Process parameter control method and system for use in chip production

Abstract

A process parameter control method and system for use in chip production. The method comprises: acquiring N dicing process production lines for a target chip; generating N device reliability factors; generating N dicing quality monitoring data sets; performing deviation analysis, and generating N quality deviation factors; determining a set of preset dicing process parameters; on the basis of the N device reliability factors and the N quality deviation factors, respectively adjusting a set of tolerance thresholds of the set of preset dicing process parameters to obtain N sets of adjusted tolerance thresholds; performing parameter optimization to obtain N sets of target optimal process parameters; and respectively transmitting the N sets of target optimal process parameters to N parameter control units of the N dicing process production lines for process parameter control. The method solves the technical problems of inconsistent chip dicing quality and low parameter control efficiency in the prior art, and achieves the technical effect of improving the quality of dicing process parameter control in chip production.

Description

Process parameter control method and system for chip production Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a process parameter control method and system for chip production. Background Art

[0002] Chip quality is affected by multiple factors. The quality of the chip dicing process affects the chip edge uniformity and the high-quality execution of multiple subsequent process steps. Currently, dicing process optimization is primarily based on feedback from the dicing quality of already produced chips. However, due to overly one-sided considerations during the optimization process, process parameter control fails to meet expectations and control feedback efficiency is poor. Existing technologies suffer from poor chip dicing quality uniformity and low parameter control efficiency.

[0003] Summary of the Invention

[0004] The present application provides a process parameter control method and system for chip production, which is used to solve the technical problems in the prior art such as poor chip dicing quality balance and low parameter control efficiency.

[0005] In view of the above problems, the present application provides a process parameter control method and system for chip production.

[0006] In a first aspect of the present application, a process parameter control method for chip production is provided, the method comprising:

[0007] Obtain N dicing process production lines for the target chip;

[0008] Traversing and collecting the fully automatic dicing machine information of the N dicing process production lines to perform reliability analysis and generate N equipment reliability factors;

[0009] Generate N monitoring dicing quality data sets, wherein the N monitoring dicing quality data sets are obtained by extracting data from a set of dicing quality inspection reports of chips produced by N dicing process production lines within a preset monitoring time domain, using dicing quality as an index;

[0010] Performing an offset analysis on the N monitoring dicing quality data sets to generate N quality offset factors;

[0011] Determining a preset dicing process parameter set based on the design information of the target chip;

[0012] Adjusting the tolerance threshold sets of the preset dicing process parameter set based on the N equipment reliability factors and the N quality deviation factors respectively to obtain N adjusted tolerance threshold sets;

[0013] Optimizing N process parameter sets of N dicing process production lines based on N adjustment tolerance threshold sets to obtain N target optimal process parameter sets;

[0014] The N target optimal process parameter sets are respectively transmitted to N parameter control units of N dicing process production lines for process parameter control.

[0015] A second aspect of the present application provides a process parameter control system for chip production, the system comprising:

[0016] A process production line acquisition module is used to obtain N dicing process production lines of the target chip;

[0017] A reliability factor generation module is used to traverse and collect information of the fully automatic dicing machines of the N dicing process production lines to perform reliability analysis and generate N equipment reliability factors;

[0018] A quality data set generation module is used to generate N monitoring scribing quality data sets, wherein the N monitoring scribing quality data sets are obtained by extracting data from a set of scribing quality inspection reports of chips produced by N scribing process production lines within a preset monitoring time domain, using scribing quality as an index;

[0019] a deviation factor generating module, configured to perform deviation analysis on the N monitoring dicing quality data sets to generate N quality deviation factors;

[0020] A process parameter set obtaining module, configured to determine a preset dicing process parameter set based on the design information of the target chip;

[0021] a tolerance threshold obtaining module, configured to adjust the tolerance threshold set of the preset dicing process parameter set based on the N equipment reliability factors and the N quality deviation factors, respectively, to obtain N adjusted tolerance threshold sets;

[0022] An optimal process parameter set acquisition module is used to optimize N process parameter sets of N dicing process production lines based on N adjustment tolerance threshold sets to obtain N target optimal process parameter sets;

[0023] The process parameter control module is used to transmit the N target optimal process parameter sets to N parameter control units of N dicing process production lines for process parameter control.

