Você está na página 1de 16

5/24/16

Analysis Mathematical
Modeling
Dr. Eng. Ir. Muhammad Rusman, ST., MT., IPM

Industrial Engineering Hasanuddin University 2016

OUTLINE
v Mathematical Model: [Empirical
Model & Theoretical Model]
v Classification of Mathematical Models
v Analysis of Mathematical
Formulations-I
v Model Testing & Sensitivity Analysis

1
5/24/16

Mathematical Model
v A mathematical model of a system
§  A symbolic representation involving a
abstract mathematical formulation
v A mathematical formulation
§  composed of symbols, and makes no
sense outside mathematics
§  not a model by itself
§  it is only by relating the mathematical
formulation to a system characterization
(i.e. variables and relationships)
3

Empirical vs
Theoretical
v Empirical Model
§  The characterization is based on no theory or
knowledge.
§  The system is viewed as a black box
§  The mathematical formulation to serve as a dummy
- must be selected on an ad-hoc basis
§  Models called empirical models

v Theoretical Model
§  The characterization is done using well established
theory
§  Models called theoretical models

2
5/24/16

Classification of
Mathematical Models
v Four categories based on
mathematical structure of the
underlying formulation
§  A: Formulations suitable for modelling
deterministic static systems
§  B: Formulations suitable for modelling
deterministic dynamic systems
§  C: Formulation suitable for modelling
probabilistic static systems
§  D: Formulations suitable for modelling
stochastic dynamic systems

The cycle of
modeling

3
5/24/16

Mathematical
Formulations
v Deterministic static Formulations
§  e.g. Linear Programming
v Deterministic dynamic Formulations
§  e.g. Dynamic Programming
v Probabilistic static Formulations
§  e.g. Regression Analysis; Design of
experiment
v Stochastic dynamic Formulations
§  e.g. Markov Process; Renewal Process

Mathematical
Formulations-I
v The math. Formulations suited for
modelling deterministic systems
§  Static Formulations
• involve either algebraic equations or
function optimization
§  Dynamic Formulations
• involve two types of variables -dependent
variables and independent variables, e.g.
X(t)

4
5/24/16

Analysis of Mathematical
Formulations-I
v Step 1:
§  Discard the association so that only the math.
formulation is retained.

v Step 2:
§  Carry out an analysis of the formulation using
appropriate math techniques

v Step 3:
§  Re-introduce the discarded association so that the
analysis can be interpreted in term of physical variables
of the system characterization to yield model behavior

Types of Analysis
v Qualitative Analysis
§  deals with the study of qualitative aspects
of a given mathematical formulation
without explicitly solving it
v Quantitative Analysis
§  concerned with finding the explicit solution
which satisfies the given math formulation.
•  Analytical methods (the solution can be exact or
approximate)
•  Computational methods (the solution is only
approximate and depends on the method used)

10

5
5/24/16

Model Testing & Sensitivity


Analysis

v Internal Validity (Verification)


§  Is the model mathematically correct
and logically consistent?
§  This also involves verifying that each
expression is dimensionally consistent.
v External Validity (Validation)
§  Is the model a sufficiently valid
representation of reality?
v Testing the solution performance
§  To determine the expected benefits,
such as net profits or net savings

11

Analysis of Sensitivity
of Solution
v Analysis of Sensitivity of Solution
§  Sensitivity analysis (evaluate the
response of the best solution to changes
in various inputs)
§  Error analysis
• The input parameters are estimated on the
basis of past data
• There is no guarantee that the future will be
similar to the past
• There could be an error in input data

12

6
5/24/16

Rules for Testing


Validity
v The evaluation of the proposed policy
has to be based on observations of
actual (or simulated) performance
v The data used for the test should
be independent of the data used to
derive the best policy
v The test should not just give
expected performance, but also
some measure of its variability,
such the standard deviation

13

Sensitivity Analysis:
[Purposes]
v If the optimal solution is relatively
insensitive, then the decision maker
and user can place more confidence in
the validity and usefulness of the model
v Sensitivity analysis provides
information about the value of
additional amounts of each scarce
resource (shadow price of the
resource)
v Sensitivity analysis is used for exploring
how the optimal solution changes as a
function of such uncertain data

14

7
5/24/16

Procedure of Error
Analysis
1.Determine the optimal policy based on the best
estimate values for all input parameters (assume
that one of these, say p, is in error).
2.Assume that the value of the input parameter p
differs from the correct value, P (p=kP). Find
the optimal policy, using the (assumed) correct
value P.
3.Compute the actual value of the objective
function if the pseudo-optimal policy determined
in (1) were implemented.
4.Find the difference between the optimal objective
function values obtained from (2) –using the
correct value and (3) –using the estimate value.

15

Keputusan Pembelian
Terigu
v  Pabrik roti membeli terigu dengan harga $1000 per
ton. Kebutuhan terigu relatif konstan selama setahun
dengan total permintaan per bulan sebesar 20 ton.
Terigu dikirim dari pabrik terigu dengan menggunakan
truk dan ongkos per sekali kirim $132, tidak
tergantung dari jumlah terigu yang diangkut.
v  Uang yang digunakan untuk membeli terigu berasal
dari suatu investasi dengan interest sebesar 8% per
tahun. Juga, terigu yang disimpan diasuransikan
dengan premi 16% yang dihitung berdasarkan nilai
rata-rata persediaan per tahun. Manajer pembelian
ingin mendapatkan kebijakan pembelian terigu yg
lebih baik dari yang terjadi sekarang.

16

8
5/24/16

System Description

17

Mathematical Model

18

9
5/24/16

Mathematical Model

19

Mathematical Model

20

10
5/24/16

EOQ Model

21

Solution

22

11
5/24/16

System Description

23

Validation:[Internal]

24

12
5/24/16

Validation:[Internal]

25

26

13
5/24/16

System/Model
Overview

27

Model building
Start Prior System Design of New
knowledge experiment knowledge
Real/physical

System System
behavior characterization

Not
Satisfactory satisfactory Make Interaction
Adequate Validation Mathematical Parameter between real
model changes model estimation and abstract

Model
behavior

Mathematical
formulation
Analysis
Abstract/
mathematical

Computational Analytical

14
5/24/16

Tugas Quiz
v Jika jaringan logistik (mulai dari supplier sampai
ke konsumen) dalam sebuah supply chain
management dianggap sebagai sebuah system,
deskripsikan selengkap mungkin karakter dari
system tersebut. Ambillah sebuah contoh kasus
yang ada dalam SCM, rumuskan masalahnya,
dan buatlah model sederhananya dengan
menjelaskan asumsi-asumsi yang anda
gunakan. Tentukan pula apa yang menjadi
variabel dan parameter (jika ada) yang anda
gunakan.

v Aini, Nia, dan Nisa pergi bersama-sama ke toko buah.


Aini membeli 2 kg apel, 2 kg anggur, dan 1 kg jeruk
dengan harga Rp 67.000,00. Nia membeli 3 kg apel, 1
kg anggur, dan 1 kg jeruk dengan harga Rp 61.000,00.
Nisa membeli 1 kg apel, 3 kg anggur, dan 2 kg jeruk
dengan harga Rp. 80.000,00. Tentukan harga 1 kg apel,
1 kg anggur, dan 4 kg jeruk.

15
5/24/16

Penyelesaian

16

Você também pode gostar