High specific modulus glass fiber composition, glass fiber and composite material

Abstract

The present invention provides a high-specific modulus glass fiber composition, glass fiber, and composite material, relating to the field of glass fiber technology. The composition comprises the following components, expressed in molar percentage: SiO2: 57.5-64.5%, Al2O3: 10.0-15.0%, CaO: 6.0-10.0%, MgO: 18.0-23.0%, Li2O: 0-1.0%, K2O: 0-1.0%, Na2O: 0-1.0%, Fe2O3: 0-1.0%, TiO2: 0-1.5%, ZrO2: 0-1.0%, Y2O3: 0-0.5%, and La2O3: 0-0.5%. The quotient of the sum of the molar percentages of CaO and ZrO2 and the molar percentage of Y2O3 is not less than 17. The composition and proportions are determined by a prediction model and a multi-objective optimization model. The glass fiber composition provided by this solution has both a high specific modulus and a low upper crystallization temperature.

Description

Technical Field

[0001] The present invention relates to the technical field of glass fibers, in particular to a high specific modulus glass fiber composition, glass fiber and composite material. Background Art

[0002] As wind turbine blades evolve toward larger sizes, higher specific modulus, and lower costs, high-specific modulus glass fiber has become the primary choice for super-large blade reinforcement. The higher the modulus of the glass fiber used in wind turbine blades, the greater its resistance to deformation. In practical applications, in addition to high modulus, lightweight glass fiber is also essential. Specific modulus is the ratio of a material's modulus to its density. A higher specific modulus indicates a higher modulus and lower density, which reduces the manufacturing complexity of large wind turbine blades.

[0003] Chinese patent CN105731813A discloses a high-modulus glass fiber composition, glass fiber, and composite material thereof. The glass fiber achieves an elastic modulus exceeding 95 GPa. This improvement is primarily achieved by introducing rare earth oxides Y2O3 and La2O3, leveraging their synergistic effect. Furthermore, the content of these two rare earth oxides can reach up to 8%, resulting in high production costs for this glass fiber.

[0004] Chinese patent CN116282934A discloses a glass fiber composition and glass fiber with high magnesium and high specific modulus, whose tensile elastic modulus reaches above 94GPa and specific modulus is greater than 3.67×10 6 However, the formula contains B2O3, which is volatile and therefore not only pollutes the environment but also increases the difficulty of controlling the batching system. Furthermore, the glass fiber's upper crystallization temperature reaches 1260-1280°C, forcing the melting temperature to remain at a higher range. This not only increases energy consumption for production but also accelerates the erosion of the furnace's refractory materials.

[0005] Therefore, in view of the high difficulty and high cost of producing existing high specific modulus glass fibers due to their high crystallization upper limit temperature and high rare earth element content, there is an urgent need for a new type of high specific modulus glass fiber composition, glass fiber and composite material. Summary of the Invention

[0006] To solve the above problems, embodiments of the present invention provide a high specific modulus glass fiber composition, glass fiber, and composite material, so that the glass fiber composition has both high specific modulus and low upper crystallization temperature.

[0007] In a first aspect, an embodiment of the present invention provides a high specific modulus glass fiber composition, comprising the following components in mole percentage: SiO2: 57.5-64.5%, Al2O3: 10.0-15.0%, CaO: 6.0-10.0%, MgO: 18.0-23.0%, Li2O: 0-1.0%, K2O: 0-1.0%, Na2O: 0-1.0%, Fe2O3: 0-1.0%, TiO2: 0-1.5%, ZrO2: 0-1.0%, Y2O3: 0-0.5% and La2O3: 0-0.5%; the mole percentage of each component satisfies:

[0008] ( Mol CaO + Mol ZrO2 ) / Mol Y2O3 ≥17

[0009] in, Mol CaO 、 Mol ZrO2 、 Mol Y2O3 are the molar percentages of CaO, ZrO2, and Y2O3, respectively;

[0010] The components and proportions of the high-specific modulus glass fiber composition are obtained through a pre-trained prediction model and a multi-objective optimization model; the input of the prediction model is multi-components and proportions and physical parameters, and the output is performance parameters; the performance parameters include modulus, density, drawing forming temperature and upper crystallization temperature.

[0011] Optionally, the molar percentages of the components satisfy:

[0012] 0.15≤ Mol 1=( Mol CaO + Mol Li2O + Mol K2O + Mol Na2O + Mol Fe2O3 ) / Mol 2≤0.30

[0013] in, Mol CaO 、 Mol Li2O 、 Mol K2O 、 Mol Na2O 、 Mol Fe2O3are the molar percentages of CaO, Li2O, K2O, Na2O, and Fe2O3, respectively; Mol 2 is the sum of the molar percentages of Al2O3 and MgO.

[0014] Optionally, the molar percentages of the components satisfy:

[0015] 15≤ Mol CaO / Mol Y2O3

[0016] in, Mol CaO 、 Mol Y2O3 are the molar percentages of CaO and Y2O3 respectively.

[0017] Optionally, the glass fiber has a glass specific modulus of ≥3.73×10 6 m, the upper limit temperature of crystallization is 1230~1255℃.

[0018] Optionally, the components and proportions of the high specific modulus glass fiber composition are obtained by a pre-trained prediction model and a multi-objective optimization model, including:

[0019] Build molecular models consisting of varying proportions of no more than four components;

[0020] Simulating each of the molecular models to obtain the modulus and density of the molecular model;

[0021] Obtaining the wire drawing forming temperature and the upper crystallization temperature of each molecular model, and determining the physical parameters of each molecular model; the physical parameters include element characteristics, oxide properties and crystal structure parameters;

[0022] Expanding the molecular model into a multicomponent format of at least twelve components, and using the performance parameters of the molecular model and the physical parameters to train a random forest model to obtain a prediction model;

[0023] The performance parameters output by the prediction model are used to solve the constructed multi-objective optimization model to obtain high specific modulus components and proportions.

