The chapter first describes the simulation environment and then detail dresses the experiment models for both PPP method and MSM method.
5.1 Simulation Environment
To convey our idea in estimating lot transport times in 300mm AMHS, simulation experiments are conducted based on realistic data from a local 300mm fab.
Our simulation models are implemented with the discrete-event simulation package -- eM-PlantTM from Tecnomatix Technologies Ltd. All the experiments are executed in a Pentium-III personal computer with Microsoft Windows XP. The eM-PlantTM is an object-oriented simulation system with characteristics of hierarchy, inheritance, and concurrency. There are some built-in objects in there for easy development.
Users can easily modify them into user-defined objects for their specific purposes.
Some of the objects defined in our simulation models are depicted in Table 3.
The Chapter 5.2 and 5.3 have some same assumptions. The running speed of an OHT is set to 2 meters per second. The time for each loading/unloading operation of a carrier is 16 seconds, respectively. Since the acceleration and deceleration of each OHT operation are relatively small, they are thus neglected.
Because the reliability is not our focus on this study, we assume that there are no failures and maintenance activities on all the entities during the simulation horizon.
Since we are interested in the effects on the performance of the OHT system, the from-to relationship between two processing tools is adopted, instead of considering the whole process flow of a semiconductor product. The inter-arrival time of transport events is probabilistic and is assumed to be of exponential distribution.
For simplifying the simulation model, stockers of infinite capacity are assumed.
No matter the capacity of stocks, the stocks perform an inter-medium of loops and the hot lots only pass through them.
Table 3. Objects in the eM-Plant Simulation Model
Items Functions Defined Object Name
Event Controller Start system
EventControler Source Start of line
Sources Stocker output port End of line
Stk1Out, Stk2Out
Frame Loop structure
Loop1, loop2, loop3 Special control method Execution of special actions
IO, and so on Raw spec. data & output Table for spec. data & record output
Performance, and so on Products Entity of products
Normal lot,
Hot lot Loadport Basic units of loadport
EQ011, EQ012 ~ EQ223 Stocker Input port Stock in a lot
Stk1In, Stk2In Delivery time record Record deliver time
Deliver trend
OHT Deliver lot
OHT Track Track
Track
5.2 PPP Simulation Model
In the simulation models, we consider an OHT loop in 79.4 meters long, where there are two stockers and 23 pieces of equipment. In order to study the impact due to the hot lots rule, a non-differentiated rule, Nearest Job First (NJF) rule, is adopted for the comparison. The NJF rule utilizes the straightforward idea of first meet, first serve and it has been suggested as a good dispatching rule in many AGV applications [21], [47].
Observing the dynamics of the OHT system, in the experiment design, we
consider three dominating control variables -- loading ratio (ρ), population of hot lots (Ω), and the number of OHTs (υ) in the loop. Systems with heavy loadings are adopted to highlight the effect of the hot lot rule in resource contention. Two loading ratios, 100% and 90% of the design specification, are used in the simulation.
As the increasing hot lots population will impose long time delays on the normal lots drastically, two distributions of hot lots, 2% and 8%, are designed for the tests. As the number of OHTs increase, system performance usually gets improved due to the increased resources. In the simulation study, we consider two configurations of the OHT numbers, 4 and 6 OHTs in the loop, respectively. Eight simulation experiments are then conducted based on the scenarios for these three control variables. The OHT delivery time, the time from the birth of a job to its completion, is considered as the performance measure. We design each experiment is simulated for three times. The total number of simulation experiments performed is 2 (hot lots ratio) x 2 (bay loading) x 2 (OHT number) x 3 (replication)
= 24. Because the delivery time for one lot is less than 2 minutes in loop, the simulation horizon is set to two weeks long with pre-run in 6 hours for each experiment. Figure 7 depicts the eM-PlantTM simulation model of PPP.
Figure 7. eM-PlantTM Simulation Model of PPP
5.3 MSM Simulation Model
The only performance measure in our simulation model is the lot delivery time.
Inputs to the simulation system include loop loading, percentage of hot lots and
number of vehicles in each loop. Without loss of generality, we build an OHT system with three loops as the control scenario, which represents the tool-to-tool transportation through several loops. Assume that all the lots in the simulation start from loop 1, and transit to loop 2, and then move to loop 3, and finally leave the system. Figure 8 shows the conceptual simulation model. The simulation horizon is set to two weeks long with time units in seconds after a warm-up of 6 hours.
Figure 8. A 3-OHT-Loops Model
Figure 9 demonstrates the simulation model of loop 1. In our simulation, each subsystem has the similar structure, but with different parameter settings for process tools. Each OHT loop is in 79.4 meters long, where there are two stockers and 23 pieces of equipment. All loops are designated with the same tool configurations.
Loops 1 and 2 have the same processing capacity of 97.2 lots per hours, and 94.2 lots for loop 3 per hours. The loop switching time is set to 16 seconds. Here, we adopt the PPP policy as the OHT dispatching rule for all loops in our simulation experiments. The PPP policy dispatched an empty OHT to a job with the highest priority. The dispatched OHT is reserved to the job once after it is dispatched and becomes empty again after completing this job. Our objective of OHT dispatching is to minimize the carrier delivery times. For each OHT loop, the OHT dispatching rule deployed for this loop remains unchanged.
Figure 9. OHT Loop Simulation Model
The models of each individual OHT loop are simulated for various combinations of OHT vehicles, loop loading, and percentage of hot lots to collect the statistics of waiting time and blocking time for each loop. Seven loop loading ratios (ρ), 90%, 92.5%, 95%, 97.5%, 100%, 102.5%, and 105% of the design specifications, are used in the simulation. Five configurations of hot lot percentage (Ω), 2%, 4%, 6%, 8% and 10%, are designed for the hot lot population tests. In the simulation study, we consider three configurations of the number of OHT vehicles (υ), 3, 4 and 5 OHTs in the loop. One hundred and five simulation experiments (ρ:7, Ω:5, υ:3, 7*5*3=105) are then conducted based on the scenarios for these three control factors.