure observation value z at the previous moment k-1 ;

[0047] The formal polishing process can be determined by filtering out the pressure values in the startup stage and the end stage. Among them, in the startup stage, since the polishing device 20 has just started, the exerted pressure value is not yet stable; in the end stage, the pressure value fluctuates due to the completion of polishing or equipment adjustment. Therefore, the formal polishing process refers to the stable stage after filtering out the startup stage and the end stage.

[0048] To determine whether it is in the formal polishing process, it is necessary to set the normal working range of the rotational speeds of the upper polishing disc 21 and the lower polishing disc 22. The normal working range of the rotational speeds of the upper polishing disc 21 and the lower polishing disc 22 can be a normal distribution based on historical formal polishing process data, and the 3σ criterion is used to determine the upper and lower limits of the fluctuations of the rotational speeds of the upper polishing disc 21 and the lower polishing disc 22. Then, the upper and lower limits of the fluctuations of the rotational speeds of the upper polishing disc 21 and the lower polishing disc 22 are respectively and .

[0049] During the real-time monitoring process, the latest n pieces of data of the current processing process are read in real time, including the rotational speed value of the upper polishing disc 21 and the rotational speed value of the lower polishing disc 22 . Then, the real-time data is compared with the preset rotational speed range. If, among these n pieces of data, each value of the rotational speed of the upper polishing disc 21 and the rotational speed of the lower polishing disc 22 is respectively within the range of and , that is, for all i (1 ≤ i ≤ n), there is belonging to , and belonging to , then it can be determined that the current wafer is in the formal polishing state.

[0050] During the formal polishing process, the controller 30 obtains the pressure observation value z at the current moment k , and the pressure observation value z at the current moment k can refer to the pressure value applied by the polishing equipment 20 to the wafer measured by the pressure sensor 10 at a certain specific time point k, that is, the pressure of the upper polishing disc 21 on the lower polishing disc 22. Obtain the pressure observation value z at the previous moment k-1 , that is, the pressure value measured at the time point k - 1.

[0051] It should be noted that the pressure of the upper polishing disc 21 on the lower polishing disc 22 and the pressure of the lower polishing disc 21 on the upper polishing disc 22 physically interact with each other. Usually, only the pressure value of one of the polishing discs needs to be measured to reflect the pressure value applied to the wafer. The pressure sensor 10 can be installed on one side of the upper polishing disc 21 or the lower polishing disc 22 to measure the pressure value applied to the wafer.

[0052] Step 202: Predict the first pressure prediction value at the current moment according to the pressure observation value z at the previous moment k-1 , and update the first pressure prediction value at the current moment according to the pressure observation value z at the current moment k to obtain the second pressure prediction value at the current moment;

[0053] According to the pressure observation value z at the previous momentk-1 , the first pressure prediction value x at the current moment can be predicted through a preset mathematical model or algorithm (such as linear regression, moving average, etc.) k∣k-1 . Further, the pressure observation value z at the current moment is obtained k , and the pressure observation value z at the current moment is used k to correct the first pressure prediction value x k∣k-1 to obtain the second pressure prediction value x at the current moment k . Among them, the correction process can be achieved through simple difference adjustment. For example, using the formula x k = x k∣k-1 + β(z k - x k∣k-1 ), where β can be an adjustment coefficient used to control the amplitude of correction.

[0054] During the polishing process, the error of the pressure observation value mainly comes from the accuracy of the polishing equipment 20, the characteristics of the wafer, and the environmental conditions. The insufficient accuracy of the polishing equipment 20 causes the polishing disc to be uneven, resulting in unstable pressure distribution; the differences in wafer characteristics affect the force distribution; the changes in environmental conditions interfere with the readings of the pressure sensor 10.

