Method and apparatus for obtaining normal control range of control chart, and computer-readable medium

Abstract

The method for obtaining the normal control range of the control diagram provided by the present invention comprises the following steps: receiving original parameter data; obtaining an index diagram; obtaining a statistical distribution curve diagram of the evaluation value in the index diagram; respectively obtaining the left center evaluation value c of the leftmost peak in the statistical distribution curve diagram l , the right center comment estimate c of the rightmost peak r , and the intermediate center evaluation value c of the intermediate peak c ; Use the left center comment valuation c l is the mean, calculated to be less than or equal to the left center comment valuation c l The left standard deviation s of all evaluation values l and use the right center comment valuation c r is the mean, calculate the value c of the review that is greater than or equal to the center of the right side r The right-hand standard deviation s of all evaluation values r ; Get the normal control range of the control chart corresponding to the indicator chart [c l ―αs l ―ρ,c r +αs r +ρ], where 3.5≤α≤4, ρ is a minimum value, which is used to adjust the error of the normal control range. The present invention improves the accuracy and reliability of the normal control range setting of the control diagram.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor manufacturing, and in particular, to a method and device for obtaining the normal control range of a control chart, and a computer-readable medium. Background Art

[0002] In the manufacturing process of semiconductor devices, a production machine is usually used to process wafers. In order to monitor the processing process of the wafers, one or more sensors are provided in the production machine. The sensors are used to collect various raw parameter data (raw trace data) during the production process of the production machine. By cooperating the sensors with software tools, a change curve of various raw parameter data over time can be collected. According to a plurality of preset main step ranges (Key Steps’ Window, KSW), statistical methods can be used to calculate evaluation values such as the average value, standard deviation, maximum value, and minimum value of various raw description data within each main step range, that is, calculate various evaluation values of each wafer within the specified step range. According to the time sequence of the production machine processing multiple wafers, a time sequence signal graph composed of single evaluation values of multiple wafers can be obtained, that is, an indicator chart (Indicator Chart). By setting a normal control range in the indicator chart, a control chart is formed. FDC (Fault Detection & Classification) analysis can be performed according to the control chart, so as to monitor the health status of the production machine.

[0003] However, currently, the normal control range of the control chart is mainly obtained by relying on empirical rules. This method requires a high labor cost, and the accuracy of the normal control range obtained by relying on empirical rules is relatively low.

[0004] Therefore, how to improve the accuracy and reliability of obtaining the normal control range of the control chart and reduce the labor cost at the same time is a technical problem that needs to be solved urgently at present. Summary of the Invention

[0005] The present invention provides a method and device for obtaining the normal control range of a control chart, and a computer-readable medium, which are used to solve the problems that the accuracy and reliability of obtaining the normal control range of the control chart in the prior art are relatively low, and at the same time reduce the labor cost of obtaining the normal control range of the control chart.

[0006] To solve the above problems, the present invention provides a method for obtaining the normal control range of a control chart, including the following steps:

[0007] Receiving raw parameter data during the semiconductor manufacturing process of the production machine for processing wafers;

[0008] Obtain an index graph based on the original parameter data, where the index graph is a curve showing the change of a single evaluation value of the production machine over time;

[0009] Obtain the statistical distribution curve graph of the evaluation value in the index graph;

[0010] Respectively obtain the left central point evaluation value c of the leftmost peak in the statistical distribution curve graph l , the right central point evaluation value c of the rightmost peak r , and the middle central point evaluation value c of the middle peak c ;

[0011] Using the left central point evaluation value c l as the mean, calculate the left standard deviation s of all the evaluation values less than or equal to the left central point evaluation value c l , and using the right central point evaluation value c l as the mean, calculate the right standard deviation s of all the evaluation values greater than or equal to the right central point evaluation value c r ; r ; r ;

[0012] Obtain the normal control range [c l -αs l -ρ, c r +αs r +ρ] of the control chart corresponding to the index graph, where 3.5 ≤ α ≤ 4 and ρ is a minimum value used to adjust the error of the normal control range.

[0013] Optionally, the specific steps for obtaining the statistical distribution curve graph of the evaluation value in the index graph include:

[0014] Obtain the initial statistical distribution curve graph based on the data of all the evaluation values in the index graph;

[0015] Denoise the initial statistical distribution curve graph to obtain the statistical distribution curve graph.

[0016] Optionally, the specific steps for denoising the initial statistical distribution curve graph include:

[0017] Remove the data within a preset proportional range at one or both ends of the initial statistical distribution curve graph.

[0018] Optionally, the specific steps for obtaining the statistical distribution curve graph of the evaluation value in the index graph include:

[0019] Judge whether there is a multiple distribution in the index graph. If so, perform differential processing on the index graph;

[0020] Obtain the statistical distribution curve graph of the index graph after differential processing.

[0021] Optionally, the specific steps for performing differential processing on the index graph include:

[0022] Determine whether there is a position in the index graph where the amplitude difference is greater than a first preset value. If so, perform first-order differential processing on the index graph.

