Title: Manufacturing Systems Modeling, Analysis and Design IME 452 IME 545 Chapter 3, Sections 3'1, 3'2
1Manufacturing SystemsModeling, Analysis and
DesignIME 452 / IME 545Chapter 3, Sections
3.1, 3.2
Amy Thompson Instructor
2Transfer Lines and General Serial Type Systems
- Transfer Line a paced, serial system subject to
breakdowns - Used for their high rates of throughput for
single, static products - Breakdown could be due to machine failure or
transport failure of system or unavailable
resources - Usually have common material handling and control
system, so usually capital intensive and
availability and effectiveness are crucial design
features
3Size of Line
- As the number of stations along a line increases,
the probability of all stations functioning
decreases. (Given that each station is not 100
reliable.) - Buffers, though expensive to install and
maintain, insulate stations from failures
elsewhere in the line. - For system design, want to determine
effectiveness of a line given buffer capacities,
failures, and repair rates
4Station Failures
- Buffered Station Operational States
- Station Failure
- Total Line Failure
- Station Blocked
- Station Starved
- Operational (cycle) Dependent Failures (_at_80)
occur only while the system is running, measured
in number of cycles between failures - Time Dependent Failures measured in time units
between failures
EmptyStarved
BlockedFull
X
5Some Typical Statistical Distributions Used in
Reliability
- Exponential
- Weibull
- Binomial
- Lognormal
6Probabilistic System Reliability
- Serial Components
- Parallel Components
- A Simple Common Cause Model
- Hybrid System
- Bridge Circuit
7System Maintainability and Supportability
- Maintenance
- Breakdown, Preventative, Productive, Total
Productive Maintenance - Availability (a measure of uptime)
8Overall Equipment Effectiveness
- Process rate ( a measure of the ability to
operate at a standard speed) - Quality rate (a measure of the ability to produce
to a standard product quality) - Equipment effectiveness (an overall measure of
the system effectiveness)
9Throughput and Productivity
- Throughput the average rate at which product
comes off the line - Effectiveness or Productivity a measure of
performance of a line configuration defined - E() operator is expected value
- Downtime period during which product doesnt
leave the line
10Productivity Evaluation
- Methods
- Discrete Event Simulation
- Analysis
- What we will cover
- Configuration and Productivity
- Paced Serial lines w/o buffers
- Two-stage paced lines w/buffers
- Unpaced Serial lines
11Configuration, Reliability and Productivity
Availability of all the machines is R
12Calculating System Productivity
- mi Production rate at state i
-
13Configuration Throughput Distribution
- Parallel systems have significantly higher
expected throughput than serial systems (given
the same machine availability).
14Productivity of Pure Serial and Pure Parallel
Lines
Normalizing the production rate to one, and
assuming the availability of each machine in the
system is the same, the productivity is
NOTE
15Paced Serial Lines w/o Buffers
- Operation Dependent Failures
- Assumptions
- Geometric distribution of failures
- Mean cycles to failure (MCTF) is and
is the failure rate - The average time to repair any station is 1/b
- All uptime and downtime variables are independent
- Idle stations do not fail
- Failures occur at the end of a cycle, dont
destroy product. - At most one station can fail on any cycle
16Paced Serial Lines w/o Buffers
- Let
- representing the probability that no machine
fails - Then with some manipulation, an m station line
behaves like a one station line with parameter ß
17Paced Serial Lines w/o Buffers
- Can approximate ?, and calculate productivity
- Time Dependent Failures exponential
distributions, failure rate ?, repair rate b,
availability
productivity
18Paced Serial Line Example
- Op dependent
- ??.02.05.03.02.12
- Exact
- Error 0.86
Average Repair time 2 cycles
19Paced Serial Line Example
mean time to failure
50
20
33 1/3
50
- Time dependent productivity
-
Average Repair Time 2 minutes
Difference from Operation Dependent 2.5
20Time-Dependent vs.Operation-Dependent Models
- Time not suspended for time-dependent failures,
so time-dependent models have a lower
effectiveness
21Assignment, Problem 1
- Operation Dependent Failure Model
- Create a 2 station series transfer line in Pro
Model. - Make both stations processing times equal to
some arbitrary value of C (cycle time), and use
no move times. - Use one repair person as a resource to perform
repairs. - The first station fails every 10 cycles and the
second station every 15 cycles. Repair time is 2
cycles for both stations. - Do not allow stations to fail on the same cycle
(dont allow station 2 to fail while station 1
has failed.) Assume a repair on station 1 and
station 2 cant occur at the same time. - Determine the line availability.