[0055] Further, due to the existence of interference factors such as errors, the pressure observation value z at the current moment k may have an obvious deviation from the actual pressure value. Similarly, it will also cause the pressure observation value z at the current moment k to have a relatively obvious deviation from the first pressure prediction value x predicted by the pressure observation value z at the previous moment k . Therefore, it is necessary to correct the first pressure prediction value x k-1 to obtain a more accurate pressure prediction value x at the current moment k∣k-1 . k∣k-1 . k

[0056] It should be noted that the first pressure prediction value x k∣k-1 can be predicted based on the pressure observation value z at the previous moment k-1 , and it cannot fully capture the latest actual pressure changes. By introducing the pressure observation value z at the current moment k , the first pressure prediction value x k∣k-1 can be corrected to make it closer to the actual pressure state. This update process can effectively reduce the prediction error and improve the reliability of the prediction value.

[0057] If the previous moment is the initial moment of formal polishing, since there is not enough pressure data accumulated for prediction calculation. Therefore, the first pressure observation value z1 can be used as the first prediction value x0 at the current moment, that is, x0 = z1.

[0058] Step 203: Obtain the target pressure prediction value at the current moment. The target pressure prediction value at the current moment is calculated based on the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment;

[0059] The second pressure prediction value at the current moment is a pressure prediction value after correction, which can reflect the comprehensive result of the latest observation data and the previous prediction; while the target pressure prediction value at the previous moment can provide reference information on the pressure change trend at the previous moment. By using a preset calculation method, such as weighted average or other fusion strategies, the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment are comprehensively considered to obtain the target pressure prediction value at the current moment.

[0060] It should be noted that the second pressure prediction value mainly reflects the pressure state at the current moment, while the target pressure prediction value takes into account the pressure change trend and historical information by introducing the target pressure prediction value at the previous moment. Comprehensively considering the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment can effectively reduce misjudgment caused by single observation error or short-term fluctuation, and improve the accuracy and reliability of the prediction. In addition, the calculation process of the target pressure prediction value can be regarded as a dynamic adjustment mechanism, which can update the prediction algorithm model in real time according to the changes of historical data and current observation values, so as to better adapt to the dynamic changes in the polishing process. Therefore, calculating the target pressure prediction value at the current moment can improve the stability of the prediction and provide a more reliable basis for the monitoring and control of the polishing process.

[0061] Step 204: Based on the observation value z at the current moment k and the target pressure prediction value at the current moment, determine the pressure offset at the current moment. If the pressure offset exceeds the dynamic range, control the polishing device 20 to stop running; the dynamic range can be dynamically adjusted based on the pressure offset within a unit time period.

[0062] The pressure offset can reflect the difference between the actual observed pressure value and the pressure prediction value. If the pressure offset exceeds the preset dynamic range, an alarm will be triggered and the polishing device 20 will be controlled to stop running to prevent potential abnormal situations from damaging the wafer.

[0063] It should be noted that the dynamic range can be dynamically adjusted according to the pressure offset data within a unit time period. The statistical characteristics (such as mean and standard deviation) of all pressure offsets within this unit time period can be calculated, and the upper and lower limits of the dynamic range can be set based on this. For example, the dynamic range can be set to several times the mean of the pressure offset within this unit time period plus or minus the standard deviation of the pressure offset.

[0064] In this embodiment, when it is confirmed that the wafer is in the formal polishing process, the pressure observation value z of the polishing equipment 20 at the current moment is obtained. k and the pressure observation value z at the previous moment k-1 ; Based on z k-1 predict the first pressure prediction value at the current moment, and use z k to update it to obtain the second pressure prediction value; Combine the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment to calculate the target pressure prediction value at the current moment; Determine the pressure offset according to the observation value z k at the current moment and the target pressure prediction value. If the offset exceeds the dynamic range, control the polishing equipment 20 to stop running. Among them, the dynamic range is dynamically adjusted based on the pressure offset data within a unit time period to adapt to the pressure changes in the actual polishing process. This method can effectively capture abnormal pressure changes through real-time monitoring and dynamic adjustment, can identify potential abnormal trends in time before the wafer is polished and broken, give early warnings and take measures, thereby improving the stability of the wafer polishing process.