[0024] Optionally, simulating each of the molecular models to obtain the modulus and density of the molecular model includes:

[0025] Determining a simulation environment of the molecular model and a simulation force field including long-range and short-range relationships between atoms;

[0026] In the simulation environment and the simulation force field, heating and melting the molecular model and then cooling it to form an amorphous glass molecular model and obtain an equilibrium volume;

[0027] determining a density of the molecular model based on the equilibrium volume, the physical parameters, and a ratio of components in the molecular model;

[0028] The modulus is obtained by applying uniaxial tensile deformation under the amorphous glass molecular model.

[0029] Optionally, the molecular model is expanded into a multicomponent format of at least twelve components, and a random forest model is trained using the performance parameters of the molecular model and the physical parameters to obtain a prediction model, comprising:

[0030] Expanding the molecular model into a multi-component format; wherein the proportions of the eight components not included in the molecular model are 0;

[0031] The multivariate component format and the corresponding performance parameters and physical parameters are used as a training set, a random forest model is trained using the training set, and the model parameters are fine-tuned using virtual multivariate data to obtain the prediction model; wherein the virtual multivariate data is generated by molecular model perturbation.

[0032] Optionally, when using the training set to train the random forest model, it also includes: introducing physical constraint features; the physical constraint features include correcting the wire drawing temperature if the upper limit crystallization temperature predicted by the model is lower than the wire drawing temperature.

[0033] Optionally, the virtual multivariate data is generated by molecular model perturbation, comprising:

[0034] Randomly adding trace proportions of at least eight components not included in the molecular model, and renormalizing the proportions of no more than four components in the molecular model; and then generating virtual multi-components and proportions that meet eutectic point conditions based on thermodynamic phase diagram rules.

[0035] Optionally, the multi-objective optimization model includes an objective function and constraints:

[0036] The objective function is:

[0037]

[0038] The constraints include:

[0039]

[0040] in, f ( x ) is the input variable of the prediction model x The objective function of E ( x )for x The corresponding modulus; r( x )for x The corresponding density; T 析 is the upper limit of crystallization temperature; Mol Y2O3 is the molar percentage of Y2O3; Mol La2O3 is the molar percentage of La2O3; Mol i For the i The mole percentage of the components; M is the number of components of the multi-component.

[0041] Optionally, the performance parameters output by the prediction model are used to solve the constructed multi-objective optimization model to obtain high specific modulus components and proportions, including:

[0042] S1: Generate an initial glass fiber component decomposition set and use it as the current population, initialized to the first database of the empty set; set a multi-objective optimization model including objective function and constraint conditions;

[0043] S2: For all solutions in the current population and the first database, calculate corresponding performance parameters using a prediction model to evaluate the objective function value and the degree of violation of the constraint conditions of each solution;

[0044] S3: Based on the objective function value and the violation degree, a density estimate and a total fitness are calculated; based on the non-dominated sorting and crowding distance, all non-dominated solutions are screened from the current population and the first database and stored in a second database; if all non-dominated solutions exceed the capacity of the second database, the solutions with the highest density estimate or the highest violation degree are gradually removed; otherwise, the dominated solutions with the lowest total fitness are gradually stored;

[0045] S4: selecting parent individuals from the second database, performing crossover and mutation on the parent individuals, and generating a set of offspring glass fiber components that meet the constraint conditions;

[0046] S5: Determine whether the current number of iterations reaches a preset upper limit. If so, output the high specific modulus component and proportion. Otherwise, decompose the descendant glass fiber component into a population for the next iteration, assign the second database to the first database, and return to step S2.

[0047] In a second aspect, the present invention further provides a glass fiber made from any one of the high specific modulus glass fiber compositions described in the first aspect.

[0048] In a third aspect, the present invention further provides a composite material prepared using any glass fiber and polymer matrix described in the second aspect.

[0049] Preferably, the composite material is used in wind turbine blades.

[0050] Compared with the prior art, the present invention has at least the following beneficial effects:

[0051] (1) The glass fiber provided by the present invention has excellent mechanical properties, and its glass specific modulus is not less than 3.73×10 6 m; The glass fiber's boron-free formula avoids environmental pollution caused by the volatile release of B2O3, meeting environmental protection requirements. Furthermore, by controlling the molar percentage of rare earth elements within 1.0%, the glass fiber composition significantly reduces raw material costs while ensuring glass network stability. Furthermore, the glass fiber's upper crystallization temperature is no higher than 1255°C, improving the glass fiber molding process and reducing production complexity, making it more suitable for industrial continuous production.

[0052] (2) The method for determining the components of high specific modulus glass fibers provided by the present invention only needs to calculate the molecular model of glass fibers with a number of components not greater than 4, and then use the collected drawing forming temperature, crystallization upper limit temperature and physical parameters to train the random forest model to obtain the performance parameters of predictable multivariate (number of components ≥ 12) glass fibers, thereby significantly reducing the consumption of computing resources, shortening the R&D cycle and cost, and improving the efficiency of determining the target glass fiber composition. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 This is a flow chart of a method for determining components of a high specific modulus glass fiber provided by one embodiment of the present invention;

[0055] Figure 2 A molecular model of a ternary glass fiber component provided by one embodiment of the present invention;

[0056] Figure 3 This is another molecular model of ternary glass fiber components provided by one embodiment of the present invention;

[0057] Figure 4 is a tensile stress-strain curve diagram of a molecular model composed of a ternary glass fiber component provided by one embodiment of the present invention;

