| |||||||||

Binary fixing heuristic for the Coco problem Description - changing bounds on variables
- basis in- and output
- setting Optimizer parameters
- saving and loading MIP solutions (file fixbvls.mos)
Source Files fixbv.mos (!******************************************************* * Mosel Example Problems * * ====================== * * * * file fixbv.mos * * `````````````` * * Example for the use of the Mosel language * * (Using the complete Coco Problem, as in coco.mos, * * this program implements a binary fixing heuristic) * * * * (c) 2008 Fair Isaac Corporation * * author: S. Heipcke, 2001, rev. Feb. 2010 * *******************************************************!) model Coco ! Start a new model uses "mmxprs" ! Load the optimizer library parameters PHASE=5 (!* Phase = 4: Mines may open/closed freely; when closed save 20000 per month * Phase = 5: Once closed always closed; larger saving !) end-parameters declarations NT=4 ! Number of time periods RP=1..2 ! Range of products (p) RF=1..2 ! factories (f) RR=1..2 ! raw materials (r) RT=1..NT ! time periods (t) REV: array(RP,RT) of real ! Unit selling price of product p CMAKE: array(RP,RF) of real ! Unit cost to make product p ! at factory f CBUY: array(RR,RT) of real ! Unit cost to buy raw material r COPEN: array(RF) of real ! Fixed cost of factory f being ! open for one period REQ: array(RP,RR) of real ! Requirement by unit of product p ! for raw material r MXSELL: array(RP,RT) of real ! Max. amount of p that can be sold MXMAKE: array(RF) of real ! Max. amount factory f can make ! over all products PSTOCK0: array(RP,RF) of real ! Initial product p stock level ! at factory f RSTOCK0: array(RR,RF) of real ! Initial raw material r stock ! level at factory f CPSTOCK = 2.0 ! Unit cost to store any product p CRSTOCK = 1.0 ! Unit cost to store any raw mat. r MXRSTOCK = 300 ! Max. amount of r that can be ! stored each f and t make: array(RP,RF,RT) of mpvar ! Amount of product p made at ! factory f sell: array(RP,RF,RT) of mpvar ! Amount of product p sold from ! factory f in period t buy: array(RR,RF,RT) of mpvar ! Amount of raw material r bought ! for factory f in period t pstock: array(RP,RF,1..NT+1) of mpvar ! Stock level of product p at ! factory f at start of period t rstock: array(RR,RF,1..NT+1) of mpvar ! Stock level of raw material r at ! factory f at start of period t openm: array(RF,RT) of mpvar ! 1 if factory f is open in ! period t, else 0 end-declarations REV :: [400, 380, 405, 350, 410, 397, 412, 397] CMAKE :: [150, 153, 75, 68] CBUY :: [100, 98, 97, 100, 200, 195, 198, 200] COPEN :: [50000, 63000] REQ :: [1.0, 0.5, 1.3, 0.4] MXSELL :: [650, 600, 500, 400, 600, 500, 300, 250] MXMAKE :: [400, 500] PSTOCK0 :: [50, 100, 50, 50] RSTOCK0 :: [100, 150, 50, 100] ! Objective: maximize total profit MaxProfit:= sum(p in RP,f in RF,t in RT) REV(p,t) * sell(p,f,t) - ! revenue sum(p in RP,f in RF,t in RT) CMAKE(p,f) * make(p,f,t) - ! prod. cost sum(r in RR,f in RF,t in RT) CBUY(r,t) * buy(r,f,t) - ! raw mat. cost sum(p in RP,f in RF,t in 2..NT+1) CPSTOCK * pstock(p,f,t) - ! p stor. cost sum(r in RR,f in RF,t in 2..NT+1) CRSTOCK * rstock(r,f,t) ! r stor. cost if PHASE=4 then ! Factory fixed cost MaxProfit -= sum(f in RF,t in RT) (COPEN(f)-20000)*openm(f,t) elif