A heat sink preparation optimization method, device and readable storage medium

Abstract

The present invention provides a heat sink preparation optimization method, device and readable storage medium. The method makes the process parameters of the process control model corresponding to at least two adjacent deposition layers in the same heat sink different. The process parameters of the process control model corresponding to each deposition layer formed during the heat sink generation process and the performance parameters of the heat sink are collected to generate multiple groups of samples. The models corresponding to different deposition layers are trained through samples and qualified judgment of the samples, so as to improve each process control model to optimize the performance of each deposition layer, and finally achieve the purpose of optimizing the comprehensive performance of the heat sink. The scheme proposed by the present invention is suitable for the preparation of heat sinks of various thicknesses, which can effectively avoid the defects and uneven performance problems in the production of heat sinks, and can reflect the effects of diversification and differentiation in each deposition layer of the heat sink, so that different deposition layers can have advantages of different directions, thereby ensuring that the final generated heat sink has better comprehensive performance.

Description

Technical Field

[0001] The present invention belongs to the technical field of heat sink preparation, and in particular relates to a heat sink preparation optimization method, device and readable storage medium. Background Art

[0002] As the power density of electronic devices continues to increase, heat dissipation has become an important issue in the design of modern electronic products, especially for high-power, high-heat flux integrated circuits (ICs) and other heat-sensitive components. Heat sinks, as an important heat dissipation component, are widely used in various electronic devices. Their performance directly affects the stability and life of the equipment.

[0003] At present, there are various methods for preparing heat sinks. Common technical means include mechanical processing, surface coating treatment and thin film deposition technology. Among these methods, sputtering deposition technology is widely used in the strengthening treatment of heat sink surfaces due to its advantages such as high precision and high uniformity. Sputtering deposition uses high-energy particles to bombard the target material in a vacuum environment, and the target atoms or molecules are impacted onto the substrate surface to form a dense film. This method has significant advantages in the preparation of heat sinks, especially for the requirements of film uniformity and high density. It provides better technical support.

[0004] However, the traditional sputtering deposition method is generally only suitable for the preparation of a single-layer film. When preparing a thicker heat sink, the fixed sputtering deposition method easily leads to unstable deposition layer thickness and high surface roughness, and it is not easy to ensure the quality of deposition. In addition, the thicker the heat sink to be prepared, the more likely the shortcomings of the deposition layer formed by the fixed sputtering deposition method will be magnified, which will cause the final heat sink to have problems such as poor density, reduced adhesion, reduced electrical and thermal conductivity, uneven performance, and even partial detachment of the deposition layer. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention proposes a heat sink preparation optimization method, the method comprising:

[0006] Pre-processing the substrate, forming a deposition layer on the surface of the processed substrate by sputtering, and sequentially forming several deposition layers on the surface of the top deposition layer by sputtering again, until the total thickness of each deposition layer reaches a preset thickness, the heat sink is generated; each deposition layer corresponds to an initial process control model, and in the process of forming each deposition layer, the process parameters of each process control model are obtained at a preset sampling frequency to be stored in a database, and the performance parameters of each heat sink are obtained and stored in the database, so as to construct a group of samples based on the performance parameters of each heat sink and the process parameters corresponding to each different deposition layer; the process parameters of the process control model corresponding to at least two adjacent deposition layers are different;

[0007] Determine a heat sink whose performance parameters meet preset requirements as a qualified heat sink, and determine a sample containing at least a preset proportion of qualified heat sinks as a qualified sample; adjust the process parameters to generate multiple groups of heat sinks accordingly and obtain the process parameters and performance parameters of each group of heat sinks respectively and store them in the database, so as to construct a sample of not less than a first preset number, and make the number of qualified samples contained in the sample not less than a second preset number;

[0008] Several groups of high-quality samples are expanded based on any one or several groups of qualified samples, and it is determined whether each of the high-quality samples is a qualified sample, so as to train the process control models corresponding to different deposition layers based on each group of samples and each group of high-quality samples, so as to improve the process parameters of each process control model to optimize the performance of each deposition layer, and further optimize the overall performance of the heat sink.

[0009] Specifically, the method of training the process control model includes:

[0010] Based on the data of the performance parameters corresponding to each of the samples and the process parameters of each process control model corresponding to each of the samples in the database, a plurality of groups of potential associated data of the same size are divided and distributed to a plurality of computing nodes through a mapping function, so as to calculate on each of the computing nodes to obtain associated data subsets corresponding to each of the potential associated data; each of the associated data subsets includes a plurality of to-be-determined associated sets;

[0011] Merging each of the associated data subsets into a candidate associated union set, calculating the support of each to-be-determined associated set in the candidate associated union set, and generating an associated data union set based on the to-be-determined associated sets whose support is not less than the minimum support threshold;

[0012] Based on each to-be-determined association set in the union of the association data, several non-empty subsets are generated to form a candidate association rule based on each of the non-empty subsets, and the confidence of each of the candidate association rules is calculated. The candidate association rules whose confidence is not less than a minimum confidence threshold are determined as association rules that can be used to determine process parameters and are retained to complete the training of each of the process control models corresponding to different deposition layers.

[0013] Specifically, the method of calculating and obtaining a subset of associated data corresponding to a potential associated data includes:

[0014] Scan the potential association data to obtain a univariate frequent item set whose support is not less than the minimum support threshold, and obtain a first association set to be determined by self-connecting the univariate frequent item set;

[0015] Scan the potential association data again to calculate the support of the first association set to be determined, so as to obtain a binary frequent item set whose support is not less than the minimum support threshold, and obtain a second association set to be determined by self-connecting the binary frequent item set;

[0016] Repeatedly iteratively scan the potential associated data to calculate the support of the currently obtained association set to be determined, so as to obtain a new frequent item set whose support is not less than the minimum support threshold, and obtain a new association set to be determined by self-connecting the latest frequent item set, until the latest frequent item set fails to meet the minimum support requirement, and then generate an associated data subset corresponding to the potential associated data based on each association set to be determined.