[0058] Figure 5It is a schematic diagram of a glass sample provided by one embodiment of the present invention undergoing a gradient crystallization temperature test. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0060] The following is the concept of the present invention. The present invention provides a high specific modulus glass fiber composition, comprising the following components in mole percentage: SiO2: 57.5-64.5% (for example, 57.5%, 58%, 58.5%, 59%, 59.5%, 60%, 60.5%, 61%, 61.5%, 62%, 62.5%, 63%, 63.5%, 64% or 64.5%), Al2O3: 10.0-15.0% (for example, 10.0%, 10.5%, 11%, 11.5%, 12%, 12.5%, 13%, 13.5%, 14%, 14.5% or 15.0%), CaO: 6.0-10.0% (for example, 6.0%, 6.5%, 7%, 7.5%, 8.0%, 8.5%, 9.0%, 9.5%). or 10%), MgO: 18.0-23.0% (for example, 18.0%, 18.5%, 19.0%, 19.5%, 20.0%, 20.5%, 21%, 21.5%, 22.0%, 22.5% or 23%), Li2O: 0-1.0%, K2O: 0-1.0%, Na2O: 0-1.0%, Fe2O3: 0-1.0%, TiO2: 0-1.5% (for example, 0%, 0.1%, 0.2%, 0.3%, 0.4%, 0.5%, 0.6%, 0.7%, 0.8%, 0.9%, 1.0%, 1.1%, 1.2%, 1.3%, 1.4% or 1.5%), ZrO2: 0-1.0%, Y2O3: 0-0.5% and La2O3: 0-0.5%; the molar percentage of each component satisfies:

[0061] ( Mol CaO + Mol ZrO2 ) / Mol Y2O3 ≥17

[0062] in, Mol CaO 、 Mol ZrO2 、 Mol Y2O3 are the mole percentages of CaO, ZrO2, and Y2O3, respectively. For example, the sum of the mole percentages of CaO and ZrO2 divided by the mole percentage of Y2O3 may be 20, 20.5, 21, 21.5, 22, 22.5, 23, 23.5, 24, 24.5, 25, or 25.5 or above;

[0063] The components and proportions of the high-specific modulus glass fiber composition are obtained through a pre-trained prediction model and a multi-objective optimization model; the input of the prediction model is multi-components and proportions and physical parameters, and the output is performance parameters; the performance parameters include modulus, density, drawing forming temperature and upper crystallization temperature.

[0064] For 0-1.0%, it can be any value between 0% and 1.0%, for example, it can be 0%, 0.05%, 0.1%, 0.2%, 0.3%, 0.4%, 0.5%, 0.6%, 0.7%, 0.8%, 0.9% or 1.0%;

[0065] For 0-0.5%, it can be any value between 0% and 0.5%, for example, it can be 0%, 0.05%, 0.1%, 0.15%, 0.2%, 0.25%, 0.3%, 0.35%, 0.4%, 0.45% or 0.5%.

[0066] The glass fiber provided by the present invention has excellent mechanical properties, and its glass specific modulus is not less than 3.73×10 6 m; The glass fiber's boron-free formula avoids environmental pollution caused by the volatile release of B2O3, meeting environmental protection requirements. Furthermore, by controlling the molar percentage of rare earth elements within 1.0%, the glass fiber composition significantly reduces raw material costs while ensuring glass network stability. Furthermore, the glass fiber's upper crystallization temperature is no higher than 1255°C, improving the glass fiber molding process and reducing production complexity, making it more suitable for industrial continuous production.

[0067] In the present invention, SiO2 is the primary component of the network former in the glass structure. This oxide possesses high chemical bond strength, effectively improving the mechanical properties, chemical stability, and thermodynamic stability of the glass material. However, excessive SiO2 concentrations increase the viscosity of the glass material, making subsequent glass fiber formation more difficult and hindering large-scale production. Therefore, the present invention limits the SiO2 molar percentage range to 57.5-64.5%.

[0068] The main system of high specific modulus glass is generally SiO2-Al2O3-MgO, which has the characteristics of high modulus and low density. Among them, Al2O3 and MgO can both enter the glass network structure, supplement the network skeleton, making it denser and thus improving the modulus and chemical stability of the glass. However, the dense network structure of this system also significantly enhances the crystallization ability of the glass and significantly accelerates the crystallization rate. This leads to a certain degree of inversion between the modulus of the glass in this system and its glass-forming ability, that is, the glass-forming ability of high modulus glass is relatively poor, and the crystallization temperature and wire drawing forming temperature are relatively high; while the specific modulus of glass in the system with lower crystallization temperature and wire drawing forming temperature usually cannot meet industrial needs. Therefore, under the premise of ensuring relatively low crystallization temperature and wire drawing forming temperature, designing high specific modulus glass fiber has always been a hot spot and difficulty in the relevant field. In the present invention, the molar percentage range of Al2O3 is limited to 10.0~15%, and the molar percentage range of MgO is limited to 18.0~23.0%.

[0069] In response to the above-mentioned problem of difficulty in designing high specific modulus glass fiber, the present invention improves the modulus and molding ability of glass fiber at the same time by adding CaO. Adding CaO to the SiO2-Al2O3-MgO system will lead to a significant mixed alkaline earth effect. This is because when two or more alkaline earth metal oxides (such as CaO and MgO) act together, their performance in the glass is more complementary or mutually reinforcing than when they exist alone. The elastic modulus of glass mainly depends on the rigidity and bonding strength of the glass network structure. CaO mainly acts as a network modifier in glass. They can replace part of the Si-O bonds in the glass network and form new Ca-O. Ca-O bonds are stronger than typical network modifiers such as Na-O and KO, which increases the local rigidity of the glass network. At the same time, due to Ca 2+ Generally, it has a high coordination number (usually ≥6), which will form a highly cross-linked structure and also increase the rigidity of the network structure. 2+ Compared with Na + or K + Alkali metal ions are small in size and high in charge, which can more effectively fill the gaps in the glass network and increase the density of the glass network and the average atomic distance. This decrease in density and atomic distance will also lead to an increase in the glass modulus. On the other hand, Ca 2+ Relative Si 4+ 、Al 3+ and Mg 2+ A larger ionic radius will break part of the original glass network structure, resulting in an increase in the number of non-bridging oxygen (NBO). This disturbance of the basic silicon-oxygen tetrahedral structure of the glass will reduce the continuity of the glass network, making the glass network easier to flow. In addition, Ca 2+The distribution of CaO within the glass network reduces the viscosity of the system, lowering the internal friction of the material, allowing the glass to be formed at relatively low temperatures and reducing the difficulty of glass fiber drawing. Therefore, the present invention limits the CaO content to a molar percentage range of 6.0 to 10.0%.

