An optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design

Abstract

This invention belongs to the field of material composition design and corrosion performance optimization technology. It proposes an optimization method for high corrosion-resistant beryllium aluminum alloys based on artificial intelligence inversion design. This method organically integrates forward predictive modeling and backward iterative optimization, constructing a closed-loop design system driven by target performance, bounded by multi-dimensional constraints, and centered on intelligent parameter space search. The method involves the systematic construction and standardized preprocessing of historical beryllium aluminum alloy data. Based on the completed dataset, a clear target corrosion resistance performance index is set as the constraint and optimization objective for the inversion design. A high-precision forward predictive model is constructed. Based on this model, the invention establishes a backward solution mechanism constrained by the target corrosion resistance performance. Advantages of this invention: This invention achieves breakthrough progress in both material inversion design theory and engineering optimization of high corrosion-resistant beryllium aluminum alloys, possessing excellent prospects for promotion and application value.

Description

Technical Field

[0001] This invention belongs to the field of material composition design and corrosion performance optimization technology. Specifically, it relates to a design method that utilizes artificial intelligence inversion technology to reverse-engineer and collaboratively optimize the composition and preparation process parameters of beryllium aluminum alloys under the constraint of target corrosion resistance performance. This method can directly start from the desired corrosion resistance performance index and quickly obtain the alloy composition ratio and process window that meet the engineering manufacturability requirements, significantly improving the research and development efficiency of high corrosion-resistant beryllium aluminum alloys. Background Technology

[0002] Beryllium aluminum alloys, with their low density, high specific stiffness, excellent thermal stability, and good dimensional stability, occupy an irreplaceable position in aerospace, precision instruments, electronic packaging, and high-end structural components. However, the corrosion resistance of these alloys in complex service environments, especially their ability to resist chloride ion-induced pitting and uniform corrosion, has always been one of the key bottlenecks restricting their wider engineering applications. Corrosion not only leads to surface deterioration and decreased mechanical properties, but may also trigger sudden failures during long-term service, endangering structural safety.

[0003] To improve the corrosion resistance of beryllium-aluminum alloys, it is usually necessary to systematically optimize their alloy composition (beryllium content, aluminum content, and the types and concentrations of trace additives) and preparation process parameters (melting temperature, cooling rate, heat treatment regime, etc.). Traditional methods mainly rely on an iterative model of "empirical proportioning - prototype preparation - corrosion testing - feedback adjustment." This trial-and-error research paradigm has three fundamental drawbacks: First, due to the high dimensionality of the parameter space and the complex nonlinear interactions between factors, relying solely on experimental exploration is extremely inefficient, often requiring months or even years, resulting in high costs; second, traditional methods are prone to getting trapped in local optima, making it difficult to find the truly optimal composition-process combination globally; third, a large amount of experimental data is not effectively utilized, leading to slow knowledge accumulation and difficulty in forming reusable design criteria.

[0004] In recent years, data-driven materials design methods, especially machine learning-assisted materials performance prediction models, have provided new insights into solving the aforementioned problems. However, most existing methods fall under the category of forward prediction, which establishes a mapping relationship from composition and process to performance, providing the corresponding performance prediction value for the given input and output. But in practical engineering scenarios, designers often face the inverse problem: given the desired target performance (e.g., corrosion current density below a certain threshold), they need to deduce the composition and process parameters that can achieve this target. Forward models cannot directly answer this question. A few studies have attempted to use exhaustive search or genetic algorithms to repeatedly call forward models to approximate the target, but these methods are computationally inefficient and lack effective constraints on the physical rationality and process feasibility of the generated scheme, resulting in output design parameters that either violate the basic principles of materials thermodynamics or cannot be achieved in actual production.

[0005] Therefore, there is an urgent need to invent a novel optimization method for highly corrosion-resistant beryllium aluminum alloys with end-to-end inversion capabilities. This method should be able to take the target corrosion resistance as input and directly solve in reverse the process to find the optimal composition ratio and process parameter combination that meets the performance requirements and conforms to the constraints of composition range and process feasibility. This would fundamentally change the paradigm of traditional material design, significantly shorten the R&D cycle, and improve design quality. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide an optimization method for high corrosion-resistant beryllium aluminum alloys based on artificial intelligence inversion design. This method aims to solve the problems of long development cycles, high costs, and insufficient parameter space exploration in traditional trial-and-error methods, while overcoming the fundamental defects of existing forward prediction models, such as the inability to directly perform inverse solving and the lack of process feasibility constraints. By constructing a forward prediction model and establishing an efficient inversion optimization mechanism based on it, end-to-end inverse design from the target corrosion resistance performance to the composition and process parameters that meet engineering constraints is achieved, significantly improving the optimization efficiency of the corrosion resistance performance of beryllium aluminum alloys and the engineering practicality of the design scheme.

