Forecasting pavement roughness conditions can facilitate decision making within a pavement management system at project and network levels. Because pavement roughness change over time is caused by some important conditions and certain stochastic factors, a parameter and dynamic forecasting model is more appropriate for forecasting roughness with respect to linear, static, and nonparameter forecasting models. Thus, an adaptive filter forecasting system is presented that forecasts pavement roughness conditions by means of an adaptive filter using roughness history. The concept of an adaptive filter forecasting system is introduced, along with its mathematical derivation and least-mean-square algorithm. In testing the system's validity, a given mathematical function is used to simulate changing pavement roughness conditions. In addition, a practical application of the adaptive filter forecasting system is presented. The roughness index used is the root-mean-square vertical acceleration of a response-type road-roughness measuring system. Finally, choice of the adaptive filter structure and its stability, based on roughness data collected from Austin Test Sections, are discussed. The structure of system should be decided before each application by experimental results with certain criteria. This is a major limitation of the system. (A)
Abstract