a study on fuzzy inventory control system under demand...
TRANSCRIPT
A STUDY ON FUZZY INVENTORY CONTROL SYSTEM UNDER DEMAND UNCERTAINTY
Master’s Thesis Presentation 2010.02.15
Author : Achmad RIADISupervisor: Prof. Saburo TSURUTA
Logistics System EngineeringCourse of Maritime Technology and Logistics
Tokyo University of Marine Science and Technology
Contents
1. Introduction
2. Review of Past Studies Related to Fuzzy Application in Inventory Control
3. Data Used for the Dynamic Simulation of Inventory Control
4. Model Design for the Dynamic Simulation of Inventory Control
5. Result of the Simulations
6. Conclusion
7. Future Research Direction 2
Introduction
Background
3
1. Inventory is needed to allow balancing between supply and demand.
2. Careful inventory control is needed to make good economic sense.
3. Considering demand uncertainty is important in inventory control.
4
Introduction (cont.)
Problem
In an ordinary inventory control system, the demand uncertainty is measured by statistical model of a normal distribution, specified by Mean and Standard Deviation.
Prediction of Mean and Standard Deviation of demand used in an ordinary inventory control system is not always precise.
Many methods have been carried out for the precise prediction of future demand:
- Adaptive forecasting, such as Moving Average, etc.
- Fuzzy system.
5
Introduction (cont.)
Objective
To investigate the effectiveness of Fuzzy inventory control system for controlling the inventory, especially with the presence of demand uncertainty.
In what condition Fuzzy inventory control system is effective?
When Fuzzy inventory control system is more effective?
6
Problem Domain:
Solving demand uncertainty by utilizing demand as Fuzzy
variable.
Showing the effectiveness by comparing the inventory control
systems (ordinary, moving average and Fuzzy inventory
control system)
Differences between this study and the other past studies:
Utilizing demand as Fuzzy variable with membership
functions.
Comparing the inventory control systems by dynamic
simulation with the possibility that the value of future
demand will change.
Review of Past Studies Related to Fuzzy Application in Inventory Control
Data Used for the Dynamic Simulation of Inventory Control
Many data have been collected and the following data have been decided which are appropriate with the needs of the study.
Demand Data
Perpetual demand with normal distribution specified by:
Mean : 651 units/day
Sta. Dev. : 200 units/day
Lead Time Data
Constant 2 days.
Relevant Costs Data
Holding Cost : 20% of unit cost per annum.
Ordering Cost : ¥688 / order (0.47% of unit cost)
Stockout Cost : 0% ~ 4% of unit cost per unit.
Unit Cost : ¥150 7
Model Design for the Dynamic Simulation of Inventory Control
There are three dynamic simulation models, which have been designed in this study:
Ordinary Inventory Control System.
Moving Average Inventory Control System.
Fuzzy Inventory Control System.
Basic method for controlling inventory is periodic review inventory control (base stock policy).
8
9
Model Design for the Dynamic Simulation of Inventory Control (cont.)
Periodic Review Inventory Control (Base Stock Policy)
Q∗ = 2CD
H T =
Q
D
∗ M∗ = d T + LT + z(s′
d)
Source: Ballou, R.H (2004)T = Re-orders period M* = Maximum
inventory level
Wrap To Zero
z-i
In
DelayOut
Variable
Integer Delay
150
Unit Cost
simout_dlvry
To Workspace 7
simout_oq
To Workspace 6
simout_oc
To Workspace 5
simout_uc
To Workspace 4
simout_hc
To Workspace 3
simout_so
To Workspace 2
simout_inv
To Workspace 1
Switch 2
Switch 1
Subtract 2
Subtract 1
|u|
Stockout
round
Rounding
Function
Review Period
688
Ordering Cost
Order
Quantity
5820
Maximum
Inventory Level
2
Lead Time
K Ts
z-1
Inventory
0.08
Holding Cost
Demand
0
Constant 2
0
Constant 1 ~= 0
Compare
To Zero
10
Model Design for the Dynamic Simulation of Inventory Control (cont.)
Dynamic Simulation Model for Ordinary Inventory Control System
M* = d(T+LT) + z(s’d)5820 units
5 days
T =Q
D
∗
Maximum
Inventory
Level
2.3
safety factor, z
Wrap To Zero
z-i
In
DelayOut
Variable
Integer Delay
150
Unit Cost
simout_dlvry
To Workspace 7
simout_oq
To Workspace 6
simout_oc
To Workspace 5
simout_uc
To Workspace 4
simout_hc
To Workspace 3
simout_so
To Workspace 2
simout_inv
To Workspace 1
Switch 2
Switch 1
Sum 3
Sum 2
Sum 1
Subtract 2
Subtract 1
|u|
Stockout
round
Rounding
Function 3
roundRounding
Function 2
round
Rounding
Function
Review Period
688
Ordering Cost
Order
Quantity
In1
In2
In3
In4
In5
In6
In7
In8
Out
1
Out
2
Out
3
Out
4
Out
5
Out
6
Out
7
Moving
Variance
In1
Out
1
Out
2
Out
3
Out
4
Out
5
Out
6
Out
7
Moving
Average
sqrtMath
Function
2
Lead Time
K Ts
z-1
Inventory
0.08
Holding Cost
Demand
0
Constant 2
0
Constant 1 ~= 0
Compare
To Zero
11
Model Design for the Dynamic Simulation of Inventory Control (cont.)