[0017] Furthermore, the method further comprises:

[0018] Before scanning the potential associated data, remove data whose support is less than the minimum support threshold in the potential associated data;

[0019] When calculating the support of the currently obtained association set to be determined and obtaining new frequent item sets whose support is not less than the minimum support threshold, the new frequent item sets whose support is less than the minimum support threshold are eliminated.

[0020] Preferably, the “scanning the potential associated data to obtain unary frequent itemsets whose support is not less than the minimum support threshold” includes:

[0021] Scan the potential associated data, generate a column data table based on the potential associated data, scan the column data table to obtain univariate frequent item sets, calculate the support of each of the univariate frequent item sets to obtain univariate frequent item sets whose support is not less than the minimum support threshold.

[0022] Specifically, the “obtaining the process parameters of each process control model at a preset sampling frequency during the process of forming each deposition layer to store them in a database” includes:

[0023] When a certain deposition layer begins to form, determining a sampling time for first acquiring process parameters, and after first acquiring the process parameters, acquiring the process parameters of each of the process control models based on the preset sampling frequency;

[0024] After the deposition layer is completely formed, the obtained process parameters and their corresponding sampling times are stored in the sub-database corresponding to the process control model in the database; any process control model is trained by the performance parameters corresponding to the sample in the database and the process parameters corresponding to the sample in the sub-database corresponding to the process control model.

[0025] Preferably, the method further comprises:

[0026] The process parameters of any two adjacent deposition layers in the same heat sink determined based on each of the process control models are different;

[0027] Alternatively, process parameters of any two deposition layers in the same heat sink determined based on the process control models are different.

[0028] Furthermore, the “expanding several groups of high-quality samples based on any one or several groups of qualified samples” includes:

[0029] For any group of qualified samples, a total of N process parameters corresponding to M deposition layers in the qualified samples are adjusted, and several groups of heat sinks are generated accordingly, and process parameters and performance parameters are obtained and stored in the database to expand several groups of high-quality samples; M is not less than one and not greater than a third preset number, and N is not less than one and not greater than a fourth preset number.

[0030] The present invention also proposes a heat sink preparation optimization device, the device comprising:

[0031] A processing module, used for pre-processing a substrate, forming a deposition layer on the surface of the processed substrate by sputtering, and forming a plurality of deposition layers in sequence on the surface of the top deposition layer by sputtering again, until the total thickness of each deposition layer reaches a preset thickness, the heat sink is generated; each deposition layer corresponds to an initial process control model, and in the process of forming each deposition layer, the process parameters of each process control model are obtained according to a preset sampling frequency to be stored in a database, and the performance parameters of each heat sink are obtained and stored in the database, so as to construct a group of samples based on the performance parameters of each heat sink and the process parameters corresponding to each different deposition layer; the process parameters of the process control model corresponding to at least two adjacent deposition layers are different;

[0032] A first sample module is used to determine a heat sink whose performance parameters meet preset requirements as a qualified heat sink, so as to determine a sample containing at least a preset proportion of qualified heat sinks as a qualified sample; adjust the process parameters to generate a plurality of groups of heat sinks accordingly and obtain the process parameters and performance parameters of each group of heat sinks respectively and store them in the database, so as to construct a sample of not less than a first preset number, and make the number of qualified samples contained in the sample not less than a second preset number;

[0033] The second sample module is used to expand several groups of high-quality samples based on any one or several groups of qualified samples, and to determine whether each of the high-quality samples is a qualified sample;

[0034] The optimization module is used to train the process control models corresponding to different deposition layers based on each group of samples and each group of high-quality samples, so as to improve the process parameters of each process control model to optimize the performance of each deposition layer, thereby optimizing the overall performance of the heat sink.

[0035] The present invention also proposes a computer-readable storage medium, which stores executable instructions for implementing the heat sink preparation optimization method as described above when executed by a processor.

[0036] The present invention has at least the following beneficial effects:

[0037] The solution proposed in the present invention deposits layer by layer by sputtering, collects the process parameters of the process control model corresponding to each deposition layer and obtains the performance parameters of the heat sink finally generated, thereby obtaining multiple groups of samples that meet the preset quantity requirements, and trains each process control model based on the samples and the eligibility of the samples, so that each process control model can respectively perform fine control on a specific deposition layer, thereby achieving precise control of the performance of each deposition layer;

[0038] Each process control model will control at least two adjacent deposition layers of the same heat sink to adopt different process parameters, which can achieve performance differentiation regulation between different deposition layers of the heat sink, so that each deposition layer can show diversified and differentiated effects, and different deposition layers have different directional advantages. It can be specially optimized according to the characteristics of different material layers to ensure that each deposition layer can play the best effect, so as to maximize the overall heat sink performance, effectively prevent the amplification of defects or accumulation of errors, and avoid the performance degradation or unevenness of the deposition layer caused by the use of a fixed sputtering deposition method. The overall solution is more flexible and has obvious advantages in energy consumption and environmental pollution. On this basis, the process parameters of each adjacent deposition layer or each deposition layer can be different, so as to more effectively avoid the accumulation of defects, decreased density, thermal stress mismatch and other problems caused by a single process parameter, and realize the functional complementarity between different layers, which can improve the thermal management performance, adhesion, reliability and durability of the heat sink, and help to improve the overall performance of the entire heat sink, so that it performs better in a variety of working environments.

