The paper develops a research already presented at ISATA whose aim is to analyse accident data to work out an accident risk model for vehicular flow in motorway environments. For this research, accident data were collected in Italian motorways over a period of about four years. Besides structural factors such as the total length of rectilinear and curve stretches for each motorway and meteorological conditions, used to homogenise accident data so as to compare them, the length of each rectilinear or curve stretches, and the exact position where the accident occurred, were collected. Particularly, rectilinear and curve stretches are subdivided into three sections: for straight roads intervals are from 0 to 150 meters, from this value to the last 150 meters and the last 150 meters; for curves intervals are little different due to their rather different planimetric development: the first interval is until 25% of the curve length, the following 50% and the last 25%. The model was built up by means of neural network and it is capable of describing how parameters influence the accident risk exposure when different environmental conditions are considered. So it is possible to measure how much fog, rain or snow affects the number of accidents when different road peculiarities (straight and curve roads) and one of the three sections are considered. For the covering abstract see ITRD E116488.
Samenvatting