• 沒有找到結果。

A Customized DBR for TFT-LCD Cell plant

Scheduling and Control of TFT-LCD Cell plants Using Drum-Buffer-Rope Systems

4. A Customized DBR for TFT-LCD Cell plant

The bottleneck or CCR in the Cell plant is identified as PI Printing station based on the discussion with the personnel in different Cell plants. According to the CCR station, the DBR system is applied to the Cell plant as follows.

4.1 Buffer Determination

Based on the PI Printing (CCR) step, the CCR buffer covers the 1st Scribe & Grinding step and the shipping buffer covers from assembly to Test steps. The size of the buffer are determined by two procedures: (1) evaluating initial buffer size based on the buffer evaluated formulas recommended by Guide(1996) and (2) tuning it based on the current CT in Cell plant.

Step 1: Evaluating initial buffer size

Schragenheim and Ronen (1990) first recommended the size of the buffer in a stable environment is normally defined as three times the average lead time to the constraint from

buffer size was determined by a constant which is then multiplied by the cumulative processing time as shown as Eq(1):

=

×

= m

j j

i MULT PT

BS

1

(1) Where

BSi = buffer size for CCR buffer (i=ccr) and Shipping buffer (i=shipping) PTj=processing time for operations covered by CCR buffer or Shipping buffer

MULT=constant buffer size multiplier which was recommended by Schragenheim and Ronen (1990) as three.

Therefore, based on Eq.(1) and the processing information shown in Fig. 1, the CCR buffer and Shipping buffer for a lot are evaluated as follows:

BSccr=3×12=36(mins)≒1(hr)

BSshipping=3×(12+12+300+240+480+120+15+12+2)

=3579(mins)≒3600=60(hrs)

Step 2: Tuning the initial buffer size based on the current CT in Cell plant.

The current manufacturing CT of TFT-LCD Cell process is about 3-5 days or average 4 days.

Because DBR restricts unlimited release of material into the system so as to prevent the growth of inventory, which would increase in the CT, the CT under DBR system will be lower than the average CT of 96 hrs. However, the CT of a lot is only about 60 hrs based on the initial buffer evaluated in step 1. In other word, the initial buffer size of BSccr and BSshipping is some small. Therefore, the initial buffer size of BSccr and BSshipping are refined as 4 hrs and 64 hrs respectively. The initial buffer size can be further made adjustments based on Buffer Management approach over time until optimal buffer sizes are identified (Goldratt and Fox 1986b, Schragenhem and Ronen 1991).

4.2 Drum Design

A daily production plan with 7-day planning horizon is assumed to be provided by the planner once a week as shown in Table 1. A customized drum design process is required to design an effective drum for Cell plant due to its specific domain, such as parallel machines and TFT glass and CF glass paired into assembly station. The procedure of the customized drum process is first to calculate ruins for each batch in daily production plan, allocate then these batches into machines, finally translate these batches into TFT and CF lots and sequence them by three lots of TFT glasses and then three lots of CF glasses.

Step 1: Generating the ruins

The ruins are the ideal production schedules for all batches at PI station that do not consider the feasibility of the production capacity at the PI station. The ideal completed time at the PI station for each batch is obtained by subtracting the shipping buffer time from its required time.

For example, the required time of 114 batch size of 14.1” is 4/1 24:00, the completed time for this batch at PI station is then 3/30 8:00(=4/1 24:00 – 64 hrs). The final ruins for batches in date 4/1 and 4/2 in Table 1 are shown in Table 2.

Step 2: Allocating batches in daily production plan into PI machines

The purposes of allocating batches in daily production plan into PI machines are to balance

these PI machines and to guarantee these batches can be completed before their completed times in ruins. The detailed steps are as follows:

Step 2.1: Sorting each batch in the daily production plan.

The batches in daily production plan are sorted by ascending its completed time in ruins. If two batches have equal completed time in ruins, the larger batch (that requires more processing time in PI machine) is assigned higher priority. The 1st priority batch is named as an allocating batch. For example, the allocating batch among ruins in Table 2 is the batch of 14.1” panel in 4/1.

