Silicon wafer automatic cargo allocation optimization method and device, electronic equipment and medium
By obtaining and processing multi-dimensional parameters of silicon wafers in real time, and using the solution optimization model to generate the best distribution solution, it solves the problems of incomplete parameter coverage, single optimization goals and insufficient real-time performance in traditional silicon wafer distribution, and achieves efficient and accurate silicon wafer distribution management.
Patent Information
- Application Number
- CN202510700661.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-28
- Publication Date
- 2025-08-29
AI Technical Summary
Traditional silicon wafer distribution methods have problems such as incomplete parameter coverage, single optimization goals and insufficient real-time performance, resulting in improper yield loss and inventory management.
By obtaining multi-dimensional parameters of silicon wafers in real time, using pre-processing methods to generate parameter feature vectors, match order requirements, input them into the pre-trained solution optimization model, filter out the best distribution solution, and realize the optimization of multi-dimensional parameters.
The distribution efficiency has been improved, the comprehensive matching degree of parameters has reached 99.5%, the yield loss has been reduced to below 0.3%, the inventory turnover rate has been increased by 40%, the cost has been reduced by 15%-25%, and the order change response time has been shortened to within 30 seconds.
Smart Images
Figure CN120563004A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of semiconductor and photovoltaic production technology, and more specifically, to a method, device, electronic equipment, and medium for optimizing automatic silicon wafer distribution. Background Art
[0002] In the semiconductor and photovoltaic industries, silicon wafer delivery must meet multi-dimensional parameter requirements, such as dimensional tolerance ≤±0.01mm, resistivity fluctuation ≤±5%, minority carrier lifetime ≥10μs, carbon content ≤10ppma, etc. Traditional delivery methods rely on manual screening, which has the following drawbacks:
[0003] Incomplete parameter coverage: Focusing only on a single or a few parameters (such as size), ignoring the comprehensive matching of electrical properties (resistivity, minority carrier lifetime) and impurity content (carbon, oxygen), resulting in yield loss (such as resistivity mismatch causing chip short circuit);
[0004] Single optimization objective: Failure to balance multiple objectives such as parameter fit, inventory turnover, and cost, which can easily lead to quality risks such as high-cost inventory backlogs or preferential consumption of low-priced inventory.
[0005] Lack of real-time performance: Inventory data is not updated synchronously with order demand, manual picking takes up to several hours, and it is difficult to cope with high-frequency order changes (such as emergency orders). Summary of the Invention
[0006] In view of this, the purpose of this application is to provide a method, device, electronic device, and medium for optimizing automatic silicon wafer allocation. This method, device, electronic device, and medium effectively solve the problems of incomplete parameter coverage, single optimization target, and insufficient real-time performance in manual allocation methods.
[0007] In a first aspect, an embodiment of the present application provides a method for optimizing automatic silicon wafer distribution, the method comprising:
[0008] Acquire parameters and corresponding parameter values of multiple dimensions of silicon wafers in stock in real time, and process the parameter values based on preprocessing methods corresponding to the multiple dimensions to obtain parameter feature vectors; different dimensions correspond to different preprocessing methods;
[0009] Extracting target parameter requirements from the order requirements, matching the parameter feature vector with the target parameter requirements of the order requirements to obtain a matching feature set; the matching feature set includes multiple matching feature combinations;
[0010] Inputting the matching feature set into a pre-trained solution optimization model to process the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimized solution set;
[0011] An optimal solution set is screened out from the optimization solution set, and an optimal distribution solution for the target quantity is extracted based on the optimal solution set to complete the optimization of the distribution solution.
[0012] In combination with the first aspect, an embodiment of the present application provides a first possible implementation of the first aspect, wherein the processing of the matching feature combinations in the matching feature set based on the solution optimization model to obtain the optimization solution set includes:
[0013] Based on the relationship between the weight of the parameter and the parameter, determining whether the parameters corresponding to the matching feature combination all meet the preset fast processing conditions;
[0014] If so, the matching feature combination is processed based on a fast processing method corresponding to the fast processing condition to obtain a corresponding optimization solution set.
[0015] In combination with the first aspect, an embodiment of the present application provides a second possible implementation of the first aspect, wherein the fast processing method corresponding to the fast processing condition is used to process the matching feature combination to obtain a corresponding optimization solution set, including:
[0016] Calculating multiple attribute values of the parameters corresponding to the matching feature combination and parameter values of the parameters to obtain parameter calculation results;
[0017] The parameter calculation results corresponding to the various parameters are aggregated to obtain a comprehensive distribution score, so as to obtain the optimization solution set based on the comprehensive distribution score.
