Method for improving flatness of semiconductor chip board of dicing machine

Abstract

The present invention discloses a method for improving the flatness of a semiconductor chip board of a dicing machine, and relates to the field related to semiconductor device manufacturing. The method comprises: receiving a semiconductor chip board dicing requirement; configuring a blade type parameter through a knife mark requirement depth and a knife mark requirement width; predicting particle breakage for a first blade dicing parameter based on the blade type parameter through the chip board hardness and the dicing requirement path; predicting particle shedding for a first blade dicing parameter based on the blade type parameter through the chip board hardness and the dicing requirement path; when the particle breakage duration is consistent with the particle shedding duration, initializing the dicing machine according to the blade type parameter and the first blade dicing parameter. The method solves the technical problem that the existing semiconductor chip board dicing has a particle breakage problem affecting the chip board flatness, and achieves the technical effect of ensuring that the broken tool particles can fall off in time during the dicing process, thereby maintaining the flatness of the chip board.

Description

Technical Field

[0001] The present application relates to the field related to semiconductor device manufacturing, and in particular to a method for improving the flatness of a semiconductor chip board of a dicing machine. Background Art

[0002] In the semiconductor manufacturing industry, the dicing of chip boards is directly related to the chip yield, performance and reliability of the final product. Any slight dicing error may cause chip failure or performance degradation. In the dicing process, the cutter used by the dicing machine is one of the key tools. The particle state of the cutter directly affects the quality of dicing. The ideal dicing process requires that the cutter particles can remain sharp to ensure the flatness and accuracy of the dicing line.

[0003] However, in actual operation, the tool particles will wear and break during the long-term cutting process, which will not only reduce the cutting efficiency, but also cause unevenness on the chip board surface during the particle repair process (i.e., self-sharpening process), affecting the overall quality of the chip board. At present, the commonly used method in the industry is to deal with the wear and breakage of tool particles by regularly replacing the tool or adjusting the cutting parameters. However, this method ignores the impact on the flatness of the chip board during the critical time period from the breakage of the tool particles to the formation of new particles. During this time period, the broken particles may continue to cut, leaving uneven knife marks on the surface of the chip board, reducing the quality of the chip board.

[0004] In the current related technologies, there is a technical problem in the cutting of semiconductor chip boards, that is, the breakage of tool particles affects the flatness of the chip board.

[0005] The present application provides a method for improving the flatness of a semiconductor chip board of a dicing machine. The method receives the dicing requirements of the semiconductor chip board, including information such as the required depth and width of the knife mark, the hardness of the chip board and the dicing path, configures the blade type parameters based on this information, and predicts particle breakage and shedding for the initial preset blade dicing parameters. When the predicted particle breakage duration is consistent with the shedding duration, the dicing machine is initialized based on these parameters, and other technical means are used to achieve the technical effect of ensuring that the broken tool particles during the dicing process can fall off in time, thereby maintaining the flatness of the chip board.

[0006] The present application provides a method for improving the flatness of a semiconductor chip board of a dicing machine, comprising:

[0007] Receive a semiconductor chip board cutting requirement, wherein the semiconductor chip board cutting requirement includes a knife mark required depth, a knife mark required width, a chip board hardness and a cutting required path; configure a blade type parameter through the knife mark required depth and the knife mark required width; predict particle breakage for a first blade cutting parameter based on the blade type parameter through the chip board hardness and the cutting required path, and obtain a particle breakage duration, wherein the first blade cutting parameter is an initial preset blade cutting parameter; predict particle shedding for a first blade cutting parameter based on the blade type parameter through the chip board hardness and the cutting required path, and obtain a particle shedding duration; when the particle breakage duration is consistent with the particle shedding duration, initialize the dicing machine according to the blade type parameter and the first blade cutting parameter.

[0008] In a possible implementation, the blade type parameter is configured by the required depth of the knife mark and the required width of the knife mark, and the following processing is performed:

[0009] According to the required cut mark depth, the blade length is configured, wherein the blade length is greater than the required cut mark depth; according to the required cut mark width, the blade thickness is configured, wherein the blade thickness and the required cut mark width are less than the cut mark tolerance width; the particle binding degree interval is obtained, and the particle binding degree is obtained by random configuration; the blade length, the blade thickness and the particle binding degree are set as the blade type parameters.

