The facility-sizing problem is formulated as follows: m facilities are to be installed, each with capacity xi ≥ 0, i = 1, ..., m. Then the random demand ξi arrives at facility i, with a known joint distribution of the random vector ξ = (ξ1, ..., ξm).
A realization of the demand, ξ = (ξ1, ..., ξm), is said to be satisfied by the capacity x if xi ≥ ξi, ∀i = 1, ..., m.
- random demand vector ξ follows a multivariate normal distribution and correlation coefficients ρi, j , i! = j .
- μ: Mean vector of the multivariate normal distribution.
- Default: [100, 100, 100]
- Σ: Variance-covariance matrix of multivariate normal distribution.
- Default: [[2000, 1500, 500], [1500, 2000, 750], [500, 750, 2000]]
- capacity: Inventory capacities of the facilities.
- Default: [150, 300, 400]
- nfacility: The number of facilities.
- Default: 3
- stockoutflag:
- 0: all facilities satisfy the demand
1: at least one of the facilities did not satisfy the demand
- nstockout:
the number of facilities which cannot satisfy the demand
- ncut:
the amount of total demand which cannot be satisfied
This model is adapted from the article Rengarajan, T., & Morton, D.P. (2009). Estimating the Efficient Frontier of a Probabilistic Bicriteria Model. Proceedings of the 2009 Winter Simulation Conference. (https://www.informs-sim.org/wsc09papers/048.pdf)
Our goal is to minimize the total costs of installing capacity while keeping the probability of stocking out low.
The probability of failing to satisfy demand ξ = (ξ1, ..., ξm) is p(x) = P(ξ! < = x). Let epsilon ∈ [0, 1] be a risk-level parameter, then we obtain the probabilistic constraint:
P(ξ! < = x) ≤ epsilon
Meanwhile, the unit cost of installing facility i is ci, and hence the total cost is
- capacity
Minimize the (deterministic) total cost of installing capacity.
1 stochastic constraint: P(Stockout) < = epsilon. Box constraints: 0 < xi < infinity for all i.
- budget: Max # of replications for a solver to take.
- Default: 10000
- epsilon: Maximum allowed probability of stocking out.
- Default: 0.05
- installation_costs: Cost to install a unit of capacity at each facility
- Default: (1, 1, 1)
None
- capacity: (300, 300, 300)
- Each facility's capacity is Uniform(0, 300).
None
None
Our goal is to maximize the probability of not stocking out subject to a budget constraint on the total cost of installing capacity.
- capacity
Maximize the probability of not stocking out.
1 deterministic constraint: sum of facility capacity installation costs less than an installation budget. Box constraints: 0 < xi < infinity for all i.
- budget: Max # of replications for a solver to take.
- Default: 10000
- installation_costs: Cost to install a unit of capacity at each facility.
- Default: (1, 1, 1)
- installation_budget: Total budget for installation costs.
- Default: 500.0
None
- capacity: (100, 100, 100)
- Use acceptance rejection to generate capacity vectors uniformly from space of vectors summing to less than installation budget.
None
None