A multi-physical field driven mini LED chip layout and wiring adaptive weight particle swarm optimization method

Abstract

The application belongs to the technical field of electronic design automation, and discloses a multi-physical field driven MiniLED chip layout and wiring adaptive weight particle swarm optimization method, which comprises the following steps: initial layout modeling and preprocessing are performed on a model to be optimized to obtain initial optimization index data; the initial optimization index data is normalized to construct a multi-objective optimization objective function; a leading contradiction coefficient is calculated to generate index initial weights, and an inner layer PSO algorithm is used for optimization and recombination to obtain optimized weights; the optimized weights are transmitted to an outer layer PSO algorithm for layout optimization; when the fitness function continuously stagnates, the inner layer PSO is restarted for weight recombination; and when the difference between the fitness of the fitness function at two times satisfies a preset condition, it is indicated that the fitness function converges, and the optimization is completed. The application can dynamically adjust the weights of various indexes according to the performance state of the current population, so that the search direction is always corrected around the main contradiction.

Description

Technical Field

[0001] This invention belongs to the field of electronic design automation technology, specifically relating to a multi-physics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing. Background Technology

[0002] With the continuous increase in power density of electronic devices, increasingly stringent requirements for energy system efficiency, and the global advancement of sustainable development strategies, thermal management has become a key bottleneck restricting the development of many high technologies. Thermal optimization, as a multidisciplinary field that uses systematic design, control, and material selection to achieve efficient heat transfer, distribution, and utilization, is becoming increasingly important.

[0003] MiniLED chip integrated embedded packaging (MEP) technology is applied to MiniLED direct-view display manufacturing, providing the industry with a brand-new technical route. The innovative solution employs a double-layer MEP substrate structure, embedding the driver IC within the substrate, effectively overcoming the bottlenecks of backlight chip placement and high-density wiring in ultra-small pitch scenarios. However, high-density chip placement can easily lead to problems such as insufficient signal integrity, weak heat dissipation management, and poor mechanical reliability. The layout and wiring design must take into account the multi-physics constraints of temperature field and thermal stress field coupling.

[0004] The core objective of chip thermal management is to maximize the system's heat dissipation performance (e.g., minimizing maximum temperature and temperature uniformity) under given constraints (such as space, weight, cost, and pump power consumption), or conversely, to minimize resource consumption while meeting heat dissipation requirements. Essentially, it is a complex system design problem involving multiple physics, multiple objectives, and multiple constraints. Traditional trial-and-error methods and empirical design are no longer sufficient to meet the needs of modern high-end equipment; model- and data-based systematic optimization methods have become the mainstream research approach.

[0005] Currently, the industry's response to chip thermal challenges has formed two major technical sectors: "thermal analysis" and "thermal sensing layout optimization." In thermal analysis, large-scale chip thermal analysis faces a sharp contradiction between accuracy and speed. Full-physics simulations such as the finite element method are too time-consuming, while acceleration methods based on reduced-order models or surrogate networks introduce non-negligible errors under non-uniform power distributions. In thermal sensing layout optimization, many researchers have transformed multi-objective optimization problems into single-objective function optimizations using weighted summation. The magnitude of the objective function is influenced by the weight values ​​and index values. Previous research has some limitations, especially in the selection of the optimal weight combination in multi-objective problems, where there is a significant degree of subjectivity.

[0006] While there is considerable research on the simultaneous thermal management and layout optimization of chips, existing methods often fail to accurately characterize the failure risk caused by thermal stress when simultaneously optimizing multiple conflicting objectives such as thermal performance and layout area. This is because they do not consider chip deformation and warpage. Furthermore, the approach of converting multi-objective optimization problems into single-objective function optimization through weighted summation relies heavily on empirically predetermined objective weights, lacking objective basis. Summary of the Invention

[0007] This invention aims to address the shortcomings of existing technologies and provides the following solutions:

[0008] A multi-physics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing includes the following steps:

[0009] The initial layout modeling and preprocessing of the model to be optimized are performed to obtain initial optimization index data;

[0010] The initial optimization index data are normalized to construct a multi-objective optimization objective function;

[0011] Calculate the dominant contradiction coefficient, generate the initial weights of the indicators, and use the inner PSO algorithm to optimize and reorganize the initial weights of the indicators to obtain the optimized weights;

[0012] The optimized weights are passed to the outer PSO algorithm for layout optimization. When the fitness function remains stagnant, the inner PSO algorithm is restarted for weight reorganization.

