A chip packaging design optimization method based on adaptive sub-problem selection strategy

Abstract

The present invention discloses a chip packaging design optimization method based on an adaptive sub-problem selection strategy, characterized in that the method includes: obtaining a number of initial parameters of the chip packaging design, and performing finite element analysis on the initial parameters to obtain prediction results of a number of targets to be optimized, thereby obtaining an initial data set; constructing a multi-objective optimization problem according to the several targets to be optimized and the initial parameters, decomposing the multi-objective optimization problem to obtain a number of single-objective optimization sub-problems, and based on a preset selection strategy, obtaining a number of candidate single-objective optimization sub-problems by screening according to the initial data set; constructing a proxy model according to the initial data set and the number of candidate single-objective optimization sub-problems, thereby obtaining an updated data set; based on a preset clustering algorithm, obtaining the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters according to the updated data set. The present invention can achieve the optimization of multiple parameters for multiple targets.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a chip packaging design optimization method based on an adaptive sub-problem selection strategy. Background Art

[0002] Integrated circuit packaging technology is developing towards precision and miniaturization. Currently, there are many studies on chip packaging parameter optimization to reduce the distribution of thermal stress and chip warping during cooling. The traditional computational-experimental method that relies on experience consumes huge manpower, material resources and time. And because the chip packaging problem has multiple optimization parameters, the traditional experimental method is not competent for this task. The finite element numerical simulation method, using the finite element model, can simulate more complex boundary conditions and material properties, which greatly reduces the cost compared to the real experiment, and thus has been widely used. When solving these optimization problems, the finite element simulation method often requires the use of high-precision numerical analysis techniques, such as finite element analysis or computational fluid dynamics simulation to evaluate the performance of candidate solutions. However, the disadvantage is that the finite element simulation requires a long running time. Each time the parameters are adjusted, a finite element simulation must be run. The time consumed for running once is very long, ranging from a few minutes to a few hours, or even a few days of CPU time. The cost of a large number of parameter tests is still high. Common finite element simulation methods combined with metaheuristic algorithms can form a cyclic workflow to optimize multiple parameters at the same time, and then optimize a single goal, but it is difficult to optimize multiple goals at the same time. In the actual chip packaging optimization problem, multiple goals need to be optimized at the same time. Therefore, the chip packaging design method of the prior art cannot achieve the optimization of multiple parameters and multiple goals.

[0003] Therefore, the existing technology still needs to be improved and developed. Summary of the invention

[0004] The technical problem to be solved by the present invention is that, in view of the above-mentioned defects of the prior art, a chip packaging design optimization method based on an adaptive sub-problem selection strategy is provided, aiming to solve the problem that the chip packaging design method in the prior art cannot optimize multiple parameters for multiple objectives.

[0005] The technical solution adopted by the present invention to solve the problem is as follows:

[0006] In a first aspect, an embodiment of the present invention provides a chip packaging design optimization method based on an adaptive sub-problem selection strategy, wherein the method comprises:

[0007] Acquire initial parameters of chip package design, and perform finite element analysis on some of the initial parameters based on preset finite element simulation software to obtain prediction results of some targets to be optimized, and use the initial parameters and the prediction results of some targets to be optimized as initial data sets;

[0008] According to the plurality of objectives to be optimized and the plurality of initial parameters, a multi-objective optimization problem is constructed, the multi-objective optimization problem is decomposed to obtain a plurality of single-objective optimization sub-problems, and based on a preset selection strategy and the initial data set, the plurality of single-objective optimization sub-problems are screened to obtain a plurality of candidate single-objective optimization sub-problems;

[0009] Constructing a proxy model according to the initial data set and the plurality of candidate single-objective optimization sub-problems, and obtaining an updated data set according to the proxy model and the initial data set;

[0010] Based on a preset clustering algorithm and according to an updated data set, optimal parameters of chip packaging design and optimal targets corresponding to the optimal parameters are obtained.

[0011] In one implementation, constructing a multi-objective optimization problem according to the plurality of objectives to be optimized and the plurality of initial parameters includes:

[0012] Normalizing the plurality of targets to be optimized to obtain a plurality of normalized targets;

[0013] Performing constraint processing on a number of the normalized targets;

[0014] A multi-objective optimization problem is obtained according to the several initial parameters and the several normalized objectives processed with constraints.

[0015] In one implementation, decomposing the multi-objective optimization problem to obtain a number of single-objective optimization sub-problems includes:

[0016] Based on the Chebyshev algorithm, the multi-objective optimization problem is decomposed into several single-objective optimization sub-problems.

[0017] In one implementation, the screening of the single-objective optimization sub-problems based on the preset selection strategy and the initial data set to obtain the candidate single-objective optimization sub-problems includes:

[0018] For each target to be optimized, calculate the distance between the vector corresponding to the target to be optimized and each single-objective optimization sub-problem, obtain the minimum distance between the vectors corresponding to the target to be optimized, and pair the single-objective optimization sub-problem corresponding to the minimum distance with the target to be optimized;

[0019] The single-objective optimization sub-problems that fail to be paired are deleted, and the single-objective optimization sub-problems paired with each target to be optimized are classified into the first sub-problem set;

[0020] For each subproblem in the first subproblem set, extract the first non-dominated solution set in the initial data set, sort the non-dominated solutions in the first non-dominated solution set in descending order based on the distance, and obtain the sorting sequence number of each non-dominated solution; obtain the solution degree value of each subproblem based on the first non-dominated solution set, the sorting sequence number and the distance, wherein the solution degree value is used to characterize the degree of solution to the problem; classify the subproblems whose solution degree values ​​are less than a first preset threshold into the second subproblem set, and delete the subproblems whose solution degree values ​​are less than the first preset threshold in the first subproblem set, to obtain the third subproblem set;

[0021] When the number of subproblems in the second subproblem set is less than a second preset threshold, the subproblem with the smallest solution degree value in the third subproblem set is classified into the second subproblem set, and the subproblems in the second subproblem set are used as several candidate single-objective optimization subproblems.

[0022] In one implementation, the step of deleting the single-objective optimization sub-problems that fail to be paired and classifying the single-objective optimization sub-problems paired with each of the objectives to be optimized into the first sub-problem set includes:

[0023] A number of single-objective optimization sub-problems are randomly selected from the single-objective optimization sub-problems that fail to pair, and the randomly selected number of single-objective optimization sub-problems are expanded to the first sub-problem set.