Dynamic Simulation Model for Moving Average Inventory Control System
Maximum
Inventory
Level
Wrap To Zero
z-i
In
DelayOut
Variable
Integer Delay
150
Unit Cost
simout_dlvry
To Workspace 7
simout_oq
To Workspace 6
simout_oc
To Workspace 5
simout_uc
To Workspace 4
simout_hc
To Workspace 3
simout_so
To Workspace 2
simout_inv
To Workspace 1
Switch 2
Switch 1
Subtract 2
Subtract 1
|u|
Stockout
Standard
Deviation
round
Rounding
Function 2
round
Rounding
Function
Review Period
688
Ordering Cost
Order
QuantityMean
2
Lead Time
K Ts
z-1
Inventory
0.08
Holding Cost
Fuzzy Logic
Controller
Demand
0
Constant 2
0
Constant 1 ~= 0
Compare
To Zero
12
Model Design for the Dynamic Simulation of Inventory Control (cont.)
Dynamic Simulation Model for Fuzzy Inventory Control System
13
Model Design for the Dynamic Simulation of Inventory Control (cont.)
Process inside the Fuzzy Logic Controller Block
Fuzzy Inference
System by ANFIS
Output Max. Inventory Level
Uncertain Inputs (mean, std. dev)
with linguistic variables
Fuzzy Inference System includes membership functions, rules, and output parameters.
ANFIS will construct a Fuzzy inference system with some parameters are tuned (adjusted).
Result of the Simulations
14
Demand mean Demand standard
deviation
Condition 1 (There is no change with the
future demand or the future
demand is same as the
expected demand)
Expected demand
mean
Expected demand
standard deviation
Condition 2-A
(Future demand mean is
uncertain, demand standard
deviation is fixed)
Condition 2-B (Future demand mean is
fixed, demand standard
deviation is uncertain)
Vary from -50% to
+50% of expected
demand mean (step
is 10%)
Expected demand
mean (fixed)
Expected demand
standard deviation
(fixed)
Vary from -50% to
+50% of expected
demand standard
deviation (step is
10%)
Simulation Conditions
15
Result of the Simulations (cont.)
Simulation Results with Condition 1 (there is no change with the future demand)
95
100
105
110
115
120
125
130
135
140
SC is 0% of unit cost
SC is 1% of unit cost
SC is 2% of unit cost
SC is 3% of unit cost
SC is 4% of unit cost
Tota
l rel
evan
t co
sts
(¥)
x103
Total Relevant Costs between Three Inventory Control Systems (Seed Num. 100)
Ordinary inventory control system
MA inventory control system
Fuzzy inventory control system
0.38% < Ordinary Inv. Contr. Syst.
9.43% > Ordinary Inv. Contr. Syst.
1.15% > Ordinary Inv. Contr. Syst.
15.47% > Ordinary Inv. Contr. Syst.
16
Result of the Simulations (cont.)
Simulation Results with Condition 1 (there is no change with the future demand)
Ordinary Moving Average Fuzzy
Service Level (%) 99.41 98.77 99.29
Conclusion of Simulation Results with Condition 1
When the stockout cost is not considered, Fuzzy inventory control system is more effective.
When the stockout cost is considered, ordinary inventory control system is more effective.
17
Result of the Simulations (cont.)
Simulation Results with Condition 2-A (demand mean is uncertain, demand std. dev is fixed)
0.0
0.5
1.0
1.5
2.0
2.5
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Tota
l Re
leva
nt
Co
sts
Demand Mean
Total Relevant Costs with Uncertain Demand Mean(In Case of Stock out Cost = 0% of Unit Cost)
OrdinaryMoving AverageFuzzy
Expected Demand Mean
Percent Different13.43%
Service Level50.83%
Service Level98.34%
18
Result of the Simulations (cont.)
Simulation Results with Condition 2-A (demand mean is uncertain, demand std. dev is fixed)
0.0
0.5
1.0
1.5
2.0
2.5
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Tota
l Re
leva
nt
Co
sts
Demand Mean
Total Relevant Costs with Uncertain Demand Mean(In Case of Stock out Cost = 4% of Unit Cost)
OrdinaryMoving AverageFuzzy
Expected Demand Mean
19
Result of the Simulations (cont.)
Simulation Results with Condition 2-B(demand mean is fixed, demand std. dev is uncertain)
0.8
0.9
1
1.1
1.2
1.3
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Tota
l Re
leva
nt
Co
sts
Demand Std. Dev.
Total Relevant Costs with Uncertain Demand Std. Dev.(In Case of Stock out Cost = 0% of Unit Cost)
OrdinaryMoving AverageFuzzy
Expected DemandStd. Dev.
Service Level98.21%
Service Level98.95%
20
Result of the Simulations (cont.)
Simulation Results with Condition 2-B(demand mean is fixed, demand std. dev is uncertain)
0.8
0.9
1
1.1
1.2
1.3
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Tota
l Re
leva
nt
Co
sts
Demand Std. Dev.
Total Relevant Costs with Uncertain Demand Std. Dev.(In Case of Stock out Cost = 4% of Unit Cost)
OrdinaryMoving AverageFuzzy
Expected DemandStd. Dev.
Conclusion
• When there is no change with the future demand:
Fuzzy inventory control system is more effective (stock out cost is not considered)
Ordinary inventory control system is more effective (stock out cost is considered)
• When the future demand mean is uncertain:
Fuzzy inventory control system is more effective.
• When the future demand standard deviation is uncertain:
Stockout cost is not considered:
Fuzzy inventory control system is effective only when the future demand sta. dev. lower than the expected demand sta. dev.
Stockout cost is considered:
Both ordinary and Fuzzy inventory control system is effective. 21
22
Future Research Direction
The other uncertainties such as supply uncertainty, lead time uncertainty and costs uncertainty can be present in inventory control.
Comparing the result of this study with the result of the other studies with different Fuzzy system.
Concerning the multi echelons/stages within the supply chain is also interesting for the future research.