[0013] When the difference between the fitness values ​​of the fitness function before and after two iterations meets a preset condition, it indicates that the fitness function has converged and the optimization is complete.

[0014] Preferably, the method for performing the initial layout modeling includes:

[0015] The geometric modeling of the model to be optimized is divided into model partitions;

[0016] Perform equivalent parameter processing on the partitioned model;

[0017] The processed model is meshed to complete the initial layout modeling.

[0018] Preferably, the initial optimization index data includes: temperature distribution, thermal displacement distribution, temperature uniformity, minimum external area of ​​the chip, and shortest line length of the chip layout.

[0019] Preferably, the method for obtaining the temperature distribution includes: constructing a transient heat conduction equation using the FDTD method, and solving for the temperature distribution using the transient heat conduction equation;

[0020] The transient heat conduction equation is:

[0021] ,

[0022] Where ρ represents the material density, C p Let T represent specific heat, K represent thermal conductivity, and g represent thermal power density. Let t represent the direction vector and t represent time. This represents the gradient operator.

[0023] Preferably, the method for obtaining the thermal displacement distribution includes: constructing a steady-state thermal displacement equation using the FDTD method, and solving the thermal displacement distribution using the steady-state thermal displacement equation;

[0024] The steady-state thermal displacement equation is:

[0025] ,

[0026] ,

[0027] Where u and v represent the displacements in the x and y directions of the grid, respectively, T' represents the temperature change, G represents the shear modulus, L represents the Lamy constant, and B represents the thermodynamic coefficient.

[0028] Preferably, the multi-objective optimization objective function is:

[0029] ,

[0030] in, T(x,y) represents the maximum temperature generated on the substrate, derived from the temperature distribution. Disp(x,y) represents the maximum displacement generated on the substrate, derived from the thermal displacement distribution. This indicates temperature uniformity, where N represents the number of grid cells. This represents the temperature of the i-th grid. The area represents the average temperature of all grids, and the area represents the minimum external area of ​​all chips. This represents the maximum value of the x-axis coordinate of the right boundary of all chips. This represents the minimum x-axis coordinate of the left boundary of all chips. This represents the maximum value of the y-axis coordinate of the upper boundary of all chips. Length represents the minimum y-axis coordinate of the lower boundary of all chips, Length represents the minimum distance between the center points of all chips, and Length(line) represents the shortest line length of the chip layout.

[0031] Preferably, the method for generating the initial weights of the indicator includes:

[0032] Construct a pressure index based on the objective function. and stress index :

[0033] ,

[0034] ,

[0035] ,

[0036] in, This represents the preset expected value of the maximum temperature. This represents the preset expected value of the maximum displacement. This represents the preset expected value for temperature uniformity. This represents the preset expected value of the minimum circumscribed area. This represents the preset expected value of the minimum distance between the center points of all chips. This represents the combined pressure in terms of temperature, thermal deformation, and temperature uniformity. This indicates the combined pressure in terms of layout area and line length. Pressure indicating maximum temperature The pressure representing the maximum displacement. Pressure indicating temperature uniformity This indicates the pressure on chip layout area. Pressure indicating the length of the interconnect between chips;

[0037] Based on the pressure index and the pressure index Calculate the dominant temperature coefficient:

[0038] ,

[0039] Where η represents the temperature-dominant coefficient;

[0040] The initial reference target weights are obtained based on the temperature-dominant coefficient:

[0041] ,

[0042] in, Indicates the initial weight of the indicator. This indicates that temperature dominates the preset weight. This indicates that area is the dominant preset weight;

[0043] The initial weights of the indicator are obtained based on the initial reference target weights:

[0044] ,

[0045] in, Indicates the initial weight of the indicator. Indicates the reference weight.

