Slicing machine cutting coordinate compensation method and device

Abstract

The present application provides a cutting coordinate compensation method and device for a dicing machine, which relates to the field of semiconductor cutting technology, including: controlling a dicing machine to cut a target wafer and obtaining wafer surface deformation data; performing a cutting path offset analysis to determine a first path offset value; if the first path offset value satisfies a predetermined three-dimensional deviation space, inputting a cutting twin model to predict the path offset, recording a second offset node and a second path offset value; performing a cutting coordinate compensation analysis based on the second path offset value to determine the second compensation data, and at the second offset node, correcting the predetermined cutting path according to the second compensation data. The present application can solve the technical problem in the prior art that the cutting path cannot be corrected in time due to the low frequency of path compensation detection and the excessive compensation delay, realize the technical goal of automatic real-time compensation of the cutting path, and achieve the technical effect of reducing the number of path adjustments and improving compensation accuracy.

Description

Technical Field

[0001] The present application relates to the field of semiconductor cutting technology, and in particular to a method and device for compensating cutting coordinates of a dicing machine. Background Art

[0002] During the wafer dicing process, the accuracy of the cutting path directly affects the processing quality and the performance of the final product.

[0003] At present, the existing technology usually uses regular or fixed time intervals to detect and compensate for path deviations. Although this method can correct deviations in the cutting process to a certain extent, it is often due to insufficient detection frequency or excessive compensation delay, which leads to the inability to correct the cutting path deviation in time, ultimately affecting the cutting quality. In addition, since the compensation operation is too dependent on manual adjustment, the number of adjustments to the cutting path is large, which not only increases the complexity of the operation, but also may cause additional error accumulation. At the same time, the compensation system is not efficient in processing multi-dimensional deviations, and the compensation accuracy is difficult to reach a high standard. Especially in the high-speed cutting scenario of wafers, the lag of traditional compensation methods further amplifies this problem, resulting in unstable cutting accuracy.

[0004] In summary, the prior art has technical problems such as the inability to timely correct the deviation of the cutting path due to low path compensation detection frequency, excessive compensation delay and excessive reliance on manual adjustment, which further affects the quality and efficiency of wafer cutting. Summary of the invention

[0005] The purpose of this application is to provide a cutting coordinate compensation method and device for a dicing machine, so as to solve the technical problems in the prior art that the cutting path cannot be corrected in time due to low path compensation detection frequency, excessive compensation delay and excessive reliance on manual adjustment, further affecting the quality and efficiency of wafer cutting.

[0006] In view of the above problems, the present application provides a cutting coordinate compensation method and device for a dicing machine.

[0007] In the first aspect, the present application provides a dicing machine cutting coordinate compensation method, which is implemented by a dicing machine cutting coordinate compensation device, including: controlling the dicing machine to cut the target wafer according to predetermined cutting parameters and a predetermined cutting path, and using a high-speed laser scanner to monitor the wafer surface state in real time during the cutting process to obtain wafer surface deformation data; performing a cutting path offset analysis based on the wafer surface deformation data to determine a first path offset value; if the first path offset value satisfies a predetermined three-dimensional deviation space, inputting the wafer surface deformation data into a cutting twin model for path offset prediction, and recording a second offset node and a second path offset value that do not satisfy the predetermined three-dimensional deviation space; performing a cutting coordinate compensation analysis based on the second path offset value to determine second compensation data, and at the second offset node, correcting the predetermined cutting path according to the second compensation data.

