Non-isothermal heat treatment process for Fe-based amorphous alloy and alloy

Abstract

The invention relates to a non-isothermal heat treatment process of Fe-based amorphous alloy and the alloy, and the method comprises the following steps: establishing a database, and collecting data of a non-isothermal crystallization process of the Fe-based amorphous alloy; establishing a machine learning model; a series of optimal machine learning models are adopted to establish the relationship among the components, the process and the performance of the Fe-based amorphous alloy; according to the prediction and verification of the optimal non-isothermal heat treatment process, the optimal non-isothermal heat treatment process of the Fe85.5 B8.5 Si2P2C2 amorphous alloy and the classic FINEMET alloy is predicted; and preparing the high-performance Fe-based amorphous and nanocrystalline dual-phase alloy, and annealing based on the predicted optimal heat treatment process. The soft magnetic performance of the Fe-based amorphous nanocrystalline alloy can be effectively improved through variable-temperature heat treatment.

Description

Technical Field

[0001] This invention relates to the field of Fe-based amorphous alloy technology, and particularly to a non-isothermal heat treatment process and alloy for Fe-based amorphous alloys. Background Technology

[0002] Due to their low cost, high thermal stability, and good corrosion resistance, amorphous alloys, especially Fe-based amorphous alloys, have attracted widespread attention in electronics, aerospace, and other high-tech fields. However, amorphous alloys are metastable materials; when heated to a certain temperature, they become highly unstable and begin to transform into a stable crystalline state, leading to significant changes in structure and properties. Therefore, studying the crystallization process is crucial for understanding the thermal stability mechanism of amorphous materials and for developing novel nanocrystalline alloys. In recent years, researchers have extensively studied the crystallization behavior of various amorphous alloys, from Pd-Ag-Si to Cu-Zr-Ti. However, research on the crystallization behavior of Fe-based amorphous alloys, especially those with high Fe content, remains incomplete.

[0003] Typically, Fe-based amorphous nanocrystalline alloys exhibit better soft magnetic properties after appropriate crystallization processes compared to their amorphous precursors. However, current reports on the crystallization kinetics of Fe-based amorphous alloys primarily focus on their crystallization behavior, neglecting the exploration of the relationship between crystallization behavior and soft magnetic properties. Optimal universal crystallization heat treatment processes for Fe-based amorphous alloys remain an issue. The ambiguous relationship between crystallization behavior and soft magnetic properties may affect the accurate guidance for annealing Fe-based amorphous precursors.

[0004] Currently, Fe-based amorphous materials typically exhibit two crystallization methods: isothermal crystallization and non-isothermal crystallization. Both methods can alter grain growth, thereby regulating the properties of amorphous alloys. Isothermal crystallization is beneficial for improving the uniformity of the crystal structure. However, it usually requires a long time and is affected by various factors such as temperature, time, and processing environment, thus requiring precise control and adjustment, making the operation complex and demanding high-precision instruments. Non-isothermal crystallization, on the other hand, offers the advantages of efficient heat treatment and easy precise control, without requiring particularly complex instruments, making it more suitable for industrial production. However, non-isothermal crystallization often leads to structural inhomogeneity, discontinuous grain boundaries, or defects, which may adversely affect the properties of nanocrystalline materials.

[0005] Currently, nanocrystalline phases are mainly obtained through isothermal crystallization heat treatment. Previous empirical methods for selecting crystallization annealing temperatures are still insufficient for precisely controlling the precipitation ratio of nanocrystals. Therefore, precisely adjusting the soft magnetic properties of nanocrystalline alloys by controlling the nanocrystal precipitation ratio through heat treatment remains a challenge.

[0006] In recent years, machine learning has demonstrated advantages in materials science, including high efficiency, low cost, pattern discovery, and multi-scale modeling. Similarly, machine learning has been applied to the design of Fe-based nanocrystals. However, previous studies have primarily focused on using machine learning to search for the compositional influences on the soft magnetic properties of Fe-based nanocrystal alloys, neglecting the effects of annealing processes and nanocrystal structure on these properties. This may significantly limit the design of novel Fe-based nanocrystal alloys with the desired soft magnetic properties.