[0024] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0025] This application obtains N target chip dicing process lines, then traverses and collects information about the fully automatic dicing machines of these N dicing process lines for reliability analysis, generates N equipment reliability factors, and then generates N monitoring dicing quality data sets. These N monitoring dicing quality data sets are obtained by extracting data from a set of dicing quality inspection reports of chips produced by the N dicing process lines within a preset monitoring time domain, using dicing quality as an index. Then, offset analysis is performed on the N monitoring dicing quality data sets to generate N quality offset factors. Then, a preset dicing process parameter set is determined based on the design information of the target chip. A tolerance threshold set of the preset dicing process parameter set is adjusted based on the N equipment reliability factors and N quality offset factors, obtaining N adjusted tolerance threshold sets. Parameter optimization is performed on the N process parameter sets of the N dicing process lines based on the N adjusted tolerance threshold sets to obtain N target optimal process parameter sets. These N target optimal process parameter sets are then transmitted to N parameter control units of the N dicing process lines for process parameter control. This achieves the technical effect of improving the process parameter control quality of the dicing process in chip production. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] FIG1 is a schematic flow chart of a process parameter control method for chip production provided by an embodiment of the present application;

[0028] FIG2 is a schematic diagram of a process flow for obtaining multiple fine-tuning step sets for multiple follow-up particle density sets in a process parameter control method for chip production provided by an embodiment of the present application;

[0029] FIG3 is a schematic diagram of the structure of a process parameter control system for chip production provided in an embodiment of the present application.

[0030] Explanation of the accompanying drawings: process production line acquisition module 11, reliability factor generation module 12, quality data set generation module 13, deviation factor generation module 14, process parameter set acquisition module 15, tolerance threshold acquisition module 16, optimal process parameter set acquisition module 17, process parameter control module 18. DETAILED DESCRIPTION

[0031] This application provides a process parameter control method and system for chip production, which is used to solve the technical problems in the prior art such as poor chip dicing quality balance and low parameter control efficiency.

[0032] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0033] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0034] Example 1

[0035] As shown in FIG1 , the present application provides a process parameter control method for chip production, wherein the method comprises:

[0036] S100: Obtain N dicing process production lines of the target chip;

[0037] In one possible embodiment, the target chip is any chip requiring scribing process parameter control. The N scribing process production lines are production lines set up in a production workshop that can simultaneously perform scribing processes on the target chips, each production line having a fully automatic scribing machine, wherein the fully automatic scribing machine is used to cut the thinned wafer into individual chips.

[0038] S200: Traversing and collecting information of the fully automatic dicing machines of the N dicing process production lines to perform reliability analysis and generate N equipment reliability factors;

[0039] Furthermore, step S200 in the embodiment of the present application further includes:

[0040] Using the design life as an index, data is extracted from the fully automatic dicing machine information of N dicing process production lines to obtain N design lives;

[0041] Using service life as an index, data is extracted from the fully automatic dicing machine information of N dicing process production lines to obtain N service lives;

[0042] Collect the number of fault repairs on N dicing process production lines and obtain N fault repair information;

[0043] The reliability identification network layer is used to identify N design life, N service life and N fault maintenance information to obtain N equipment reliability factors.

[0044] In one embodiment, N equipment reliability factors are determined by collecting information from fully automatic dicing machines on each of N dicing process lines. The N equipment reliability factors reflect the degree of influence of the equipment on the chip dicing quality. The higher the equipment reliability factor, the more reliable the corresponding dicing process line.