[0065] In one embodiment, obtaining the target pressure prediction value at the current moment based on the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment includes:

[0066] Using the moving weighted average algorithm, based on the second pressure prediction value at the current moment associated with the first coefficient and the target pressure prediction value at the previous moment associated with the second coefficient, obtain the target pressure prediction value at the current moment; where the first coefficient is a dynamically adjustable coefficient, and the greater the first coefficient, the greater the influence of the pressure observation value z k at the current moment on the target pressure prediction value at the current moment, and the second coefficient is adjusted following the adjustment of the first coefficient.

[0067] By assigning different weights (i.e., the first coefficient and the second coefficient) to the second pressure prediction value and the target pressure prediction value at the previous moment, combining the latest observation data and the historical pressure change trend, obtain the target pressure prediction value at the current moment.

[0068] Among them, the first coefficient can be a dynamically adjustable parameter, and the larger its value, the greater the influence of the pressure observation value z k at the current moment on the target pressure prediction value. Correspondingly, the second coefficient changes according to the adjustment of the first coefficient to ensure that the sum of the two coefficients is 1.

[0069] In this embodiment, the dynamic adjustment of the first coefficient and the second coefficient can more flexibly balance the latest observation data and historical data, thereby improving the accuracy and reliability of the target pressure prediction value.

[0070] In one embodiment, the moving weighted average algorithm is used to obtain the predicted value of the target pressure at the current moment. The formula for the moving weighted average is:

[0071] ,

[0072] where EWMA k is the predicted value of the target pressure at the current moment, and EWMA k-1 is the predicted value of the target pressure at the previous moment. is the predicted value of the second pressure at the current moment, is the first coefficient, is the second coefficient.

[0073] Specifically, the moving weighted average algorithm (EWMA) is used to obtain the predicted value of the target pressure EWMA k at the current moment, and the calculation formula is .

[0074] where the larger the first coefficient , the greater the weight of the pressure observation value x k at the current moment in calculating the predicted value of the target pressure EWMA k at the current moment, while the second coefficient correspondingly decreases to ensure that the predicted value of the target pressure can flexibly reflect the current observation data and the trend of historical pressure changes.

[0075] It should be noted that the moving weighted average algorithm can smoothly fuse the observation data and historical data at the current moment through weighted averaging, reduce the influence of short-term fluctuations on the prediction result, and maintain the ability to track long-term trends. The moving weighted average algorithm can effectively capture the trend of pressure changes and provide a more stable predicted value for subsequent monitoring and early warning.

[0076] In this embodiment, the method can flexibly balance the latest observation data and historical data by dynamically adjusting the first coefficient , thereby improving the accuracy and reliability of the predicted value of the target pressure and enhancing its adaptability and stability.

[0077] In one embodiment, dynamically adjusting the first coefficient includes:

[0078] Inputting the pressure observation value z k at the current moment into a pre-trained Gaussian regression model to obtain the predicted variance , and dynamically updating the first coefficient according to the predicted variance ; where the calculation formula for updating the first coefficient is:

[0079] ,

[0080] is the updated first coefficient, is the set pressure, is the first coefficient before update.

[0081] During the wafer polishing process, the aging and wear of the equipment are inevitable, and these factors will cause the performance of the polishing equipment 20 to gradually decline, thereby affecting the stability and accuracy of the pressure observation value. Predict the variance of the current data through the Gaussian regression process . Among them, the Gaussian regression model can learn the relationship between input and output based on historical data and output the pressure prediction variance at the current moment .

[0082] Based on the predicted variance , use the formula to dynamically update the first coefficient. When the predicted variance is large, it means that the Gaussian regression model has a high uncertainty about the current observation value, and the updated first coefficient will decrease and rely more on historical data; conversely, when the predicted variance is small, will increase and pay more attention to the current observation value.