[0046] Preferably, the fitness function is:

[0047] , ,

[0048] in, Represents the overall fitness function. The weight value representing the maximum temperature. The weight value represents the maximum displacement. The weight value represents the minimum distance between the center points of all chips. This represents the weight value of the minimum external area of ​​all chips. Weighting values ​​representing temperature uniformity.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0050] Compared to fixed-weight optimization methods, this algorithm dynamically adjusts the weights of each indicator based on the current performance state of the population, ensuring that the search direction is always corrected around the primary challenge. Because it can reduce search bias caused by unreasonable weight allocation through weight reorganization during the optimization process and redirect particles to better regions during stagnation phases, it can achieve a smaller overall objective function value. Attached Figure Description

[0051] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0053] Figure 2 This is a flowchart illustrating the method according to an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the initial chip layout according to an embodiment of the present invention;

[0055] Figure 4 The diagram shows the layout after optimization by different optimization methods in the embodiments of the present invention, wherein (a) is fixed weight GA, (b) is fixed weight PSO, (c) is fixed weight SA, and (d) is adaptive weight PSO.

[0056] Figure 5This is a schematic diagram of the fitness curves of different optimization methods in embodiments of the present invention;

[0057] Figure 6 The diagram shows the temperature field distribution after optimization by different methods in the embodiments of the present invention, where (a) is the initial layout, (b) is the fixed weight GA, (c) is the fixed weight PSO, (d) is the fixed weight SA, and (e) is the adaptive weight PSO.

[0058] Figure 7 The diagram shows the displacement field distribution after optimization by different methods in the embodiments of the present invention, where (a) is the initial layout, (b) is the fixed weight GA, (c) is the fixed weight PSO, (d) is the fixed weight SA, and (e) is the adaptive weight PSO. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] Example 1:

[0062] In this embodiment, as Figure 1 , Figure 2 As shown, a multi-physics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing includes the following steps:

[0063] S1. Perform initial layout modeling and preprocessing on the model to be optimized to obtain initial optimization index data.

[0064] The method for initial layout modeling includes: dividing the geometric model of the model to be optimized into model partitions; performing equivalent parameter processing on the partitioned model; and meshing the processed model to complete the initial layout modeling.

[0065] Initial optimization metrics include: temperature distribution, thermal displacement distribution, temperature uniformity, minimum external chip area, and shortest chip layout line length.

[0066] Methods for obtaining the temperature distribution include: constructing a transient heat conduction equation using the FDTD method, and then solving for the temperature distribution using the transient heat conduction equation; the transient heat conduction equation is:

[0067] ,

[0068] Where ρ represents the material density, C p Let T represent specific heat, K represent thermal conductivity, and g represent thermal power density. Let t represent the direction vector and t represent time. This represents the gradient operator.

[0069] For a two-dimensional heat conduction problem, the above formula can be expanded as follows:

[0070] ,

[0071] By performing central differences in time and space on the heat conduction equation, and assuming that the heat source does not change with time, we can obtain the following:

[0072] ,

[0073] in:

[0074] ,

[0075] ,

[0076] ,

[0077] ,

[0078] in, This indicates the temperature at the current moment. Indicates the temperature at the next moment. This represents the coefficient after temperature difference in the x-direction. This represents the coefficient after temperature difference in the y-direction. Operators representing temperature differences in the x-direction. The operator representing the temperature difference in the y-direction. This represents the mesh size in the x-direction of the discrete element. This represents the mesh size in the y-direction of the discrete element.

[0079] This is the explicit FDTD algorithm formula for heat conduction. It needs to meet the stability condition to operate effectively, namely the time step. satisfy:

[0080] .

[0081] Methods for obtaining thermal displacement distribution include: constructing a steady-state thermal displacement equation using the FDTD method, and solving for the thermal displacement distribution using the steady-state thermal displacement equation.