[0007] The information disclosed in the background section is only for enhancing the understanding of the background of this invention, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] This invention provides a non-isothermal heat treatment process and alloys for Fe-based amorphous alloys, combining machine learning and crystallization kinetics to predict their optimal annealing process. A novel, universal variable-temperature characteristic annealing process is introduced, applicable to Fe-based amorphous alloys, differing from traditional methods of post-isothermal crystallization for nanocrystalline alloys. Automating the optimal universal variable-temperature characteristic annealing process may provide new insights into the preparation of Fe-based nanocrystalline alloys with desired soft magnetic properties.

[0009] A non-isothermal heat treatment process for Fe-based amorphous alloys includes: Step 1: Establish a database and collect data on the non-isothermal crystallization process of Fe-based amorphous alloys. The database includes input features and output features. The input features include element content, heating rate, temperature, initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, and crystallization volume fraction. The output features include the initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, crystallization volume fraction, and Avrami index. The initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, and crystallization volume fraction are used as both input and output features. Step 2: Establish a machine learning model. The non-isothermal crystallization process data is divided into a training set and a test set. A limiting gradient boosting decision tree, nearest neighbor algorithm, and Gaussian process regression model are used to predict the initial crystallization temperature of the first crystallization peak, the characteristic temperature of the first crystallization peak, the crystallization volume fraction, and the Avrami index of the Fe-based amorphous alloy. Step 3: Predict and validate the optimal non-isothermal heat treatment process, Fe 85.5 B 8.5 Si2P2C2 amorphous alloy and Fe 73.5 Cu1Nb3Si 13.5 The non-isothermal crystallization process of B9 amorphous alloy was predicted based on a well-established machine learning model, and the non-isothermal crystallization process was calculated to verify the accuracy of the machine learning model prediction. Step 4: Prepare Fe-based amorphous alloy by melting Fe in a vacuum melting furnace. 85.5 B 8.5 Si2P2C2 master alloy, and Fe 73.5 Cu1Nb3Si 13.5 The B9 master alloy was used to prepare an alloy strip that was completely amorphous. Then, the amorphous alloy strip was heat-treated based on the optimal non-isothermal heat treatment process to obtain an Fe-based amorphous alloy.

[0010] In the non-isothermal heat treatment process of Fe-based amorphous alloy, in step S4, the obtained master alloy is broken into pieces, the master alloy is placed in anhydrous ethanol for ultrasonic cleaning and drying, placed in the bottom of a quartz glass tube and fixed in the center of a copper roller coil, and under the atmosphere of argon gas as a protective gas after vacuuming, the master alloy in the strip spinning machine is heated to a molten state, and then sprayed onto the high-speed rotating copper roller through a nozzle for cooling, thus preparing an alloy strip that is completely amorphous.

[0011] In the aforementioned non-isothermal heat treatment process for Fe-based amorphous alloys, the vacuum condition is a vacuum level below 5*10⁻⁶. -3 Pa, the melting process is repeated more than 5 times to ensure the uniformity of the alloy, and the strip spinning speed is 40m / s.

[0012] In the aforementioned non-isothermal heat treatment process for Fe-based amorphous alloys, the optimal heat treatment process predicted by machine learning is used, i.e., Fe... 85.5 B 8.5 The Si2P2C2 amorphous alloy was subjected to non-isothermal crystallization heat treatment in a vacuum environment, with the temperature being increased from room temperature to 413 °C at a rate of 5 K / min and then water quenched.

[0013] In the non-isothermal heat treatment process of Fe-based amorphous alloys, based on the set hyperparameters, the overall goodness of fit of the limit gradient boosting decision tree model predicting the initial crystallization temperature of the first crystallization peak is 0.98, the overall goodness of fit of the nearest neighbor algorithm predicting the characteristic temperature of the first crystallization peak is 0.94, the overall goodness of fit of the Gaussian process regression model predicting the crystallization volume fraction is 0.99, and the overall goodness of fit of the Gaussian process regression model predicting the Avrami exponent is 0.98.

[0014] In the non-isothermal heat treatment process for Fe-based amorphous alloys, the ratio of training set to test set is 4:1, and 10-fold cross-validation is introduced to ensure the predictive ability of the model.

[0015] An Fe-based amorphous alloy prepared by the method described above.

[0016] In the Fe-based amorphous nanocrystalline alloy described above, the Fe-based amorphous alloy is an amorphous nanocrystalline dual-phase alloy.

[0017] The Fe-based amorphous nanocrystalline alloy has a saturation magnetic induction intensity of 1.72-1.86 T.