[0023] Optionally, the specific steps for performing differential processing on the index graph include:

[0024] Determine whether there is a position in the index graph where the amplitude surge is greater than a second preset value. If so, perform second-order differential processing on the index graph.

[0025] Optionally, the specific steps for performing differential processing on the index graph include:

[0026] Determine whether there is a position in the index graph where the amplitude turn is greater than a third preset value. If so, perform third-order differential processing on the index graph.

[0027] Optionally, respectively obtain the left center point evaluation value c l of the leftmost peak, the right center point evaluation value c r of the rightmost peak, and the middle center point evaluation value c c of the middle peak. The specific steps include:

[0028] Determine whether the statistical distribution curve graph is a skewed peak distribution curve. If so, obtain the peak center point evaluation value of the statistical distribution curve graph, and use the peak center point evaluation value as the left center point evaluation value c l , the right center point evaluation value c r , and the middle center point evaluation value c c .

[0029] Optionally, respectively obtain the left center point evaluation value c l of the leftmost peak, the right center point evaluation value c r of the rightmost peak, and the middle center point evaluation value c c of the middle peak. The specific steps include:

[0030] Use the K-mean clustering method to respectively obtain the left center point evaluation value c l of the leftmost peak, the right center point evaluation value c r of the rightmost peak, and the middle center point evaluation value c c .

[0031] To solve the above problems, the present invention also provides a device for obtaining the normal control range of a control chart, including a processor, and the processor includes:

[0032] A receiving circuit for receiving the original parameter data during the semiconductor process treatment of a wafer by a production machine tool;

[0033] A storage circuit for storing an index chart obtained according to the original parameter data, where the index chart is a change curve of a single evaluation value of the production machine tool over time;

[0034] A first calculation circuit for obtaining a statistical distribution curve chart of the evaluation value in the index chart;

[0035] A second calculation circuit for respectively obtaining the left center point evaluation value c of the leftmost peak in the statistical distribution curve chart l , the right center point evaluation value c of the rightmost peak r , and the middle center point evaluation value c of the middle peak c ;

[0036] A third calculation circuit for using the left center point evaluation value c l as the mean value and calculating the left standard deviation s of all the evaluation values less than or equal to the left center point evaluation value c l , and using the right center point evaluation value c l as the mean value and calculating the right standard deviation s of all the evaluation values greater than or equal to the right center point evaluation value c r ; r ; r ;

[0037] An acquisition circuit for obtaining the normal control range [c l -αs l -ρ, c r +αs r +ρ] of the control chart corresponding to the index chart, where 3.5 ≤ α ≤ 4, ρ is a minimum value for adjusting the error of the normal control range.

[0038] Optionally, the first calculation circuit is used to obtain an initial statistical distribution curve chart according to all the evaluation value data in the index chart, and perform noise reduction processing on the initial statistical distribution curve chart to obtain the statistical distribution curve chart.

[0039] Optionally, the first calculation circuit is used to remove the data within a preset proportion range at one or both ends of the initial statistical distribution curve chart to implement noise reduction processing on the initial statistical distribution curve chart.

[0040] Optionally, the first calculation circuit is further configured to determine whether there is a multiple distribution in the index graph. If so, perform differential processing on the index graph, and obtain a statistical distribution curve graph of the index graph after the differential processing.

[0041] Optionally, the first calculation circuit is configured to determine whether there is a position in the index graph where the amplitude difference is greater than a first preset value. If so, perform first-order differential processing on the index graph.

[0042] Optionally, the first calculation circuit is configured to determine whether there is a position in the index graph where the amplitude surge is greater than a second preset value. If so, perform second-order differential processing on the index graph.

[0043] Optionally, the first calculation circuit is configured to determine whether there is a position in the index graph where the amplitude turning point is greater than a third preset value. If so, perform third-order differential processing on the index graph.

[0044] Optionally, the second calculation circuit is configured to determine whether the statistical distribution curve graph is a skewed peak distribution curve. If so, obtain an evaluation value of the peak center point in the statistical distribution curve graph, and use the evaluation value of the peak center point as both the left center point evaluation value c l 、the right center point evaluation value c r 、and the middle center point evaluation value c c .

[0045] Optionally, the second calculation circuit is configured to respectively obtain the left center point evaluation value c of the leftmost peak in the statistical distribution curve graph by using the K-mean clustering method l 、the right center point evaluation value c of the rightmost peak r 、and the middle center point evaluation value c of the middle peak c .

[0046] To solve the above problems, the present invention further provides a computer-readable medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in any one of the above is implemented.