- Add a second repair person and allow simultaneous
failures. What is the new line availability? How
does adding a second repair person and allowing
simultaneous failures effect the calculation of
the expected availability?
22Assignment, Problem 2
- Time Dependent Failure Model
- Create a 2 station series transfer line in Pro
Model. - Make both stations processing times equal to
some arbitrary value of C (cycle time), and use
no move times. - Use one repair person as a resource to perform
repairs. - The first station fails every 10 cycles according
to time and the second station every 15 cycles
according to time. The station can not fail again
until it is repaired. (Time to failure measured
since station was last repaired.) Repair time is
2 cycles according to time for both stations. - Assume a repair on station 1 and station 2 cant
occur at the same time. - Determine the line availability.
- How does this line availability compare with that
of the operation-dependent model?
23Manufacturing SystemsModeling, Analysis and
DesignIME 452 / IME 545Chapter 3, Sections 3.3
Amy Thompson Instructor
24Parallel-Serial Production Lines
No Crossover
25Parallel-Serial Production Lines
Crossover
26Parallel-Serial Production Line Example
Width m3 (parallel lines), Length n2
(operations)
Let mttf 270 minutes, mttr 30 minutes The
availability R
Crossover
No Crossover
27Parallel-Serial Production Line with Crossover
Example
28Productivity Comparison Parallel-Serial
LineMachine Reliability 0.9
29Productivity Ratio Comparison Parallel-Serial
LineMachine Reliability 0.9
30Parallel-Serial Observations
- Synergistic Improvements to Productivity
- Lower Variability to Throughput
- Additional Parallel Lines produce Diminishing
Returns - More Value at Lower Machine Availability
- Productivity/ Capital Cost Trade-off vs. Buffers
31Two-stage paced lines w/buffer with
operation-dependent failures
- Describe model with a Markov Chain
- States of the Markov Chain are (S1,S2, z) where
S(i) is the station status for station i - Book uses W for operational and R for needs
repair. Other nomenclature uses A and B for
operational machines, respectively, and A and B
for machines needing repair, respectively. - Whether buffers are full are empty effect whether
machines become idle.
32The Buffer-Machine Interaction
- Basic Building Block
- Assumptions
- View System at Start of Cycle
- Failure and Repairs occur at end of cycle
- Part is delivered downstream at end of cycle
before failure or repair - If a station starts a cycle under repair, it will
not deliver a part downstream and its end. - If station is operational at start, will send a
part downstream, unless station is starved. - Z is the buffer capacity, x is the buffer level
- Buffer full Station A Blocked
- Buffer Empty Station B Starved
- If Both Stations Failed or Operational, Buffer
size unaffected - Both machines never fail at the same time
33The Buffer-Machine Interaction
- State Diagram (Internal Conditions) See Table 3.1
for transition states
B fails to repair
A repaired
A fails to repair
A,B Functional
B fails Buffer Full
A,B A Idle
A,B B Idle
A fails Buffer Empty
A,B Xx
B repaired
Buffer Empty
Buffer Empty
B fails to repair
A fails
A fails to repair
A,B Xx1
B fails
A repaired
A,B Xx-1
B repaired
A fails
AB fail to repair
A,B Xx
A repairs
B repairs
B fails
34Two-stage paced lines w/buffer with
operation-dependent failures
- Let S be the set of states of the system. Define
the steady-state balance equations by applying
the Chapman-Kolmogorov result - P(s) is the probability of being in state s and
p(u,v) is the transition probability for ending
in state v given that we began the cycle in state
u - We group the rows of Table 3.1 with similar
resultant states to obtain steady-state equations.
35Two-stage paced lines w/buffer with
operation-dependent failures
- Example state AB0 or (WW0)
- 3 possible ways to have reached this state, what
are they? - RW1
- WR0
- WW0
- All equations are shown on p. 75
36Two-stage paced lines w/buffer with
operation-dependent failures
- Using steady-state probability theory and the
fact that states are mutually exclusive and
exhaustive, these sets of equations can be
reduced to the following Productivity
equation(eq. 3-9) - Closed Form Solution for non-simultaneous station
failures (see eq. 3-10) by Buzacott. - Limit theory on increasing buffer size as a
buffer capacity is increased, asymptotic
effectiveness approaches the capacity of the
least effective station.
37Deterministic Failure Repair Time and Buffers
- Time spent waiting for repair is when buffers get
filled or emptied - Rule of Thumb
- Buffers should be large enough to accommodate at
least the average repair time - See Figure 3.4 on page 79.
