Proposal of a segmentation procedure for skid resistance data.

Abstract

Skid resistance of pavements presents a high spatial variability along a road. This pavement characteristic is related directly to wet weather accidents; therefore, it is important to identify and characterize the skid resistance of homogeneous segments along a road in order to implement proper road safety management.
Several data segmentation methods have been applied to other pavement characteristics (e.g. roughness). However, no application to skid resistance data was found during the literature review for this study. Typical segmentation methods are rather too general or too specific to ensure a detailed segmentation of skid resistance data, which can be used for managing pavement performance.
The main objective of this paper is to propose a procedure for segmenting skid resistance data, based on existing data segmentation methods. The procedure needs to be efficient and to fulfill road management requirements.
The proposed procedure considers the Leverage method to identify outlier data, the CUSUM method to accomplish initial data segmentation and a statistical method to group consecutive segments that are statistically similar. The statistical method applies the Student’s t-test of mean equities, along with analysis of variance and the Tuckey test for the multiple comparison of means. The proposed procedure was applied to a sample of skid resistance data measured with SCRIM (Side force Coefficient Routine Investigatory Machine) on a 4.2 km section of Chilean road and was compared to conventional segmentation methods.
Results show that the proposed procedure is more efficient than the conventional segmentation procedures, achieving the minimum weighted sum of square errors (SSE(p)) with all the identified segments statistically different.
Due to its mathematical basis, the proposed procedure can be easily adapted and programmed for use in road safety management.