[0070] In a preferred embodiment, the molar percentages of the components satisfy:

[0071] 0.15≤ Mol 1=( Mol CaO + Mol Li2O + Mol K2O + Mol Na2O + Mol Fe2O3 ) / Mol 2≤0.30

[0072] in, Mol CaO 、 Mol Li2O 、 Mol K2O 、 Mol Na2O 、 Mol Fe2O3 are the molar percentages of CaO, Li2O, K2O, Na2O, and Fe2O3, respectively; Mol 2 is the sum of the molar percentages of Al2O3 and MgO. For example, Mol 1 can be 0.15, 0.16, 0.17, 0.18, 0.20, 0.22, 0.25, 0.26, 0.28 or 0.30.

[0073] In the embodiments of the present invention, the oxide combination in the above formula helps reduce the glass viscosity and crystallization tendency, thereby reducing the difficulty of subsequent glass fiber forming. However, excessive content will significantly destroy the glass network structure and reduce the mechanical properties of the glass fiber.

[0074] In a preferred embodiment, the molar percentages of the components satisfy:

[0075] 15≤ Mol 3= Mol CaO / Mol Y2O3

[0076] in, Mol CaO 、 Mol Y2O3 are the molar percentages of CaO and Y2O3 respectively.

[0077] In the present invention, the addition of oxide Y2O3 can increase the glass fiber drawing window temperature, but this will also lead to an increase in the upper limit of crystallization temperature and the drawing forming temperature; while CaO can effectively reduce the melting point and high temperature viscosity of glass fiber, reduce the upper limit of crystallization temperature, and reduce the energy consumption required for production. It has been confirmed by experiments that when 0.15≤ Mol 1≤0.30, and 15≤ Mol 3, it helps to control the upper limit of crystallization temperature of glass fiber to no more than 1255℃ while ensuring that the glass fiber has a good drawing window temperature.

[0078] In a preferred embodiment, the glass modulus of the glass fiber is ≥3.73×10 6 m, the upper limit temperature of crystallization is 1230~1255℃.

[0079] In a preferred embodiment, the components and proportions of the high specific modulus glass fiber composition are obtained by a pre-trained prediction model and a multi-objective optimization model. Specifically, the high specific modulus glass fiber component determination method is as follows: Figure 1 Shown, including:

[0080] Step 100, establishing a molecular model composed of different proportions of no more than four components;

[0081] Step 102, simulating each molecular model to obtain the modulus and density of the molecular model;

[0082] Step 104, obtaining the wire drawing forming temperature and the upper crystallization temperature of each molecular model, and determining the physical parameters of each molecular model; the physical parameters include element characteristics, oxide properties and crystal structure parameters;

[0083] Step 106: Expand the molecular model into a multicomponent format of at least twelve components, and train a random forest model using the performance parameters and physical parameters of the molecular model to obtain a prediction model; the prediction model inputs the multicomponents and their proportions and physical parameters, and outputs the performance parameters; the performance parameters include modulus, density, wire drawing temperature, and upper crystallization temperature;

[0084] Step 108 : Solve the constructed multi-objective optimization model using the performance parameters output by the prediction model to obtain high specific modulus components and proportions.

[0085] In an embodiment of the present invention, only a molecular model of glass fibers with no more than four components needs to be calculated. The collected drawing temperature, crystallization upper limit temperature, and physical parameters are then used to train a random forest model to obtain predictable performance parameters for multivariate (number of components ≥ 12) glass fibers. Finally, a multi-objective optimization model is used to screen and determine the high-specific modulus glass fiber components and proportions from the predicted performance parameters. This significantly reduces the consumption of computing resources, shortens the R&D cycle and costs, and improves the efficiency of determining the high-specific modulus glass fiber composition.

[0086] Described below Figure 1 How to perform the steps shown.

[0087] For step 100, the components include SiO2, Al2O3, CaO, MgO, Li2O, K2O, Na2O, Fe2O3, TiO2, ZrO2, Y2O3, and La2O3, wherein the variation range of each component is greater than 15% (in terms of molar percentage) (for example, 15.5%, 16%, 18%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, etc.). The molecular model includes components composed of the same component in different proportions and components composed of different components in any proportion. For example, Figure 2 A molecular model consisting of 20% SiO2, 55% Al2O3, and 25% Y2O3 by mole percentage is shown. Figure 3 A molecular model consisting of 73.5% SiO2, 14.1% Na2O, and 12.4% ZrO2 in molar percentages is shown.

[0088] Specifically, for example, a range of compositional variations for different components is designed, different oxides (i.e., components) are placed at random locations in the simulation box, and the file is converted using VMD molecular simulation visualization software into a data file that can be input into open-source molecular dynamics software for simulation in step 102. For example, in step 100, no more than 1,000 molecular models composed of different proportions of the quaternary components are established.

[0089] In step 102, each molecular model is simulated to obtain the modulus and density of the molecular model, including:

[0090] Determining a simulation environment of the molecular model and a simulation force field including long-range and short-range relationships between atoms;

[0091] In a simulated environment and simulated force field, the molecular model is heated and melted and then cooled to form an amorphous glass molecular model and obtain an equilibrium volume;

[0092] Determine the density of the molecular model based on the equilibrium volume, physical parameters and the ratio of components in the molecular model;

[0093] The modulus is obtained by applying uniaxial tensile deformation under the amorphous glass molecular model.