[0018] In the aforementioned Fe-based amorphous nanocrystalline alloy, Fe 73.5 Cu1Nb3Si 13.5 The saturation magnetic induction of B9 nanocrystalline alloy is increased by at least 15%, while maintaining coercivity below 1 A / m.

[0019] Compared with existing technologies, this invention has the following advantages: This invention utilizes machine learning technology to establish a data model related to the crystallization kinetics of Fe-based amorphous alloys. Through database establishment and model training, the optimal non-isothermal heat treatment process for Fe-based amorphous alloys is predicted. This machine learning-based method can efficiently process and analyze large amounts of data, providing a more accurate and optimized heat treatment process. Combined with the theory of non-isothermal crystallization kinetics, experimental verification is used to validate the accuracy and practicality of the model. In practical operation, experiments have verified the feasibility of the optimal heat treatment process predicted based on machine learning, and have also demonstrated the effectiveness of this method for Fe-based amorphous alloys. 85.5 B 8.5 Si2P2C2 and Fe 73.5 Cu1Nb3Si 13.5 The soft magnetic properties of B9 (FINEMET) amorphous nanocrystalline alloys have been optimized. Specifically, the FINEMET nanocrystalline alloy obtained through predictive non-isothermal heat treatment exhibits a 15% increase in saturation magnetic induction while maintaining a coercivity below 1 A / m, making it superior to all similar alloys treated with conventional annealing. This invention differs from traditional trial-and-error material design methods, employing a novel variable-temperature heat treatment process based on crystallization kinetics and aided by high-throughput computation, which can accelerate the design of high-performance Fe-based amorphous nanocrystalline alloys. Attached Figure Description

[0020] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0021] In the attached diagram: Figure 1 This is a schematic diagram of a machine learning model for predicting crystallization behavior built according to an embodiment of the present invention. Figure 2Fe prepared in the embodiments of the present invention 85.5 B 8.5 Schematic diagram of the change trend of soft magnetic material of Si2P2C2 after non-isothermal heat treatment; Figure 3 The hysteresis loop diagram of the FINEMET alloy prepared in the embodiments of the present invention; Figure 4 This is a graph showing the predicted Avrmai index variation of FINEMET nanocrystalline alloys according to an embodiment of the present invention.

[0022] The present invention will be further explained below with reference to the accompanying drawings and embodiments. Detailed Implementation

[0023] Specific embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While specific embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0024] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that different terms may be used to refer to the same component. This specification and claims do not distinguish components based on differences in terminology, but rather on differences in function. The terms "comprising" or "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising but not limited to." The following descriptions are preferred embodiments for carrying out the invention; however, these descriptions are for the purpose of understanding the general principles of the specification and are not intended to limit the scope of the invention. The scope of protection of this invention is determined by the appended claims.

[0025] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. The accompanying drawings do not constitute a limitation on the embodiments of the present invention.

[0026] like Figures 1 to 4 As shown, the non-isothermal heat treatment process for Fe-based amorphous alloys includes the following steps: Step 1: Establish a database to collect data on the non-isothermal crystallization process of Fe-based amorphous alloys. This includes input and output features. The input features include element content, heating rate, temperature, initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, and crystallization volume fraction. The output features include the initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, crystallization volume fraction, and Avrami index. The initial crystallization temperature, characteristic temperature of the first crystallization peak, and crystallization volume fraction of the first crystallization peak are simultaneously used as both input and output features. Step 2: Build the machine learning model. First, begin with data collection, gathering multi-dimensional data on the non-isothermal crystallization process (such as temperature, time, and material composition). Second, perform data cleaning, removing duplicate data and feature columns with a missing rate exceeding 15%. During model training, the XGBoost model constructs 200 decision trees of depth 10. Each tree splits nodes by calculating gradient gain, with the prediction residual of the previous tree serving as the input to the next tree. Finally, the prediction results of all trees are weighted and summed with a learning rate of 0.2. k The NN model uses a BallTree data structure to store samples. During prediction, it calculates the Mahalanobis distance between the target point and its four nearest neighbors, and performs weighted voting with a weight of 1 / (d²+0.1). Gaussian process regression uses the RBF kernel function to construct the covariance matrix and solves for the posterior distribution through Cholesky decomposition. Measures to address non-convergence include: XGBoost gradually decays the learning rate by multiplying it by 0.9 every 20 rounds; if the validation loss does not decrease for 10 consecutive rounds, it triggers an early stopping mechanism and automatically adjusts the strength of the L2 regularization term; Gaussian process regression switches the Matern kernel function when the kernel matrix calculation is abnormal and maintains numerical stability by truncating eigenvalues ​​below 95%.