[0007] To achieve the aforementioned objectives and technical effects, this invention proposes an optimization method for high corrosion-resistant beryllium aluminum alloys based on artificial intelligence inverse design. This method organically integrates forward predictive modeling with backward iterative optimization, constructing a closed-loop design system driven by target performance, bounded by multi-dimensional constraints, and centered on intelligent search of the parameter space. Specifically, this method includes the following interrelated and progressively advancing technical steps: First, a systematic construction and standardized preprocessing of historical data on beryllium-aluminum alloys were performed. This invention extensively collected multi-source data from existing studies on the composition, preparation process, and corresponding corrosion resistance of beryllium-aluminum alloys. Specifically, the composition data included the mass percentage of beryllium (typically ranging from 55% to 70%), the aluminum content, and the concentration of any trace elements that may be present (such as nickel, iron, silicon, magnesium, etc.); the process data included the melting temperature (typically 700℃ to 750℃), holding time, solidification cooling rate (ranging from furnace cooling, air cooling to water quenching, corresponding to 5℃ / s to 50℃ / s), and subsequent heat treatment regimes (solution temperature, solution time, aging temperature, and aging time); the corrosion resistance data mainly included the corrosion current density (I_corr, unit: A / cm²) obtained through electrochemical testing. 2 The dataset includes the corrosion potential (E_corr, in V relative to the reference electrode) and the uniform corrosion rate (mm / year) calculated using the weight loss method. Data sources cover three channels: first, raw test data obtained through controlled experiments; second, peer-reviewed data systematically extracted from high-quality academic journals and conference papers both domestically and internationally; and third, historical R&D data accumulated within enterprises or research institutions. All data underwent rigorous quality review, outlier removal, and missing value imputation. Continuous variables were normalized or standardized to ultimately construct a structured, unified dataset suitable for subsequent modeling.

[0008] Based on the completed dataset, this invention sets specific target corrosion resistance performance indicators as constraints and optimization objectives for the inversion design. These target performance indicators are typically given as target or upper limit values ​​for corrosion current density and upper limits for corrosion rate. For example, the target could be set as "corrosion current density below 1.0 × 10⁻⁶". -7 A / cm 2 The target value can be set as "or 'the corrosion rate is reduced to less than 50% of that of the unoptimized material'". These target values ​​can be flexibly set according to specific engineering application scenarios. For example, marine engineering structures require higher corrosion resistance (lower corrosion current density), while the requirements in inland atmospheric environments can be appropriately relaxed. The performance target can be a specific numerical point or an acceptable numerical range. This target setting step is the "bullseye" of the entire inversion design, and all subsequent searches and optimizations revolve around how to achieve this target.

[0009] To achieve positive prediction of components, processes, and corrosion resistance, this invention constructs a high-precision positive prediction model. This model takes the concatenation of component feature vectors and process parameter feature vectors as input, and outputs corrosion resistance indicators (typically corrosion current density or corrosion rate). The component feature vectors include beryllium content, aluminum content, and the concentrations of various trace elements; the process parameter feature vectors include melting temperature, cooling rate, heat treatment temperature, and time. The positive prediction model can be implemented using various machine learning regression algorithms, including but not limited to: random forest regression (robust to nonlinear relationships, providing feature importance ranking), gradient boosting regression trees (high prediction accuracy and strong generalization ability), support vector regression (suitable for small sample sizes), or deep neural networks (suitable for large-scale data and complex mappings). The model is trained using supervised learning, taking component-process combinations from historical datasets as input and corresponding corrosion resistance performance as output labels, optimizing model parameters by minimizing prediction errors (such as mean squared error). After training, the positive prediction model can quickly and accurately predict the corrosion resistance of any given candidate component and process parameter combination, thus providing a "proxy" for performance evaluation for subsequent inversion optimization.