[0039] Furthermore, this solution adopts the method of association analysis. By dividing the potential associated data through the sample data in the database, the mapping function and multi-computing node distribution calculation can be used to achieve parallel calculation, reduce the calculation burden and deviation and improve the calculation efficiency, and ensure that there will be no calculation bottleneck due to excessive data volume. Finally, sufficiently reliable association rules are generated based on the confidence rule, ensuring that the process control model can mine effective association rules for controlling process parameters from large-scale sample data, making the production process of heat sinks more accurate and reliable, and avoiding the limitations of manual adjustment; through frequent item set algorithm and association rule analysis, potential rules can be mined from a large amount of process data, so as to find the optimal process parameter combination in the heat sink preparation process, provide a clear process parameter relationship for the process control model, and enhance the interpretability of the model;

[0040] By gradually scanning potential data sets and generating frequent item sets, we can more accurately find the close relationship between data, improve the accuracy and effectiveness of process models, and adaptively adjust the size and characteristics of frequent item sets to ensure that the model can flexibly respond to different production conditions and data changes. In addition, multiple rounds of scanning and self-connection can reveal the complex nonlinear relationship between process parameters, helping to build a more accurate process control model.

[0041] By eliminating data that has no practical significance for the generation of the final frequent itemsets in the potential associated data, the computation and memory consumption can be reduced, the computational efficiency can be greatly improved, and the influence of low-frequency data on the final results can be avoided, thereby improving the quality of association rule mining; and by eliminating itemsets whose support does not meet the minimum threshold, redundant calculations can also be reduced, the generation of invalid itemsets can be avoided, the execution of the algorithm can be accelerated, and the quality of the final generated frequent itemsets can be further improved, ensuring that the mined association rules have sufficient statistical significance;

[0042] In addition, this solution can also expand additional high-quality samples based on qualified samples, and limit the number of deposition layers and the number of adjusted process parameters corresponding to the process parameters adjusted by the high-quality samples, thereby increasing the diversity and comprehensiveness of training samples, providing more high-quality data for the process control model, and ensuring that the newly generated high-quality samples have a higher probability of passing as much as possible, so that the process control model can more easily find the association rules for optimizing the process parameters, forming a good optimization closed loop, and improving the effect of heat sink preparation.

[0043] Therefore, the present invention provides a heat sink preparation optimization method, device and readable storage medium. The scheme proposed in the present invention is suitable for the preparation of heat sinks of various thicknesses, which can effectively avoid the defects and performance unevenness problems in the production of heat sinks, and can reflect the diversified and differentiated effects in each deposition layer of the heat sink, so that different deposition layers can have different directional advantages, thereby ensuring that the final generated heat sink has more excellent comprehensive performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 A schematic diagram of a method flow chart of a heat sink preparation optimization method provided in Example 1;

[0046] Figure 2 is an example diagram of a heat sink structure;

[0047] Figure 3 A schematic flow chart of a method for training a process control model;

[0048] Figure 4 A flowchart for determining available association rules;

[0049] Figure 5 A flowchart of a method for calculating a subset of associated data;

[0050] Figure 6 This is a schematic diagram of the module structure of the heat sink preparation optimization device provided in Example 2.

[0051] Reference numerals

[0052] 10-processing module; 20-first sample module; 31-second sample module; 32-optimization module; 40-calculation module; 41-first acquisition unit; 42-second acquisition unit; 43-iterative acquisition unit; 50-merging module; 60-rule determination module. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Hereinafter, various embodiments of the present invention will be described more fully. The present invention may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of the present invention to the specific embodiments disclosed herein, but the present invention should be understood to cover all adjustments, equivalents and / or alternatives falling within the spirit and scope of the various embodiments of the present invention.

[0055] Hereinafter, the term "include" or "may include" used in various embodiments of the present invention indicates the presence of disclosed functions, operations or elements, and does not limit the addition of one or more functions, operations or elements. In addition, as used in various embodiments of the present invention, the terms "include", "have" and their cognates are intended only to indicate specific features, numbers, steps, operations, elements, components or combinations of the foregoing items, and should not be understood as first excluding the presence of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing items or the possibility of adding one or more features, numbers, steps, operations, elements, components or combinations of the foregoing items.

[0056] In various embodiments of the present invention, the expression "or" or "at least one of A or / and B" includes any combination or all combinations of the words listed at the same time. For example, the expression "A or B" or "at least one of A or / and B" may include A, may include B, or may include both A and B.

[0057] The expressions (such as "first", "second", etc.) used in various embodiments of the present invention may modify the various constituent elements in various embodiments, but may not limit the corresponding constituent elements. For example, the above expressions do not limit the order and / or importance of the elements. The above expressions are only used for the purpose of distinguishing an element from other elements. For example, the first user device and the second user device indicate different user devices, although both are user devices. For example, without departing from the scope of various embodiments of the present invention, the first element may be referred to as the second element, and similarly, the second element may also be referred to as the first element.

[0058] It should be noted that in the present invention, unless otherwise clearly specified and defined, the terms such as "installation", "connection", "fixation" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0059] In the present invention, those skilled in the art need to understand that the terms indicating orientation or positional relationship herein are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0060] The terms used in various embodiments of the present invention are only used to describe the purpose of specific embodiments and are not intended to limit the various embodiments of the present invention. As used herein, the singular form is intended to also include the plural form, unless the context clearly indicates otherwise. Unless otherwise limited, all terms used here (including technical terms and scientific terms) have the same meaning as the meaning generally understood by those of ordinary skill in the art to which the various embodiments of the present invention belong. The terms (such as the terms defined in the dictionary generally used) will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning, unless clearly defined in various embodiments of the present invention.