[0083] Exemplarily, assume that the set pressure = 100 units, the current first coefficient = 0.5, and the predicted variance = 25 unit² output by the Gaussian regression model, then the updated first coefficient is = 100 + 25100×0.5 ≈ 0.4.

[0084] It should be noted that the Gaussian regression model can obtain the predicted variance in the following way: Assume that there is a historical data set containing past pressure observation values and their corresponding actual pressure values. The Gaussian regression model learns the relationship between these data points and establishes a Gaussian process. In the Gaussian process, each input point z k corresponds to an output mean and variance.

[0085] In this embodiment, this dynamic adjustment mechanism effectively improves the adaptability to pressure changes and the prediction accuracy, and enhances the stability and reliability of the early warning.

[0086] In one embodiment, as Figure 3 shown, update the first pressure prediction value at the current moment according to the pressure observation value z k at the current moment to obtain the second pressure prediction value at the current moment, including the following steps:

[0087] Step 301: Obtain the process noise covariance Q and the observation noise covariance R, and determine the error covariance at the current moment based on the process noise covariance Q and the error covariance at the previous moment;

[0088] Specifically, the process noise covariance Q can describe the uncertainty of the internal state change, while the observation noise covariance R can reflect the error degree in the measurement process. Through the formula P k = P k-1 + Q, add the error covariance P k-1 at the previous moment to the process noise covariance Q to obtain the error covariance P k .

[0089] Step 302: Substitute the error covariance at the current moment and the observation noise covariance R into the gain equation to obtain the Kalman gain;

[0090] The calculation formula for the Kalman gain is . This formula can reflect the degree of trust in the observed data at the current moment. Among them, K k is the Kalman gain, P k is the error covariance at the current moment, and R is the observation noise covariance. The value of the Kalman gain ranges from 0 to 1. When the observation noise covariance R is large, the Kalman gain K k will decrease, indicating a lower degree of trust in the observed data; on the contrary, when the observation noise covariance R is small, the Kalman gain K k will increase, indicating a higher degree of trust in the observed data.

[0091] Exemplarily, assume that the error covariance P k = 4 (unit²) at the current moment and the observation noise covariance R = 1 (unit²), then the Kalman gain is calculated as: K k = 0.8. It means that at the current moment, the degree of trust in the observed data is relatively high (because K k is close to 1), and the observed data will play a greater role in state update. On the contrary, if the observation noise covariance R increases, for example, R = 9, then the Kalman gain will decrease. K k ≈ 0.308. At this time, the degree of trust in the observed data is relatively low, and the influence of the observed data on state update is small, and more reliance will be placed on the previous predicted value.

[0092] Step 303: Substitute the Kalman gain, the pressure observation value z k at the current moment, and the first pressure prediction value at the current moment into the update equation to obtain the second pressure prediction value at the current moment.

[0093] According to the Kalman gain K k , the pressure observation value z kand the first pressure prediction value x at the current moment k∣k-1 , through the update equation x k = x k∣k-1 + K k (z k - x k∣k-1 ). The second pressure prediction value x at the current moment is calculated k .

[0094] Among them, z k - x k∣k-1 represents the deviation between the actual observation and the prediction. The Kalman gain K k can determine the weight of the deviation between the actual observation and the prediction in the update process. If the Kalman gain K k is larger, it means a higher degree of trust in the observed data, and the observed value has a greater impact on the second pressure prediction value; on the contrary, if the Kalman gain K k is smaller, it means a lower degree of trust in the observed data, and the predicted value has a smaller impact on the second pressure prediction value.

[0095] In this embodiment, by obtaining the process noise covariance Q and the observation noise covariance R, and combining the error covariance at the previous moment, the error covariance at the current moment is determined, and then the Kalman gain is calculated. Finally, using the Kalman gain, the pressure observation value z k at the current moment and the first pressure prediction value, the second pressure prediction value at the current moment is obtained through the update equation. This process not only considers the internal dynamic changes and the uncertainty of the observed data, but also dynamically adjusts the weights of the predicted value and the observed value through the Kalman gain, thereby optimizing the accuracy of the pressure prediction.