[0010] The core innovation of this invention lies in the construction of the inversion optimization process. Based on the forward prediction model, this invention establishes an inversion solution mechanism constrained by the target corrosion resistance performance. Unlike traditional exhaustive search or random sampling, this invention employs an intelligent optimization algorithm to efficiently search the parameter space to find the optimal combination of components and process parameters that meets the target performance requirements. Specifically, the inversion optimization process includes the following sub-steps: First, defining the search space—based on the fundamental principles of materials physicochemistry and engineering experience, setting reasonable value ranges for each design variable (such as beryllium content, cooling rate, etc.), for example, limiting the beryllium content to between 55% and 70%, and the cooling rate to between 5℃ / s and 50℃ / s; Second, initializing the candidate solution population—randomly generating a batch of initial candidate parameter combinations within the search space; Third, performance evaluation—for each candidate combination, calling the pre-trained forward prediction model to calculate its corresponding predicted corrosion resistance performance; Fourth, constraint judgment—comparing the predicted performance with the preset target performance indicators. The process involves several steps: First, if the candidate combination meets the objective (e.g., the predicted corrosion current density is below a set threshold), it is marked as a feasible solution. If not, the fitness function (e.g., the absolute value of the difference between the target value and the predicted value) is calculated based on the difference between the candidate and the objective. Second, candidate solutions are updated using heuristic optimization algorithms (e.g., selection, crossover, and mutation operations in genetic algorithms, or velocity and position updates in particle swarm optimization) to generate a new generation of candidate solutions, allowing the population to evolve towards a better solution (i.e., a predicted performance closer to the objective). Third, convergence is determined by repeating steps three through five until the preset maximum number of iterations is reached or the fitness function value is below the convergence threshold. Through this iterative evolution process, the inversion optimization algorithm can efficiently search for candidate parameter combinations that satisfy the objective performance constraints in a high-dimensional parameter space, avoiding the combinatorial explosion problem of exhaustive search.

[0011] To ensure that the design scheme output by the inversion optimization not only meets performance requirements but is also physically reasonable and engineering-manufacturable, this invention introduces two important types of constraints during the inversion process. The first type is composition range constraints, which limit the content of each alloying element to within its thermodynamically stable range to avoid generating unrealizable compositional combinations. For example, too low a beryllium content (<55%) will lead to insufficient material stiffness, while too high a content (>70%) will increase brittleness and make processing difficult. The addition of trace elements should also be controlled within the solid solubility limit to prevent the precipitation of harmful intermetallic compound phases. These constraints can be implemented by setting upper and lower bounds for variables in the optimization algorithm, or by imposing a penalty term in the fitness function—when a candidate solution's composition exceeds a reasonable range, a large penalty value is applied to its fitness, thus naturally eliminating it during the evolution process. The second type is process feasibility constraints, which ensure that the output process parameters (such as cooling rate and heat treatment temperature) can be achieved in actual production. For example, the cooling rate cannot exceed the equipment's capacity (e.g., the maximum cooling rate for water quenching is approximately 50℃ / s, and the minimum for furnace cooling is approximately 0.5℃ / s); the heat treatment temperature cannot exceed the alloy's overheating temperature; and the holding time should be within a reasonable industrial cycle (usually not exceeding 24 hours). The introduction of process feasibility constraints allows the final output design scheme to be directly used to guide actual production, significantly improving the engineering practicality of this invention. The above two types of constraints can be used as hard constraints (directly eliminating infeasible solutions during parameter sampling) or as soft constraints (guiding the search to avoid infeasible regions through a penalty function).

[0012] Finally, this invention outputs the optimal design scheme that meets the target corrosion resistance requirements. The output includes at least three parts: First, the optimal alloy composition ratio, given as the mass percentage of each element (e.g., recommended values ​​or ranges for beryllium content, aluminum content, and trace elements); second, the optimal combination of process parameters, including specific values ​​or windows for key parameters such as melting temperature, cooling rate, and heat treatment temperature and time; third, the predicted corrosion resistance values ​​(usually corrosion current density and corrosion rate) corresponding to this design scheme, to verify that it meets the target requirements. The output can be in the form of numerical tables, parameter cards, or a graphical interface, facilitating direct use by researchers for sample preparation and process verification. Furthermore, this invention can also output multiple suboptimal candidate schemes, allowing designers to flexibly choose based on other unmodeled engineering requirements (such as cost and processability).