[0082] Specifically, most materials within a chip cell can be considered rigid bodies. According to the principles of elasticity, the stress within a rigid body must satisfy the following equilibrium differential equation, which, under two-dimensional steady-state conditions, is:

[0083] ,

[0084] in, This represents the normal stress in the x-direction of the mesh. This represents the normal stress in the y-direction of the mesh. This represents the shear stress of the mesh. This represents the external load applied in the x-direction. This represents the external load applied in the y-direction.

[0085] The calculation of strain (deformation that occurs during thermal expansion of the model) satisfies the physical equation (generalized Hooke's law):

[0086] ,

[0087] in, This represents the normal strain in the x-direction on the mesh. The normal strain in the y-direction is represented by μ, Poisson's ratio is represented by α, the coefficient of thermal expansion of the material is represented by E, and the elastic modulus is represented by E. Let G represent the change in temperature, and let G represent the shear modulus, satisfying:

[0088] ,

[0089] Stress is expressed in terms of strain as:

[0090] ,

[0091] in:

[0092] ,

[0093] Where L represents the Lamy constant and B represents the thermodynamic coefficient. The relationship between strain and displacement satisfies the geometric equation:

[0094] .

[0095] Combining the above three equations, we obtain the two-dimensional steady-state thermal displacement equation:

[0096] ,

[0097] ,

[0098] Where u and v represent the displacements in the x and y directions of the grid, respectively.

[0099] S2. Normalize the initial optimization index data and construct a multi-objective optimization objective function.

[0100] In this embodiment, the selected optimization indicators and their calculation methods are shown in Table 1:

[0101] Table 1 .

[0102] The thermal sensing layout optimization problem is a classic multi-objective optimization problem, where each objective aims to minimize the impact of heat. Therefore, the multi-objective optimization objective function is: ,

[0103] in, T(x,y) represents the maximum temperature generated on the substrate, derived from the temperature distribution. Disp(x,y) represents the maximum displacement generated on the substrate, derived from the thermal displacement distribution. This indicates temperature uniformity, where N represents the number of grid cells. This represents the temperature of the i-th grid. The area represents the average temperature of all grids, and the area represents the minimum external area of ​​all chips. This represents the maximum value of the x-axis coordinate of the right boundary of all chips. This represents the minimum x-axis coordinate of the left boundary of all chips. This represents the maximum value of the y-axis coordinate of the upper boundary of all chips. Length represents the minimum y-axis coordinate of the lower boundary of all chips, Length represents the minimum distance between the center points of all chips, and Length(line) represents the shortest line length of the chip layout.

[0104] Furthermore, when constructing the objective functions, the various objectives not only have drastically different physical dimensions, but their fluctuation ranges also vary by orders of magnitude. This heterogeneity in dimensions and scale can severely interfere with the optimization algorithm's judgment of superior and inferior solutions. Therefore, this paper normalizes each objective function using the initial layout as a benchmark to eliminate the interference of physical units, rather than being constrained by the absolute magnitude of specific index values.

[0105] S3. Calculate the dominant contradiction coefficient, generate the initial weights of the indicators, and use the inner PSO algorithm to optimize and reorganize the initial weights of the indicators to obtain the optimized weights.

[0106] Methods for generating initial weights for indicators include:

[0107] In this embodiment, the algorithm first calculates various performance indicators based on the initial layout, including maximum temperature, maximum displacement, interconnect length, layout area, and temperature uniformity. To determine whether "thermal issues" or "layout issues" are more prominent in the initial layout, the program constructs a temperature pressure index and an area pressure index. Each indicator is first compared to a preset expected value:

[0108] ,

[0109] in, Pressure indicating maximum temperature The pressure representing the maximum displacement. Pressure indicating temperature uniformity This indicates the pressure on chip layout area. Pressure indicating the length of the interconnect between chips.

[0110] The preset expected value is set as follows:

[0111] ,

[0112] in, This represents the preset expected value of the maximum temperature. This represents the preset expected value of the maximum displacement. This represents the preset expected value for temperature uniformity. This represents the preset expected value of the minimum circumscribed area. This represents the preset expected value of the minimum distance between the center points of all chips.