[0018] In combination with the first aspect, the embodiment of the present application provides a third possible implementation of the first aspect, wherein, after determining whether all parameters corresponding to the matching feature combination meet preset fast processing conditions based on the relationship between the parameter weight and the parameter, the method further includes:
[0019] If not, then select a partial matching feature combination from the matching feature combination and call different types of iteration methods;
[0020] The different types of iterative methods are executed to iterate the partial matching feature combination to obtain an optimization solution set.
[0021] In combination with the first aspect, the embodiment of the present application provides a fourth possible implementation of the first aspect, wherein the step of selecting the best solution set from the set of optimization solutions includes:
[0022] Presetting a screening rule for the set of optimization solutions, and calling the screening rule based on the set of optimization solutions;
[0023] The optimization solution set is screened based on the screening rules to obtain the best solution set.
[0024] In combination with the first aspect, an embodiment of the present application provides a fifth possible implementation of the first aspect, wherein the matching the parameter feature vector with the target parameter requirement of the order requirement to obtain a matching feature set includes:
[0025] Convert the target parameter requirements of the order requirements based on a preset feature conversion method to obtain an order feature vector;
[0026] The multi-dimensional similarity between the order feature vector and the parameter feature vector is calculated, and a matching feature combination is obtained based on the multi-dimensional similarity.
[0027] In combination with the first aspect, an embodiment of the present application provides a sixth possible implementation of the first aspect, wherein the preprocessing method corresponding to multiple dimensions is used to process the parameter values to obtain the parameter feature vector, including:
[0028] Based on the preprocessing methods corresponding to the multiple dimensions, the parameter values are processed to obtain an intermediate interval matrix;
[0029] The intermediate interval matrices corresponding to the multiple dimensions are fused to obtain a parameter feature vector in matrix form corresponding to the parameter value.
[0030] In a second aspect, an embodiment of the present application provides a device for optimizing automatic silicon wafer distribution, the device comprising:
[0031] An acquisition module is used to obtain parameters and corresponding parameter values of multiple dimensions of the silicon wafers in stock in real time, and process the parameter values based on the preprocessing methods corresponding to the multiple dimensions to obtain parameter feature vectors; different dimensions correspond to different preprocessing methods;
[0032] An extraction module is used to extract target parameter requirements from the order requirements, and match the parameter feature vector with the target parameter requirements of the order requirements to obtain a matching feature set; the matching feature set includes multiple matching feature combinations;
[0033] An input module, configured to input the matching feature set into a pre-trained solution optimization model, so as to process the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimized solution set;
[0034] The optimization module is used to screen out the best solution set from the optimization solution set, and extract the best distribution solution of the target quantity based on the best solution set to complete the optimization of the distribution solution.
[0035] In a third aspect, an embodiment of the present application provides an electronic device comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of any one of the methods for optimizing automatic silicon wafer distribution are performed.
[0036] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer program executes any one of the steps of the method for optimizing automatic silicon wafer distribution.
[0037] An embodiment of the present application provides a method for optimizing automatic silicon wafer allocation. The method first obtains parameters and corresponding parameter values of multiple dimensions of silicon wafers in stock in real time, and processes the parameter values based on preprocessing methods corresponding to the multiple dimensions to obtain parameter feature vectors; different dimensions correspond to different preprocessing methods; secondly, extracts target parameter requirements in order requirements, matches the parameter feature vectors with the target parameter requirements of the order requirements to obtain a matching feature set; the matching feature set includes multiple matching feature combinations; then, inputs the matching feature set into a pre-trained solution optimization model, processes the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimized solution set; finally, selects the best solution set from the optimized solution set, generates a corresponding optimal allocation solution based on the best optimized combination, and completes allocation optimization based on the best allocation solution, thereby optimizing allocation based on the parameters of multiple dimensions of silicon wafers. Furthermore, after obtaining order requirements, an optimized allocation solution is generated in a timely manner based on the silicon wafers in stock, effectively solving the problems of incomplete parameter coverage, single optimization target, and insufficient real-time performance of manual allocation methods, and ensuring the timeliness and effectiveness of silicon wafer allocation in the semiconductor and photovoltaic industries. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.
[0039] Figure 1 A schematic diagram of the process of the first silicon wafer automatic distribution optimization method provided in an embodiment of the present application is shown;
[0040] Figure 2A schematic diagram of a process for optimizing automatic silicon wafer distribution according to an embodiment of the present application is shown;
[0041] Figure 3 A schematic diagram of a process for obtaining a parameter feature vector according to an embodiment of the present application is shown;
[0042] Figure 4 The following is a structural block diagram of a first silicon wafer automatic distribution optimization device provided in an embodiment of the present application;
[0043] Figure 5 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION
[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of illustration and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts can be implemented out of sequence, and steps without logical context can be reversed or implemented simultaneously. In addition, those skilled in the art, under the guidance of the contents of this application, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.