[0094] Specifically, the simulation environment includes the unit, spatial dimension, boundary conditions, atom type, simulation time step, and atomic neighbor parameters of the simulation system; the simulation force field uses damped shifted force to simulate the long-range relationship between atoms, and uses Buckingham potential function to simulate the short-range relationship between atoms; then the system is heated to 3000K to fully melt the molecular model, and then cooled to 300K to form an amorphous glass molecular model. The NPT ensemble is used in the simulation process; then the size of the entire system box is counted to calculate the density of the molecular model; finally, by applying uniaxial tensile deformation, the stress-strain curve of the molecular model is obtained (for example, Figure 4 Shown Figure 2 The modulus of the molecular model (i.e., Young's modulus) was calculated based on the stress-strain curve in the strain range of 0% to 0.5%.

[0095] Specifically, the density of the molecular model is determined by the following formula:

[0096]

[0097] in, r is the density; Mol i For the i The mole percentage of the components; M 0 is the number of elements of the multicomponents included in the molecular model; m i For the i The relative molecular mass of the components; V The volume is obtained by converting the equilibrium volume into units, in cm 3 ; N A is Avogadro's constant.

[0098] In step 104 , the wire drawing temperature and the upper crystallization temperature limit of each molecular model are obtained from the process parameters, not from simulation.

[0099] In step 106, the molecular model is expanded into a multicomponent format of at least twelve components, and the performance parameters and physical parameters of the molecular model are used to train a random forest model to obtain a prediction model, including:

[0100] Expand the molecular model to a multi-component format; the proportions of the eight components not included in the molecular model are 0;

[0101] The multivariate component format and the corresponding performance parameters and physical parameters are used as the training set. The random forest model is trained using the training set, and the model parameters are fine-tuned using virtual multivariate data to obtain a prediction model; wherein, the virtual multivariate data is generated by molecular model perturbation.

[0102] Specifically, when training a random forest model, it is necessary to determine the number of components to be predicted. For example, a molecular model can be expanded to a twelve-component format and then trained on a random forest model. The resulting prediction model takes twelve components, their proportions, and physical parameters as input, and outputs performance parameters. It should be noted that this prediction model has been validated using a validation set.

[0103] In the present invention, by introducing physical parameters including element properties (such as the atomic weight, electronegativity, ionic radius, oxidation state, etc. of each element), oxide properties (such as the melting point, thermal expansion coefficient, bond energy, etc. of the oxide) and crystal structure parameters, the training process can fully extract the real physical parameter characteristics and learn the synergistic effects between components, which not only enhances the physical interpretability of the final prediction model, but also improves the generalization ability and prediction accuracy of the prediction model.

[0104] In a preferred embodiment, when training the random forest model using the training set, it also includes: introducing physical constraint features; the physical constraint features include correcting the wire drawing temperature if the upper limit crystallization temperature predicted by the model is lower than the wire drawing temperature.

[0105] In the present invention, physical constraint features are further introduced to improve the prediction accuracy and reliability of the prediction model.

[0106] In a preferred embodiment, virtual multivariate data are generated by molecular model perturbation, comprising:

[0107] At least eight components not included in the molecular model are randomly added in trace proportions, and the proportions of no more than four components in the molecular model are renormalized; then, virtual multi-component components and proportions that meet the eutectic point conditions are generated based on the thermodynamic phase diagram rules.

[0108] Specifically, the trace ratio is 0 to 5% (for example, 0%, 1%, 2%, 3%, 4% or 5%).

[0109] In step 108, the multi-objective optimization model includes the objective function and constraints:

[0110] The objective function is:

[0111]

[0112] Constraints include:

[0113]

[0114] in, f ( x ) is the input variable of the prediction model x The objective function of E ( x )for x The corresponding modulus; r ( x )for x The corresponding density; T 析 is the upper limit of crystallization temperature; Mol Y2O3 is the molar percentage of Y2O3; Mol La2O3 is the molar percentage of La2O3; Mol i For the i The mole percentage of the components.

[0115] Specifically, the value of the objective function is the specific modulus. While satisfying the constraints, maximizing the specific modulus can solve the components and proportions with high specific modulus. At this time, the glass fiber component solved can simultaneously have a high specific modulus, a lower upper crystallization temperature, and a lower rare earth element content.

[0116] For step 108, the performance parameters output by the prediction model are used to solve the constructed multi-objective optimization model to obtain high specific modulus components and proportions, including:

[0117] S1: Generate the initial glass fiber component decomposition and use it as the current population P 0, initialized to the first database of the empty set, set the objective function f ( x ) and the above constraints;

[0118] Among them, the initial population P 0={ x 1, x 2,…, x N}, x k Indicates the k Multi-tuple decomposition; N Indicates the population size. x k =( c k1 , c k2 ,…, c kM ), c ki Indicates thek The first solution i The percentage of the components; M Indicates the total number of possible glass fiber composition types, M ≥12.

[0119] S2: For all solutions in the current population and the first database, the prediction model is used to calculate the performance parameters corresponding to each solution, and the objective function value and the degree of violation of the constraint conditions of each solution are evaluated accordingly;

[0120] The degree of violation is determined by the following formula:

[0121]

[0122] in, CV ( x ) represents the solution x The degree of violation of the constraint, g y ( x ) indicates the y The amount of constraint violations, Y represents the total number of constraints, Y =4.

