A tin-based solder alloy composition optimization method and system for microelectronic interconnection based on wetting force prediction
By constructing a multi-dimensional feature database and an improved wetting force prediction model, combined with SHAP analysis and Bayesian optimization algorithm, the problem of optimizing the wetting performance of tin-based solder sheets was solved, achieving efficient and economical alloy composition optimization, which is suitable for the high-density, miniaturization, and high-frequency requirements of microelectronic interconnect materials.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- KUNMING UNIV OF SCI & TECH
- Filing Date
- 2026-04-10
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies struggle to accurately quantify the wetting properties of tin-based solder sheets, especially the synergistic mechanism of trace elements in the Sn-Ag-Cu-X multi-element system. This results in long development cycles and high costs for new alloys, and makes it difficult to achieve targeted optimization of wetting properties.
A multi-dimensional feature database was constructed, and an improved wetting force prediction model was adopted, combined with SHAP analysis and goal-oriented optimization. The composition of tin-based solder sheet alloy was predicted by machine learning. Temperature-composition coupled feature mapping, multi-scale weighted standardization and wetting force specific loss function were introduced, and the alloy composition was optimized by combining Bayesian optimization algorithm.
It significantly improves the efficiency of tin-based solder alloy composition optimization, provides more accurate guidance for wetting mechanism research, ensures that the alloy composition meets the requirements of production process and economy, and reduces R&D costs.
Smart Images
Figure CN122436076A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of microelectronic packaging material design and artificial intelligence technology, specifically to a method and system for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction. Background Technology Microelectronic soldering, a core process in electronics manufacturing, directly determines the reliability and lifespan of electronic devices through its connection quality. As electronic devices evolve towards higher density, miniaturization, and higher frequency, extremely stringent requirements are placed on the comprehensive performance of microelectronic interconnect materials. Tin-based solder pads, as a key basic material in electronic packaging, have their wetting properties being one of the most critical indicators for evaluating their solderability and soldering quality. Wetting force (Fmax) directly reflects the interfacial bonding ability and spreading behavior of the solder with the substrate (such as copper or nickel) in the molten state, and is a key factor determining solder joint formation, void ratio, and long-term reliability.
[0002] Currently, the development of high-performance tin-based solder sheets mainly relies on the traditional "thermodynamic calculation-experimental trial and error" model. Although thermodynamic software such as Thermo-Calc can assist in phase diagram calculations and initial composition screening, the synergistic mechanism of trace elements (such as Ga, Ce, Ni, Bi, etc.) in Sn-Ag-Cu-X multi-component systems is complex. In particular, the wetting process is strongly coupled with the nonlinear effect of temperature and composition, making it difficult to accurately quantify their influence using traditional methods. For example, within the welding temperature range of 230℃ to 280℃, the contributions of different alloying elements to the surface tension, viscosity, and interfacial reaction of the melt vary greatly, resulting in complex wetting force variations that are difficult to accurately predict using empirical formulas or simple linear models. This leads to long development cycles and high costs for new alloys, and makes it difficult to achieve targeted optimization of wetting performance.
[0003] Machine learning technology offers a novel solution to these challenges. By learning the complex mapping relationship between "composition-process-wetting properties" from historical experimental data, machine learning models can quickly predict the wetting force of new compositions and guide optimization directions. However, effectively applying machine learning to the design of tin-based solder wetting properties still faces three major challenges: First, wetting force is significantly affected by soldering temperature, requiring the construction of feature engineering that incorporates the temperature-composition coupling effect; second, the model's prediction results need to be interpretable to guide metallurgists in understanding the interaction mechanism between temperature and elements; and third, it is necessary to closely integrate prediction with goal-oriented reverse optimization to directly output an engineered composition scheme that meets a specific wetting force threshold.
[0004] Therefore, developing an intelligent design method that integrates temperature-composition coupling modeling, physical information-driven machine learning, interactive interpretability analysis, and goal-oriented reverse optimization is of great significance for breaking through the research and development bottleneck of high-performance tin-based solder sheets. Summary of the Invention
[0005] To overcome the above-mentioned technical problems, the present invention provides a method and system for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnection based on wetting force prediction.
