ScienceAsia 27 (2001)
PROPOSED DEMAND FORECASTING AND PRODUCTION PLANNING METHODS
The proposed demand forecasting and production planning methods are depicted in a step-by-step fashion in Fig. 1.
Most factories produce a variety of products that can be categorized into product groups or families. Individual products in the same product group generally have some common characteristics. For example, they may have the same demand pattern and a relatively stable product mix. As a result, it is possible to forecast the aggregate demand of the product group first, and then disaggregate it into the demand of individual products. Since the forecast of the aggregate demand is more accurate than that of the individual demand1, it is initially determined in Step 1. Then the demands of individual products are determined in Step 2 by multiplying the aggregate demand with the corresponding product mix that is normally known and quite constant. Since the demand forecasts are always subject to forecast errors, safety stocks are provided to avoid stock-out problems. Based on the standard deviation of the forecast errors and the required service level, the safety stocks for individual products are determined in Step 3.
Production planning decisions are so complicated and important that they should not be subjectively and intuitively made. Consequently, an appropriate production planning model should be formulated to determine the optimal decisions. With this model, its parameters, eg, demand forecasts, safety stocks, holding cost, overtime cost, machine capacity, inventory capacity, and available regular time and overtime, are entered or updated (Step 4). In step 5,
the optimal decisions regarding the production quantities, inventory levels, and regular production time and overtime for each product in each pro- duction stage are obtained by solving the production planning model. Step 6 indicates that only the optimal production plan of the current month will be implemented. After one month has elapsed, the demand forecasts and the production plan will be revised (by repeating Steps 1 to 5) according to a rolling horizon concept.
BACKGROUND OF THE CASE STUDY
The pressure container factory manufactures 15 products, ranging from 1.25 to 50 kg of the capacity of pressurized gas. The products are divided into eight product groups, namely, Group 1 to Group 8. The first six groups have only two components, “head” and “bottom”, while the last two groups have three components, “head”, “bottom”, and “body”. The production process can be divided into five stages as shown in Fig. 2. Stage 3 is only required to produce the products having three components (ie, those in Groups 7 and 8). Stage 4, the circumference welding, is found to be a bottleneck stage due to its long processing time.
Presently monthly demand forecasts are sub- jectively determined by the Marketing Department based on past sales and expected future market conditions. No systematic method is used in fore- casting. Using these forecasts and other constraints, such as availability of raw materials, equipment, and production capacity, the monthly production plan for a three-month period is intuitively determined without considering any cost factor. This results in inaccurate demand forecasts and, subsequently, an inefficient production plan.
1) Forecast the monthly demands of each product group throughout the planning horizon of 12 months
Stage 1 Blanking
2) Determine the demand for each individual product
3) Determine the safety stock for each individual product
Stage 2 Forming of bottom and head
Stage 3 Forming of body
4) Update the parameters in the production planning model
Stage 4 Circumference welding
5) Run the planning model to obtain the optimal planning dicisions
6) Roll the plan by repeating Steps 1 to 5 after one month has elapsed
Stage 5 Finishing
Fig 1. Proposed forecasting and planning steps.
Fig 2. The production process to manufacture a pressure container.