[0024] In one implementation, obtaining an updated data set according to the proxy model and the initial data set includes:

[0025] Obtaining predicted posterior information of the proxy model according to the proxy model, and obtaining recommended parameters through a preset acquisition function according to the predicted posterior information; wherein the predicted posterior information is used to characterize the posterior distribution of the proxy model;

[0026] Expanding the recommended parameters to the several initial parameters to obtain updated initial parameters, iteratively performing finite element analysis on the several initial parameters based on preset finite element simulation software to obtain prediction results of several targets to be optimized, and using the several initial parameters and the prediction results of several targets to be optimized as the initial data set, until a preset condition is met, and then stopping the iteration;

[0027] The initial parameters in the initial data set are replaced with the expanded initial parameters to obtain an updated data set.

[0028] In one implementation, obtaining the optimal parameters of the chip package design and the optimal target corresponding to the optimal parameters based on the updated data set based on the preset clustering algorithm includes:

[0029] Obtaining a second non-dominated solution set of the updated data set;

[0030] Based on a preset clustering algorithm, clustering the non-dominated solutions in the second non-dominated solution set to obtain cluster centers;

[0031] Calculate the distance between the parameters in the updated data set and the cluster center;

[0032] The parameter corresponding to the minimum distance is taken as the optimal parameter for chip packaging design;

[0033] The target to be optimized corresponding to the minimum distance is taken as the optimal target corresponding to the optimal parameter.

[0034] In a second aspect, an embodiment of the present invention further provides a chip packaging design optimization device based on an adaptive sub-problem selection strategy, wherein the device comprises:

[0035] An initial data set determination module is used to obtain initial parameters of chip packaging design, and perform finite element analysis on several of the initial parameters based on preset finite element simulation software to obtain prediction results of several targets to be optimized, and use the several initial parameters and the prediction results of several targets to be optimized as the initial data set; a single-objective optimization sub-problem screening module is used to construct a multi-objective optimization problem based on several targets to be optimized and several initial parameters, decompose the multi-objective optimization problem to obtain several single-objective optimization sub-problems, and screen several of the single-objective optimization sub-problems based on a preset selection strategy and the initial data set to obtain several candidate single-objective optimization sub-problems;

[0036] An updated data set acquisition module is used to construct a proxy model based on the initial data set and the plurality of candidate single-objective optimization sub-problems, and obtain an updated data set based on the proxy model and the initial data set;

[0037] The optimal parameter and optimal target determination module is used to obtain the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters based on the preset clustering algorithm and the updated data set.

[0038] In a third aspect, an embodiment of the present invention further provides an intelligent terminal, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, wherein the one or more programs include a method for executing a chip packaging design optimization method based on an adaptive sub-problem selection strategy as described in any one of the above.

[0039] In a fourth aspect, an embodiment of the present invention also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute a chip packaging design optimization method based on an adaptive sub-problem selection strategy as described in any one of the above.

[0040] Beneficial effects of the present invention: The embodiment of the present invention first obtains the initial parameters of the chip packaging design, and performs finite element analysis on several of the initial parameters based on preset finite element simulation software to obtain prediction results of several targets to be optimized, and uses the several initial parameters and the prediction results of several targets to be optimized as the initial data set; then, according to the several targets to be optimized and the several initial parameters, a multi-objective optimization problem is constructed, the multi-objective optimization problem is decomposed to obtain several single-objective optimization sub-problems, and based on the preset selection strategy and the initial data set, several of the single-objective optimization sub-problems are screened to obtain several candidate single-objective optimization sub-problems; then, according to the initial data set and the several candidate single-objective optimization sub-problems, a proxy model is constructed, and according to The proxy model and the initial data set are used to obtain an updated data set; finally, based on a preset clustering algorithm, according to the updated data set, the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters are obtained; it can be seen that in the embodiment of the present invention, the optimization problem of the target of the chip packaging design is decomposed into multiple single-objective optimization sub-problems, and the multiple single-objective optimization sub-problems are screened to select the sub-problem with the best potential, thereby improving the optimization efficiency, and then an updated data set is obtained by constructing a proxy model for the multiple single-objective optimization sub-problems, and the updated data set is processed based on the preset clustering algorithm to obtain the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters, thereby ensuring the representativeness and diversity of the obtained optimal parameters and optimal targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 A schematic flow chart of a chip packaging design optimization method based on an adaptive sub-problem selection strategy provided in an embodiment of the present invention.

[0043] Figure 2 A cross-sectional view of a finite element model of a chip package to be optimized according to an implementation manner provided in an embodiment of the present invention.

[0044] Figure 3A schematic diagram of a coordinate system of a finite element model of a chip package to be optimized according to an implementation manner provided in an embodiment of the present invention.

[0045] Figure 4 A schematic diagram of simulation of the first target warping in the finite element analysis model provided in an embodiment of the present invention.

[0046] Figure 5 A schematic diagram of the simulation of the second target von Mises stress in the finite element analysis model provided in an embodiment of the present invention.

[0047] Figure 6 A principle block diagram of a chip packaging design optimization device based on an adaptive sub-problem selection strategy provided by an embodiment of the present invention.

[0048] Figure 7 This is a block diagram of the internal structure principle of the intelligent terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention discloses a chip package design optimization method based on an adaptive sub-problem selection strategy. In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0050] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0051] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as herein.

[0052] In the prior art, the chip packaging design method cannot achieve the optimization of multiple parameters for multiple objectives.

[0053] In order to solve the problems of the prior art, the present embodiment provides a chip packaging design optimization method based on an adaptive sub-problem selection strategy, which decomposes the optimization problem of the chip packaging design objective into multiple single-objective optimization sub-problems, and selects the sub-problem with the best potential by screening the multiple single-objective optimization sub-problems to improve the optimization efficiency. Then, an updated data set is obtained by constructing an agent model for the multiple single-objective optimization sub-problems, and the updated data set is processed based on a preset clustering algorithm to obtain the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters, thereby ensuring the representativeness and diversity of the obtained optimal parameters and optimal targets. In the specific implementation, the initial parameters of the chip packaging design are first obtained, and finite element analysis is performed on several of the initial parameters based on preset finite element simulation software to obtain prediction results of several targets to be optimized, and the prediction results of several initial parameters and several targets to be optimized are used as the initial data set; then, based on the several targets to be optimized and the several initial parameters, a multi-objective optimization problem is constructed, the multi-objective optimization problem is decomposed to obtain several single-objective optimization sub-problems, and based on the preset selection strategy and the initial data set, several of the single-objective optimization sub-problems are screened to obtain several candidate single-objective optimization sub-problems; then, based on the initial data set and several of the candidate single-objective optimization sub-problems, a proxy model is constructed, and based on the proxy model and the initial data set, an updated data set is obtained; finally, based on the preset clustering algorithm, according to the updated data set, the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters are obtained.