[0006] To achieve the above-mentioned technology, the present invention is demonstrated in two aspects. Firstly, the present invention provides a method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction, comprising the following steps: S1. Construct a multi-dimensional feature database containing tin-based solder sheet alloy composition, welding process characteristics, and wetting performance indicators; S2. Wetting force prediction is performed based on a multi-dimensional feature database and a wetting force prediction model. The wetting force prediction model takes alloy composition, welding process characteristics and coupling feature mapping results as inputs, and wetting force as the prediction target. It combines an improved multi-scale weighted normalization method and an improved wetting force-specific loss function for model training. The prediction accuracy is evaluated by 5-fold cross-validation, and the prediction results are output. S3. Based on the wetting force prediction model, the improved SHAP analysis method is used for interpretation; S4. Based on the prediction results of the wetting force prediction model and the SHAP analysis results, construct an objective function to minimize the deviation between the predicted wetting force and the target wetting force and the raw material cost, and solve it to obtain the optimal alloy composition; S5. Verify the optimal alloy composition through experiments and iterative model calculations, and complete the closed-loop design of the tin-based solder alloy composition optimization method.
[0007] Specifically, the alloy composition includes Sn, Ag, Cu and trace elements, the welding process feature is the welding temperature, and the wetting performance index is the wetting force. After performing outlier processing, extreme value interference removal, standard normalization, and dimensional difference elimination on the collected alloy composition, welding process feature, and wetting performance index, a multi-dimensional feature database is obtained.
[0008] Specifically, the coupling feature map is calculated using an improved temperature-composition coupling feature map method, as shown in the following expression: In the formula, The effective wetting temperature is T, where T is the actual welding temperature. For the first feature map The content of each element This is the coupling coefficient for the element's sensitivity to wetting temperature. A collection of elements The index of the element.
[0009] Specifically, the improved multi-scale weighted normalization method is used to normalize input features before machine learning to preserve feature importance weights, and its expression is as follows: In the formula, Indicates the first The final value after standardization of each feature; Representing the The weight coefficients of each feature; Indicates the first The median of the features; Representing the The range of each characteristic; Indicates the first The mean of each feature; Indicates the first Standard deviation of each feature; Indicates the first The original values of each feature.
[0010] Specifically, the improved wetting force-specific loss function integrates four factors—absolute error, relative error, model complexity penalty, and component out-of-bounds penalty—to improve model prediction accuracy while preventing overfitting.
[0011] Specifically, the improved expression for the wetting force specificity loss function is as follows: In the formula, For the first Measured wetting power of each sample For the first The predicted wettability of each sample, where β represents the weighting coefficient that emphasizes the overall numerical deviation; and γ represents the weighting coefficient that emphasizes the relative error. λ is a small constant to prevent the denominator from being zero; λ is the regularization coefficient, and Ω(θ) is the model complexity penalty term. This indicates a penalty for exceeding the composition limit, which constrains the predicted alloy composition to conform to the feasible range of industrial processes, thus avoiding formulations that have no practical processing significance. This represents the total number of samples.
[0012] Specifically, the improved SHAP analysis method first calculates the improved SHAP weights, as shown in the following expression: In the formula, Indicates the first The dynamic SHAP value of each feature; For the first Traditional SHAP values for each feature; Represents the nonlinear sensitivity coefficient; The model output represents the first... The second-order partial derivatives of each characteristic are used to quantify the nonlinear effect of the element on the wetting force; Indicates the first The coefficient of variation of each feature, and the relative dispersion of element content; Next, the importance of SHAP features is calculated, as shown in the following expression: In the formula, Indicates the first The feature in the first Dynamic SHAP values for each sample; Finally, the interpretation result of each feature is obtained through SHAP feature importance.
[0013] Specifically, the objective function is expressed as follows: In the formula, This is the predicted value for wetting power. Target wetting power; This represents the cost regularization coefficient; Representing the The cost coefficient of an element reflects its market cost.