[0061] Example 1

[0062] This embodiment proposes a heat sink preparation optimization method, please refer to Figure 1-Figure 2 , methods include:

[0063] S100: Pre-treating the substrate, forming a deposition layer on the surface of the processed substrate by sputtering, and sequentially forming several deposition layers on the surface of the top deposition layer by sputtering again, until the total thickness of each deposition layer reaches a preset thickness to complete the generation of the heat sink; in the process of forming each deposition layer, obtaining the process parameters of each process control model at a preset sampling frequency and storing them in a database, and respectively obtaining the performance parameters of each heat sink and storing them in the database, so as to construct a group of samples based on the performance parameters of each heat sink and the process parameters corresponding to each different deposition layer.

[0064] It should be noted that the pretreatment of the substrate may include but is not limited to placing in a vacuum environment, cleaning, surface roughening, removing oxides, preheating, activation and / or charge adjustment; the pretreatment in step S100 can remove pollutants or an oxide layer on the surface of the substrate to ensure that the deposited layer can be evenly and firmly attached to the substrate;

[0065] Process parameters may include, but are not limited to, vacuum degree, substrate temperature, target temperature, sputtering power, sputtering time, sputtering atmosphere, substrate movement speed, target distance and / or target movement speed;

[0066] Performance parameters may include, but are not limited to, thickness, density, surface roughness, grain size, degree of crystallinity, tensile strength, electrical conductivity, compactness, and / or uniformity of the deposited layer.

[0067] It should also be noted that if adjacent deposition layers use the same process parameters, the same defects may continue to accumulate. When the thickness exceeds a certain range, the density and conductivity of the deposition layers will be reduced, and interlayer thermal stress mismatch will occur, thereby affecting the stability and performance of the heat sink. For this reason, the method proposed in this embodiment will make the process parameters of the process control model corresponding to at least two adjacent deposition layers different. By using differentiated process parameters, the physical properties of each deposition layer can be adjusted to optimize the overall performance of the heat sink. In addition, different process parameters can be used to adjust the physical and chemical properties of different deposition layers, and the functional properties of each layer can be accurately controlled, so that the deposition layers are more complementary in terms of thermal management, oxidation resistance, conductivity, etc., and ultimately the heat sink performs better under different working environments.

[0068] For example, by adjusting the process parameters such as temperature, atmosphere, power, etc. of different deposition layers, the crystal structure, density, hardness and thermal conductivity of each layer can be affected, so that different layers of the heat sink can complement each other in terms of heat conduction to improve the overall thermal management performance; by adjusting the process parameters of different deposition layers to give them different toughness and hardness, the adhesion between the layers can be enhanced, thereby increasing the reliability and durability of the heat sink and reducing peeling or cracking that may occur during use.

[0069] Specifically, process parameters of the process control model corresponding to at least two adjacent deposition layers are different;

[0070] Preferably, the process parameters of any two adjacent deposition layers in the same heat sink determined based on each process control model are different;

[0071] Or, the process parameters of any two deposition layers in the same heat sink determined based on each process control model are different; by adjusting the process parameters of each deposition layer in the heat sink prepared by the process control model, adjacent deposition layers have differentiated physical and chemical properties, which can effectively avoid problems such as defect accumulation, density reduction, thermal stress mismatch, etc. caused by a single process parameter, and achieve functional complementarity between different layers, which can improve the thermal management performance, adhesion, reliability and durability of the heat sink, and help improve the overall performance of the entire heat sink, making it perform better in a variety of working environments.

[0072] S200: Determine a heat sink whose performance parameters meet preset requirements as a qualified heat sink, and determine a sample that contains at least a preset proportion of qualified heat sinks as a qualified sample; adjust the process parameters to generate multiple groups of heat sinks accordingly and obtain the process parameters and performance parameters of each group of heat sinks and store them in a database to construct a sample of no less than a first preset number, and the number of qualified samples contained in the sample is no less than a second preset number.

[0073] In this embodiment, the preset requirements may include but are not limited to indicators such as thermal conductivity, strength, and durability. The first preset number is not less than the second preset number. Through step S200, the method proposed in this embodiment can obtain a sufficient number of representative and diverse samples, and the samples contain a sufficient number of qualified samples, so that the process control model trained by the samples can show better predictive ability.

[0074] S300: Expand several groups of high-quality samples based on any one or several groups of qualified samples, and determine whether each of the high-quality samples is a qualified sample, so as to train the process control models corresponding to different deposition layers based on each group of samples and each group of high-quality samples, thereby improving the process parameters of each process control model to optimize the performance of each deposition layer, and further optimize the overall performance of the heat sink.

[0075] Furthermore, for any group of qualified samples, a total of N process parameters corresponding to M deposition layers in the qualified samples are adjusted, and several groups of heat sinks are generated accordingly, and process parameters and performance parameters are obtained and stored in the database to expand several groups of high-quality samples, and high-quality samples that at least contain a preset proportion of qualified heat sinks are also determined as qualified samples; M is not less than one and not greater than a third preset number, and N is not less than one and not greater than a fourth preset number.

[0076] Preferably, in step S100, when a certain deposition layer begins to form, a sampling time for first acquiring the process parameters is determined, and after the process parameters are first acquired, the process parameters of each of the process control models are periodically acquired based on a preset sampling frequency;

[0077] After the deposition layer is completely formed, the acquired process parameters and their corresponding sampling times will be stored in the sub-database of the corresponding process control model in the database; any process control model is trained through the performance parameters of the corresponding samples in the database and the process parameters of the corresponding samples in the sub-database of the corresponding process control model.

