Slag　powder　is　a　process　with　characters　of　multivariables,　strongly　coupling　and　nonlinearity.　The　material　layer　thickness　plays　an　important　role　in　the　process.　It　can　reflect　the　dynamic　balance　between　the　feed　volume　and　discharge　volume　in　the　vertical　mill.　Keeping　the　material　layer　thickness　in　a　suitable　range　can　not　only　improve　the　quality　of　powder,　but　also　save　electrical　power.　Previous　studies　on　the　material　layer　thickness　did　not　consider　the　relationship　among　the　material　layer　thickness,　quality　and　yield.　In　this　paper,　the　yield　and　quality　factors　are　taken　into　account　and　the　variables　that　affect　the　material　layer　thickness,　yield　and　quality　are　analyzed.　Then　the　models　of　material　layer　thickness,　yield　and　quality　are　established　based　on　generalized　regression　neural　network.　The　production　process　demands　for　highest　yield,　best　production　quality　and　smallest　error　of　material　layer　thickness　at　the　same　time.　From　this　point　of　view,　the　slag　powder　process　can　be　regarded　as　a　multi-objective　optimization　problem.　To　improve　the　diversity　of　solutions,　a　CT-NSGAII　algorithm　is　proposed　by　introducing　the　clustering-based　truncation　mechanism　into　solution　selection　process.　Simulation　shows　that　the　proposed　method　can　solve　the　multi-objective　problem　and　obtain　solutions　with　good　diversity.　(C)　2020　The　Author(　s).　Published　by　Elsevier　B.V.