Automatic detection of outlier data received in multi-parametric capillary sensors of diesel fuels fit for use
Michał Borecki , P. Prus , Michael L. Korwin-Pawlowski , Piotr Doroz , Jan Szmidt
AbstractThe multi-parametric capillary sensor with local sample heating has been shown as an effective tool for diesel fuel fit for use classification at the laboratory level of technology, where a trained operator performs the experiments. The sensor consists of disposable capillary optrode, head and measurement control unit. An increase of the technology level of the sensor requires automation of samples handling and implementing automatic rejection of uncertain outlier data. Such data uncertainty may come from variations of capillary optrode diameters, inaccuracy of optrode filling with sample, inaccuracy of corking the sample as well as inaccuracy of optrode positioning in the head. Mentioned inaccuracies of preparation of the measurement may lead to outlier data, which impacts the correctness of sample classification. In this paper automatic detection of outlier data received in multi-parametric capillary sensors of diesel fuels is proposed and examined with data collected by untrained and trained operators. Performed experiments show that direct statistical tools applied to raw data lead to improper results of outlier data pointing. The proper outlier data pointing taking place for raw data converted to vector pattern of data on the base of physical phenomena described by experimental data or with the use of analysis of first derivative of raw data characteristic points course.
* presented citation count is obtained through Internet information analysis and it is close to the number calculated by the Publish or Perish system.