Reliability — MTTF, MTBF, exponential failure model

3.6.31 · Physics › Spacecraft Structures & Systems Engineering

Overview

Spacecraft systems engineering mein, reliability ek component ya system ki probability hai ki wo specified time ke liye, stated conditions ke andar, bina failure ke apna required function perform karta rahega. MTTF (Mean Time To Failure), MTBF (Mean Time Between Failures), aur exponential failure model ko samajhna mission planning, redundancy design, aur risk assessment ke liye critical hai.

Reliability ko is sawaal ka jawaab dene wali cheez samjho: "Agar main aaj ye satellite launch karun, to 5 saal baad bhi kaam karne ki kitni chance hai?"

Core Concepts

Reliability Kya Hai?

$R (t) = P (survival until time t)$

Properties:

$R (0) = 1$ (system kaam karna shuru karta hai)
$R (t)$ monotonically decrease karta hai
$0 \leq R (t) \leq 1$

Failure rate $λ (t)$ time t par failures ki instantaneous rate describe karta hai:

$λ (t) = - \frac{1}{R ( t )} \frac{d R}{d t}$

Ye definition kyun? Failure rate batata hai ki "surviving population ka kaunsa fraction per unit time fail hota hai." Negative sign ensure karta hai ki $λ > 0$ kyunki $R (t)$ decrease hoti hai.

Exponential Failure Model

Assumption: Bahut saare electronic components ki useful life ke dauran constant failure rate $λ$ hoti hai (bathtub curve ka "flat" hissa, infant mortality ke baad aur wear-out se pehle).

Failure rate definition se shuru karte hain: $λ = - \frac{1}{R ( t )} \frac{d R}{d t}$

Kyunki $λ$ constant hai: $\frac{d R}{d t} = - λ R$

Ye ek separable differential equation hai: $\frac{d R}{R} = - λ d t$

Dono sides integrate karo: $\int_{R (0)}^{R (t)} \frac{d R}{R} = - λ \int_{0}^{t} d t$

$ln R (t) - ln R (0) = - λ t$

Kyunki $R (0) = 1$ : $ln R (t) = - λ t$

$R (t) = e^{- λ t}$

Physical meaning: Survival ki probability exponentially decay hoti hai. Zyada $λ$ matlab zyada tez decay (kam reliable). Ye radioactive decay jaisa hi mathematical form hai!

Mean Time To Failure (MTTF)

MTTF, time-to-failure random variable $T$ ki expected value hai:

$MTTF = E [T] = \int_{0}^{\infty} t \cdot f (t) d t$

jahan $f (t)$ , failure time ki probability density function hai.

f(t) aur R(t) ke beech relationship: PDF, cumulative distribution function ki derivative hai: $f (t) = \frac{d}{d t} [1 - R (t)] = - \frac{d R}{d t}$

Exponential reliability $R (t) = e^{- λ t}$ ke liye: $f (t) = - \frac{d}{d t} (e^{- λ t}) = λ e^{- λ t}$

Ab MTTF compute karo: $MTTF = \int_{0}^{\infty} t \cdot λ e^{- λ t} d t$

Integration by parts: Maan lo $u = t$ , $d v = λ e^{- λ t} d t$

$d u = d t$
$v = - e^{- λ t}$

$MTTF = [- t e^{- λ t}]_{0}^{\infty} + \int_{0}^{\infty} e^{- λ t} d t$

Pehla term: $lim_{t \to \infty} - t e^{- λ t} = 0$ (exponential dominate karta hai), aur $t = 0$ par ye zero hai.

Doosra term: $\int_{0}^{\infty} e^{- λ t} d t = [- \frac{1}{λ} e^{- λ t}]_{0}^{\infty} = \frac{1}{λ}$

$MTTF = \frac{1}{λ}$

Intuition: Agar failures $λ$ per hour ki rate se hoti hain, to average mein pehli failure ke liye $1/ λ$ hours wait karte hain. Agar $λ = 0.001$ failures/hour hai, to MTTF = 1000 hours.

