Reliability — MTTF, MTBF, exponential failure model

Overview

In spacecraft systems engineering, reliability is the probability that a component or system will perform its required function without failure for a specified time under stated conditions. Understanding MTTF (Mean Time To Failure), MTBF (Mean Time Between Failures), and the exponential failure model is critical for mission planning, redundancy design, and risk assessment.

Think of reliability as answering: "If I launch this satellite today, what's the chance it's still working 5 years from now?"

Core Concepts

What is Reliability?

$R(t) = P(\text{survival until time } t)$

Properties:

• $R(0) = 1$ (system starts working)
• $R(t)$ is monotonically decreasing
• $0 \leq R(t) \leq 1$

The failure rate $\lambda(t)$ describes the instantaneous rate of failures at time t:

$\lambda(t) = -\frac{1}{R(t)} \frac{dR}{dt}$

Why this definition? The failure rate tells us "what fraction of the surviving population fails per unit time." The negative sign ensures $\lambda > 0$ since $R(t)$ decreases.

The Exponential Failure Model

Assumption: Many electronic components have a constant failure rate $\lambda$ during their useful life (the "flat" part of the bathtub curve, after infant mortality and before wear-out).

Starting from the failure rate definition: $\lambda = -\frac{1}{R(t)} \frac{dR}{dt}$

Since $\lambda$ is constant: $\frac{dR}{dt} = -\lambda R$

This is a separable differential equation: $\frac{dR}{R} = -\lambda \, dt$

Integrate both sides: $\int_{R(0)}^{R(t)} \frac{dR}{R} = -\lambda \int_0^t dt$

$\ln R(t) - \ln R(0) = -\lambda t$

Since $R(0) = 1$: $\ln R(t) = -\lambda t$

$\boxed{R(t) = e^{-\lambda t}}$

Physical meaning: The probability of survival decays exponentially. Higher $\lambda$ means faster decay (less reliable). This is the same mathematical form as radioactive decay!

Mean Time To Failure (MTTF)

The MTTF is the expected value of the time-to-failure random variable $T$:

$\text{MTTF} = E[T] = \int_0^\infty t \cdot f(t) \, dt$

where $f(t)$ is the probability density function of failure time.

Relationship between f(t) and R(t): The PDF is the derivative of the cumulative distribution function: $f(t) = \frac{d}{dt}[1 - R(t)] = -\frac{dR}{dt}$

For exponential reliability $R(t) = e^{-\lambda t}$: $f(t) = -\frac{d}{dt}(e^{-\lambda t}) = \lambda e^{-\lambda t}$

Now compute MTTF: $\text{MTTF} = \int_0^\infty t \cdot \lambda e^{-\lambda t} \, dt$

Integration by parts: Let $u = t$, $dv = \lambda e^{-\lambda t} dt$

• $du = dt$
• $v = -e^{-\lambda t}$

$\text{MTTF} = \left[ -t e^{-\lambda t} \right]_0^\infty + \int_0^\infty e^{-\lambda t} \, dt$

The first term: $\lim_{t \to \infty} -t e^{-\lambda t} = 0$ (exponential dominates), and at $t=0$ it's zero.

The second term: $\int_0^\infty e^{-\lambda t} \, dt = \left[ -\frac{1}{\lambda} e^{-\lambda t} \right]_0^\infty = \frac{1}{\lambda}$

$\boxed{\text{MTTF} = \frac{1}{\lambda}}$

Intuition: If failures happen at rate $\lambda$ per hour, on average we wait $1/\lambda$ hours for the first failure. If $\lambda = 0.001$ failures/hour, MTTF = 1000 hours.

Mean Time Between Failures (MTBF)

For systems with constant failure rate and negligible repair time:

$\boxed{\text{MTBF} = \frac{1}{\lambda}}$

MTTF vs MTBF distinction:

• MTTF: Non-repairable items (satellites, consumables) — time until death
• MTBF: Repairable systems (spacecraft subsystems with ground control, IS modules) — average uptime between fixes

In practice, for exponential models: $\text{MTBF} = \text{MTTF} + \text{MTTR}$

where MTTR (Mean Time To Repair) is the average repair time. For spacecraft with long/impossible repair, we often just use MTTF.

Worked Examples

A spacecraft's solar array has a failure rate of $\lambda = 0.00005$ failures/hour.

a) What is the reliability for a 5-year mission?

Convert 5 years to hours: $t = 5 \times 365.25 \times 24 = 43,830 \text{ hours}$

Apply exponential model: $R(t) = e^{-\lambda t} = e^{-0.00005 \times 43,830}$ $R(t) = e^{-2.1915} = 0.112$

Answer: 11.2% chance of surviving 5 years (88.8% chance of failure).

Why this step? We directly apply the exponential reliability formula. The product $\lambda t$ is dimensionless (failures/hour × hours = failures), giving a pure number for the exponent. Note $0.00005 \times 43{,}830 = 2.1915$.

b) What is the MTTF?

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

Convert to years: $\text{MTTF} = \frac{20,000}{24 \times 365.25} \approx 2.28 \text{ years}$

Answer: Expected lifetime is 2.28 years. Since the mission is 5 years and MTTF is only 2.28 years, we'd need redundancy!

A Mars orbiter needs 95% reliability for a 2-year mission. What maximum failure rate is acceptable?