[0096] In one embodiment, the polishing device 20 includes an upper polishing disc 21 and a lower polishing disc 22. As Figure 4 shown, determining that the wafer is in the formal polishing process includes the following steps:

[0097] Step 401: Obtain the rotation speed range of the upper polishing disc 21 and the rotation speed range of the lower polishing disc 22. Among them, the rotation speed ranges of the upper and lower polishing discs are determined based on the rotation speed of the upper polishing disc 21 and the rotation speed of the lower polishing disc in the historical formal polishing process;

[0098] By statistically analyzing the rotation speed data of the upper polishing disc 21 and the lower polishing disc in the historical formal polishing process, the mean and standard deviation of the rotation speed of the upper polishing disc 21 and the mean and standard deviation of the rotation speed of the lower polishing disc are calculated. Based on the mean and standard deviation of the rotation speed of the upper polishing disc 21 and the mean and standard deviation of the rotation speed of the lower polishing disc, combined with the process requirements and the equipment characteristics, the normal fluctuation range of the rotation speed of the upper polishing disc 21 and the normal fluctuation range of the rotation speed of the lower polishing disc are set, that is, the rotation speed range of the upper polishing disc 21 and the rotation speed range of the lower polishing disc.

[0099] Exemplarily, the method of mean plus or minus three times the standard deviation can be used to determine the rotational speed range of the upper polishing disc 21 and the rotational speed range of the lower polishing disc, ensuring that during normal polishing, the rotational speed values mostly fall within the rotational speed range.

[0100] Step 402: Obtain the latest n1 pieces of data during the current wafer polishing process. If the rotational speeds of the upper polishing disc 21 in the latest n1 pieces of data are all within the rotational speed range of the upper polishing disc 21 and the rotational speeds of the lower polishing disc are all within the rotational speed range of the lower polishing disc, it is determined that the wafer is in the formal polishing process.

[0101] The latest n1 pieces of data can include the rotational speed information of the upper polishing disc 21 and the rotational speed information of the lower polishing disc. Further, check whether the rotational speeds of the upper polishing disc 21 in the n1 pieces of data are all within the preset rotational speed range of the upper polishing disc 21, and at the same time check whether the rotational speeds of the lower polishing disc are all within the preset rotational speed range of the lower polishing disc. If both of these conditions are met, that is, all the checked rotational speeds of the upper polishing disc 21 and the lower polishing disc are within their respective normal ranges, it can be determined that the current wafer is in the formal polishing process.

[0102] In this embodiment, this method uses the statistical characteristics of historical data as a benchmark for the current processing state, ensuring that only when the rotational speed meets the normal process range is it recognized as the formal processing process, thereby improving the stability and reliability of the polishing process and providing an important basis for subsequent quality control and anomaly warning.

[0103] In one embodiment, as Figure 5 shown, the method further includes the following steps:

[0104] Step 501: Obtain the latest sequence of pressure observation values within a preset time period, and perform a sliding window operation on the sequence of pressure observation values;

[0105] First, determine a preset time period, such as the most recent 10 seconds or 30 seconds. Then obtain all the pressure observation values within the preset time period to form the latest sequence of pressure observation values.

[0106] Further, define a sliding window that can slide on the sequence of pressure observation values, with a certain step size (such as 1 second) for each slide, and extract the pressure observation values within the window. In this way, the long sequence of pressure observation values can be divided into multiple shorter subsequences, and each subsequence represents the change of pressure within a specific time window. The sliding window operation helps to perform local analysis on the pressure change, thereby more effectively capturing the dynamic characteristics of the pressure.