[0113] Constructing a pressure index based on the objective function and stress index :

[0114] ,

[0115] ,

[0116] in, This represents the combined pressure in terms of temperature, thermal deformation, and temperature uniformity. This indicates the combined pressure in terms of layout area and line length; if If the initial layout shows a positive correlation, it indicates that the main problem is thermal performance; conversely, if the initial layout shows a negative correlation, it indicates that the main problem is area compression and interconnect length.

[0117] Based on stress index and stress index Calculate the temperature-dominant coefficient, which is used to generate the initial weights:

[0118] ,

[0119] Where η represents the temperature-dominant coefficient.

[0120] The initial reference target weights are obtained based on the temperature-dominant coefficient. Specifically, the temperature-dominant coefficient is preset. and area-dominant preset weight Then, the initial reference target weights are obtained by linear mixing using the dominant coefficient η:

[0121] ,

[0122] in, Indicates the initial weight of the indicator. This indicates that temperature dominates the preset weight. This indicates that area is the dominant preset weight;

[0123] The initial weights of the indicators are obtained based on the initial reference target weights. Specifically, they are related to the reference weights. Perform smooth fusion and normalization so that the sum of all weights is 1, to obtain the smoothed reference target weights:

[0124] ,

[0125] in, Indicates the initial weight of the indicator. Indicates the reference weight.

[0126] S4. Pass the optimized weights to the outer PSO algorithm for layout optimization. When the fitness function remains stagnant, restart the inner PSO algorithm to reorganize the weights.

[0127] The outer PSO (Particle Layout Sort) represents the chip placement scheme; each particle contains the x and y coordinates and rotation states of all chips. In each generation, the placement constraint penalty and normalized performance index for all particles are first calculated, and then the fitness function is calculated based on the current weight w.

[0128] , ,

[0129] in, Represents the overall fitness function. The weight value representing the maximum temperature. The weight value represents the maximum displacement. The weight value represents the minimum distance between the center points of all chips. This represents the weight value of the minimum external area of ​​all chips. Weighting values ​​representing temperature uniformity.

[0130] After identifying the initial primary contradiction and obtaining the initial weights, using fixed weights for optimization can easily lead to the algorithm getting stuck in local optima or biased towards a single objective. When the fitness stagnates or the objective function does not change significantly, a new objective weight reorganization is performed. (Here, θ=0.35 is chosen to reduce the impact of dynamic weights and the occurrence of outliers.) Dynamic weights consist of the mean and variance terms of each normalized indicator. The mathematical basis is that if an indicator is generally too large (indicating that it has not been optimized well), its weight should be increased; if an indicator fluctuates greatly (indicating instability), its weight should also be increased.

[0131] The new weights are not accepted unconditionally. Instead, the following criteria are used to determine whether to update them: if the objective function value of the new weights is significantly better than that of the old weights (an improvement exceeding a threshold), the new weights are accepted; otherwise, the original weights remain unchanged. This avoids "ineffective reorganization" or "oscillating updates." The program also sets a lower limit for the weights. In other words, the weight of each objective cannot be too low. This prevents an objective from being completely ignored during the optimization process.

[0132] In the adaptive weighted optimization path, particle swarm optimization (PSO) is used as the outer layer layout optimizer to search for the optimal spatial arrangement of chips on the substrate. Each particle represents a complete chip layout scheme, and its position vector is composed of the planar coordinates and rotation states of all chips. The optimization objective of the outer layer PSO is to minimize the overall fitness function while satisfying substrate boundary constraints, inter-chip safety spacing constraints, and rotation constraints.

[0133] S5. When the difference between the fitness values ​​of the fitness function before and after two iterations meets a preset condition, it indicates that the fitness function has converged and the optimization is complete.

[0134] In this embodiment, the preset condition is: the difference between the fitness values ​​of the fitness function at two consecutive iterations is less than 10 for 15 consecutive generations. -5 hour.

