Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
IEC 61703:2016 — Reliability — Mathematical Expressions
IEC 61703:2016 is the definitive mathematical reference for reliability engineering, providing standardized expressions for reliability, availability, maintainability, and supportability (RAMS) metrics used across all engineering disciplines.
Introduction
IEC 61703:2016, titled “Reliability — Mathematical expressions for reliability, availability, maintainability and maintenance support terms,” provides a comprehensive mathematical framework for quantifying and communicating reliability characteristics of equipment, systems, and components. The standard unifies the mathematical definitions used across the IEC 60300 (Dependability management) series and the IEC 60050-192 (Dependability vocabulary), ensuring consistent interpretation of reliability metrics worldwide.
The standard addresses a critical need: before IEC 61703, different industries and even different groups within the same industry used incompatible definitions for fundamental metrics such as MTBF (Mean Time Between Failures), availability, and failure rate. A reliability figure quoted as “99.9%” could mean inherent availability, operational availability, or achieved availability depending on who was calculating it. IEC 61703 eliminates this ambiguity by providing unambiguous mathematical definitions for each term.
The use of non-standardized reliability expressions has been a major source of contractual disputes and safety misinterpretations in engineering projects. A 99.9% availability guarantee means very different things depending on whether downtime includes preventive maintenance, logistics delay, or only corrective maintenance.
Fundamental Reliability Functions
IEC 61703 defines the core reliability functions that form the basis for all derived metrics:
| Function | Symbol | Mathematical Expression | Description |
|---|---|---|---|
| Reliability function | R(t) | R(t) = P(T > t) | Probability of survival beyond time t |
| Failure distribution (CDF) | F(t) | F(t) = 1 – R(t) | Probability of failure by time t |
| Failure density function | f(t) | f(t) = dF(t)/dt | Instantaneous probability density of failure |
| Hazard (failure rate) function | λ(t) | λ(t) = f(t) / R(t) | Instantaneous failure rate at time t |
| Cumulative hazard | H(t) | H(t) = ∫λ(t) dt | Integrated hazard over interval [0, t] |
| Mean Time To Failure | MTTF | MTTF = ∫R(t) dt (0 to ∞) | Expected operating time to first failure |
| Mean Time Between Failures | MTBF | MTBF = MTTF + MTTR | Expected time between consecutive failures |
Exponential Distribution
The exponential distribution (constant failure rate) is the most widely used model in electronic reliability:
R(t) = exp(-λt), MTBF = 1/λ
The standard emphasizes that the exponential model is valid only during the “useful life” period (after infant mortality and before wear-out). The constant failure rate assumption implies that the component is memoryless — a non-repairable component that has survived for 1000 hours has the same probability of failing in the next hour as a brand-new component.
Weibull Distribution
The Weibull distribution provides flexibility for modeling different failure regimes:
R(t) = exp[-(t/η)β], λ(t) = (β/η)(t/η)β-1
| Shape Parameter β | Failure Regime | Application Examples |
|---|---|---|
| β < 1 | Decreasing failure rate (infant mortality) | Early life failures, manufacturing defects |
| β = 1 | Constant failure rate (random failures) | Electronic components during useful life |
| 1 < β < 2 | Gradually increasing failure rate | Mechanical wear, gradual degradation |
| β = 2 | Linear increasing failure rate (Rayleigh) | Corrosion, erosion processes |
| β > 2 | Rapidly increasing failure rate | Bearing wear, fatigue failure |
The Weibull distribution is the preferred model for mechanical and electromechanical components because its shape parameter captures the failure physics of wear, fatigue, and degradation. The exponential distribution, with its constant failure rate assumption, is generally inappropriate for mechanical systems.
Availability and Maintainability Metrics
IEC 61703 provides a structured taxonomy of availability metrics that explicitly account for maintenance and support factors:
| Availability Type | Expression | Includes | Typical Use Case |
|---|---|---|---|
| Inherent Availability | Ai = MTBF / (MTBF + MTTR) | Corrective maintenance only, ideal support | Design comparison, specification |
| Achieved Availability | Aa = MTBM / (MTBM + MDT) | Corrective + preventive maintenance | Factory acceptance testing |
| Operational Availability | Ao = OT / (OT + TMT) | All downtime including logistics, admin | Field performance measurement |
where MTTR = Mean Time To Repair, MTBM = Mean Time Between Maintenance, MDT = Mean Downtime, OT = Operating Time, and TMT = Total Maintenance Time.