[0027] Step 3: Predict and validate the optimal non-isothermal heat treatment process, Fe 85.5 B 8.5 Si2P2C2 amorphous alloys and FINEMET amorphous alloys (specifically, Fe...) 73.5 Cu1Nb3Si 13.5 B9) Based on the established machine learning model, predict its non-isothermal crystallization process and calculate its non-isothermal crystallization process to verify the accuracy of the machine learning model prediction. We compare the difference between the actual calculated crystallization volume fraction and the crystallization volume fraction predicted by the machine learning. The smaller the difference, the more accurate the prediction result. Step 4: Prepare Fe-based amorphous alloy by melting Fe in a vacuum melting furnace. 85.5 B 8.5The Si2P2C2 master alloy and the FINEMET alloy were used to prepare an alloy strip that was completely amorphous. Then, the amorphous alloy strip was heat-treated based on the optimal non-isothermal heat treatment process to obtain an Fe-based amorphous alloy.

[0028] In the non-isothermal heat treatment process of Fe-based amorphous alloy, in step S4, the obtained master alloy is broken into pieces, the master alloy is placed in anhydrous ethanol for ultrasonic cleaning and drying, placed in the bottom of a quartz glass tube and fixed in the center of a copper roller coil, and under the atmosphere of argon gas as a protective gas after vacuuming, the master alloy in the strip spinning machine is heated to a molten state, and then sprayed onto the high-speed rotating copper roller through a nozzle for cooling, thus preparing an alloy strip that is completely amorphous.

[0029] The vacuum condition in the non-isothermal heat treatment process of the Fe-based amorphous alloy is a vacuum degree lower than 5*10. -3 Pa, the melting process is repeated more than 5 times to ensure the uniformity of the alloy, and the strip spinning speed is 40m / s.

[0030] The aforementioned Fe-based amorphous alloy utilizes machine learning to predict the optimal non-isothermal heat treatment process to process Fe... 85.5 B 8.5 The Si2P2C2 amorphous alloy was subjected to non-isothermal crystallization heat treatment in a vacuum environment, with the temperature increased from room temperature to 413 °C at a rate of 5 K / min and then water quenched.

[0031] In the non-isothermal heat treatment process of Fe-based amorphous alloys, the overall goodness of fit of the limiting gradient boosting decision tree model predicting the initial crystallization temperature of the first crystallization peak is 0.98; the overall goodness of fit of the nearest neighbor algorithm predicting the characteristic temperature of the first crystallization peak is 0.94; the overall goodness of fit of the Gaussian process regression model predicting the crystallization volume fraction is 0.99; and the overall goodness of fit of the Gaussian process regression model predicting the Avrami exponent is 0.98. These high goodness of fits indicate that the machine learning model established in this invention has good predictive ability.

[0032] In the non-isothermal heat treatment process of Fe-based amorphous alloys, the ratio of training set to test set is 4:1, and 10-fold cross-validation is introduced to ensure the predictive ability of the model.

[0033] An Fe-based amorphous alloy is prepared by the method described above.

[0034] In a preferred embodiment of the Fe-based amorphous alloy, the Fe-based amorphous alloy is an amorphous nanocrystalline dual-phase alloy.

[0035] In a preferred embodiment of the Fe-based amorphous alloy, the saturation magnetic induction intensity is 1.72-1.86 T.

[0036] In a preferred embodiment of the Fe-based amorphous alloy, the saturation magnetic induction intensity of the FINEMET nanocrystalline alloy is increased by at least 15%, while maintaining the coercivity below 1 A / m.

[0037] In one embodiment, machine learning-assisted crystallization kinetics can be used to propose a novel variable-temperature heat treatment process for Fe-based amorphous alloys. By establishing a database of the non-isothermal crystallization process of Fe-based amorphous alloys, machine learning can efficiently and accurately predict the optimal non-isothermal crystallization heat treatment process for other novel Fe-based amorphous alloys, thereby preparing Fe-based amorphous nanocrystalline alloys with desired soft magnetic properties. When designing material compositions, using machine learning based on isothermal crystallization kinetics can accelerate the development of high-performance soft magnetic materials. The crystallization process of amorphous alloys is diffusion-controlled. In the crystallization mechanism, the relationship between the crystallization volume fraction and the Avrami index shows that when the Avrami index is higher than 2.5, the nucleus growth rate increases over time; conversely, when the Avrami index is between 1.5 and 2.5, the nucleus growth rate decreases over time. We hypothesize that during crystal growth, rapidly growing nanocrystals contribute to the formation of ordered magnetic domain structures, thereby improving soft magnetic properties. However, as the growth rate of nanocrystals gradually decreases, it may lead to incomplete crystal structures and irregular magnetic domain arrangement, thus reducing soft magnetic properties. Therefore, crystallization parameters are crucial for determining the optimal annealing process.