[0010] In a possible implementation, the chip board hardness and the cutting requirement path are used to predict particle breakage for the first blade cutting parameter based on the blade type parameter to obtain the particle breakage duration, and the following processing is performed:

[0011] Through the chip board hardness and the cutting requirement path, adjacent samples are collected for the blade type parameters and the first blade cutting parameters to obtain a first adjacent sampling result; the first adjacent sampling result is traversed to perform adjacent sample collection to obtain a second adjacent sampling result; until the N-1th adjacent sampling result is traversed to perform adjacent sample collection to obtain the Nth adjacent sampling result, 15≥N≥2; particle rupture prediction is performed based on the first adjacent sampling result, the second adjacent sampling result until the Nth adjacent sampling result to obtain the particle rupture duration.

[0012] In a possible implementation, particle rupture prediction is performed according to the first adjacent sampling result, the second adjacent sampling result, and up to the Nth adjacent sampling result to obtain the particle rupture duration, and the following processing is performed:

[0013] Step 1: Extract a first group of adjacent sampling results according to the Nth adjacent sampling result, wherein the first group of adjacent sampling results belongs to a first parent node; Step 2: Perform particle rupture duration mode statistics on the first group of adjacent sampling results to obtain a first particle rupture aggregation duration; Step 3: Add the first particle rupture aggregation duration to the first parent node, and delete the first group of adjacent sampling results from the Nth adjacent sampling result; Execute steps 1 to 3 from the Nth adjacent sampling result to the first adjacent sampling result until only a single particle rupture aggregation duration remains, which is set as the particle rupture duration.

[0014] In a possible implementation, when the particle rupture duration is consistent with the particle shedding duration, the following processing is performed:

[0015] When the duration deviation between the particle rupture duration and the particle shedding duration is greater than or equal to the duration deviation threshold, the particle rupture duration is inconsistent with the particle shedding duration; when the duration deviation between the particle rupture duration and the particle shedding duration is less than the duration deviation threshold, the particle rupture duration is consistent with the particle shedding duration.

[0016] In a possible implementation, when the time deviation between the particle rupture duration and the particle shedding duration is greater than or equal to a time deviation threshold, and the particle rupture duration is inconsistent with the particle shedding duration, the following processing is performed:

[0017] Adjust the first blade cutting parameter and the first particle binding degree of the blade type parameter to obtain the second blade cutting parameter and the second particle binding degree; predict particle breakage based on the blade type parameter according to the second blade cutting parameter and the second particle binding degree to obtain an updated particle breakage time; predict particle shedding based on the blade type parameter according to the second blade cutting parameter and the second particle binding degree to obtain an updated particle shedding time; when the updated particle breakage time is consistent with the updated particle shedding time, initialize the dicing machine according to the second blade cutting parameter, the second particle binding degree and the blade type parameter.

[0018] In a possible implementation, the following processing is performed:

[0019] After the blade cutting parameters and the blade type parameters are updated k times, a blade control parameter particle space is constructed, k≥25; the blade control parameters are directed updated according to the blade control parameter particle space to obtain extended blade cutting parameters and extended blade type parameters; the particle rupture time and the particle shedding time are set as the convergence condition; when the extended blade cutting parameters and the extended blade type parameters meet the convergence condition, the dicing machine is initialized according to the extended blade cutting parameters and the extended blade type parameters.

[0020] In a possible implementation, the blade control parameter is updated in a directional manner according to the blade control parameter particle space to obtain the extended blade cutting parameter and the extended blade type parameter, and the following processing is performed:

[0021] According to the blade control parameter particle space, a target blade control parameter particle with a smaller time deviation is selected; according to the blade control parameter particle space, a reference blade control parameter particle with a larger time deviation is selected; according to the target blade control parameter particle, the reference blade control parameter particle is guided to adjust, and the extended blade cutting parameter and the extended blade type parameter are generated.

