Research On Multi-parameter Ultrasonic Evaluation Method Of Grain Size Based On Evolutionary Algorithm

The microstructure of alloy materials such as superalloys and titanium alloys can reflect some performance characteristics,such as: presence or absence of defects,service life,internal uniformity,etc.,while grain size is most sensitive to material properties.Ultrasonic detection is used to detect the material and extract corresponding characteristic parameters,such as: sound velocity,attenuation coefficient,backscattering signal,nonlinear coefficient,etc.,to construct a linear evaluation model of ultrasonic characteristic parameters and grain size,and input corresponding ultrasonic characteristics.The parameters obtain the corresponding grain size values,thereby achieving a process of non-destructive characterizing the grain size.At present,the grain size is mainly evaluated by single ultrasonic characteristic parameters,and the characterization relationship is not comprehensive enough.When the microstructure of the material is more complicated,the accuracy of the evaluation results and the linear relationship of the curve are poor,and even the evaluation failure occurs.In response to these problems,consider merging multiple ultrasound parameters and establishing a new ultrasound evaluation relationship model with grain size.When dealing with multiple ultrasonic characteristic parameters,it also faces the difficulty of selection and control,changing the single evaluation mechanism of the traditional evaluation model,forming a controllable multiple ultrasonic parameters and finding the pattern recognition problem of the model decision coefficients.In this paper,the initial screening of multiple ultrasonic parameters is carried out by correlation measurement method.The mapping function is constructed to map multiple parameters into single-dimensional parameters,and the evolutionary algorithm is introduced to deal with the optimization problem.A multi-parameter ultrasonic evaluation of grain size relationship model is established.The main research work and achievements are as follows:(1)The basic properties of alloy materials such as GH4169 and TC4 are introduced.The metallographic and ultrasonic detection methods are used to extract the corresponding characteristic parameters.The traditional linear relationship model of ultrasonic evaluation of grain size is analyzed by experiments.(2)Considering that the characteristic information of single ultrasonic characteristic parameters reflecting the grain size is not comprehensive,it is proposed to combine multiple ultrasonic parameters to construct a multi-parameter ultrasonic evaluation method for non-destructive quantitative characterization of GH4169 grain size.According to the correlation metric,the effective parameters are selected from the ultrasonic parameters such as sound velocity,attenuation coefficient and nonlinear coefficient.The mapping function is mapped into single-dimensional ultrasonic parameters and normalized.For the fitting process of single-dimensional parameters and grain size,Setting the average absolute error as the optimization goal and setting the target to the minimum optimization problem,introducing the evolutionary algorithm to solve the optimization problem,finding the best mapping function coefficient and fitting function coefficient,and finally establishing a multi-parameter ultrasonic evaluation model.The above method can effectively solve the problems of large error and low evaluation accuracy of the traditional evaluation model,and has a good evaluation effect.(3)With the increase of the complex information of the microstructure of the object under test,the evaluation curve established with the minimum error is getting worse and worse.To solve this problem,a primary ? phase crystal of TC4 titanium alloy with monotonicity is proposed.Ultrasonic evaluation method for particle size.According to the correlation metrics,the initial dimension reduction of the ultrasonic parameters is carried out,and then the mapping function is reduced to a single-dimensional parameter and fitted to the grain size to construct a monotonicity(the sequence parameter difference of the characteristic parameter samples is positive or negative at the same time)The optimization problem of optimization target is to introduce the differential evolution algorithm to solve the optimization problem,find the optimal mapping function and the fitting function coefficients,and then establish a multi-parameter ultrasonic evaluation model with monotonicity of mapping.This method exhibits good monotonic performance and small error.(4)In order to ensure that the ultrasonic evaluation model guarantees small error while reflecting good monotonic performance,the multi-objective optimization algorithm(NSGA-II)is introduced to optimize the target,and the optimization problem of multi-objective minimization is established,and monotonicity is also at the same time.The constraint mode is transformed into the Spearman rank correlation coefficient,which leads to a series of optimal solution sets on the Pareto frontier,forming an intelligent selectable evaluation model set,which is convenient for the actual application process to meet the current problem set as the evaluation model.The method considers multiple objectives at the same time,provides selectable models and forms an intelligent evaluation model set.The performance is more stable,the model is more diverse,and the advantages of the multi-parameter ultrasonic evaluation model under single target are integrated.(5)The fitting method of the evaluation model is improved,and the evaluation method for monotonicity and accuracy is proposed.In the first-order fitting relationship model,the error is large,and the second-order fitting relationship model is treated with unevaluable phenomena.The monotonicity and accuracy are combined and the second-order fitting model is constrained.This method ensures the advantage of the second-order fitting model with small error and has good monotonicity,which provides a feasible idea for the ultrasonic evaluation model fitting method.