A Sample Size Statistical Analysis and Its Impact on Decarburization Measurements Metrics

Steel’s transformation is crucial for quality improvement. The high-temperature reheating process is usually performed under oxidizing atmospheres. Therefore, decarburization is present, resulting in material losses and depreciating the material’s quality. The decarburization depth is commonly used to authenticate the quality of steel or develop prediction models. This depth is commonly measured by optical microscope techniques and following the standard E1077-01. Nonetheless, not much care has been paid to the sample size impact upon measurement metrics. The present study uses 1045-steel metallographies to analyze the influence on standard statistical tools, such as Kolmogorov-Smirnov and Anderson-Darling normality tests. Then, data were used to study the analysis of variance and honestly significant difference Tukey tests. The analysis shows that the optimal sample size for statistical analysis occurred between 20 and 30. In contrast, models must observe the RMSE’s performance as a function of the size.