[0038] Therefore, this invention employs machine learning to predict the crystallization behavior of Fe-based amorphous alloys, including the initial crystallization temperature of the first crystallization peak, the characteristic temperature of the first crystallization peak, the crystallization volume fraction, and the Avrami index. For non-isothermal crystallization processes, excessively slow heating rates affect the efficiency of heat treatment, while faster heating rates can disrupt the uniformity and stability of the crystallization nuclei. Therefore, we selected a heating rate of 5 K / min. Then, based on the machine learning prediction results, we infer the crystallization process of the Fe-based amorphous alloy. When the Avrami index corresponding to a certain crystallization volume fraction is close to 2.5, it is considered that the degree of crystallization is optimal for soft magnetic properties, thus automatically outputting the optimal non-isothermal heat treatment process.

[0039] In one embodiment, a novel variable-temperature heat treatment process for Fe-based amorphous materials based on crystallization kinetics, aided by high-throughput computation, is employed. This process uses elemental metals Fe, Si, and B, as well as alloys of compounds Fe3P and Fe3C, as raw materials. An electronic balance is used to determine the chemical composition of Fe... 85.5 B 8.5 Si2P2C2 and Fe 73.5 Si 13.5B9Cu1Nb3 raw material was weighed. After melting the raw material into a master alloy in an induction melting furnace, the resulting master alloy was broken into small pieces. A piece of master alloy of suitable size was ultrasonically cleaned and dried in anhydrous ethanol. This small piece of master alloy was then slowly placed into the bottom of a quartz glass tube and fixed in the center of a copper roller coil. Under an atmosphere of argon gas as a protective gas after vacuuming, the master alloy in the strip spinning machine was heated to a molten state. It was then rapidly cooled by blowing air through a nozzle onto a high-speed rotating copper roller, producing an amorphous alloy strip. Following this, a non-isothermal crystallization heat treatment process was performed to obtain an alloy strip with a nanocrystalline structure. The above-mentioned Fe... 85.5 B 8.5 The saturation magnetic induction of the Si2P2C2 nanocrystalline alloy is 1.82 T, and the magnetic induction of the FINEMET nanocrystalline alloy is 1.38 T.

[0040] Step 1) Database establishment: Extensive data collection on the non-isothermal crystallization process of Fe-based amorphous alloys. Input features include element content, heating rate, temperature, initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, and crystallization volume fraction. Output features include the initial crystallization temperature of the first crystallization peak, the characteristic temperature of the first crystallization peak, crystallization volume fraction, and Avrami index. It is important to note that the initial crystallization temperature of the first crystallization peak, the characteristic temperature of the first crystallization peak, and the crystallization volume fraction are used as both input features and output performance. Step 2) Establishing the machine learning model: Extreme gradient boosting decision tree, nearest neighbor algorithm, and Gaussian process regression model are used to predict the initial crystallization temperature, characteristic temperature of the first crystallization peak, crystallization volume fraction, and Avrami index of the Fe-based amorphous alloy. The ratio of training set to test set is 4:1, and 10-fold cross-validation is introduced to ensure the model's predictive ability.

[0041] Step 3) Prediction and verification of the optimal non-isothermal heat treatment process, selecting a high Fe content Fe... 85.5 B 8.5 The non-isothermal crystallization process of Si2P2C2 amorphous alloy and a classic FINEMET amorphous alloy was predicted based on a well-established machine learning model, and the non-isothermal crystallization process was calculated to verify the accuracy of the machine learning model prediction. Step 4) Preparation of iron-based amorphous nanocrystalline soft magnetic alloy: Fe was melted in a vacuum melting furnace. 85.5 B 8.5The Si2P2C2 and FINEMET master alloys were used. The resulting master alloy was broken into small pieces, and a piece of suitable size was ultrasonically cleaned in anhydrous ethanol, then dried and slowly placed into the bottom of a quartz glass tube. It was then fixed in the center of a copper roller coil. Under a vacuum and argon atmosphere as a protective gas, the master alloy in the spinning machine was heated to a molten state and rapidly cooled by blowing air through a nozzle onto a high-speed rotating copper roller, producing an alloy strip. Testing showed it to be completely amorphous. Next, based on the optimal non-isothermal heat treatment process proposed in step 3), the amorphous alloy strip was heat-treated to obtain an amorphous nanocrystalline dual-phase alloy.