[0045] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.
[0046] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.
[0047] Currently, in the semiconductor and photovoltaic industries, silicon wafer selection must meet multi-dimensional parameter requirements, such as dimensional tolerance ≤±0.01mm, resistivity fluctuation ≤±5%, minority carrier lifetime ≥10μs, and carbon content ≤10ppma. Traditional selection relies on manual screening, which has the following drawbacks: incomplete parameter coverage, a single optimization target, and insufficient real-time performance.
[0048] Based on this, the embodiments of the present application provide a method, device, electronic device and medium for optimizing automatic silicon wafer distribution, which are described below through embodiments.
[0049] Example 1
[0050] To facilitate understanding of this embodiment, a silicon wafer automatic distribution optimization method disclosed in the embodiment of this application is first introduced in detail. Figure 1 The flowchart of a method for optimizing automatic silicon wafer distribution is shown in FIG. Figure 2 A flow chart of another method for optimizing automatic silicon wafer distribution provided by the present application is shown. The method for optimizing automatic silicon wafer distribution provided by the present application includes:
[0051] S101, acquiring parameters and corresponding parameter values of multiple dimensions of the silicon wafers in stock in real time, and processing the parameter values based on preprocessing methods corresponding to the multiple dimensions to obtain parameter feature vectors; different dimensions correspond to different preprocessing methods;
[0052] S102: extracting target parameter requirements from the order requirements, matching the parameter feature vector with the target parameter requirements of the order requirements to obtain a matching feature set; the matching feature set includes multiple matching feature combinations;
[0053] S103, inputting the matching feature set into a pre-trained solution optimization model, so as to process the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimized solution set;
[0054] S104: Filter out an optimal solution set from the optimization solution set, and extract an optimal distribution solution for the target quantity based on the optimal solution set to complete the optimization of the distribution solution.
[0055] In step S101, the parameters of various dimensions of the silicon wafers in stock and the corresponding parameter values are acquired in real time. The acquisition can also be performed at a fixed acquisition frequency, such as one day or half a day. The specific acquisition methods include sensor acquisition, production data acquisition, or online query, etc. The number of silicon wafers in stock is not unique. The collected parameters include physical parameters, electrical parameters, and impurity parameters. The physical parameters include diameter (accuracy ±0.01mm), thickness (accuracy ±0.001mm), surface flatness (accuracy ±0.1μm); crystal orientation (accuracy ±0.5°), electrical parameters include resistivity (Ω·cm, accuracy ±1%), minority carrier lifetime (μs, accuracy ±5%), impurity parameters Including carbon content level (ppma, accuracy ±2%), oxygen content (ppma, accuracy ±2%), and the above eight parameters belong to continuous dimensions and discrete dimensions respectively. The continuous dimension is such as resistivity, and the discrete dimension is such as carbon content. This application corresponds to different preprocessing methods for different dimensions. After obtaining the corresponding parameter values of the above parameters of the silicon wafers in stock, the parameter values are processed based on the preprocessing methods corresponding to multiple dimensions to obtain parameter feature vectors. The parameter feature vectors are used to characterize multiple data features of the silicon wafers in stock. This application also establishes a dynamic inventory database based on the corresponding parameter values of the above parameters of the silicon wafers in stock to store the corresponding parameter values of the above parameters.
[0056] In the specific implementation process of step S101, there is an embodiment: Figure 3 As shown, the preprocessing method based on multiple dimensions corresponding to the parameter values is used to process the parameter values to obtain the parameter feature vector, including:
[0057] S1011. Process the parameter values based on the preprocessing methods corresponding to the multiple dimensions to obtain an intermediate interval matrix;
[0058] S1012. Fusion of the intermediate interval matrices corresponding to the multiple dimensions to obtain a parameter feature vector in matrix form corresponding to the parameter value.
[0059] In steps S1011 - S1012, based on the preprocessing methods corresponding to the multiple dimensions, the parameter values are processed to obtain an intermediate interval matrix. For example, for the parameters of the continuous dimension, the set preprocessing method is interval discretization. Specifically, based on the parameter values, they are graded and then mapped. For the parameters of the discrete dimension, the set preprocessing method is one - hot encoding. Specifically, a one - hot encoding matrix is first created, and then the data is converted. Then, based on the above two methods, the parameter values are processed correspondingly to obtain an intermediate interval matrix. The intermediate interval matrix of the continuous dimension and the intermediate interval matrix of the discrete dimension are fused to obtain the parameter feature vector in matrix form corresponding to the parameter values. The parameter feature vector is used to construct an n×m feature vector matrix, where n is the number of silicon wafers in stock and m is the parameter dimension (≥8).
