We　study　the　train　aggregation　behavior　in　a　railway　subsystem　using　the　Markov　approach.　This　approach　is　different　from　the　queue　theory,　which　is　a　main　method　to　design　some　railway　subsystems　in　the　literature.　With　some　simple　assumptions,　we　solve　the　corresponding　master　equation　describing　the　train　aggregation　behavior.　Using　the　least　square　trigonometric　approximation,　we　construct　a　continuous　rate　function　for　a　real　example,　and　show　the　evolution　of　the　size　of　train　cluster　with　time.　Also　the　variance　and　relative　error　of　the　train　aggregation　are　calculated.