[0042] The conditions for the non-isothermal crystallization heat treatment process are as follows: for Fe... 85.5 B 8.5 The Si2P2C2 amorphous alloy was subjected to non-isothermal crystallization heat treatment in a vacuum environment, with the temperature increased from room temperature to 413 °C at a rate of 5 K / min and then water quenched.

[0043] The conditions for the non-isothermal crystallization heat treatment process are as follows: for FINEMET amorphous alloys, non-isothermal crystallization heat treatment is performed in a vacuum environment, with the temperature increasing from 501 °C to 507 °C at a heating rate of 5 K / min, followed by water quenching.

[0044] To study its magnetic properties, the above Fe 85.5 B 8.5 Magnetic hysteresis loop tests were performed on Si2P2C2 amorphous nanocrystalline alloys. The results show that Fe exhibits a nanocrystalline structure after undergoing non-isothermal heat treatment. 85.5 B 8.5 The saturation magnetic induction intensity of the Si2P2C2 amorphous nanocrystalline alloy is 1.85T.

[0045] To study its magnetic properties, the above Fe 85.5 B 8.5 Magnetic hysteresis loop tests were performed on Si2P2C2 amorphous nanocrystalline alloys. The results show that Fe exhibits a nanocrystalline structure after undergoing non-isothermal heat treatment. 85.5 B 8.5 The saturation magnetic induction of the Si2P2C2 amorphous nanocrystalline alloy is 1.38T, and the coercivity is less than 1 A / m. Example 1

[0046] A Fe-based amorphous variable-temperature heat treatment process includes the following steps: Step 1) Database establishment: Extensive data collection on the non-isothermal crystallization process of Fe-based amorphous alloys. Input features include element content, heating rate, temperature, initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, and crystallization volume fraction. Output features include the initial crystallization temperature of the first crystallization peak, the characteristic temperature of the first crystallization peak, crystallization volume fraction, and Avrami index. It is important to note that the initial crystallization temperature of the first crystallization peak, the characteristic temperature of the first crystallization peak, and the crystallization volume fraction are used as both input features and output performance. Step 2) Establishing the machine learning model: Extreme gradient boosting decision tree, nearest neighbor algorithm, and Gaussian process regression model are used to predict the initial crystallization temperature, characteristic temperature of the first crystallization peak, crystallization volume fraction, and Avrami index of the Fe-based amorphous alloy. The ratio of training set to test set is 4:1, and 10-fold cross-validation is introduced to ensure the model's predictive ability.

[0047] Step 3) Prediction and verification of the optimal non-isothermal heat treatment process, selecting a high Fe content Fe... 85.5 B 8.5 The non-isothermal crystallization process of Si2P2C2 amorphous alloy was predicted based on a well-established machine learning model, and the non-isothermal crystallization process was calculated to verify the accuracy of the machine learning model prediction. Step 4) Preparation of iron-based amorphous nanocrystalline soft magnetic alloy: Fe was melted in a vacuum melting furnace. 85.5 B 8.5 The Si2P2C2 master alloy was broken into small pieces. A piece of master alloy of suitable size was ultrasonically cleaned in anhydrous ethanol, dried, and then slowly placed into the bottom of a quartz glass tube. It was then fixed in the center of a copper roller coil. Under a vacuum atmosphere with argon as a protective gas, the master alloy in the spinning machine was heated to a molten state and rapidly cooled by blowing air through a nozzle onto the high-speed rotating copper roller, thus preparing an alloy strip. Testing showed it to be completely amorphous. Next, based on the optimal non-isothermal heat treatment process proposed in step 3), the amorphous alloy strip was heat-treated to obtain an amorphous nanocrystalline dual-phase alloy.