[0013] Advantages of this invention: Compared with existing technologies, the optimization method for high corrosion-resistant beryllium aluminum alloys based on artificial intelligence inversion design proposed in this invention has the following significant and substantial beneficial effects: First, this invention is the first in the field to achieve end-to-end reverse design from target corrosion resistance performance to material composition and process parameters, fundamentally changing the traditional "trial and error-verification" R&D paradigm. Designers can directly input desired performance indicators, and the system automatically outputs the optimal material solution, greatly shortening the derivation path from performance targets to process solutions. Second, by combining intelligent optimization algorithms (such as genetic algorithms and particle swarm optimization) with high-precision forward prediction models, this invention can efficiently search for the global optimum in a high-dimensional, nonlinear parameter space, effectively avoiding the shortcomings of traditional methods that are prone to getting trapped in local optima, and the search efficiency is far higher than exhaustive methods or random sampling. Third, by introducing composition range constraints and process feasibility constraints, this invention ensures that the design scheme output by the inversion is physically reasonable and engineeringally manufacturable, overcoming the drawback of pure data-driven methods that are prone to outputting infeasible solutions, and significantly improving the engineering practicality of the design results. Fourth, this invention can fully utilize existing historical data (experimental data, literature data, database data), and through training a forward model, aggregate this scattered knowledge into a reusable predictive agent. Then, through inversion optimization, it achieves proactive application of knowledge, forming a closed loop of "data → model → design," exhibiting strong knowledge reuse and iteration capabilities. Fifth, experimental verification shows that the beryllium-aluminum alloy optimized using the method of this invention exhibits a reduction in corrosion current density of approximately 30% or more, a significant decrease in corrosion rate, and a substantial improvement in corrosion resistance. Simultaneously, this method significantly reduces the number of trial-and-error experiments in physical experiments, shortening the R&D cycle from several months using traditional methods to several days, and significantly reducing R&D costs. In summary, this invention has achieved breakthrough progress in both material inversion design theory and engineering optimization of high corrosion-resistant beryllium-aluminum alloys, demonstrating excellent prospects for promotion and application value. Attached Figure Description

[0014] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the overall process of the optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design proposed in this invention; Figure 2 This is a schematic diagram of the structure of the positive prediction model in this invention; Figure 3 This is a flowchart illustrating the inversion optimization process in this invention; Figure 4 This is a schematic diagram of the parameter search process in inversion optimization; Figure 5 A comparison chart showing the corrosion resistance performance before and after optimization. Detailed Implementation

[0015] The present invention will be further explained below with reference to specific implementation schemes, but it is not limited to the present invention. The structures, proportions, sizes, etc. shown in the accompanying drawings are only used to complement the content disclosed in the specification, so as to enable those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modification of the structure, change of the proportion relationship or adjustment of the size, without affecting the effect and purpose that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings. The embodiments described herein are for illustrative purposes only and do not constitute any limitation on the scope of protection of this invention.

[0017] Figure 1 It fully demonstrates the complete logical chain from historical data construction, target performance setting, positive prediction model training, inversion optimization search to constraint introduction and final solution output.

[0018] Figure 2 The model depicts the mapping relationship with component characteristics and process characteristics as inputs and corrosion resistance as output, and shows the network structure or decision path inside the model.

[0019] Figure 3 The closed-loop optimization mechanism, which includes candidate solution initialization, performance evaluation, constraint judgment, fitness calculation, update iteration, and convergence judgment, is presented in detail.

[0020] Figure 4 Taking a two-dimensional parameter space as an example, this paper demonstrates how a group of candidate solutions can iteratively approach the optimal parameter region corresponding to the target performance.

[0021] Figure 5 The significant differences in corrosion current density, corrosion rate, and other indicators between the beryllium aluminum alloy optimized using the method of this invention and the alloy before optimization (or the traditional method) are shown in the form of bar charts or box plots.