[0047] The method and device for obtaining the normal control range of a control chart, and a computer-readable medium provided by the present invention automatically obtain the normal control method of the control chart through an objective calculation method, avoiding the problems of high labor cost and large subjectivity caused by manually setting the normal control range in the prior art. The present invention calculates the normal control range by converting a statistical distribution curve of a non-normal distribution into multiple normal distribution curves. On the one hand, it avoids setting the normal control range too loosely, reducing the false negative rate of FDC, enabling engineers to be aware of abnormal conditions of the machine in a timely manner; on the other hand, it avoids setting the normal control range too tightly, reducing the false positive rate of FDC, thereby reducing the manual inspection cost and also reducing the missed inspection problem caused by engineers omitting manual inspection due to a high false positive rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Attached Figure 1A is a flowchart of the method for obtaining the normal control range of a control chart in the specific embodiment of the present invention;

[0049] Attached Figure 1B is a schematic structural diagram of a production machine in the specific embodiment of the present invention;

[0050] Attached Figure 2 is a schematic diagram of converting a skewed distribution into two normal distributions in the specific embodiment of the present invention;

[0051] Attached Figure 3 is a schematic diagram of converting a multimodal distribution into two normal distributions in the specific embodiment of the present invention;

[0052] Attached Figures 4A - 4B is a schematic diagram of an index chart with a multiple distribution in the specific embodiment of the present invention;

[0053] Attached Figures 5A - 5C is a schematic diagram of differentiating an index chart with different characteristics in the specific embodiment of the present invention;

[0054] Attached Figure 6 is a corresponding relationship diagram between different differentiated index charts and statistical distribution curves in the specific embodiment of the present invention;

[0055] Attached Figure 7 is a structural block diagram of the device for obtaining the normal control range of a control chart in the specific embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The following will describe in detail the specific embodiments of the method and device for obtaining the normal control range of a control chart, and a computer-readable medium provided by the present invention with reference to the accompanying drawings.

[0057] The current method for obtaining the normal control range in a control chart is based on empirical rules, that is, the normal control range of each control chart is limited to [average - α × standard deviation, average + α × standard deviation], where α is a constant determined according to manual experience; or, the normal control range of each control chart is limited to [average - average × k%, average + average × k%], where k is a constant determined according to manual experience. However, the current method of setting the normal control range based on empirical rules is only suitable for the case where the statistical distribution curve corresponding to the index chart is a normal distribution. When the statistical distribution curve is a skewed peak distribution, due to the inability to accurately obtain the interval definition of normal values and deviation values, the normal control range limited by the empirical rule will result in too tight a limit on one side of the control chart and a high false alarm rate, while too loose a limit on the other side and a high miss rate. When the statistical distribution curve is a multi-peak distribution, due to the inability to accurately obtain the interval definition of normal values and deviation values, the normal control range limited by the empirical rule will result in too loose limits on both sides of the control chart and a high miss rate. When the index chart is a multiple distribution, the empirical rule cannot be directly used to limit the normal control range of the control chart.

[0058] To improve the accuracy and reliability of the normal control range limit in the control chart, this specific embodiment provides a method for obtaining the normal control range of the control chart, as follows Figure 1A is a flowchart of the method for obtaining the normal control range of the control chart in the specific embodiment of the present invention. As Figure 1A shown, the method for obtaining the normal control range of the control chart includes the following steps:

[0059] Step S11, receiving the original parameter data during the semiconductor process treatment of the wafer by the production machine.

[0060] Step S12, obtaining an index chart according to the original parameter data, where the index chart is a curve of the single evaluation value of the production machine changing with time.

[0061] Specifically, the abscissa of the index chart is the wafer numbers arranged in chronological order, and the ordinate is the evaluation value. The evaluation value in the index chart in this specific embodiment can be an average value, a standard deviation, a maximum value, or a minimum value, etc. The production machine can be a machine for performing any process treatment on the wafer, such as a lithography machine, a film deposition machine, etc. As follows Figure 1B is a structural schematic diagram of a production machine in the specific embodiment of the present invention, Figure 1B in which the solid arrow represents the transfer path for the wafer to enter the production machine for semiconductor process treatment, and the dashed arrow represents the transfer path for the wafer to exit the production machine after the semiconductor process treatment. Figure 1BTaking the shown production machine as an example, the transfer structure transfers the wafer 20 to the loading port 21 of the production machine, and then enters the transfer chamber 22 inside the production machine through the loading port 21. After the wafer 20 entering the transfer chamber 22 completes the direction adjustment in the orienter 23 inside the production machine, it is then sent into the processing chamber 24, and the processing chamber 24 is used for performing semiconductor manufacturing processes on the wafer 20. After the wafer undergoes semiconductor manufacturing processes in the processing chamber 24, it is transferred out of the production machine. One or more sensors are provided in the processing chamber 24 inside the production machine, and the sensors are used to collect various raw parameter data during the production process of the production machine (i.e., during the process of performing semiconductor manufacturing processes on the wafer). Multiple wafers enter the production machine in sequence for semiconductor manufacturing processes, and according to the raw parameter data of the multiple wafers collected by the sensors during the semiconductor manufacturing processes, the index diagram can be obtained. After setting the normal control range (including the upper limit and the lower limit of the normal control range) in the index diagram, the control chart of the production machine can be obtained. Through the control chart, FDC analysis can be performed on the production machine, thereby monitoring the health status of the production machine.

[0062] Step S13, obtain the statistical distribution curve of the evaluation value in the index diagram.

