PSI - Issue 64

Isabella Mazzatura et al. / Procedia Structural Integrity 64 (2024) 114–121 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

120

7

Wall 2 Wall 3 Wall 4 Wall 5 Wall 6 Wall 7

100

90

80

Linear regression

Wall 2 Wall 3 Wall 4 Wall 5 Wall 6 Wall 7

y = -18.313x + 86.585 y = -15.703x + 64.233 y = -15.028x + 54.835 y = -15.692x + 70.063 y = -20.79x + 63.315 y = -14.383x + 61.851 y = -19.489x + 68.186

70

60

50

Heigth [cm]

40

30

All configurations

20

10

0

-2

-1

0

1

2

3

4

Error [cm]

Fig. 8. Linear regression in between error and height of the ducts, all configurations and all concrete cover depths are considered. The red line represents the regression of the data considered all together (all configurations, all concrete covers, all walls).

4. Conclusion It can be therefore concluded that the modality used to perform the test and the existing concrete cover does not influence the result. Indeed, in all cases the trend is in having more relevant errors in correspondence to the base or to the top of the specimens, therefore suggesting a potential influence of the geometry and/or boundary conditions (at the top of the walls is the air, while at the base, the ground) on the electromagnetic technique. This aspect, anyway, cannot be further investigated since the samples are very similar from the geometrical point of view. In the future, it is advisable to provide new samples geometrically different to investigate better this trend. The range of errors obtained lies always within the half-diameter of the ducts employed, and thus is acceptable. In the absence of further information on the influence of geometric/boundary conditions, it is possible to consider all the data together and obtain the histogram shown in Fig. 9a.

Q-Q plot

5

4

3

2

1

0

-4

-3

-2

-1

0

1

2

3

4

Empirical data [cm]

-1

-2

-3

-4

-5

Theoretical data [cm]

a) b) Fig. 9. a) shows the histogram of the data and the fitting of the beta, normal, and gamma distribution. b) show the Q-Q plot within the empirical data and the theoretical ones considering them normally distributed.