Thickness matching optimization method and system for batch assembly of semiconductor cold plates

Abstract

The invention discloses a thickness matching optimization method and system for batch assembly of semiconductor cold plates, and relates to the technical field of batch assembly of semiconductor cold plates. S2, acquiring thickness data; s3, constructing a thickness data array; s4, performing full combination traversal and objective function calculation; and S5, optimal assembly combination screening. According to the method, based on the cold plate assembly matching thought of thickness data sorting combination optimization, the thickness deviation superposition effect of multiple layers of thin plates is counteracted through ascending / descending sort combination of the thicknesses of the thin plates of different specifications, the fluctuation value of the total thickness of the cold plates of the same batch before welding is greatly reduced, the optimal assembly combination can be rapidly screened out, and the welding quality is improved. According to the method, the high-efficiency process of the chemical etching runner and the consistency of the batch performance of the cold plate are both considered, the machining process for normalizing the thickness of the thin plate does not need to be additionally added, the production efficiency of the etching process is kept, and meanwhile the problem of thickness consistency degradation caused by the etching process is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor cold plate batch assembly, in particular to a thickness matching optimization method and system for semiconductor cold plate batch assembly. BACKGROUND

[0002] With the rapid development of the semiconductor industry, the heat dissipation demand of semiconductor devices is increasingly stringent, and the cold plate as the core thermal management component has entered the batch production stage, which puts forward very high requirements for its performance consistency. Traditional cold plate processing adopts machining method to process the thickness of multi-layer thin plates, and then forms through diffusion welding. Although it can guarantee the performance stability of single cold plate, the machining process is complicated and inefficient, and it cannot adapt to the rhythm of large-scale production. In order to improve production efficiency, the chemical etching process is gradually popularized in the industry to prepare cold plate runner. This process can realize the rapid and batch forming of the runner, but due to the limitation of etching process characteristics, the thickness consistency of the metal thin plate where the runner is located is poor, and local thickness deviation is easy to occur. During batch welding assembly, the thickness deviation of multi-layer thin plates will have a cumulative effect, resulting in significant differences in the total thickness of the cold plates before welding in the same batch, and then causing significant differences in the compression amount during the welding process, ultimately causing batch fluctuations in the core performance of the cold plate, such as heat dissipation efficiency and pressure resistance, which seriously affects the overall operation stability of the semiconductor device, and becomes a key technical bottleneck restricting the batch production of chemical etching cold plate. SUMMARY

[0003] In view of the deficiencies of the prior art, the present application provides a thickness matching optimization method and system for semiconductor cold plate batch assembly, which solves the problems raised in the background art.

[0004] To achieve the above purpose, the present application realizes the following technical scheme: a thickness matching optimization method for semiconductor cold plate batch assembly, comprising the following steps: S1, preliminary preparation: Determine the number of thin plate layers of the cold plate to be produced, the specifications and models of each layer of thin plate, and the number of products in this batch, and calculate the total number of each specification thin plate required; prepare thickness detection equipment and data processing tools that meet the assembly tolerance requirements; S2, thickness data acquisition: Measure the thickness of all the thin plates to be assembled in this batch one by one, record the original thickness data according to the thin plate specifications, and ensure that the data is accurate and traceable; S3, constructing a thickness data array: Establish a two-dimensional data array in the data processing tool, with the number of thin plate layers as the row and the number of batch products as the column, and fill the thickness data of the same specification thin plate into the corresponding column cells in sequence; for each column of thickness data in the array, two arrangement methods of ascending order and descending order are preset; S4, full combination traversal and target function calculation: If the cold plate has layers of thin plates, the total number of combinations is determined as kinds; for each permutation and combination scheme, the multi-layer thin plate thickness data of each row in the transverse direction is accumulated to obtain the total pre-weld thickness of each product; the maximum value and the minimum value of the total pre-weld thickness of all products under the scheme are counted, and the difference between the two is calculated as the first target function; S5, optimal assembly combination screening: Compare the first target function values of all permutation and combination schemes, and select the combination scheme with the smallest target function value as the optimal assembly relationship; according to the thickness sorting of each specification thin plate in the scheme, the thin plates are assembled layer by layer to complete the pre-welding superposition.