[0063] Specifically, by analyzing the evaluation value data in the index diagram, the statistical distribution curve is obtained. The abscissa of the statistical distribution curve is the evaluation value, and the ordinate is the occurrence times of the evaluation value in the index diagram or the proportion of the occurrence times of the evaluation value in the index diagram. The statistical distribution curve can be a normal distribution curve, a skewed peak distribution diagram, a multi-peak distribution diagram, or a multiple distribution diagram.

[0064] Optionally, the specific steps for obtaining the statistical distribution curve of the evaluation value in the index diagram include:

[0065] Obtain the initial statistical distribution curve according to all the evaluation value data in the index diagram;

[0066] Perform denoising processing on the initial statistical distribution curve to obtain the statistical distribution curve.

[0067] Specifically, since there may be small peaks in the initial statistical distribution curve graph due to deviation values (i.e., evaluation values generated due to various abnormal conditions, such as evaluation values generated when the production machine stops operating and then restarts due to reasons such as preventive maintenance of the machine, manual shutdown, downtime, etc.), therefore, in order to avoid the influence of such small peaks on the accurate setting of the normal control range of the control chart, the initial statistical distribution curve graph can be denoised to remove the small peaks caused by the deviation values, thereby forming the statistical distribution curve graph. Among them, those skilled in the art can select the specific method of denoising according to actual needs as long as the denoising effect can be achieved.

[0068] Optionally, the specific steps for denoising the initial statistical distribution curve graph include:

[0069] Removing the data within a preset proportional range at one or both ends of the initial statistical distribution curve graph.

[0070] Specifically, since the normal control range in the control chart is continuous, the small peaks generated by the deviation values appear at both ends of the initial statistical distribution curve. In order to simplify the denoising operation, the data within a preset proportional range at one or both ends of the initial statistical distribution curve graph can be removed, thereby forming the statistical distribution curve graph. Among them, the preset proportional range can be, but is not limited to, 5%.

[0071] Optionally, the specific steps for obtaining the statistical distribution curve graph of the evaluation value in the index graph include:

[0072] Judging whether the index graph has a multiple distribution. If so, performing differential processing on the index graph;

[0073] Obtaining the statistical distribution curve graph of the index graph after differential processing.

[0074] Appendix Figures 4A - 4B is a schematic diagram of the index graph with a multiple distribution in the specific embodiment of the present invention. The mean value of the evaluation values in different production time periods of the same production machine will produce a higher or lower translation, and the obtained index graph is as Figure 4A shown. In addition, under long-term use of the production machine, the overall evaluation values of the same type will show an upward or downward trend, and the obtained index graph is as Figure 4B shown. Directly performing statistical analysis on the evaluation values in the index graph with a multiple distribution will still result in a statistical distribution graph with a multiple distribution. In order to convert the statistical distribution graph with a multiple distribution into a skewed peak distribution or a multi-peak distribution to remove the influence of the translation of the mean value of the evaluation values caused by different time periods in the index graph, thereby facilitating the subsequent setting of the normal control range of the control chart, the index graph with a multiple distribution is first subjected to differential processing, and then the statistical distribution curve graph of the index graph after differential processing is obtained.

[0075] Optionally, the specific steps for differentiating the index graph include:

[0076] Determine whether there is a position in the index graph where the amplitude difference is greater than a first preset value. If so, perform a first-order differentiation on the index graph.

[0077] Optionally, the specific steps for differentiating the index graph include:

[0078] Determine whether there is a position in the index graph where the amplitude surge is greater than a second preset value. If so, perform a second-order differentiation on the index graph.

[0079] Optionally, the specific steps for differentiating the index graph include:

[0080] Determine whether there is a position in the index graph where the amplitude turn is greater than a third preset value. If so, perform a third-order differentiation on the index graph.

[0081] Appendix Figures 5A - 5C is a schematic diagram for differentiating index graphs with different characteristics in the specific embodiment of the present invention. Appendix Figure 6 is a corresponding relationship diagram between the index graphs with different differentiations and the statistical distribution curve graphs in the specific embodiment of the present invention. For multiple distribution index graphs with different characteristics, different differentiation methods are adopted to better define the normal control range of the control chart. For example, as Figure 6 shown, perform statistical analysis on the original index graph composed of original data. Since the original index graph shows a multiple distribution, the obtained statistical distribution curve graph also shows a multiple distribution. When there is a position with a relatively large amplitude difference (for example, a position where the amplitude difference is greater than the first preset value) in the original index graph, perform a first-order differentiation on the original index graph, as Figure 5A shown. Through statistical analysis of the index graph after the first-order differentiation, obtain the statistical distribution curve graph with a multi-peak or skewed-peak distribution, as Figure 6 shown. When there is a position with a relatively large amplitude surge (for example, a position where the amplitude surge is greater than the second preset value) in the original index graph, perform a second-order differentiation on the original index graph, as Figure 5B shown. Through statistical analysis of the index graph after the second-order differentiation, obtain the statistical distribution curve graph with a multi-peak or skewed-peak distribution, as Figure 6 shown. When there is a position with a relatively large amplitude turn (for example, a position where the amplitude turn is greater than the third preset value) in the original index graph, perform a third-order differentiation on the original index graph, as Figure 5CAs shown. Through statistical analysis of the index diagram after third-order differential processing, the statistical distribution curve graph with multi-peak or skewed-peak distribution is obtained, as Figure 6 shown. Among them, for the specific values of the first preset value, the second preset value, and the third preset value, those skilled in the art can set them according to actual needs.

