PSI - Issue 61

Jose Beltra Mira et al. / Procedia Structural Integrity 61 (2024) 156–163 Author name / Structural Integrity Procedia 00 (2024) 000–000

160

5

(a) ABS Cases

(b) PC Cases

(c) PLA Cases

7

7

7

6

6

6

0.5 )

0.5 )

0.5 )

5

5

5

4

4

4

1 Mean K Q (MPa ·m 2 3

1 Mean K Q (MPa ·m 2 3

1 Mean K Q (MPa ·m 2 3

0

0

0

W = 30 mm W = 40 mm W = 50 mm

W = 30 mm W = 40 mm W = 50 mm

W = 30 mm W = 40 mm W = 50 mm

Fig. 4. Mean conditional fracture toughness K Q for each material and sample size. Note that the error bars indicate standard error SE = σ/ √ n .

attempted, but the dataset failed the model adequacy test (Anderson-Darling for residual normality (AD = 0.926, p = 0.016) and Levene’s Test for equal variances (L = 1.19, p = 0.360)), so a non-parametric method must be used to analyze the data. For this, the Kruskal-Wallis test (based on medians instead of means) was used. The analysis shown in Table 5 indicates that there is no statistically significant di ff erence inK Q between sample sizes. The main e ff ects plot is shown in Figure 5 gives another perspective of the e ff ect of material selection and sample size. The e ff ect can be clearly noted using a traditional main e ff ects plot (based on data means) as well as one driven by medians.

Table 5. Results of the non-parametric statistical analysis.

Factor

Test

H-value p-value Conclusion

n

Material

Kruskal-Wallis 27 18.84

< 0.001 Statistically significant impact 0.588 Not statistically significant impact

Sample Size Kruskal-Wallis 27 1.06

Main Effects Versus Material

Main Effects Versus Sample Size

6

6

5

5

0.5 )

0.5 )

Median Mean

4

4

3 K Q (MPa ·m

3 K Q (MPa ·m

2

2

ABS

PC

PLA

W = 30 mm W = 40 mm W = 50 mm

Fig. 5. Main e ff ects plots relative to both the means and medians of the conditional fracture toughness K Q .

4.2. Versus Material Properties

If there was a significant dependence on specimen size, it would be expected that the mechanical properties would have a non-linear relationship with the sample size. To test this and further validate the results from the statistical analysis, a comparison of the breaking force relative to yield stress, elastic modulus, and elongation. If these mechan ical properties are consistent between sample sizes, it is expected that the intervals between the breaking strength values will depend mostly on the area of material bring broken in the test (i.e., the crack area). Table 6 shows the broken areas for each case, which match very nicely with the intervals between breaking strength values sorted by major material property (Figure 6). Based on this analysis, it was concluded that the strength of the CT samples was