[0008] In the second aspect, the present application also provides a dicing machine cutting coordinate compensation device for executing the dicing machine cutting coordinate compensation method as described in the first aspect, including: a wafer cutting module, the wafer cutting module is used to control the dicing machine to cut the target wafer according to predetermined cutting parameters and a predetermined cutting path, and a high-speed laser scanner is used to monitor the surface state of the wafer in real time during the cutting process to obtain the wafer surface deformation data; an offset analysis module, the offset analysis module is used to perform a cutting path offset analysis based on the wafer surface deformation data to determine a first path offset value; an offset prediction module, the offset prediction module is used to input the wafer surface deformation data into the cutting twin model for path offset prediction if the first path offset value satisfies a predetermined three-dimensional deviation space, and record a second offset node and a second path offset value that do not satisfy the predetermined three-dimensional deviation space; a compensation analysis module, the compensation analysis module is used to perform a cutting coordinate compensation analysis based on the second path offset value to determine the second compensation data, and at the second offset node, correct the predetermined cutting path according to the second compensation data.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] The target wafer is cut by controlling the dicing machine according to the predetermined cutting parameters and the predetermined cutting path. During the cutting process, a high-speed laser scanner is used to monitor the surface state of the wafer in real time to obtain the surface deformation data of the wafer; a cutting path offset analysis is performed based on the wafer surface deformation data to determine a first path offset value; if the first path offset value satisfies the predetermined three-dimensional deviation space, the wafer surface deformation data is input into the cutting twin model to predict the path offset, and the second offset node and the second path offset value that do not satisfy the predetermined three-dimensional deviation space are recorded; a cutting coordinate compensation analysis is performed based on the second path offset value to determine the second compensation data, and at the second offset node, the predetermined cutting path is corrected according to the second compensation data to achieve the technical goal of automatic real-time compensation of the cutting path, and achieve the technical effect of reducing the number of path adjustments, improving compensation accuracy and cutting efficiency.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0013] Figure 1 A schematic diagram of the flow chart of the dicing machine cutting coordinate compensation method of the present application;

[0014] Figure 2 This is a schematic diagram of the structure of the cutting coordinate compensation device of the dicing machine of this application.

[0015] Description of reference numerals:

[0016] Wafer cutting module 11 , offset analysis module 12 , offset prediction module 13 , compensation analysis module 14 . DETAILED DESCRIPTION

[0017] This application provides a dicing machine cutting coordinate compensation method and device to solve the technical problem in the prior art that the cutting path cannot be corrected in time due to low frequency of path compensation detection, too large compensation delay, and excessive reliance on manual adjustment, which further affects the quality and efficiency of wafer cutting. The technical goal of achieving automatic real-time compensation of the cutting path is achieved, achieving the technical effect of reducing the number of path adjustments, improving compensation accuracy and cutting efficiency.

[0018] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0019] For example, please refer to the attached Figure 1 The present application provides a dicing machine cutting coordinate compensation method, which is applied to a dicing machine cutting coordinate compensation device, and specifically includes the following steps:

[0020] Step 1: According to the predetermined cutting parameters and the predetermined cutting path, the dicing machine is controlled to cut the target wafer. During the cutting process, a high-speed laser scanner is used to monitor the surface state of the wafer in real time to obtain the deformation data of the wafer surface.

[0021] Specifically, when cutting wafers, first ensure that the operation meets the process requirements according to the pre-set cutting parameters, such as cutting speed, cutting angle, coolant flow, cutting depth, blade speed, etc., and set the cutting path as the predetermined cutting path. Among them, the predetermined cutting path is the trajectory of the tool or laser moving along the specified route during the cutting process. Then, control the dicing machine to cut the target wafer according to the predetermined cutting parameters and the predetermined cutting path. In order to ensure the accuracy and quality of the cutting, a high-speed laser scanner is used to monitor the surface state of the wafer in real time during the cutting process. For example, by obtaining the deformation data of the wafer surface through the scanner, stress concentration or fine cracks that appear on the wafer during the cutting process can be detected in time. If the deformation data of a certain area exceeds the preset threshold, such as the surface displacement exceeds 0.01 mm, it may indicate that there is a structural problem with the wafer, so that the cutting process can be adjusted in time to avoid wafer damage.

[0022] Step 2: Perform cutting path deviation analysis based on the wafer surface deformation data to determine a first path deviation value.

[0023] Specifically, during the wafer cutting process, the obtained wafer surface deformation data is used to analyze the cutting path offset, obtain the actual offset of the cutting path relative to the original route, and then calculate and determine the first path offset value. For example, if the height change of a certain area on the wafer surface increases from 0.005 mm to 0.008 mm, it is judged whether the current cutting path deviates from the predetermined path. For example, if the predetermined path should be cut at a certain point, and the actual cutting position is offset by 0.02 mm, then 0.02 mm is the first path offset value.

