Model for Predicting Human

Reliability
Arief Rahman, 2021
A. CLASICCAL DECISION THEORY
Elements of Decision Making
The rational decision making
process
• Define the problem
• Information gathering
• Identify alternatives
• Evaluate alternatives
• Select and implement decision
Why decisions often go wrong:
• Certainty vs. Uncertainty
• Bounded rationality
• Nonlinearity
• Habits & heuristics
• Path of least resistance
 Classic Decision Theory
• Maximizing expected value of the outcome
• Mathematical models of human decision
making
• Following assumptions are made about how
people make decisions:
• All alternatives are considered
• Information acquisition is perfect
• Probabilities are calculated correctly
• Problems:
• Humans are not rational decision makers
• No universal agreement on the worth
associated with various outcomes.
Basic Concepts in Decision Analysis
• State : A description of the world
• Events : A set of states {S1, S2,…, Sj}
• Consequences : A set of states
• Acts : Results from decisions
Basic Terminology
• Ordering of preferences
• Alternatives can be quantified and ordered
• A > B, A preferred to B
• A = B, A is equivalent to B
• A ≥B, B is not preferred to A
• Transitivity of preference
• if A1 ≥A2, & A2 ≥A3, then A1 ≥A3
Decision Trees
• Decisions over time and/or events

Decision Making Under Certainty:

• Each alternative leads to one and only one
consequence (Consequences are known)
• Lexicographic ordering
• Dominance
• Satisficing
• Maximin
• Minimax
Lexicographic Ordering

• All options are first compared in terms of the

criterion deemed most important.
• If there is a unique best performing option, that
option is selected as the most preferred.
• In case of a tie, the selection process moves
to the second ranked criterion
• Seeks the remaining option which scores best on
the second criterion.
• In case of a tie on the second criterion, the
process is repeated for options tying in terms
of both the first and second ranked criteria,
• And so on until a unique option is identified or
all criteria have been considered.
Dominance

• Effective for both quick qualitative and quantitative

comparisons
• In the real world, solutions rarely are outwardly dominant
Satisficing

• Otherwise known as Minimum Aspiration

• Select any Ai over all events j such that
• Cij ≥ aspiration level
• Cease to search for alternatives when you find
an alternative whose expected utility or level
of preference satisfaction exceeds some
previously determined threshold.
• Stopping rule
• Is this in keeping with the rational approach to
decision making?
Maximin

• Select any Ai over all events j such that you minimize the maximum loss
• Maximum of the (row)minima.
• Conservative approach but overly pessimistic
Minimax

• Avoid regrets that could result from making a non-optimal decision.

• Regret: the opportunity loss if i is chosen
• Opportunity loss is the payoff difference between the best possible
outcome under Sj and the actual outcome from choosing Ai.
• Convert payoff matrix to opportunity loss (regret) table
 Maximin vs. Minimax

• Minimax contains more problem information through opportunity

losses (e.g. actual monetary losses plus unrealized potential
profits)
• Still conservative but more middle-of-the-road
The Uncertain State of the World

Decision Making Under Risk

• Each alternative has several
possible consequences
• The probability of occurrence for
each consequence, C, is known.
• Expected value: E (Ai) =Σjpj(Cij)

Decision Making Under Uncertainty

• Alternatives are known
• Consequences are not known
• Probability of outcomes are not known
• Minimax/Maximin can still be used.
• Utility a key concept
 Utility Theory

• Utility theory is an attempt to infer CONTOH MAUT:

subjective value, or utility, from
choices. Want to buy a car based on price, fuel economy, &
• Applies to both decision making under reliability.
risk & decision making under uncertainty
• Two types
• Expected utility
• Same as EV but utility is the value
associated with some outcome, not
necessarily monetary.
• Subjective Expected Utility (SEU)
• Subjective focuses on decision making
behavior
• Risk neutral/adverse/seeking
• Multi-attribute utility (MAUT)
• Multiple objective extension of SEU
• Interval scale is critical Major drawback: Assumption of rationality
B. BAYES RULE
Conjunctions and Disjunctions
Venn Diagrams