[0005] Further, the data processing tool is Excel software, and the two-dimensional data array is a matrix structure of "number of thin plate layers x number of batch products".

[0006] Further, the thickness detection device is a high-precision thickness gauge, and the detection accuracy meets the cold plate assembly tolerance requirements.

[0007] Further, the cold plate is prepared by using a chemical etching flow channel process, and the number of thin plate layers is a positive integer ≥ 2.

[0008] Further, the calculation formula of the first target function is: wherein, is the total pre-weld thickness of the i-th product, is the total pre-weld thickness of the i-th product, , is the number of batch products, is the maximum value of the total thickness of all products, is the minimum value of the total thickness of all products.

[0009] A thickness matching optimization method for batch assembly of semiconductor cold plates further includes the following steps: S6, first target function weight and second target function weight setting: Set the cold plate pre-welding target total thickness , determine the first target function weight and the second target function weight based on the cold plate design parameters and material characteristics, and satisfy ( , ); S7, construction of double-target weighted optimization function: Construct a double-target weighted optimization function , wherein Let the first objective function be... This is a function representing the average deviation between the total thickness of a single product and the target total thickness. S8. Genetic Algorithm for Fast Finding of Minimum Combination Solution: The arrangement of the plate thicknesses in each column is binary-encoded, either in ascending or descending order, to construct an initial population; a bi-objective weighted optimization function is then used. The fitness function is used to iteratively update the population through selection, crossover, and mutation operations until the iteration termination condition is met. S9. The combination scheme that minimizes the output fitness function value: The combination scheme with the minimum fitness function value at the end of the output iteration is taken as the optimal assembly relationship, and the thin plate is assembled layer by layer and pre-welding stacked according to this scheme.

[0010] Furthermore, the second objective function The calculation formula is: in, For the first Total thickness of each product before welding This refers to the quantity of products in a batch. The target total thickness.

[0011] Furthermore, the weights The determination method is as follows: When prioritizing overall consistency of total thickness, take , ; When prioritizing the accuracy of thickness for individual products, take , .

[0012] Furthermore, the iteration termination condition of the genetic algorithm is: The fitness function value of the population fluctuates within ≤0.001mm for 10 consecutive generations, or the number of iterations reaches the preset maximum value, which is 50-100 generations.

[0013] A thickness matching optimization system for batch assembly of semiconductor cold plates, which applies the aforementioned thickness matching optimization method for batch assembly of semiconductor cold plates, the system comprising: The thickness detection module is used to measure the thickness of each sheet of thin plate to be assembled in this batch, and output accurate and traceable raw thickness data. The data processing module receives raw data from the thickness detection module, constructs a two-dimensional data array with the vertical axis representing the number of thin plate layers and the horizontal axis representing the batch product quantity, and presets ascending and descending sorting attributes for each column of data; The combinatorial optimization module is used to determine... An arrangement combination scheme, For the number of sheet layers, traverse each scheme and accumulate the thickness data of each row to obtain the total thickness of the product before welding, and calculate the first target function , filter out The smallest optimal assembly combination; The target function optimization module is used to set the target total thickness , determine the weight , build a double-target weighted optimization function ; The intelligent optimization module is used to realize the rapid optimization of the double-target function through the genetic algorithm, and output the optimal assembly combination with the minimum fitness function value; The assembly execution module is used to complete the layer-by-layer assembly and pre-welding superposition according to the optimal assembly relationship output by the combination optimization module or the intelligent optimization module.

