Series and Parallel Reliability Systems

Why System Structure Matters

A system can contain several components but still have a simple success rule. A series system works only if every required component works; a parallel system works if at least one of its redundant components works. Those different rules lead to different probability calculations, even when the component reliabilities are identical.

Reliability calculations begin by stating exactly what counts as system success. Then represent each component’s success as an event, use the system’s structure to form the system-success event, and apply the assumed independence of components. This tutorial uses a binary model: each component either succeeds or fails during the period being studied.

Definition: For component \(i\), let \(S_i\) be the event that the component succeeds during the specified period, and let its reliability be \(r_i=P(S_i)\). Its failure probability is \(1-r_i\). A series system succeeds if all its components succeed. A parallel system succeeds if at least one of its components succeeds.

The period and meaning of success must be clear. For example, a hypothetical backup pump might be counted as successful if it can deliver the required flow when called upon during a specified operating period. Reliability is the probability of that event under the stated model; it is not a claim that the component will work for every possible use or condition.

Series Systems

Suppose a system contains \(n\) components in series, and every component is required for the system to work. Its success event is the intersection of all component-success events:

S_{\text{series}}=\bigcap_{i=1}^{n} S_i.

If the component-success events are mutually independent, the probability of that intersection is the product of their probabilities. Thus, the system reliability is:

R_{\text{series}}=P(S_{\text{series}})=\prod_{i=1}^{n}r_i.

The independence assumption is essential to this product formula. Knowing only the individual component reliabilities is not generally enough to find the probability that all components succeed if their outcomes may be dependent.

Worked Example: A Three-Component Series System

A hypothetical monitoring unit requires a sensor, a processor, and a power module all to operate during a one-month period. Assume their success events are mutually independent. The component reliabilities for that period are \(0.98\), \(0.95\), and \(0.90\), respectively. Find the system reliability and determine whether it meets a target reliability of \(0.84\).

Method and model. This is a series system, so all three components must succeed. Under the stated independent Bernoulli success model, the system reliability is the product of the three component reliabilities:

R_{\text{series}}=0.98(0.95)(0.90).

First, \(0.98(0.95)=0.931\). Then:

R_{\text{series}}=0.931(0.90)=0.8379.

The probability that the system fails is the complement:

$$ 1-0.8379=0.1621. $$

The reliability is \(0.8379\), which is below \(0.84\), so the system does not meet the target. No rounding is needed. The common mistake to avoid is adding the component failure probabilities as if that were the exact system failure probability. Failures can occur together, so the exact system reliability follows from the intersection and the independence product.

Parallel Systems

A parallel system succeeds as long as at least one component succeeds. Its failure event is therefore that every component fails. If the component-success events are mutually independent, their failure events are also independent, so the probability of joint failure is the product of the individual failure probabilities. Taking the complement gives:

R_{\text{parallel}} =1-P(\text{all components fail}) =1-\prod_{i=1}^{n}(1-r_i).

The complement is usually the simplest route: calculate the probability that all parallel components fail, then subtract it from one. This formula assumes that any one component’s success is enough for the system to meet its success criterion. If the system requires two components to work, or if components share a resource whose failure disables them all, the simple parallel model does not apply as stated.

Worked Example: Two Independent Backup Units

A hypothetical safety installation has two independent backup units. During a specified test period, the first unit succeeds with probability \(0.80\) and the second succeeds with probability \(0.90\). The installation succeeds if at least one unit succeeds. Find its reliability and the probability that both units fail.

Method and model. This is a parallel system with independent Bernoulli success events. The failure probabilities are \(1-0.80=0.20\) and \(1-0.90=0.10\). Both units must fail for the installation to fail:

P(\text{both fail})=(0.20)(0.10)=0.02.

Therefore, the installation reliability is:

R_{\text{parallel}}=1-0.02=0.98.

The two requested probabilities are \(0.02\) for both units failing and \(0.98\) for system success; they sum to \(1\), as complementary outcomes must. The common mistake to avoid is multiplying the success probabilities, \(0.80(0.90)\). That product is the probability that both units succeed, not the probability that at least one succeeds.

Combining Series and Parallel Blocks

Many systems use both structures. The reliable approach is to translate the whole configuration into its success condition before calculating. A parallel block succeeds if at least one of its components succeeds; a series connection of blocks succeeds only if every block succeeds.

When blocks contain disjoint sets of mutually independent components, their outcomes are independent. The block reliabilities can then be combined using the series or parallel rule. In particular, a series connection of two independent blocks with reliabilities \(R_1\) and \(R_2\) has reliability \(R_1R_2\). A parallel connection of those blocks has reliability \(1-(1-R_1)(1-R_2)\). If blocks share components, their outcomes need not be independent, so multiplying block reliabilities may be invalid.

Write the success rule.
State which components or blocks must work and where one working component is enough.

Calculate simple block reliabilities.
Use a product for a series block and one minus the product of failure probabilities for a parallel block.

