MaxiMax: best of best = 80 Facility Layout

Decision Theory MaxiMin: best of worst= 50

Lapalace: best row avg= 55
MiniMax Regret: highest # for each column. Subtract LP: Facility Layout Big M:
other # in that column from it. Find max
- for each row then Set: i= departments (A to E); k = locations (1 to 6) “if this happens, then this happens”
choose minimum from it. SUM
Variables: 𝑋 ≥ 𝑃𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 ∗ 𝑌 ;
Avg. Regret: regret x risk % 𝑦 1 if department is assigned, 0 if not 𝑦 can only be 1 if 𝑥 is at least equal to the
EVPI = Certainty – Risk 𝑋 total distance travelled parameter (could be anything.)
Certainty: Highest of each column x probability Objective Function: Min Sum(𝑋 ) x ≤ M ∗ y ; we can produce up to 0 if 𝑦 is 0
Risk: A1xPa+B1xPb (for all rows – look for biggest value) Constraints:
Assign each department once: and up to M if 𝑦 is 1.
Sum 𝑦 =1, …, Sum 𝑦 =1 (For all departments)
Sensitivity Analysis: Make sure each location is used maximum once
transform rows into equations: Sum𝑦 ≤ 1, …, Sum𝑦 ≤ 1 (For all locations)
Graph the scenarios and set the two optimal outcomes Distance (example A2)
equal: 𝑋 + 𝑀 ∗ (1 − 𝑦 ) ≥ 𝑇𝑟𝑖𝑝𝑠𝐴𝑡𝑜𝐵 ∗ (5𝑦 + 5𝑦 +. . +60𝑦 ) +
Ex. (Do N.) = 50+(60-50)x etc. . . +𝑇𝑟𝑖𝑝𝑠𝐴𝑡𝑜𝐸 ∗ (5𝑦 + 5𝑦 +. . +60𝑦 )
Aggregated Planning Solve for X and multiply with the Point Estimate (Given in M= Sum of largest amounts of trips x furthest distances from baseline department to other
Exam) departments
Answer: for value < PE*x we choose scenario 1, for value
> PE*x we choose scenario 2
Chase Plan Level Plan Mixed Model

In the chase plan one tries to exactly match the forecast demand. In the level plan it is important that regular production is constant. LP: Mixed Model
Inventory/backlog are not used. Instead, overtime/subcontracting are used. Fluctuation in demand is covered using inventory and backlog. Set: i = months (1 to 6)
Decision Variables:
𝑋 = # of units produced regular
LP: Chase Plan LP: Chase Plan 𝑋 = # of units produced using overtime
Set: i= months (1 to 6) Set: i= months (1 to 6) 𝑋 = # of units using subcontractor
Variables: Variables: 𝑋 = # of units kept in ending inventory
𝑋 = # of units produced regular 𝑋 = # of units produced 𝑋 = # of backorders
𝑋 = # of units produced using overtime 𝑋 = # of units kept in ending Inventory Objective Function: Min (30 ∗ (𝑋 +. . +𝑋 ) + 40 ∗ (𝑋 +. . +𝑋 ) +
𝑋 = # of units produced using subcontractor Objective Function: Min (30 ∗ (𝑋 +. . +𝑋 ) + 10 ∗ ( + 50 ∗ (𝑋 +. . +𝑋 ) + (70 ∗ (𝑋 +. . +𝑋 ) + 10 ∗ ( +
Objective Function: Min (30 ∗ (𝑋 +. . +𝑋 ) + 40 ∗ (𝑋 +. . +𝑋 ) + +. . + )
50 ∗ (𝑋 +. . +𝑋 )) +. . + )
Constraints: Constraints: Constraints:
𝑋 ≤ 4500 ,…, 𝑋 ≤ 4500 (Capacity Constraints) 𝑋 = 𝑋 ,.., 𝑋 = 𝑋 (Constant production Constraints) 𝑋 ≤ 4500, 𝑋 ≤ 5000, 𝑋 ≤ 3000 for i=1-6 (Capacity Constraints)
𝑋 + 𝑋 + 𝑋 = 4000, …, 𝑋 + 𝑋 + 𝑋 = 4000 (Flow Constraints) 𝑋 ≤ 4500 ,…, 𝑋 ≤ 4500 (Capacity Constraints) 𝑋 +𝑋 +𝑋 +𝑋 = 8000 + 𝑋
All variables are non-negative integers (Domain Constraints) 𝑋 = 4000+𝑋 , 𝑋 + 𝑋 = 4000 + 𝑋 ,.., 𝑋 + 𝑋 = 4000 + 𝑋 𝑋 +𝑋 +𝑋 +𝑋 +𝑋 = 7000 + 𝑋 + 𝑋
(Flow Constraints) ..
All variables are non-negative integers (Domain Constraints) 𝑋 + 𝑋 + 𝑋 + 𝑋 = 8000 + 𝑋 (Balance Constraints)
All variables are non-negative integers (Domain Constraints)
Location Planning
Center of Gravity Optimal Location
LP: Location Planning
Variables: 𝑋=
( ∗ )
;𝑌=
( ∗ ) How to check for better Location other than optimal location:
(𝑋 ): 𝑤𝑒𝑖𝑔𝑡ℎ ∗ |𝑋 − 𝑋 | + 𝑤𝑒𝑖𝑔ℎ𝑡 ∗ |𝑋 − 𝑋 | 𝑢𝑠𝑤.
𝑋 = # of of orders transported from production facility [i] (1 to 4) to the
Calculate for all X and Y & choose the coordinates with the lowest values as the better solution.
retail locations [j] (A to C)
𝑌 = (1 if new location (3 and 4) is opened, 0 otherwise) W
Objective Function: Min (0.5𝑋 + 1.1𝑋 + 0.7𝑋 + ⋯ + 0.5𝑋 +
300𝑌 + 1000𝑌 )
Constraints:
LP: Center of Gravity (Example Range [5,4] to [7,7])
Variables: * y=5=x 12
=