[0024] Step three: If the first path offset value satisfies the predetermined three-dimensional deviation space, the wafer surface deformation data is input into the cutting twin model for path offset prediction, and the second offset node and the second path offset value that do not satisfy the predetermined three-dimensional deviation space are recorded.

[0025] Specifically, when the first path offset value of the wafer is detected, it is determined whether the first path offset value meets the predetermined three-dimensional deviation space, that is, the deviation range allowed in the three-dimensional coordinate system. For example, the predetermined three-dimensional deviation space can be set to an X, Y, and Z coordinate range of ±0.01 mm. If the first path offset value is within the predetermined three-dimensional deviation space, the cutting process continues normally. If the first path offset value is not within the predetermined three-dimensional deviation space, that is, the deviation exceeds the predetermined three-dimensional deviation space, such as an offset of 0.015 mm in the X direction, then the wafer surface deformation data obtained from the wafer surface needs to be input into the cutting twin model. The cutting twin model is a virtual model for predicting path offsets. By analyzing deformation data, larger errors that may occur during wafer cutting can be predicted in advance. Next, for nodes that do not meet the predetermined three-dimensional deviation space, such as points with an offset exceeding 0.02 mm, they are recorded as second offset nodes, and the second path offset value corresponding to the point is also recorded for further analysis and adjustment of the cutting path.

[0026] Step 4: Perform cutting coordinate compensation analysis based on the second path offset value to determine second compensation data, and correct the predetermined cutting path according to the second compensation data at the second offset node.

[0027] Specifically, after obtaining the second path offset value, it is necessary to perform compensation analysis of the cutting coordinates and adjust the cutting path based on the second path offset value to ensure the accuracy of the cutting. For example, if the second path offset value is 0.02 mm in the X direction and 0.01 mm in the Y direction, the compensation analysis calculates the compensation amount that can eliminate the second path offset value, which is the second compensation data. Specifically, for the adjustment values ​​in the X and Y directions, such as compensation of -0.02 mm in the X direction and -0.01 mm in the Y direction. Then, at the second offset node, that is, the specific location where the offset occurs, such as a point on the wafer, the originally predetermined cutting path is corrected by applying the second compensation data, so that the cutting path is realigned with the ideal cutting path to ensure the accuracy and quality of the final cutting result.

[0028] The dicing machine cutting coordinate compensation method is applied to the dicing machine cutting coordinate compensation device, which can achieve the technical goal of automatic real-time compensation of the cutting path, and achieve the technical effect of reducing the number of path adjustments, improving compensation accuracy and cutting efficiency.

[0029] Furthermore, the present application also includes:

[0030] The first path offset value includes a first lateral offset value, a first longitudinal offset value and a first height offset value.

[0031] Specifically, the first lateral offset value in the first path offset value refers to the displacement difference of the wafer in the X-axis direction. For example, if the predetermined cutting path should move 1 mm in the X-direction, but the actual cutting is offset by 0.005 mm, 0.005 mm is the first lateral offset value. By recording the lateral changes, the lateral accuracy of the cutting process can be analyzed.

[0032] Next, the first longitudinal offset value in the first path offset value refers to the displacement difference of the wafer in the Y-axis direction, that is, the lateral offset value. Assuming that the cutting path should advance 2 mm in the Y direction, but the actual offset is 0.01 mm, then 0.01 mm is the longitudinal offset value, and then monitor whether there is an error in the longitudinal cutting.

[0033] Then, the first height offset value is the height change value of the wafer in the Z-axis direction, which represents the offset of the cutting depth. For example, if the target cutting depth is 0.2 mm, and the actual depth deviation is 0.003 mm, then 0.003 mm is the height offset value, thereby monitoring the vertical accuracy of the cutting process.

[0034] By analyzing the offset values ​​in each direction separately, we can fully understand the overall offset of the wafer during the cutting process, thereby making more accurate path corrections and providing a basis for subsequent compensation and corrections.