[0107] Exemplarily, assume that the preset time period is 10 seconds, the sequence of pressure observation values is [1, 2, 3, 4, 5, 6, 7, 8, 9, 10], the sliding window size is 3, and the step size is 1. Then the sliding window operation will sequentially extract the following subsequences: the first window: [1, 2, 3], the second window: [2, 3, 4], the third window: [3, 4, 5], the fourth window: [4, 5, 6], the fifth window: [5, 6, 7], the sixth window: [6, 7, 8], the seventh window: [7, 8, 9], the eighth window: [8, 9, 10].

[0108] Step 502: Determine the first-order difference values diff of each pressure observation value within the window. If the window size is m, the set of first-order difference values P within the window diff = [diff1, diff2, …… diff m−1 ;

[0109] Assume that the size of the sliding window is m, and the sequence of pressure observation values within the window is [z1, z2, …, z m , then the set of first-order difference values P diff is [diff1, diff2, …, diff m-1 , where each difference value diff i is calculated from the difference between adjacent observations, that is, diff i = z i+1 - z i . Exemplarily, if the sequence of observation values within the window is [z1, z2, z3, z4], then the set of first-order difference values is P diff = [z2 - z1, z3 - z2, z4 - z3].

[0110] Step 503: If there are L consecutive first-order difference values within the window that exceed the differential dynamic threshold, control the polishing device 20 to stop running; wherein, the differential dynamic threshold is determined according to the mean and standard deviation of the multiple first-order difference values collected within the window.

[0111] Calculate the mean μ i and the standard deviation σ diff of all the first-order difference values diff diff within the window, and set an adjustment coefficient based on business requirements. Further, the differential dynamic threshold can be μ diff + σ diff .

[0112] If there are L consecutive first-order difference values diff iIf the differential dynamic threshold is exceeded, it indicates that the pressure change is abnormally drastic and there is a sudden problem in the polishing process. At this time, the polishing device 20 needs to be controlled to stop running to prevent wafer damage or degradation of polishing quality.

[0113] In this embodiment, the method monitors the pressure change trend in real time and uses statistical methods to dynamically adjust the threshold, which can effectively identify abnormal pressure fluctuations and stop equipment operation in time, thereby ensuring the quality and stability of the wafer polishing process.

[0114] In one embodiment, Figure 6 As shown, dynamically adjusting the dynamic interval based on the pressure deviation within a unit time period includes the following steps:

[0115] Step 601: acquiring a plurality of pressure observation values ​​in a sliding window within a unit time period, and obtaining each pressure offset in the sliding window based on each pressure observation value in the sliding window and a corresponding target pressure prediction value;

[0116] A plurality of pressure observation values ​​in a sliding window within a unit time period are obtained, and for each pressure observation value in the sliding window within the unit time period, the difference between the pressure observation value and the corresponding target pressure prediction value is calculated, that is, each pressure offset in the sliding window is obtained.

[0117] For example, it is assumed that the pressure observation sequence in the sliding window within a unit time period is [z1, z2, …, z k ], the corresponding target pressure prediction value sequence is [EWMA1, EWMA2, …, EWMA k ], then each pressure offset is . Furthermore, the pressure offset set within the sliding window [ , ,…, ].

[0118] Step 602: Calculate the offset mean and offset standard deviation of the sliding window based on each pressure offset in the sliding window, and adjust the dynamic range based on the offset mean and offset standard deviation of the sliding window.

[0119] For the pressure offset value set within the sliding window [ , ,…, ] to perform statistical analysis and calculate the mean value μ and standard deviation σ of the offset. The mean value μ can reflect the average level of pressure offset in the window, while the standard deviation σ can indicate the degree of dispersion of these offset values.

[0120] According to the offset mean μ and the offset standard deviation σ, the upper and lower limits of the dynamic range can be dynamically adjusted. For example, according to business requirements, an adjustment coefficient η (such as 3) can be set, and the upper limit of the dynamic range can be set to , and the lower limit is set to .

[0121] In this embodiment, the method can reflect the deviation between the pressure observation value and the target prediction value in real time, capture the dynamic characteristics of the pressure change through statistical analysis, and thus flexibly adjust the dynamic range to adapt to the pressure fluctuation in the actual polishing process.