One binary (opened or not) for every possible center’s location (12 in total)
𝑋 + 𝑋 + 𝑋 + 𝑋 ≥ 450(Demand Constraints) 𝑦 , 𝑦 , 𝑦 , . . 𝑦 , 𝑦 (𝑒. 𝑔. 54 𝑠𝑡𝑎𝑛𝑑𝑠 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝑋 = 5 𝑎𝑛𝑑 𝑌 = 4)
𝑋 + 𝑋 + 𝑋 + 𝑋 ≥ 600 One binary for every possible sending path (48 in total) (store options production centus)
-> x

𝑋 + 𝑋 + 𝑋 + 𝑋 ≥ 700 𝑦 (𝑚𝑒𝑎𝑛𝑠 𝑠𝑒𝑛𝑑𝑖𝑛𝑔 𝑓𝑟𝑜𝑚 [7,4] 𝑡𝑜 [5,4]

54 , 𝑦 5 , 𝑦5 d , . . 𝑦 ,𝑦
𝑋 + 𝑋 + 𝑋 ≤ 1500 (Supply Constraints) Objective Function: Min (𝑦 ∗ 2 ∗ 13 + 𝑦 5 ∗ 3 ∗ 13+. . +𝑦 ∗ 5 ∗ 24)
𝑋 + 𝑋 + 𝑋 ≤ 1500 Constraints:
𝑋 + 𝑋 + 𝑋 ≤ 300 ∗ 𝑌 Sum of 𝑦 = 2 (Assign two centers)
𝑋 + 𝑋 + 𝑋 ≤ 1500 ∗ 𝑌
Sum of 𝑦 = 1 (Assign each station to one center)
𝑋 𝑎𝑟𝑒 𝑛𝑜𝑛 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠 (Domain Constraints)
Sum of 𝑦 =1
𝑌 𝑎𝑟𝑒 𝑏𝑖𝑛𝑎𝑟𝑖𝑒𝑠 Predecessor Constraints:
Sum of 𝑦 =1
𝑋 ≥ 𝑜 (𝑟𝑒𝑑𝑢𝑛𝑑𝑎𝑛𝑡)
Sum of 𝑦 =1 𝑋 ≥ 𝑋 + 𝐷 − 1𝑦 − 1𝑦 → 𝑤𝑒 𝑐𝑎𝑛 𝑐𝑟𝑎𝑠ℎ 𝑡𝑤𝑖𝑐𝑒
𝑦 ≤ 𝑦 (Make sure the center to which the stations sent waste is opened) …
… 𝑋 ≥ 𝑋 + 𝐷 − 1𝑦 (if 𝑿𝑱 has multiple predecessors)
𝑦 ≤𝑦 𝑋 ≥ 𝑋 + 𝐷 − 1𝑦
Project Management …
Crash 2 can only happen after Crash 1
𝑦 ≥𝑦
Set up Network Diagram: LP: Project Management ...
Forward Method: Earliest Finish= Earliest Start + Activities Duration Variables: Late is longer than X weeks
Backward Method: Latest Finish – Activities Duration 15 binaries for crashing (1 if crashed, 0 if not) 𝑋 ≥𝑋 −𝑋
Slack = (LS-ES) or (LF-EF) Activities with 0 Slack are critical!! 𝑦 ,𝑦 ,..,𝑦 𝑋 , 𝑋 , . . , 𝑋 𝑎𝑛𝑑 𝑋 𝑎𝑛𝑑
Critical Path: Longest Duration – Determines Project Duration (no 𝑋 𝑎𝑟𝑒 𝑛𝑜𝑛 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠 (Domain Constraints)
10 variables for the starting times of the activities
slack on the critical path -LS&ES are the same) 𝑦 , 𝑦 , . . , 𝑦 𝑎𝑟𝑒 𝑏𝑖𝑛𝑎𝑟𝑖𝑒𝑠
Crashing: Reduction of duration of an activity
𝑋 , 𝑋 , . . , 𝑋 𝑎𝑛𝑑 𝑋 T
1 variable measuring the lateness of the project
Only crash activities on the CP! Check after each crash if the CP has 𝑋
changed (if other path has a longer duration)
S Objective Function: Min (550𝑦 + 560𝑦 + ⋯ + 1500𝑦 + 525𝑋 )

