A method for analyzing wafer manufacturing process data based on Ftest

Abstract

The present invention discloses a method for analyzing wafer manufacturing process data based on Ftest, belonging to the technical field of wafer manufacturing process data analysis. This method is implemented based on a wafer manufacturing process data analysis system, including a data acquisition module, a database, a data processing module, a critical value confirmation module, a calculation and analysis module, and an intelligent adjustment module; this method logically groups the detection parameters after the wafer processing is completed, and through Ftest analysis, effectively solves the problem of accurately identifying and analyzing the complex relationships between groups in the massive wafer manufacturing process data, reflecting the integrity and comprehensiveness of wafer manufacturing process data analysis; aiming at the specific requirements of the wafer manufacturing industry, calculates the variances of different detection parameter groups and the F statistic between the variances of different detection parameter groups, can quickly identify the abnormal setting parameters affecting the wafer performance and control the wafer processing machine to automatically correct the abnormal setting parameters, improving the production and processing quality of the wafer.

Description

Technical Field

[0001] The present invention relates to the technical field of wafer process data analysis, and specifically provides a method for analyzing wafer process data based on Ftest inspection. Background Art

[0002] Wafer process data refers to a series of data generated and collected during the semiconductor wafer manufacturing process, which covers information from raw material preparation, wafer processing, testing to packaging; by analyzing the wafer process data, problems existing in the wafer processing production process can be effectively discovered, so as to make rapid adjustments; among them, Ftest, as a classic statistical test method, is crucial for evaluating the consistency and significance of variances between wafer process data; in the prior art, due to relying on single-machine computing resources, with the explosion of data volume, the analysis process of wafer process data cannot be efficiently parallelized; moreover, there are differences in the machines for wafer processing, and the ability to analyze and mine wafer process data is insufficient, which easily causes inaccurate processing parameters and affects the wafer quality. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for analyzing wafer process data based on Ftest inspection to solve the problems proposed in the above background art.

[0004] To solve the above technical problems, the present invention provides the following technical solution: A method for analyzing wafer process data based on Ftest inspection, the method includes the following steps:

[0005] S1. Use a wafer processing machine to process the wafer and collect wafer process data, where the wafer process data includes the setting parameters of the wafer processing machine during the wafer processing process and the detection parameters after the wafer processing is completed;

[0006] S2. Process the collected wafer process data, and logically group the detection parameters according to the wafer processing machine ID;

[0007] S3. Conduct Ftest inspection on the logically grouped detection parameters, calculate the variances of different detection parameter groups, and calculate the F statistic between the variances of different detection parameter groups; compare the calculated F statistic with the critical value of the F distribution to determine whether there is an abnormal detection parameter group; when there is an abnormal detection parameter group, execute step S4;

[0008] S4. Determine the corresponding abnormal setting parameters according to the abnormal detection parameter group, and automatically correct the abnormal setting parameters in the wafer processing machine.

[0009] A wafer manufacturing process data analysis system based on Ftest, the system comprising: a data acquisition module, a database, a data processing module, a critical value confirmation module, a calculation and analysis module, and an intelligent adjustment module;

[0010] The data acquisition module is used to acquire wafer manufacturing process data;

[0011] The database is used to store the acquired wafer manufacturing process data as historical data and continuously update the database;

[0012] The data processing module is used to process the acquired wafer manufacturing process data and logically group the wafer manufacturing process data;

[0013] The critical value confirmation module is used to determine the critical values of the F distribution under different detection parameters;

[0014] The calculation and analysis module is used to perform Ftest on the detection parameters after logical grouping, calculate the variances of different detection parameter groups, and calculate the F statistic between the variances of different detection parameter groups; determine whether there are abnormal detection parameter groups;

[0015] The intelligent adjustment module is used to automatically correct the abnormal set parameters in the wafer processing machine.