[0014] The present application provides a kind of thickness matching optimization method and system for batch assembly of semiconductor cold plate, with the following beneficial effects: 1.The thickness matching optimization method and system for batch assembly of semiconductor cold plate based on the cold plate assembly matching idea of thickness data sorting combination optimization, through the ascending / descending sequence arrangement combination of different specifications sheet thickness, offsets the thickness deviation superposition effect of multi-layer sheet, greatly reduces the fluctuation value of total thickness of cold plate before welding in the same batch, and converts the thickness data of multi-layer sheet of cold plate into two-dimensional array, with total thickness extreme difference as target function, quickly filters out the optimal assembly combination, balances the operation convenience and optimization accuracy, realizes the balance of chemical etching flow channel efficient process and the consistency of cold plate batch performance, without additional machining process of sheet thickness normalization, while retaining the production efficiency of etching process, solves the problem of thickness consistency degradation.

[0015] 2.The thickness matching optimization method and system for batch assembly of semiconductor cold plate also breaks through the limitation of traditional single-target optimization through double-target weighted optimization function, not only pays attention to the overall consistency of total thickness of cold plate in the same batch, but also considers the accurate matching of single product thickness and design target value, realizes double performance guarantee, makes the assembly effect more in line with the multi-element demand of precision in actual production, and the introduction of genetic algorithm effectively solves the problem of low efficiency of full combination traversal in multi-layer sheet scene, greatly improves the optimization speed, and the flexible adjustment characteristics of weight makes the technical scheme adapt to different production priority demand, further enhances the practicality and adaptability of its industrial application. BRIEF DESCRIPTION OF DRAWINGS

[0016] Fig. 1 It is the flowchart of steps S1-S5 of the thickness matching optimization method of the semiconductor cold plate batch assembly of the present application; Fig. 2A flowchart of steps S6-S9 of the thickness matching optimization method for batch assembly of a semiconductor cold plate of the present application. DETAILED DESCRIPTION

[0017] The embodiments of the present application will be further described below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application, but cannot be used to limit the scope of the present application.