[0035] Furthermore, the present application also includes:

[0036] The predetermined three-dimensional deviation space is set based on the cutting accuracy requirement, including a lateral deviation threshold, a longitudinal deviation threshold and a height deviation threshold; if the first lateral offset value does not meet the lateral deviation threshold and / or the first longitudinal offset value does not meet the longitudinal deviation threshold and / or the first height offset value does not meet the height deviation threshold, a cutting coordinate compensation analysis is performed based on the first path offset value to determine the first compensation data; the predetermined cutting path is corrected in real time according to the first compensation data.

[0037] Specifically, the predetermined three-dimensional deviation space is set according to the accuracy required in the cutting process, and includes a lateral deviation threshold, a longitudinal deviation threshold, and a height deviation threshold. The lateral deviation threshold refers to the maximum deviation range allowed in the X-axis direction. For example, if the allowable lateral deviation is ±0.005 mm, then the actual deviation of the cutting path in the X-direction should not exceed ±0.005 mm. Similarly, the longitudinal deviation threshold is for the allowable error range in the Y-axis direction, such as ±0.01 mm. The height deviation threshold sets the allowable error of the cutting depth, such as ±0.002 mm. The lateral deviation threshold, longitudinal deviation threshold, and height deviation threshold determine the allowable error range in the cutting process to ensure the accuracy and quality of the cutting.

[0038] Next, if during the cutting process, when any one or two of the conditions that the first lateral offset value does not meet the predetermined lateral deviation threshold, the first longitudinal offset value does not meet the longitudinal deviation threshold, and the first height offset value does not meet the height deviation threshold are met, or all the conditions are met at the same time, the offset value that does not meet the corresponding deviation threshold condition is analyzed for cutting coordinate compensation to determine the first compensation data. For example, if the lateral offset value reaches 0.006 mm, which exceeds the predetermined range of ±0.005 mm, then the numerical analysis of the lateral offset value exceeding the threshold is performed to obtain the first compensation data for cutting coordinate compensation analysis, and then the cutting path is adjusted to correct the deviation.

[0039] Finally, the original predetermined cutting path is corrected in real time based on the calculated first compensation data. Real-time correction means that the cutting process will not stop, but the cutting path will be dynamically adjusted according to the change of the offset value to ensure the cutting accuracy, and then the deviation will be corrected immediately to avoid excessive error accumulation. For example, when the lateral offset reaches 0.006 mm, after compensating -0.001 mm, the path will be restored to the set error range.

[0040] By determining compensation data to adjust the cutting path in real time, it is ensured that even if errors occur during the cutting process, the accuracy and quality of the final cutting can be maintained through real-time correction.

[0041] Furthermore, the present application also includes:

[0042] Read the device property data of the dicing machine and the wafer property data of the target wafer; before cutting the target wafer, use a high-speed laser scanner to collect the three-dimensional data of the wafer surface; use digital twin technology to simulate and construct the cutting twin model based on the device property data, wafer property data, three-dimensional data of the wafer surface, predetermined cutting parameters and predetermined cutting path.

[0043] Specifically, the device attribute data of the dicing machine and the wafer attribute data of the target wafer are read. The device attribute data includes the hardware information and performance parameters of the dicing machine, such as cutting speed, type and angle of the tool. The wafer attribute data includes information such as the size, material properties, thickness, etc. of the wafer. For example, the thickness of the wafer is 0.5 mm and the diameter is 150 mm, thereby ensuring that the cutting process is adapted to different wafer materials and structures.

[0044] Next, before cutting the target wafer, a high-speed laser scanner is used to collect 3D data on the wafer surface to ensure an accurate understanding of the wafer shape before cutting. The 3D data can record the tiny bumps and curvatures on the wafer surface, such as the height difference of a certain area on the surface is 0.002 mm, thereby improving the cutting accuracy and avoiding cutting deviations due to surface irregularities.

[0045] Then, digital twin technology is used to build a cutting twin model based on the equipment attribute data, wafer attribute data, and wafer surface three-dimensional data. Digital twin technology is a virtual simulation technology that can simulate the real cutting process through predetermined cutting parameters and predetermined cutting paths. For example, the cutting speed of the dicing machine is set to 0.5 mm per second, and the path is a curved shape. It is input into the simulation model, and the cutting effect of the wafer is simulated through digital twin technology, so as to predict problems in cutting in advance.