Maintainability Function
The standard defines maintainability M(t) as the probability that a repair is completed within time t. For lognormally distributed repair times (the most common model):
M(t) = Φ[ln(t / tmed) / σ]
where Φ is the standard normal CDF, tmed is the median repair time, and σ is the logarithmic standard deviation. The standard notes that repair times are typically right-skewed — most repairs are quick, but a small fraction take much longer due to diagnostic complexity, parts availability, or access issues.
A common pitfall in availability engineering is using MTTR (mean) as the repair time input without considering the distribution shape. If the repair time distribution has a long tail (high σ), the operational availability can be significantly lower than predicted by MTBF/(MTBF+MTTR), even when the mean values appear adequate.
Engineering Insights for Reliability Modeling
1. Confidence Intervals, Not Point Estimates. IEC 61703 emphasizes that reliability metrics should always be reported with confidence intervals. A single-point MTBF estimate from a test with few failures is highly uncertain. For example, if 2 failures are observed in 10,000 hours of testing, the MTBF point estimate is 5,000 hours, but the 90% confidence interval spans approximately 2,100 to 16,000 hours. Engineers should use the chi-squared distribution to compute confidence bounds based on the number of observed failures and total test time.
2. Competing Risk and Mixture Models. Real systems experience failures from multiple mechanisms simultaneously (e.g., electronic component failures, mechanical wear, software defects). The standard provides guidance on competing risk models where the system fails when the first of several independent failure mechanisms occurs. The overall reliability is the product of the individual survival functions: R(t) = R1(t) × R2(t) × … × Rn(t). Mixture models (where a fraction of the population has one failure distribution and the remainder has another) are also covered.
3. Censored Data Handling. Reliability data is almost always censored — units may be removed from test before failure (right-censored), or failure may occur before the first inspection (left-censored). IEC 61703 references the Kaplan-Meier estimator (product-limit method) for non-parametric reliability estimation from censored data. The standard also covers maximum likelihood estimation (MLE) for parametric models with censored data, which provides more precise estimates when the assumed distribution is correct.
4. System Reliability from Component Data. For series systems (where any component failure causes system failure), the system reliability is Rs(t) = ΠRi(t). For parallel (redundant) systems, Rs(t) = 1 – Π[1 – Ri(t)]. The standard provides expressions for k-out-of-n systems, standby redundancy, and complex configurations that can be evaluated using reliability block diagrams (RBD) or fault tree analysis (FTA) — both of which are addressed in companion standards IEC 61078 and IEC 61025.
Frequently Asked Questions
1. What is the difference between MTTF and MTBF?
MTTF (Mean Time To Failure) applies to non-repairable items — the expected operating time until the first and only failure. MTBF (Mean Time Between Failures) applies to repairable items — the expected time between consecutive failures, which includes the operating time after repair. For the same item with exponential failures, MTTF = MTBF numerically, but they refer to different concepts. IEC 61703 explicitly distinguishes these terms and cautions against their interchangeable use.
2. How do I select the appropriate failure distribution for my data?
The standard recommends a three-step process: (1) Plot the empirical failure data on probability paper (or use software) for candidate distributions (exponential, Weibull, lognormal). (2) Use goodness-of-fit tests (Anderson-Darling, Kolmogorov-Smirnov) to quantiatively assess the fit quality. (3) Consider the physics of failure — the selected distribution should be consistent with the known failure mechanism. For electronic components during useful life, exponential is usually appropriate. For mechanical fatigue, Weibull with β > 1 or lognormal is recommended.
3. What is the relationship between reliability and availability?
Reliability R(t) is the probability that a system operates without failure for a specified period. Availability A(t) is the probability that the system is operational at a given time, considering both failures and repairs. For a repairable system with high reliability but long repair times, availability can be poor. Conversely, a system with moderate reliability but very fast repair can have high availability. The relationship is: steady-state availability = MTBF / (MTBF + MTTR), assuming exponential distributions for both failure and repair.
4. How does IEC 61703 relate to IEC 60300 dependability management?
IEC 61703 provides the mathematical foundation for IEC 60300, the overarching dependability management series. IEC 60300 defines the dependability program and processes, while IEC 61703 supplies the precise mathematical expressions needed to implement those processes. They are complementary: IEC 60300 says “what to do,” while IEC 61703 says “how to calculate it.” Both standards reference IEC 60050-192 for terminology.