[0018] As shown in Figs. 1-2 The present application provides a technical solution: a thickness matching optimization method for batch assembly of a semiconductor cold plate, comprising the following steps: S1, preliminary preparation: Determine the number of thin plates, the specifications and models of each layer of thin plates, and the number of products in this batch, and calculate the total number of each specification of thin plates required (for example, five layers of thin plate structure, eight batches of products require five specifications of a total of 40 thin plates); prepare thickness detection equipment and data processing tools that meet the assembly tolerance requirements, the thickness detection equipment is a high-precision thickness gauge, and the detection accuracy meets the cold plate assembly tolerance requirements; S2, thickness data acquisition: Measure the thickness of each thin plate to be assembled in this batch, record the original thickness data according to the thin plate specifications, and ensure that the data is accurate and traceable; S3, constructing a thickness data array: Establish a two-dimensional data array in the data processing tool, the vertical direction is represented by the number of thin plate layers (different thin plate categories), and the horizontal direction is represented by the number of batch products. The thickness data of thin plates of the same specification are sequentially filled into the corresponding column cells. For each column of thickness data in the array, two arrangement methods of ascending order and descending order are preset. The data processing tool is Excel software, and the two-dimensional data array is a matrix structure of "number of thin plate layers x number of batch products"; S4, full combination traversal and target function calculation: If the cold plate has layers of thin plates, the total number of combinations is ; For each arrangement combination scheme, the thickness data of the multiple layers of thin plates in each row is accumulated to obtain the total thickness before welding of each product. The maximum value and the minimum value of the total thickness before welding of all products under this scheme are calculated, and the difference between the two is taken as the first target function. The cold plate is prepared by chemical etching channel process, and the number of thin plate layers is a positive integer ≥2; S5, optimal assembly combination screening: Compare the first target function values of all arrangement combination schemes, and select the combination scheme with the smallest target function value as the optimal assembly relationship. According to the thickness sorting of each specification of thin plate in the scheme, the thin plates are assembled layer by layer to complete the pre-welding superposition; The calculation formula of the first target function is: in, For the first Total thickness of each product before welding , This refers to the quantity of products in a batch. The maximum total thickness of all products. This is the minimum total thickness of all products; Example 1: A chemically etched flow channel cold plate with a five-layer thin plate structure and eight batches of products is used as an example; Preliminary preparations: It was determined that the cold-rolled steel plate is a five-layer thin plate structure. This batch will produce 8 products, and it is calculated that 40 thin plates of 5 different specifications are needed; a high-precision thickness gauge (accuracy 0.001mm) and Excel software are prepared. Thickness data acquisition: The thickness of 40 thin plates was measured one by one using a high-precision thickness gauge. Data was recorded for 5 specifications. The measured thickness data of the thin plate of specification 1 (top cover plate) were 1.002mm, 1.005mm, and 0.998mm; the thickness of the thin plate of specification 2 (flow channel layer 1) was 0.803mm, 0.801mm, and 0.799mm, etc. Each specification corresponds to a different tolerance range (for example, the tolerance of specification 1 is ±0.005mm, and the tolerance of specification 2 is ±0.003mm). Constructing the data array: Create a 5×8 two-dimensional array in Excel. Column A corresponds to the top cover plate (specification 1), and columns B and E correspond to the flow channel layer 1 (specification 2), isolation layer (specification 3), flow channel layer 2 (specification 4), and bottom substrate (specification 5), respectively. Rows 1-8 correspond to 8 product numbers. The data range in column A is 0.995-1.005mm, and in column B it is 0.797-0.803mm, etc. Use the custom sorting function to set column A to ascending order, column B to descending order, and the remaining columns to be sorted according to the production process requirements. Full combinatorial traversal and Calculate: The total number of combinations is This method uses Excel VBA macro programming to automatically iterate through rows, calculating the thickness of each row using the SUM function after each combination. ,according to Calculate the combination of Values, for example: A combination of 8 products The thicknesses are 4.998mm, 5.003mm, 5.001mm, 4.999mm, 5.002mm, 5.000mm, 4.997mm, and 5.004mm respectively. The calculation results will be automatically filled into the specified cells; Optimal combination screening and assembly: selection The smallest (0.004mm) combination scheme is used to assemble the thin plates according to the arrangement order of the scheme, and the pre-welding stacking is completed; Actual welding verification showed that the welding compression deviation of this batch of cold plates was controlled within 0.02mm, and the batch fluctuation of heat dissipation performance was significantly reduced compared with the traditional assembly method, achieving a significant improvement in performance consistency. Based on the above description, the thickness matching optimization method and system for batch assembly of semiconductor cold plates is based on the cold plate assembly matching idea of ​​thickness data sorting and combination optimization. By arranging and combining the thicknesses of thin plates of different specifications in ascending / descending order, the cumulative effect of thickness deviation of multi-layer thin plates is offset, which greatly reduces the fluctuation value of the total thickness of the cold plates in the same batch before welding. Furthermore, the thickness data of the multi-layer thin plates of the cold plate is transformed into a two-dimensional array. With the extreme difference of the total thickness as the objective function, the optimal assembly combination is quickly screened out. It takes into account both the ease of operation and the optimization accuracy, and achieves the balance between the high efficiency of chemical etching flow channel process and the consistency of cold plate batch performance. There is no need to add an additional machining process for thin plate thickness normalization. While retaining the production efficiency of the etching process, it solves the problem of thickness consistency degradation caused by it. Considering the deviation between the total thickness of a single product and the design target thickness, a situation may arise where "the overall fluctuation is small, but the deviation of a single product from the target value is large," affecting welding accuracy; at the same time, when the number of layers of cold-rolled thin plates is large (e.g., n≥8), the number of combinations traversed in the full combination process ( The number of types (of semiconductor cold plates) is growing exponentially, leading to a significant increase in optimization time and insufficient adaptability. To address this problem, the present invention provides a thickness matching optimization method for batch assembly of semiconductor cold plates, which further includes the following steps: S6. Weight settings for the first objective function and the second objective function: Set the target total thickness of the cold plate before welding. The weights of the first objective function are determined based on the design parameters and material properties of the cold plate. Weights of the second objective function ,satisfy ( , ); Weight The determination method is as follows: When prioritizing overall consistency of total thickness, take , ; When prioritizing the accuracy of thickness for individual products, take , ; S7. Construction of the bi-objective weighted optimization function: Construct a dual-objective weighted optimization function ,in Let the first objective function be... This is a function representing the average deviation between the total thickness of a single product and the target total thickness. Second objective function The calculation formula is: Wherein, is the total thickness of the first product before welding, is the number of products in the batch, is the target total thickness; S8, genetic algorithm fast optimization minimum combination scheme: The arrangement of each column of sheet thickness is binary coded, and the arrangement is ascending order or descending order, 0 represents ascending order and 1 represents descending order, and an initial population is constructed; a double objective weighted optimization function is used as the fitness function, and the population is iteratively updated through selection, crossover and mutation operations until the iteration termination condition is met; The iteration termination condition of the genetic algorithm is: The fluctuation range of the fitness function value of the population of 10 generations in succession is less than or equal to 0.001 mm, or the iteration number reaches a preset maximum value, and the preset maximum value is 50-100 generations, and the specific process of the genetic algorithm fast optimization is as follows: The arrangement of the layer sheet is binary coded, wherein "0" represents ascending order and "1" represents descending order, and finally a chromosome with a length of is formed, for example, the coding "01011001" indicates that the 1st, 3rd, 6th and 7th layers are arranged in ascending order, and the 2nd, 4th, 5th and 8th layers are arranged in descending order; By randomly generating 50 chromosomes that meet the above coding rules, an initial population is constructed, providing basic data for subsequent iterative optimization; A double objective function is used as the fitness function, and in the evaluation, the smaller the value, the higher the fitness of the sheet arrangement scheme represented by the corresponding chromosome, and the closer to the optimization target; The roulette selection method is used to select the initial population or the population in the iteration process based on the fitness of the chromosome, and the excellent chromosomes with high fitness are selected into the next generation population, the selected chromosomes are crossed with a crossover probability of 0.8, new chromosome combinations are generated by exchanging part of the gene fragments of the chromosomes, the diversity of the population is increased, the genes of the chromosomes are randomly mutated according to a mutation probability of 0.05, i.e. changing the gene coding (0 to 1 or 1 to 0), further exploring the solution space, and avoiding the algorithm from falling into local optimum; The above selection, crossover and mutation operations are repeated to generate a new population, and when the double objective function When the value fluctuation range is less than or equal to 0.001 mm, or the number of iterations reaches 80, the iteration process is stopped, and the thin plate arrangement scheme corresponding to the optimal chromosome in the obtained population at this time is the result after the genetic algorithm optimization; S9, output the combination scheme with the minimum fitness function value: The combination scheme with the minimum fitness function value at the iteration termination is output as the optimal assembly relationship, and the thin plate is assembled layer by layer and overlapped before welding according to the scheme; A thickness matching optimization system for batch assembly of semiconductor cold plates applies the above-mentioned thickness matching optimization method for batch assembly of semiconductor cold plates, and the system comprises: A thickness detection module is configured to measure the thickness of each thin plate in the batch to be assembled and output accurate and traceable original thickness data; A data processing module is configured to receive the original data from the thickness detection module, construct a two-dimensional data array with the number of thin plate layers in the vertical direction and the number of batch products in the horizontal direction, and preset two arrangement properties of ascending order and descending order for each column of data; A combination optimization module is configured to determine a plurality of arrangement combination schemes, for the number of thin plate layers, traverse each scheme and accumulate the product pre-welding total thickness of each row of thickness data, and calculate the first objective function , and screen out the optimal assembly combination with the minimum value; An objective function optimization module is configured to set the target total thickness , determine the weight , and construct a double-objective weighted optimization function ; An intelligent optimization module is configured to realize fast optimization of the double-objective function through a genetic algorithm, and output the optimal assembly combination with the minimum fitness function value; An assembly execution module is configured to complete the layer-by-layer assembly and pre-welding overlap according to the optimal assembly relationship output by the combination optimization module or the intelligent optimization module based on the thickness sorting of each specification thin plate.

