Vous êtes sur la page 1sur 17

Supply chain management Lecture Inventory Management: case Grolsch

Marjan van den Akker

March 31, 2009

Problem description Grolsch

14 production lines available n dierent products to be made sequence dependent set up times no backlogging (delivery from stock) Goal: minimize inventory cost

Complications

Demand not exactly known; Stochasticity in set up times; Disruptions in production process.

Temporal simplications

Demand is exactly known in the rst T periods Ignore uncertainty in production process Suppose there is only one production line. Approach Solve the simplied problem for the rst T periods Use the obtained knowledge to solve the real problem.

Decompostion approach

Two subproblems: 1. Selection problem: Which products should be made on day t? 2. Lot-sizing problem: How much should be made per run? Approach Solve problem 1 Given the solution to problem 1, solve problem 2 Given the solution to problem 2, adjust the solution of problem 1 Given the new solution to problem 1, solve problem 2, etc

Goal of nding a solution of problem 1


to know which products to produce in period t (t = 1, . . . , T ) to know the net production time for doing so.

Therefore, we need to know The products to be made in period t; The rst and last product in period t; The order in which the remaining products are produced in period t; When the interperiodical set up between periods t and t + 1 is done.

Data for solving problem 2

D(i, t) = demand product i in period t; P P M (i) = production per minute of product i; CAP = capacity of warehouse; IC(i) = inventory cost per unit of product i per time unit; N P T (t) = net production time in period t (known quantity, if the solution to problem 1 is known). IN V (i, 0) = initial inventory product i.

Getting the LP formulation

Decision variables: ???? Decision variables: AP (i, t) = amount to produce of product i in period t (xed at 0 if i is not produced in t); IN V (i, t) = inventory of product i at the end period t.

Assumption: products are delivered at the end of the day; you can use both the inventory at the start of the day and the production of the day.

LP formulation (in words)

Objective function ???? Constraints ????

LP formulation
T n

min
t=1 i=1

IC(i)IN V (i, t)

subject to i, t

IN V (i, t) = IN V (i, t 1) + AP (i, t) D(i, t)


n i=1 n

AP (i, t) N P T (t) P P M (i)

IN V (i, t) CAP
i=1

t i, t

IN V (i, t), AP (i, t) 0

10

LP formulation with penalty terms

If the LP is infeasible, then you get NO information. Introduce additional decision variables OT (t) = amount of over time needed in period t ED(i, t) = amount of external deliveries needed in period t for product i Penalty cost P OT = cost per unit of over time P ED = cost per unit of any product bought elsewhere. If necessary, dierent penalty costs can be specied for dierent products and time periods.

11

New model
T

min
t,i

IC(i)IN V (i, t) + P OT
t=1

OT (t) + P ED
t,i

ED(i, t)

subject to IN V (i, t) = IN V (i, t 1) + AP (i, t) D(i, t) + ED(i, t)


n i=1

i, t

AP (i, t) N P T (t) + OT (t) P P M (i)


n

IN V (i, t) CAP
i=1

t i, t

IN V (i, t), AP (i, t), OT (t), ED(i, t) 0

12

Adjusting the solution of problem 1

Use a local search routine like Simulated Annealing. Possible neighborhoods ????

13

Adjusting the solution of problem 1 (2)


Possible neighborhoods: Add a product to the list of products that will be produced in period t. Remove a product from the list of products that will be produced in period t. Change the rst or last product to be produced in period t. Do the interperiodical set up between periods t and t + 1 in the other period. N.B.: Use sensitivity analysis to the linear program in the local search approach.

14

Getting rid of the temporal simplications

14 instead of 1 production line: more of the same (nothing to worry about). Stochasticity of set up times: This will average out (not much to worry about). Disturbances in the production process: Keep some production capacity idle. Unknown demand: ????

15

How to arm yourself against uncertain demand

Rules of thumb: ????

16

Questions?

17

Vous aimerez peut-être aussi