[0022] The method for improving the flatness of a semiconductor chip board of a dicing machine proposed in this application first receives a semiconductor chip board dicing requirement, wherein the semiconductor chip board dicing requirement includes a knife mark required depth, a knife mark required width, a chip board hardness and a dicing required path, then configures a blade type parameter through the knife mark required depth and the knife mark required width, and then predicts particle breakage for the first blade dicing parameter based on the blade type parameter through the chip board hardness and the dicing required path, and obtains a particle breakage duration, wherein the first blade dicing parameter is an initial preset blade dicing parameter, and at the same time, predicts particle shedding for the first blade dicing parameter based on the blade type parameter through the chip board hardness and the dicing required path, and obtains a particle shedding duration, when the particle rupture duration is consistent with the particle shedding duration, the dicing machine is initialized according to the blade type parameter and the first blade dicing parameter, thereby achieving the technical effect of ensuring that the broken tool particles during the dicing process can fall off in time, thereby maintaining the flatness of the chip board. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the method according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0024] Figure 1 A schematic flow chart of a method for improving the flatness of a semiconductor chip board of a dicing machine provided in an embodiment of the present application.

[0025] Figure 2 A schematic diagram of the process of obtaining the particle rupture time in the method for improving the flatness of a semiconductor chip board of a dicing machine provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] 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 in accordance with the contents of the specification. 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 listed below.

[0027] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0028] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations 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 inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0029] The present application embodiment provides a method for improving the flatness of a semiconductor chip board of a dicing machine, such as Figure 1 As shown, the method includes:

[0030] Step S100, receiving the semiconductor chip board dicing requirements, wherein the semiconductor chip board dicing requirements include the required knife mark depth, required knife mark width, chip board hardness and required dicing path. Specifically, the user or system inputs the semiconductor chip board dicing requirements to the dicing machine, including the required knife mark depth, required knife mark width, chip board hardness and required dicing path. Among them, the required knife mark depth refers to the depth that needs to be reached during the dicing process, which affects the cutting effect and chip performance; the required knife mark width refers to the width of the knife mark left after cutting, which affects the flatness and appearance of the chip board; the chip board hardness refers to the physical hardness of the semiconductor chip board, which affects the difficulty of cutting and tool wear; the dicing required path refers to the specific path or pattern on the chip board that needs to be cut, which determines the shape and layout of the final chip.

[0031] Step S200, configuring blade type parameters according to the required depth of the knife mark and the required width of the knife mark. Specifically, the blade type parameters include blade length, blade thickness, and particle bonding, etc., which are used to describe the physical properties and cutting performance of the tool. Among them, the length of the blade is configured according to the required depth of the knife mark to ensure the cutting depth; the thickness of the blade is configured according to the required width of the knife mark to ensure that the cutting width meets the requirements; based on empirical data or experimental results, a suitable particle bonding is selected to balance the wear resistance and self-sharpening of the tool. The blade type parameters are used to configure the cutting parameters.