[0122] In one embodiment, the embodiment of the present application designs a dual warning mechanism. During the wafer polishing process, the dual warning mechanism jointly determines whether to issue a burst disk warning through two links: prediction warning and real-time monitoring warning.

[0123] Exemplarily, once an abnormal change occurs in the real-time monitoring value, even if the pressure deviation between the target pressure prediction value at the current moment and the pressure observation value at the current moment is small, a warning will be immediately issued, directly triggering the shutdown of the polishing device 20 to ensure that the polishing device 20 stops running in case of abnormality. Among them, the real-time monitoring warning has a higher priority and directly responds to the current state of the device. The dual warning mechanism can reduce the error risk when solely relying on the prediction value and ensure timely response in case of sudden changes in the actual environment.

[0124] Based on the same concept, the present application also provides an electronic device, including a memory and a processor, where the memory stores a computer program, and the processor executes the method for monitoring wafer polishing in the computer program.

[0125] In one embodiment, an electronic device is provided, including a memory and a processor. The electronic device can be a terminal, and its internal structure diagram can be as Figure 7As shown in the figure. The electronic device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it realizes a method for monitoring wafer polishing. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the electronic device, or an external keyboard, touchpad, or mouse, etc.

[0126] Based on the same concept, as Figure 8 shown in the figure, the present application further provides a wafer polishing system 800. The system includes the above-mentioned electronic device 810. In addition, the wafer polishing system 800 further includes:

[0127] An upper polishing plate 820 for polishing one side of the wafer; a lower polishing plate 830 for polishing the other side of the wafer; a supporting device 840 for fixing the wafer. The supporting device 840 applies a supporting force to the wafer to make the wafer contact with the upper polishing plate 820 and the lower polishing plate 830; a polishing liquid pipeline 850 for coating the polishing liquid on the wafer and / or the upper polishing plate 820 and the lower polishing plate 830; a rotational speed sensor 860 for monitoring the instantaneous rotational speed of the upper polishing plate 820 and the lower polishing plate 830; a pressure sensor 870 for monitoring the instantaneous pressure applied by the upper polishing plate 820 and the lower polishing plate 830.

[0128] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0129] The above embodiments only represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for monitoring wafer polishing, characterized in that: The method is applied to a wafer polishing system, the system comprising a pressure sensor, the pressure sensor being mounted on a polishing device and being used to obtain a pressure value applied by the polishing device to the wafer, the method being executed by a controller, the method comprising: Get the pressure observation value z of the polishing equipment at the current moment when the wafer is in the formal polishing process k And the pressure observation value z at the previous moment k-1 ; According to the pressure observation value z at the previous moment k-1 Predict the first pressure prediction value at the current moment, and according to the pressure observation value z at the current moment k Updating the first pressure prediction value at the current moment to obtain a second pressure prediction value at the current moment; Obtaining a target pressure prediction value at a current moment, wherein the target pressure prediction value at the current moment is calculated based on the second pressure prediction value at the current moment and the target pressure prediction value at a previous moment; Based on the observed value z at the current moment k and the target pressure prediction value at the current moment, determine the pressure deviation at the current moment, and if the pressure deviation exceeds the dynamic interval, control the polishing device to stop running; the dynamic interval can be dynamically adjusted based on the pressure deviation within a unit time period; The target pressure prediction value at the current moment is obtained based on the second pressure prediction value at the current moment and the target pressure prediction value at the previous moment, including: using a moving weighted average algorithm, based on the second pressure prediction value at the current moment associated with the first coefficient and the target pressure prediction value at the previous moment associated with the second coefficient, the target pressure prediction value at the current moment is obtained; wherein the first coefficient is a dynamically adjustable coefficient, and the larger the first coefficient, the larger the pressure observation value z at the current moment k The greater the impact on the target pressure prediction value at the current moment, the second coefficient is adjusted following the adjustment of the first coefficient.