[0060] Suppose there is a set of resistivity data of silicon wafers in stock, ranging from 1.0 - 1.2 Ω·cm, which is divided into 3 intervals and corresponds to grades A / B / C. The parameter values are [1.05, 1.12, 1.18, 1.07, 1.15], with the unit of (Ω·cm). First, determine the interval boundaries:
[0061] Interval 1: 1.0 - 1.0667 → Grade A
[0062] Interval 2: 1.0667 - 1.1333 → Grade B
[0063] Interval 3: 1.1333 - 1.2 → Grade C;
[0064] And map the parameter values to the intervals. The obtained intermediate interval matrix is:
[0065]
[0066] Suppose there is a set of carbon content grade data of silicon wafers in stock, which are respectively in three grades: low, medium, and high. Then the carbon content grades are: ['low','medium', 'high', 'low','medium']. First, create a one - hot encoding matrix:
[0067] 'low' corresponds to [1, 0, 0]
[0068] 'medium' corresponds to [0, 1, 0]
[0069] 'high' corresponds to [0, 0, 1];
[0070] Then convert the one - hot encoding matrix data. The obtained intermediate interval matrix is:
[0071]
[0072] Assume there are 5 silicon wafers in stock, each with 8 parameters (including resistivity, minority carrier lifetime, carbon content level, etc.). The parameter vector matrix at the construction site is:
[0073]
[0074] Parameter 1 is the grade (A / B / C) corresponding to resistivity, parameter 2 is other continuous parameters such as minority carrier lifetime, and parameter 8 is the one-hot encoding value corresponding to the carbon content grade.
[0075] In step S102, after receiving the order demand imported through the API or file, the target parameter requirements in the order demand are extracted from the order demand, such as the order demand requires a target diameter of 156.75±0.1mm and a target resistivity of 1.0-1.2Ω·cm for the silicon wafer. The priority weights of multiple parameters can also be extracted from the order demand, that is, the priorities of the above eight parameters are not exactly the same, and can be based on the order demand or customized, such as the weight of the minority carrier lifetime parameter is 40%. After obtaining the order demand, the parameter feature vector is matched with the target parameter requirement of the order demand to obtain a matching feature set; the matching feature set includes multiple matching feature combinations; a matching feature combination is a distribution plan, each distribution plan includes a combination of silicon wafers in stock in the quantity required by the order demand, and the parameters of multiple dimensions in the silicon wafers in stock are matched with the target parameter requirements in the order demand to determine the silicon wafers in the silicon wafers in stock that meet the order demand.
[0076] In a specific implementation process of step S102, there is an embodiment in which the matching of the parameter feature vector with the target parameter requirement of the order requirement to obtain a matching feature set includes:
[0077] S1021. Convert the target parameter requirements of the order requirements based on a preset feature conversion method to obtain an order feature vector;
[0078] S1022. Calculate the multi-dimensional similarity between the order feature vector and the parameter feature vector, and obtain a matching feature combination based on the multi-dimensional similarity.
[0079] In steps S1021-S1022, the target parameter requirements of the order requirements are converted based on a preset feature conversion method to obtain an order feature vector. The conversion can be performed in a feature conversion method similar to the parameter value, or in other ways. The target parameter requirements of the order requirements are converted to obtain an order feature vector, and the multi-dimensional similarity between the order feature vector and the parameter feature vector is calculated respectively. The calculation method of the multi-dimensional similarity includes a Euclidean distance calculation network and a cosine similarity calculation network. For details, reference can be made to the Euclidean distance calculation network and the cosine similarity calculation network in the prior art. Based on the multi-dimensional similarity, a matching feature combination of multiple parameter matches is obtained. If the similarities corresponding to the two calculation methods are both 0, it means that the silicon wafers in stock do not match the order requirements, and another batch of silicon wafers in stock needs to be replaced.
[0080] In step S103, after obtaining the matching combination, the matching feature set is input into a pre-trained solution optimization model. The solution optimization model is preset and is used to optimize the matching feature combination in the matching feature set to obtain a better matching feature combination. The matching feature combination in the matching feature set is processed based on the solution optimization model to obtain an optimized solution set. The optimized solution set is obtained by processing the optimized feature combination that is more in line with the order requirements from various aspects or angles compared with the matching feature combination by the solution optimization model, that is, a set of optimized solutions that are more in line with the order requirements. Since each optimized solution can be a candidate distribution solution, the optimized feature set is also a candidate distribution solution set.