[0054] Example

[0055] Chip packaging is a process that uses membrane technology and micro-machining technology to arrange, fix and connect chips and other electronic components on a frame or substrate, lead out the wiring terminals and fix them through potting with plastic insulating media to form an overall structure. It plays the role of placing, fixing, sealing, and protecting circuit chips from physical and chemical influences of the surrounding environment, and enhancing electrical and thermal performance.

[0056] Fine Pitch Ball Grid Array (fpBGA) is an array-molded, cost-effective, space-saving, laminate-based chip-scale packaging method with a plastic overmolded package and a series of fine-pitch solder ball terminals. fpBGA is a widely used package for space-constrained applications such as mobile and handheld computing devices. The smaller size, lower cost, and higher density options of fpBGA make it an ideal advanced technology packaging solution for high-performance or portable applications.

[0057] In the chip packaging process, due to the large differences in the thermodynamic performance indicators of various materials such as substrates, chips, adhesives used to bond substrates and chips, and epoxy resins, the device will produce large and unevenly distributed stress and strain in the cooling process from the molding temperature to room temperature, causing uneven device warping, which seriously affects the reliability, welding performance and yield rate of electronic devices. This problem is more prominent under the trend of precision and miniaturization of packaging.

[0058] Therefore, an efficient design optimization method is proposed for fpBGA chip packaging, which has very important engineering significance for achieving high reliability and cost-effective chip-level packaging design.

[0059] Exemplary Methods

[0060] This embodiment provides a chip packaging design optimization method based on an adaptive sub-problem selection strategy, which can be applied to an artificial intelligence smart terminal. Figure 1 As shown, the method includes:

[0061] Step S100, obtaining initial parameters of chip package design, and performing finite element analysis on several of the initial parameters based on preset finite element simulation software to obtain prediction results of several targets to be optimized, and using the several initial parameters and the prediction results of several targets to be optimized as initial data sets;

[0062] In practice, based on the chip package design to be optimized, a finite element structure is established in the finite element simulation software. The finite element structure components include a substrate, a chip, a binder between the substrate and the chip, and an epoxy resin for encapsulating the chip. Figure 2 As shown in the "fpBGA cross-section diagram" in the figure, it includes the substrate, the epoxy resin encapsulating the chip, the adhesive between the substrate and the chip, and the chip. Establish a spatial rectangular coordinate system, including the origin O, the X direction, the Y direction, and the Z direction. Figure 3As shown in the "coordinate diagram" in the figure, the X and Y directions of the entire device are square. Along the positive direction of the z-axis, there are substrate, adhesive, chip, and epoxy resin in turn. The origin O is at the center of the bottom surface of the substrate. The side length of the substrate is 90mm, the side length of the adhesive is 40mm, the side length of the chip is 40mm, and the side length of the epoxy resin is 90mm, covering the chip and the adhesive. Set the constraint relationship, which includes setting binding constraints on the contact surfaces between the substrate and the adhesive, the adhesive and the chip, the substrate and the epoxy resin, the adhesive and the epoxy resin, and the chip and the epoxy resin. Set the boundary conditions: fix the four right-angled tops of the bottom surface of the substrate, set the load: set the temperature load, the initial temperature is 175°C, and the end temperature is 25°C. Set the finite element software analysis output result: set the amount that needs to be output after the software performs simulation calculations. The data that needs to be output in this embodiment are the von Mises stress and deformation (used to calculate the warpage) of each mesh unit and node of the model under the action of the temperature load. Set the material properties, including Young's modulus, Poisson's ratio, and coefficient of thermal expansion (CTE) of epoxy resin, substrate, chip, and adhesive, as shown in Table 1.

[0063] Table 1

[0064]

[0065] Then the engineer defines the optimization goal, initial design parameters (ie variables) and design constraints in advance, and inputs them into the finite element simulation software, which can be obtained by the finite element simulation software. In this embodiment, the optimization goal can be the warpage and von Mises stress of the chip, which are denoted as Q and F respectively. The warpage of the chip is specifically defined as the absolute value of the difference between the deformation of the center of the bottom surface of the substrate in the Z-axis direction and the deformation of the right-angled top of the bottom surface of the substrate in the Z-axis direction. The von Mises stress of the chip is defined as the maximum value of the von Mises stress at each point inside the entire device during the cooling process from 175°C to 25°C. The optimization goal is to minimize the warpage and von Mises stress of the chip during the cooling process. The ranges of warpage and stress are defined as [0mm, 0.025mm], [130N, 230N] respectively. There are several initial parameters, which can be the thickness of EMC, denoted as x_1, with a value range of [0.55mm, 0.95mm]; the thickness of the substrate, denoted as x_2, with a value range of [0.2mm, 0.3mm]; the thickness of the chip, denoted as x_3, with a value range of [0.2mm, 0.32mm]; the thickness of the adhesive, denoted as x_4, with a value range of [0.02mm, 0.04mm]; the thermal expansion coefficient (CTE) of EMC, denoted as x_5, with a value range of [8ppm / ℃, 12ppm / ℃]. The domain (i.e., value range) of the five initial parameters is denoted as Ω. In order to avoid the influence of different dimensions on different parameters, we use the normalization method, and the specific formula is as follows:

[0066] P n =(PP l ) / (P u -P l )

[0067] Among them, P n is the normalized value, P is the original parameter, P u ,P l are the upper and lower bounds respectively.

[0068] The design constraints are the initial parameter constraints and the number of algorithm iterations N iter , the initial parameter constraint refers to the value range of the initial parameters mentioned above. Let the d initial parameters be (x1,…,x d ) T . The domain formed by the value ranges of the n initial parameters is denoted as Ω. Based on the preset finite element simulation software, finite element analysis is performed on several of the initial parameters to obtain prediction results of several targets to be optimized. In this embodiment, the initial parameters (EMC thickness, substrate thickness, chip thickness and adhesive thickness) are input into the finite element analysis software to obtain the warpage value and the von Mises stress value, and then the prediction results of several of the initial parameters and several of the targets to be optimized are used as the initial data set. If there is no data set, a sampling method is used to randomly generate a set of design variables in the design space. The sampling method can be Latin hypercube sampling or random sampling, and the data is sent to the finite element analysis software for analysis to obtain the prediction results of several targets to be optimized (that is, the results of different optimization targets). The initial data set contains 11d-1 data, denoted as Each data consists of a set of initial and corresponding prediction results of the target to be optimized.