[0046] Simulation can be used to optimize the cutting process and ensure cutting accuracy, such as avoiding cutting errors caused by uneven wafer surface or inappropriate equipment parameters.

[0047] Furthermore, the present application also includes:

[0048] The wafer surface deformation data is input into the cutting twin model, and the cutting path offset predictions for multiple time nodes in a future period are performed in sequence from first to last, and the predicted offset values ​​are recorded, wherein the predicted offset values ​​include lateral predicted offset values, longitudinal predicted offset values ​​and height predicted offset values; if the lateral predicted offset value does not meet the lateral deviation threshold and / or the longitudinal predicted offset value does not meet the longitudinal deviation threshold and / or the height predicted offset value does not meet the height deviation threshold, the recorded offset time node is set as the second offset node, and the predicted offset value of the second offset node is set as the second path offset value.

[0049] Specifically, the wafer surface deformation data is input into the cutting twin model, and the future cutting offset is predicted through simulation technology. In this process, the cutting path offset is predicted for multiple time nodes in the future cycle in turn. Each time node represents a specific moment in the cutting process. By simulating the offset at multiple moments, it is known in advance whether the cutting will deviate. For example, at some point in the future, the wafer may deviate by 0.004 mm in the X direction, which is the lateral predicted offset value, and by 0.006 mm in the Y direction, which is the longitudinal predicted offset value. The height predicted offset value represents the height deviation in the Z direction.

[0050] Next, record the predicted offset value and determine whether the predicted offset value exceeds the set allowable deviation range. For example, if the set lateral deviation threshold is ±0.003 mm, if the lateral predicted offset value reaches 0.004 mm, the lateral deviation threshold is no longer met. Similarly, the longitudinal and height predicted offset values ​​are also compared based on the preset longitudinal and height deviation thresholds. If the predicted offset value in any direction exceeds the threshold, there is a large cutting deviation at that time node, which may affect the cutting quality.

[0051] Then, when the predicted offset value in any direction does not meet the corresponding deviation threshold, the time node is marked as the second offset node, that is, the moment when a large offset is expected to occur during the cutting process. At this time, the predicted offset value of the time node is recorded as the second path offset value. For example, if the predicted offset in the X direction of a time node is 0.005 mm, in the Y direction is 0.007 mm, and in the height direction is 0.002 mm, and the X and Y directions exceed the threshold, then the predicted offset values ​​in the X and Y directions will be recorded as the second path offset value of the node.

[0052] By discovering potential cutting errors in advance and making timely adjustments, cutting accuracy can be ensured, thereby preventing subsequent problems from occurring.

[0053] Furthermore, the present application also includes:

[0054] Using the device attribute data, wafer attribute data and predetermined cutting parameters as similarity retrieval constraints, a positive sample retrieval is performed to obtain a sample compensation data set that meets a predetermined similarity threshold, wherein the sample compensation data includes a sample path offset value and a sample three-dimensional compensation data; using the sample path offset value as input and the sample three-dimensional compensation data as supervision, a neural network prediction operator is trained and verified to construct a coordinate compensation analyzer; using the coordinate compensation analyzer to perform cutting coordinate compensation analysis on the second path offset value, and outputting the second compensation data.

[0055] Specifically, the equipment attribute data, wafer attribute data and predetermined cutting parameters are used as constraints for similarity retrieval to retrieve positive samples, that is, to find historical data similar to the current cutting conditions in the database. The predetermined similarity threshold refers to the lower limit of the similarity between two sets of data. Only the results of the positive sample retrieval that reaches the predetermined similarity threshold will be considered to be of reference value. For example, if the similarity threshold is set to 90%, only those samples with a similarity greater than or equal to 90% will be selected. The retrieval result is a sample compensation data set, which contains the sample path offset values ​​recorded in the previous cutting process and the three-dimensional compensation data of the sample. For example, the X-direction offset of a historical sample is 0.003 mm, and the Y-direction offset is 0.004 mm. The compensation data is the corresponding adjustment amount.

