Infeasibility of model due to new constraints
AnsweredWhen I add certain constraints, my model shows infeasibility.
Initial Model
import gurobipy as gp
from gurobipy import GRB
# Input data
plants = ['Plant1', 'Plant2', 'Plant3']
warehouses = ['Warehouse1', 'Warehouse2']
transport_cost = {
('Plant1', 'Warehouse1'): 10,
('Plant1', 'Warehouse2'): 15,
('Plant2', 'Warehouse1'): 20,
('Plant2', 'Warehouse2'): 25,
('Plant3', 'Warehouse1'): 30,
('Plant3', 'Warehouse2'): 35
}
discounts = {
'Plant1': {
(0, 100): 0.9,
(101, float('inf')): 0.8
},
'Plant2': {
(0, 50): 0.95,
(51, float('inf')): 0.85
},
'Plant3': {
(0, 75): 0.92,
(76, float('inf')): 0.82
}
}
plant_volumes = {
'Plant1': 80,
'Plant2': 120, # Adjusted volume value outside the defined range
'Plant3': 90
}
warehouse_demands = {
'Warehouse1': 100,
'Warehouse2': 120
}
# Create model
model = gp.Model('Transportation')
# Create variables
x = model.addVars(plants, warehouses, name='x')
# Set objective
obj = gp.quicksum(
transport_cost[plant, warehouse] * (1 - discount)
* x[plant, warehouse]
for plant in plants
for warehouse in warehouses
for (volume_range, discount) in discounts[plant].items()
if volume_range[0] <= plant_volumes[plant] <= volume_range[1]
)
model.setObjective(obj, GRB.MINIMIZE)
# Add constraints
supply_constraints = {
plant: model.addConstr(gp.quicksum(x[plant, warehouse] for warehouse in warehouses) <= plant_volumes[plant])
for plant in plants
}
demand_constraints = {
warehouse: model.addConstr(gp.quicksum(x[plant, warehouse] for plant in plants) >= warehouse_demands[warehouse])
for warehouse in warehouses
}
# Optimize model
model.optimize()
# Print solution
if model.status == GRB.OPTIMAL:
print('Optimal solution found.')
for plant in plants:
for warehouse in warehouses:
print(f'x[{plant}, {warehouse}] = {x[plant, warehouse].x}')
else:
print('No solution found.')
Now the challenge with this model is it fulfills the demand of a warehouse with multiple plants which is not desired hence introduced 2 new variables which are mentioned below
single_assignment_constraints = {
warehouse: model.addConstr(gp.quicksum(x[plant, warehouse] for plant in plants) == 1,name='singleflag')
for warehouse in warehouses
}
warehouse_assignment_constraints = {
warehouse: model.addConstr(gp.quicksum(x[plant, warehouse] for plant in plants) <= y[plant] * plant_volumes[plant],name='assignment')
for warehouse in warehouses
for plant in plants
}
The moment I do this, the result becomes infeasible in though the demand is below the supply and warehouse capacities.
So i ran IIS with below result but not able to pinpoint the problem
IIS computed: 2 constraints, 0 bounds
IIS runtime: 0.02 seconds (0.00 work units)
Constraints in the Irreducible Inconsistent Subsystem (IIS):
demand singleflag
0
-
Official comment
This post is more than two years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support. -
Your issue is not reproducible because variable \(\texttt{y[plant]}\) is not defined.
After computing the IIS, you can write it to a file
model.write(iis.ilp)
and open the IIS file \(\texttt{iis.ilp}\) in any standard text editor. You can then analyze the constraints in the IIS and try to find out why they together are infeasible.
You are saying that the 2 constraint found by the IIS algorithm are demand and singleflag
model.addConstr(gp.quicksum(x[plant, warehouse] for plant in plants) >= warehouse_demands[warehouse])
model.addConstr(gp.quicksum(x[plant, warehouse] for plant in plants) == 1These 2 constraint are incompatible because \(\texttt{warehouse_demands[warehouse] > 1}\).
I guess your \(\texttt{y}\) variable is a binary assignment constraint which you wanted to use in the singleflag constraint instead of \(\texttt{x}\).
0
Post is closed for comments.
Comments
2 comments