An adaptive neuro-fuzzy inference system for assessing runway friction decay

Abstract

The study describes a method of modelling runway friction decay using an adaptive neuro-fuzzy inference system (ANFIS). Using a given input/output data set a fuzzy inference system (FIS) whose membership function parameters are tuned using an optimization algorithm was obtained. This allows the fuzzy inference system to learn from the data it is modelling, i.e., the ANFIS is used to train a FIS model to emulate the data presented to it, by modifying the membership function parameters according to a chosen error criterion. The input variables are the friction measured immediately after each maintenance operation of surface degumming, and a parameter on current traffic; the output is the current or future friction, between one maintenance operation and the next. The model thus provides effective and efficient forecasting, representing the evolutionary behavior of pavement friction on the specific runways. Testing the model performance in terms of errors and dimension of the rule base indicates satisfactory effectiveness, sufficient for its application as a decision-making tool in airport maintenance management systems (APMS), also considering the efficiency of the model, since it does not require the operator to carry out on-site measurements or carry out difficult calculations for prediction of current or future levels of friction. The prediction of current or even future levels of friction becomes much more readily possible for airport managers.