[0045] Preferably, the design life is used as an index to retrieve data from the fully automatic dicing machine information of N dicing process production lines to obtain N design lifespans, and then the service life is used as an index to extract data from the fully automatic dicing machine information of N dicing process production lines to obtain N service lives. The design life is the life of the fully automatic dicing machine under normal working conditions. The service life is the length of time the fully automatic dicing machine has been in use. Furthermore, the number of fault repairs of the N dicing process production lines is collected to obtain N fault repair information, wherein the N fault repair information reflects the equipment loss of the N dicing process production lines. The reliability identification network layer is used to identify the N design lifespans, N service lives, and N fault repair information to obtain N equipment reliability factors.

[0046] Preferably, by obtaining multiple sample design life, multiple sample service life, multiple sample fault repair information and multiple sample equipment reliability factors as training data, the feedforward neural network is supervised trained until the output reaches convergence, thereby obtaining the trained reliability identification network layer.

[0047] S300: Generate N monitoring scribing quality data sets, wherein the N monitoring scribing quality data sets are obtained by extracting data from a set of scribing quality inspection reports of chips produced by N scribing process production lines within a preset monitoring time domain, using scribing quality as an index;

[0048] In one embodiment, a data search is performed on a set of dicing quality inspection reports of chips produced by N dicing process production lines within a preset monitoring time domain using dicing quality as an index, thereby obtaining the N monitoring dicing quality data sets. The N monitoring dicing quality data sets reflect the dicing quality of the N dicing process production lines within the preset monitoring time domain. The preset detection time domain is a pre-set time period for dicing quality monitoring.

[0049] S400: performing an offset analysis on the N monitoring dicing quality data sets to generate N quality offset factors;

[0050] Furthermore, step S400 in the embodiment of the present application further includes:

[0051] Randomly select a monitoring dicing quality dataset from the N monitoring dicing quality datasets as a first monitoring dicing quality dataset;

[0052] Constructing a first offset analysis space based on the first monitoring scribing quality data set, wherein the first offset analysis space has a plurality of particle points, and the plurality of particle points correspond one-to-one to the monitoring scribing quality data in the first monitoring scribing quality data set;

[0053] performing an offset analysis based on the plurality of particle points and the first offset analysis space to generate a first mass offset factor;

[0054] N offset analysis spaces are generated according to the N monitoring dicing quality data sets, and after performing offset analysis, N quality offset factors are generated.

[0055] Furthermore, as shown in FIG2 , step S400 in the embodiment of the present application further includes:

[0056] Randomly generating a plurality of guiding particles and a plurality of following particle sets from the plurality of particle points, wherein the following particles in the plurality of following particle sets are generated within a guiding distance threshold range of the plurality of guiding particles;

[0057] Calculating a plurality of guiding particle densities and a plurality of following particle density sets of the plurality of guiding particles and the plurality of following particle sets within a preset step size;

[0058] The inverse of the ratio of the multiple guiding particle densities of the multiple guiding particles to the sum of the multiple guiding particle densities is traversed and calculated, and a preset fine-tuning step size is multiplied according to the calculation result to obtain multiple fine-tuning step size sets of the multiple following particle density sets.

[0059] Furthermore, step S400 in the embodiment of the present application further includes:

[0060] Performing N iterations in any direction on the multiple following particle sets according to the multiple fine-tuning step sets to obtain N iterative following particle sets;

[0061] Selecting N iterative following particle density sets, a plurality of guiding particle densities, and a plurality of following particle density sets corresponding to the N iterative following particle sets, and determining a particle with the largest particle density as a target particle;

[0062] The difference between the monitored dicing quality data and the preset dicing quality data corresponding to the target particle is calculated, and the ratio of the difference to the preset dicing quality data is calculated, and a first mass shift factor is generated according to the calculation result.