[0123] S3: Based on the objective function value and the degree of violation, the density estimate and total fitness of each solution are calculated; according to the non-dominated sorting and crowding distance, all non-dominated solutions are screened from the current population and the first database and stored in the second database; if the number of all non-dominated solutions exceeds the capacity of the second database, the solutions with the highest density estimate or the highest degree of violation are preferentially removed; if not, some dominated solutions with the lowest total fitness are appropriately stored to maintain population diversity;

[0124] The dominance relationship is defined as x 1 Dominant Solution x 2 (denoted as x 1≺ x 2) If and only if f ( x 1)≥ f ( x 2)∧CV( x 1)≤CV( x 2);

[0125] The density estimate is determined by the following formula:

[0126]

[0127] in D k Represents the solution x k The density estimate of Sp represents the current population, d ( x k , x j ) represents the solution x k and x j The crowding distance between e is a small positive number to prevent division by zero errors;

[0128] The calculation formula for congestion distance is:

[0129]

[0130] in d k Represents the solution x k The crowding distance; f max 、 f min Represent the maximum and minimum objective function values ​​among all solutions respectively;

[0131] The total fitness calculation formula is:

[0132]

[0133] in F ( x k ) represents the solution x k The total fitness of rank ( x k ) represents the solution x k non-dominant class of; l Represents the weight coefficient of the density estimate; m Indicates the penalty coefficient for constraint violation.

[0134] S4: Select parent individuals from the second database and use crossover and mutation operators to generate offspring glass fiber component decomposition sets that meet the constraints. Repair operators or constraint feasibility tests can be used to ensure that the offspring solutions meet the basic constraints.

[0135] The crossover operation is:

[0136]

[0137] in C child,i Indicates the first iThe content of the component (i.e., glass fiber component); C parent1,i and C parent2,i Represents the first i The content of the ingredients; α ∈[0,1] is a randomly generated crossover coefficient;

[0138] The mutation operation is:

[0139]

[0140] in C new,i Indicates the mutation i The content of the ingredients; C old,i Indicates the first i The content of the ingredients; s Indicates variable asynchronous length; N (0,1) represents a standard normal distribution random number;

[0141] The repair operations are:

[0142]

[0143] in, C repaired,i Indicates that after repair i The content of the ingredients; C raw,i Indicates the first i The content of the ingredients.

[0144] S5: Determine whether the current number of iterations reaches the preset upper limit. If so, output the high specific modulus component and proportion. If not, decompose the descendant glass fiber component into a new current population, assign the second database to the first database, and return to step S2.

[0145] Specifically, a multi-objective optimization algorithm is used to optimize and calculate the modulus, density, drawing forming temperature, and upper crystallization temperature of the multi-component output of the prediction model based on the set objective function, and multiple glass fiber components that meet the constraints are obtained, that is, glass fiber composition solutions that meet the constraints. Then, mathematical statistics are performed on the multiple groups of glass fiber composition solutions to generate the optimal content range of each component of high modulus glass fiber.

[0146] The smaller the total fitness in step S2, the better the solution.

[0147] In step S3, it should be noted that when the capacity of the second database is exceeded, the solutions with the highest density estimate or the highest violation are gradually removed until the capacity of the second database is reached (i.e., the second database is filled). If the capacity of the second database is not reached, the dominated solutions with the highest total fitness values ​​are sorted from low to high, and the top-ranked dominated solutions are preferentially stored in the second database until the capacity of the second database is reached.

[0148] In step S5, when the preset number of iterations is reached, the solution set included in the second database of the current iteration is output, namely, the high specific modulus components and proportions.

[0149] In the present invention, by combining the above-mentioned solution method with the governing strength and density calculation fitness, it is possible to balance convergence and diversity, thereby efficiently approximating the optimal solution set of the multi-objective optimization model through iteration, quickly determining the high specific modulus glass fiber components and proportions that meet the multi-objective optimization model, shortening the research and development cycle and cost of the glass fiber components, and improving research and development efficiency.

[0150] The present invention also provides a glass fiber made from any of the above-mentioned high specific modulus glass fiber compositions.

[0151] The present invention also provides a composite material prepared by using the glass fiber and polymer matrix.

[0152] In order to more clearly illustrate the technical solutions and advantages of the present invention, a high specific modulus glass fiber composition is described in detail below through several embodiments.

[0153] In the following Examples 1 to 14 and Comparative Examples 1 to 8, the required amounts of various raw materials were calculated according to the composition of the high specific modulus glass fiber component in each Example or Comparative Example, accurately weighed and uniformly mixed, and then placed in a platinum crucible for melting glass at 1550° C. The glass was melted and formed and processed to obtain the glass samples corresponding to each Example or Comparative Example.

[0154] Among them, Comparative Example 8 is Example A16 in Chinese patent CN111807707A, including the following components in weight percentage: SiO2: 52.0%, Al2O3: 18.9%, CaO: 1.0%, MgO: 10.7%, Y2O3: 16.0%, Na2O: 0.3%, K2O: 0.2%, Li2O: 0%, Fe2O3: 0.4%, TiO2: 0.4%, SrO: 0%, La2O3: 0%, CeO2: 0%. Converting this weight percentage to molar percentage yields: SiO2: 60.9%, Al2O3: 13.02%, CaO: 1.26%, MgO: 18.8%, Y2O3: 4.98%, Na2O: 0.34%, K2O: 0.15%, Li2O: 0%, Fe2O3: 0.18%, TiO2: 0.35%, SrO: 0%, La2O3: 0%, CeO2: 0%. Obviously, weight percentages and molar percentages are not completely comparable, with the weight percentages of Al2O3 and Y2O3 differing significantly from their molar percentages. The Young's modulus of this comparative example is 106.3 GPa, and the upper crystallization temperature is 1255°C. Mol 1 is 0.061; Mol 3 is 0.25.

[0155] The high-specific modulus glass fiber components and ratios, as well as the properties of the glass samples, in Examples 1-14 are shown in Tables 1 and 2. The component formulations and properties of the glass samples in Comparative Examples 1-8 are shown in Table 3. In Tables 1-3, density determination was performed on the glass bulk samples using the Archimedean method in accordance with GB / T 5432. The density of the glass bulk sample was calculated by first weighing the glass bulk sample in air to obtain a mass, then weighing it in an auxiliary liquid of known density (distilled water) to obtain another mass.