A B
• P(A  B) = P(A) x P(B)

• P(A  B) = P(A) + P(B)

(mutually exclusive) A B

• P(A  B)
=P(A)+P(B) - P(A  B)
(not mutually exclusive)
A B
 Conditional Probability
The probability of an event B occurring when it is CONTOH:
known that some event A has occurred is called a Suppose that our sample space S is the population of
conditional probability and is denoted by adults in a small town who have completed the
P(B|A). The symbol P(B|A) is usually read “the requirements for a college degree. We shall categorize
probability that B occurs given that A occurs” or them according to gender and employment status
simply “the probability of B, given A.”

M : a man is chosen,
E : the one chosen is employed
Theorem of total probability /
Rule of Elimination
CONTOH Rule of Elimination:

In a certain assembly plant, three machines, B1, B2, and B3, make 30%, 45%, and 25%,
respectively, of the products. It is known from past experience that 2%, 3%, and 2% of
the products made by each machine, respectively, are defective. Now, suppose that a
finished product is randomly selected. What is the probability that it is defective?
PENYELESAIAN :
A: the product is defective,
B1: the product is made by machine B1 P(A) = P(B1)P(A|B1) + P(B2)P(A|B2) + P(B3)P(A|B3)
B2: the product is made by machine B2
B3: the product is made by machine B3 P(A) = 0.006 + 0.0135 + 0.005
P(B1)P(A|B1) = (0.3)(0.02) = 0.006
= 0.0245
P(B2)P(A|B2) = (0.45)(0.03) = 0.0135 • By the rule of elimination, suppose that
P(B3)P(A|B3) = (0.25)(0.02) = 0.005 we now consider the problem of finding
the conditional probability P(Bi|A).
• In other words, suppose that a product
was randomly selected and it is
defective.
• What is the probability that this product
was made by machine Bi ? Questions of
this type can be answered by using the
following theorem, called Bayes’ rule.
 Bayes’s Rule
CONTOH :
A manufacturing firm employs three analytical plans for
the design and development of a particular product. For
cost reasons, all three are used at varying times. In fact,
plans 1, 2, and 3 are used for 30%, 20%, and 50% of the
products, respectively. The defect rate is different for the
three procedures as follows:
P(D|P1) = 0.01
With reference to CONTOH Rule of Elimination, if a product P(D|P2) = 0.03
was chosen randomly and found to be defective, what is the P(D|P3) = 0.02
probability that it was made by machine B3 ? where P(D|Pj) is the probability of a defective product,
given plan j. If a random product was observed and
found to be defective, which plan was most likely used
and thus responsible?

P(P1) = 0.30, P(P2) = 0.20, and P(P3) = 0.50

C. Statistical Density Function
 Continuous Random Variables
Probability Density Function (PDF)
Means and Variance
Example: Uniform PDF
Cumulative Distribution Function
CDF for Discrete r.v.’s
Standard Gaussian (Normal) PDF
• Standard Normal:
General Gaussian (Normal) PDF
• General Normal:

• It turns out that:

• Let then:

• Fact:
 Calculating Normal Probabilities

• No closed form available for CDF.

– But, there are tables (for standard normal).

• If then

• If :
 DISTRIBUSI WEIBULL
Distribusi Weibull ini diperkenalkan oleh ahli Perubah acak kontinyu X terdistribusi Weibull dengan parameter  dan ,
fisikawan swedia Waloddi Weibull pada tahun jika fungsi :
1939. Grafik distribusi weibull untuk =1 dan   x  −1e− x ; x  0
berbagai nilai parameter  berikut ini: f(x) = 
Distribusi Weibull 0 ; x yanglain
dengan   0 dan   0
 = 1,  = 5
Jika =1 maka distribusi weibull menjadi distribusi eksponensial.
2.0

•
 = 1,  = 3 • Jika >1 maka kurvanya mirip lonceng dan menyerupai kurva
 = 1,  = 1 normal namun agak miring/tendensi ke kanan atau kekiri.
1.5

 = 1,  = 2
f(x)

1.0

Distribusi Weibull dapat digunakan dalam persoalan keandalan

dan pengujian panjang umur (life testing) seperti waktu sampai
rusak atau panjang umur suatu kompenen, diukur dari suatu
0.5

waktu tertentu sampai rusak.