[0081] In the specific implementation process of step S103, there is an embodiment in which the processing of the matching feature combinations in the matching feature set based on the solution optimization model to obtain the optimization solution set includes:
[0082] S1031. Based on the relationship between the weight of the parameter and the parameter, determine whether the parameters corresponding to the matching feature combination all meet the preset rapid processing conditions;
[0083] S1032: If yes, process the matching feature combination based on a fast processing method corresponding to the fast processing condition to obtain a corresponding optimization solution set.
[0084] In steps 1031-S1032, in order to achieve the effect of timely response and timely generation of distribution plans, the solution optimization model first makes a judgment based on the weight assigned to each parameter in the order demand and the relationship between the parameters, that is, based on the weight of the parameter and the relationship between the parameters, it judges whether the parameters corresponding to the matching feature combination all meet the preset quick processing conditions. The quick processing conditions are whether the weights of the parameters are clear or whether there is an uncooperative relationship between the parameters. Clear means that the weights are numbers that do not need to be questioned. According to the uncooperative relationship, it is judged whether it is a single parameter processing scenario, otherwise it is a complex parameter processing scenario. The complex parameter processing scenario is such as a scenario that requires simultaneous optimization of yield and cost. If so, for the single parameter processing scenario corresponding to the quick processing condition, the solution optimization model quickly optimizes the matching feature combination based on the quick processing method corresponding to the preset quick processing condition to obtain a corresponding set of optimization solutions.
[0085] In the specific implementation process of step S1032, there is an embodiment in which the matching feature combination is processed in a fast processing manner corresponding to the fast processing condition to obtain a corresponding optimization solution set, including:
[0086] S10321. Calculate multiple attribute values of the parameters corresponding to the matching feature combination and parameter values of the parameters to obtain parameter calculation results;
[0087] S10322. Summarize parameter calculation results corresponding to multiple parameters to obtain a comprehensive distribution score, and obtain the optimization solution set based on the comprehensive distribution score.
[0088] In steps S10321-S10322, when it is determined that the optimization scenario is a single-parameter scenario, processing is performed based on the preset fast processing method of the single-parameter scenario, that is, calculation is performed based on the fast processing method described in formula (1):
[0089] f=w1·M+w2·I+w3·C(1);
[0090] Among them, M is the parameter matching degree of the matching feature combination (0-1, 1 means complete match); I is the inventory priority (calculated based on inventory days and batch size, the higher the turnover rate, the higher the priority); C is the cost coefficient (including procurement cost and transportation cost, the lower the cost, the higher the coefficient); among them, w1+w2+w3=1, the weight can be dynamically configured through the human-computer interface (accuracy 0.1%), the parameter matching degree is the sum of the product of the parameter priority weights and the similarity of the multiple parameters corresponding to the matching feature combination, that is, the parameter matching degree Mi=∑(wp·similarity), among which wp is the parameter priority weight, the inventory priority is Ii=max(inventory turnover rate) inventory turnover rate i (the higher the turnover rate, the closer Ii is to 1), and the cost coefficient is cost. This coefficient Ci = cost i / min (cost) (the lower the cost, the closer Ci is to 1); that is, the multiple attribute values of the parameters corresponding to the matching feature combination and the parameter values of the parameters are calculated to obtain parameter calculation results, and the attribute values include similarity, parameter priority and cost coefficients. The parameter calculation results corresponding to the multiple parameters are summarized to obtain a comprehensive distribution score, and the comprehensive distribution score is sorted in descending order, so as to obtain the optimized feature set including multiple optimized feature combinations based on the comprehensive distribution score, that is, the matching feature combination with a higher comprehensive distribution score is the optimized feature set that meets the order requirements, and the solution optimization model also generates a corresponding optimization solution based on the optimized feature combination in the optimized feature set, and obtains a corresponding optimization solution set based on the obtained optimization solution.
[0091] Regarding the cost coefficient: Assume there are three silicon wafers in stock, namely silicon wafer A, silicon wafer B, and silicon wafer C, with costs of $50, $40, and $60 respectively. First, find the minimum cost: min(cost) = 40. Then calculate the cost coefficient Ci for each silicon wafer:
[0092] For wafer A:
[0093] CA = cost A / min (cost) = 50 / 40 = 1.25
[0094] For wafer B:
[0095] CB = cost B / min (cost) = 40 / 40 = 1
[0096] For wafer C:
[0097] CC = cost C / min (cost) = 60 / 40 = 1.5
[0098] Result Analysis
[0099] The cost coefficient CA of silicon wafer A is 1.25, which means that its cost is 25% higher than the minimum cost.