[0032] In a possible implementation, the blade type parameters are configured through the required depth of the cutter mark and the required width of the cutter mark, and step S200 further includes step S210, configuring the blade length according to the required depth of the cutter mark, wherein the blade length is greater than the required depth of the cutter mark. Specifically, the required depth of the cutter mark refers to the specific requirements of the customer or process requirements for the depth of the trace left on the chip board after dicing, which determines the depth of the trace left on the chip board after the final dicing; the blade length is the size of the tool perpendicular to the dicing direction, which determines the maximum depth that the tool can penetrate the chip board. In order to ensure the dicing effect, the blade length must be greater than the required depth of the cutter mark to ensure that the blade can completely penetrate the chip board to the required depth and leave enough margin to deal with possible wear or incomplete penetration. Step S220, configuring the blade thickness according to the required width of the cutter mark, wherein the blade thickness and the required width of the cutter mark are less than the tolerance width of the cutter mark. Specifically, the required width of the cutter mark is the specific requirements of the customer or process for the width of the cutter mark after dicing. The configuration of blade thickness is based on the required width of the cut. The blade thickness is the size of the tool in the cutting direction, which determines the width of the cut after cutting. In order to ensure that the width of the cut after cutting meets the requirements, and taking into account the process tolerance (i.e., allowing a certain error range), the blade thickness will be slightly smaller than the required width of the cut plus a preset cut tolerance width (a preset error range), which is used to ensure both the cutting accuracy and the cutting quality problems caused by the tool being too wide. Step S230, obtain the particle bonding interval, randomly configure, and obtain the particle bonding. Specifically, the particle bonding is an indicator of the bonding strength between the tool particles and the matrix, which directly affects the wear resistance, self-sharpening and service life of the tool. Determine a reasonable particle bonding interval based on experience and experimental data, and then randomly select a value in this interval as the particle bonding of the current tool. Step S240, set the blade length, the blade thickness and the particle bonding as the blade type parameters. Specifically, after completing the configuration of blade length, blade thickness and particle bonding, these three parameters are integrated into blade type parameters. This implementation method customizes the parameters of the blade according to specific cutting requirements, achieving the technical effect of maximizing the cutting accuracy and flatness of the chip board.

[0033] Step S300, using the chip board hardness and the cutting demand path, based on the blade type parameter, predicting particle fracture for the first blade cutting parameter, and obtaining particle fracture duration, wherein the first blade cutting parameter is an initial preset blade cutting parameter. Specifically, using a particle fracture prediction model based on the chip board hardness and the cutting demand path, the first blade cutting parameter is simulated and calculated, and according to the simulation results, the expected fracture duration of the tool particles during the cutting process is calculated.

[0034] like Figure 2 As shown, in a possible implementation, through the chip board hardness and the cutting requirement path, based on the blade type parameter, the first blade cutting parameter is predicted for particle breakage, and the particle breakage duration is obtained. Step S300 further includes step S310, through the chip board hardness and the cutting requirement path, the blade type parameter and the first blade cutting parameter are adjacently sampled to obtain a first adjacent sampling result. Specifically, according to the current chip board hardness and cutting requirement path, combined with the configured blade type parameter and the first blade cutting parameter (i.e., the initial preset cutting parameter), similar or close cutting cases are searched in the database or historical records, and the parameters of these cases (such as hardness, path, blade type, cutting parameters, etc.) should be consistent with the parameters of the current task or have little deviation. Through comparative analysis, the most satisfactory groups of adjacent samples are selected to form a first adjacent sampling result. Step S320, traverse the first adjacent sampling result to collect adjacent samples and obtain a second adjacent sampling result. Specifically, based on the first adjacent sampling result, these samples are further traversed, and adjacent samples are collected again for each sample. Through this process, the second adjacent sampling result is obtained. Step S330, until the N-1th adjacent sampling result is traversed to collect adjacent samples, and the Nth adjacent sampling result is obtained, 15≥N≥2. Specifically, the above process (step S320) is repeated multiple times until the preset sampling level N is reached. Each traversal performs a wider level of adjacent sample collection based on the previous result, where the value range of N is set to 2 to 15, indicating that a maximum of 15 adjacent sample collection iterations can be performed. Step S340, according to the first adjacent sampling result, the second adjacent sampling result until the Nth adjacent sampling result, the particle rupture prediction is performed to obtain the particle rupture duration. Specifically, after completing multiple adjacent sample collections, all the collected adjacent sampling results (from the first to the Nth adjacent sampling results) are used as input data, and the particle rupture duration is predicted using a statistical model, and the predicted particle rupture duration is output. This implementation method makes full use of existing data and empirical knowledge through multiple adjacent sample collection and iterative analysis, reduces the impact of randomness and uncertainty on the prediction results, and improves the accuracy and reliability of particle rupture time prediction.