[0082] Step S14: respectively obtain the left central point evaluation value c l of the leftmost peak, the right central point evaluation value c r of the rightmost peak, and the middle central point evaluation value c c of the middle peak.

[0083] Optionally, the specific steps of respectively obtaining the left central point evaluation value c l of the leftmost peak, the right central point evaluation value c r of the rightmost peak, and the middle central point evaluation value c c of the middle peak include:

[0084] Adopt the K-mean clustering method to respectively obtain the left central point evaluation value c l of the leftmost peak, the right central point evaluation value c r of the rightmost peak, and the middle central point evaluation value c c of the middle peak.

[0085] Optionally, the specific steps of respectively obtaining the left central point evaluation value c l of the leftmost peak, the right central point evaluation value c r of the rightmost peak, and the middle central point evaluation value c c of the middle peak include:

[0086] Judge whether the statistical distribution curve graph is a skewed-peak distribution curve. If so, obtain the peak central point evaluation value of the statistical distribution curve graph, and use the peak central point evaluation value as the left central point evaluation value c l , the right central point evaluation value c r , and the middle central point evaluation value c c at the same time.

[0087] App Figure 2 is a schematic diagram of converting a skewed-peak distribution into two normal distributions in the specific implementation manner of the present invention. The following takes the statistical distribution curve graph as a skewed-peak distribution curve as an example for illustration. As Figure 2 shown, since there is only one peak in the skewed-peak distribution curve, it can be considered that the leftmost peak, the rightmost peak, and the middle peak in the skewed-peak distribution curve coincide, that is, the left central point evaluation value c l, the evaluation value c of the right center point of the rightmost peak r , and the evaluation value c of the middle center point of the middle peak c are all equal, and are all the evaluation value c of the peak center point in the skewed peak distribution curve. In this specific embodiment, the normal distribution can also be regarded as a subset of the skewed peak distribution.

[0088] Appendix Figure 3 is a schematic diagram of converting a multi-peak distribution into two normal distributions in a specific embodiment of the present invention. The following takes the statistical distribution curve as a multi-peak distribution curve as an example for illustration. Figure 3 μ in it represents the mean of all evaluation values in the index diagram. As Figure 3 shown, the small peak on the rightmost side of the statistical distribution curve of the multi-peak distribution (i.e., Figure 3 the peak pointed to by the dotted arrow) is a small peak caused by the deviation value, that is, noise. Therefore, the small peak caused by the deviation value at the rightmost end of the statistical distribution curve shown in Figure 3 showing a multi-peak distribution can be excluded. After that, by using the K-mean clustering method or other statistical algorithms, the evaluation value c of the left center point of the leftmost peak in the statistical distribution curve is obtained respectively l , the evaluation value c of the right center point of the rightmost peak r , and the evaluation value c of the middle center point of the middle peak c ( Figure 3 not shown in the figure).

[0089] Step S15, using the evaluation value c of the left center point l as the mean, calculate the left standard deviation s of all the evaluation values less than or equal to the evaluation value c of the left center point l , and using the evaluation value c of the right center point l as the mean, calculate the right standard deviation s of all the evaluation values greater than or equal to the evaluation value c of the right center point r r . r

[0090] Specifically, all the original data in the statistical distribution curve is X. Based on all the evaluation values less than or equal to the evaluation value c of the left center point in the statistical distribution curve l , all the evaluation values are Y, and the following formula is used to obtain the left standard deviation s with the evaluation value c of the left center point l as the mean l :

[0091]

[0092]

[0093] Among them, i is a positive integer.

[0094] Based on all the evaluation values in the statistical distribution curve that are greater than or equal to the right central point evaluation value c r For all the evaluation values of Z, the following formula is used to obtain the left standard deviation s with the right central point evaluation value c r as the mean: r :

[0095]

[0096]

[0097] Among them, i is a positive integer.

[0098] For example, for the statistical distribution curve with a skewed peak distribution as in Figure 2 , obtain the peak central point evaluation value c in the statistical distribution curve, and use the peak central point evaluation value as both the left central point evaluation value c l , the right central point evaluation value c r , and the middle central point evaluation value c c . Consider the data to the left of the peak central point as a normal distribution, and consider the data to the right of the peak central point as another normal distribution. Figure 2 The solid line in the normal distribution curve represents the actual data in the index diagram. The dashed line in the normal distribution curve represents the virtual data supplemented when considering the data to the left of the peak central point as a normal distribution for calculation and the data to the right of the peak central point as another normal distribution for calculation (this part of the data does not actually exist in the index diagram). When considering the skewed peak distribution curve as two normal distribution curves, the proportion of the actual data in each normal distribution curve reaches 99.73%. Therefore, the accuracy and reliability of the normal control range obtained by considering the skewed peak distribution curve as two normal distribution curves are relatively high.