[0056] Next, the sample path offset value is used as input, and the three-dimensional compensation data of the sample is used as a supervisory signal to train and verify the neural network prediction operator. Among them, the sample path offset value and the sample three-dimensional compensation data are divided into a training set and a verification set. The neural network prediction operator is trained with the training set. When the output accuracy of the neural network prediction operator exceeds the output accuracy threshold, the neural network prediction operator is verified using the verification set. If the verification result meets the verification accuracy threshold, the verification of the neural network prediction operator is completed, and the coordinate compensation analyzer is obtained. The output accuracy threshold and the verification accuracy threshold are customized by technicians in this field according to actual conditions, such as the output accuracy threshold is 85% and the verification accuracy threshold is 90%.

[0057] Finally, the trained coordinate compensation analyzer is used to perform cutting coordinate compensation analysis on the second path offset value, and the newly detected path offset value is compensated and analyzed to output the corresponding second compensation data. For example, the second path offset value is 0.004 mm in the X direction. The coordinate compensation analyzer calculates a compensation value based on the training results, which may be -0.004 mm, so as to perform real-time path correction.

[0058] The coordinate compensation analyzer is used to compensate and analyze the currently detected path deviation to ensure the accuracy of the cutting path and avoid excessive deviation, so that the cutting returns to the predetermined path.

[0059] Furthermore, the present application also includes:

[0060] The historical compensation log is called to analyze and determine the path compensation delay; and the second offset node is updated forward according to the path compensation delay.

[0061] Specifically, call the historical compensation log to analyze the delay problem that occurs during the path compensation process. The historical compensation log records the previous cutting compensation data, including the specific time of each compensation execution and the path offset adjustment. By analyzing the historical compensation log, determine whether there is a delay when performing path compensation, and the size of the delay. For example, if a compensation operation is planned to be performed at an offset of 0.005 mm, but due to the system response time, the actual compensation occurs when the offset reaches 0.007 mm, then the difference is the path compensation delay.

[0062] Next, after determining the path compensation delay, the second offset node is updated forward according to the delay. The second offset node refers to the time point at which a large offset is predicted to occur during the cutting process. Because path compensation has a certain delay, in order to ensure that the compensation can take effect in time, the time point is appropriately advanced. For example, if the delay is 0.002 seconds, the original second offset node may need to be moved forward 0.002 seconds so that the compensation operation can be successfully completed before the offset occurs.

[0063] By calling the historical compensation log, analyzing the possible delay in the path compensation process, and then moving the second offset node forward according to the delay, it can be ensured that the compensation operation takes effect in time and avoid cutting errors caused by delays.

[0064] In summary, the dicing machine cutting coordinate compensation method provided in this application has the following technical effects:

[0065] The target wafer is cut by controlling the dicing machine according to the predetermined cutting parameters and the predetermined cutting path. During the cutting process, a high-speed laser scanner is used to monitor the surface state of the wafer in real time to obtain the surface deformation data of the wafer; a cutting path offset analysis is performed based on the wafer surface deformation data to determine a first path offset value; if the first path offset value satisfies the predetermined three-dimensional deviation space, the wafer surface deformation data is input into the cutting twin model to predict the path offset, and the second offset node and the second path offset value that do not satisfy the predetermined three-dimensional deviation space are recorded; a cutting coordinate compensation analysis is performed based on the second path offset value to determine the second compensation data, and at the second offset node, the predetermined cutting path is corrected according to the second compensation data to achieve the technical goal of automatic real-time compensation of the cutting path, and achieve the technical effect of reducing the number of path adjustments, improving compensation accuracy and cutting efficiency.

[0066] Embodiment 2, based on the same inventive concept as the dicing machine cutting coordinate compensation method in the previous embodiment, this application also provides a dicing machine cutting coordinate compensation device, please refer to the attached Figure 2 ,include:

[0067] A wafer cutting module 11, wherein the wafer cutting module 11 is used to control the dicing machine to cut the target wafer according to predetermined cutting parameters and a predetermined cutting path, and a high-speed laser scanner is used to monitor the surface state of the wafer in real time during the cutting process to obtain the surface deformation data of the wafer; an offset analysis module 12, wherein the offset analysis module 12 is used to perform a cutting path offset analysis based on the wafer surface deformation data to determine a first path offset value; an offset prediction module 13, wherein the offset prediction module 13 is used to input the wafer surface deformation data into the cutting twin model to predict the path offset if the first path offset value satisfies a predetermined three-dimensional deviation space, and record a second offset node and a second path offset value that do not satisfy the predetermined three-dimensional deviation space; a compensation analysis module 14, wherein the compensation analysis module 14 is used to perform a cutting coordinate compensation analysis based on the second path offset value to determine the second compensation data, and at the second offset node, correct the predetermined cutting path according to the second compensation data.