[0035] In a possible implementation, particle rupture prediction is performed based on the first adjacent sampling result, the second adjacent sampling result, and up to the Nth adjacent sampling result to obtain the particle rupture duration, and step S340 further includes step 1: extracting a first group of adjacent sampling results based on the Nth adjacent sampling result, wherein the first group of adjacent sampling results belongs to a first parent node; step 2: performing particle rupture duration mode statistics on the first group of adjacent sampling results to obtain a first particle rupture aggregation duration; step 3: adding the first particle rupture aggregation duration to the first parent node, and deleting the first group of adjacent sampling results from the Nth adjacent sampling result; and executing steps 1 to 3 from the Nth adjacent sampling result to the first adjacent sampling result until only a single particle rupture aggregation duration remains, which is set as the particle rupture duration.

[0036] Specifically, starting from the Nth adjacent sampling result (i.e., the sampling result obtained from the last iteration), the sampling results are grouped, and each group of sampling results is regarded as belonging to a parent node. The first group of adjacent sampling results is the set of sampling results belonging to the first parent node. For the first group of adjacent sampling results, the particle rupture duration in each sampling result is counted, and the duration with the most occurrences (i.e., the mode) is found. This mode is regarded as the particle rupture aggregation duration of the group of sampling results, representing the typical value or central trend of the particle rupture duration in the group of sampling results. After determining the first particle rupture aggregation duration, it is associated with the first parent node, and the first group of adjacent sampling results is removed from the Nth adjacent sampling result to reduce the amount of data to be processed subsequently. Subsequently, step one and step two are continued to be performed according to the first particle rupture aggregation duration and the N-1th adjacent sampling result, but this time the operation is performed for the second parent node (i.e., the next group of adjacent sampling results). This process will be repeated until all adjacent sampling results have been processed and only one unique particle rupture aggregation duration remains, which is set as the final particle rupture duration. This implementation method utilizes the concept of mode in statistics. It reduces the impact of noise and inconsistency by finding the duration with the highest number of occurrences. It gradually narrows the prediction range through layer-by-layer aggregation and screening, thereby improving the accuracy and stability of particle rupture duration prediction.

[0037] Step S400, using the chip board hardness and the cutting requirement path, based on the blade type parameter, the first blade cutting parameter is predicted to have a particle shedding time. Specifically, the particle shedding time prediction method is similar to the particle rupture time prediction method, which will not be described here.

[0038] Step S500, when the particle rupture time is consistent with the particle shedding time, the dicing machine is initialized according to the blade type parameter and the first blade cutting parameter. Specifically, when the particle rupture time is consistent with or close to the particle shedding time, it is considered that the currently configured blade type and blade cutting parameters can meet the flatness requirements. If the prediction result does not meet expectations, adjust the blade type parameters or blade cutting parameters according to the prediction data, and re-predict. Using the finalized blade type parameters and blade cutting parameters, the dicing machine is initialized, including tool installation, speed adjustment, path planning, etc. After confirming that all settings are correct, start the dicing machine to perform chip board dicing operations. The embodiment of the present application adopts technical means such as receiving the cutting requirements of the semiconductor chip board, including the required depth and width of the knife mark, the hardness of the chip board and the cutting path, configuring the blade type parameters based on this information, and predicting the particle breakage and shedding of the initial preset blade cutting parameters. When the predicted particle breakage time is consistent with the shedding time, the dicing machine is initialized according to these parameters, thereby achieving the technical effect of ensuring that the broken tool particles during the cutting process can fall off in time, thereby maintaining the flatness of the chip board.

[0039] In one possible implementation, when the particle rupture duration is consistent with the particle shedding duration, step S500 further includes step S510, when the duration deviation between the particle rupture duration and the particle shedding duration is greater than or equal to a duration deviation threshold, the particle rupture duration is inconsistent with the particle shedding duration; step S520, when the duration deviation between the particle rupture duration and the particle shedding duration is less than the duration deviation threshold, the particle rupture duration is consistent with the particle shedding duration.