[0078] Specifically, the method of training the process control model includes:

[0079] S400: Based on the performance parameters of each sample and the process parameters of each process control model corresponding to each sample in the database, several groups of potential associated data of the same size are divided and distributed to multiple computing nodes through a mapping function, so as to calculate the associated data subset corresponding to each potential associated data on each computing node.

[0080] It should be noted that each associated data subset contains several associated sets to be determined. The distributed concept is adopted in step S500, and based on the performance parameters of each sample in the database and the process parameters of different deposition layers corresponding to each sample in the sub-database, several groups of potential associated data corresponding to different deposition layers can be divided out respectively, so as to train process control models corresponding to different deposition layers.

[0081] S500: Merge each associated data subset into a candidate associated union set, calculate the support of each to-be-determined associated set in the candidate associated union set, and generate an associated data union set based on the to-be-determined associated sets whose support is not less than a minimum support threshold.

[0082] It should be noted that support is an indicator that measures the frequency of occurrence of a certain association rule in a data set. In this embodiment, support specifically refers to the proportion of certain process parameter combinations or the relationship between process parameters and performance in all potential associated data; the minimum support threshold is a predetermined standard for determining which association sets are meaningful. Association sets with support lower than the minimum support threshold may represent relationships that do not appear frequently enough in the data, that is, the credibility of their association relationships is low.

[0083] S600: Generate several non-empty subsets based on each to-be-determined association set in the union of associated data, form a candidate association rule based on each non-empty subset, calculate the confidence of each candidate association rule, determine the candidate association rule with a confidence not less than a minimum confidence threshold as an association rule that can be used to determine process parameters and retain it, so as to complete the training of each process control model corresponding to different deposition layers.

[0084] The non-empty subset in step S600 is a subdivision of the association set. The non-empty subset can be used as the basis for candidate association rules. For each non-empty subset, the system generates a candidate association rule. Each candidate association rule represents a possible process control relationship;

[0085] For example, an association set may be "when the temperature is X, the pressure is Y", then multiple non-empty subsets may be generated from this set. For example, if "temperature is X" or "pressure is Y" is taken as a subset, then a candidate association rule may be "if the process parameter temperature of the process control model corresponding to the deposition layer is X and the pressure is Y, then the final performance parameter Z of the heat sink formed by the deposition layer may reach the best";

[0086] The confidence of each candidate association rule indicates the probability of the rule being true, that is, the frequency of the result occurring under the premise. The higher the confidence, the more reliable the rule is. The candidate rules with lower confidence can be filtered out through the minimum confidence threshold, so that each process control model can determine the process parameters for preparing the heat sink through association rules with high confidence.

[0087] Specifically, the method of calculating and obtaining a subset of associated data corresponding to a potential associated data in step S400 includes:

[0088] S410: Scan the potential associated data to obtain univariate frequent item sets whose support is not less than a minimum support threshold, and obtain a first to-be-determined associated set by self-connecting the univariate frequent item sets.

[0089] Preferably, step S410 can scan the potential associated data, generate a column data table based on the potential associated data, scan the column data table to obtain unary frequent item sets, calculate the support of each unary frequent item set, and obtain unary frequent item sets whose support is not less than a minimum support threshold; thus, step S410 converts the potential associated data into a column data table so that the ID of each item set is listed, so that the support of each item set can be calculated by a simple set operation, and therefore, only the potential associated data needs to be scanned once to obtain the unary frequent item sets, thereby reducing unnecessary operations and being more efficient than directly scanning the potential associated data to obtain the unary frequent item sets.

[0090] S420: Scan the potential associated data again to calculate the support of the first association set to be determined, to obtain binary frequent itemsets whose support is not less than the minimum support threshold, and obtain the second association set to be determined by self-connection of the binary frequent itemsets.

[0091] S430: Repeat the iterative scanning of potential associated data to calculate the support of the currently obtained association set to be determined, so as to obtain a new frequent item set whose support is not less than the minimum support threshold, and obtain a new association set to be determined by self-connecting the latest frequent item set, until the latest frequent item set does not meet the minimum support requirement, and then generate an associated data subset corresponding to the potential associated data based on each association set to be determined.

[0092] Furthermore, before scanning the potential associated data in steps S410-S430, data in the potential associated data whose support is less than the minimum support threshold can be eliminated; accordingly, when calculating the support of the currently obtained association set to be determined and obtaining new frequent item sets whose support is not less than the minimum support threshold, new frequent item sets whose support is less than the minimum support threshold can be eliminated; this helps to save computing resources and improve mining efficiency in the process of executing step S400, and avoid repeated calculations on irrelevant item sets in subsequent steps.

[0093] Therefore, the method proposed in the present invention reduces the computation and memory consumption by eliminating the data that has no practical significance for the generation of the final frequent itemsets in the potential associated data, greatly improves the computation efficiency, and avoids the influence of low-frequency data on the final result, thereby improving the quality of association rule mining;

[0094] By eliminating item sets whose support does not meet the minimum threshold, redundant calculations can also be reduced, the generation of invalid item sets can be avoided, the execution of the algorithm can be accelerated, and the quality of the final generated frequent item sets can be further improved, ensuring that the mined rules have sufficient statistical significance.