[0156] Determination of the upper limit of crystallization temperature: The temperature gradient is used according to ASTM C829. First, the broken glass sample is evenly spread in a platinum boat, and then the platinum boat is placed in a gradient furnace. It is kept at a certain temperature for 2 hours and then cooled to room temperature. The upper limit of crystallization temperature of the glass can be calculated based on the crystallization position of the glass sample in the platinum boat and the temperature function of the gradient furnace during the holding stage. Figure 5 A schematic diagram showing a glass sample undergoing a gradient crystallization temperature test.

[0157] Young's modulus: Glass bulk samples were tested using the ultrasonic echo method according to GB / T 38897. By measuring the longitudinal and transverse propagation velocities of elastic waves in solid samples, and based on the theory of elastic wave propagation in solids, the propagation speed of different modes of sound waves in a solid is related to its elastic modulus and density, the Young's modulus of the glass bulk sample can be calculated. Specific modulus = Young's modulus / density / 9.8.

[0158] Table 1

[0159]

[0160] Table 2

[0161]

[0162] Table 3

[0163]

[0164] As shown in Tables 1 and 2, the glass samples prepared by the present invention have extremely low rare earth element content and do not contain boron, but have both high specific modulus and extremely low upper crystallization temperature. The glass specific modulus is not less than 3.73×10 6 m, and the upper limit of crystallization temperature is 1230~1255℃. The comparative examples are all high modulus and high specific modulus glass fiber components well known in the field of glass fiber. Mol 1 is less than 0.1, and Mol 3<10, so the upper limit of crystallization temperature is greater than 1255℃. Mol 1 only has 0.019 and 0.027, and Mol 3 are all less than 1, so their upper crystallization temperature limits are as high as 1364°C and 1372°C respectively. Compared with the embodiments, the ZrO2 content in Comparative Examples 1 to 4 is significantly higher, which will lead to higher production costs. Compared with the embodiments, Comparative Examples 5 to 7 all contain B2O3, which is volatile and will not only cause environmental pollution, but also increase the difficulty of controlling the batching system. In this way, the test data of the above embodiments further confirm that the high specific modulus glass fiber components and proportions determined by the high specific modulus glass fiber component determination method provided by the present invention have both high specific modulus and extremely low upper crystallization temperature limits, confirming the reliability and accuracy of the method. Compared with the above embodiments, Comparative Example 8 is obviously Mol 1<0.15、 Mol 3<15, which does not meet the requirements of the present invention; and it obviously contains more Y2O3, which will lead to a significant increase in cost, thereby limiting its market size. At the same time, the trained prediction model provided by the present invention was used to verify the comparative example 8, and the obtained crystallization upper limit temperature was 1260℃ and the specific modulus was 3.64×106 m, which is lower than the minimum specific modulus of 3.73×10 6 m.

[0165] An embodiment of the present invention further provides a device for determining the composition of a high specific modulus glass fiber. As a device in a logical sense, the device is formed by the CPU of the computing device in which the device is located reading the corresponding computer program from the non-volatile memory into the internal memory and executing the program. The device is used to implement the method for determining the composition of the high specific modulus glass fiber of the first aspect described above, comprising:

[0166] Building blocks for constructing molecular models composed of varying proportions of up to four components;

[0167] A simulation module is used to simulate each molecular model to obtain the modulus and density of the molecular model;

[0168] The training module is used to obtain the wire drawing temperature and upper crystallization temperature of each molecular model and determine the physical parameters of each molecular model; the physical parameters include elemental characteristics, oxide properties, and crystal structure parameters; and expand the molecular model into a multi-component format and use the performance parameters and physical parameters of the molecular model to train a random forest model to obtain a prediction model; the prediction model inputs are multi-components and their proportions and physical parameters, and the output is performance parameters; the performance parameters include modulus, density, wire drawing temperature, and upper crystallization temperature;

[0169] The solution module is used to solve the constructed multi-objective optimization model using the performance parameters output by the prediction model to obtain high specific modulus components and proportions.

[0170] In some specific implementations, the construction module may be used to perform step 100, the simulation module may be used to perform step 102, the training module may be used to perform steps 104 and 106, and the solution module may be used to perform step 108. Detailed description is omitted here.

[0171] It should be understood that the structure illustrated in the embodiment of the present invention does not constitute a specific limitation on the device for determining the composition of a high-specific modulus glass fiber. In other embodiments of the present invention, the device for determining the composition of a high-specific modulus glass fiber may include more or fewer components than shown, or may combine or separate certain components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0172] The information interaction, execution process, etc. between the modules in the above-mentioned device are based on the same concept as the embodiment of the method of the present invention. For specific contents, please refer to the description in the embodiment of the method of the present invention and will not be repeated here.

[0173] An embodiment of the present invention further provides a computing device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, a method for determining a high specific modulus glass fiber component in any embodiment of the present invention is implemented.

[0174] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor executes a method for determining a high specific modulus glass fiber component according to any embodiment of the present invention.

[0175] Specifically, a system or device equipped with a storage medium can be provided, on which software program codes that implement the functions of any of the above-mentioned embodiments are stored, and a computer (or CPU or MPU) of the system or device can be enabled to read and execute the program codes stored in the storage medium.

[0176] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute part of the present invention.

[0177] Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (e.g., CD-ROMs, CD-Rs, CD-RWs, DVD-ROMs, DVD-RAMs, DVD-RWs, and DVD+RWs), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer via a communications network.

[0178] In addition, it should be clear that the functions of any of the above embodiments can be achieved not only by executing the program code read by the computer, but also by enabling the operating system operating on the computer to complete part or all of the actual operations based on the instructions of the program code.

[0179] In addition, it can be understood that the program code read from the storage medium is written into a memory provided in an expansion board inserted into the computer or into a memory provided in an expansion module connected to the computer, and then based on the instructions of the program code, a CPU installed on the expansion board or expansion module is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above embodiments.