Table 2. The ruins for batches in date 4/1 & 4/2 in Table 1.

Current time(Tnow): 3/29 0:00

Batches Batch Size Start time at ruins Completed time at ruins 14.1” for 4/1 114 3/28 10:24 3/30 08:00

15” for 4/1 72 3/29 03:12 3/30 08:00 17” for 4/1 27 3/29 21:12 3/30 08:00 14.1” for 4/2 117 3/29 09:12 3/31 08:00

15” for 4/2 75 3/30 02:00 3/31 08:00 17” for 4/2 30 3/30 20:00 3/31 08:00 Step 2.2: Selecting a highest priority PI machine.

Based on the allocating batch, next step is to select a best feasible PI machine for allocation of this batch. The purpose of this selection is to satisfy the scheduling goal that is first to meet the requirement of daily production plan and then to reduce the setup time, which then increase throughput. The machine selection procedures are as follows:

Step 2.2.1: Calculating first the available capacity (AC) for each PI machine.

The AC of a machine equals the completed time of the allocating batch in ruins minus the available time (AT) of a PI machine. The AT of a machine is defined as the time the latest completion time of batches allocated in this machine. However, the initial AT depends on the initial WIP in a PI machine. For example, an initial WIP for four PI machines are given in the first three rows in Table 3. Based on the WIPs in Table 3, the initial AT for the 1st PI machine is 3/29 8:00 which equals to the summation of current time(3/29 0:00) and the processing time of 40 lots of 14.1” panel in PI machine(i.e., 40×12 mins). The initial AT for the other PI machines are shown in the 4th row in Table 3.

Step 2.2.2: Selecting the candidate machines.

The candidate machines are first defined as the machines whose AC are more than processing time of three lots (i.e., 72 mins). If candidate machines do not exist, then all machines are candidate machines. For example, based on AT in Table 3 and the completed time of allocating batch in Table 2, the AC for four PI machines are 24 hrs, 26hrs, 24hrs and 12 mins, and 23 hrs and 36 mins, respectively. Therefore, these four machines are all the candidate machines.

Step 2.2.3:Revising candidate machines.

If none of the latest batch allocated in these candidate machines are same as the type of the allocating batch, go to Step 2.2.4; otherwise the candidate machines is revised as these machines whose latest allocated batch are same as the type of this allocating batch. For example, because latest batch allocated in PI machine #1 and #2 are same as the type of the allocating batch as shown in Table 3, the candidate machines are revised as PI machine #1 and

#2

.

Step 2.2.2: Selecting the allocating machine.

An allocating machine is defined as a machine with earliest AT in candidate machines. For example, because the AT of machine #2 is earlier than that of machine #1 as shown in Table 3, the machine #2 is selected as the allocating machine.

Table 3. An illustration of the initial WIP in four PI machines.

Current time(Tnow): 3/29 0:00

PI Machine # 1 2 3 4

WIP Item 14.1” 14.1” 15” 17”

WIP Qty 40 30 36 42

Initial Available Time 3/29 08:00 3/29 06:00 3/29 07:12 3/29 08:24 Step 2.3: Allocating the allocating batch into the allocating machine.

Because TFT glass and CF glass must be processed by PI station respectively, the process batch size in PI station is double size of the batch in the daily production plan. For example, the batch size of the allocating batch (14.1” panel in 4/1) is 114, therefore the processing time of this batch is 2736 mins (=114 × 2 × 12mins). When the allocating batch is allocated into the allocating machine, the completed time of this allocating batch in the allocating machine can be calculated as the summation of AT of this allocating machine and the processing time of this allocating batch. However, if the type of this allocating batch is different to the type of the latest batch in the allocating machine, a setup time (120 mins) is required. For example, the completed time of the allocating batch (14.1” panel in 4/1) is 3/31 3:36 which equals 3/29 06:00 (the AT of the PI machine #2) + 2736(mins).

相關文件