[0069] After obtaining the initial data set, you can execute Figure 1 The following steps are shown: S200, constructing a multi-objective optimization problem according to the plurality of objectives to be optimized and the plurality of initial parameters, decomposing the multi-objective optimization problem to obtain a plurality of single-objective optimization sub-problems, and screening the plurality of single-objective optimization sub-problems based on a preset selection strategy and the initial data set to obtain a plurality of candidate single-objective optimization sub-problems;

[0070] In step S200, constructing a multi-objective optimization problem based on the several objectives to be optimized and the several initial parameters includes the following steps: normalizing the several objectives to be optimized to obtain several normalized objectives; constraining the several normalized objectives; and obtaining a multi-objective optimization problem based on the several initial parameters and the several normalized objectives that have been constrained.

[0071] To avoid bias caused by different dimensions, we normalize each target to obtain several normalized targets; constrain the normalized targets, that is, set their corresponding value ranges. Then, we obtain the multi-objective optimization problem based on the initial parameters and the normalized targets that have been constrained:

[0072] Minimize f(x)=(f1(x),…,f n (x) T ,

[0073]

[0074] Among them, Ω is the domain of the value range of n design variables, (x1,…,x d ) T are d initial parameters, R d means: R is a set of real numbers, d is the number of design variables, R d It is the d-dimensional real number space.

[0075] After obtaining the multi-objective optimization problem, the multi-objective optimization problem needs to be decomposed to obtain a number of single-objective optimization sub-problems; accordingly, the decomposition of the multi-objective optimization problem to obtain a number of single-objective optimization sub-problems includes the following steps: based on the Chebyshev algorithm, the multi-objective optimization problem is decomposed into a number of single-objective optimization sub-problems.

[0076] Specifically, sub-problem decomposition methods other than using the Chebyshev algorithm are also within the scope of the present invention. In this embodiment, the Chebyshev method decomposes the multi-objective optimization problem into a series of single-objective sub-problems. The Chebyshev method defines the sub-problems as follows:

[0077] minimize

[0078] x∈Ω

[0079] in is the reference vector, set to is for all i=1,…, and Satisfy w i ≥0, so we get a set of N subproblems, denoted as where g i is the i-th subproblem. Regarding the weight vector, we can define a direction line passing through the reference point and the optimal solution of the subproblem, expressed as:

[0080] z=φv+z *

[0081] Where φ is the proportionality coefficient, v=(v1,…,vm ) T is the direction vector. k Expressed as:

[0082]

[0083] In this embodiment, the direction vector is made to evenly divide the entire sub-problem space. In addition, the method of using other methods to generate evenly distributed direction vectors is also within the protection scope of the present invention. According to the above formula, the denominator is a constant, and each component of the direction vector is known, thereby obtaining each component of the weight vector.

[0084] After obtaining a number of single-objective optimization subproblems, based on a preset selection strategy and the initial data set, the single-objective optimization subproblems are screened to obtain a number of candidate single-objective optimization subproblems. Accordingly, the screening of the single-objective optimization subproblems based on the preset selection strategy and the initial data set to obtain a number of candidate single-objective optimization subproblems includes the following steps: for each target to be optimized, calculating the distance between the vector corresponding to the target to be optimized and each single-objective optimization subproblem, obtaining the minimum distance between the vectors corresponding to the target to be optimized, and pairing the single-objective optimization subproblem corresponding to the minimum distance with the target to be optimized; deleting the single-objective optimization subproblems that fail to pair, and classifying the single-objective optimization subproblems paired with each target to be optimized into the first subproblem set; for each subproblem in the first subproblem set, Extract the first non-dominated solution set in the initial data set, and sort the non-dominated solutions in the first non-dominated solution set in descending order based on the distance to obtain a sorting number for each non-dominated solution; obtain a solution degree value for each sub-problem based on the first non-dominated solution set, the sorting number and the distance, wherein the solution degree value is used to characterize the degree of problem solution; classify the sub-problems whose solution degree values ​​are less than a first preset threshold into a second sub-problem set, and delete the sub-problems whose solution degree values ​​are less than the first preset threshold in the first sub-problem set to obtain a third sub-problem set; when the number of sub-problems in the second sub-problem set is less than the second preset threshold, classify the sub-problem with the smallest solution degree value in the third sub-problem set into the second sub-problem set, and use the sub-problems in the second sub-problem set as several candidate single-objective optimization sub-problems.

[0085] Specifically, for each target to be optimized, the distance between the vector corresponding to the target to be optimized and each single-target optimization sub-problem is calculated, that is, the distance between each target vector f i , i = 1, ..., n and each sub-problem g j ,j=1,…,N distance matrix D. Each element d in D ij The calculation formula is:

[0086]

[0087] in, Yes j The direction vector of .

[0088] Then, the minimum distance of the vector corresponding to the target to be optimized is obtained, and the single-objective optimization sub-problem corresponding to the minimum distance is paired with the target to be optimized; for example, each target vector f i Assign it to the subproblem closest to it, delete the single-objective optimization subproblem that fails to pair, and classify the single-objective optimization subproblem paired with each target to be optimized into the first subproblem set G I ; For example: the sub-problems that are not assigned are initially screened out, and the sub-problems that are assigned to the target vector form a new sub-problem set, denoted as G I . After deleting the single-objective optimization subproblems that failed to pair, and classifying the single-objective optimization subproblems that are paired with each target to be optimized into the first subproblem set, the following steps are included: randomly selecting a number of single-objective optimization subproblems from the single-objective optimization subproblems that failed to pair, and expanding the randomly selected single-objective optimization subproblems to the first subproblem set. For example: At the same time, in order to prevent the influence of misjudgment of correlation to a certain extent, randomly select 5 subproblems that are screened out this time and add them to G I For each subproblem in the first subproblem set, extract the first non-dominated solution set in the initial data set, and sort the non-dominated solutions in the first non-dominated solution set in descending order based on the distance D to obtain the sorting number r of each non-dominated solution. i,j , based on the first non-dominated solution set, the sorting sequence number and the distance, obtain the solution degree value SD of each sub-problem i , S.D. i The calculation formula is:

[0089]