[0135] Example 2:

[0136] In this embodiment, the effectiveness of the present invention will be evaluated through examples.

[0137] Three chips were placed on a 40mm × 40mm substrate. A safety distance of 5mm was set between the chips and the substrate edge, meaning the chips could be placed within the maximum reasonable area of ​​30mm × 30mm. The safety distance between the chips was 3mm. The material parameters of the chips and substrate are shown in Table 2. The chip size and thermal power density (W / m³) are also listed. 2 As shown in Table 3:

[0138] Table 2 ,

[0139] Table 3 .

[0140] Initial layout as follows Figure 3 As shown. For the adaptive weighted PSO algorithm, the parameters of the outer PSO algorithm are set as follows: population size of 50, maximum number of iterations of 200; maximum inertia weight of 0.90, minimum of 0.40; individual learning factor of 1.80, swarm learning factor of 1.80, and maximum velocity scaling factor of 0.20. These parameters enable PSO to have a larger search step size in the early stage of optimization, which is beneficial to expanding the search range; as the iteration progresses, the inertia weight gradually decreases, the particle search tends to stabilize, thereby improving the local convergence ability of the algorithm in the later stage.

[0141] At the start of the algorithm, various performance metrics are calculated based on the initial layout, and initial weights are constructed for each metric. When the fitness stagnates or the objective function does not change significantly, a new objective weight reorganization is performed. Finally, optimization ends when the fitness remains unchanged for an extended period or the maximum number of optimization rounds is reached.

[0142] The following section compares the optimization performance of the adaptive weighted PSO algorithm for this problem with other algorithms using fixed weights. The weight combination chosen for fixed weight optimization is as follows: To facilitate comparison of the optimization curves, each optimization algorithm was run 200 times, ensuring convergence. The layouts optimized by different methods are shown below. Figure 4 As shown.

[0143] The specific parameter changes before and after optimization are shown in Table 4:

[0144] Table 4 .

[0145] It can be seen that, within the area where the chip can be placed, the adaptive weighted PSO algorithm has the best optimization effect on the highest temperature, temperature uniformity and maximum displacement. This ensures that the chip operates within a better operating temperature range, reduces the risk of device thermal failure and thermal deformation failure, and thus fully utilizes the chip's performance.

[0146] Fitness curves for different optimization methods are as follows: Figure 5 As shown, the Adaptive Weighted PSO algorithm outperforms the frameworks using the other three algorithms in both convergence and optimization. Specifically, the Adaptive Weighted PSO algorithm exhibits excellent convergence, stabilizing after approximately 100 iterations. The GA-based framework shows similar convergence behavior to the Adaptive Weighted PSO algorithm, but its optimization performance is relatively poor. The SA and PSO-based frameworks are weaker in both convergence and optimization results.

[0147] Calculating the thermal displacement and using mechanical parameters as one of the optimization indicators better reflects the thermodynamic parameters of the chip, resulting in a more comprehensive optimization index.

[0148] By using the FDTD algorithm to calculate heat transfer and thermal displacement, the temperature and displacement field distributions before and after optimization can be obtained, such as... Figure 6 , Figure 7 As shown.

[0149] The temperature and displacement fields also show that the optimized layout has a more uniform temperature compared to the initial layout, effectively reducing the maximum temperature of the substrate and chip, and generating a smaller maximum thermal displacement.

[0150] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A multi-physics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing, characterized in that, Includes the following steps: The initial layout modeling and preprocessing of the model to be optimized are performed to obtain initial optimization index data; The initial optimization index data are normalized to construct a multi-objective optimization objective function; Calculate the dominant contradiction coefficient, generate the initial weights of the indicators, and use the inner PSO algorithm to optimize and reorganize the initial weights of the indicators to obtain the optimized weights; The optimized weights are passed to the outer PSO algorithm for layout optimization. When the fitness function remains stagnant, the inner PSO algorithm is restarted for weight reorganization. When the difference between the fitness values ​​of the fitness function before and after two iterations meets a preset condition, it indicates that the fitness function has converged and the optimization is complete.