[0063] In one possible embodiment, N quality offset factors are generated by analyzing the offset between the dicing quality in N monitoring dicing quality data sets and the preset dicing quality data. Preferably, a monitoring dicing quality data set is randomly selected from the N monitoring dicing quality data sets as the first monitoring dicing quality data set, and a first offset analysis space is constructed based on the first monitoring dicing quality data set, wherein the first offset analysis space has a plurality of particle points, and the plurality of particle points correspond one-to-one to the monitoring dicing quality data in the first monitoring dicing quality data set. Preferably, the first offset analysis space is a two-dimensional coordinate system, and each coordinate point corresponds to a monitoring dicing quality data set. Therefore, a plurality of particle points can be obtained based on the first monitoring dicing quality data set. An offset analysis is performed based on the plurality of particle points and the first offset analysis space to generate a first quality offset factor. The first quality offset factor reflects the dicing quality of the dicing process production line corresponding to the first monitoring dicing quality data set. N offset analysis spaces are generated based on the N monitoring dicing quality data sets, and after the offset analysis is performed, N quality offset factors are generated.

[0064] In one embodiment, multiple guiding particles and multiple following particle sets are randomly generated from the multiple particle points, wherein the following particles in the multiple following particle sets are generated within a guiding distance threshold range of the multiple guiding particles. The guiding particles are used for iterative particle guidance, and the following particles are used for iteration. Furthermore, multiple guiding particle densities and multiple following particle density sets are calculated for the multiple guiding particles and the multiple following particle sets within a preset step size. The particle density is calculated by calculating the ratio of the number of particles within the preset step size to the particle's distance to the particle in the first offset analysis space divided by the area formed by the outermost particles, reflecting the density of particles clustered around the particle. Furthermore, the inverse of the ratio of the multiple guiding particle densities to the sum of the multiple guiding particle densities is calculated for the multiple guiding particles. The calculated result is multiplied by a preset fine-tuning step size to obtain multiple fine-tuning step size sets for the multiple following particle sets. The preset step size is a density calculation range set by one skilled in the art. The preset fine-tuning step size is a preset distance for fine-tuning the following particles. By obtaining multiple fine-tuning step size sets, the multiple following particle sets are adaptively adjusted, achieving the technical effect of improving iteration accuracy.

[0065] The multiple following particle sets are iterated N times in any direction according to the multiple fine-tuning step sets to obtain N iterative following particle sets. Then, N iterative following particle density sets, multiple guide particle densities, and multiple following particle density sets corresponding to the N iterative following particle sets are selected, and the particle with the largest particle density is determined as the target particle, wherein the target particle is the particle that best represents the first offset analysis space. A first quality offset factor is generated based on the calculation result by calculating the difference between the monitored dicing quality data corresponding to the target particle and the preset dicing quality data, and calculating the ratio of the difference to the preset dicing quality data.

[0066] S500: Determine a preset dicing process parameter set based on the design information of the target chip;

[0067] S600: adjusting the tolerance threshold set of the preset dicing process parameter set based on the N equipment reliability factors and the N quality deviation factors respectively to obtain N adjusted tolerance threshold sets;

[0068] In one possible embodiment, a preset set of dicing process parameters is determined for dicing based on the design information of the target chip. The preset set of dicing process parameters is pre-set by a person skilled in the art and includes parameters such as tool material, tool speed, and cutting fluid ratio. Furthermore, the tolerance threshold set of the preset dicing process parameter set is adjusted for each of N equipment reliability factors and N quality deviation factors to obtain N sets of adjusted tolerance thresholds. The tolerance threshold set represents the allowable range of parameters corresponding to the process parameters.

[0069] S700: Optimizing N process parameter sets of N dicing process lines based on N adjustment tolerance threshold sets to obtain N target optimal process parameter sets;

[0070] S800: Transmitting the N target optimal process parameter sets to N parameter control units of N dicing process production lines respectively for process parameter control.