[0016] Compared with the prior art, the beneficial effects achieved by the present invention are: logically grouping the detection parameters after wafer processing, and through Ftest analysis, effectively solving the problem of accurately identifying and analyzing the complex relationships between groups in a large amount of wafer manufacturing process data, reflecting the integrity and comprehensiveness of wafer manufacturing process data analysis; aiming at the specific requirements of the wafer manufacturing industry, calculating the variances of different detection parameter groups and the F statistic between the variances of different detection parameter groups, being able to quickly identify the abnormal set parameters affecting wafer performance and controlling the wafer processing machine to automatically correct the abnormal set parameters, improving the production and processing quality of wafers. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a schematic structural diagram of a wafer manufacturing process data analysis system based on Ftest of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0019] Please refer to Figure 1, the present invention provides a technical solution:

[0020] Please refer to Figure 1 , in this embodiment: A method for analyzing wafer manufacturing process data based on Ftest is provided, and this method includes the following steps:

[0021] S1. Use a wafer processing machine to process the wafer and collect wafer manufacturing process data, where the wafer manufacturing process data includes the setting parameters of the wafer processing machine during the wafer processing and the detection parameters after the wafer processing is completed.

[0022] It should be noted that the wafer processing machine is used to process the production process of the wafer, including but not limited to wafer polishing and wafer cutting; the setting parameters and detection parameters are the parameters corresponding to the action type of the wafer processing machine on the wafer.

[0023] In this embodiment, the wafer processing machine is used to perform cutting processing on the wafer. At this time, the setting parameter represents the cutting depth of the wafer processing machine for the wafer, and the detection parameter represents the actual cutting depth of the wafer material; n wafer processing machines are used to process wafers of the same batch, and the initial setting parameters and other condition parameters of each wafer processing machine are the same; during the process of collecting the detection parameter group, the wafer manufacturing process data stored in Parquet format is read from the Hadoop Distributed FileSystem or a similar distributed file system through the API of PySpark; the Parquet file format is particularly suitable for fast reading in big data scenarios due to its columnar storage and efficient compression characteristics.

[0024] S2. Process the collected wafer manufacturing process data and logically group the detection parameters according to the wafer processing machine ID.

[0025] Specifically, the method steps for logically grouping the wafer manufacturing process data are as follows:

[0026] S21. Determine the number n of wafer processing machines and determine the different IDs of the n wafer processing machines;

[0027] S22. Use the wafer processing machine to process wafers of the same batch, and logically group the detection parameters after the wafer processing is completed under the same setting parameters according to the wafer processing machine ID to obtain a logically grouped detection parameter group, denoted as set H; H = {H1, H2,..., H n};

[0028] where, H1, H2,..., H n respectively represent the detection parameter groups after the wafer processing of different wafer processing machines; Hi It represents the set of detection parameters after the wafer processing of the i-th wafer processing machine; i = 1, 2,..., n; They respectively represent different detection parameters in the i-th set of detection parameters; m i It represents the number of detection parameters in the i-th set of detection parameters; the number of detection parameters in the set of detection parameters is the same as the number of wafers in the wafer processing machine.

[0029] It should be noted that the set H is the detection parameters logically grouped under the same set parameters during the wafer processing of the same batch; the number of detection parameters in the set of detection parameters is the same as the number of wafers in the wafer processing machine, so that each wafer has corresponding detection parameters.

[0030] In this embodiment, by using the groupby operation of PySpark, according to the pre-set wafer processing machine ID, the wafer process data is logically grouped to prepare the set of detection parameters of different wafer processing machines for the subsequent Ftest inspection. By logically grouping the wafer process data and analyzing through the Ftest inspection, the problem of accurately identifying and parsing the complex relationships between groups in the massive wafer process data is effectively solved. Thus, according to the actual cutting depth of the wafer material in the detection parameters, the cutting depth of the wafer by the wafer processing machine in the set parameters is adjusted, improving the analysis ability of the wafer process data and reflecting the integrity and comprehensiveness of the wafer process data analysis.

[0031] S3. Conduct Ftest inspection on the logically grouped detection parameters, calculate the variances of different sets of detection parameters, and calculate the F statistic between the variances of different sets of detection parameters; compare the calculated F statistic with the critical value of the F distribution to determine whether there is an abnormal set of detection parameters; when there is an abnormal set of detection parameters, execute step S4.

