Cohesive　zone　model　(CZM)　has　gained　considerable　attention　to　investigate　the　fracture　mechanism　of　asphalt　mixture.　The　parameter　of　CZM　was　generally　determined　through　manually　adjustment　to　match　the　numerical　simulation　to　the　experimental　measurement.　However,　the　method　is　time　consuming　and　precision　uncontrolled.　This　study　introduced　a　novel　approach　to　determine　the　CZM　parameters　based　on　an　optimization　approach.　The　semicircular　bending　test　was　conducted　at　an　intermediate　temperature　to　obtain　the　fracture　mechanism　of　asphalt　mixture.　The　Kriging　model　was　implemented　into　the　genetic　algorithm　as　a　surrogate　model　to　predict　the　CZM　parameters　of　bilinear　cohesive　law.　A　pre-select　operation　was　proposed　to　enhance　the　computational　performance.　The　result　showed　that　the　simulation　with　CZM　parameters　obtained　from　the　optimization　method　matched　well　with　the　experimental　measurement,　indicating　that　the　method　could　precisely　characterize　the　fracture　mechanics　of　asphalt　mixture.　The　pre-select　operation　could　achieve　a　more　efficient　optimization　using　few　initial　samples.　The　proposed　approach　provides　an　efficient　procedure　to　characterize　the　fracture　behavior　of　asphalt　mixture.