[0048] To predict the crystallization behavior of Fe-based amorphous alloys, a limiting gradient boosting decision tree, a nearest neighbor algorithm, and a Gaussian process regression model were used to predict the initial crystallization temperature, characteristic temperature, crystallization volume fraction, and Avrami index of the first crystallization peak of Fe-based amorphous alloys. Figure 2 The diagram shows a scatter plot of the machine learning models established in this invention. All models have a goodness of fit exceeding 0.9, demonstrating their excellent predictive and generalization abilities.

[0049] To investigate the microstructure of the soft magnetic properties of the samples, the hysteresis loop of the iron-based nanocrystalline alloy thin strips prepared in the embodiments of this invention was measured using a vibrating sample magnetometer. Transmission electron microscopy and selected area electron diffraction analysis were then employed. The test results are as follows: Figure 3 As shown in the figure, the saturation magnetic induction of the sample in the amorphous state is 1.68 T; after heat treatment, the saturation magnetic induction is 1.72-1.86 T. Figure 4 The variation of the Avrmai index of FINEMET nanocrystalline alloys predicted by embodiments of the present invention is shown. Example 2

[0050] A Fe-based amorphous variable-temperature heat treatment process includes the following steps: Step 1) Prediction and verification of the optimal non-isothermal heat treatment process: A classic FINEMET amorphous alloy was selected, and its non-isothermal crystallization process was predicted based on the established machine learning model. The non-isothermal crystallization process was calculated to verify the accuracy of the machine learning model prediction. Step 2) Preparation of the iron-based amorphous nanocrystalline soft magnetic alloy: The FINEMET master alloy was melted in a vacuum melting furnace. The resulting master alloy was broken into small pieces. A piece of master alloy of suitable size was ultrasonically cleaned in anhydrous ethanol, then dried, and slowly placed into the bottom of a quartz glass tube. It was fixed in the center of a copper roller coil. Under an atmosphere of argon as a protective gas after vacuuming, the master alloy in the strip spinning machine was heated to a molten state. The molten alloy was then rapidly cooled by blowing air through a nozzle onto a high-speed rotating copper roller, producing an alloy strip. Testing showed it to be completely amorphous. Next, based on the optimal non-isothermal heat treatment process proposed in Step 1), the amorphous alloy strip was heat-treated to obtain an amorphous nanocrystalline dual-phase alloy.

[0051] To test the soft magnetic properties of the samples, the hysteresis loop of the iron-based nanocrystalline alloy thin strip prepared in the embodiments of this invention was measured using a vibrating sample magnetometer. The test results are as follows: Figure 3 As shown in the figure, the saturation magnetic induction intensity of the sample in the amorphous state is 1.2 T; after heat treatment, the saturation magnetic induction intensity can gradually increase to 1.38 T.

[0052] To compare the practicality of the non-isothermal heat treatment proposed in this invention, we compared the soft magnetic properties of the FINEMET nanocrystalline alloy obtained by non-isothermal heat treatment in this invention with other FIMEMET-like type nanocrystalline alloys. The results showed that the FINEMET nanocrystalline alloy obtained by non-isothermal heat treatment exhibited the highest saturation magnetic induction intensity and maintained a coercivity of less than 1 A / m.

[0053] In one embodiment, the method includes, 1) Database establishment: Extensive data collection on the non-isothermal crystallization process of Fe-based amorphous alloys; 2) Machine learning model establishment: A series of optimal machine learning models were used to establish the relationship between the composition, process, and properties of Fe-based amorphous alloys; 3) Prediction and verification of the optimal non-isothermal heat treatment process: Predicting the Fe... 85.5 B 8.5 The optimal non-isothermal heat treatment process for Si2P2C2 amorphous alloys and classic FINEMET alloys; 4) Preparation of high-performance Fe-based amorphous nanocrystalline dual-phase alloys, annealed based on the optimal heat treatment process predicted in step 2). The advantages of this invention are: based on non-isothermal crystallization kinetics, machine learning is used to efficiently predict the optimal variable-temperature crystallization heat treatment process for Fe-based amorphous alloys to prepare Fe-based amorphous nanocrystalline alloys. Results show that variable-temperature heat treatment can effectively improve the soft magnetic properties of Fe-based amorphous nanocrystalline alloys. Specifically, the FINEMET nanocrystalline alloy obtained through predictive non-isothermal heat treatment exhibits a 15% increase in saturation magnetic induction while maintaining a coercivity below 1 A / m, making its performance superior to all similar alloys treated with conventional annealing.

