Intel Five-Machine Six Step Mini-Fab Description: Dr. Karl Kempf

0.0. MINI-FAB BLOCK DIAGRAM:

1.1. THE PRODUCTS AND TEST WAFER (volumes):

Note that four lots will be lost each week to emergency breakdowns at Mc and Md (see section 1.7).

1.2. THE PROCESS FLOW (steps and sequencing):

All production and test wafers follow this sequence subject to all machine restrictions such as batching and setups.

Test wafers have two further restrictions - two or more lots of test wafers may never be in the same batch, and a test lot may never run through the same machine twice, except at a unique machine that runs multiple steps. These test wafers run through the full process, require setups, can be included in batches, and are intended to broadly monitor equipment performance, hence the restrictions on batching and machine selection.

(An obvious variation here is that test wafers alternate step S1 and S5 machines (see below) every time they start, and alternate S2 and S4 machines (see below) every other time they arrive at step S2 - this would assure even coverage of all trajectories.)

1.3. THE EQUIPMENT SET (machines, batch sizes, set tip times):

This is a diffusion-like batching where the machine can run if three appropriate lots are available to load at the same time (as opposed to etch-like batching). When batching step S1, one can mix products and one test lots. For example, Pa/Pa/Pa, Pa/Pa/Pb, Pa/Pa/TW, Pa/Pb/TW, Pb/Pb/TW, Pb/Pb/Pa, Pb/Pb/Pb are acceptable. However, when batching step S5, one cannot mix products, bur one can mix one test lot. For example, Pa/Pa/Pa, Pa/Pa/TW, Pb/Pb/TW, Pb/Pb/Pb are acceptable, but Pa/Pa/Pb, Pa/Pb/TW, Pb/Pb/Pa are not acceptable. It is never acceptable to mix lots waiting for S1 and lots waiting for S5 into the same batch. (An obvious extension here is to be able to batch either 2 or 3, or, 3 or 4.)

This is a lithography-like setup where the machine can run as soon as a lot is available and it is likely that other lots will arrive before the first lot finishes processing (as opposed to an implant-like setup). At either step S3 or step S6, one can be running either product Pa or product Pa or a test wafer lot. Three different changes each require a setup. In one possibility, the lot type stays the same and the step changes (S3 to S6 or S6 to S3). In another possibility, the step stays the same and the lot type changes (among Pa, Pb, and TW). In the third possibility, both the step and the lot type change (for example, going from Pa on S6 to Pb on S3). Note that all setup times are symmetric (for example, going from Pa/S3 to TW/S3 or going from TW/S3 to Pa/S3 both take 5 minutes). Note also that all setups require a machine and an operator for the setup time. Finally note that a setup can only be done immediately prior to execution of the run that the setup is intended to enable.

1.4. PROCESS : EQUIPMENT MAPPING (with processing time):

Equipment preemption does not occur. Once a machine begins the execution of a step, it must complete before it starts any other activity. Note that the load and unload times, and setup times where applicable, are not included in these numbers - this is purely the run time of the process step.

1.5. PERSONNEL (including production operators and maintenance techs):

P01, P02

Two production operators are available for 540 minutes each per shift. Each gets two 60 minute breaks and one 60 minute meeting/training session per shift, and the two operators do not have to synchronize their off time in any way.

MT1

One maintenance tech is available for 600 minutes per shift. This tech gets two 45 minute breaks and one 30 minute meeting/training session per shift that do not have to synchronies with the off time of the operators.

Note that no preemption can occur with personnel. Once they begin a task, that task must complete before any other task can begin.

1.6. PROCESS : EQUIPMENT : OPERATOR MAPPING (with assist times):

Each machine run includes a load time which requires an operator at the beginning of a run and includes an unload time which requires an operator at the end of a run. An operator is not required at the machine between the end of the load time and the beginning of the unload time (the machine processing time given above). Within a given run of Mc or Md, the same operator does not have to perform both the load and the unload.

1.7. PROCESS : EQUIPMENT : TECHNICIAN MAPPING (with repair times) :

Emergency maintenance (EM) requests are surprises - that is, it is not know with much accuracy when they will begin, except that a machine must be running for the surprise breakdown to happen. The EM specified here happens randomly once each half week, with the assumption that the execution of the previous EM holds for roughly 2 shifts - hence, with 168 hours per week, a half week is 84 hours, two shifts is 24 hours, so every 54 hours + or - 30 hours (54 + 30 = 84, 54 - 30 = 24). But the repair, once started, takes between 6 and 8 hours or 420 min + or - 60 min. In order to keep the EM window within half a week with a worst case repair, this moves the 54 +/- 30 to 50 +/- 26 (50 + 26 + 8 = 84, 50 - 26 = 24)

Preventive maintenance (PM) requests are expected - there is a known window in which they must be started. After the window opens, operators might decide to run lots or technicians might decide to do the PM. (If a machine is in any stage of its running cycle when the window opens, the technician can not start the PM until the whole cycle is finished.) After the window closes, if the PM has not been done, operators can no longer run lots but the technician can still decide to do the PM. (If a machine is in any stage of its running cycle when the window closes, it can finish the whole cycle, but no new work can be started, and the technician can not start the PM until the cycle is finished.) Once the PM is started, it takes a known time to complete. In the examples given here, the PM window opens at the beginning of each shift or 6 hours after the previous PM whichever is later, and the PM must be completed by the end of the shift or at the beginning of each day or 12 hours after the previous PM whichever is later, and the PM must be completed by the end of the day.