[0040] Specifically, the time deviation between the particle rupture time and the particle shedding time is calculated, and the time deviation is obtained by subtracting the two time lengths and taking the absolute value. This time deviation is compared with a preset time deviation threshold (a preset threshold determined based on experience, experimental data or process requirements, used to determine whether the particle rupture time and the particle shedding time are close enough). If the time deviation is greater than or equal to the threshold, it is determined that the particle rupture time is inconsistent with the particle shedding time. If the comparison result in step S510 shows that the time deviation is less than the time deviation threshold, then it is determined that the particle rupture time is consistent with the particle shedding time. In this case, the dicing machine is initialized according to the previously determined blade type parameters and the first blade cutting parameters to ensure that the rupture and shedding of the tool particles can be coordinated during the dicing process, thereby maintaining the flatness of the chip board. This implementation method introduces duration deviation and duration deviation threshold to make a more detailed comparison and analysis of particle rupture duration and particle detachment duration, allowing the system to tolerate deviations within a certain range, and thus more flexibly judge whether two durations are close enough, achieving the technical effect of improving the accuracy and practicality of the prediction.

[0041] In one possible implementation, when the time deviation between the particle rupture time and the particle shedding time is greater than or equal to a time deviation threshold, the particle rupture time is inconsistent with the particle shedding time, and step S510 further includes step S511, adjusting the first blade cutting parameter and the first particle binding degree of the blade type parameter to obtain the second blade cutting parameter and the second particle binding degree. Specifically, if the time deviation between the particle rupture time and the particle shedding time is greater than or equal to a time deviation threshold, it is necessary to adjust the blade cutting parameter and the particle binding degree to improve this situation. At this time, based on the current first blade cutting parameter and the first particle binding degree of the blade type parameter, a new second blade cutting parameter and the second particle binding degree are calculated through an adjustment strategy guided by experimental data. Step S512, according to the second blade cutting parameter and the second particle bonding degree, based on the blade type parameter, predict particle breakage, and obtain an updated particle breakage duration; step S513, according to the second blade cutting parameter and the second particle bonding degree, based on the blade type parameter, predict particle shedding, and obtain an updated particle shedding duration. Specifically, after obtaining the second blade cutting parameter and the second particle bonding degree, use these new parameters and bonding degrees, based on the blade type parameter, re-predict particle breakage and particle shedding, and evaluate whether the adjusted blade cutting parameter and particle bonding degree can effectively improve the particle breakage duration, making it closer to the particle shedding duration. Step S514, when the updated particle breakage duration is consistent with the updated particle shedding duration, initialize the dicing machine according to the second blade cutting parameter, the second particle bonding degree, and the blade type parameter. Specifically, if the updated particle breakage duration is consistent with the updated particle shedding duration, then initialize the dicing machine according to the second blade cutting parameter, the second particle bonding degree, and the blade type parameter. In this way, the breakage and shedding of tool particles can be better coordinated during the dicing process, thereby ensuring the flatness of the chip board. This implementation method optimizes the breakage and shedding behavior of tool particles by predicting and adjusting the duration of particle breakage and shedding, thereby improving the flatness of the chip board.

[0042] In a possible implementation, step S510 further includes step S515, after the blade cutting parameters and blade type parameters are updated k times, constructing a blade control parameter particle space, k ≥ 25. Specifically, based on the previous multiple adjustments (at least 25 times) to the blade cutting parameters and blade type parameters, construct a particle space containing multiple possible parameter combinations. This particle space is a space composed of parameter set combinations that appeared in the previous iteration process, and is used to find parameter combinations that meet the convergence conditions during the search process. Step S516, according to the blade control parameter particle space, perform a directed update of the blade control parameters to obtain extended blade cutting parameters and extended blade type parameters. Specifically, after the particle space is constructed, the parameters in the particle space are directed updated according to a certain strategy to generate new particles, so as to find a parameter combination that is closer to the ideal state in the particle space. Step S517, setting the particle rupture time and the particle shedding time to be consistent as the convergence condition; Step S518, when the extended blade cutting parameters and the extended blade type parameters meet the convergence condition, the dicing machine is initialized according to the extended blade cutting parameters and the extended blade type parameters. Specifically, the particle rupture time and the particle shedding time are set to be consistent as the convergence condition of the algorithm. Only when the algorithm finds a set of parameters under which the particle rupture and shedding occur almost simultaneously, the algorithm will stop searching and consider that the optimal solution has been found. When the parameter combination after the directional update meets the convergence condition, the dicing machine is initialized according to this set of parameters. In this way, during the dicing process, the tool particles will fall off quickly after rupture, thereby reducing the impact on the flatness of the chip board and improving the dicing quality. This implementation method fully explores the parameter space by constructing a particle space and performing directional updates, achieving a technical effect that helps to find a more accurate and stable parameter combination.