[0014] Specifically, the optimal alloy composition is obtained as follows: an optimization algorithm is used to search the objective function to obtain candidate alloy compositions, a comprehensive score is calculated based on the candidate alloy compositions, and the alloy formula with the best overall performance is output; the expression for calculating the comprehensive score is as follows: In the formula, This represents the overall score of alloy wetting force; κ is the synergistic effect weighting coefficient. This represents the synergistic effect term of an element.
[0015] Secondly, the present invention provides a tin-based solder alloy composition optimization system for microelectronic interconnects based on wetting force prediction, comprising a data management module for performing step S1; a model training and prediction module for performing step S2; an interpretability analysis module for performing step S3; a composition optimization module for performing step S4; and a user interface for performing step S5.
[0016] Beneficial effects of the present invention This invention significantly improves optimization efficiency: the innovative wetting force-driven inverse optimization function improves the convergence speed of optimization by introducing cost regularization and Bayesian optimization strategies, and effectively avoids getting trapped in local optima.
[0017] The physical significance of this invention is clearer: the combination of dynamic SHAP weight update mechanism and temperature-composition coupling characteristics can better capture the nonlinear synergistic effect and temperature sensitivity between trace elements, providing more accurate theoretical guidance for the study of wetting mechanism.
[0018] This invention enhances engineering practicality: by using a component out-of-bounds penalty term and cost regularization, it ensures that the recommended components not only meet wetting performance requirements but also comply with actual production processes and economic constraints. Attached Figure Description
[0019] Figure 1 This is a flowchart of the steps of the present invention; Figure 2 Training set prediction results; Figure 3 Test set prediction results; Figure 4 Feature importance histogram; Figure 5 Importance distribution pie chart; Figure 6 Ag's SHAP dependency graph; Figure 7 The interaction SHAP dependency graph between temperature and Bi; where, Figure 7 (a) is a temperature SHAP dependency plot; Figure 7 (b) is the SHAP dependency graph of Bi. Detailed Implementation
[0020] The present invention will be further described in detail below with reference to specific embodiments.
[0021] See Figure 1 A method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction includes the following steps: S1. Data Acquisition and Feature Engineering; Wetness test data of tin-based solder alloys were collected, and a multi-dimensional feature database containing alloy composition, welding process characteristics and wetting performance indicators was constructed. Historical data on tin-based solder alloys were obtained by collecting wetting force test data of tin-based solders from historical experiments, literature, and thermodynamic simulations. The alloy composition includes: Sn, Ag, Cu and trace elements (Ga, Ce, Ni, Bi, In, Mn, etc.); The welding process characteristics are: welding temperature (T, unit: °C); The wetting performance index is: wetting force (F) max (Unit: mN) Outlier processing, extreme value removal, standard normalization, and dimension difference elimination were performed on the collected data. In this embodiment, a total of 273 sets of valid experimental data were collected, covering variable ranges such as Ag content 0.3-4.0 wt%, Cu content 0-1.2 wt%, and welding temperature 230-280℃.