[0090] Wherein n is a control parameter, which is 100 in this embodiment; ST i Yes i The number of times G is selected during the iterative optimization process. I The solution value SD i Sub-problems with a value less than the first preset threshold (such as 1) are classified into the second sub-problem set G II , and delete the sub-problems whose solution degree values ​​in the first sub-problem set are less than the first preset threshold value to obtain the third sub-problem set; the third sub-problem set contains the remaining sub-problems in the first sub-problem set; when the number of sub-problems in the second sub-problem set is less than the second preset threshold value T, the sub-problem with the smallest solution degree value in the third sub-problem set is classified into the second sub-problem set G II , until the second sub-problem set G IIThe number of subproblems is T, where T is the number of pre-set subproblems after screening, and is set to 5 in this method. In this embodiment, the subproblems in the second subproblem set are used as several candidate single-objective optimization subproblems. Then, the K-means algorithm is used to select G II Divide into T subsets to avoid the subproblems to be selected being too close. Select the subproblem with the smallest SD in each subset, and call the set of the selected T subproblems G s . s Each subproblem in the T data as training set. T The data consists of two parts, 90% of the data is in g i The remaining 10% of the data are randomly selected from the rest of the data. T Set to the minimum value of the current data number and 100.

[0091] After obtaining several candidate single-objective optimization sub-problems, we can perform the following Figure 1 The following steps are shown: S300, constructing a proxy model according to the initial data set and the plurality of candidate single-objective optimization sub-problems, and obtaining an updated data set according to the proxy model and the initial data set;

[0092] In step S300, the proxy model can be an independent Gaussian process model, a multi-task Gaussian process model, a Bayesian neural network, or a radial basis function. The proxy model is used to mathematically model the initial data set and several of the candidate single-objective optimization sub-problems. In this embodiment, the proxy model uses CoMOGP in the multi-task Gaussian process model to mathematically model several of the candidate single-objective optimization sub-problems, specifically: the CoMOGP model uses Q shared random processes to help share knowledge between different sub-problems, and also uses T independent random processes for T sub-problems to capture the unique information of each sub-problem. For T sub-problems with unknown similarities, the CoMOGP proxy model uses Q shared processes to assist in information transfer between sub-problems, and uses T specific output processes to capture the individual features of each sub-problem. This solves the problem that the general multi-task Gaussian process model has poor performance when the sub-problems have low similarity. The covariance function represents the covariance between each pair of decision vectors, and the exponential square (SE) covariance function is usually used:

[0093]

[0094] in, is the signal variance, Including the characteristic length scale, x and x' are both design variables. The functional form of various k(x,x') refers to the calculation of the covariance function for any two design variables x and x'.

[0095] For t, t′=1,…,T, subproblem g t (x) and g t′ The covariance function between (x′) is defined as follows:

[0096]

[0097] where a t,q and a t′,q is the relevant parameter. k tt′ The value of (x,x′) reflects g t (x) and g t′ The correlation degree between (x′), k t (x,x′) is the covariance function used by T independent random processes, k q (x, x′) is the covariance function used by the shared random process. In this embodiment, both covariance functions use the exponential square covariance function k SE The T sub-problems can be expressed as Assume that they follow a Gaussian random process:

[0098]

[0099] The MOGP covariance of T×T is K g (x,x′) is defined as:

[0100]

[0101] Given a data set of T sub-problems Using a matrix of nT×d and the nT×1 vector Represent the initial parameters and target variables of the training data respectively. The covariance matrix K y It is expressed as:

[0102]

[0103] Among them, K tt′ (X t ,X t′ ) is an n×n matrix. For t, t′=1,…,T,x i ∈X t And x j ∈X t′ , I n It is an identity matrix with 1 on the diagonal and 0 on the rest of the elements. The (i,j) element in the matrix represents k tt′ (x i ,x j ); is a T×T diagonal noise matrix, represents the Kronecker product.

[0104] After the proxy model is obtained, an updated data set is obtained according to the proxy model and the initial data set; correspondingly, the step of obtaining the updated data set according to the proxy model and the initial data set includes the following steps: obtaining predicted posterior information of the proxy model according to the proxy model, and obtaining recommended parameters through a preset acquisition function according to the predicted posterior information; wherein the predicted posterior information is used to characterize the posterior distribution of the proxy model; expanding the recommended parameters to several of the initial parameters to obtain several updated initial parameters, iteratively performing finite element analysis on several of the initial parameters based on preset finite element simulation software to obtain predicted results of several targets to be optimized, and using the predicted results of several of the initial parameters and several of the targets to be optimized as the initial data set, until a preset condition is met and the iteration is stopped; replacing the initial parameters in the initial data set with the expanded initial parameters to obtain an updated data set.

[0105] Specifically, the probabilistic proxy model used in this embodiment is essentially a Bayesian statistical inference method, based on existing data to calculate a posterior distribution information that obeys a Gaussian distribution. The posterior distribution given by the proxy model has a mean function and a variance function. The mean function can be considered as a prediction of the objective function under any design parameters based on the current data, and the variance function reflects the uncertainty of this predicted value under any design parameters. The mean function and the variance function are information obtained by the proxy model established using real data (design parameters and finite element real results), which can be considered as the output of the proxy model. The output here is in the form of two functions, that is, the predicted value of the objective function and the uncertainty of this predicted value can be given for any design parameter. The recommendation of the next set of design parameters for simulation analysis requires a comprehensive prediction value and the uncertainty of the prediction, which is reflected by the acquisition function. The acquisition function will use the function form of the posterior mean and variance. By maximizing the acquisition function, the parameter corresponding to the maximum value of the acquisition function is the next parameter recommended. Using the trained proxy model information, the next set of parameters for simulation analysis is recommended through the acquisition function. The acquisition function uses the proxy model information with low evaluation cost to make design parameter recommendations by balancing the tendency to explore unknown design areas and use known information; increasing the weight of the mean function in the acquisition function can be considered as using known information (believing the current prediction results), and increasing the weight of the variance function in the acquisition function can be considered as a preference for exploring unknown areas. The acquisition function can use other acquisition function forms, such as maximum expected improvement (EI), maximum probability improvement (PI), and any other acquisition function that can be applied to Gaussian process models. Since the proxy model information is used, the computational cost is low, and the classical optimization algorithm can be used to find the optimal use. In this embodiment, at a new sampling point x* The predicted mean and variance of are expressed as:

[0106]

[0107]

[0108] in is a T×nT matrix. For t=1,…,T,K ** =Kg ( x * ,x * ),matrix The Tth term in is expressed as Σ * is a T×T matrix, where the tth diagonal element corresponds to g t (x * ). That is, for t=1,…,T

[0109] To make predictions using the CoMOGP model, it is necessary to infer the hyperparameter θ in the covariance matrix and the additional parameter β in the mean function. The CoMOGP model used in the present invention has only one shared process, i.e., Q=1, and uses the SE covariance function. Therefore, θ contains T(T+1) / 2 related parameters, covariance parameters (T+1)(d+1) and noise parameters T. In addition, in order to improve the prediction accuracy of the model, the present invention uses a quadratic model as the common mean function of T output after using the CoMOGP model. β consists of 2d+1 parameters.