[0068] Furthermore, the dicing machine cutting coordinate compensation device is also used for:

[0069] The first path offset value includes a first lateral offset value, a first longitudinal offset value and a first height offset value.

[0070] Furthermore, the dicing machine cutting coordinate compensation device is also used for:

[0071] The predetermined three-dimensional deviation space is set based on the cutting accuracy requirement, including a lateral deviation threshold, a longitudinal deviation threshold and a height deviation threshold; if the first lateral offset value does not meet the lateral deviation threshold and / or the first longitudinal offset value does not meet the longitudinal deviation threshold and / or the first height offset value does not meet the height deviation threshold, a cutting coordinate compensation analysis is performed based on the first path offset value to determine the first compensation data; the predetermined cutting path is corrected in real time according to the first compensation data.

[0072] Furthermore, the dicing machine cutting coordinate compensation device is also used for:

[0073] Read the device property data of the dicing machine and the wafer property data of the target wafer; before cutting the target wafer, use a high-speed laser scanner to collect the three-dimensional data of the wafer surface; use digital twin technology to simulate and construct the cutting twin model based on the device property data, wafer property data, three-dimensional data of the wafer surface, predetermined cutting parameters and predetermined cutting path.

[0074] Furthermore, the dicing machine cutting coordinate compensation device is also used for:

[0075] The wafer surface deformation data is input into the cutting twin model, and the cutting path offset predictions for multiple time nodes in a future period are performed in sequence from first to last, and the predicted offset values ​​are recorded, wherein the predicted offset values ​​include lateral predicted offset values, longitudinal predicted offset values ​​and height predicted offset values; if the lateral predicted offset value does not meet the lateral deviation threshold and / or the longitudinal predicted offset value does not meet the longitudinal deviation threshold and / or the height predicted offset value does not meet the height deviation threshold, the recorded offset time node is set as the second offset node, and the predicted offset value of the second offset node is set as the second path offset value.

[0076] Furthermore, the dicing machine cutting coordinate compensation device is also used for:

[0077] Using the device attribute data, wafer attribute data and predetermined cutting parameters as similarity retrieval constraints, a positive sample retrieval is performed to obtain a sample compensation data set that meets a predetermined similarity threshold, wherein the sample compensation data includes a sample path offset value and a sample three-dimensional compensation data; using the sample path offset value as input and the sample three-dimensional compensation data as supervision, a neural network prediction operator is trained and verified to construct a coordinate compensation analyzer; using the coordinate compensation analyzer to perform cutting coordinate compensation analysis on the second path offset value, and outputting the second compensation data.

[0078] Furthermore, the dicing machine cutting coordinate compensation device is also used for:

[0079] The historical compensation log is called to analyze and determine the path compensation delay; and the second offset node is updated forward according to the path compensation delay.