Combine the blocks.
Apply the rule matching the overall structure, checking that the relevant block outcomes are independent before multiplying.

Check the result.
A reliability must lie between zero and one. Also verify that the chosen success event matches the system’s stated operating rule.

Worked Example: A Series Component and a Parallel Pair

A hypothetical device operates if a controller works and at least one of two independent valves works. The controller reliability is \(0.96\); the two valve reliabilities are \(0.85\) and \(0.90\). Assume all three component-success events are mutually independent. Find the device reliability.

Method and model. First calculate the reliability of the parallel valve pair. Both valves fail with probability \(1-0.85=0.15\) and \(1-0.90=0.10\). Their joint failure probability is:

$$ (0.15)(0.10)=0.015. $$

Thus, the valve pair succeeds with probability:

R_{\text{valves}}=1-0.015=0.985.

The controller is in series with this pair, so both the controller and the valve block must succeed. The components are mutually independent, and the blocks use separate components. Therefore:

R_{\text{device}}=(0.96)(0.985)=0.9456.

The device reliability is \(0.9456\), and its failure probability is \(1-0.9456=0.0544\). These values sum to \(1\), providing a check. The common mistake to avoid is treating the three components as all being in series; the device does not require both valves to work.

Why the Product Rules Hold

The formulas come directly from the event definitions and independence. For a series system, success is the intersection of all component-success events. Mutual independence means the probability of that intersection is the product of the individual probabilities. For a parallel system, it is often easier to take the complement of system success: the complement occurs exactly when every component fails. Independence of the component-success events also gives independence of their complements, so the probability of joint failure is the product of the failure probabilities.

\begin{aligned} R_{\text{series}} &=P\left(\bigcap_{i=1}^{n} S_i\right) =\prod_{i=1}^{n}P(S_i) =\prod_{i=1}^{n}r_i,\\ R_{\text{parallel}} &=1-P\left(\bigcap_{i=1}^{n} S_i^c\right) =1-\prod_{i=1}^{n}P(S_i^c) =1-\prod_{i=1}^{n}(1-r_i). \end{aligned}

Here \(S_i^c\) denotes the event that component \(i\) fails. The derivation uses mutual independence, not merely the fact that component reliabilities have been supplied. A shared power supply, common environmental exposure, or another common cause can make failures dependent. In that situation, the system structure still identifies the relevant event, but joint probabilities must be determined from an appropriate dependence model rather than inferred from the marginal reliabilities alone.

Common Reliability Traps

Using the wrong success condition. A series system requires every component; a parallel system requires at least one. Translate the words into an event before calculating.
Multiplying success probabilities for a parallel system. That gives the probability that all components succeed. To find parallel reliability, multiply failure probabilities and subtract from one.
Assuming independence without justification. Individual reliabilities do not determine a joint probability when component outcomes are dependent.
Combining blocks that share components as if independent. Independence of disjoint component events supports the usual block calculation; shared components may create dependence.
Ignoring the specified period. A reliability must refer to a defined operating period and success criterion. Probabilities for different periods or definitions should not be combined as though they describe the same event.

These calculations are useful in evaluating redundancy and identifying bottlenecks in hypothetical systems. A series system’s reliability cannot exceed the reliability of its least reliable component, because system success requires that component to succeed. A parallel system’s reliability cannot be below that of any individual component, because the system succeeds whenever that component succeeds. Such bounds provide quick checks on arithmetic and on whether the configuration has been interpreted correctly.

Key takeaway: Model component success events, state the independence assumption, and match the calculation to the system logic. Series reliability is the product of success probabilities; parallel reliability is one minus the product of failure probabilities. Combined systems require both the correct block structure and justified independence.

Check Your Understanding

For each problem, assume the component success events are mutually independent unless stated otherwise.

A series system has three components with reliabilities \(0.97\), \(0.96\), and \(0.98\). Find the system reliability. (Answer: \(0.912576\).)
Three backup sensors have reliabilities \(0.88\), \(0.75\), and \(0.60\). The system succeeds if at least one sensor works. Find its reliability. (Answer: \(0.9880\).)
A device succeeds if at least one component in the first parallel pair works and at least one component in a second parallel pair works. The first pair has reliabilities \(0.90\) and \(0.80\); the second pair has reliabilities \(0.70\) and \(0.60\). Find the device reliability. (Answer: \(0.8624\).)
A controller in series with a parallel pair is required for a system to work. Its reliability is \(0.95\), and the two parallel components have reliabilities \(0.80\) and \(0.70\). Find the system reliability. (Answer: \(0.8930\).)
A system has two parallel branches, each branch consisting of two components in series. The first branch component reliabilities are \(0.90\) and \(0.80\); the second branch reliabilities are \(0.75\) and \(0.60\). Find the system reliability. (Answer: \(0.8460\).)
A series system has component reliabilities \(0.90\), \(0.95\), and \(x\). What is the smallest value of \(x\) that gives system reliability at least \(0.80\)? (Answer: \(x\geq 0.9357\), approximately.)

What You'll Learn