Mean Time Between Failures (MTBF)

Constant failure rate aur negligible repair time wale systems ke liye:

$MTBF = \frac{1}{λ}$

MTTF vs MTBF ka fark:

MTTF: Non-repairable items (satellites, consumables) — death tak ka time
MTBF: Repairable systems (ground control wale spacecraft subsystems, IS modules) — fixes ke beech average uptime

Practice mein, exponential models ke liye: $MTBF = MTTF + MTTR$

jahan MTTR (Mean Time To Repair) average repair time hai. Lambi/impossible repair wale spacecraft ke liye, hum aksar sirf MTTF use karte hain.

Worked Examples

Ek spacecraft ke solar array ki failure rate $λ = 0.00005$ failures/hour hai.

a) 5-year mission ke liye reliability kya hai?

5 saal ko hours mein convert karo: $t = 5 \times 365.25 \times 24 = 43, 830 hours$

Exponential model apply karo: $R (t) = e^{- λ t} = e^{- 0.00005 \times 43, 830}$ $R (t) = e^{- 2.1915} = 0.112$

Answer: 5 saal survive karne ki 11.2% chance hai (failure ki 88.8% chance).

Ye step kyun? Hum directly exponential reliability formula apply karte hain. Product $λ t$ dimensionless hai (failures/hour × hours = failures), exponent ke liye ek pure number deta hai. Note karo $0.00005 \times 43, 830 = 2.1915$ .

b) MTTF kya hai?

$MTTF = \frac{1}{λ} = \frac{1}{0.00005} = 20, 000 hours$

Saalon mein convert karo: $MTTF = \frac{20 , 000}{24 \times 365.25} \approx 2.28 years$

Answer: Expected lifetime 2.28 saal hai. Kyunki mission 5 saal ka hai aur MTTF sirf 2.28 saal hai, humein redundancy chahiye hogi!

Ek Mars orbiter ko 2-year mission ke liye 95% reliability chahiye. Maximum acceptable failure rate kya hai?

Given: $R (t) = 0.95$ , $t = 2 \times 365.25 \times 24 = 17, 532$ hours

Find: $λ$

$R (t) = e^{- λ t}$ se: $0.95 = e^{- λ \times 17, 532}$

Natural log lo: $ln (0.95) = - λ \times 17, 532$ $- 0.05129 = - λ \times 17, 532$ $λ = \frac{0.05129}{17 , 532} = 2.926 \times 1 0^{- 6} failures/hour$

Answer: Maximum acceptable failure rate $2.93 \times 1 0^{- 6}$ failures/hour hai, ya equivalently MTTF ≥ 342,000 hours (39 saal).

Ye step kyun? Humne logarithms use karke exponential function ko invert kiya. Log, R aur $λ$ ke beech exponential relationship ko "unwrap" karta hai.

Do identical power supplies parallel mein use ho rahe hain (system kaam karta hai agar kam se kam ek kaam kare). Har ek ki $λ = 0.0001$ failures/hour hai. 10,000 hours ke liye system reliability kya hai?

Single component reliability: $R_{1} (t) = e^{- 0.0001 \times 10, 000} = e^{- 1} = 0.368$

Parallel system: System tabhi fail hota hai jab DONO fail ho jayein. $R_{sys} = 1 - P (both fail)$

Independent components ke liye: $P (both fail) = [1 - R_{1} (t)]^{2}$

$R_{sys} = 1 - (1 - 0.368)^{2} = 1 - (0.632)^{2}$ $R_{sys} = 1 - 0.399 = 0.601$

Answer: System reliability 60.1% hai, jo single unit ke 36.8% se kaafi behtar hai.

Redundancy kyun kaam karta hai: Bhale hi har unit ke survive hone se zyada fail hone ki chance ho, DONO ke fail hone ki chance do probabilities ka product hai, jo chhota hota hai. Long-duration missions ke liye redundancy bahut zaroori hai.

Common Mistakes

Galat soch: "MTTF = 1000 hours matlab component 1000 hours chalega."