[0022] Example 1 This embodiment aims to improve the corrosion resistance of beryllium-aluminum alloys in a simulated marine environment (3.5% sodium chloride solution), and details the implementation process and verification results of the invention. The alloy system is a typical high-beryllium-content beryllium-aluminum alloy, with an initial composition ranging from 60% to 65% beryllium (mass percentage), and the balance being aluminum and trace impurities. The initial process conditions are: melting temperature 710℃, cooling rate 10℃ / s, and no subsequent heat treatment.

[0023] Step 1: Historical data construction and preprocessing.

[0024] This study constructed a historical dataset containing 420 valid samples. The data sources include: Part 1 (approximately 250 sets) consists of systematic experimental data from previous studies, covering corrosion current density test results under different combinations of beryllium contents (58%, 60%, 62%, 64%, 66%), cooling rates (5, 10, 15, 20, 30℃ / s), and heat treatment temperatures (400℃, 450℃, 500℃, holding time 2 hours); Part 2 (approximately 120 sets) comes from publicly published academic literature from the past fifteen years, extracting corrosion data of beryllium-aluminum alloys that meet the above composition and process ranges; Part 3 (approximately 50 sets) comes from the historical R&D database of collaborating institutions. The data underwent unified preprocessing: obvious outliers (such as extreme values ​​caused by pitting corrosion during the testing process) were removed, and missing individual process parameters were filled using the K-nearest neighbor interpolation method based on similar samples. Z-score standardization was applied to all continuous input features (Be content, cooling rate, and heat treatment temperature) to ensure that each feature has a mean of 0 and a standard deviation of 1. The target variable, corrosion current density, spans multiple orders of magnitude, underwent logarithmic transformation before standardization to make its distribution closer to a normal distribution. The final structured dataset includes the following fields: Be content (wt%), Al content (wt%), cooling rate (°C / s), heat treatment temperature (°C), heat treatment time (h), and corrosion current density (A / cm²). 2 ) and its logarithm.

[0025] Step 2: Setting the target corrosion resistance.

[0026] The target corrosion resistance performance set in this embodiment is: the corrosion current density is reduced to less than 50% of the initial level, and the absolute value does not exceed 1.5 × 10⁻⁶. -7 A / cm 2 The initial level (corresponding to a Be content of 62%, a cooling rate of 10℃ / s, and no heat treatment) showed a measured corrosion current density of approximately 3.2 × 10⁻⁶. -7 A / cm 2 Therefore, the target performance is specifically quantified as: corrosion current density ≤ 1.5 × 10⁻⁶. -7 A / cm 2 Meanwhile, to ensure the practicality of the design scheme, it is not required to be lower than a certain excessively low value; rather, this value will be used as an upper limit. This objective will serve as a constraint in the inversion optimization process.

[0027] Step 3: Building a positive prediction model.

[0028] This invention employs random forest regression as the positive prediction model. The input features are a four-dimensional vector: Be content (wt%), Al content (wt%), the logarithm of cooling rate (°C / s), and heat treatment temperature (°C) (if the data is as-cast without heat treatment, the heat treatment temperature is set to 25°C as the baseline). The target output is the logarithm of the corrosion current density. 420 sets of data are randomly divided into a training set (336 sets) and a test set (84 sets) in an 8:2 ratio. The random forest model is configured as follows: 200 decision trees, a maximum depth of 8, a minimum number of leaf samples of 5, and a feature subset size equal to all features. The model is trained on the training set, and five-fold cross-validation is used to optimize the hyperparameters. After training, the prediction performance on the test set is: Mean Absolute Percentage Error (MAPE) of 9.6%, and a coefficient of determination R0. 2 The value of 0.87 indicates that the positive prediction model can predict corrosion current density with high accuracy from composition and process parameters. Feature importance analysis shows that Be content and cooling rate are the two most important influencing factors, consistent with the corrosion physical mechanism.

[0029] Step 4: Inversion optimization process and constraint introduction.