[0100] The cost coefficient CB of silicon wafer B is 1, which means that its cost is exactly the minimum cost, so CB is close to 1.
[0101] The cost coefficient CC of silicon wafer C is 1.5, which means that its cost is 50% higher than the minimum cost.
[0102] In the specific implementation process of step S1031, there is an embodiment in which, after determining whether all parameters corresponding to the matching feature combination meet the preset fast processing conditions based on the relationship between the weight of the parameter and the parameter, the method further includes:
[0103] S10311. If not, select a partial matching feature combination from the matching feature combination and call different types of iteration methods;
[0104] S10312. Execute the different types of iterative methods to iterate the partial matching feature combinations to obtain an optimization solution set.
[0105] In steps S10311-S10312, if not, it is determined that the optimization scenario is a complex parameter scenario. At this time, a partial matching feature combination is selected from the matching feature combination, and this partial matching feature combination is the better matching feature combination. Different types of iteration methods are called, and the iteration methods include selection, crossover, mutation and other methods for iterative optimization. The specific method of the selection can be roulette, and the specific method of the crossover can be a matching feature combination of a single-point exchange of silicon wafers. The mutation is a random replacement of silicon wafers in inventory. The different types of iteration methods are executed to iterate the partial matching feature combination, and the comprehensive distribution score of the iterated matching feature combination is calculated, thereby obtaining an optimized feature set including multiple optimized feature combinations. The optimized feature set is a matching feature combination with a higher comprehensive distribution score, that is, an optimized feature combination. The solution optimization model also generates a corresponding optimization solution based on the optimized feature combination in the optimized feature set, and obtains a corresponding optimization solution set based on the obtained optimization solution. The iteration termination condition is to terminate the iteration when the preset number of iterations is reached or the fitness meets certain conditions.
[0106] Selection involves selecting a subset of matching feature combinations from the current set as parents, which are then used to generate the next generation of optimized feature combinations. Common selection methods include roulette wheel selection and tournament selection. Roulette wheel selection selects based on the fitness ratio of individuals, with individuals with higher fitness being more likely to be selected.
[0107] Crossover involves pairing multiple selected matching feature combinations, generating optimized feature combinations for the new offspring through a crossover operation. Common crossover methods include point crossover and uniform crossover. Point crossover involves randomly selecting one or more crossover points and exchanging the gene segments between the two matching feature combinations at those points.
[0108] Mutation involves mutating the newly generated optimized feature combinations to increase population diversity. Common mutation methods include bit flipping and insertion. Bit flipping: Randomly select one or more gene bits and change their values.
[0109] In step S104, the best solution set is screened out from the optimization solution set. That is, in order to allow users who generate order requirements to have optionality, this application does not set the number of best distribution solutions in the best solution set to be unique. The number is set specifically according to actual conditions and can be 3. The best distribution solution includes silicon wafer ID, quantity, parameter details, comprehensive score, and advantage description. The advantages of each best distribution solution are marked in Table 1, such as 98% parameter matching for Solution 1, 12% cost reduction for Solution 2, and 20% increase in inventory turnover for Solution 3. The best distribution solution supports API push to the ERP system or manual review, so that users can make decisions or automatically execute distribution, and also realizes the optimization of distribution solutions based on the best distribution solution. This application also feeds back the best distribution solution to the dynamic inventory database for storage.
[0110] Table 1 Best distribution plan
[0111]
[0112] Based on the above, this application has achieved the reduction of distribution efficiency from an average of 2 hours for manual work to ≤5 minutes, an efficiency improvement of ≥80%, a comprehensive parameter matching degree of ≥99.5%, and a yield loss reduction from 5% of the traditional method to below 0.3%; inventory turnover rate has increased by 40%, and obsolete inventory (unused for more than 60 days) has been reduced by 60%; through multi-objective optimization, the cost of a single distribution is reduced by 15%-25% compared with the manual solution. Through configurable weight parameters, it supports differentiated scenarios such as "semiconductor-level high-precision matching" (increasing the weights of resistivity and minority carrier lifetime) and "photovoltaic-level cost-priority matching" (increasing carbon content and cost weight), adapting to the needs of different industries; realizing the synchronous processing of ≥8 core parameters such as size, resistivity, minority carrier lifetime, carbon content, etc., and solving the limitations of traditional single-parameter matching (such as chip failure caused by qualified size but substandard minority carrier lifetime) through eigenvector matrix modeling; inventory data update delay ≤10 seconds, order change response time ≤30 seconds, this application also supports real-time inventory silicon wafer warning (such as automatically triggering alternative plans when a certain parameter is insufficient in inventory); modular design supports parameter expansion, and when adding new parameters (such as metal impurity content, crystal orientation deviation), only data acquisition and eigenvector definition need to be updated, without the need for system reconstruction.