[0054] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A non-isothermal heat treatment process for Fe-based amorphous alloys, characterized in that, Includes the following steps: Step 1: Establish a database and collect data on the non-isothermal crystallization process of Fe-based amorphous alloys. The database includes input features and output features. The input features include element content, heating rate, temperature, initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, and crystallization volume fraction. The output features include the initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, crystallization volume fraction, and Avrami index. The initial crystallization temperature of the first crystallization peak, characteristic temperature of the first crystallization peak, and crystallization volume fraction are used as both input and output features. Step 2: Establish a machine learning model. The non-isothermal crystallization process data is divided into a training set and a test set. A limiting gradient boosting decision tree, nearest neighbor algorithm, and Gaussian process regression model are used to predict the initial crystallization temperature of the first crystallization peak, the characteristic temperature of the first crystallization peak, the crystallization volume fraction, and the Avrami index of the Fe-based amorphous alloy. Step 3: Predict and validate the optimal non-isothermal heat treatment process, Fe 85.5 B 8.5 Si2P2C2 amorphous alloy and Fe 73.5 Cu1Nb3Si 13.5 The non-isothermal crystallization process of B9 amorphous alloy was predicted based on a well-established machine learning model, and the non-isothermal crystallization process was calculated to verify the accuracy of the machine learning model prediction. Step 4: Prepare Fe-based amorphous alloy by melting Fe in a vacuum melting furnace. 85.5 B 8.5 Si2P2C2 master alloy, and Fe 73.5 Cu1Nb3Si 13.5 The B9 master alloy was used to prepare an alloy strip that was completely amorphous. Then, the amorphous alloy strip was heat-treated based on the optimal non-isothermal heat treatment process to obtain an Fe-based amorphous alloy.

2. The non-isothermal heat treatment process for Fe-based amorphous alloys according to claim 1, characterized in that, Preferably, in step S4, the obtained master alloy is broken into pieces, the master alloy is placed in anhydrous ethanol for ultrasonic cleaning and drying, placed in the bottom of a quartz glass tube and fixed in the center of a copper roller coil, and under the atmosphere of argon gas as a protective gas after vacuuming, the master alloy in the tape spinning machine is heated to a molten state, and then sprayed onto the high-speed rotating copper roller through a nozzle for cooling, thus preparing an alloy thin strip that is completely amorphous.

3. The non-isothermal heat treatment process for Fe-based amorphous alloys according to claim 2, characterized in that, The condition for vacuuming is a vacuum level below 5*10. -3 Pa, the melting process is repeated more than 5 times to ensure the uniformity of the alloy, and the strip spinning speed is 40m / s.

4. The non-isothermal heat treatment process for Fe-based amorphous alloys according to claim 1, characterized in that, The optimal heat treatment process based on machine learning prediction, namely Fe 85.5 B 8.5 The Si2P2C2 amorphous alloy was subjected to non-isothermal crystallization heat treatment in a vacuum environment, with the temperature increased from room temperature to 413 °C at a rate of 5 K / min and then water quenched.

5. The non-isothermal heat treatment process for Fe-based amorphous alloys according to claim 1, characterized in that, Based on the set hyperparameters, the overall goodness of fit of the limiting gradient boosting decision tree model for predicting the initial crystallization temperature of the first crystallization peak is 0.98, the overall goodness of fit of the nearest neighbor algorithm for predicting the characteristic temperature of the first crystallization peak is 0.94, the overall goodness of fit of the Gaussian process regression model for predicting the crystallization volume fraction is 0.99, and the overall goodness of fit of the Gaussian process regression model for predicting the Avrami exponent is 0.

98.

6. The non-isothermal heat treatment process for Fe-based amorphous alloys according to claim 1, characterized in that, The ratio of training set to test set is 4:1, and 10-fold cross-validation is introduced to ensure the predictive power of the model.

7. An Fe-based amorphous alloy, characterized in that, It is prepared by the method described in any one of claims 1-6.

8. The Fe-based amorphous nanocrystalline alloy according to claim 7, characterized in that, Fe-based amorphous alloys are amorphous and nanocrystalline dual-phase alloys.

9. The Fe-based amorphous nanocrystalline alloy according to claim 7, characterized in that, Its saturation magnetic induction intensity is 1.72-1.86 T.

10. The Fe-based amorphous alloy according to claim 7, characterized in that, Fe 73.5 Cu1Nb3Si 13.5 The saturation magnetic induction of B9 nanocrystalline alloy is increased by at least 15%, while maintaining coercivity below 1 A / m.

Images