[0071] Furthermore, step S700 in the embodiment of the present application further includes:

[0072] Randomly adjusting the N process parameter sets according to a preset adjustment method to obtain N initial process parameter neighborhoods, wherein the preset adjustment method is to randomly increase or decrease the parameters in the process parameter set;

[0073] Eliminating out-of-specification parameters from the N initial process parameter neighborhoods based on the N adjustment tolerance threshold sets to obtain N target process parameter neighborhoods;

[0074] N first process parameter sets are randomly sampled without replacement from the N target process parameter neighborhoods, and the N first process parameter sets are identified using a fitness identification network layer to obtain N first fitnesses.

[0075] Furthermore, step S700 in the embodiment of the present application further includes:

[0076] Again, N second process parameter sets are randomly sampled from the N target process parameter neighborhoods without replacement, and the N second process parameter sets are identified using the fitness recognition network layer to obtain N second fitnesses;

[0077] Determine whether the N first fitnesses are greater than the N second fitnesses respectively, and if so, accept the N second process parameter sets as the optimal process parameter sets for the N stages according to a certain probability;

[0078] If not, the N second process parameter sets are used as the N-stage optimal process parameter sets;

[0079] Multiple iterations are performed based on the N optimal process parameter sets of the stages, and the N process parameter sets corresponding to the maximum fitness values ​​during the iterations are used as the N target optimal process parameter sets.

[0080] In one possible embodiment, the parameter optimization process for N process parameter sets of N dicing process lines is constrained according to N sets of adjustment tolerance thresholds, thereby improving optimization accuracy. The N target optimal process parameter sets are parameter sets that best match the actual dicing conditions of the N dicing process lines. Furthermore, the N target optimal process parameter sets are transmitted to N parameter control units of the N dicing process lines, respectively, to control the process parameters of the dicing process of the target chip.

[0081] In one embodiment, the N process parameter sets are randomly adjusted according to a preset adjustment method to obtain N initial process parameter neighborhoods, wherein the preset adjustment method is to randomly increase or decrease the parameters in the process parameter set. Out-of-specification parameters are eliminated from the N initial process parameter neighborhoods according to the N adjustment tolerance threshold sets, that is, parameters that do not meet the adjustment tolerance threshold sets are eliminated, thereby obtaining N target process parameter neighborhoods. The target process parameter neighborhoods are the ranges within which process control parameters in the N dicing process lines can be selected.

[0082] Then, N first process parameter sets are randomly extracted from the N target process parameter neighborhoods without replacement, and the N first process parameter sets are identified using a fitness recognition network layer to obtain N first fitnesses, wherein the N first fitnesses reflect the adaptability of the N first process parameter sets to the N dicing process production lines. Then, N second process parameter sets are randomly extracted from the N target process parameter neighborhoods without replacement, and the N second process parameter sets are identified using a fitness recognition network layer to obtain N second fitnesses. It is determined whether the N first fitnesses are greater than the N second fitnesses. If so, the N second process parameter sets are accepted as the optimal process parameter sets for the N stages with a certain probability, thereby avoiding falling into a local optimal solution. If not, the N second process parameter sets are used as the optimal process parameter sets for the N stages. Multiple iterations are then performed based on the N optimal process parameter sets for the N stages, and the N process parameter sets corresponding to the maximum fitness values ​​during the iterations are used as the N target optimal process parameter sets. Preferably, the convolutional neural network is supervised and trained by obtaining multiple sample process parameter sets and multiple sample fitnesses until the output reaches convergence, thereby obtaining the trained fitness recognition network layer, achieving the technical effect of intelligently identifying the fitness of the process parameter set.

[0083] In summary, the embodiments of the present application have at least the following technical effects:

[0084] This application analyzes multiple dicing process production lines for target chips from the perspectives of dicing quality and equipment reliability, determines the basis for adjusting the tolerance threshold set of a preset dicing process parameter set, and improves the accuracy of iterative optimization of process parameters. Furthermore, by optimizing N process parameter sets, N target optimal process parameter sets are obtained, and these are transmitted to N parameter control units of N dicing process production lines for process parameter control. This achieves the technical effect of improving the quality of process parameter control in chip production.

