چكيده به لاتين
An appropriate interchange type selection would help tackle traffic problems and avoid the decreasing level of traffic service and safety. Design alternatives are usually generated according to an engineer’s experience. However, some useful types of them may be neglected. The criteria for interchange type selection vary from one study to another, but, the criterion that should be considered first is the traffic performance. In this thesis, an algorithm is suggested to auto-generate design alternatives. A Fuzzy-based method was also proposed to investigate traffic performance as well as to compare and prioritize the interchange design alternatives and an index called Traffic Operation Index (TOI) was introduced. In this investigation, four types of surrogate safety measures were combined using fuzzy tools for the first time in the world and an index called No-Collision Potential Index (NCPI) was proposed to determine the safety level. VISSIM simulation, Artificial Neural Network (ANN), Particle Swarm Optimization (PSO), and Fuzzy Inference System (FIS) were used to estimate the value of effective parameters. Finally, using the ANN and PSO models, the parameters could be estimated just by knowing simple traffic and geometrical characteristics of the interchange parts. The method is applied to two real-world interchanges. The findings showed an acceptable precision and validity of the models. It was also found that using the ANN model, the driving time, density, NCPI, and TOI could be estimated with maximum error of ±1 seconds, ±2 veh/ln/km, ±8.5 percent, and ±3.5 percent, respectively. The maximum error for estimating the driving time, density, NCPI, and TOI by PSO model was ±7 seconds, ±5 veh/ln/km, ±9 percent, and ±10 percent, respectively. Totally, results indicated that the ANN model was more precise than PSO model, however, the PSO model was more valid.
