23.2 System Reliability Configurations: Series, Parallel, and k-out-of-n Systems

Industrial systems—ranging from automated manufacturing transfer lines to petrochemical refineries and aerospace flight controls—consist of interconnected assemblies of individual components. System reliability analysis uses Reliability Block Diagrams (RBD) to model how component survival or failure propagates to the overall system level, assuming component failures are statistically independent.

---

1. Series System Reliability (Weakest-Link Architecture)

In a series configuration, every component is mission-critical: the system functions if and only if all $n$ components operate successfully. If any single component fails, the entire system fails.

                         Series Configuration RBD

     ┌──────────┐     ┌──────────┐     ┌──────────┐         ┌──────────┐
  ──►│ Comp 1   │────►│ Comp 2   │────►│ Comp 3   │─── ··· ─►│ Comp n   ├──►
     │  R₁(t)   │     │  R₂(t)   │     │  R₃(t)   │         │  R_n(t)  │
     └──────────┘     └──────────┘     └──────────┘         └──────────┘

Mathematical Formulation

Assuming component failures are statistically independent events: $R_{\text{sys}}(t) = P(T_1 > t \cap T_2 > t \cap \dots \cap T_n > t) = \prod_{i=1}^{n} R_i(t)$ Because each individual reliability $R_i(t) \le 1.0$: $R_{\text{sys}}(t) \le \min\{R_1(t), R_2(t), \dots, R_n(t)\}$ Adding components in series monotonically degrades overall system reliability. For example, a system with 50 components, each having $R = 0.99$, achieves a total system reliability of only $R_{\text{sys}} = (0.99)^{50} \approx 0.605$ (60.5%).

Series Systems with Exponential Components

When all $n$ components exhibit constant failure rates $\lambda_1, \lambda_2, \dots, \lambda_n$: $R_{\text{sys}}(t) = \prod_{i=1}^{n} e^{-\lambda_i t} = \exp\left( -\left[ \sum_{i=1}^{n} \lambda_i \right] t \right) = e^{-\lambda_{\text{sys}} t}$ The equivalent system failure rate is the direct arithmetic sum of individual failure rates: $\lambda_{\text{sys}} = \sum_{i=1}^{n} \lambda_i$ Correspondingly, the system Mean Time to Failure is: $\text{MTTF}_{\text{sys}} = \frac{1}{\lambda_{\text{sys}}} = \frac{1}{\sum_{i=1}^{n} \lambda_i}$

---

2. Parallel System Reliability (Active Redundancy)

In a parallel configuration with active redundancy, all $n$ components operate concurrently. The system succeeds if at least one component survives, and fails only if all $n$ components fail simultaneously.

                         Parallel Configuration RBD

                        ┌──────────┐
                   ┌───►│  Comp 1  ├───┐
                   │    │  R₁(t)   │   │
                   │    └──────────┘   │
                   │    ┌──────────┐   │
                 ──┼───►│  Comp 2  ├───┼──►
                   │    │  R₂(t)   │   │
                   │    └──────────┘   │
                   │         :         │
                   │    ┌──────────┐   │
                   └───►│  Comp n  ├───┘
                        │  R_n(t)  │
                        └──────────┘

Mathematical Formulation

System failure occurs when all components fail. By independence, system unreliability $F_{\text{sys}}(t)$ is the product of component unreliabilities: $F_{\text{sys}}(t) = \prod_{i=1}^{n} F_i(t) = \prod_{i=1}^{n} \left[1 - R_i(t)\right]$ Applying the fundamental identity $R_{\text{sys}}(t) = 1 - F_{\text{sys}}(t)$: $R_{\text{sys}}(t) = 1 - \prod_{i=1}^{n} \left[1 - R_i(t)\right]$ For $n$ identical components, each with reliability $R$: $R_{\text{sys}} = 1 - (1 - R)^n$

Parallel System MTTF (Identical Exponential Components)

For $n$ identical components with constant failure rate $\lambda$, the system reliability over time is: $R_{\text{sys}}(t) = 1 - (1 - e^{-\lambda t})^n$ Integrating $R_{\text{sys}}(t)$ from $0$ to $\infty$ yields the classic harmonic series equation for active parallel MTTF: $\text{MTTF}_{\text{sys}} = \frac{1}{\lambda} \sum_{i=1}^{n} \frac{1}{i} = \frac{1}{\lambda} \left( 1 + \frac{1}{2} + \frac{1}{3} + \dots + \frac{1}{n} \right)$