[0022] S2, Machine Learning Model Construction and Training; Based on traditional standardization methods, traditional loss functions, and SHAP importance results for each alloying element, we construct an improved temperature-composition coupled feature mapping method, an improved wetting force-specific loss function, an improved multi-scale weighted standardization method, and an improved SHAP weight calculation method, and use machine learning to predict wetting force. Machine learning includes gradient boosting decision trees, random forests, and support vector machines. This invention selects the gradient boosting decision tree (XGBoost) as the machine learning model. The model takes alloy composition, welding process characteristics, and coupling feature mapping results as inputs, and wetting force as the prediction target. It combines an improved multi-scale weighted normalization method and an improved wetting force-specific loss function for model training. The prediction accuracy is evaluated through 5-fold cross-validation, and the predicted wetting force is output. The results are as follows: Figure 2 and Figure 3 As shown; The improved temperature-composition coupled feature mapping method is calculated as follows: In the formula, The effective wetting temperature is T, where T is the actual welding temperature. For the first feature map The content of each element (wt%). The coupling coefficient of this element to the sensitivity of wetting temperature is obtained by fitting experimental data. This feature mapping method physically couples welding temperature and alloy composition, reflecting the synergistic effect of different elements on melt wetting behavior at a specific temperature, and solving the defect of isolated modeling of temperature and composition in traditional methods. In this embodiment, the contribution of different elements to the sensitivity of wetting temperature is obtained by fitting 273 sets of experimental data. Before performing machine learning, an improved multi-scale weighted normalization method is used to replace the traditional Z-score normalization to standardize the input features, in order to preserve the feature importance weights. The expression is as follows: In the formula, Indicates the first The final value after standardization of each feature is used to eliminate the dimensions while retaining the feature importance weights. Representing the The weight coefficients of each feature are calculated as follows: , Indicates the first The contribution weights of various alloying elements to wettability (obtained by normalization of traditional SHAP values) reflect the core role of elements in metallurgy. Indicates the first The median of a feature is more resistant to extreme values than the mean; Representing the The range of each characteristic; Indicates the first The mean of each feature; Indicates the first Standard deviation of each feature; Indicates the first The original values of each feature; The prediction is performed using an improved wetting force-specific loss function instead of the traditional mean square error formula, as shown in the following expression: In the formula, For the first Measured wetting power of each sample For the first The predicted wettability of each sample is given by β, which represents the weighting coefficient that emphasizes the overall numerical deviation and is set to 0.5 in this invention; γ represents the weighting coefficient that emphasizes the relative error (to avoid the influence of dimensions) and is set to 0.4 in this invention. λ is a small constant to prevent the denominator from being zero; λ is the regularization coefficient, and Ω(θ) is the model complexity penalty term. This represents a penalty term for out-of-bounds composition, constraining the predicted alloy composition to conform to the feasible range of industrial processes, and avoiding formulations that have no practical processing significance. The expression is as follows: In the formula, Representing the The absolute value of the SHAP feature importance of each element, i.e. the penalty weight: the core element (such as Ag with a high SHAP value) is penalized more severely when it goes out of bounds, which is in line with the laws of metallurgy. For the first The content of each element; and They represent the first The industrial feasibility upper and lower limits of the elements; η is the out-of-bounds penalty coefficient, which is taken as 10 in this invention; The method for calculating the importance of SHAP features is as follows: In the formula, Indicates the first The feature in the first The dynamic SHAP values of each sample; where, The SHAP weights are obtained through an improved calculation method, and their expression is as follows: In the formula, Indicates the first The dynamic SHAP value of each feature; For the first Traditional SHAP values for each feature; This represents the nonlinear sensitivity coefficient, ranging from 0.1 to 0.3, which controls the correction magnitude of the nonlinear term to the SHAP value (the stronger the nonlinearity of the alloy system, the higher the sensitivity coefficient). (The larger the value) The model output represents the first... The second-order partial derivatives of each feature quantify the nonlinear effect of the element on wetting force (e.g., after the Ag content increases to a certain threshold, the rate of increase in wetting force changes abruptly). Indicates the first The coefficient of variation of each characteristic, and the relative dispersion of element content.
[0023] S3. Model interpretability analysis and feature selection; An improved SHAP analysis method was used to interpret the trained wetting force prediction model, quantify the contribution of each input feature (alloying element and welding temperature) to the output wetting force, identify key elements and their optimal content ranges, and visualize the interaction between temperature and elements. The results are as follows: Figure 4 (Feature Importance Bar Chart) Figure 5 (Importance distribution pie chart) Figure 6 (Ag's SHAP dependency graph) and Figure 7 middle Figure 7 (a) and Figure 7 (b) (SHAP dependency graph of temperature and Bi interaction) is shown.
