清华大学-生产计划与控制：PPC-L10A-Optimal-Lot-Sizing.pdfVIP

Ye Cheng Dept. of Industrial Engineering Tsinghua University 1 Optimal Lot Sizing for Time-Varying Demand  Compute true optimal lot sizes  Minimize the total cost of holding and setup cost over the planning horizon Casting the problem as shortest-path problem Dynamic programming  Assumption  Forecasted demands over the next n periods are known r = (r , r , ……, r ) 1 2 n  Holding cost $h per unit per period Charged against ending inventory of each period $K per setup Example 7A.1  A local semiconductor fabrication shop of Hi-Tech Co.  Forecasted demands over the next four weeks: 52, 87, 23, 56  Only on setup each week  No backordering  Enough capacity  Consider the total number of feasible production policies  y , …, y , the order quantities in each week 1 4  y 1 ≥ 52  No stock out in week 1  y 1 ≤ 218 Zero ending inventory in week 4  So, y 1 can take any one of 167 values  y 1 + y 2 ≥ 52 + 87 = 139  No stock out in week 2  If y 1 ≤ 139, then 139 - y 1 ≤ y 2 ≤ 218 - y 1  If y 1 139, then 0 ≤ y 2 ≤ 218 - y 1  10200 different values of just the pair ( y , y ) 1 2  Enormous number of feasible solutions for even moderate sized problems Wagner-Whitin Algorithm  Reduce the number of policies to be considered as candidates for optimality  Based on the observation An optimal policy is an exact requirements policy Each value of y is exactly the sum of a set of future demands y = r , or y = r +