PHASE=5 then MaxProfit -= sum(f in RF,t in RT) COPEN(f)* openm(f,t) end-if ! Product stock balance forall(p in RP,f in RF,t in RT) PBal(p,f,t):= (pstock(p,f,t+1) = pstock(p,f,t) + make(p,f,t) - sell(p,f,t)) ! Raw material stock balance forall(r in RR,f in RF,t in RT) RBal(r,f,t):= rstock(r,f,t+1) = rstock(r,f,t) + buy(r,f,t) - sum(p in RP) REQ(p,r)*make(p,f,t) ! Capacity limit at factory f forall(f in RF,t in RT) MxMake(f,t):= sum(p in RP) make(p,f,t) <= MXMAKE(f)*openm(f,t) ! Limit on the amount of prod. p to be sold forall(p in RP,t in RT) MxSell(p,t):= sum(f in RF) sell(p,f,t) <= MXSELL(p,t) ! Raw material stock limit forall(f in RF,t in 2..NT+1) MxRStock(f,t):= sum(r in RR) rstock(r,f,t) <= MXRSTOCK if PHASE=5 then ! Once closed, always closed forall(f in RF,t in 1..NT-1) Closed(f,t):= openm(f,t+1) <= openm(f,t) end-if ! Initial product levels forall(p in RP,f in RF) pstock(p,f,1) = PSTOCK0(p,f) ! Initial raw material levels forall(r in RR,f in RF) rstock(r,f,1) = RSTOCK0(r,f) forall(f in RF,t in RT) openm(f,t) is_binary !************************************************************************ ! Solution heuristic: ! solve the LP and save the basis ! fix all open variables which are integer feasible at the relaxation ! solve the MIP and save the best solution value ! reset all variables to their original bounds ! load the saved basis ! solve the MIP using the solution value found previously as cutoff !************************************************************************ declarations TOL=5.0E-4 ! Zero-tolerance osol: array(RF,1..2) of real ! Solution values for openm variables bas: basis end-declarations setparam("zerotol", TOL) ! Set Mosel comparison tolerance setparam("XPRS_verbose",true) ! Enable message printing in mmxprs setparam("XPRS_CUTSTRATEGY",0) ! Disable automatic cuts - we use a heuristic setparam("XPRS_PRESOLVE",0) maximize(XPRS_LPSTOP,MaxProfit) ! Solve the LP problem savebasis(bas); ! Save the current basis writeln("\n>>>>LP relaxation objective: ", getobjval) ! Fix all variables in the first two time periods which are integer feasible forall(f in RF, t in 1..2) do osol(f,t):= openm(f,t).sol if osol(f,t) = 0 then openm(f,t).ub:= 0.0 elif osol(f,t) = 1 then openm(f,t).lb:= 1.0 end-if end-do maximize(XPRS_CONT,MaxProfit) ! Solve the MIP problem ifgsol:=false if getprobstat=XPRS_OPT then ! If an integer feas. solution was found writeln("\n>>>>Heuristic MIP solution: ", getobjval) ifgsol:=true solval:=getobjval ! Get the value of the best solution end-if ! Reset variables to their original bounds forall(f in RF, t in 1..2) if ((osol(f,t) = 0) or (osol(f,t) = 1)) then openm(f,t).lb:= 0.0 openm(f,t).ub:= 1.0 end-if loadbasis(bas) ! Load the saved basis: bound changes are ! immediately passed on to the optimizer ! if the problem has not been modified ! in any other way, so that there is no ! need to reload the matrix if ifgsol then ! Set the cutoff to the best known solution setparam("XPRS_MIPABSCUTOFF", solval) end-if maximize(MaxProfit) ! Solve the MIP problem if getprobstat=XPRS_OPT then writeln("\n>>>>Best integer solution: ", getobjval) else writeln("\n>>>>Best integer solution: ", solval) end-if end-model | |||||||||

Copyright 2017 Fair Isaac Corporation. |