38Longer Line with a Single Buffer
- System Reduction
- Rule 1 A set of stations can be aggregated into
a single, virtual station, using equation 3.4,
assuming all stations have a common repair rate,
b, and all stations must stop if any individual
station fails (no buffers between stations). - Rule 2 Median Buffer Location if only one
buffer is placed, place buffer in middle of line - This is w.r.t. the failure rates up and
downstream of the buffer - Rule 3 Reversibility If direction is
reversed, production rate stays the same. (Dont
have to investigate a reverse design, gives same
effectiveness.)
39Single Buffer Location
40Manufacturing SystemsModeling, Analysis and
DesignIME 452 / IME 545Chapter 3, Sections 3.3
Amy Thompson Instructor
41Time vs. Operation Dependent Errors
- Time failures occur independent of the number of
cycles since the last failure. - Operation dependent errors only occur when the
station is running. - Most failures (80) are operation dependent
errors.
42Compare Time and Operation Dependent Models
- Specify in Downtime/Locations in Pro Model
- Specify frequency between downtimes in terms of
cycle time (c) - Specify to use the operator for 2 cycles in terms
of c - Operation Dependent Model in Pro Model
- Use Usage function in Downtime/Locations
- Time Dependent Model in Pro Model
- Use Clock function in Downtime/Locations
43Operation Dependent Model 2-Machine Example
- Operation Dependent
- Total Throughput Time for 1000 parts 126,600
using usage function - Resource used 166 times
- Time Dependent
- Total Throughput Time for 1000 parts 149,800
using clock function - Resource used 248 times
44Buffer Placement Example
- Average Repair Duration 16 cycles
- Total Failure Rate
- Median Rule
- Place Buffer after 2nd Station
- How Big? Rule of Thumb Accommodate average
repair time say 16 pieces
45Buffer Placement Example
16
.003
.004
.005
.003
.002
46Assignment
- Redo this problem, placing the buffer after
station 3. Does availability improve?
47Manufacturing SystemsModeling, Analysis and
DesignIME 452 / IME 545Chapter 3, Sections 3.5
Amy Thompson Instructor
48Unpaced Serial LineVariation in Production Rate
- Even though average workload may be balanced,
variation in processing time can occur. - Throughput with no buffers or breakdowns
49Unpaced Serial LineVariation in Production Rate
- If CV1.0, then if production rate is 0.1, actual
throughput is 0.10.5 0.05 units completed per
minute, 3 per hour - If CV0.1, then if production rate is 0.1, actual
throughput is 0.10.9 0.09 units completed per
minute, 5.4 per hour -
50Unpaced Serial LineVariation in Production Rate
- What effect does adding stations to this type of
line have? - What effect does lowering coefficient of
variation have on this type of line? - What happens when you lower the average time it
takes to produce a part, without lowering the
standard deviation of time for producing the
part? -
51Unpaced Serial LinesVariation in Production Rate
- If you place buffers of the same size between
successive stations, assume the first station is
never starved, and the first buffer tends to be
full. - Since the last workstation is never blocked, the
last buffer will tend to be empty. - In general, buffer utilization decreases from the
front to the rear of the line, with the middle
buffer being about ½ full on the average. - Throughput is dependent upon the ratio of the
buffer capacity (Z) to processing time
coefficient of variation (cv).
52Unpaced Serial LinesVariation in Production Rate
- For Z/cv 10, 80 of capacity lost because of
variability of processing times is recovered. - For Z/cv 20, 90 of capacity lost because of
variability of processing times is recovered. (2
times larger buffer gives only 10 increase in
capacity) - The capacity lost in the unbuffered line that is
recovered by adding buffers is only marginally
dependent on line length.
53Unpaced Serial LinesVariation in Production Rate
- The capacity lost in the unbuffered line that is
recovered by adding buffers is only marginally
dependent on line length. - Variation in processing time with equal buffers
and no breakdowns
54Unpaced Serial Lines Variation in Production
Rate
- Blumenfeld Equation (approximation) (eq. 3-18)
T is mean service time at each station
55Buffers for Unpaced Lines
- For lines with identical workstations, the best
allocation of buffers is to have many buffers of
nearly equal size between stations vs. one buffer
in the middle of the line. - The largest buffers should be in the middle of
the line, but the difference in size should be no
larger than one slot. - Buffer size should be symmetrical around line.
- As stations become non-identical, the less
reliable stations should have larger input and
output buffers. - Buffers are less useful in unbalanced lines.
- If breakdowns or high processing time variability
occurs at a workstation other than the
bottleneck, input and output buffers at the
bottleneck are more important.