• For $n = 1$: $\text{MTTF} = 1/\lambda = 1.000 / \lambda$
• For $n = 2$: $\text{MTTF} = (1 + 1/2)/\lambda = 1.500 / \lambda$
• For $n = 3$: $\text{MTTF} = (1 + 1/2 + 1/3)/\lambda = 1.833 / \lambda$
• For $n = 4$: $\text{MTTF} = (1 + 1/2 + 1/3 + 1/4)/\lambda = 2.083 / \lambda$

[!NOTE] Notice the dramatic diminishing marginal returns of adding redundant active units: adding a second unit increases MTTF by 50%, a third adds 33.3%, and a fourth adds only 25% of baseline life.

Non-Constant Hazard Rate in Parallel Systems

Even when each individual component possesses a strictly constant failure rate $\lambda$, a parallel system does not have a constant failure rate. At $t = 0$, all units are healthy, so the instantaneous probability of total system failure is near zero ($h_{\text{sys}}(0) = 0$). As operating time accumulates and individual units fail, the hazard rate monotonically increases over time. Therefore, never use $1/\lambda_{\text{sys}}$ for parallel systems.

---

3. Standby Redundancy (Cold, Warm, and Hot)

In standby redundancy, secondary (backup) units do not operate concurrently under full load. Instead, a sensing and switching circuit detects failure of the primary unit and routes electrical or mechanical power to an offline backup.

                      Standby Redundancy Architecture

                           ┌──────────────┐
                           │ Primary Unit │
                     ┌────►│    (λ)       ├─────┐
                     │     └──────────────┘     │
                     │                          ▼
  ─── Power / Signal ┤                 ┌─────────────────┐
                     │                 │ Sensing & Switch│──► System Output
                     │     ┌───────────┤   Mechanism     │
                     └────►│  Standby  └─────────────────┘
                           │   Unit    │
                           └───────────┘

Standby Operational Classifications

• Cold Standby: The backup unit remains unpowered and completely dormant until the primary fails. Its failure rate while dormant is zero ($\lambda_d = 0$). It suffers zero wear or aging while in storage.
• Warm Standby: The backup is energized in a reduced-power or idling state, allowing faster transition than cold standby. Its dormant failure rate is non-zero but lower than active operation ($0 < \lambda_d < \lambda$).
• Hot Standby: The backup is continuously energized and synchronized with the primary, providing instantaneous transfer without process interruption. Mathematically, hot standby behaves like active parallel redundancy.

Cold Standby with Perfect Switching (Exponential Components)

Consider a system with 1 primary unit and $n-1$ cold standby backups (total $n$ units), each with operating failure rate $\lambda$. Assume the sensing/switching mechanism is 100% reliable ($p_s = 1.0$). Because units operate sequentially, total time to failure is the sum of $n$ independent, identically distributed exponential random variables. The number of failures occurring by time $t$ follows a Poisson distribution with parameter $\lambda t$. The system survives time $t$ if the total number of failures is at most $n-1$: $R_{\text{sys}}(t) = \sum_{i=0}^{n-1} \frac{(\lambda t)^i e^{-\lambda t}}{i!} = e^{-\lambda t} \sum_{i=0}^{n-1} \frac{(\lambda t)^i}{i!}$ For the common case of 1 active unit and 1 cold standby ($n = 2$): $R_{\text{sys}}(t) = e^{-\lambda t} (1 + \lambda t)$ The MTTF of $n$ sequential cold standby units is strictly additive: $\text{MTTF}_{\text{sys}} = \int_{0}^{\infty} R_{\text{sys}}(t)\,dt = \frac{n}{\lambda}$ For $n = 2$, $\text{MTTF} = 2/\lambda$. Comparing cold standby to active parallel redundancy: $\text{MTTF}_{\text{cold standby}} = \frac{2}{\lambda} > \text{MTTF}_{\text{active parallel}} = \frac{1.5}{\lambda}$ Cold standby always yields higher reliability and MTTF than active parallel redundancy because the backup unit does not experience operational wear while the primary is functioning.