[0180] Those skilled in the art will understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk, etc. Various media that can store program codes.

[0181] It should be noted that, in this article, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprising a ..." do not exclude the presence of other identical factors in the process, method, article or device comprising the elements.

[0182] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for determining the components of a high specific modulus glass fiber composition, characterized in that: include: Build molecular models consisting of varying proportions of no more than four components; Simulating each of the molecular models to obtain the modulus and density of the molecular model; Obtaining the wire drawing forming temperature and the upper crystallization temperature of each molecular model, and determining the physical parameters of each molecular model; the physical parameters include element characteristics, oxide properties and crystal structure parameters; Expanding the molecular model into a multi-component format; wherein the proportion of at least eight components not included in the molecular model is 0; The multivariate component format and the corresponding performance parameters and the physical parameters are used as a training set, a random forest model is trained using the training set, and model parameters are fine-tuned using virtual multivariate data to obtain a prediction model; The virtual multivariate data is generated by perturbing the molecular model, including: randomly adding trace proportions of at least eight components not included in the molecular model, and renormalizing the proportions of no more than four components in the molecular model; and then generating virtual multivariate components and proportions that meet eutectic point conditions based on thermodynamic phase diagram rules; The performance parameters output by the prediction model are used to solve the constructed multi-objective optimization model to obtain the high specific modulus components and proportions; the input of the prediction model is the multi-components and proportions and physical parameters, and the output is the performance parameters; the performance parameters include modulus, density, wire drawing temperature and crystallization upper limit temperature; the multi-objective optimization model includes the objective function and constraints: The objective function is: The constraints include: in, f ( x ) is the input variable of the prediction model x The objective function of E ( x )for x The corresponding modulus; ρ ( x )for x The corresponding density; T 析 is the upper limit of crystallization temperature; Mol Y2O3 is the molar percentage of Y2O3; Mol La2O3 is the molar percentage of La2O3; Mol i For the i The mole percentage of the components; M is the number of components of the multi-component.

2. The method according to claim 1, characterized in that The high specific modulus components and proportions include the following components in molar percentage: SiO2: 57.5-59.5%, Al2O3: 10.0-15.0%, CaO: 6.0-10.0%, MgO: 19.49-23.0%, Li2O: 0-1.0%, K2O: 0-1.0%, Na2O: 0-1.0%, Fe2O3: 0-1.0%, TiO2: 0-1.5%, ZrO2: 0-1.0%, Y2O3: 0.02-0.5% and La2O3: 0-0.5%; the molar percentages of the components satisfy: ( Mol CaO + Mol ZrO2 ) / Mol Y2O3 ≥17 in, Mol CaO 、 Mol ZrO2 、 Mol Y2O3 are the molar percentages of CaO, ZrO2, and Y2O3, respectively.

3. The method according to claim 2, characterized in that The molar percentage of each component satisfies: 0.15≤( Mol CaO + Mol Li2O + Mol K2O + Mol Na2O + Mol Fe2O3 ) / Mol 2≤0.30 in, Mol CaO 、 Mol Li2O 、 Mol K2O 、 Mol Na2O 、 Mol Fe2O3 are the molar percentages of CaO, Li2O, K2O, Na2O, and Fe2O3, respectively; Mol 2 is the sum of the molar percentages of Al2O3 and MgO.

4. The method according to claim 2, characterized in that The molar percentage of each component satisfies: 15≤ Mol CaO / Mol Y2O3 in, Mol CaO 、 Mol Y2O3 are the molar percentages of CaO and Y2O3 respectively.

5. The method according to claim 2, characterized in that The glass modulus of glass fiber is ≥3.73×10 6 m, the upper limit temperature of crystallization is 1230~1255℃.

6. The method according to claim 1, characterized in that Simulating each of the molecular models to obtain the modulus and density of the molecular model includes: Determining a simulation environment of the molecular model and a simulation force field including long-range and short-range relationships between atoms; In the simulation environment and the simulation force field, heating and melting the molecular model and then cooling it to form an amorphous glass molecular model and obtain an equilibrium volume; determining a density of the molecular model based on the equilibrium volume, the physical parameters, and a ratio of components in the molecular model; The modulus is obtained by applying uniaxial tensile deformation under the amorphous glass molecular model.

7. The method according to claim 1, characterized in that When using the training set to train the random forest model, it also includes: introducing physical constraint features; the physical constraint features include correcting the wire drawing temperature if the upper limit crystallization temperature predicted by the model is lower than the wire drawing temperature.

8. The method according to any one of claims 1 to 7, characterized in that The performance parameters output by the prediction model are used to solve the constructed multi-objective optimization model to obtain high specific modulus components and proportions, including: S1: Generate an initial glass fiber component decomposition set and use it as the current population, initialized to the first database of the empty set; set a multi-objective optimization model including objective function and constraint conditions; S2: For all solutions in the current population and the first database, calculate corresponding performance parameters using a prediction model to evaluate the objective function value and the degree of violation of the constraint conditions of each solution; S3: Based on the objective function value and the violation degree, a density estimate and a total fitness are calculated; based on the non-dominated sorting and crowding distance, all non-dominated solutions are screened from the current population and the first database and stored in a second database; if all non-dominated solutions exceed the capacity of the second database, the solutions with the highest density estimate or the highest violation degree are gradually removed; otherwise, the dominated solutions with the lowest total fitness are gradually stored; S4: selecting parent individuals from the second database, performing crossover and mutation on the parent individuals, and generating a set of offspring glass fiber components that meet the constraint conditions; S5: Determine whether the current number of iterations reaches a preset upper limit. If so, output the high specific modulus component and proportion. Otherwise, decompose the descendant glass fiber component into a population for the next iteration, assign the second database to the first database, and return to step S2.

Images