[0095] Example 2

[0096] This embodiment proposes a heat sink preparation optimization device, which is used to implement the heat sink preparation optimization method proposed in Example 1. The device includes:

[0097] The processing module 10 is used to pre-process the substrate, form a deposition layer on the surface of the processed substrate by sputtering, and sequentially form several deposition layers on the surface of the top deposition layer by sputtering again, until the total thickness of each deposition layer reaches a preset thickness to complete the generation of the heat sink; each deposition layer corresponds to an initial process control model, and in the process of forming each deposition layer, the process parameters of each process control model are obtained according to a preset sampling frequency to store in a database, and the performance parameters of each heat sink are obtained and stored in the database, so as to construct a set of samples based on the performance parameters of each heat sink and the process parameters corresponding to each different deposition layer;

[0098] The first sample module 20 is used to determine the heat sinks whose performance parameters meet the preset requirements as qualified heat sinks, and to determine the samples containing at least a preset proportion of qualified heat sinks as qualified samples; adjust the process parameters to generate multiple groups of heat sinks accordingly and obtain the process parameters and performance parameters of each group of heat sinks respectively and store them in the database, so as to construct a sample of not less than a first preset number, and make the number of qualified samples contained in the sample not less than a second preset number;

[0099] The second sample module 31 is used to expand several groups of high-quality samples based on any one or several groups of qualified samples, and to determine whether each high-quality sample is a qualified sample;

[0100] The optimization module 32 is used to train each process control model corresponding to different deposition layers based on each group of samples and each group of high-quality samples, so as to improve the process parameters of each process control model to optimize the performance of each deposition layer, and further optimize the overall performance of the heat sink.

[0101] Specifically, the process parameters of the process control model corresponding to at least two adjacent deposition layers are different. By using differentiated process parameters, the physical properties of each deposition layer can be adjusted to optimize the overall performance of the heat sink. Moreover, different process parameters can be used to adjust the physical and chemical properties of different deposition layers, and the functional properties of each layer can be accurately controlled, so that the deposition layers are more complementary in terms of thermal management, oxidation resistance, conductivity, etc., and ultimately the heat sink performs better under different working environments. By adjusting the process parameters of each deposition layer in the heat sink prepared by the process control model, the adjacent deposition layers have differentiated physical and chemical properties, which can effectively avoid problems such as defect accumulation, decreased density, and thermal stress mismatch caused by a single process parameter, and achieve functional complementarity between different layers, which can improve the thermal management performance, adhesion, reliability and durability of the heat sink, and help improve the overall performance of the entire heat sink, so that it performs better in a variety of working environments.

[0102] Preferably, the process parameters of any two adjacent deposition layers in the same heat sink determined based on each process control model are different;

[0103] Alternatively, process parameters of any two deposition layers in the same heat sink determined based on each process control model are different.

[0104] The preset requirements may include but are not limited to indicators such as thermal conductivity, strength, and durability. The first preset number is not less than the second preset number. Through the first sample module 20, the device proposed in this embodiment can obtain a sufficient number of representative and diverse samples, and the samples contain a sufficient number of qualified samples, so that the process control model trained by the samples can show better predictive ability.

[0105] Furthermore, for any group of qualified samples, the second sample module 31 can adjust a total of N process parameters corresponding to M deposition layers in the qualified samples, and accordingly generate several groups of heat sinks and obtain process parameters and performance parameters and store them in the database to expand several groups of high-quality samples, and similarly determine high-quality samples that contain at least a preset proportion of qualified heat sinks as qualified samples; M is not less than one and not greater than a third preset number, and N is not less than one and not greater than a fourth preset number.

[0106] Preferably, when a certain deposition layer begins to form, the processing module 10 determines a sampling time for first acquiring the process parameters, and after first acquiring the process parameters, acquires the process parameters of each of the process control models based on a preset sampling frequency;

[0107] After the deposition layer is completely formed, the processing module 10 will store the acquired process parameters and their corresponding sampling times in the sub-database corresponding to the process control model in the database; any process control model is trained through the performance parameters of the corresponding samples in the database and the process parameters of the corresponding samples in the sub-database of the corresponding process control model.

[0108] Specifically, the device also includes:

[0109] The computing module 40 is used to divide a plurality of groups of potential associated data of the same size based on the performance parameters of each sample and the process parameters of each process control model corresponding to each sample in the database, and distribute the groups to a plurality of computing nodes through a mapping function, so as to calculate the associated data subset corresponding to each potential associated data on each computing node;

[0110] A merging module 50 is used to merge the associated data subsets into a candidate associated union set, calculate the support of each to-be-determined associated set in the candidate associated union set, and generate an associated data union set based on the to-be-determined associated sets whose support is not less than a minimum support threshold;

[0111] The rule determination module 60 is used to generate a number of non-empty subsets based on each to-be-determined association set in the association data union, to form a candidate association rule based on each non-empty subset, to calculate the confidence of each candidate association rule, to determine the candidate association rule whose confidence is not less than the minimum confidence threshold as the association rule that can be used to determine the process parameters and to retain it, so as to complete the training of each process control model corresponding to different deposition layers.

[0112] It should be noted that each associated data subset contains several to-be-determined associated sets, and the calculation of the calculation module 40 adopts a distributed concept, and can be based on the performance parameters of each sample in the database and the data of the process parameters of different deposition layers corresponding to each sample in the sub-database, respectively divide several groups of potential associated data corresponding to different deposition layers, thereby training the process control models corresponding to different deposition layers;

[0113] In this embodiment, support specifically refers to the proportion of certain process parameter combinations or the relationship between process parameters and performance in all potential associated data; the minimum support threshold is a predetermined standard for determining which association sets are meaningful, and the relationship represented by the association set with a support lower than the minimum support threshold may not appear frequently enough in the data, that is, the credibility of its association relationship is low;

[0114] A non-empty subset is a subdivision of an association set. A non-empty subset can be used as the basis for candidate association rules. For each non-empty subset, the system generates a candidate association rule. Each candidate association rule represents a possible process control relationship.