[0099] For another example, for the statistical distribution curve with a multi-peak distribution as in Figure 3 , after determining the leftmost peak, the rightmost peak, and the middle peak, directly obtain the left central point evaluation value c of the leftmost peak in the statistical distribution curve l , the right central point evaluation value c of the rightmost peak r , and the middle central point evaluation value c of the middle peak c . Combine the part to the left of the leftmost peak and the part to the right of the rightmost peak to form a skewed peak distribution curve, and then consider the formed skewed peak distribution curve as two normal distributions, that is, consider the part to the left of the left central point evaluation value c l as a skewed peak distribution, and consider the part to the right of the right central point evaluation value cr The part to the right is regarded as another skewed peak distribution. Evaluate the value c at the left center point l As the mean, calculate the left standard deviation s of all the evaluation values less than or equal to the evaluation value c at the left center point l of all the said evaluation values l , and use the evaluation value c at the right center point r As the mean, calculate the right standard deviation s of all the evaluation values greater than or equal to the evaluation value c at the right center point r of all the said evaluation values r . Figure 3 The solid line in the normal distribution curve represents the actual data in the index diagram. The dashed line in the normal distribution curve represents in order to regard the data on the left side of the evaluation value c at the left center point l as a normal distribution for calculation, and the evaluation value c at the right center point r The data on the right side is regarded as another normal distribution for calculation, and the virtual data supplemented (this part of data does not actually exist in the index diagram) is calculated. The proportion of the actual data in each normal distribution curve has reached 99.73%. Therefore, the accuracy and reliability of the normal control range obtained by regarding the skewed peak distribution curve as two normal distribution curves are relatively high.

[0100] Step S16, obtain the normal control range [c l -αs l -ρ, c r +αs r +ρ] corresponding to the control chart of the index diagram, where 3.5 ≤ α ≤ 4, and ρ is a minimum value used to adjust the error of the normal control range.

[0101] Specifically, c r +αs r +ρ is the upper limit of the normal control range in the control chart, and c l -αs l -ρ is the lower limit of the normal control range in the control chart. ρ is a minimum value, and its specific value can be determined according to the actual control accuracy requirements or the accuracy of the sensor used by the production machine for data acquisition. In an example, the value of ρ can be 0. When α = 3, the expected sample proportion range is 0.9973002, and the approximate expected frequency is 1 in 370 times; when α = 3.5, the expected sample proportion range is 0.9995347, and the approximate expected frequency is 1 in 2149 times; when α = 4, the expected sample proportion range is 0.9999367, and the approximate expected frequency is 1 in 15787 times; when α = 4.5, the expected sample proportion range is 0.9999932, and the approximate expected frequency is 1 in 147160 times.

[0102] Moreover, this specific embodiment also provides a device for obtaining the normal control range of a control chart. Attached Figure 7 is a structural block diagram of the device for obtaining the normal control range of a control chart in the specific embodiment of the present invention. The device for obtaining the normal control range of a control chart provided in this specific embodiment can adopt, for example, Figure 1A , Figure 1B , Figures 2 - 3 , Figures 4A - 4B , Figures 5A - 5C and Figure 6 shown in the method for obtaining the normal control range of a control chart to obtain the normal control range of a control chart. The device for obtaining the normal control range of a control chart includes:

[0103] A receiving circuit 75, configured to receive the original parameter data during the semiconductor manufacturing process of a wafer by a production machine;

[0104] A storage circuit 70, configured to store an index chart obtained according to the original parameter data, where the index chart is a curve of the change of a single evaluation value of the production machine over time;

[0105] A first calculation circuit 71, configured to obtain a statistical distribution curve chart of the evaluation value in the index chart;

[0106] A second calculation circuit 72, configured to respectively obtain the left center point evaluation value c l of the leftmost peak, the right center point evaluation value c r of the rightmost peak, and the middle center point evaluation value c c of the middle peak;

[0107] A third calculation circuit 73, configured to use the left center point evaluation value c l as the mean value, calculate the left standard deviation s l of all the evaluation values less than or equal to the left center point evaluation value c l , and use the right center point evaluation value c r as the mean value, calculate the right standard deviation s r of all the evaluation values greater than or equal to the right center point evaluation value c r ;

[0108] An acquisition circuit 74, configured to acquire the normal control range [c l -αs l -ρ, c r +αs r +ρ] of the control chart corresponding to the index chart, where 3.5≤α≤4, and ρ is a minimum value used to adjust the error of the normal control range.

[0109] Optionally, the first calculation circuit 71 is configured to obtain an initial statistical distribution curve graph based on data of all the evaluation values in the index graph, and perform denoising processing on the initial statistical distribution curve graph to obtain the statistical distribution curve graph.