0.0

0.0 0.5 1.0 1.5 2.0 2.5

x
Contoh :
Seorang pengemudi truk sedang mengendarai truknya dari Jakarta ke Surabaya. Berapa error rate nya 10 jam
setelah dia mengendarai truknya dari Jakarta dan hitung mean time to human error dari pengemudi tersebut?
Jika terdistribusi Weibull dengan =500 dan =2
Penyelesaian :

Fungsi densitas distribusi Weibull :

Reliability performance dari pengemudi tersebut pada waktu t adalah:

Mean Time to Human Error :

Error rate pengemudi tersebut

selama waktu 10 jam adalah 0.00008
error/jam dan MTTHE pengemudi
tersebut adalah 12500 jam
 DISTRIBUSI POISSON
Distribusi peluang peubah acak Poisson X, yang Contoh :
menyatakan banyaknya sukses yang terjadi dalam
suatu selang waktu atau daerah tertentu Diketahui probabilitas terjadinya kecelakaan kerja adalah
dinyatakan dengan t, diberikan oleh: 0,0005. Kalau di perusahaan jumlah pekerja yang mengalami
kecelakaan kerja sebanyak 400. Hitunglah peluang tepat tiga
orang yang mengalami kecelakaan.
Penyelesaian :
x = 0, 1, 2, … dan λ > 0
λ= n.p = 400 x 0,0005 = 0.2
Distribusi Poisson digunakan untuk menghitung
peluang jumlah khusus “kejadian” selama jangka
waktu atau selang waktu tertentu (berbentuk peubah
acak).
Contoh:
• Seorang insinyur Teknik Industri akan meneliti waktu
T antara kendaraan tiba di suatu simpang empat
yang padat selama waktu sibuk di suatu kota besar.
Waktu tiba merupakan kejadian Poisson.
• Mahasiswa menemukan dompet di jalan Jawa. Jadi peluang tepat tiga orang mengalami kecelakaan kerja adalah
Banyak mahasiswa yang lewat jalan tersebut 1,09x10^-3
merupakan contoh ditribusi Poisson.
 DISTRIBUSI EKSPONENSIAL
Distribusi ekponensial termasuk didalam distribusi
yang bersifat probabilistik. Distribusi eksponensial
• Distribusi eksponensial banyak diaplikasikan pada proses
dapat diketahui apabila kita mendeskripsikan antrian layanan seperti proses telepon masuk pada system
panjang dari waktu inter-arrival secara homogen layanan.
pada proses Poisson. Distribusi eksponensial juga • Distribusi eksponensial juga digunakan pada teori dan teknik
dapat dilihat sebagai pasangan yang bersifat kontinu reabilitas.
pada distribusi geometri. Distribusi eksponensial • Distribusi eksponensial juga mudah untuk ditambahkan pada
mendeskripsikan sebuah waktu secara kontinu yang
failure rates pada model reabilitas.
dibutuhkan untuk proses perpindahan state.
Karakteristik Distribusi Eksponensial