Kyun sahi lagta hai: MTTF ek warranty ya design life specification ki tarah lagta hai.

Sach: MTTF bahut saare components ka average hai. Aadhe MTTF se pehle fail ho jayenge, aadhe baad mein. Specifically, exponential model ke liye: $R (MTTF) = R (1/ λ) = e^{- λ \cdot 1/ λ} = e^{- 1} \approx 0.37$

Sirf 37% components apne MTTF tak survive karte hain! 63% pehle fail ho jaate hain.

Fix: Mission planning ke liye sirf MTTF nahi, reliability curves $R (t)$ use karo. 90% reliability achieve karne ke liye, $t ≪ MTTF$ chahiye.

Galat soch: "Do redundant systems jinki $λ = 0.001$ hai unki combined $λ = 0.002$ hogi."

Kyun sahi lagta hai: Hum aksar cheezein series mein add karte hain, aur failure rates series systems ke liye add hoti bhi hain.

Sach: Parallel redundancy ke liye, tum simply $λ$ values add nahi kar sakte. Ye use karna hoga: $R_{parallel} = 1 - \prod_{i = 1}^{n} (1 - R_{i} (t))$

Redundancy factor $n$ wale identical components ke liye: $R_{parallel} = 1 - (1 - e^{- λ t})^{n}$

Iska koi simple constant failure rate nahi hoti!

Fix: Redundant systems ke liye hamesha reliability functions $R (t)$ se kaam karo, phir zaroorat ho to effective failure rate compute karo.

Galat soch: "Saare spacecraft components exponential reliability follow karte hain."

Kyun sahi lagta hai: Exponential model simple hai aur widely padhaya jaata hai.

Sach: Exponential model constant failure rate assume karta hai (memoryless failures, jaise random cosmic ray strikes). Mechanical parts (bearings, actuators) mein wear ki wajah se increasing failure rate hoti hai (Weibull distribution). Batteries predictably degrade hoti hain (alag model).

Bathtub curve:

Infant mortality (decreasing $λ$ ) — manufacturing defects
Useful life (constant $λ$ ) — random failures, exponential model apply hota hai
Wear-out (increasing $λ$ ) — mechanical aging

Fix: Exponential model sirf electronic components ke liye unke useful life phase mein use karo. Mechanisms ke liye Weibull distribution use karo; batteries ke liye cycle-life models use karo.

Active Recall Questions

#flashcards/physics

Reliability R(t) ki definition kya hai? :: Ye probability ki ek system time 0 se time t tak bina failure ke operate kare. $R (t) = P (survival until t)$ , jahan $R (0) = 1$ aur $0 \leq R (t) \leq 1$ .

Kaun sa assumption exponential reliability model ki taraf le jaata hai?

Constant failure rate

λ

time ke saath. Iska matlab hai failures memoryless hain — agli ghante mein failure ki probability is baat par depend nahi karti ki component kitne time se already operate kar raha hai.

Constant failure rate se exponential reliability formula derive karo.

λ = - \frac{1}{R} \frac{d R}{d t}

se shuru karo, constant

λ

deta hai

\frac{d R}{R} = - λ d t

. Integrate karo:

ln R (t) = - λ t

, isliye

R (t) = e^{- λ t}

MTTF kya hai aur failure rate se kaise relate karta hai?

Mean Time To Failure — ek non-repairable component ki pehli failure tak expected time. Exponential model ke liye

MTTF = \frac{1}{λ}

Kitne fraction components apne MTTF tak survive karte hain?

Sirf 37% (lagbhag 1/e). Ye isliye hai kyunki

R (MTTF) = R (1/ λ) = e^{- 1} \approx 0.368

MTTF aur MTBF mein kya fark hai?

MTTF non-repairable items ke liye hai (permanent failure tak ka time). MTBF repairable systems ke liye hai (failures ke beech average uptime). Negligible repair time wale exponential models ke liye, numerically equal hote hain.

Ek component ki

λ = 0.0002

failures/hour hai. 5000 hours ke liye reliability kya hai?