[0032] It should be noted that the critical value of the F-distribution is a series of values generated by statistical software or mathematical functions, and these values depend on the significance level set by the manager and the degrees of freedom in the detection parameters; the significance level represents the risk that the manager can accept for wrongly rejecting the null hypothesis; based on historical wafer process data, the manager determines the significance level of the Ftest, and under the given significance level and degrees of freedom, it is judged whether the calculated F-statistic is extreme enough through the critical value of the F-distribution, so as to decide whether to reject the null hypothesis; when the calculated F-statistic is greater than the critical value of the F-distribution, the null hypothesis is rejected, and it is considered that there are significant differences in the detection parameter groups participating in the calculation, which are abnormal detection parameter groups; when the calculated F-statistic is less than or equal to the critical value of the F-distribution, the null hypothesis is not rejected, and it is considered that there is not enough evidence to show that there are significant differences in the means of the detection parameter groups, which are normal detection parameter groups; by splitting the calculation process of one Ftest into the calculations of multiple detection parameter groups, the variances of different detection parameter groups and the F-statistics between the variances of different detection parameter groups are obtained, and a highly optimized data processing flow and algorithm are designed, which can quickly identify the abnormal setting parameters affecting the wafer performance and control the wafer processing machine to automatically correct the abnormal setting parameters, improving the production and processing quality of the wafer.

[0033] In this implementation, the multi-threaded technology is used to submit the calculation tasks of multiple detection parameter groups of the Ftest in parallel to the Spark cluster, and multiple independent Ftest jobs can be submitted to the Spark cluster at the same time, so that the computing resources of the entire cluster are more fully utilized, significantly improving the overall processing speed; within each detection parameter group, a customized UDF is applied to perform the Ftest, and this UDF encapsulates the data processing flow and algorithm of the Ftest; when comparing the variances of the detection parameters in different detection parameter groups pairwise, the Python multiprocessing module is used to start child processes to execute in parallel, and multiple child processes jointly share the calculation tasks of these detection parameter groups.

[0034] Specifically, the method steps are as follows:

[0035] S31. According to the different detection parameters in different detection parameter groups, calculate the variances of the detection parameter groups after the wafer processing of different wafer processing machines respectively According to the calculation formula:

[0036]

[0037] Among them, represents the variance of the detection parameters in the i-th detection parameter group; h ij represents the j-th detection parameter in the i-th detection parameter group; represents m in the i-th detection parameter group iThe average value of the detection parameters; j = 1, 2,..., m i ;

[0038] S32. Compare the variances of different groups of detection parameters pairwise, calculate the F statistic between the variances of different groups of detection parameters, according to the calculation formula:

[0039]

[0040] where, F ab represents the F statistic of the variance of the a-th group of detection parameters and the variance of the b-th group of detection parameters; represents the variance of the detection parameters in the a-th group of detection parameters; represents the variance of the detection parameters in the b-th group of detection parameters; a, b = 1, 2,..., n, and a ≠ b;

[0041] S33. According to the significance level set by the management personnel, the numerator degrees of freedom and the denominator degrees of freedom in F ab , determine the critical value K of the corresponding F distribution ab , compare K ab with F ab ; when K ab ≥ F ab , the detection parameter groups of the a-th and b-th wafer processing machines are normal; when K ab < F ab , regard the detection parameter groups of the a-th and b-th wafer processing machines as abnormal detection parameter groups.

[0042] In this implementation, the numerator degrees of freedom df a = m a - 1; the denominator degrees of freedom df b = m b - 1; where, m a represents the number of detection parameters in the a-th group of detection parameters; m b represents the number of detection parameters in the b-th group of detection parameters; through the determined significance level, the numerator degrees of freedom df a and the denominator degrees of freedom df b , determine the critical value K of the corresponding F distribution ab , so as to judge whether there is a significant difference in the variances of the detection parameter groups of the a-th and b-th wafer processing machines.