[0110] Optionally, the first calculation circuit 71 is configured to remove data within a preset proportional range at one or both ends of the initial statistical distribution curve graph, so as to implement denoising processing on the initial statistical distribution curve graph.

[0111] Optionally, the first calculation circuit 71 is further configured to determine whether there is a multiple distribution in the index graph. If so, perform differential processing on the index graph, and obtain a statistical distribution curve graph of the index graph after the differential processing.

[0112] Optionally, the first calculation circuit 71 is configured to determine whether there is a position in the index graph where the amplitude difference is greater than a first preset value. If so, perform first-order differential processing on the index graph.

[0113] Optionally, the first calculation circuit 71 is configured to determine whether there is a position in the index graph where the amplitude surge is greater than a second preset value. If so, perform second-order differential processing on the index graph.

[0114] Optionally, the first calculation circuit 71 is configured to determine whether there is a position in the index graph where the amplitude turn is greater than a third preset value. If so, perform third-order differential processing on the index graph.

[0115] Optionally, the second calculation circuit 72 is configured to determine whether the statistical distribution curve graph is a skewed peak distribution curve. If so, obtain an evaluation value of the peak center point in the statistical distribution curve graph, and use the evaluation value of the peak center point as both the left center point evaluation value c l , the right center point evaluation value c r , and the middle center point evaluation value c c .

[0116] Optionally, the second calculation circuit 72 is configured to respectively obtain the left center point evaluation value c l of the leftmost peak, the right center point evaluation value c r of the rightmost peak, and the middle center point evaluation value c c of the middle peak in the statistical distribution curve graph by using the K-mean clustering method.

[0117] This specific embodiment also provides a computer-readable medium, on which a computer program is stored. When the computer program is executed by a processor, the method for obtaining the normal control range of the control chart as described in any one of the above is implemented.

[0118] The method and device for obtaining the normal control range of a control chart and the computer-readable medium provided by this specific implementation manner automatically obtain the normal control method of the control chart through objective calculation, avoiding the problems of high labor cost and large subjectivity caused by relying on manual subjective setting of the normal control range in the prior art. The present invention calculates the normal control range by converting the statistical distribution curve of non-normal distribution into multiple normal distribution curves. On the one hand, it avoids setting the normal control range too loose, reducing the FDC false negative rate and enabling engineers to be informed in a timely manner when abnormal conditions occur on the machine. On the other hand, it avoids setting the normal control range too tight, reducing the FDC false positive rate, thereby reducing the manual inspection cost and also reducing the missed inspection problem caused by engineers omitting manual inspection due to the high false positive rate.

[0119] The above are only the preferred implementation manners of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for obtaining the normal control range of a control chart, characterized in that, it includes the following steps: Receiving the original parameter data during the semiconductor process treatment of wafers by a production machine; Obtaining an index chart according to the original parameter data, where the index chart is a curve of the change of a single evaluation value of the production machine over time; Obtaining a statistical distribution curve chart of the evaluation value in the index chart; Obtain the left central point evaluation value c of the leftmost peak in the statistical distribution curve diagram respectively l and the right central point evaluation value c of the rightmost peak r as well as the middle central point evaluation value c of the middle peak c ; With the left center point evaluation value c l Calculate the left standard deviation s of all the evaluation values whose mean is less than or equal to the left center point evaluation value c l And calculate the right standard deviation s of all the evaluation values whose mean is greater than or equal to the right center point evaluation value c l With the right center point evaluation value c r r r ;​​ Obtain the normal control range of the control chart corresponding to the said metric chart [c l -αs l -ρ,c r +αs r +ρ], where 3.5 ≤ α ≤ 4, ρ is a minimum value used to adjust the error of the said normal control range.

2. The method for obtaining the normal control range of a control chart according to claim 1, characterized in that, The specific steps for obtaining the statistical distribution curve chart of the evaluation value in the index chart include: Obtaining an initial statistical distribution curve chart according to all the data of the evaluation values in the index chart; Performing denoising processing on the initial statistical distribution curve chart to obtain the statistical distribution curve chart.

3. The method for obtaining the normal control range of a control chart according to claim 2, characterized in that, The specific steps for performing denoising processing on the initial statistical distribution curve chart include: Removing the data within a preset proportional range at one or both ends of the initial statistical distribution curve chart.

4. The method for obtaining the normal control range of a control chart according to claim 1, characterized in that, The specific steps for obtaining the statistical distribution curve chart of the evaluation value in the index chart include: Judging whether there is a multiple distribution in the index chart. If so, performing differential processing on the index chart; Obtaining the statistical distribution curve chart of the index chart after the differential processing.

5. The method for obtaining the normal control range of a control chart according to claim 4, characterized in that, The specific steps for performing differential processing on the index chart include: Judging whether there is a position in the index chart where the amplitude difference is greater than a first preset value. If so, performing first-order differential processing on the index chart.

6. The method for obtaining the normal control range of a control chart according to claim 4, characterized in that, The specific steps for performing differential processing on the index chart include: Judging whether there is a position in the index chart where the amplitude surge is greater than a second preset value. If so, performing second-order differential processing on the index chart.