[0085] Example 2

[0086] Based on the same inventive concept as the process parameter control method for chip production in the aforementioned embodiment, as shown in FIG3 , the present application provides a process parameter control system for chip production. The system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0087] A process production line acquisition module 11 is used to obtain N dicing process production lines of the target chip;

[0088] A reliability factor generating module 12 is configured to traverse and collect information of the fully automatic dicing machines of the N dicing process production lines to perform reliability analysis and generate N equipment reliability factors;

[0089] The quality data set generation module 13 is used to generate N monitoring scribing quality data sets, wherein the N monitoring scribing quality data sets are obtained by extracting data from a set of scribing quality inspection reports of chips produced by N scribing process production lines within a preset monitoring time domain, using scribing quality as an index;

[0090] A deviation factor generating module 14 is configured to perform deviation analysis on the N monitoring dicing quality data sets to generate N quality deviation factors;

[0091] A process parameter set obtaining module 15 is configured to determine a preset dicing process parameter set based on the design information of the target chip;

[0092] a tolerance threshold obtaining module 16, configured to adjust the tolerance threshold set of the preset dicing process parameter set based on the N equipment reliability factors and the N quality deviation factors, respectively, to obtain N adjusted tolerance threshold sets;

[0093] An optimal process parameter set obtaining module 17 is configured to perform parameter optimization on N process parameter sets of N dicing process production lines based on N adjustment tolerance threshold sets to obtain N target optimal process parameter sets;

[0094] The process parameter control module 18 is used to transmit the N target optimal process parameter sets to N parameter control units of N dicing process production lines respectively for process parameter control.

[0095] Furthermore, the deviation factor generation module 14 is configured to perform the following steps:

[0096] Randomly select a monitoring dicing quality dataset from the N monitoring dicing quality datasets as a first monitoring dicing quality dataset;

[0097] Constructing a first offset analysis space based on the first monitoring scribing quality data set, wherein the first offset analysis space has a plurality of particle points, and the plurality of particle points correspond one-to-one to the monitoring scribing quality data in the first monitoring scribing quality data set;

[0098] performing an offset analysis based on the plurality of particle points and the first offset analysis space to generate a first mass offset factor;

[0099] N offset analysis spaces are generated according to the N monitoring dicing quality data sets, and after performing offset analysis, N quality offset factors are generated.

[0100] Furthermore, the deviation factor generation module 14 is configured to perform the following steps:

[0101] Randomly generating a plurality of guiding particles and a plurality of following particle sets from the plurality of particle points, wherein the following particles in the plurality of following particle sets are generated within a guiding distance threshold range of the plurality of guiding particles;

[0102] Calculating a plurality of guiding particle densities and a plurality of following particle density sets of the plurality of guiding particles and the plurality of following particle sets within a preset step size;

[0103] The inverse of the ratio of the multiple guiding particle densities of the multiple guiding particles to the sum of the multiple guiding particle densities is traversed and calculated, and a preset fine-tuning step size is multiplied according to the calculation result to obtain multiple fine-tuning step size sets of the multiple following particle density sets.

[0104] Furthermore, the deviation factor generation module 14 is configured to perform the following steps:

[0105] Performing N iterations in any direction on the multiple following particle sets according to the multiple fine-tuning step sets to obtain N iterative following particle sets;

[0106] Selecting N iterative following particle density sets, a plurality of guiding particle densities, and a plurality of following particle density sets corresponding to the N iterative following particle sets, and determining a particle with the largest particle density as a target particle;

[0107] The difference between the monitored dicing quality data and the preset dicing quality data corresponding to the target particle is calculated, and the ratio of the difference to the preset dicing quality data is calculated, and a first mass shift factor is generated according to the calculation result.