[0043] Furthermore, store the collected wafer process data as historical data in the database and continuously update the database; based on the historical wafer process data stored in the database, the management personnel set the significance level of the detection parameters and determine the critical values of the F distribution under different detection parameters according to the degrees of freedom in the detection parameters.

[0044] It should be noted that the method for determining the critical value of the F-distribution under different detection parameters is as follows: according to the degrees of freedom of the numerator df a , the degrees of freedom of the denominator df b , taking the F-statistic as the independent variable, the probability density function of the F-distribution is obtained According to the determined significance level, determine the critical value of the F-distribution; among them, the critical value of the F-distribution makes the right area of the curve formed by the probability density function equal to the determined significance level.

[0045] S4. Determine the corresponding abnormal setting parameters according to the abnormal detection parameter group, and automatically correct the abnormal setting parameters in the wafer processing machine.

[0046] Specifically, the method steps are as follows:

[0047] S41. Analyze the abnormal detection parameter group, determine the detection parameter group corresponding to the abnormal detection parameter group that is abnormal after pairwise comparison with other detection parameter groups in steps S32 - S33, and use this detection parameter group as the adjustment analysis data. Determine the wafer processing machine corresponding to the adjustment analysis data, and use the setting parameters of this wafer processing machine as the abnormal setting parameters;

[0048] S42. According to the set H of the detection parameter groups after logical grouping, determine the total average value of the detection parameters where h ij represents the j-th detection parameter in the i-th detection parameter group;

[0049] S43. Compare the detection parameters in the adjustment analysis data with to determine the set X1 of detection parameters greater than and the set X2 of detection parameters less than . Calculate the dispersion degrees L1 and L2 of the detection parameters in the detection parameter sets X1 and X2 from respectively according to the square difference, and adjust the abnormal setting parameters in the wafer processing machine according to the difference degree between L1 and L2.

[0050] In this embodiment, determine the detection parameter group corresponding to the abnormal detection parameter group that is abnormal after comparison with other detection parameter groups, and use this detection parameter group as the adjustment analysis data; for example, when analyzing the abnormal detection parameter group, it is found that the detection parameter group in the second wafer processing machine is abnormal after pairwise comparison with the detection parameter groups of other wafer processing machines. Therefore, it is judged that the setting parameters in the second wafer processing machine are inaccurate. Use the detection parameter group in the second wafer processing machine as the adjustment analysis data, and use the setting parameters in the second wafer processing machine as the abnormal setting parameters; compare the detection parameters in the second detection parameter group with Compare and screen out the set of detection parameters X1 greater than and the set of detection parameters X2 less than . Sum the squared differences of the detection parameters in set X1 with to obtain the degree of dispersion L1; sum the squared differences of the detection parameters in set X2 with to obtain the degree of dispersion L2; since in this implementation, the set parameter represents the cutting depth of the wafer processing machine for the wafer, and the detection parameter represents the actual cutting depth of the wafer material, the differences in wafer processing machines cause the actual cutting depth of the wafer material to not reach the ideal conditions. By calculating the difference between L1 and L2, when L1 - L2 is greater than 0, at this time the actual cutting depth of the wafer processing machine for the wafer material is too high, and the set parameter in the second wafer processing machine is corrected to reduce the cutting depth of the wafer. Moreover, the greater L1 - L2 is, the more the cutting depth is reduced; when L1 - L2 is less than 0, at this time the actual cutting depth of the wafer processing machine for the wafer material is too low, and the set parameter in the second wafer processing machine is corrected to increase the cutting depth of the wafer processing machine for the wafer. Moreover, the smaller L1 - L2 is, the more the cutting depth is increased; by automatically adjusting the set parameters of the wafer processing machine, the production quality of the wafer is thus improved.

[0051] It should be noted that the steps S1 - S4 are the process of automatically correcting the set parameters in the wafer processing machine after processing a batch of wafers. When processing the next batch of wafers, repeat steps S1 - S4 until the wafer processing machine is shut down.

