Outliers Detection Using Control Charts for Oil Wells

The presence of moderate values in a normal population is more likely than the presence of extreme values. Within this context, the assumption of normality of any population is due to the high probability of data to be normally distributed [1, 2]. The definition of outliers is subject to analysis and interpretation of results. Decisions regarding the identification of outliers should be taken individually and depend on a specific experiment [1]. Control charts are records of observations in statistical process built in a Cartesian coordinate system. The measurements obtained are represented in a time/space order and compared with the control limits. If any measurement exceeds the control limits, the process is considered to be out of bounds of statistical control and the value identified is defined as an outlier. Thus, this work aims at identifying outliers by control charts using data from drilling oil wells in order to improve the generation of synthetic sonic profile. This work was supported by FAPITEC and CNPq.