[0115] The confidence of each candidate association rule indicates the probability of the rule being true, that is, the frequency of the result occurring under the premise. The higher the confidence, the more reliable the rule is. The candidate rules with lower confidence can be filtered out through the minimum confidence threshold, so that each process control model can determine the process parameters used in the preparation of heat sinks through association rules with high confidence.

[0116] Specifically, the calculation module 40 includes:

[0117] A first acquisition unit 41 is used to scan the potential associated data to obtain a univariate frequent item set whose support is not less than a minimum support threshold, and obtain a first to-be-determined associated set by self-connection of the univariate frequent item set;

[0118] The second acquisition unit 42 is used to scan the potential association data again to calculate the support of the first association set to be determined, so as to obtain a binary frequent item set whose support is not less than a minimum support threshold, and obtain a second association set to be determined by self-connection of the binary frequent item set;

[0119] The iterative acquisition unit 43 is used to repeatedly iteratively scan the potential associated data to calculate the support of the currently obtained association set to be determined, so as to obtain a new frequent item set whose support is not less than the minimum support threshold, and obtain a new association set to be determined by self-connecting the latest frequent item set, until the latest frequent item set does not meet the minimum support requirement, and then generate an associated data subset corresponding to the potential associated data based on each association set to be determined.

[0120] Preferably, the first acquisition unit 41 can scan the potential associated data, generate a column data table based on the potential associated data, scan the column data table to obtain uniary frequent item sets, calculate the support of each uniary frequent item set, and obtain uniary frequent item sets whose support is not less than a minimum support threshold; thus, the first acquisition unit 41 converts the potential associated data into a column data table so that the ID of each item set is listed, so that the support of each item set can be calculated by a simple set operation. Therefore, when obtaining the uniary frequent item sets, it is only necessary to scan the potential associated data once, which reduces unnecessary operations and is more efficient than directly scanning the potential associated data to obtain the uniary frequent item sets.

[0121] Furthermore, before the first acquisition unit 41, the second acquisition unit 42 and the iterative acquisition unit 43 scan the potential associated data, the data whose support is less than the minimum support threshold in the potential associated data can be eliminated; accordingly, when the support of the currently obtained association set to be determined is calculated and a new frequent item set whose support is not less than the minimum support threshold is obtained, the new frequent item set whose support is less than the minimum support threshold can be eliminated; this helps the calculation module 40 to save computing resources, improve mining efficiency, and avoid repeated calculations on irrelevant item sets in subsequent steps.

[0122] By eliminating data that has no practical significance for the generation of the final frequent itemsets, the computation and memory consumption are reduced, the computational efficiency is greatly improved, and the influence of low-frequency data on the final results is avoided, thereby improving the quality of association rule mining;

[0123] By eliminating item sets whose support does not meet the minimum threshold, redundant calculations can also be reduced, the generation of invalid item sets can be avoided, the execution of the algorithm can be accelerated, and the quality of the final generated frequent item sets can be further improved, ensuring that the mined rules have sufficient statistical significance.

[0124] Example 3

[0125] This embodiment proposes a computer-readable storage medium, which stores executable instructions for implementing the heat sink preparation optimization method proposed in Embodiment 1 when executed by a processor.

[0126] It should be noted that computer-readable media include permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carriers.

[0127] In summary, the present invention provides a method, device and readable storage medium for optimizing the preparation of a heat sink. The scheme proposed in the present invention is suitable for the preparation of heat sinks of various thicknesses, can effectively avoid defects and uneven performance problems occurring in the production of heat sinks, and can reflect diversified and differentiated effects in each deposition layer of the heat sink, so that different deposition layers can have advantages of different directions, thereby ensuring that the final generated heat sink has better comprehensive performance.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A heat sink preparation optimization method, characterized in that: The method comprises: Pre-processing the substrate, forming a deposition layer on the surface of the processed substrate by sputtering, and sequentially forming several deposition layers on the surface of the top deposition layer by sputtering again, until the total thickness of each deposition layer reaches a preset thickness, the heat sink is generated; each deposition layer corresponds to an initial process control model, and in the process of forming each deposition layer, the process parameters of each process control model are obtained at a preset sampling frequency to be stored in a database, and the performance parameters of each heat sink are obtained and stored in the database, so as to construct a group of samples based on the performance parameters of each heat sink and the process parameters corresponding to each different deposition layer; the process parameters of the process control model corresponding to at least two adjacent deposition layers are different; Determine a heat sink whose performance parameters meet preset requirements as a qualified heat sink, and determine a sample containing at least a preset proportion of qualified heat sinks as a qualified sample; adjust the process parameters to generate multiple groups of heat sinks accordingly and obtain the process parameters and performance parameters of each group of heat sinks respectively and store them in the database, so as to construct a sample of not less than a first preset number, and make the number of qualified samples contained in the sample not less than a second preset number; Based on any one or several groups of qualified samples, several groups of high-quality samples are expanded, and it is determined whether each of the high-quality samples is a qualified sample, so as to train the process control models corresponding to different deposition layers based on the samples of not less than the first preset number previously constructed in each group and the high-quality samples in each group, so as to improve the process parameters of each process control model to optimize the performance of each deposition layer, and further optimize the overall performance of the heat sink.