[0110] Similar to the single-output Gaussian process model, these parameters can be inferred by maximizing the marginal likelihood function. The formula is as follows:

[0111]

[0112] Wherein, X is the design variable matrix, y is the target variable vector, n is the dimension of X; θ is the vector representing the hyperparameters; Ky is the covariance matrix. In view of the availability of the partial derivatives of the marginal likelihood function, the present invention adopts an efficient gradient descent algorithm to solve, that is, the minimization function of the GPML toolbox. The function uses a conjugate gradient optimization algorithm with a maximum number of iterations of 500. For the initial values ​​of the parameters, the present invention sets the relevant parameters, covariance parameters and regression coefficients to 1, and the noise parameter to 0.01.

[0113] The predicted posterior information of the proxy model is obtained according to the proxy model, wherein the predicted posterior information is used to characterize the posterior distribution of the proxy model; for example: i ∈G S , given the predicted posterior distribution of the surrogate model The commonly used acquisition function ALCB is defined as:

[0114]

[0115] Among them, γ is a parameter that defines the degree of exploration. The larger the γ value, the more inclined to explore the unknown design space, and the smaller the γ value, the more inclined to trust the current optimal region as the optimal.

[0116] Then, according to the predicted posterior information, the recommended parameters are obtained through a preset acquisition function (in this embodiment, the acquisition function is the ALCB acquisition function), such as optimizing the original fixed γ and using adaptive values ​​instead of fixed values, which is called the ALCB acquisition function, which can effectively alleviate the phenomenon of reduced optimization efficiency caused by information bias in the multi-objective optimization process. The definition of γ is as follows:

[0117]

[0118] where |NP| represents the number of non-dominated solutions obtained so far, and

[0119] By minimizing the ALCB acquisition function through the predicted distribution information obtained by the CoMOGP model, we obtain T sampling points, that is, T groups of recommended parameters. The above process uses the Matlab programming language.

[0120] In this embodiment, the recommended five sets of parameters are stored together with the original parameters in a local csv file "design.csv". By modifying the corresponding file with the suffix "py" in the finite element model working file directory, it can automatically read the data and changes of "design.csv", set the design parameters in the model to the new recommended values ​​and perform simulation, and store the predicted results of the target to be optimized analyzed by the software in another csv file "target.csv". The iterative optimization algorithm file obtains the predicted results of the target to be optimized corresponding to the new parameters by reading the updated data in "target.csv".

[0121] Expand the recommended parameters to several initial parameters to obtain several updated initial parameters, and update the data set EP, the number of evaluations FE←FE+T, and the reference vector z * , the number of times each sub-question is selected ST i. Then determine whether the number of evaluations FE is less than the set maximum number of evaluations (such as 120 times). If the number of evaluations FE is less than the set maximum number of evaluations, iteratively perform a finite element analysis on the initial parameters based on the preset finite element simulation software to obtain the prediction results of the several targets to be optimized, and use the several initial parameters and the prediction results of the several targets to be optimized as the initial data set. If the number of evaluations FE is greater than or equal to the set maximum number of evaluations, stop the iteration. Replace the initial parameters in the initial data set with the expanded initial parameters to obtain an updated data set.

[0122] In one implementation, the updated initial parameters are automatically modified by a preset code, that is, the updated initial parameters are replaced by the original initial parameters by the code, and then the updated initial parameters are input into the finite element analysis software for simulation, and several updated prediction results of the target to be optimized are obtained, and the several updated prediction results of the target to be optimized are also input into the preset code. In this way, the entire processing process is a fully automatic iterative optimization process, saving time and labor costs in actual chip optimization design projects.

[0123] After getting the updated data set, you can execute Figure 1 The following steps are shown: S400, based on a preset clustering algorithm, according to an updated data set, obtaining optimal parameters of chip package design and optimal targets corresponding to the optimal parameters.

[0124] Step S400 includes the following steps: obtaining a second non-dominated solution set of the updated data set;

[0125] Based on a preset clustering algorithm, clustering the non-dominated solutions in the second non-dominated solution set to obtain cluster centers;

[0126] Calculate the distance between the parameters in the updated data set and the cluster center;

[0127] The parameter corresponding to the minimum distance is taken as the optimal parameter for chip packaging design;

[0128] The target to be optimized corresponding to the minimum distance is taken as the optimal target corresponding to the optimal parameter.

[0129] Specifically, a clustering algorithm is used to cluster the non-dominated solutions NP in the second non-dominated solution set. The number of clusters is the number of optimal solutions p that the user hopes to obtain, and the cluster center, that is, the center point of each category, is output. The center point is the most representative solution in the optimal solution set recommended to the user and retains diversity. Calculate the distance between the parameters in the updated data set and the cluster center; use the parameters corresponding to the minimum distance as the optimal parameters for chip packaging design; for example, output the parameters closest to the center point of each category and the corresponding prediction results of the target to be optimized as the final recommended design. Figure 4 is a warping simulation diagram of an embodiment of the present invention, such as Figure 5 This is a von Mises stress simulation diagram of an embodiment of the present invention.

[0130] The advantages of the present invention are:

[0131] 1. The present invention uses a low-computational-cost proxy model to simulate the relationship between design variables and optimization targets, and uses an heuristic acquisition function to recommend design parameters. Compared with traditional chip packaging methods, it can produce better optimization results in a shorter time and with less computational cost;

[0132] 2. The present invention introduces a solution for jointly optimizing multiple objectives in the machine learning model. Compared with other machine learning methods used in the packaging field, the present invention can solve more problems, and the coexistence of multiple objectives is more practical;

[0133] 3. The adaptive sub-problem selection strategy proposed in the present invention can identify and select the most promising sub-problems through two-step screening to further replace auxiliary optimization, avoiding the situation in which some unnecessary sub-problems or solved sub-problems are processed in existing similar methods, thereby greatly improving the optimization efficiency.

