14927

1426 دكتري

دكتري

ماده چگال

اصفهان : دانشگاه صنعتي اصفهان

1397

يازده، 106 ص.،: مصور، جدول، نمودار

فرهاد شهبازي، ميثم اكبرزاده

وحيد سالاري

محمد صادق موحد، امير لهراسبي، فرهاد فضيله

1398/05/26

كتابنامه

فيزيك

فيزيك

1398/05/26

2552878

AbstractIn this thesis we study and work on traffic data analysis of highways and urban areas Using physics science besides transportation engineering enable us to extract informationfrom traffic data set which can be speed density or flow Using clustering methods suchas K mean clustering algorithm and also statistical analysis methods such as Fokker Planckequation and Markov chain enable us to categorize and define traffic phases We showedthat for highways under study the three phase classification works better than two phaseclassification Also the border of phases were found Plotting the phase diagram helpedus to understad the traffics situations at the highways Using the knowledge of phases andborders of them besides Shannon entropy equation equip us to define the online phase of thehighway which can be a step to predict traffic jam After that to have a better vision of traffic traffic modeling named TASEP dynamic borrowed from physics is used First we showed that this simple model can show and explainthe traffic features With the help of TASEP dynamic we show that wide moving jam statehas a multiple stable states and this is the reason of seeing scatter data at this stage Wealso showed that TASEP dynamic can explain hysteresis effect We showed hysteresis likeeffect can happen in the system not only because of the structure of the highway network but also the different ranges of in out rates This moding showed that the hystersis effecthappens bacause of multiple stable states in wide moving jam state Any small changes inin out rate value can change the state of the system which has a multiple state state andcause hysteresis effect As a final work we worked on speed data of 40 urban streets of Philadelphia PA We chose adowntown area Using the definitions of information flow we found the connectivity betweenstreets i e we found how streets can affect each other We show that we can cluster streetsin a way that only streets with high affection on each other will be in the same group Atthe final step we showed that the information of the clustering method can help us to havea better prediction by using the right features for prediction method Keywords Data Analysis Traffic Data Clustering Markov Chain Fokker PlanckEquation Shannon Entropy Traffic Modeling Hysteresis Three Phase Traffic Model Infor mation Flow Prediction

فرهاد شهبازي، ميثم اكبرزاده

وحيد سالاري

محمد صادق موحد، امير لهراسبي، فرهاد فضيله