[0108] Furthermore, the reliability factor generation module 12 is configured to perform the following steps:

[0109] Using the design life as an index, data is extracted from the fully automatic dicing machine information of N dicing process production lines to obtain N design lives;

[0110] Using service life as an index, data is extracted from the fully automatic dicing machine information of N dicing process production lines to obtain N service lives;

[0111] Collect the number of fault repairs on N dicing process production lines and obtain N fault repair information;

[0112] The reliability identification network layer is used to identify N design life, N service life and N fault maintenance information to obtain N equipment reliability factors.

[0113] Furthermore, the optimal process parameter set obtaining module 17 is configured to perform the following steps:

[0114] Randomly adjusting the N process parameter sets according to a preset adjustment method to obtain N initial process parameter neighborhoods, wherein the preset adjustment method is to randomly increase or decrease the parameters in the process parameter set;

[0115] Eliminating out-of-specification parameters from the N initial process parameter neighborhoods based on the N adjustment tolerance threshold sets to obtain N target process parameter neighborhoods;

[0116] N first process parameter sets are randomly sampled without replacement from the N target process parameter neighborhoods, and the N first process parameter sets are identified using a fitness identification network layer to obtain N first fitnesses.

[0117] Furthermore, the optimal process parameter set obtaining module 17 is configured to perform the following steps:

[0118] Again, N second process parameter sets are randomly sampled from the N target process parameter neighborhoods without replacement, and the N second process parameter sets are identified using the fitness recognition network layer to obtain N second fitnesses;

[0119] Determine whether the N first fitnesses are greater than the N second fitnesses respectively, and if so, accept the N second process parameter sets as the optimal process parameter sets for the N stages according to a certain probability;

[0120] If not, the N second process parameter sets are used as the N-stage optimal process parameter sets;

[0121] Multiple iterations are performed based on the N optimal process parameter sets of the stages, and the N process parameter sets corresponding to the maximum fitness values ​​during the iterations are used as the N target optimal process parameter sets.

[0122] It should be noted that the above-mentioned order of the embodiments of the present application is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. The above description is of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0124] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. Process parameter control method for chip production, It is characterized in that The method comprises: Obtain N dicing process production lines of the target chip; Traversing and collecting the fully automatic dicing machine information of the N dicing process production lines to perform reliability analysis, and generating N equipment reliability factors; Generate N monitoring dicing quality data sets, wherein the N monitoring dicing quality data sets are obtained by extracting data from a set of dicing quality inspection reports of chips produced by N dicing process production lines within a preset monitoring time domain, with dicing quality as an index; Performing an offset analysis on the N monitoring dicing quality data sets to generate N quality offset factors; Determining a preset dicing process parameter set based on the design information of the target chip; Adjusting the tolerance threshold set of the preset dicing process parameter set based on the N equipment reliability factors and the N quality deviation factors respectively to obtain N adjusted tolerance threshold sets; Based on N adjustment tolerance threshold sets, N process parameter sets of N dicing process production lines are optimized to obtain N target optimal process parameter sets; The N target optimal process parameter sets are respectively transmitted to N parameter control units of N dicing process production lines for process parameter control.

2. The method according to claim 1, It is characterized in that The method comprises: Randomly select a monitoring dicing quality data set from the N monitoring dicing quality data sets as a first monitoring dicing quality data set; Constructing a first offset analysis space based on the first monitoring scribing quality data set, wherein the first offset analysis space has a plurality of particle points, and the plurality of particle points correspond one-to-one to the monitoring scribing quality data in the first monitoring scribing quality data set; Performing an offset analysis based on the plurality of particle points and the first offset analysis space to generate a first mass offset factor; N offset analysis spaces are generated according to the N monitoring dicing quality data sets, and after performing offset analysis, N quality offset factors are generated.

3. The method according to claim 2, It is characterized in that The method comprises: Randomly generate a plurality of guiding particles and a plurality of following particle sets from the plurality of particle points, wherein the following particles in the plurality of following particle sets are generated within a guiding distance threshold range of the plurality of guiding particles; Calculating a plurality of guiding particle densities and a plurality of following particle density sets of the plurality of guiding particles and the plurality of following particle sets within a preset step length; The inverse of the ratio of the multiple guiding particle densities of the multiple guiding particles to the sum of the multiple guiding particle densities is traversed and calculated, and a preset fine-tuning step is multiplied by the calculation result to obtain multiple fine-tuning step sets of the multiple following particle density sets.