• Sebuah variabel acak X dikatakan merupakan
distribusi lognormal, dengan parameter μ
dan σ, jika ln (X) merupakan distribusi
normal dengan μ mean dan deviasi standar
σ.
• Kurva lognormal merupakan kurva yang
miring ke kanan. Untuk μ tertentu,
kemiringan meningkatkan sebanding dengan
peningkatan σ.
DISTRIBUSI LOGNORMAL
• Karena lognormal berdasarkan pada distribusi normal, maka setiap sifat
lognormal dapat diturunkan dari sifat distribusi normal.
• PDF distribusi normal adalah:
f(x) = exp[−(x − μ)2 / (2σ2)] / [(2π)1/2σ], for x in R.
• Lognormal density function, dgn parameter μ dan σ, adalah:
f(x) = exp{−[ln(x) − μ]2 / (2σ2) }/ [x (2π)1/2 σ]
untuk x > 0
• Perbedaan pdf distribusi normal dan pdf lognormal tidak hanya
penggantian x dengan ln (x), tetapi juga merupakan faktor x tambahan
dalam x (2π) 1 / 2 σ dikarenakan perubahan variabel dari X ke ln [X].

✓ Pada biologi, variabel yang cenderung ke distribusi lognormal adalah pengukuran

kesehatan individu misalnya kelangsungan hidup penderita kanker.
✓ Pada finance, menurut keterangan Black–Scholes model, perubahan pada logaritma
dari penaksiran kurs, indeks harga, and indeks saham pasar diasumsikan normal.
✓ Distribusi dari ukuran kota adalah lognormal. Menurut masing-masing ukurannya,
semua kota mengikuti proses pertumbuhan stokastik yang sama. Sehingga hasilnya,
logaritma dari ukuran kota adalah distribusi normal.
✓ Pada analisis realibility, distribusi lognormal sering digunakan untuk memodelkan
waktu dalam memperbaiki sistem yang maintainable.
 
Fungsi gamma : ( ) = x −1e − x dx
 DISTRIBUSI GAMMA
0

Dengan sifat Γ(α)= (α-1) Γ(α-1), sehingga

untuk α=n yg berupa bilangan bulat positif,
maka Γ(n) = (n-1)! Distribusi Gamma
Variabel random X memiliki distribusi gamma
dengan parameter α dan β, jika fungsi rapat 1.2

probabilitasnya diberikan oleh: 1

 −x
 1 x −1e  ; x0 0.8
f(x) =    ( ) alfa=1 beta=1

 0 ; x yanglain 0.6 alfa=2 beta=1
alfa=4 beta=1
dengan   0 dan   0 0.4

Distribusi gamma memiliki peran yang 0.2

sangat penting di bidang teori antrian

0
dan teori keandalan (reliabilitas). 0 1 2 3 4 5
Distribusi gamma mendapat namanya x
dari fungsi gamma yang sudah dikenal
luas.
Contoh:
Dalam studi terhadap kesehatan pekerja pabrik, dipelajari efek polusi asap pabrik industri terhadap waktu
survival-nya. Diketahui bahwa untuk tingkat polusi udara tertentu, waktu survival-nya mengikuti pola distribusi
gamma dengan α=5 dan β=10 dalam satuan minggu. Berapakah probabilitasnya bahwa pekerja tersebut akan
survive terhadap polutan pabrik tersebut tak lebih dari 60 minggu.
Penyelesaian :
Misal X adalah variabel random yg menyatakan waktu hidup (survival time), berarti probabilitasnya
bahwa X ≤ 60 adalah:
60 60
1 1
P( X  60; = 5,  = 10) =   x −1e − x /  dx =  5 x 4e − x /  dx
0
 ( ) 0
 (5)
Integral ini sulit dievaluasi
secara langsung. Akan tetapi
Jadi, dapat didefinisikan x/β=y, berarti x= βy dx= βdy dan x=60 jadi
dapat dievaluasi dengan
y= 60/10=6 sehingga:
perantaraan tabel fungsi
gamma tak lengkap F: 1
6
1 4 −x /  1 1
6 6
1 4 −y
P( X  60) = 5  x e dx = 5  ( y ) 4 e − y  dy =  y e dy
 0 (5)  0 (5)  (5)
y −1e − y
x 0

F ( x; ) =  dy
0
( ) Dengan definisi fungsi gamma tak lengkap F(x;α) jadi:
P (X≤60)= F(x=6; α=5) = 0.715
Terima Kasih