[0113] In the specific implementation process of step S104, there is an embodiment in which the step of selecting the best solution set from the optimization solution set includes:
[0114] S1041. Pre-setting a screening rule for the optimization solution set, and calling the screening rule based on the optimization solution set;
[0115] S1042: Filter the optimization solution set based on the screening rule to obtain the best solution set.
[0116] In steps S1041-S1042, the screening rules are pre-set based on order requirements, and can also be set based on user requirements, such as "Top 3 comprehensive distribution scores", "Top 3 lowest costs", "Top 3 inventory turnover priorities", etc. Based on the above screening rules, multiple optimization schemes in the optimization scheme set are screened to obtain the best solution set obtained by combining the best distribution schemes. The best feature set is obtained by combining the three optimization schemes screened based on the screening rules, that is, each best distribution scheme in the best solution set is a distribution scheme that meets the order requirements.
[0117] Example 2
[0118] This application also provides a silicon wafer automatic distribution optimization device, such as Figure 4The following is a block diagram of a device for optimizing automatic wafer distribution. The functions implemented by this device correspond to the steps of executing the aforementioned method for optimizing automatic wafer distribution on a terminal device. This device can be understood as a component of a server including a processor. The device described in this application includes:
[0119] Acquisition module 401 is used to acquire parameters and corresponding parameter values of multiple dimensions of the inventory silicon wafers in real time, and process the parameter values based on preprocessing methods corresponding to the multiple dimensions to obtain parameter feature vectors; different dimensions correspond to different preprocessing methods;
[0120] Extraction module 402, configured to extract target parameter requirements from the order requirements, and match the parameter feature vector with the target parameter requirements of the order requirements to obtain a matching feature set; the matching feature set includes multiple matching feature combinations;
[0121] An input module 403 is configured to input the matching feature set into a pre-trained solution optimization model, so as to process the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimized solution set;
[0122] The optimization module 404 is configured to select an optimal solution set from the optimization solution set, and extract an optimal distribution solution for the target quantity based on the optimal solution set to complete the optimization of the distribution solution.
[0123] In a feasible implementation, the input module includes:
[0124] A judgment module, configured to judge whether all parameters corresponding to the matching feature combination meet a preset rapid processing condition based on the relationship between the weight of the parameter and the parameter;
[0125] The first processing module is used for the module, and is used for processing the matching feature combination based on the fast processing method corresponding to the fast processing condition to obtain a corresponding optimization solution set.
[0126] In a feasible implementation manner, the input module further includes:
[0127] A first calculation module is used to calculate multiple attribute values of the parameters corresponding to the matching feature combination and parameter values of the parameters to obtain parameter calculation results;
[0128] The summarizing module is used to summarize the parameter calculation results corresponding to multiple parameters to obtain a comprehensive distribution score, so as to obtain the optimization solution set based on the comprehensive distribution score.
[0129] In a feasible implementation, the input module also includes:
[0130] A selection module, for selecting a partial matching feature combination from the matching feature combination if no, and calling different types of iteration methods;
[0131] An execution module is used to execute the different types of iterative methods to iterate the partial matching feature combination to obtain an optimization solution set.
[0132] In a feasible implementation, the optimization module includes:
[0133] a setting module, configured to pre-set screening rules for the set of optimization solutions and call the screening rules based on the set of optimization solutions;
[0134] A screening module is used to screen the optimization solution set based on the screening rules to obtain the best solution set.
[0135] In a feasible implementation, the extraction module includes:
[0136] A conversion module, configured to convert the target parameter requirements of the order requirements based on a preset feature conversion method to obtain an order feature vector;
[0137] The second calculation module is used to calculate the multi-dimensional similarity between the order feature vector and the parameter feature vector, and obtain a matching feature combination based on the multi-dimensional similarity.
[0138] In a feasible implementation, the acquisition module includes:
[0139] A second processing module is configured to process the parameter values to obtain an intermediate interval matrix based on preprocessing methods corresponding to the multiple dimensions;
[0140] A fusion module is used to fuse the intermediate interval matrices corresponding to the multiple dimensions to obtain a parameter feature vector in matrix form corresponding to the parameter value.