4. The method according to claim 3, It is characterized in that The method comprises: According to the multiple fine-tuning step sets, the multiple follower particle sets are iterated N times in any direction to obtain N iterative follower particle sets; Selecting N iterative following particle density sets, a plurality of guiding particle densities, and a plurality of following particle density sets corresponding to the N iterative following particle sets, and determining a particle with the largest particle density as a target particle; The difference between the monitored dicing quality data and the preset dicing quality data corresponding to the target particle is calculated, and the ratio of the difference to the preset dicing quality data is calculated, and a first mass shift factor is generated according to the calculation result.

5. The method according to claim 1, It is characterized in that The method comprises: Taking the design life as the index, data is extracted from the fully automatic dicing machine information of N dicing process production lines to obtain N design lives; Taking the service life as an index, data is extracted from the fully automatic dicing machine information of N dicing process production lines to obtain N service lives; Collect the number of fault repairs of N dicing process production lines to obtain N fault repair information; The reliability identification network layer is used to identify N design life, N service life and N fault maintenance information to obtain N equipment reliability factors.

6. The method according to claim 1, It is characterized in that The method comprises: Randomly adjusting the N process parameter sets according to a preset adjustment method to obtain N initial process parameter neighborhoods, wherein the preset adjustment method is to randomly increase or decrease the parameters in the process parameter set; Eliminating out-of-specification parameters from the N initial process parameter neighborhoods based on the N adjustment tolerance threshold sets to obtain N target process parameter neighborhoods; N first process parameter sets are randomly selected without replacement from the N target process parameter neighborhoods, and the N first process parameter sets are identified using a fitness identification network layer to obtain N first fitnesses.

7. The method according to claim 6, It is characterized in that The method comprises: Again, randomly extracting N second process parameter sets from the N target process parameter neighborhoods without replacement, and identifying the N second process parameter sets using a fitness recognition network layer to obtain N second fitnesses; Determine whether the N first fitnesses are greater than the N second fitnesses respectively, and if so, accept the N second process parameter sets as the optimal process parameter sets for the N stages according to a certain probability; If not, the N second process parameter sets are used as the N-stage optimal process parameter sets; Multiple iterations are performed according to the optimal process parameter sets of the N stages, and the N process parameter sets corresponding to the maximum fitness values ​​in the iteration process are used as the N target optimal process parameter sets.

8. Process parameter control system for chip production, It is characterized in that The system comprises: A process production line acquisition module is used to obtain N dicing process production lines of the target chip; A reliability factor generation module, used for traversing and collecting the information of the fully automatic dicing machines of the N dicing process production lines to perform reliability analysis and generate N equipment reliability factors; A quality data set generation module is used to generate N monitoring dicing quality data sets, wherein the N monitoring dicing quality data sets are obtained by extracting data from a set of dicing quality inspection reports of chips produced by N dicing process production lines within a preset monitoring time domain, with dicing quality as an index; The deviation factor generation module is used to generate the N monitoring dicing quality data sets. Offset analysis, generating N mass offset factors; A process parameter set acquisition module, used to determine a preset dicing process parameter set based on the design information of the target chip; A tolerance threshold acquisition module, used to adjust the tolerance threshold set of the preset dicing process parameter set based on the N equipment reliability factors and the N quality deviation factors, respectively, to obtain N adjusted tolerance threshold sets; An optimal process parameter set acquisition module is used to optimize N process parameter sets of N dicing process production lines based on N adjustment tolerance threshold sets to obtain N target optimal process parameter sets; The process parameter control module is used to transmit the N target optimal process parameter sets to N parameter control units of N dicing process production lines respectively for process parameter control.

Images