PSI - Issue 59

Victor Aulin et al. / Procedia Structural Integrity 59 (2024) 436–443

441

Victor Aulin et al. / Structural Integrity Procedia 00 (2019) 000 – 000

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

2

3

4

5

6

7

8

9

10

c - number of neurons in hidden layer Кількість нейронів у прихованому шарі

- розпізнаних навчальних прикладів, %

- розпізнаних тестових прикладів, %

–– recognized educational tasks, % – – recognized test tasks%

Fig. 2. Graphic display of convergence of recognized defects of resource-determining elements of the engine: the cylinder-piston group(a), the slider-crank linkage(b), the valvetrain(c) during training and testing of the ANNs.

Figure of the convergence of the values of the maximum error in the recognition of defects, clearly shows that the condition of the objective function of the minimum of errors is fulfilled at l = 7 for the valvetrain, for the slider crank linkage at l = 5, and for the cylinder-piston group at l = 7. The obtained results are presented in summary table 2 , where the number of neurons of the input layer ( N x ) is equal to the number of controlled parameters, and the number of neurons of the output layer ( N y ) is the number of possible defects.

Table 2. Determination of the optimal number of neurons in the hidden layer for detecting combinations of defects in resource-determining elements of engines

Size of the study sample

Maximum number of neurons in the hidden layer

Optimal number of neurons in the hidden layer

Name of resource determining elements

N x

N y

The cylinder-piston group 9

4 4 6

36 16 36

22 23 46

7 5 7

The slider-crank linkage

4 6

The valvetrain

The results are also presented in the form of ANN graph models for resource-determining elements of the engine. The achieved combinations of defects for the main resource elements of the YMZ-238 engine were formed in the form of a matrix of technological states and graph models of an artificial neural network (fig. 3-5).

Fig. 3. Graph-model of an artificial neural network of recognition of defects in the slide-crank linkage.