The objective of this investigation was to develop a model to predict on-board sound intensity (OBSI) on hot mix asphalt pavements using on-site and laboratory data. The data used included noise and physical property data collected on 25 asphalt-surfaced roadway test sections at the MnROAD pavement testing facility. These test sections were constructed mainly in 2007 and 2008 using a variety of materials, mixtures and layer thicknesses. A modeling approach called the mechanism decomposition approach was used to develop the models. In this approach, the contributions of different noise mechanisms to the overall noise level and to noise in certain frequency ranges are modeled separately then are combined to form the total noise spectrum. Ultimately, two nonlinear statistical models were developed that predict one-third octave band and overall sound intensity levels on asphalt-surfaced pavements. The models incorporate the pavement parameters that were found to have the most significant effects on tire-pavement noise including pavement macrotexture, air temperature, modulus of the pavement surface layer, and the combined effect of temperature and modulus. The models differ in the type of texture data used as an input parameter. The models have been found to predict the overall OBSI sound intensity level to within 1.5 dB and the onethird octave bands to within 2 dB for most of the pavements tested. Other metrics and evaluation of the model accuracy by cell, year, temperature and other factors are also reported.