Scheduling

Johnson’s Rule: -> use to minimise completion time w/ 2 workstations Hungarian Method: > LP: Hungarian Method
Create Sequence by ordering by lowest processing time: Step 1: Row min. subtracted from each value in row (for all rows) Variables: (all variables are binaries)
If on Machine A: Job is placed at beginning. Step 2: (new table) Column min subtracted from each value in Binary taking 1 (0 if not) if the worker I (A to E) is allocated to the job
If on Machine B: Job is placed at end (repeat until all jobs are allocated) column (for all) (1 to 5)
Jobs are always allocated for both Machines at once! (If there is a 0 in the column you don’t have to do anything as 0 is 𝑦
When mapping out the times: Jobs on Machine B can only start after the same your min) Objective Function: Min (31𝑦 +. . . +23𝑦 )
Job on Machine A is done! Step 3: Cross out all 0 with the lowest possible number of lines. If Constraints:
Idle Time: If Machine B could potentially start but Machine A is not done yet the number of lines is equal to the number of jobs, you are done Sum of 𝑦 = 1, ..., Sum of 𝑦 = 1 (assign each worker once)
Job Splitting: Look at the Idle times in the middle of Machine B. Try to split the and can allocate the workers. If a worker has a 0 for a job it
Sum of 𝑦 = 1, ..., Sum of 𝑦 = 1 (assign each job once)
previous job on Machine A that is causing the Idle time on Machine B in half means that he is most efficient for that job.
All variables are binaries (Domain Constraints)
and redo the Mapping and see whether Idle time has been reduced. If less lines than jobs: find the min among the numbers that are
If you don’t want to assign a worker for a certain job:
Idle time and total makespan are correlated as if one goes up the other not covered by a line, subtract min. from uncovered numbers and
𝑦 = 0 (for whichever I and j)
one goes up as well and vice versa. add to numbers that are covered by two lines (crossings points)

Notations in Testvision General LP Key Knowledge

Sets: Objective Objective

There is no specific requirement for how to write down a set, as long as it is a
proper definition that specifies:
1. Which index you use Final Value: the value the variable takes in the optimal solution. Final Value: the value the constrained resource takes in the
2. Which values it can take Objective Coefficient: the coefficient multiplying the variable to optimal solution.
3. What it refers to reach the objective function value. Constraint R.H.: the value constraining the final value.
For example: I = 1, 2, 3 represents the suppliers 1, 2 and 3; Reduced Cost: the amount by which the objective coefficient Shadow price: the amount by which the objective function value
Parameters and Decision Variables: should be reduced to have the variable used in the model. It can would change if the constraint R.H. is increased by 1.
You can choose between the following options: also be seen as the marginal benefit. Allowable increase and decrease: (refers to an increase in the
Allowable increase and decrease: (refers to an increase in the constraint R.H.) the range within which the shadow price can be
objective coefficient) the range within which the final value will relied upon.
not be affected (the objective function value will be impacted
(except if the final value is 0))

un
 Inventory Management better SD per
->
Daily SD, year
d b2 4ac
3 Etate"
-

is
x
=

Economic Order Quantity

= -

Reorder Point #ofdays per

∗ ∗ ROP = 𝑑̅𝐿𝑡 + (𝑧 𝐿𝑡 𝜎 + 𝑑̅ 𝜎 ) (when should we reorder) year
EOQ = ; D: annual demand, S: ordering cost, H: annual holding cost per unit
year)
in a