[0052] A wafer manufacturing process data analysis system based on Ftest inspection, the system includes: a data acquisition module, a database, a data processing module, a critical value confirmation module, a calculation and analysis module, and an intelligent adjustment module;

[0053] The data acquisition module is used to acquire wafer manufacturing process data;

[0054] The database is used to store the acquired wafer manufacturing process data as historical data and continuously update the database;

[0055] The data processing module is used to process the acquired wafer manufacturing process data and logically group the wafer manufacturing process data;

[0056] The critical value confirmation module is used to determine the critical values of the F distribution under different detection parameters;

[0057] The calculation and analysis module is used to perform F - test on the detected parameters after logical grouping, calculate the variances of different groups of detected parameters, and calculate the F - statistic between the variances of different groups of detected parameters; and determine whether there is an abnormal group of detected parameters.

[0058] The intelligent adjustment module is used to automatically correct the abnormal set parameters in the wafer processing machine.

[0059] Furthermore, the intelligent adjustment module includes a human - machine interaction platform, which is used to digitally display the collected wafer process data, the variances of different groups of detected parameters calculated, and the F - statistic between the variances of different groups of detected parameters; among which, the management personnel can manually adjust the set parameters of the wafer processing machine through the human - machine interaction platform.

[0060] In this embodiment:

[0061] The wafer processing machine is used to process wafers. The data acquisition module acquires the wafer process data and sends the acquired wafer process data to the database and the data processing module.

[0062] The data processing module processes the acquired wafer process data, logically groups the detected parameters according to the wafer processing machine ID, and sends the logically grouped detected parameters to the calculation and analysis module.

[0063] The critical value confirmation module determines the critical values of the F - distribution under different detected parameters based on the historical wafer process data stored in the database, according to the significance level set by the management personnel and the degrees of freedom in the detected parameters.

[0064] The calculation and analysis module performs F - test on the logically grouped detected parameters, calculates the variances of different groups of detected parameters, and calculates the F - statistic between the variances of different groups of detected parameters; compares the calculated F - statistic with the critical values of the F - distribution determined in the critical value confirmation module to determine whether there is an abnormal group of detected parameters; when there is an abnormal group of detected parameters, sends the abnormal group of detected parameters to the intelligent adjustment module; if the group of detected parameters is normal, no operation is performed.

[0065] The intelligent adjustment module determines the corresponding abnormal set parameters according to the abnormal group of detected parameters and automatically corrects the abnormal set parameters in the wafer processing machine.

[0066] Among which, when the management personnel manually adjust the set parameters of the wafer processing machine through the human - machine interaction platform, the intelligent adjustment module keeps the set parameters in the wafer processing machine fixed.

[0067] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A wafer manufacturing process data analysis method based on Ftest, characterized in that The method includes the following steps: S1. Use a wafer processing machine to process the wafer and collect wafer process data, where the wafer process data includes the setting parameters of the wafer processing machine during the wafer processing and the detection parameters after the wafer processing is completed; S2. Process the collected wafer process data and logically group the detection parameters according to the wafer processing machine ID; S3. Conduct an Ftest on the logically grouped detection parameters, calculate the variances of different detection parameter groups, and calculate the F statistic between the variances of different detection parameter groups; compare the calculated F statistic with the critical value of the F distribution to determine whether there is an abnormal detection parameter group; when there is an abnormal detection parameter group, execute step S4; S4. Determine the corresponding abnormal setting parameters according to the abnormal detection parameter group and automatically correct the abnormal setting parameters in the wafer processing machine; The method steps of step S4 are as follows: S41. Analyze the abnormal detection parameter group, determine the detection parameter group corresponding to the detection parameter group that is an abnormal detection parameter group after pairwise comparison with other detection parameter groups, and use this detection parameter group as the adjustment analysis data. Determine the wafer processing machine corresponding to the adjustment analysis data and use the setting parameters of this wafer processing machine as the abnormal setting parameters; S42. Determine the total average value of the detection parameters according to the set H of the detection parameter groups grouped by logic S43. Compare the detection parameters in the adjusted analysis data with to determine the set X1 of detection parameters greater than and the set X2 of detection parameters less than . Calculate the dispersion degrees L1 and L2 of the detection parameters in the detection parameter sets X1 and X2 respectively from , and adjust the abnormal setting parameters in the wafer processing machine according to the difference degree between L1 and L2.