[0024] S4. Goal-oriented component reverse optimization; With the target wetting force as the optimization direction, and combined with the obtained alloying elements and their optimal content range, a single-objective optimization model is established. A Bayesian optimization algorithm is used to search for the global optimal solution set in the composition space. A multi-starting point L-BFGS-B optimization algorithm is used, with five sets of empirical initial composition combinations preset (including intermediate value combination, high wetting tendency combination, low cost combination, high Ag combination, and low Ag high Bi combination). The local optimal solution is searched within the feasible boundary of element concentration, and the recommended alloy composition is output. The optimization algorithm aims to minimize the deviation between the predicted wetting force and the target wetting force, while simultaneously minimizing raw material costs. First, based on the traditional optimization objective function, the improved expression for constructing the wetting force-driven optimization command is as follows: In the formula, This is the predicted value for wetting power. Target wetting power; This represents the cost regularization coefficient; Representing the The cost coefficient of an element reflects its market cost. Next, based on the optimization instructions, to achieve precise and targeted optimization of wetting force, a core reasoning formula is constructed that integrates the nonlinear contributions, dynamic weights, and interaction effects of elements. This quantifies the comprehensive contribution of the alloy formulation to the target wetting force. The comprehensive score of the alloy wetting force is a weighted fusion of the predicted wetting force component and the elemental synergistic effect term, as shown in the following expression: In the formula, This represents the overall score of alloy wetting force; κ is the synergistic effect weighting coefficient. Represents the synergistic effect term of elements; Among them, the elemental synergistic effect term The expression is as follows: In the formula, Indicates the first Type of element and the first Synergy coefficients among elements; The index function is a determination function that determines the pair of elements. The interaction importance is 1 when it exceeds the threshold τ, and 0 otherwise; where τ=0.01, only element pairs with significant SHAP interaction importance (such as Ag-Cu, Ga-Ce) are retained. The synergy coefficient is determined by experimentally verified elemental interaction rules (e.g., Ag-Cu eutectic synergy). =0.6, Ga-Ce synergistic take =0.4); Finally, an optimization algorithm is used to search for the global optimal solution set in the composition space. Specifically, the Bayesian optimization algorithm is used to search with the comprehensive score of wetting force as the optimization objective, and the candidate alloy composition that achieves the best balance between wetting force and cost is output.
[0025] S5. Experimental verification and model iteration; The optimal alloy composition recommended by S4 was experimentally prepared and its wetting performance was tested. The obtained experimental data was fed back into the database to retrain and optimize the wetting performance prediction model, forming a closed-loop iterative optimization system.
[0026] To verify the present invention, for the Sn-Ag-Cu-Bi-In alloy with high wetting force, the target performance was set to achieve a wetting force of more than 6.5 mN at a welding temperature of 260℃. The results are shown in Table 1. Table 1: Prediction Results of High Wetting Strength Sn-Ag-Cu-Bi-In Alloys Performance prediction results: Wetting force: 6.48 mN (excellent); Welding temperature: 260℃; The invention demonstrates that by adding appropriate amounts of Bi and In, combined with optimized Ag content, the wetting performance of solder at high temperatures is significantly improved, making it suitable for the requirements of high-density microelectronic packaging for excellent wetting properties.
[0027] To verify the present invention, for the low-cost wettability Sn-Ag-Cu-Bi alloy, the target performance was set as a wetting force of not less than 4.5 mN at a welding temperature of 250℃, while minimizing the Ag content to reduce costs. The results are shown in Table 2. Table 2: Prediction results of low-cost wettability Sn-Ag-Cu-Bi alloy Performance prediction results: Wetting force: 4.62 mN (good); Welding temperature: 250℃; Estimated cost index: 18.2 (significantly lower than the 28.5 of the high wetting power formulation); The present invention demonstrates that by reducing the Ag content and appropriately increasing the Bi content, the raw material cost is effectively controlled while ensuring that the wetting force meets the basic requirements, making it suitable for large-scale industrial applications.
[0028] To verify the present invention, for the Sn-Ag-Cu-Ga-Ce alloy with high wetting force, the target performance was set to achieve a wetting force of more than 5.0 mN under low temperature welding conditions of 240℃. The results are shown in Table 3. Table 3: Prediction results of high wettability Sn-Ag-Cu-Ga-Ce alloys Performance prediction results: Wetting force: 5.13 mN (good); Welding temperature: 240℃; The invention demonstrates that by adding trace amounts of Ga and Ce elements, good wetting properties can still be obtained at lower welding temperatures, making it suitable for low-temperature welding scenarios involving heat-sensitive components.