[0019] In Example 2, eight-layer thin plate structures and 10 batch products of chemical etching flow channel cold plates are taken as examples. Preparation and data collection: determine that the cold plate is an eight-layer thin plate structure, produce 10 products, and prepare 80 thin plates of 8 specifications; after thickness measurement, an 8x10 Excel data array is constructed according to the basic method; Parameter setting: set the target total thickness , preferentially guarantee the accuracy of a single product, and take , ; calculate according to the formula : Build The dual objective function; Genetic algorithm optimization: Generate binary chromosomes of length 8 (e.g., "00110101"), initial population size 50; crossover probability 0.8, mutation probability 0.05, iterate to the 35th generation, for 10 consecutive generations... If the value fluctuation is ≤0.001mm, the iteration is terminated; Final selection The combination scheme with the smallest value (0.003mm) has... , ; The thin plates were assembled and welded according to the optimal solution. The test results showed that the average thickness deviation of a single product was ≤0.01mm, the welding compression deviation was ≤0.015mm, and the optimization time was only 2 minutes. Based on the above description, the thickness matching optimization method and system for batch assembly of semiconductor cold plates also breaks through the limitations of traditional single-objective optimization by using a dual-objective weighted optimization function. It focuses on both the overall consistency of the total thickness of the cold plates in the same batch and the precise matching of the thickness of individual products with the design target value, achieving dual performance assurance. This makes the assembly effect more in line with the diverse precision requirements of actual production. Furthermore, the introduction of the genetic algorithm effectively solves the problem of low efficiency in full combination traversal in multi-layer thin plate scenarios, greatly improving the optimization speed. At the same time, the flexible adjustment of weights allows the technical solution to adapt to different production priority requirements, further enhancing its practicality and adaptability for industrial applications.