7. The method for obtaining the normal control range of a control chart according to claim 4, characterized in that, The specific steps for performing differential processing on the index chart include: Judging whether there is a position in the index chart where the amplitude turn is greater than a third preset value. If so, performing third-order differential processing on the index chart.

8. The method for obtaining the normal control range of a control chart according to claim 1, characterized in that, Obtain the left central point evaluation value c of the leftmost peak in the statistical distribution curve diagram respectively l and the right central point evaluation value c of the rightmost peak r and the middle central point evaluation value c of the middle peak c . The specific steps are as follows: Determine whether the statistical distribution curve is a skewed peak distribution curve. If so, obtain the peak center point evaluation value in the statistical distribution curve, and use the peak center point evaluation value as both the left center point evaluation value c l , the right center point evaluation value c r , and the middle center point evaluation value c c .

9. The method for obtaining the normal control range of a control chart according to claim 1, characterized in that, Obtain the left center point evaluation value c of the leftmost peak in the statistical distribution curve diagram respectively l and the right center point evaluation value c of the rightmost peak r as well as the middle center point evaluation value c of the middle peak c The specific steps are as follows: Use the K-mean clustering method to obtain the left center point evaluation value c of the leftmost peak, the right center point evaluation value c of the rightmost peak, and the middle center point evaluation value c of the middle peak in the statistical distribution curve respectively. l and the right center point evaluation value c of the rightmost peak r and the middle center point evaluation value c of the middle peak c .

10. A device for obtaining the normal control range of a control chart, characterized in that, it includes: A receiving circuit for receiving the original parameter data during the semiconductor process treatment of wafers by a production machine; A storage circuit for storing an index chart obtained according to the original parameter data, where the index chart is a curve of the change of a single evaluation value of the production machine over time; A first calculation circuit for obtaining the statistical distribution curve chart of the evaluation value in the index chart; A second computing circuit for respectively obtaining a left central point evaluation value c of the leftmost peak in the statistical distribution curve graph l and a right central point evaluation value c of the rightmost peak r and a middle central point evaluation value c of the middle peak c ; A third calculation circuit for evaluating the value c at the left center point l Calculate the left standard deviation s of all the evaluation values that are less than or equal to the evaluation value c at the left center point l for the mean calculation l and calculate the right standard deviation s of all the evaluation values that are greater than or equal to the evaluation value c at the right center point r for the mean calculation r ; r ​ An acquisition circuit for acquiring the normal control range of a control chart corresponding to the index chart [c l -αs l -ρ,c r +αs r +ρ], where 3.5 ≤ α ≤ 4 and ρ is a minimum value used to adjust the error of the normal control range.

11. The device for obtaining the normal control range of the control chart according to claim 10, wherein, the first calculation circuit is configured to obtain an initial statistical distribution curve graph based on the data of all the evaluation values in the index graph, and perform denoising processing on the initial statistical distribution curve graph to obtain the statistical distribution curve graph.

12. The device for obtaining the normal control range of the control chart according to claim 11, wherein, the first calculation circuit is configured to remove data within a preset proportion range at one or both ends of the initial statistical distribution curve graph, so as to implement the denoising processing on the initial statistical distribution curve graph.

13. The device for obtaining the normal control range of the control chart according to claim 10, wherein, the first calculation circuit is further configured to determine whether there is a multiple distribution in the index graph. If so, perform differential processing on the index graph, and obtain the statistical distribution curve graph of the index graph after the differential processing.

14. The device for obtaining the normal control range of the control chart according to claim 13, wherein, the first calculation circuit is configured to determine whether there is a position in the index graph where the amplitude difference is greater than a first preset value. If so, perform first-order differential processing on the index graph.

15. The device for obtaining the normal control range of the control chart according to claim 13, wherein, the first calculation circuit is configured to determine whether there is a position in the index graph where the amplitude surge is greater than a second preset value. If so, perform second-order differential processing on the index graph.

16. The device for obtaining the normal control range of the control chart according to claim 13, wherein, the first calculation circuit is configured to determine whether there is a position in the index graph where the amplitude turning is greater than a third preset value. If so, perform third-order differential processing on the index graph.

17. The device for obtaining the normal control range of the control chart according to claim 10, wherein, The second calculation circuit is used to determine whether the statistical distribution curve is a skewed peak distribution curve. If so, the evaluation value of the peak center point in the statistical distribution curve is obtained, and the evaluation value of the peak center point is used as both the evaluation value c of the left center point l , the evaluation value c of the right center point r and the evaluation value c of the middle center point c .

18. The device for obtaining the normal control range of the control chart according to claim 10, wherein, The second calculation circuit is used to respectively obtain the left center point evaluation value c of the leftmost peak in the statistical distribution curve graph by using the K-mean clustering method l , the right center point evaluation value c of the rightmost peak r , and the middle center point evaluation value c of the middle peak c .

19. A computer-readable medium, on which a computer program is stored, wherein, the computer program, when executed by a processor, implements the method according to any one of claims 1-9.

Images