Influence of data non-normality on the statistical quality control of pavement works

ABSTRACT

The DNER-PRO 277/97 standard describes the method of statistical quality control for road works and services in Brazil, assuming a normal distribution of executive characteristics. However, the literature regularly documents the violation of this assumption. In this context, this study investigated the influence of non-normality on the acceptance/rejection decisions according to the aforementioned standard. Information on asphalt binder content from the execution of three projects was used as a reference. Using different statistical tests (Shapiro-Wilk, Kolmogorov-Smirnov and Anderson-Darling), the possibility of deviations from normality was confirmed, with the log-logistic probability distribution showing greater fit to the data. For this particular case, a Monte Carlo simulation pointed out a severe rejection bias for small samples when adopting the assumption of data normality. Therefore, in cases where non-normality cannot be disregarded, the risk for the executor may be higher than stated in the standard.

Keywords:
Asphalt paving; Quality control; Normal distribution; Statistical control