Imperfect Switching

If the sensing and switching mechanism has a probability of successful operation $p_s < 1.0$, the reliability for 1 primary and 1 cold standby becomes: $R_{\text{sys}}(t) = e^{-\lambda t} + p_s \lambda t e^{-\lambda t} = e^{-\lambda t} (1 + p_s \lambda t)$

---

4. $k$-out-of-$n$ System Reliability

A $k$-out-of-$n$ system represents a generalized redundancy model where the system functions if and only if at least $k$ out of $n$ identical, independent components are operating successfully.

Mathematical Formulation

Because the components are identical and independent, the number of surviving components $X$ follows a Binomial Distribution with parameters $n$ and component reliability $R$: $R_{k/n} = P(X \ge k) = \sum_{i=k}^{n} \binom{n}{i} R^i (1 - R)^{n-i}$ where the binomial coefficient is: $\binom{n}{i} = \frac{n!}{i! (n - i)!}$

Boundary and Special Cases

• When $k = n$: $R_{n/n} = \binom{n}{n} R^n (1-R)^0 = R^n$ (Series System).
• When $k = 1$: $R_{1/n} = 1 - \binom{n}{0} R^0 (1-R)^n = 1 - (1 - R)^n$ (Active Parallel System).
• Triple Modular Redundancy (TMR) / 2-out-of-3 System: Used extensively in voting logic for flight computers and nuclear safety shutdown systems ($k = 2, n = 3$): $R_{2/3} = \binom{3}{2} R^2 (1 - R)^{3-2} + \binom{3}{3} R^3 (1 - R)^{3-3} = 3 R^2 (1 - R) + R^3 = 3 R^2 - 2 R^3$

---

5. Combined Series-Parallel System Reduction

Most industrial systems combine series and parallel sub-networks. Reliability block diagram reduction solves these problems through step-by-step modular substitution.

                      System Block Diagram Reduction Process

  Original Network:                     Step 1: Reduce Parallel Sub-blocks
  ┌───[ A ]───┐                           ┌───[ R_A ]───┐
  │           │                           │             ├───[ R_C ]───►
──┼───[ B ]───┼───[ C ]───►             ──┴───[ R_P ]───┘
  │           │                           where R_P = 1 - (1-R_A)(1-R_B)
  └───[ B' ]──┘
                                        Step 2: Reduce Remaining Series
                                        ──►[ R_sys = R_P · R_C ]──►

Low-Level vs. High-Level Redundancy

A classic principle in industrial reliability design is that component-level (low-level) redundancy is always superior to system-level (high-level) redundancy:

  High-Level Redundancy (System Level):       Low-Level Redundancy (Component Level):
  ┌───[ A₁ ]───[ B₁ ]───┐                     ┌───[ A₁ ]───┐   ┌───[ B₁ ]───┐
──┼                     ├──►                ──┼            ├───┼            ├──►
  └───[ A₂ ]───[ B₂ ]───┘                     └───[ A₂ ]───┘   └───[ B₂ ]───┘
  R_high = 1 - (1 - R_A · R_B)²               R_low = [1 - (1 - R_A)²] · [1 - (1 - R_B)²]

For example, if $R_A = R_B = 0.80$:

• Series branch reliability: $R_{\text{branch}} = (0.80)(0.80) = 0.64$.
• High-level system redundancy: $R_{\text{high}} = 1 - (1 - 0.64)^2 = 1 - (0.36)^2 = 1 - 0.1296 = 0.8704$.
• Low-level component redundancy: $R_{\text{comp}, A} = 1 - (1 - 0.80)^2 = 0.96$. System reliability: $R_{\text{low}} = (0.96)(0.96) = 0.9216$.
• Result: $R_{\text{low}} (92.16%) > R_{\text{high}} (87.04%)$. Duplicating lower-level components isolates failures so that a failure in component $A_1$ does not disable component $B_1$.

---

6. Step-by-Step Worked Engineering Calculations

Worked Example 23.2.1: Hybrid Network Reduction

Problem: An automated paint-curing facility relies on a control system structured as follows: an incoming power stage consists of two parallel supplies ($S_1$ and $S_2$, each with $R_S = 0.95$). Downstream from the power supplies is a processing unit comprising three temperature sensors arranged in a 2-out-of-3 voting configuration ($R_{\text{sensor}} = 0.90$ each). Finally, the entire assembly connects in series to a master emergency shutdown valve ($R_V = 0.98$). Find the overall system reliability $R_{\text{sys}}$.

