Abstarct: Pavement management system enables choosing the most economic and efficient method for maintaining and recovery by deep indication of pavement condition and prediction its condition in the future. For a pavement management system, condition prediction models are like engines for a vehicle. Generally, pavement performance model express the pavement deterioration process during the period of using the road. In this paper, family model has been used for estimating performance of passages in the sari city. Pavement of different pieces which are similar in technical and loading settings are placed in the same family using modeling by family, and decrement in pavement quality is modeled baxsed on pavement condition index in set of pieces of one family. During a case study, passages of the sari city have been divided into two groups baxsed on their traffic data, including: first, passages with heavy traffic and second, passages with light traffic. Each group was modeled using regression model and artificial neural network. In regression model, 4th degree model had a better quality and comparing regression model with artificial neural network, it was observed that artificial neural network roughly had better results than regression model. In all cases correlation coefficient was in 0.87 - 0.925 range which is mostly because of input data limitation. Still, considering the amount of data, time and money spent for maintaining them, it can be stated that models have a very good accuracy.