[0043] In a possible implementation, the blade control parameters are updated in a directional manner according to the blade control parameter particle space to obtain the extended blade cutting parameters and the extended blade type parameters. Step S516 further includes step S5161, selecting the target blade control parameter particles with a smaller time deviation according to the blade control parameter particle space. Specifically, after the blade control parameter particle space is constructed, the deviation between the particle rupture time and the particle shedding time corresponding to each particle in the space is evaluated, and the particles with a smaller time deviation are screened out as the target blade control parameter particles. The target blade control parameter particles represent a better parameter combination in the current search space, and their particle rupture and shedding behaviors are closer to the ideal state, which can better balance the particle rupture and shedding process, thereby reducing the impact on the flatness of the chip board. Step S5162, according to the blade control parameter particle space, select the reference blade control parameter particles with a larger time deviation. Specifically, a particle with a relatively large time deviation is selected as a reference blade control parameter particle. Although the performance of the reference blade control parameter particle is not as good as that of the target blade control parameter particle, it is widely distributed in the parameter space and represents different search directions, which helps to introduce more diversity in the subsequent parameter update process and avoid falling into a local optimal solution. Step S5163, guide the adjustment of the reference blade control parameter particle according to the target blade control parameter particle, and generate the extended blade cutting parameter and the extended blade type parameter. Specifically, guide the adjustment of the reference blade control parameter particle according to the characteristics of the target blade control parameter particle, that is, fine-tune some parameters of the reference particle so that its performance gradually approaches that of the target particle, and gradually improve the performance of the reference particle in an iterative manner. By calculating the parameter difference between the target particle and the reference particle, and generating a new parameter combination according to these differences, namely the extended blade cutting parameter and the extended blade type parameter. This implementation method selects target particles with smaller duration deviation as guides, ensuring that the direction of parameter update is towards a more optimal solution. Selecting reference particles with larger duration deviation as adjustment objects helps to introduce more diversity in the search process and prevent the algorithm from falling into a local optimal solution, achieving the technical effect of increasing the possibility of finding the optimal solution while ensuring search efficiency.

[0044] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does 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.

Claims

1. A method for improving the flatness of a semiconductor chip board of a dicing machine, characterized in that: include: Receiving semiconductor chip board dicing requirements, wherein the semiconductor chip board dicing requirements include a required cut depth, a required cut width, a chip board hardness, and a required dicing path; Configure blade type parameters according to the required depth of the knife mark and the required width of the knife mark; Through the chip board hardness and the cutting requirement path, based on the blade type parameter, the particle fracture prediction is performed on the first blade cutting parameter to obtain the particle fracture duration, wherein the first blade cutting parameter is the initial preset blade cutting parameter; According to the chip board hardness and the cutting requirement path, based on the blade type parameter, the particle shedding prediction is performed on the first blade cutting parameter to obtain the particle shedding time; When the particle rupture time is consistent with the particle shedding time, initializing the dicing machine according to the blade type parameter and the first blade dicing parameter; According to the chip board hardness and the cutting requirement path, based on the blade type parameter, the particle breakage prediction is performed on the first blade cutting parameter to obtain the particle breakage duration, including: According to the chip board hardness and the cutting requirement path, adjacent sample collection is performed on the blade type parameter and the first blade cutting parameter to obtain a first adjacent sampling result; Traversing the first adjacent sampling result to collect adjacent samples, and obtaining a second adjacent sampling result; Until the N-1th adjacent sampling result is traversed to collect adjacent samples, the Nth adjacent sampling result is obtained, 15≥N≥2; Predicting particle rupture according to the first adjacent sampling result, the second adjacent sampling result, and up to the Nth adjacent sampling result to obtain the particle rupture duration; The particle rupture prediction is performed according to the first adjacent sampling result, the second adjacent sampling result, and up to the Nth adjacent sampling result to obtain the particle rupture duration, including: Step 1: extracting a first group of adjacent sampling results according to the Nth adjacent sampling result, wherein the first group of adjacent sampling results belongs to the first parent node; Step 2: performing mode statistics of particle rupture duration on the first group of adjacent sampling results to obtain the first particle rupture aggregation duration; Step 3: adding the first particle fracture aggregation time to the first parent node, and deleting the first group of adjacent sampling results from the Nth adjacent sampling result; Execute steps 1 to 3 from the Nth adjacent sampling result to the first adjacent sampling result until only a single particle rupture aggregation time is left, which is set as the particle rupture time; By analogy, based on the blade type parameter, the particle shedding prediction is performed on the first blade cutting parameter to obtain the particle shedding time.