1.8. CELL LAYOUT AND TRANSPORTATION (transport and buffer limits):

Note - when a buffer is full, the feeding step must not start to run more work
Note - only one lot can be in transport at any one time

Any load transaction (any buffer to transporter) : 1 min
Any unload transaction (transporter to any buffer) : 1 min

Transport from one cell to another is simply the sum of the relevant times
(ex. moving one lot from C1 to C3 = 1 min load + 4 min C1 > C2 + 4 min C2 > C3 + 1 min unload = 10 min)

1.9. THE CAPACITY ANALYSIS

Are there enough machines ? (not including setups)

Utilization / Availability = 1200 / 1365 = 87.9%

Utilization / Availability = 840 / 1080 = 77.8%

Utilization / Availability = 630 / 690 = 91.3%

Are there enough operators ? (without setups or operator transport times)

Capacity available OP1 = OP2 = 540 min/shift

Utilization / Availability OP1 = OP2 = 77.8%

Are there enough techs ? (without technician transport times)

Total Capacity required = 75 + 360 + 30 = 465 min/shift (= 64.6%)

Capacity availability 600 min/shift (= 83.3%)

Required Capacity / Available Capacity = 77.5 %

Is there enough transport ?

Utilization / availability 444 / 720 = 61.7 %

2.1 Lots, States, and State Transistions

As shown in the diagram of logical connections between possible states of lots, the first break is between busy and idle. Whenever a lot is busy, it is the focus of the transportation system (if busy:being_transported) or of a machine (if busy:being_run) or of a machine and an operator (if busy:being_loaded or busy:being_unloaded). Whenever a lot is idle, it is either waiting for as operator to load it for processing (if idle:waiting_to_load) or for an operator to unload it after processing (if idle:waiting_to_unload).

The following information completely defines the states of a lot.

As shown in the diagram of states and state transitions for lots, the following state transitions are possible.

2.2. Machines, States, and State Transitions

As shown in the diagram of logical connections between possible states of machines, the first break is between up and down. Whenever a machine is up, it is the focus of operators, and whenever it is down it is the focus of technicians. The two categories of up machines include full and empty. It is generally the case that operators focus on getting work to an up_empty machine, and work away from an up_full machine. The two categories of down machines include pm (preventive maintenance) and em (emergency maintenance). It is the case that technicians have an easier time with down_pm machines since the occurrence and duration are known, and have a harder time with down_em machines since the occurrence is, to some degree, a surprise and the duration is, to some degree, unknown.

The following information completely defines the states of a machine.

As shown in the diagram of states and state transitions to machines, the following state transitions are possible.

2.3. Operators, States, and State Transitions

As shown in the diagram of logical connections between possible states of operators, the first break is between up and down. Whenever an operator is up, the operator is on the factory floor. Whenever the operator is up and busy, the operator is doing something to tend the machines. If there are no machines which need tended, then the operator is up but idle. Whenever an operator is down, the operator is off of the factory floor and is therefore idle from the point of this production model.

The following information completely defines the states of an operator.

As shown in the diagram of states and state transitions for operators, the following state transitions are possible.

Technicians

Transporters

3.0 Decisions

There are a variety of decisions that the decision makers in the system need to be prepared to make as the simulation progresses on a minute to minute basis, including.....

3.1 The Operators

The operators must decide .....

- which machine/which step/which lot (in that order)

- when to break

3.2 The Technicians

The technicians must decide .....

- which machine/(maybe which maintenance, em or pm)

- when to break

3.3 The Transporter

The transporter must decide .....

- which lot

3.4 The Starter

Some entity must decide .....

- what to start at the begiinning of each shift

3.5 Murphy

Some entity must decide .....

- when to initiate an emergency maintanence

- how long the emergency maintenance should last

It is imaginable that the decision makers hold all of the logic to make all of these decisions. It is also imaginable that here is a chance to model an "automation" system - i.e. some decision support (or some automated decision making) - which the decision makers can use (or which force decisions on the active agents).

4.0 Approach

In our terminology, this is the high level control approach.

5.0 Metrics

Return to Dr. Rodriguez's Home Page

Email questions for Dr. Rodriguez