[0020] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A thickness matching optimization method for batch assembly of semiconductor cold plates, characterized in that: It comprises the following steps: S1, preliminary preparation: Determine the number of thin plate layers of the cold plate to be produced, the specifications and models of each layer of thin plate, and the quantity of the batch product, and calculate the total quantity of each specification of thin plate required; prepare thickness detection equipment and data processing tools that meet the assembly tolerance requirements; S2, thickness data acquisition: Measure the thickness of each thin plate in the batch, record the original thickness data according to the thin plate specifications, and ensure that the data is accurate and traceable; S3, constructing a thickness data array: In the data processing tool, a two-dimensional data array is established, with the vertical direction representing the number of thin plate layers and the horizontal direction representing the number of batch products. The thickness data of thin plates of the same specification are sequentially filled into the corresponding column cells. For each column of thickness data in the array, two arrangement methods, ascending and descending, are preset; S4, full combination traversal and target function calculation: If the cold plate is shared The total number of combinations is determined For each arrangement combination scheme, the multi-layer sheet thickness data of each row in the transverse direction is accumulated to obtain the total thickness before welding of each product; the maximum value and the minimum value of the total thickness before welding of all products under the scheme are counted, and the difference between the two is calculated as a first objective function; S5, optimal assembly combination screening: Compare the first target function values of all permutation and combination schemes, and select the combination scheme with the smallest target function value as the optimal assembly relationship. According to the thickness sorting of each specification of thin plate in the scheme, the thin plates are assembled layer by layer to complete the pre-welding lamination.

