Development of a safety index for egyptian highways using artificial intelligence

Highway safety Deterioration is one of the major problems facing the highway Industry.The increase of accident rates is a matter of concern that disturbs highway users, authorities, engineers, and communities. Researchers in the last 30 years focused on developing techniques for pavement condition evaluation. Non of them investigated the possibility of developing a highway safety evaluation tool, which is needed for evaluatingthe safety aspects of any highway section. Egyptian rural highways facing the same problem and decision-makers use their own experience to performe highway safety improvement plan and allocate funds to upgrade safety condition of rural network.To improve safety condition of Egyptian rural highways, this research will therefore provide a detailed highway safety evaluation technique. This technique will take intoconsideration the highway features which affect highway safety. The developed technique relies on expert’s opinions and employs Artificial Intelligence (AI) technique to estimatesafety index.The developed safety index algorithm consists of two main parts. The first is the safety deducts point’s curves while, the second is the correction curve. Safety deducts points curves (29.curves) relate the highway features measurements or conditions with thecorresponding safety deduct points based on experts opinions. Experts opnions have been collected using specially desigened questionnaire forms ..A correction curve relates the Total Safety Deducts Points (TSDP) with the Total Modified Safety Deducts Points (TMSDP) has been developed to take care of any error due to any difference between field and questionnaire measuerments.Neural network has been utilized to develop the correction curve. An artificial highway sections data set has been produced using a specially written computer program. This dataset has been used to train the neural network program (input represent features values and output represents TSDP calculated using 29 safety urves).