2. The method for wafer manufacturing process data analysis based on Ftest as claimed in claim 1, wherein, The method steps for logically grouping the wafer process data are as follows: S21. Determine the number n of wafer processing machines and determine the different IDs of the n wafer processing machines; S22. Use a wafer processing machine to process wafers of the same batch. According to the ID of the wafer processing machine, logically group the detection parameters after the processing of the wafers under the same set parameters to obtain a set of detection parameters after logical grouping, denoted as set H; H = {H1, H2,..., H n}; Among them, H1, H2, ..., H n respectively represent the detection parameter groups after the wafer processing of different wafer processing machines; H i represents the detection parameter group after the wafer processing of the i-th wafer processing machine; i = 1, 2, ..., n; respectively represent different detection parameters in the i-th detection parameter group; m i represents the number of detection parameters in the i-th detection parameter group; the number of detection parameters in the detection parameter group is the same as the number of wafers in the wafer processing machine.

3. A method for wafer process data analysis based on Ftest, as claimed in claim 2, wherein The method steps of step S3 are as follows: S31. Calculate the variances of the detection parameter groups after the wafer processing of different wafer processing machines respectively according to different detection parameters in different detection parameter groups According to the calculation formula: Among them, represents the variance of the detection parameters in the i-th detection parameter group; h ij represents the j-th detection parameter in the i-th detection parameter group; represents the average value of m i detection parameters in the i-th detection parameter group; j = 1, 2,..., m i ; S32. Compare the variances of different detection parameter groups pairwise, calculate the F statistic between the variances of different detection parameter groups, according to the calculation formula: Among them, F ab represents the F statistic of the variance of the a-th detection parameter group and the variance of the b-th detection parameter group; represents the variance of the detection parameters in the a-th detection parameter group; represents the variance of the detection parameters in the b-th detection parameter group; a, b = 1, 2,..., n, and a ≠ b; S33. Determine the critical value K of the corresponding F distribution according to the significance level set by the management personnel, the numerator degrees of freedom and the denominator degrees of freedom in F ab , and compare K ab with F ab ; when K ab ≥F ab , the detection parameter groups of the a-th and b-th wafer processing machines are normal; when K ab <F ab <F ab , the detection parameter groups of the a-th and b-th wafer processing machines are used as abnormal detection parameter groups.

4. A method for wafer process data analysis based on Ftest, as claimed in claim 3, wherein, Store the collected wafer process data as historical data in the database and continuously update the database; based on the historical wafer process data stored in the database, the management personnel set the significance level of the detection parameters and determine the critical values of the F distribution under different detection parameters according to the degrees of freedom in the detection parameters.

5. A wafer manufacturing process data analysis system based on Ftest, applying a wafer manufacturing process data analysis method based on Ftest as described in any one of claims 1-4, characterized in that, The system includes: a data acquisition module, a database, a data processing module, a critical value confirmation module, a calculation and analysis module, and an intelligent adjustment module; The data acquisition module is used to collect wafer process data; The database is used to store the collected wafer process data as historical data and continuously update the database; The data processing module is used to process the collected wafer process data and logically group the wafer process data; The critical value confirmation module is used to determine the critical values of the F distribution under different detection parameters; The calculation and analysis module is used to conduct an Ftest on the logically grouped detection parameters, calculate the variances of different detection parameter groups, and calculate the F statistic between the variances of different detection parameter groups; determine whether there is an abnormal detection parameter group; The intelligent adjustment module is used to automatically correct the abnormal setting parameters in the wafer processing machine.

6. The wafer manufacturing process data analysis system based on the Ftest inspection according to claim 5, characterized in that, The intelligent adjustment module includes a human-computer interaction platform for digitally displaying the variances of different sets of detected parameters obtained by collecting wafer manufacturing process data and calculations and the F statistic between the variances of different sets of detected parameters; among them, management personnel can manually adjust the setting parameters of the wafer processing machine tool through the human-computer interaction platform.

Images