Solution:

Step 1: Reduce the Parallel Power Stage ($S$) $R_{\text{power}} = 1 - (1 - R_S)^2 = 1 - (1 - 0.95)^2 = 1 - (0.05)^2 = 1 - 0.0025 = 0.9975$

Step 2: Reduce the 2-out-of-3 Sensor Stage ($K$) $R_{\text{sensor-stage}} = 3 R_{\text{sensor}}^2 - 2 R_{\text{sensor}}^3 = 3(0.90)^2 - 2(0.90)^3 = 3(0.81) - 2(0.729) = 2.43 - 1.458 = 0.9720$

Step 3: Combine All Stages in Series The overall system is a series combination of the power stage, sensor stage, and shutdown valve: $R_{\text{sys}} = R_{\text{power}} \times R_{\text{sensor-stage}} \times R_V$ $R_{\text{sys}} = (0.9975) \times (0.9720) \times (0.98) = 0.95018 \approx 0.9502 \quad (95.02\%)$

---

Worked Example 23.2.2: Active Parallel vs. Cold Standby MTTF

Problem: A critical avionics flight computer has a primary processor with a constant failure rate $\lambda = 0.0005\text{ failures/hr}$. An engineering team can add one identical backup unit as either (a) an active parallel redundant processor or (b) a cold standby processor with a perfect sensor/switch mechanism. Compare the resulting system reliability over a $1,000\text{-hour}$ mission and the system MTTF under both configurations.

Solution:

Component Baseline Values: $\text{MTTF}_{\text{single}} = \frac{1}{0.0005} = 2,000\text{ hours}$ $R(1,000) = e^{-(0.0005)(1,000)} = e^{-0.50} \approx 0.60653$

Option A: Active Parallel Redundancy: $R_{\text{parallel}}(1,000) = 1 - [1 - R(1,000)]^2 = 1 - (1 - 0.60653)^2 = 1 - (0.39347)^2 = 1 - 0.15482 = 0.84518 \approx 0.8452$ $\text{MTTF}_{\text{parallel}} = \frac{1}{\lambda}\left(1 + \frac{1}{2}\right) = 2,000 \times 1.5 = 3,000\text{ hours}$

Option B: Cold Standby Redundancy (Perfect Switch): $R_{\text{standby}}(1,000) = e^{-\lambda t}(1 + \lambda t) = e^{-0.50}(1 + 0.50) = (0.60653)(1.50) = 0.90980 \approx 0.9098$ $\text{MTTF}_{\text{standby}} = \frac{n}{\lambda} = \frac{2}{0.0005} = 4,000\text{ hours}$

Comparison: Cold standby yields a mission reliability of 90.98% (versus 84.52% for active parallel) and an MTTF of 4,000 hours (versus 3,000 hours for active parallel). Standby eliminates in-service aging of the backup while the primary operates.

---

7. NCEES Reference Handbook Tips & Realistic Exam Traps

• Series Systems Always Weaken Reliability: If an exam problem adds any component in series—regardless of whether its reliability is 0.999—the system reliability must decrease. If your computed $R_{\text{sys}}$ is higher than any series component, check your math immediately.
• Diminishing Returns in Active Redundancy: Redundancy adds cost, weight, and volume. Moving from $n=1$ to $n=2$ increases MTTF by 50%, but going from $n=2$ to $n=3$ yields only an additional 33.3% gain, and $n=3$ to $n=4$ adds only 25%.
• Common Cause Failures (CCF): Standard RBD calculations assume component failures are completely independent. In real-world industrial settings, extreme external conditions (e.g., a plant fire, a flood, power bus collapse, or common software bug) can disable parallel units simultaneously. In such cases, actual reliability is lower than the independence model predicts.
• Don't Forget the Binomial Combinations in $k$-out-of-$n$: A frequent error on the FE exam is computing $R^k(1-R)^{n-k}$ without multiplying by the combinatorial coefficient $\binom{n}{i}$. For a 2-out-of-3 system, there are $\binom{3}{2} = 3$ distinct ways for exactly two components to survive.