Fault Tree Probability Calculation: How-To

Fault tree probability calculation works out the chance that a top event (usually a system failure) occurs by combining the probabilities of basic events using Boolean logic. You enter the failure probabilities of the basic events and the tree’s logical structure (AND/OR gates), and the calculator returns the top event probability. This article explains the math, shows a worked example, and covers the common pitfalls.

Formula

The top event probability is computed by reducing the fault tree to a set of Boolean equations. For an AND gate, the output occurs only if all inputs occur, so the probability is the product of the input probabilities:

P(AND) = P(A) × P(B)

For an OR gate, the output occurs if at least one input occurs. Assuming independent basic events, the probability is:

P(OR) = 1 − (1 − P(A)) × (1 − P(B))

In the example above, the tree has two branches: an AND gate with B1 and B2, and a single basic event B3, both feeding an OR gate. The formula becomes:

P(top) = 1 − (1 − P(B1)×P(B2)) × (1 − P(B3))

Each P(Bi) is the probability that basic event i occurs during the mission time (e.g., a component fails). All probabilities are unitless numbers between 0 and 1.

Worked Example

Using the default values: P(B1) = 0.01, P(B2) = 0.02, P(B3) = 0.03.

1. Calculate the AND gate: 0.01 × 0.02 = 0.0002
2. Calculate the OR gate: 1 − (1 − 0.0002) × (1 − 0.03) = 1 − 0.9998 × 0.97 = 1 − 0.969806 = 0.030194
3. Rounded to 4 decimals: 0.0302

The top event probability is 0.0302, or about 3.02%.

When to Use a Fault Tree

Fault tree analysis (FTA) is used in reliability engineering, safety analysis, and risk assessment—especially in aerospace, nuclear power, chemical processing, and automotive safety. Use it when you need to understand how combinations of lower-level failures lead to a specific top event (e.g., “loss of coolant” or “airbag fails to deploy”). It is a top-down, deductive approach: you start with the undesired event and work backward to root causes.

FTA is particularly useful when:

• You have a well-defined top event that can be broken into logical causes.
• You have probability data for basic events (from field data, testing, or expert judgment).
• You need to identify single points of failure or common cause vulnerabilities.
• You want to quantify the improvement from adding redundancy or changing maintenance schedules.

How to Build a Fault Tree

1. Define the top event precisely (e.g., “pump system fails to deliver flow”).
2. Identify the immediate causes and connect them with AND or OR gates. An AND gate means all causes must occur; an OR gate means any one cause suffices.
3. Continue decomposing each cause until you reach basic events that have known probabilities (component failures, human errors, etc.).
4. Assign a probability to each basic event. This is often a failure rate (λ) multiplied by mission time (t) for non-repairable components: P = 1 − e^(−λt), which approximates λt when λt is small.
5. Compute the top event probability using Boolean algebra, either by hand for small trees or with software for large ones.

Common Mistakes and Misconceptions

A frequent error is treating OR gates as simple addition. For two events, P(A OR B) = P(A) + P(B) only if the events are mutually exclusive, which is rarely true in fault trees. The correct formula subtracts the intersection: P(A OR B) = P(A) + P(B) − P(A AND B). If you use the product for the intersection (assuming independence), you get the 1 − (1−P(A))(1−P(B)) form.

Another common mistake is ignoring dependencies. The formulas above assume basic events are independent. In real systems, common cause failures (e.g., a single power supply affects multiple components) violate this assumption. If dependencies exist, you need more advanced methods like beta factor or Markov analysis.

Also, be careful with units of time. Failure probabilities are often given per hour or per year. If your mission time is 10 hours and your failure rate is per 1,000 hours, convert appropriately. Mixing units leads to wildly wrong results.

Finally, remember that the fault tree only models the logic you put into it. If you omit a failure mode or a gate, the result will be optimistic. The output is only as good as the tree structure and the input probabilities.

Interpreting the Result

The top event probability is a number between 0 and 1. A value of 0.03 means there is a 3% chance the top event occurs during the specified mission time. This is not a frequency; it is a probability. If you need a failure rate (failures per hour), you would divide by the mission time, but only if the system is non-repairable and the probability is small.

Often, you also want to know which basic events contribute most to the top event. That is called importance analysis. The most common measure is Fussell-Vesely importance, which is the fraction of top event probability that involves a given basic event. This helps prioritize maintenance or design changes. The calculator above does not compute importance, but you can get a sense by varying each input and seeing how the output changes.

Limitations and Alternatives

Fault tree analysis assumes that events are binary (occur or not) and that the system is in a steady state (or mission time is fixed). It does not handle time-dependent behavior well, such as repairable systems where components can fail and be repaired. For those, Markov chains or Petri nets are more appropriate.

Another limitation is the need for accurate input probabilities. In practice, these are often uncertain. A sensitivity analysis (varying inputs within plausible ranges) can show how robust your conclusion is. If the top event probability changes drastically with small input changes, you may need better data.

Alternatives to FTA include event tree analysis (which is inductive, starting from an initiating event and exploring consequences), reliability block diagrams (which are equivalent to fault trees for many systems), and failure mode and effects analysis (FMEA, which is bottom-up and qualitative).

Frequently Asked Questions

How do I convert a failure rate to a probability?

For a non-repairable component with constant failure rate λ (failures per hour) and mission time t (hours), the probability of failure is P = 1 − e^(−λt). If λt is small (less than 0.1), you can approximate P ≈ λt. For example, λ = 1e-5 per hour and t = 100 hours gives P ≈ 0.001.

What if I have more than two inputs to a gate?

The formulas generalize. For an AND gate with n inputs, P = ∏ P_i. For an OR gate with n inputs, P = 1 − ∏ (1 − P_i). You can chain the two-input formulas, but the direct product form is simpler and avoids rounding errors.

Can I use fault trees for human errors?

Yes, but human error probabilities are harder to estimate. You can use generic databases (e.g., THERP) or expert judgment. The same Boolean logic applies; just treat the human error as a basic event with its own probability.

Why does my OR gate probability seem too high?

If you are using the formula 1 − (1−P1)(1−P2), the result is always less than P1 + P2 (unless one is zero). If you are getting a number close to the sum, you may be double-counting the intersection. For rare events (P < 0.01), the approximation P1 + P2 is often acceptable, but it overestimates the true value.

How do I handle common cause failures?

Common cause failures break the independence assumption. One simple method is the beta factor model: assume a fraction β of each component’s failure is due to a shared cause. Then model a separate basic event that causes all components to fail simultaneously. More advanced methods include the multiple Greek letter model or explicit dependency gates.

What is the difference between a fault tree and a reliability block diagram?

They are mathematically equivalent for most systems. A fault tree is top-down (failure-oriented), while a reliability block diagram is success-oriented (shows what must work). The same system can be modeled either way; choose based on which is easier to communicate.

Related

• AI Threat Detection Workflow for Critical Systems
• Roguelike Mechanics in a Classic Pokemon Adventure
• Special Weapons Case: Does It Affect Rank or Trophies?

in p1 | P(B1) | - | 0.01 | 0
in p2 | P(B2) | - | 0.02 | 0
in p3 | P(B3) | - | 0.03 | 0
out top | Top event probability | - | 1 - (1 - p1 * p2) * (1 - p3) | 4
out and_gate | P(AND gate) | - | p1 * p2 | 4