[0080] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The dicing machine cutting coordinate compensation method and specific examples in the aforementioned embodiment 1 are also applicable to the dicing machine cutting coordinate compensation device of this embodiment. Through the aforementioned detailed description of the dicing machine cutting coordinate compensation method, those skilled in the art can clearly know the dicing machine cutting coordinate compensation device of this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0081] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0082] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A dicing machine cutting coordinate compensation method, characterized in that: include: According to the predetermined cutting parameters and the predetermined cutting path, the dicing machine is controlled to cut the target wafer. During the cutting process, a high-speed laser scanner is used to monitor the surface state of the wafer in real time to obtain the deformation data of the wafer surface; Performing a cutting path deviation analysis based on the wafer surface deformation data to determine a first path deviation value; If the first path offset value satisfies a predetermined three-dimensional deviation space, the wafer surface deformation data is input into the cutting twin model for path offset prediction, and a second offset node and a second path offset value that do not satisfy the predetermined three-dimensional deviation space are recorded; Perform cutting coordinate compensation analysis based on the second path offset value to determine second compensation data, and correct the predetermined cutting path according to the second compensation data at the second offset node; The wafer surface deformation data is input into the cutting twin model for path deviation prediction, which also includes: Reading device attribute data of the dicing machine and wafer attribute data of the target wafer; Before cutting the target wafer, a high-speed laser scanner is used to collect three-dimensional data of the wafer surface; Using digital twin technology, the cutting twin model is constructed by simulation according to the device attribute data, wafer attribute data, wafer surface three-dimensional data, predetermined cutting parameters and predetermined cutting path; Performing a cutting coordinate compensation analysis based on the second path offset value to determine second compensation data includes: Taking the device attribute data, wafer attribute data and predetermined cutting parameters as similarity retrieval constraints, a positive sample retrieval is performed to obtain a sample compensation data set that meets a predetermined similarity threshold, wherein the sample compensation data includes a sample path offset value and a sample three-dimensional compensation data; Taking the sample path offset value as input and the sample three-dimensional compensation data as supervision, the neural network prediction operator is trained and verified to construct a coordinate compensation analyzer; The coordinate compensation analyzer is used to perform cutting coordinate compensation analysis on the second path offset value, and the second compensation data is output.

2. The dicing machine cutting coordinate compensation method according to claim 1, characterized in that: The first path offset value includes a first lateral offset value, a first longitudinal offset value and a first height offset value.

3. The dicing machine cutting coordinate compensation method according to claim 2, characterized in that: The predetermined three-dimensional deviation space is set based on the cutting accuracy requirement, and includes a lateral deviation threshold, a longitudinal deviation threshold, and a height deviation threshold; If the first lateral offset value does not satisfy the lateral deviation threshold and / or the first longitudinal offset value does not satisfy the longitudinal deviation threshold and / or the first height offset value does not satisfy the height deviation threshold, performing cutting coordinate compensation analysis based on the first path offset value to determine first compensation data; The predetermined cutting path is corrected in real time according to the first compensation data.

4. The dicing machine cutting coordinate compensation method according to claim 3, characterized in that: Inputting the wafer surface deformation data into the cutting twin model to predict the path deviation includes: Inputting the wafer surface deformation data into the cutting twin model, predicting the cutting path offset of multiple time nodes in the future cycle in order from the earliest to the latest, and recording the predicted offset values, wherein the predicted offset values ​​include a lateral predicted offset value, a longitudinal predicted offset value, and a height predicted offset value; If the lateral predicted offset value does not satisfy the lateral deviation threshold and / or the longitudinal predicted offset value does not satisfy the longitudinal deviation threshold and / or the height predicted offset value does not satisfy the height deviation threshold, the recorded offset time node is set to the second offset node, and the predicted offset value of the second offset node is set to the second path offset value.

5. The dicing machine cutting coordinate compensation method according to claim 1, characterized in that: The method also includes: Call historical compensation logs to analyze and determine path compensation delay; The second offset node is updated forward according to the path compensation delay.

6. A dicing machine cutting coordinate compensation device, characterized in that: The steps for implementing the dicing machine cutting coordinate compensation method according to any one of claims 1 to 5 include: A wafer cutting module, which is used to control a dicing machine to cut a target wafer according to predetermined cutting parameters and a predetermined cutting path, and to use a high-speed laser scanner to monitor the surface state of the wafer in real time during the cutting process to obtain deformation data of the wafer surface; An offset analysis module, the offset analysis module is used to perform a cutting path offset analysis based on the wafer surface deformation data to determine a first path offset value; An offset prediction module, wherein if the first path offset value satisfies a predetermined three-dimensional deviation space, the wafer surface deformation data is input into a cutting twin model for path offset prediction, and a second offset node and a second path offset value that do not satisfy the predetermined three-dimensional deviation space are recorded; A compensation analysis module is used to perform cutting coordinate compensation analysis based on the second path offset value, determine second compensation data, and correct the predetermined cutting path according to the second compensation data at the second offset node.

Images