R (5000) = e^{- 0.0002 \times 5000} = e^{- 1} \approx 0.368

ya 36.8% survival probability.

Parallel redundant systems ke liye failure rates add kyun nahi kar sakte?

Parallel systems tabhi fail hote hain jab SAARE components fail ho jayein. Reliability hai

R_{sys} = 1 - \prod (1 - R_{i})

, jo simple sum of

λ

values mein reduce nahi hoti. Effective failure rate non-constant hoti hai.

Exponential model kab apply NAHI hota?

Mechanical wear-out mein (increasing

λ

, Weibull use karo), infant mortality mein (decreasing

λ

), battery degradation mein (cycle-based), ya kisi bhi component mein jiska failure rate time-dependent ho.

Ek mission ko 10 saal ke liye 90% reliability chahiye. Component ka MTTF = 15 saal hai. Kya ye kaafi hai?

Check karo:

R (10 yr) = e^{- 10/15} = e^{- 0.667} = 0.513

ya 51.3%. KAAFI NAHI hai! Sirf 51% reliable, chahiye 90%. Design improve karna hoga ya redundancy add karni hogi.

Recall Ek 12-saal ke bachche ko explain karo

Socho tumhare paas ek toy robot hai. "Reliability" wo chance hai ki tumhara robot bina toote kaam karta rahe. Kuch toys randomly tootte hain — shayad vibration se ek wire disconnect ho jaati hai. Agar tumhare robot ke har saal tootne ki 10% chance hai, to hum calculate kar sakte hain ki 5 saal survive karne ki kitni probability hai.

MTTF (Mean Time To Failure) ye poochne jaisa hai: "Agar mere paas 100 identical robots hote, to average mein har ek ke tootne mein kitna time lagta?" Agar MTTF 10 saal hai, to matlab average mein 10 saal chalte hain. Lekin important baat: kuch 2 saal mein toot jaate hain, kuch 20 saal chalte hain! Ye ek average hai, guarantee nahi.

Exponential model har ghante dice roll karne jaisa hai. Agar 1 aaya, to robot toot gaya (low chance). Har ghanta independent hai — robot ko yaad nahi ki wo 5 saal se kaam kar raha hai aur thak gaya ho. Ye electronics ke liye (random failures) bahut badhiya kaam karta hai lekin un cheezoon ke liye nahi jo wear out hoti hain jaise bike chains (wo use se thakti hain).

Spacecraft engineers ye formulas isliye use karte hain kyunki wo space mein cheezein fix nahi kar sakte! Agar satellite toot jaata hai, to game over. Isliye wo calculate karte hain: "Iski 5 saal chalte rehne ki kitni chance hai?" aur zaroorat ho to backup systems add karte hain.

Mnemonic: Low Lambda = Long Life

Chhota λ (failure rate) matlab bada MTTF (lambi life). Ye reciprocals hain — fraction ko flip karo!

Visual: $λ$ failures per hour hai (teekha neeche jaana), MTTF hours per failure hai (lambi flat road). Inverse relationship.

Connections

Bathtub Curve — failure rate vs. time, teen life phases
Weibull Distribution — wear-out ke liye generalized failure model
Series vs. Parallel System Reliability — component reliabilities kaise combine karein
Fault Tree Analysis — system se components tak top-down reliability analysis
Redundancy Design — high reliability ke liye active vs. standby redundancy
Probability and Statistics Fundamentals — exponential distribution, expected value
Mission Design Constraints — reliability kaise mass, power, cost budgets ko drive karti hai
Poisson Process — constant-rate random events ka mathematical foundation
Availability vs. Reliability — repair time aur uptime percentage incorporate karna
Safety-Critical Systems — jab 99.999% reliability chahiye (five nines)

In reliability concepts ko master karo aur tum mission risk quantify kar sakte ho, redundancy decisions justify kar sakte ho, aur systems engineers ki language mein baat kar sakte ho. Har spacecraft in exponentials ke khilaf ek gamble hai — apne odds jaano.