Quantitative Study On Small Angle Scattering Of Nano-Precipitates In Aluminum Alloy

In the field of microstructure characterization in materials science,small-angle scattering is a commonly used means,which makes up for the shortcomings of electron microscope: small field of view and no statistical significance.In addition,small-angle scattering can achieve rapid,in-situ characterization of the microstructure.However,the energy of ordinary Xrays is too low so that X-rays can only penetrate several microns into alloy which makes it impossible to effectively characterize the microstructure of alloys.The high-energy X-rays of a synchrotron radiation source can easily penetrate hundreds of microns into alloy,quickly completing the in-situ characterization of the alloy.In addition,for larger metal parts,a neutron source can be used instead of a high-energy X-ray source,so that the penetration depth of the beam is increased from the micrometer scale to the centimeter scale.Aluminum scandium alloy has very good mechanical properties.It has high strength,excellent plasticity and fracture toughness.It has a wide range of applications in aerospace,machining and other fields.Adding a small amount of zirconium to the aluminum scandium alloy can reduce the cost of the alloy,increase the processing temperature,and prevent the aggregation of precipitates.The structure of the nano-precipitates in aluminum alloy is the key to determine the mechanical properties of the alloy.In order to study the effect of aluminum alloy manufacturing process and heat treatment parameters on the size and its distribution of nano-precipitates,high-energy X-rays and small neutron scattering on the aluminum alloy are used in this subject to characterize the aluminum alloy.This project has carried out small-angle X-ray scattering experiments on aluminum-scandium alloy in Shanghai synchrotron radiation facility.Its advantages are high accuracy and short test time,which can be used for realtime in-situ experiments.The small-angle neutron scattering experiment of aluminum-scandium-zirconium alloy has the advantage that it can perform non-destructive and statistically significant detection on large block samples.The TEM electron microscope is used to characterize the experimental samples in advance,and it is observed that the nano-precipitates inside the sample has spherical or core-shell structure.This project establishes a multiparameter model of the nano-precipitates structure,and then processes the small-angle scattering experimental data of the sample.However,the processing of small-angle scattering experimental data is relatively complicated.For the metal materials small-angle scattering data fitting problem,the current methods mostly use manual adjustment of parameter values to fit unknown parameters in the model.This method has obvious disadvantages: manual adjustment efficiency is low,the process is complicated,the accuracy is not high enough,and human subjectivity has a greater impact on the results.A good alternative to manual adjustment of parameters is to automate the fitting process of the experimental data.The data fitting problem can be transformed into an optimization problem.Therefore,the strategy of automatic data fitting is an automated process of solving the transformed optimization problem.The automated data fitting process can greatly improve the fitting efficiency and accuracy.It can also greatly weaken the impact of human subjectivity in the fitting process.A new automated data fitting strategy is conceived in this paper,which consists of three parts: Bayesian optimization,gradient descent method,and grid search method.This automated data fitting strategy solves the two shortcomings of the gradient descent method: the sensitivity to the initial value and the easiness of falling into the local optimal value.In addition,this paper proposes a paradigm for constructing loss functions that the loss function can be composed of the following ways: loss function = main standard + gradient provision + regular term.The loss function paradigm solves the problem that the gradient descent method fails on the integer-type loss function Cor Map.By combining the automated data fitting strategy and new loss functions paradigm,this paper successfully designs a new small-angle scattering data fitting strategy and,based on this strategy,develops a software for small-angle scattering data fitting,completing fully automated small-angle scattering data fitting for the first time in the world.This fully automated fitting function enables to greatly improves the efficiency of small-angle scattering data fitting,expand the search range of parameter values to be fitted,and weaken the influence of human subjectivity on the fitting results.The use of the automated data fitting strategy is easy and simple,and users can easily use them without deep expertise,which shows certain advantage in cross-disciplinary fields.In addition,in this paper,the software is used to fit two sets of small-angle scattering experimental data of aluminum-scandium alloy and aluminum-scandium-zirconium alloy to verify the effectiveness and reliability of the software.At present,the data fitting optimization strategy has been patented.It has been open to users of Shanghai synchrotron radiation facility and Mianyang research reactor,and has been recognized by these users.