[0029] To verify the present invention, the target performance of the Sn-Ag-Cu-Ni-Ce-Ga alloy was set: to achieve the best balance between wetting force (≥5.5 mN), welding temperature (250℃) and cost. The results are shown in Table 4. Table 4: Prediction Results of the Overall Optimal Sn-Ag-Cu-Ni-Ce-Ga Alloy Performance prediction result: Wetting force 5.67 mN (good); Welding temperature 250℃; estimated cost index 24.3.
[0030] The invention demonstrates that by using the synergistic effect of multiple trace elements, a good balance can be achieved between wetting performance and cost, making it suitable for general-purpose high-precision solder pad requirements.
[0031] A system for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction includes: a data management module, a model training and prediction module, an interpretability analysis module, a composition optimization module, and a user interface. The data management module is used to execute the steps described in S1, specifically including: reading the wetting force experimental data file, removing outliers and filling missing values for alloy composition, welding temperature and measured wetting force values, and constructing a multi-dimensional feature database containing the contents of trace elements such as Ag, Cu, Ga, Ce, Ni, Bi, In, and Mn and welding temperature; The model training and prediction module is used to perform the steps described in S2, specifically including: performing temperature-composition coupled feature mapping calculations to generate effective wetting temperatures. As an extended feature, an XGBoost regressor is used as the base learner, with alloy composition, welding temperature, and coupling features as the basis. The model is trained by combining an improved multi-scale weighted normalization method and an improved wetting force-specific loss function with the input of the model and the output of the wetting force Fmax. The prediction accuracy is evaluated by 5-fold cross-validation and the predicted wetting force is output. The interpretability analysis module is used to perform the steps described in S3, specifically including: calculating the traditional SHAP value based on the trained XGBoost model, further calculating the dynamic SHAP value by introducing the second-order partial derivative and the coefficient of variation, generating a feature importance ranking bar chart, an importance distribution pie chart, and a SHAP dependency graph of key elements (Ag, Bi, temperature, etc.), and visualizing the nonlinear contribution and interaction effect of each alloy element and welding temperature on wetting force; The component optimization module is used to perform the steps described in S4, specifically including: achieving a target wetting power set by the user. The objective function is constructed with welding temperature T as a constraint; a Bayesian optimization algorithm is used to search for the global optimal solution set in the composition space; a multi-starting point L-BFGS-B optimization algorithm is used, with five sets of empirical initial composition combinations preset (including intermediate value combination, high wetting tendency combination, low cost combination, high Ag combination, and low Ag high Bi combination); local optimal solutions are searched within the feasible boundary of element concentration, and recommended alloy composition is output. The user interface is used to execute the steps described in S5, specifically including: providing a graphical interface that supports user-defined numerical input of target wetting force and welding temperature, real-time display of optimization calculation progress and iteration count, and displaying recommended alloy composition (content of each element), predicted wetting force, error with target value and cost index in a structured table format. It also supports feeding back optimization results and experimental verification data to the database to trigger incremental model training and closed-loop iteration.
[0032] The specific embodiments of the present invention have been described in detail above with reference to examples. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.
Claims
1. A method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction, characterized in that, Includes the following steps: S1. Construct a multi-dimensional feature database containing tin-based solder sheet alloy composition, welding process characteristics, and wetting performance indicators; S2. Wetting force prediction is performed based on a multi-dimensional feature database and a wetting force prediction model. The wetting force prediction model takes alloy composition, welding process characteristics and coupling feature mapping results as inputs, and wetting force as the prediction target. It combines an improved multi-scale weighted normalization method and an improved wetting force-specific loss function for model training. The prediction accuracy is evaluated by 5-fold cross-validation, and the prediction results are output. S3. Based on the wetting force prediction model, the improved SHAP analysis method is used for interpretation; S4. Based on the prediction results of the wetting force prediction model and the SHAP analysis results, construct an objective function to minimize the deviation between the predicted wetting force and the target wetting force and the raw material cost, and solve it to obtain the optimal alloy composition; S5. Verify the optimal alloy composition through experiments and iterative model calculations, and complete the closed-loop design of the tin-based solder alloy composition optimization method.
2. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 1, characterized in that, The alloy composition includes Sn, Ag, Cu and trace elements, the welding process feature is welding temperature, and the wetting performance index is wetting force. After performing outlier processing, extreme value interference removal processing, standard normalization processing and dimension difference elimination processing on the collected alloy composition, welding process feature and wetting performance index, a multi-dimensional feature database is obtained.
3. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 2, characterized in that, The coupling feature map is calculated using an improved temperature-composition coupling feature map method, and the expression is as follows: In the formula, The effective wetting temperature is T, where T is the actual welding temperature. For the first feature map The content of each element This is the coupling coefficient for the element's sensitivity to wetting temperature. A collection of elements The index of the element.
4. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 3, characterized in that, The improved multi-scale weighted normalization method is used to normalize input features before machine learning to preserve feature importance weights, and its expression is as follows: In the formula, Indicates the first The final value after standardization of each feature; Representing the The weight coefficients of each feature; Indicates the first The median of the features; Representing the The range of each characteristic; Indicates the first The mean of each feature; Indicates the first Standard deviation of each feature; Indicates the first The original values of each feature.
5. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 4, characterized in that, The improved wetting force-specific loss function integrates four factors: absolute error, relative error, model complexity penalty, and component out-of-bounds penalty, to improve model prediction accuracy while preventing overfitting.
6. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 5, characterized in that, The improved wetting force specificity loss function is expressed as follows: In the formula, For the first Measured wetting power of each sample For the first The predicted wettability of each sample, where β represents the weighting coefficient that emphasizes the overall numerical deviation; and γ represents the weighting coefficient that emphasizes the relative error. λ is a small constant to prevent the denominator from being zero; λ is the regularization coefficient, and Ω(θ) is the model complexity penalty term. This indicates a penalty for exceeding the composition limit, which constrains the predicted alloy composition to conform to the feasible range of industrial processes, thus avoiding formulations that have no practical processing significance. This represents the total number of samples.
7. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 6, characterized in that, The improved SHAP analysis method first calculates the improved SHAP weights, expressed as follows: In the formula, Indicates the first The dynamic SHAP value of each feature; For the first Traditional SHAP values for each feature; Represents the nonlinear sensitivity coefficient; The model output represents the first... The second-order partial derivatives of each characteristic are used to quantify the nonlinear effect of the element on the wetting force; Indicates the first The coefficient of variation of each feature, and the relative dispersion of element content; Next, the importance of SHAP features is calculated, as shown in the following expression: In the formula, Indicates the first The feature in the first Dynamic SHAP values for each sample; Finally, the interpretation result of each feature is obtained through SHAP feature importance.
8. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 7, characterized in that, The objective function is expressed as follows: In the formula, This is the predicted value for wetting power. Target wetting power; This represents the cost regularization coefficient; Representing the The cost coefficient of an element reflects its market cost.
9. The method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction according to claim 8, characterized in that, The method for obtaining the optimal alloy composition is as follows: An optimization algorithm is used to search the objective function to obtain candidate alloy compositions; a comprehensive score is calculated based on the candidate alloy compositions; and the alloy formula with the best overall performance is output. The expression for calculating the comprehensive score is as follows: In the formula, This represents the overall score of alloy wetting force; κ is the synergistic effect weighting coefficient. This represents the synergistic effect term of an element.
10. A system for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction, the system being used to implement the method for optimizing the alloy composition of tin-based solder sheets for microelectronic interconnects based on wetting force prediction as described in claims 1-9, characterized in that... The system includes a data management module for performing step S1; a model training and prediction module for performing step S2; an interpretability analysis module for performing step S3; a component optimization module for performing step S4; and a user interface for performing step S5.