[0030] This embodiment uses a genetic algorithm as the inversion optimization engine. Four design variables are used: Be content (ranging from 58% to 66%), Al content (automatically calculated by subtracting the Be content and a preset trace element total of 0.2% from 100%, therefore not considered an independent variable), cooling rate (ranging from 5℃ / s to 30℃ / s, logarithmic), and heat treatment temperature (ranging from 25℃ to 500℃, where 25℃ represents the as-cast state). The population size is set to 50, the maximum number of generations is 100, the crossover probability is 0.8, and the mutation probability is 0.1. The fitness function is defined as: if the corrosion current density predicted by the forward model is ≤ 1.5 × 10⁻⁶... -7 A / cm 2 Then fitness = 1 / (predicted value + 1 × 10) -9 (i.e., the smaller the predicted value, the higher the fitness); if the predicted value > 1.5 × 10⁻⁶ -7 A / cm 2 Then fitness = 1 / (|predicted value - 1.5 × 10⁻¹⁰) -7 | + 0.1×10 -7(i.e., penalizing the portion exceeding the target). Simultaneously, hard constraints are applied during initialization and mutation: if the Be content exceeds the range [58, 66], resampling is required; if the cooling rate exceeds the range [5, 30], resampling is required; if the heat treatment temperature exceeds the range [25, 500], resampling is required. For process feasibility constraints, this embodiment additionally requires: if the heat treatment temperature exceeds 450℃, the holding time cannot exceed 2 hours (this constraint is achieved through fitness penalty); the cooling rate cannot exceed 25℃ / s (to prevent excessive thermal stress leading to cracking). After 100 generations of evolution, the fitness of the optimal solution in the population tends to stabilize.

[0031] Step 5: Optimize the output results.

[0032] The optimal design scheme obtained from the inversion optimization is as follows: beryllium content is 61.2% (mass percentage), aluminum content is 38.6%, and the total amount of trace elements is controlled below 0.2%; the cooling rate is 22.5℃ / s; the heat treatment regime is: holding at 420℃ for 1.5 hours, followed by air cooling. The corrosion current density predicted by the forward prediction model for this scheme is 1.32×10⁻⁶. -7 A / cm 2 The target (≤1.5×10) is met. -7 A / cm 2 In comparison, the baseline scheme before optimization (Be 62%, cooling rate 10℃ / s, as-cast) predicted a corrosion current density of 3.21 × 10⁻⁶. -7 A / cm 2 The optimized scheme reduced the predicted corrosion current density by approximately 59%. Furthermore, the algorithm output three suboptimal candidate schemes, corresponding to different Be contents (61.5%, 60.8%, and 62.0%) and process combinations, for flexible selection during subsequent experimental verification.

[0033] Step 6: Experimental verification.

[0034] To verify the effectiveness of this invention, beryllium-aluminum alloy samples were prepared strictly according to the optimized output scheme (Be 61.2%, cooling rate 22.5℃ / s, heat treatment at 420℃ / 1.5h). Simultaneously, a control sample was prepared according to the baseline scheme before optimization. The corrosion current density of both samples was measured under the same electrochemical testing conditions (3.5% sodium chloride solution, room temperature, potentiodynamic polarization scan). Experimental results show that the corrosion current density of the optimized sample is 1.28 × 10⁻⁶. -7 A / cm 2 =1.32×10 -7 A / cm 2 The relative error was only 3.0%; the measured value of the reference sample was 3.35 × 10⁻⁶. -7 A / cm2 , compared with the predicted value of 3.21×10 -7 A / cm 2 The error is 4.2%. Compared to the baseline sample, the optimized sample actually reduced the corrosion current density by approximately 61.8% (from 3.35 × 10⁻⁶). -7 Reduced to 1.28×10 -7 A / cm 2 The corrosion rate decreased from approximately 4.0 μm / year to approximately 1.5 μm / year, resulting in a significant improvement in corrosion resistance. Simultaneously, the optimized process parameters remained within achievable industrial ranges (a cooling rate of 22.5℃ / s could be achieved through forced air cooling, and a heat treatment temperature of 420℃ / 1.5h was a conventional heat treatment regime), verifying the good manufacturability of the design scheme output by this invention. In contrast, using traditional empirical methods (orthogonal experimental optimization within a Be content range of 60-65% and a cooling rate of 5-25℃ / s) would require at least 16-25 samples and approximately two months to obtain a relatively optimal scheme (corrosion current density approximately 1.9 × 10⁻⁶). -7 A / cm 2 In contrast, after data construction, the total time for model training and inversion optimization of this invention is only about 3 hours, and the optimization effect is significantly better than that of traditional methods (corrosion current density reduced by about 32% vs. 61.8%). Experimental comparison fully demonstrates the dual superiority of this invention in terms of optimization efficiency and optimization effect.

