Page 153 - Data Science Algorithms in a Week

Simulation Optimization Using a Hybrid Scheme … 137

x  (t)= f (x (t ), p ) (This notation represents the SD model equations)
x (Vector with initial values of all state variables)
T
0
0.5 ≤ Desired Days Supply of Parts Inventory ≤ 5
0.5 ≤ Time to Correct Parts Inventory ≤ 5
0.5 ≤ Preforms Cycle Time ≤ 3
0.5 ≤ Presses Cycle Time ≤ 3
0.5 ≤ Time to Correct Inventory ≤ 5
0.5 ≤ Supplier Delivery Delay ≤ 5
0.5 ≤ Time to Adjust Labor ≤ 5
0.5 ≤ Labor Recruiting Delay ≤ 5
5000 ≤ a1 ≤ 50000
5000 ≤ a2 ≤ 50000
1000 ≤ a3 ≤ 50000
10 ≤ a4 ≤ 100

Stabilization Policy

The stabilization policy is obtained after solving the optimization problem presented
in the previous section. The optimization algorithm was run at time 0 using the following
settings: swarm size = 30 particles, neighborhood size = 3 particles, initial inertia
weight = 0.5, iteration lag = 5, cognitive coefficient = 1.2, social coefficient = 1.2. The
time to obtain the optimal policy (after 150 PSO iterations and 1,243 PHC iterations) was
89 seconds.

9,300 Units
22,000 Units
10,000 Units
60 People
8,000

6,000 Units
10,000 Units
2,500 Units
20 People
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Time (week)

Preforms WIP Level Units
Presses WIP Level Units
Finished Goods Inventory Units
Labor People
ADE
Figure 5. Behavior of variables of interest for the stabilization policy.