[0134] 4. The present invention introduces the CoMOGP model to better model the selected sub-problems. The CoMOGP model can not only capture the specific characteristics of each sub-problem, but also transfer useful information between sub-problems. Regardless of the similarity between sub-problems, the CoMOGP model can maintain good prediction quality.

[0135] 5. The present invention proposes a new acquisition function ALCB suitable for the current optimization framework, which can better balance the exploration and development processes in the multi-objective optimization scenario of chip packaging and improve the final optimization effect.

[0136] 6. The present invention uses a clustering algorithm to recommend the most representative solution that maintains diversity from a large number of optimization results.

[0137] 7. The process of the present invention utilizes the programmable characteristics of finite element software. Through data communication between finite element software and algorithm code, it not only provides an efficient optimization method, but also realizes a fully automatic optimization process, saving manpower and time costs in actual chip packaging design tasks.

[0138] Exemplary Devices

[0139] like Figure 6 As shown in , an embodiment of the present invention provides a chip packaging design optimization device based on an adaptive sub-problem selection strategy, the device includes an initial data set determination module 501, a single-objective optimization sub-problem screening module 502, an updated data set acquisition module 503 and an optimal parameter and optimal target determination module 504, wherein:

[0140] The initial data set determination module 501 is used to obtain initial parameters of the chip package design, and perform finite element analysis on the initial parameters based on the preset finite element simulation software to obtain prediction results of the targets to be optimized, and use the initial parameters and the prediction results of the targets to be optimized as the initial data set;

[0141] The single-objective optimization sub-problem screening module 502 is used to construct a multi-objective optimization problem according to the plurality of objectives to be optimized and the plurality of initial parameters, decompose the multi-objective optimization problem to obtain a plurality of single-objective optimization sub-problems, and screen the plurality of single-objective optimization sub-problems based on a preset selection strategy and the initial data set to obtain a plurality of candidate single-objective optimization sub-problems;

[0142] An updated data set acquisition module 503 is used to construct a proxy model according to the initial data set and the plurality of candidate single-objective optimization sub-problems, and obtain an updated data set according to the proxy model and the initial data set;

[0143] The optimal parameter and optimal target determination module 504 is used to obtain the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters based on the preset clustering algorithm and the updated data set.

[0144] Based on the above embodiments, the present invention further provides an intelligent terminal, whose principle block diagram can be shown as follows: Figure 7As shown. The intelligent terminal includes a processor, a memory, a network interface, a display screen, and a temperature sensor connected through a system bus. Among them, the processor of the intelligent terminal is used to provide computing and control capabilities. The memory of the intelligent terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the intelligent terminal is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a chip packaging design optimization method based on an adaptive sub-problem selection strategy is implemented. The display screen of the intelligent terminal can be a liquid crystal display screen or an electronic ink display screen, and the temperature sensor of the intelligent terminal is pre-set inside the intelligent terminal to detect the operating temperature of the internal device.

[0145] Those skilled in the art will understand that Figure 7 The schematic diagram in the figure is only a block diagram of a part of the structure related to the scheme of the present invention, and does not constitute a limitation on the smart terminal to which the scheme of the present invention is applied. The specific smart terminal may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0146] In one embodiment, a smart terminal is provided, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, and the one or more programs include instructions for performing the following operations:

[0147] Acquire initial parameters of chip package design, and perform finite element analysis on some of the initial parameters based on preset finite element simulation software to obtain prediction results of some targets to be optimized, and use the initial parameters and the prediction results of some targets to be optimized as initial data sets;

[0148] According to the plurality of objectives to be optimized and the plurality of initial parameters, a multi-objective optimization problem is constructed, the multi-objective optimization problem is decomposed to obtain a plurality of single-objective optimization sub-problems, and based on a preset selection strategy and the initial data set, the plurality of single-objective optimization sub-problems are screened to obtain a plurality of candidate single-objective optimization sub-problems;

[0149] Constructing a proxy model according to the initial data set and the plurality of candidate single-objective optimization sub-problems, and obtaining an updated data set according to the proxy model and the initial data set;

[0150] Based on a preset clustering algorithm and according to an updated data set, optimal parameters of chip packaging design and optimal targets corresponding to the optimal parameters are obtained.

[0151] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0152] In summary, the present invention discloses a chip packaging design optimization method based on an adaptive sub-problem selection strategy, characterized in that the method includes: obtaining a number of initial parameters of the chip packaging design, and performing finite element analysis on the initial parameters to obtain prediction results of a number of targets to be optimized, thereby obtaining an initial data set; constructing a multi-objective optimization problem according to the several targets to be optimized and the initial parameters, decomposing the multi-objective optimization problem to obtain a number of single-objective optimization sub-problems, and based on a preset selection strategy, obtaining a number of candidate single-objective optimization sub-problems by screening according to the initial data set; constructing a proxy model according to the initial data set and the number of candidate single-objective optimization sub-problems, thereby obtaining an updated data set; based on a preset clustering algorithm, obtaining the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters according to the updated data set. The present invention can achieve the optimization of multiple parameters for multiple targets.

[0153] Based on the above embodiments, the present invention discloses a chip packaging design optimization method based on an adaptive sub-problem selection strategy. It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, it can be improved or transformed according to the above description. All these improvements and transformations should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A chip packaging design optimization method based on an adaptive sub-problem selection strategy, characterized in that: The method comprises: Acquire several initial parameters of the chip package design, and perform finite element analysis on the several initial parameters based on preset finite element simulation software to obtain prediction results of several targets to be optimized, and use the several initial parameters and the prediction results of the several targets to be optimized as initial data sets; According to the plurality of objectives to be optimized and the plurality of initial parameters, a multi-objective optimization problem is constructed, the multi-objective optimization problem is decomposed to obtain a plurality of single-objective optimization sub-problems, and based on a preset selection strategy and the initial data set, the plurality of single-objective optimization sub-problems are screened to obtain a plurality of candidate single-objective optimization sub-problems; Constructing a proxy model according to the initial data set and the plurality of candidate single-objective optimization sub-problems, and obtaining an updated data set according to the proxy model and the initial data set; Based on a preset clustering algorithm and according to an updated data set, optimal parameters of chip package design and optimal targets corresponding to the optimal parameters are obtained; The step of obtaining an updated data set according to the proxy model and the initial data set includes: Obtaining predicted posterior information of the proxy model according to the proxy model, and obtaining recommended parameters through a preset acquisition function according to the predicted posterior information; wherein the predicted posterior information is used to characterize the posterior distribution of the proxy model; Expanding the recommended parameters to the initial parameters to obtain the updated initial parameters, iteratively performing finite element analysis on the initial parameters based on the preset finite element simulation software to obtain the prediction results of the targets to be optimized, and taking the initial parameters and the prediction results of the targets to be optimized as the initial data set, until the preset conditions are met, and stopping the iteration; Replacing the initial parameters in the initial data set with the expanded initial parameters to obtain an updated data set; The method of obtaining the optimal parameters of the chip package design and the optimal target corresponding to the optimal parameters based on the preset clustering algorithm and the updated data set includes: Obtaining a second non-dominated solution set of the updated data set; Based on a preset clustering algorithm, clustering the non-dominated solutions in the second non-dominated solution set to obtain cluster centers; Calculate the distance between the parameters in the updated data set and the cluster center; The parameter corresponding to the minimum distance is taken as the optimal parameter for chip packaging design; The target to be optimized corresponding to the minimum distance is taken as the optimal target corresponding to the optimal parameter.