2. The multiphysics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing according to claim 1, characterized in that, The method for performing the initial layout modeling includes: The geometric modeling of the model to be optimized is divided into model partitions; Perform equivalent parameter processing on the partitioned model; The processed model is meshed to complete the initial layout modeling.

3. The multiphysics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing according to claim 1, characterized in that, The initial optimization metrics include: temperature distribution, thermal displacement distribution, temperature uniformity, minimum external area of ​​the chip, and shortest line length of the chip layout.

4. The multiphysics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing according to claim 3, characterized in that, The method for obtaining the temperature distribution includes: constructing a transient heat conduction equation using the FDTD method, and solving for the temperature distribution using the transient heat conduction equation; The transient heat conduction equation is: ， Where ρ represents the material density, C p Let T represent specific heat, K represent thermal conductivity, and g represent thermal power density. Let t represent the direction vector and t represent time. This represents the gradient operator.

5. The multiphysics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing according to claim 3, characterized in that, The method for obtaining the thermal displacement distribution includes: constructing a steady-state thermal displacement equation using the FDTD method, and solving the thermal displacement distribution using the steady-state thermal displacement equation; The steady-state thermal displacement equation is: ， ， Where u and v represent the displacements in the x and y directions of the grid, respectively, T' represents the temperature change, G represents the shear modulus, L represents the Lamy constant, and B represents the thermodynamic coefficient.

6. The multiphysics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing according to claim 5, characterized in that, The multi-objective optimization objective function is: ， in, T(x,y) represents the maximum temperature generated on the substrate, derived from the temperature distribution. Disp(x,y) represents the maximum displacement generated on the substrate, derived from the thermal displacement distribution. This indicates temperature uniformity, where N represents the number of grid cells. This represents the temperature of the i-th grid. The area represents the average temperature of all grids, and the area represents the minimum external area of ​​all chips. This represents the maximum value of the x-axis coordinate of the right boundary of all chips. This represents the minimum x-axis coordinate of the left boundary of all chips. This represents the maximum value of the y-axis coordinate of the upper boundary of all chips. Length represents the minimum y-axis coordinate of the lower boundary of all chips, Length represents the minimum distance between the center points of all chips, and Length(line) represents the shortest line length of the chip layout.

7. The multiphysics-driven adaptive weighted particle swarm optimization method for MiniLED chip placement and routing according to claim 6, characterized in that, The method for generating the initial weights of the indicator includes: Construct a pressure index based on the objective function. and stress index : ， ， ， in, This represents the preset expected value of the maximum temperature. This represents the preset expected value of the maximum displacement. This represents the preset expected value for temperature uniformity. This represents the preset expected value of the minimum circumscribed area. This represents the preset expected value of the minimum distance between the center points of all chips. This represents the combined pressure in terms of temperature, thermal deformation, and temperature uniformity. This indicates the combined pressure in terms of layout area and line length. Pressure indicating maximum temperature The pressure representing the maximum displacement. Pressure indicating temperature uniformity This indicates the pressure on chip layout area. Pressure indicating the length of the interconnect between chips; Based on the pressure index and the pressure index Calculate the temperature-dominant coefficient: ， Where η represents the temperature-dominant coefficient; The initial reference target weights are obtained based on the temperature-dominant coefficient: ， in, Indicates the initial weight of the indicator. This indicates that temperature dominates the preset weight. This indicates that the area is the primary factor in the preset weighting; The initial weights of the indicator are obtained based on the initial reference target weights: ， in, Indicates the initial weight of the indicator. Indicates the reference weight.

8. The multiphysics-driven MiniLED chip placement and routing adaptive weighted particle swarm optimization method according to claim 6, characterized in that, The fitness function is: ， ， in, Represents the overall fitness function. The weight value representing the maximum temperature. The weight value represents the maximum displacement. The weight value represents the minimum distance between the center points of all chips. This represents the weight value of the minimum external area of ​​all chips. Weighting values ​​representing temperature uniformity.

Images