Dynamic and econometric models for predicting the automobile roads condition

Abstract

Essential components of the system of pavement condition management (PMS) of highways are the assessment of the current state of both individual sections and the network as a whole and the prediction of quantitative characteristics describing the process of changing the parameters of the design of road. With the development of technical diagnostic tools, radiophysical tools, geo-radars, have become increasingly used to obtain primary data. Their advantages consist in the possibility of conducting non-destructive testing during movement, as well as in the amount of initial information that other means of control do not provide. At the same time, the primary data sets obtained using GPR, in addition to primary processing, require subsequent interpretation in order to convert them to the form adopted in the road industry. Based on the analysis of the literature data, it follows that the actual task of improving the reliability of local and network level forecasts has not yet been fully resolved. In particular, monitoring even using a GPR installed on a car-laboratory, requires not only the cost of the measurements themselves, but also the preliminary calibration of the instrument (GPR) on these sections and the subsequent in-depth interpretation of the data obtained. In this paper, we propose to add, in addition to the existing criteria within the framework of the general operator model, a set of elements in the form of an operator, providing for the correction of the model, taking into account econometric indicators. In addition, an algorithm was first formulated and a solution was proposed for the problem of optimal selection of a set of sites in order to obtain the most reliable forecasts with the necessary condition for the minimum number of such sites. Also, a solution to the optimization problem with cost constraints is given, provided that the maximum predictive efficiency is achieved. As a result, both of these decisions are combined into a one integrated approach to forecasting at the network level, taking into account the results of the use of GPR diagnostics and dynamic models of the local level.