Solving The Freight Train Formation Plan Problem:A Case Study On The Planned Kenya Railway Network

This thesis presents a formulation and solution of the railway freight Train Formation Plan(TFP)network problem based on the planned Kenya railway network as a case study.Different sections of the planned network are planned to be implemented in phases in the years 2020,2025 and 2040.For the network in each of the years,the freight train formation plan problem is formulated as a network design problem and the mathematical model proposed is an INLP that is solved using the global solver of the optimization software,LINGO 11.The optimization approach used aims at minimizing the total time spent by railway cars(total car-hour consumption)in the network.The projected freight attractions and productions data was collected from a feasibility study report and a gravity model was used to carry out freight distribution to obtain the origin-destination matrix.Global optimal solutions for the network TFP were found which can be used directly to inform the railway operation and planning.For the year 2040,the network consists of 9 yards and the result shows that 26 train services must be built per day in order to satisfy the railway freight transportation demand in the network.Ten are direct train connection services which serve non-adjacent yards.For the year 2025,two direct train connection services must be built;one from Mombasa to Malaba operated at a frequency of 3.44 trains per day and the other from Mombasa to Nairobi at a frequency of 4.60 trains per day.For the initial plan in year 2020,one direct train connection service from Mombasa to Nairobi must be built and operated at a frequency of 3.61 trains per day.The results obtained by the optimization approach can be directly used to inform network operation planning as well as form a framework for future studies on freight train operation planning for Kenya's future network.Although traditional solvers as well as the global solver used in this thesis are powerful and can find optimum solutions in short computer time in simple networks,the drawback is that they may take too long to solve complex non-linear problems for complex or large scale networks especially when a global optimum is sought.For such networks,fast sophisticated heuristics algorithms are capable of finding acceptable solutions in short computation time,which can be investigated in future research.