2. The method according to claim 1, characterized in that Methods for training process control models include: Based on the data of the performance parameters corresponding to each of the samples and the process parameters of each process control model corresponding to each of the samples in the database, a plurality of groups of potential associated data of the same size are divided and distributed to a plurality of computing nodes through a mapping function, so as to calculate on each of the computing nodes to obtain associated data subsets corresponding to each of the potential associated data; each of the associated data subsets includes a plurality of to-be-determined associated sets; Merging each of the associated data subsets into a candidate associated union set, calculating the support of each to-be-determined associated set in the candidate associated union set, and generating an associated data union set based on the to-be-determined associated sets whose support is not less than a minimum support threshold; Based on each to-be-determined association set in the union of the association data, several non-empty subsets are generated to form a candidate association rule based on each of the non-empty subsets, and the confidence of each of the candidate association rules is calculated. The candidate association rules whose confidence is not less than a minimum confidence threshold are determined as association rules that can be used to determine process parameters and are retained to complete the training of each of the process control models corresponding to different deposition layers.

3. The method according to claim 2, characterized in that The method of calculating and obtaining a subset of associated data corresponding to a potential associated data includes: Scan the potential association data to obtain a univariate frequent item set whose support is not less than the minimum support threshold, and obtain a first association set to be determined by self-connecting the univariate frequent item set; Scan the potential association data again to calculate the support of the first association set to be determined, so as to obtain a binary frequent item set whose support is not less than the minimum support threshold, and obtain a second association set to be determined by self-connecting the binary frequent item set; Repeatedly iteratively scan the potential associated data to calculate the support of the currently obtained association set to be determined, so as to obtain a new frequent item set whose support is not less than the minimum support threshold, and obtain a new association set to be determined by self-connecting the latest frequent item set, until the latest frequent item set fails to meet the minimum support requirement, and then generate an associated data subset corresponding to the potential associated data based on each association set to be determined.

4. The method according to claim 3, characterized in that The method further comprises: Before scanning the potential associated data, remove data whose support is less than the minimum support threshold in the potential associated data; When calculating the support of the currently obtained association set to be determined and obtaining new frequent item sets whose support is not less than the minimum support threshold, the new frequent item sets whose support is less than the minimum support threshold are eliminated.

5. The method according to claim 3, characterized in that: The "scanning the potential associated data to obtain unary frequent itemsets whose support is not less than the minimum support threshold" includes: Scan the potential associated data, generate a column data table based on the potential associated data, scan the column data table to obtain univariate frequent item sets, calculate the support of each of the univariate frequent item sets to obtain univariate frequent item sets whose support is not less than the minimum support threshold.

6. The method according to any one of claims 1 to 5, characterized in that: The “obtaining the process parameters of each process control model at a preset sampling frequency during the process of forming each deposition layer to store them in a database” includes: When a certain deposition layer begins to form, determining a sampling time for first acquiring process parameters, and after first acquiring the process parameters, acquiring the process parameters of each of the process control models based on the preset sampling frequency; After the deposition layer is completely formed, the obtained process parameters and their corresponding sampling times are stored in the sub-database of the database corresponding to the process control model; any process control model is trained through the performance parameters of the samples in the database and the process parameters of the samples in the sub-database corresponding to the process control model.

7. The method according to claim 1, characterized in that The method further comprises: The process parameters of any two adjacent deposition layers in the same heat sink determined based on each of the process control models are different; Alternatively, process parameters of any two deposition layers in the same heat sink determined based on the process control models are different.

8. The method according to claim 1, characterized in that The "expanding several groups of high-quality samples based on any one or several groups of qualified samples" includes: For any group of qualified samples, a total of N process parameters corresponding to M deposition layers in the qualified samples are adjusted, and several groups of heat sinks are generated accordingly, and process parameters and performance parameters are obtained and stored in the database to expand several groups of high-quality samples; M is not less than one and not greater than a third preset number, and N is not less than one and not greater than a fourth preset number.

9. A heat sink preparation optimization device, characterized in that: The device comprises: A processing module, used for pre-processing a substrate, forming a deposition layer on the surface of the processed substrate by sputtering, and forming a plurality of deposition layers in sequence on the surface of the top deposition layer by sputtering again, until the total thickness of each deposition layer reaches a preset thickness, the heat sink is generated; each deposition layer corresponds to an initial process control model, and in the process of forming each deposition layer, the process parameters of each process control model are obtained according to a preset sampling frequency to be stored in a database, and the performance parameters of each heat sink are obtained and stored in the database, so as to construct a group of samples based on the performance parameters of each heat sink and the process parameters corresponding to each different deposition layer; the process parameters of the process control model corresponding to at least two adjacent deposition layers are different; A first sample module is used to determine a heat sink whose performance parameters meet preset requirements as a qualified heat sink, so as to determine a sample containing at least a preset proportion of qualified heat sinks as a qualified sample; adjust the process parameters to generate a plurality of groups of heat sinks accordingly and obtain the process parameters and performance parameters of each group of heat sinks respectively and store them in the database, so as to construct a sample of not less than a first preset number, and make the number of qualified samples contained in the sample not less than a second preset number; The second sample module is used to expand several groups of high-quality samples based on any one or several groups of qualified samples, and to determine whether each of the high-quality samples is a qualified sample; The optimization module is used to train the process control models corresponding to different deposition layers based on each group of samples previously constructed and having a number not less than a first preset number and each group of high-quality samples, so as to improve the process parameters of each process control model to optimize the performance of each deposition layer, thereby optimizing the overall performance of the heat sink.

10. A computer-readable storage medium, characterized in that: Executable instructions are stored, and when executed by a processor, the instructions are used to implement the method according to any one of claims 1 to 8.

Images