Resource　scarcity　has　driven　growing　interest　in　circular　economy　(CE).　Closed-loop　supply　chain　(CLSC)　with　returnable　transport　items　(RTIs)　in　the　food　industry　is　an　important　component　of　CE.　However,　existing　works　on　food　CLSC　with　RTIs　have　not　simultaneously　considered　the　perishability,　facility　location,　and　uncertain　demand　under　limited　information.　Therefore,　this　work　addresses　a　new　food　CLSC　optimisation　problem.　We　first　propose　a　non-linear　chance-constrained　programming　model.　It　is　then　transformed　into　a　mixed-integer　linear　programming　model　via　using　the　distribution-free　(DF)　method　and　sample　average　approximation　(SAA)　method,　respectively.　An　illustrative　example　reveals　that　the　DF　method　needs　only　10.50%　of　the　computation　time　of　the　SAA　method.　To　address　large-scale　problems,　an　improved　Lagrangian　relaxation　(LR)　method　is　developed.　To　address　the　computational　challenge　in　large-scale　problems,　an　improved　Lagrangian　relaxation　(LR)　algorithm　is　developed.　Results　show　that　CPLEX　achieves　a　gap　of　75.57%,　while　the　LR　surpasses　it　by　finding　near-optimal　solutions　with　a　gap　of　1.22%,　using　only　31.82%　of　the　computation　time　required　by　CPLEX.　For　this　work,　the　main　insights　are　summarised:　(1)　extending　product　shelf　life　can　reduce　the　total　cost;　and　(2)　to　alleviate　uncertain　demand　and　production　risks,　production　capacity　and　product　inventory　capacity　can　be　appropriately　expanded,　but　excessive　investment　may　not　improve　returns.