Tuned　particle　dampers　(TPDs)　are　recognized　for　their　innovative　approach　to　nonlinear　energy　dissipation,　maintaining　the　dynamic　characteristics　of　the　systems　they　protect.　However,　the　inherent　complexity　of　particle　system　nonlinearities　poses　substantial　challenges　in　developing　accurate　dynamical　models.　Traditional　analysis　methods　are　inadequate　in　addressing　the　dynamic　characteristics　of　such　complex　coupled　systems.　This　study　introduces　a　novel　integration　of　multibody　dynamics　with　discrete　element　methods　through　a　bidirectional　coupling　approach　to　a　comprehensive　model　of　the　TPD　system.　Then,　extensive　investigations　into　the　system＇s　responses　and　its　vibration　control　efficacy　under　diverse　external　forces　and　frequencies　are　conducted,　and　the　energy　dissipation　attributes　and　dynamic　adjustments　are　also　delved　into　system　mass.　Given　the　extensive　simulation　times　and　substantial　resource　demands　of　such　systems,　a　neural　network-based　predictive　model　has　been　devised　to　forecast　the　vibration　absorption　capabilities　of　TPDs,　offering　essential　insights　that　enhance　engineering　applications.