2. The method for improving the flatness of a semiconductor chip board of a dicing machine according to claim 1, characterized in that: The blade type parameters are configured according to the required depth of the knife mark and the required width of the knife mark, including: According to the required depth of the knife mark, the blade length is configured, wherein the blade length is greater than the required depth of the knife mark; According to the required width of the knife mark, the blade thickness is configured, wherein the blade thickness and the required width of the knife mark are less than the tolerance width of the knife mark; Obtain particle binding degree interval, randomly configure, and obtain particle binding degree; The blade length, the blade thickness and the particle bonding degree are set as the blade type parameters.

3. The method for improving the flatness of a semiconductor chip board of a dicing machine according to claim 1, characterized in that: When the particle rupture duration is consistent with the particle shedding duration, it includes: When the time deviation between the particle rupture time and the particle shedding time is greater than or equal to the time deviation threshold, the particle rupture time is inconsistent with the particle shedding time; When the duration deviation between the particle rupture duration and the particle shedding duration is less than the duration deviation threshold, the particle rupture duration is consistent with the particle shedding duration.

4. The method for improving the flatness of a semiconductor chip board of a dicing machine as claimed in claim 3, characterized in that: When the time deviation between the particle rupture duration and the particle shedding duration is greater than or equal to the time deviation threshold, the particle rupture duration is inconsistent with the particle shedding duration, and then the method includes: Adjusting the first blade cutting parameter and the first particle bonding degree of the blade type parameter to obtain a second blade cutting parameter and a second particle bonding degree; According to the second blade cutting parameter and the second particle bonding degree, particle breakage prediction is performed based on the blade type parameter to obtain an updated particle breakage duration; According to the second blade cutting parameter and the second particle bonding degree, particle shedding prediction is performed based on the blade type parameter to obtain an updated particle shedding time; When the duration of the renewal particle rupture is consistent with the duration of the renewal particle shedding, the dicing machine is initialized according to the second blade cutting parameter, the second particle bonding degree, and the blade type parameter.

5. The method for improving the flatness of a semiconductor chip board of a dicing machine as claimed in claim 4, characterized in that: Also includes: After the blade cutting parameters and blade type parameters are updated k times, the blade control parameter particle space is constructed, k ≥ 25; Performing a directional update of blade control parameters according to the blade control parameter particle space to obtain an extended blade cutting parameter and an extended blade type parameter; The particle rupture time and the particle shedding time are consistent as a convergence condition; When the extended blade dicing parameters and the extended blade type parameters meet the convergence condition, the dicing machine is initialized according to the extended blade dicing parameters and the extended blade type parameters.

6. The method for improving the flatness of a semiconductor chip board of a dicing machine as claimed in claim 5, characterized in that: According to the blade control parameter particle space, the blade control parameter is updated in a directional manner to obtain the extended blade cutting parameter and the extended blade type parameter, including: According to the blade control parameter particle space, a target blade control parameter particle with a smaller time deviation is selected; According to the blade control parameter particle space, a reference blade control parameter particle with a large time deviation is selected; The reference blade control parameter particles are guided to adjust according to the target blade control parameter particles, so as to generate the extended blade cutting parameters and the extended blade type parameters.

Images