2. The chip packaging design optimization method based on the adaptive sub-problem selection strategy according to claim 1 is characterized in that: The constructing of a multi-objective optimization problem according to the plurality of objectives to be optimized and the plurality of initial parameters comprises: Normalizing the plurality of targets to be optimized to obtain a plurality of normalized targets; Performing constraint processing on a number of the normalized targets; A multi-objective optimization problem is obtained according to the several initial parameters and the several normalized objectives processed with constraints.

3. The chip packaging design optimization method based on the adaptive sub-problem selection strategy according to claim 1 is characterized in that: Decomposing the multi-objective optimization problem to obtain several single-objective optimization sub-problems includes: Based on the Chebyshev algorithm, the multi-objective optimization problem is decomposed into several single-objective optimization sub-problems.

4. The chip packaging design optimization method based on the adaptive sub-problem selection strategy according to claim 1 is characterized in that: The single-objective optimization sub-problems are screened based on the preset selection strategy and the initial data set to obtain the candidate single-objective optimization sub-problems including: For each target to be optimized, calculate the distance between the vector corresponding to the target to be optimized and each single-objective optimization sub-problem, obtain the minimum distance between the vectors corresponding to the target to be optimized, and pair the single-objective optimization sub-problem corresponding to the minimum distance with the target to be optimized; The single-objective optimization sub-problems that fail to be paired are deleted, and the single-objective optimization sub-problems paired with each target to be optimized are classified into the first sub-problem set; For each subproblem in the first subproblem set, extract the first non-dominated solution set in the initial data set, sort the non-dominated solutions in the first non-dominated solution set in descending order based on the distance, and obtain the sorting sequence number of each non-dominated solution; obtain the solution degree value of each subproblem based on the first non-dominated solution set, the sorting sequence number and the distance, wherein the solution degree value is used to characterize the degree of solution to the problem; classify the subproblems whose solution degree values ​​are less than a first preset threshold into the second subproblem set, and delete the subproblems whose solution degree values ​​are less than the first preset threshold in the first subproblem set, to obtain the third subproblem set; When the number of subproblems in the second subproblem set is less than a second preset threshold, the subproblem with the smallest solution degree value in the third subproblem set is classified into the second subproblem set, and the subproblems in the second subproblem set are used as several candidate single-objective optimization subproblems.

5. The chip packaging design optimization method based on the adaptive sub-problem selection strategy according to claim 4 is characterized in that: The step of deleting the single-objective optimization sub-problems that fail to pair and classifying the single-objective optimization sub-problems that pair with each target to be optimized into the first sub-problem set includes: A number of single-objective optimization sub-problems are randomly selected from the single-objective optimization sub-problems that fail to pair, and the randomly selected number of single-objective optimization sub-problems are expanded to the first sub-problem set.

6. A chip packaging design optimization device based on an adaptive sub-problem selection strategy, characterized in that: The device comprises: An initial data set determination module is used to obtain initial parameters of chip packaging design, and perform finite element analysis on several of the initial parameters based on preset finite element simulation software to obtain prediction results of several targets to be optimized, and use the several initial parameters and the prediction results of several targets to be optimized as the initial data set; A single-objective optimization sub-problem screening module is used to construct a multi-objective optimization problem according to the plurality of objectives to be optimized and the plurality of initial parameters, decompose the multi-objective optimization problem to obtain a plurality of single-objective optimization sub-problems, and screen the plurality of single-objective optimization sub-problems based on a preset selection strategy and the initial data set to obtain a plurality of candidate single-objective optimization sub-problems; An updated data set acquisition module is used to construct a proxy model based on the initial data set and the plurality of candidate single-objective optimization sub-problems, and obtain an updated data set based on the proxy model and the initial data set; An optimal parameter and optimal target determination module, used to obtain the optimal parameters of the chip packaging design and the optimal targets corresponding to the optimal parameters based on a preset clustering algorithm and an updated data set; The step of obtaining an updated data set according to the proxy model and the initial data set includes: Obtaining predicted posterior information of the proxy model according to the proxy model, and obtaining recommended parameters through a preset acquisition function according to the predicted posterior information; wherein the predicted posterior information is used to characterize the posterior distribution of the proxy model; Expanding the recommended parameters to the initial parameters to obtain the updated initial parameters, iteratively performing finite element analysis on the initial parameters based on the preset finite element simulation software to obtain the prediction results of the targets to be optimized, and taking the initial parameters and the prediction results of the targets to be optimized as the initial data set, until the preset conditions are met, and stopping the iteration; Replacing the initial parameters in the initial data set with the expanded initial parameters to obtain an updated data set; The method of obtaining the optimal parameters of the chip package design and the optimal target corresponding to the optimal parameters based on the preset clustering algorithm and the updated data set includes: Obtaining a second non-dominated solution set of the updated data set; Based on a preset clustering algorithm, clustering the non-dominated solutions in the second non-dominated solution set to obtain cluster centers; Calculate the distance between the parameters in the updated data set and the cluster center; The parameter corresponding to the minimum distance is taken as the optimal parameter for chip packaging design; The target to be optimized corresponding to the minimum distance is taken as the optimal target corresponding to the optimal parameter.

7. An intelligent terminal, characterized in that: The device comprises a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, and the one or more programs include being used to execute the method according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method as claimed in any one of claims 1 to 5.

Images