2. The thickness matching optimization method for batch assembly of a semiconductor cold plate according to claim 1, wherein: The data processing tool is Excel software, and the two-dimensional data array is a matrix structure of "number of thin plate layers x number of batch products".

3. The thickness matching optimization method for batch assembly of a semiconductor cold plate according to claim 1, wherein: The thickness detection equipment is a high-precision thickness gauge with detection accuracy meeting the cold plate assembly tolerance requirements.

4. The thickness matching optimization method for batch assembly of a semiconductor cold plate according to claim 1, wherein: The cold plate is prepared by chemical etching process, and the number of thin plate layers is ≥2.

5. The thickness matching optimization method for batch assembly of a semiconductor cold plate according to claim 1, wherein: The calculation formula of the first target function is: wherein, is the total thickness of the product before welding, is the total thickness of the product before welding, , is the number of products in the batch, is the maximum value of the total thickness of all products, is the minimum value of the total thickness of all products.

6. The thickness matching optimization method for batch assembly of a semiconductor cold plate of claim 1, wherein: It also comprises the following steps: S6, setting the first target function weight and the second target function weight: Setting a target total thickness of the cold plate before welding Determining a first objective function weight based on the cold plate design parameters and material properties and a second objective function weight Satisfying ( , ) S7, constructing a double-target weighted optimization function: Constructing a dual-objective weighted optimization function wherein is a first objective function, is a function of average deviation of the total thickness of the single product from a target total thickness; S8, genetic algorithm for rapid optimization of the minimum combination scheme: The arrangement of the thickness of each column of thin plates is binary coded, the arrangement is ascending order or descending order, and an initial population is constructed; a double-target weighted optimization function The population is iteratively updated by selection, crossover and mutation operations until the iteration termination condition is met. S9, output the combination scheme with the smallest fitness function value: Output the combination scheme with the smallest fitness function value at the iteration termination as the optimal assembly relationship, and complete the thin plate assembly layer by layer and pre-welding lamination according to the scheme.

7. The thickness matching optimization method for batch assembly of a semiconductor cold plate according to claim 6, wherein: The second objective function The calculation formula is: wherein, is the pre-weld total thickness of the product, is the number of products in the batch, is the target total thickness.

8. The thickness matching optimization method for batch assembly of a semiconductor cold plate according to claim 6, wherein: The weights are determined as follows: When the overall consistency of the total thickness is preferentially guaranteed, take , ; When the single product thickness precision is preferentially guaranteed, take , . 9.The thickness matching optimization method for batch assembly of a semiconductor cold plate according to claim 6, wherein: The iteration termination condition of the genetic algorithm is: The fluctuation amplitude of the fitness function value of the population is ≤0.001mm for 10 consecutive generations, or the iteration number reaches the preset maximum value, which is 50-100 generations.

10. A thickness matching optimization system for semiconductor cold plate mass assembly, applying the thickness matching optimization method for semiconductor cold plate mass assembly of any one of claims 1-9, characterized in that: The system comprises: A thickness detection module for measuring the thickness of each thin plate in the batch, outputting accurate and traceable original thickness data; A data processing module for receiving the original data from the thickness detection module, constructing a two-dimensional data array with the vertical direction representing the number of thin plate layers and the horizontal direction representing the number of batch products, and presetting ascending and descending arrangement attributes for each column of data; a combination optimization module for determining a permutation and combination scheme, for the number of thin plate layers, traverse each scheme and accumulate each row thickness data to obtain the total thickness of the product before welding, calculate the first objective function , screen out the optimal assembly combination with the smallest an objective function optimization module for setting a target total thickness , determining weights , constructing a dual objective weighted optimization function ; An intelligent optimization module for realizing rapid optimization of double-target functions through genetic algorithm and outputting the optimal assembly combination with the smallest fitness function value; An assembly execution module for completing the thin plate assembly layer by layer and pre-welding lamination according to the thickness sorting of each specification of thin plate in the optimal assembly relationship output by the combination optimization module or the intelligent optimization module.

Images