[0035] In summary, this invention provides an optimization method for high corrosion-resistant beryllium aluminum alloys based on artificial intelligence inversion design. By constructing a forward prediction model and an intelligent inversion optimization algorithm, and introducing constraints on composition range and process feasibility, efficient and accurate inverse design from target corrosion resistance performance to optimal composition and process parameters is achieved. The described embodiments are merely one specific application form. Any modifications, equivalent substitutions, or improvements made to the above technical solutions within the spirit and principles of this invention should be included within the protection scope of the claims of this invention.

[0036] Matters not covered in this invention are common knowledge.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design, characterized in that, The method includes the following steps: First, historical data of beryllium aluminum alloy is obtained, including composition data, process data and corrosion resistance data. The data is then preprocessed to construct a unified dataset. The composition data includes at least the content ratio of each element, the process data includes at least the heat treatment temperature, time and cooling method, and the corrosion resistance data includes at least the corrosion current density and corrosion rate. Secondly, a target corrosion resistance performance index is set, which serves as a constraint condition for the inversion design; then, input features are constructed based on the data, which include at least a composition feature vector and a process parameter feature vector. Furthermore, a forward prediction model is constructed to describe the relationship between composition and process parameters and corrosion resistance. This model maps composition and process to corrosion resistance. Based on this, under the constraint of the target corrosion resistance, the composition and process parameters are solved through inverse search or iterative optimization to obtain candidate parameter combinations that meet the performance requirements. The inverse optimization process includes parameter space search, performance constraint judgment, and result screening and updating. At the same time, composition range constraints and process feasibility constraints are introduced in the inverse process. The composition range constraints are used to limit the content of each element within a reasonable range, and the process feasibility constraints are used to ensure that the heat treatment parameters are feasible in actual production. Finally, the optimal design scheme that meets the target corrosion resistance requirements is output. The scheme includes the alloy composition ratio, process parameter combination and corresponding corrosion resistance prediction value.

2. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1, characterized in that, The content of each element in the composition data includes the mass percentage of beryllium, the mass percentage of aluminum, and the concentration of trace added elements, which are selected from one or more of nickel, iron, silicon, and magnesium.

3. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1, characterized in that, The heat treatment parameters in the process data include solution temperature, solution time, aging temperature, and aging time, and the cooling methods include furnace cooling, air cooling, wind cooling, and water quenching.

4. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1, characterized in that, The positive prediction model is implemented using random forest regression, gradient boosting regression tree, support vector regression, or deep neural network. It is trained on the dataset through supervised learning to minimize the error between the predicted corrosion resistance performance and the actual corrosion resistance performance.

5. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1, characterized in that, The reverse search or iterative optimization method in the inversion optimization process is implemented using genetic algorithm, particle swarm optimization algorithm, Bayesian optimization or differential evolution algorithm. By iteratively updating the candidate solution population and evaluating its positive prediction performance, it gradually converges to the optimal parameter combination that satisfies the target performance constraint.

6. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1 or 5, characterized in that, The parameter space search in the inversion optimization process includes the following sub-steps: initializing a candidate solution population in the search space; calling the forward prediction model to calculate the predicted corrosion resistance performance of each candidate solution; comparing the predicted performance with the target performance index and calculating the fitness function based on the difference; generating a new generation of candidate solutions using a heuristic optimization algorithm; repeating the above steps until the convergence condition is met.

7. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1, characterized in that, The reasonable range within the composition range constraint is determined based on the phase stability and mechanical property requirements of beryllium aluminum alloy, wherein the beryllium content is limited to between 55% and 70% by mass, and the concentration of trace added elements is limited to below their solid solubility limit.

8. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1, characterized in that, The feasible conditions in the process feasibility constraints include: the cooling rate cannot exceed the equipment capacity range, the heat treatment temperature cannot exceed the overheating temperature of the alloy, and the holding time is within the preset industrial cycle range.

9. The optimization method for high corrosion-resistant beryllium aluminum alloy based on artificial intelligence inversion design according to claim 1, characterized in that, The optimal design scheme output also includes multiple candidate schemes for suboptimal output and their corresponding corrosion resistance performance prediction values, so that designers can flexibly choose according to additional engineering requirements.

Images