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

Find: $\lambda$

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

Take natural log: $\ln(0.95) = -\lambda \times 17,532$ $-0.05129 = -\lambda \times 17,532$ $\lambda = \frac{0.05129}{17,532} = 2.926 \times 10^{-6} \text{ failures/hour}$

Answer: Maximum acceptable failure rate is $2.93 \times 10^{-6}$ failures/hour, or equivalently MTTF ≥ 342,000 hours (39 years).

Why this step? We inverted the exponential function using logarithms. The log "unwraps" the exponential relationship between R and $\lambda$.

Two identical power supplies are used in parallel (system works if at least one works). Each has $\lambda = 0.0001$ failures/hour. What is system reliability for 10,000 hours?

Single component reliability: $R_1(t) = e^{-0.0001 \times 10,000} = e^{-1} = 0.368$

Parallel system: System fails only if BOTH fail. $R_{\text{sys}} = 1 - P(\text{both fail})$

For independent components: $P(\text{both fail}) = [1 - R_1(t)]^2$

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

Answer: System reliability is 60.1%, much better than the 36.8% for a single unit.

Why redundancy works: Even though each unit is more likely to fail than survive, the chance that BOTH fail is the product of two probabilities, which is smaller. Redundancy is crucial for long-duration missions.

Common Mistakes

Wrong thinking: "MTTF = 1000 hours means the component will last 1000 hours."

Why it feels right: MTTF sounds like a warranty or design life specification.

The truth: MTTF is an average over many components. Half will fail before MTTF, half after. Specifically, for exponential model: $R(\text{MTTF}) = R(1/\lambda) = e^{-\lambda \cdot 1/\lambda} = e^{-1} \approx 0.37$

Only 37% of components survive to their MTTF! 63% fail earlier.

Fix: Use reliability curves $R(t)$ for mission planning, not just MTTF. To achieve 90% reliability, you need $t \ll \text{MTTF}$.

Wrong thinking: "Two redundant systems with $\lambda = 0.001$ each have combined $\lambda = 0.002$."

Why it feels right: We often add things in series, and failure rates do add for series systems.

The truth: For parallel redundancy, you cannot simply add $\lambda$ values. You must use: $R_{\text{parallel}} = 1 - \prod_{i=1}^n (1 - R_i(t))$

For identical components with redundancy factor $n$: $R_{\text{parallel}} = 1 - (1 - e^{-\lambda t})^n$

This does NOT have a simple constant failure rate!

Fix: Always work with reliability functions $R(t)$ for redundant systems, then compute effective failure rate if needed.

Wrong thinking: "All spacecraft components follow exponential reliability."

Why it feels right: Exponential model is simple and widely taught.

The truth: Exponential model assumes constant failure rate (memoryless failures, like random cosmic ray strikes). Mechanical parts (bearings, actuators) have increasing failure rate due to wear (Weibull distribution). Batteries degrade predictably (different model).

Bathtub curve:

• Infant mortality (decreasing $\lambda$) — manufacturing defects
• Useful life (constant $\lambda$) — random failures, exponential model applies
• Wear-out (increasing $\lambda$) — mechanical aging

Fix: Use exponential model only for electronic components in their useful life phase. For mechanisms, use Weibull distribution; for batteries, use cycle-life models.

Active Recall Questions

#flashcards/physics

What is the definition of reliability R(t)? :: The probability that a system operates without failure from time 0 to time t. $R(t) = P(\text{survival until } t)$, with $R(0) = 1$ and $0 \leq R(t) \leq 1$.

MTTF (Mean Time To Failure) is like asking: "If I had 100 identical robots, on average how long until each one breaks?" If MTTF is 10 years, that means on average they last 10 years. But important: some break at 2 years, some last 20 years! It's an average, not a guarantee.

The exponential model is like rolling dice every hour. If you roll a 1, the robot breaks (low chance). Each hour is independent — the robot doesn't "remember" it's been working for 5 years and get tired. This works great for electronics (random failures) but NOT for things that wear out like bike chains (those DO get tired from use).

Spacecraft engineers use these formulas because they can't fix things in space! If a satellite breaks, it's game over. So they calculate: "What's the chance this lasts 5 years?" and add backup systems if needed.

Mnemonic: Low Lambda = Long Life

Small λ (failure rate) means big MTTF (long life). They're reciprocals — flip the fraction!

Visual: $\lambda$ is failures per hour (steep downhill), MTTF is hours per failure (long flat road). Inverse relationship.

Connections

• Bathtub Curve — failure rate vs. time, three life phases
• Weibull Distribution — generalized failure model for wear-out
• Series vs. Parallel System Reliability — how to combine component reliabilities
• Fault Tree Analysis — top-down reliability analysis from system to components
• Redundancy Design — active vs. standby redundancy for high reliability
• Probability and Statistics Fundamentals — exponential distribution, expected value
• Mission Design Constraints — how reliability drives mass, power, cost budgets
• Poisson Process — mathematical foundation for constant-rate random events
• Availability vs. Reliability — incorporating repair time and uptime percentage
• Safety-Critical Systems — when you need 99.999% reliability (five nines)

Master these reliability concepts and you can quantify mission risk, justify redundancy decisions, and speak the language of systems engineers. Every spacecraft is a gamble against these exponentials — know your odds.

Concept Map

Hinglish (regional understanding)