2. The method for monitoring wafer polishing according to claim 1, characterized in that: The moving weighted average algorithm is used to obtain the target pressure prediction value at the current moment. The moving weighted average calculation formula is: ， Among them, EWMA k is the target pressure prediction value at the current moment, EWMA k-1 is the predicted value of the target pressure at the previous moment, is the second pressure prediction value at the current moment, is the first coefficient, is the second coefficient.

3. The method for monitoring wafer polishing according to claim 1, characterized in that: Dynamically adjusting the first coefficient includes: The pressure observation value z at the current moment k Input the pre-trained Gaussian regression model to get the prediction variance , according to the prediction variance Dynamically update the first coefficient; wherein the calculation formula for updating the first coefficient is: ， is the first coefficient after update, To set the pressure, is the first coefficient before updating.

4. The method for monitoring wafer polishing according to any one of claims 1 to 3, characterized in that: According to the current pressure observation value z k The first pressure prediction value at the current moment is updated to obtain a second pressure prediction value at the current moment, including: Obtaining a process noise covariance Q and an observation noise covariance R, and determining an error covariance at a current moment according to the process noise covariance Q and an error covariance at a previous moment; Substitute the error covariance and the observation noise covariance R at the current moment into the gain equation to obtain the Kalman gain; According to the Kalman gain, the pressure observation value z at the current moment k And the first pressure prediction value at the current moment is substituted into the update equation to obtain the second pressure prediction value at the current moment.

5. The method for monitoring wafer polishing according to claim 1, characterized in that: The polishing device comprises upper and lower polishing plates; Confirm that the wafer is in the formal polishing process, including: Obtaining an upper polishing plate speed interval and a lower polishing plate speed interval, wherein the upper and lower polishing plate speed intervals are determined based on the upper polishing plate speed and the lower polishing plate speed in a historical formal polishing process; The latest n1 data of the current wafer polishing process are obtained. If the upper polishing plate speeds in the latest n1 data are all within the upper polishing plate speed range and the lower polishing plate speeds are all within the lower polishing plate speed range, it is determined that the wafer is in the formal polishing process.

6. The method for monitoring wafer polishing according to claim 1, characterized in that: The method further comprises: Obtaining the latest pressure observation value sequence within a preset time period, and performing a sliding window operation on the pressure observation value sequence; Determine the first-order difference value diff of each pressure observation value in the window. If the window size is m, the first-order difference value set P in the window diff =[diff1, diff2, ... diff m-1 ]; If there are L consecutive first-order difference values ​​in the window that exceed the differential dynamic threshold, the polishing device is controlled to stop running; wherein the differential dynamic threshold is determined based on the mean and standard deviation of multiple first-order difference values ​​collected in the window.

7. The method for monitoring wafer polishing according to claim 1, characterized in that: Dynamically adjusting the dynamic interval based on the pressure deviation within a unit time period includes: Acquire multiple pressure observation values ​​in a sliding window within a unit time period, and obtain each pressure offset in the sliding window based on each pressure observation value in the sliding window and a corresponding target pressure prediction value; The mean offset and the standard offset of the sliding window are calculated based on each pressure offset in the sliding window, and the dynamic range is adjusted based on the mean offset and the standard offset of the sliding window.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for monitoring wafer polishing according to any one of claims 1 to 7 are implemented.

9. A wafer polishing system, characterized in that: The system includes the electronic device of claim 8, and the wafer polishing system further includes: An upper polishing plate, used for polishing one side of the wafer; a lower polishing plate, used for polishing the other side of the wafer; A supporting device, used to fix the wafer, the supporting device applies a supporting force to the wafer so that the wafer contacts the upper and lower polishing plates; A polishing liquid pipeline, used for applying polishing liquid to the wafer and / or the upper and lower polishing plates; Speed ​​sensor, used to monitor the instantaneous speed of the upper and lower polishing discs; Pressure sensor for monitoring the instantaneous pressure applied by the upper and lower polishing plates.

Images