[0141] Example 3
[0142] The present application also provides an electronic device, such as Figure 5 As shown, it includes: a processor 501, a memory 502 and a bus 503, the memory 502 stores machine-readable instructions executable by the processor 501, when the electronic device is running, the processor 501 and the memory 502 communicate through the bus 503, when the machine-readable instructions are executed by the processor 501, any one of the steps of the silicon wafer automatic distribution optimization method is performed.
[0143] Example 4
[0144] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the methods for optimizing automatic silicon wafer distribution are executed.
[0145] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.
[0146] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.
[0147] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0148] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, platform server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.
[0149] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.
Claims
1. A method for optimizing automatic silicon wafer distribution, characterized in that: The method comprises: Acquire parameters and corresponding parameter values of multiple dimensions of silicon wafers in stock in real time, and process the parameter values based on preprocessing methods corresponding to the multiple dimensions to obtain parameter feature vectors; different dimensions correspond to different preprocessing methods; Extracting target parameter requirements from the order requirements, matching the parameter feature vector with the target parameter requirements of the order requirements to obtain a matching feature set; the matching feature set includes multiple matching feature combinations; Inputting the matching feature set into a pre-trained solution optimization model to process the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimized solution set; An optimal solution set is screened out from the optimization solution set, and an optimal distribution solution for the target quantity is extracted based on the optimal solution set to complete the optimization of the distribution solution.
2. The method according to claim 1, characterized in that The step of processing the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimization solution set includes: Based on the relationship between the weight of the parameter and the parameter, determining whether the parameters corresponding to the matching feature combination all meet the preset fast processing conditions; If so, the matching feature combination is processed based on a fast processing method corresponding to the fast processing condition to obtain a corresponding optimization solution set.
3. The method according to claim 2, characterized in that The processing of the matching feature combination in a rapid processing manner corresponding to the rapid processing condition to obtain a corresponding optimization solution set includes: Calculating multiple attribute values of the parameters corresponding to the matching feature combination and parameter values of the parameters to obtain parameter calculation results; The parameter calculation results corresponding to the various parameters are aggregated to obtain a comprehensive distribution score, so as to obtain the optimization solution set based on the comprehensive distribution score.
4. The method according to claim 2, characterized in that After determining whether all parameters corresponding to the matching feature combination meet a preset fast processing condition based on the relationship between the weight of the parameter and the parameter, the method further includes: If not, then select a partial matching feature combination from the matching feature combination and call different types of iteration methods; The different types of iterative methods are executed to iterate the partial matching feature combination to obtain an optimization solution set.
5. The method according to claim 1, wherein The step of selecting the best solution set from the optimization solution set includes: Presetting a screening rule for the set of optimization solutions, and calling the screening rule based on the set of optimization solutions; The optimization solution set is screened based on the screening rules to obtain the best solution set.
6. The method according to claim 1, characterized in that The matching of the parameter feature vector with the target parameter requirement of the order requirement to obtain a matching feature set includes: Convert the target parameter requirements of the order requirements based on a preset feature conversion method to obtain an order feature vector; The multi-dimensional similarity between the order feature vector and the parameter feature vector is calculated, and a matching feature combination is obtained based on the multi-dimensional similarity.
7. The method according to claim 1, characterized in that The preprocessing method based on multiple dimensions corresponding to the parameter values is used to process the parameter values to obtain the parameter feature vector, including: Based on the preprocessing methods corresponding to the multiple dimensions, the parameter values are processed to obtain an intermediate interval matrix; The intermediate interval matrices corresponding to the multiple dimensions are fused to obtain a parameter feature vector in matrix form corresponding to the parameter value.
8. A silicon wafer automatic distribution optimization device, characterized in that: The device comprises: An acquisition module is used to obtain parameters and corresponding parameter values of multiple dimensions of the silicon wafers in stock in real time, and process the parameter values based on the preprocessing methods corresponding to the multiple dimensions to obtain parameter feature vectors; different dimensions correspond to different preprocessing methods; An extraction module is used to extract target parameter requirements from the order requirements, and match the parameter feature vector with the target parameter requirements of the order requirements to obtain a matching feature set; the matching feature set includes multiple matching feature combinations; An input module, configured to input the matching feature set into a pre-trained solution optimization model, so as to process the matching feature combinations in the matching feature set based on the solution optimization model to obtain an optimized solution set; The optimization module is used to screen out the best solution set from the optimization solution set, and extract the best distribution solution of the target quantity based on the best solution set to complete the optimization of the distribution solution.
9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the machine-readable instructions are executed by the processor, the steps of the method for optimizing automatic silicon wafer distribution as described in any one of claims 1 to 7 are performed.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for optimizing automatic silicon wafer distribution as claimed in any one of claims 1 to 7.