ROP without variability: “d and LT are constant” = 𝑑𝑥𝐿𝑡

Number of orders per year: D/EOQ
Total Annual Ordering Cost = ∗ 𝑆 + ∗ 𝐻
̅
ROP with variability: “higher reorder point (SS)” = 𝑑𝐿𝑡 + 𝑆𝑆
ROP D varies= 𝒅𝑳𝑻 + 𝒛𝝈𝒅 √𝑳𝑻 ROP LT varies= 𝑑𝐿𝑡 + (𝑧𝜎 𝑑)
2
Daily si=SD per month
Safety Stock = (𝑧 𝐿𝑡 𝜎 + 𝑑̅ 𝜎 ) SS normal: 𝑧𝜎 √𝐿𝑇
Stock out Risk= 1- Service Level; 100%-Stockout Risk = Service Level days
#of month
per

monthly a.
𝑑̅: average demand per period 12
Risk Stock out with delayed order by 1 day: On Hand = ROP – used
=yeaty
demand gear
*
I
𝐿𝑡 : average lead time Fixed Interval Order Size: Order size = 𝑑̅(𝑂𝐼 + 𝐿𝑇) + 𝑧 ∗ 𝜎 √𝑂𝐼 + 𝐿𝑇 − OR ofdays in agree
𝜎 : variance of demand per time period 𝑆𝑡𝑜𝑐𝑘 𝑜𝑛 𝐻𝑎𝑛𝑑
𝜎 : variance of lead time ( )
Performance Measures z: depends on desired service level Find Stock out Risk (find z): Quality Control
√

Job Flow Time: Accumulated Job Time Ratios: Quality Management:

Makespan: Sum of Job Time Process in control: As long as values are within range
Critical Ratio:
Job Lateness: (actual completion date – due date) Process not in control: If values are outside the range which is
Average Tardiness = Average Lateness CR>1 Job is not late; CR<1 Job is late more than the 3-sigma limit allows for 3 outof 1000
FCFS (First Come First Serve): process jobs in order they arrive Slack per Operation(S/O): # Mean Control Limit (known 𝜎): UCL/LCL = 𝑥 ± 𝑧𝜎
SPT (Shortest Processing Time): processed according to the time needed (shortest first) Mean Control Limit (unknown 𝜎): UCL/LCL= 𝑥 ± 𝐴 𝑅
EDD (Earliest Due Date): processed according to due date Avg. Flow Time: # Range Control Limit: LCL =𝐷 ∗ 𝑅; UCL =𝐷 ∗ 𝑅
Avg. Tardiness: # Process Capability:
Critical Ratio Sequence: Avg. # of Jobs: ( ) Centered Process
Step 1: Divide Due Date by the Job Time for all Jobs and choose lowest ratio
Step 2: Make a new table without the job you chose in Step 1. Repeat Step 1 but subtract Tardiness: (flow time – due date) ∗ ∗√
𝜎= “How many defects on a certain # of products”
the Job Time from the chosen Job in Step 1 from the Due Dates before you divide by Job (Process >1.33 to be capable)
Time for every row. Choose lowest again. Cpk
Uncentered Process ;
-

Step 3: Same as Step 2 but the Job Time that you subtract from the due dates is the sum
of the Job Times from Step 1 and Step 2. 𝑅 : Average Range
Step 4: Repeat until you have a sequence. X= Average Mean

Graphs/Figures and Tables

Sigma level;3(37.00), 4(31.33),5k 1.57), e(32.00)

Sensitivity Analysis Network Diagram

Distances Appendix to location Planning

Manhattan: Manhattan cost
x=1X, -x1x +, IX xx--
-

[Ixi x
((y: -3)
=
-
+

Muther Grid
y 1y, y)xc
=
-

+,(y2-y)x...
Facility Layout:
Euclidean: Euclidean cost:
Muther Grid: (heuristic appro)
(x, (y, y))).
+
A= absolutely necessary x)
y)"
-

E= very important S(xi x)

-
+
(yi -

i
I= important
O= ordinary important
U= unimportant
Critical Pairs: 1-4; 3-5; 6-2
X= undesirable (only if stated)
Sequence: 3-5-1-4-6-2

Appendix to
Facility Layout
->
Extra formulas

Appendix scheduling to

Cj completion
di due
date
linen
=

=
time
* 2

11
Y

mir
2

min.
Tardiness
Pi Processing Flowtie
=

Find:
Av. Flow the E G/n 5j2Cj
-

d;
Finish AFAP
Makespan create new variable [max? C=> Min
=> (E
Min Cmax
Au Tardiness => ceate new
non-neg
Variable Tj? j-d;
Min Tj/n
late 8 otherwise
Total late jobs binary (j for
1 =

=>
new

(4M+2j cj dj
= -

un
MinL;
X
late XEND
2 -

Deadline

Powered by TCPDF (www.tcpdf.org)