| In the 21 st century,with the increasing demands for lightweight in the fields of aerospace,aerospace,electronics,and especially automotive,magnesium alloys have attracted attention due to their many properties,especially low density performance characteristics.However,the two technical bottlenecks of magnesium alloy in the application process of corrosion resistance and poor shaping ability seriously restrict its wide application.Therefore,breaking through its inherent technological bottlenecks has become a hot topic of research.The influencing factors of traditional research and processing are very complex and changeable,resulting in unstable test results.In particular,the vivacity of magnesium alloys leads to metallurgical quality instability.These factors will inevitably lead to the large fluctuation of experimental data,low accuracy of conclusion,heavy workload,long time and high cost during the research process.For this reason,artificial intelligence research method has been widely used in the field of material science.Although it covers the research field of magnesium alloys,it has hardly been reported for the corrosion performance research.Therefore,based on the previous work of the research group,this paper uses the principles and characteristics of BP neural network and genetic algorithms to establish a prediction model for the corrosion and mechanical properties of magnesium alloys.The main results are as follows:1)Based on the data of corrosion properties of magnesium alloy,the 4-8-1 three layer BP neural network prediction model of two magnesium alloys-AZ31 and AZ91,was optimized by genetic algorithm with output value including the centrifugal pressure,corrosion time,content of Fe elements and content of the second phase,and the output value containing corrosion depth as the output.The difference of corrosion depth caused by different corrosion time under the same conditions is used to quantitatively characterize the corrosion resistance.The prediction model can predict the depth of corrosion more accurately,thus using the error between the comparison sample and the expected value of the corrosion depth to characterize the corrosion performance.The average absolute error of the corrosion depth is reduced from 7.3% to 2.3%,the maximum absolute error is less than 3%,the root mean square error of the traditional BP neural network is 0.6835,and the root mean square error of the BP neural network optimized by genetic algorithm is 0.3924,after the BP neural network prediction model optimized by genetic algorithm.Therefore,compared with the traditional BP neural network prediction model,the neural network prediction model optimized by the genetic algorithm has higher prediction accuracy,so that it can better predict the corrosion performance of the magnesium alloy.2)Based on the experimental results of 196 groups of mechanical properties of AZ31 and AZ91 magnesium alloys and related data,on the basis of data analysis and arrangement,alloy elements,deformation temperature,deformation rate,deformation coefficient,solution temperature,solid solution time,aging temperature,and aging time are input,and the tensile strength(UTS),yield strength(YS)and elongation(ELO)are output.The 8-8-1 three layer BP neural network prediction model optimized by genetic algorithm is used to train the data.The training results show that: The prediction model can accurately predict the tensile strength,yield strength and elongation.The absolute average error of the tensile strength of two magnesium alloys AZ31 and AZ91 is reduced from 2.80% to 0.88% and 1.2%;the absolute average error of yield strength is determined by 6.62% decreased to 3.3% and 3.0%;the absolute average error of elongation decreased from 10.16% to 8.0% and 7.8%.The root mean square error of the BP neural network is 0.7502,and the root mean square error of the BP neural network optimized by genetic algorithm is 0.4139.Compared with the general BP neural network prediction model,it has a higher prediction accuracy. |
PDF Full Text Request