PSI - Issue 48

6

Oleh Yasniy et al. / Procedia Structural Integrity 48 (2023) 183–189 Yasniy et al / Structural Integrity Procedia 00 (2019) 000 – 000

188

0,0000016

0,00012

Train data Test data

0,00011

0,0000014

Train data Test data

0,00010

0,0000012

0,00009

0,0000010

0,00008

0,0000008

0,00007

0,0000006 Average Squared Error

Average Squared Error

0,00006

0,0000004

0,00005

0,0000002

0,00004

10 20 30 40 50 60 70 80 90 100 Number of Trees

10

20

30

40

50

60

70

80

90 100

Number of Trees

Fig. 7. Dependences of the root mean square error on the number of trees for the linear (left) and non-linear (right region obtained by random forests

Fig. 8. The dependences of importance of input parameters for the linear (left) and non-linear (right) region obtained by random forests

The number of k -nearest neighbors in the linear and non-linear sections is 10 and 1, respectively. The number of trees in two cases is equal to 100. 4. Conclusions The stress-strain diagrams of 6061- T651 aluminum alloy at temperatures T = 20, 100, 150, 200, 250, and 300ºС were built by machine learning methods, particularly by k -nearest neighbors and random forests. It was found that the prediction results agreed with the experimental ones. It was determined that the errors of 9.6% and 5.9% for linear and non-linear regions were obtained by the k -nearest neighbors method in the test sample. The errors of the random forest method were 15% and 13.7%.