1. Introduction to Weibull Distribution
Originally formulated by Swedish mathematician Waloddi Weibull in 1951, the Weibull distribution is the universal workhorse of reliability engineering. Unlike the exponential distribution, which enforces an unrealistic memoryless assumption (constant failure rate), the Weibull distribution models the entire physical lifecycle of components—from manufacturing burn-in to late-stage metal fatigue.
The 2-parameter Weibull Cumulative Distribution Function (CDF), representing cumulative unreliability F(t), is defined by:
where β (beta) is the dimensionless shape parameter and η (eta) is the scale parameter (characteristic life in hours, cycles, or miles).
2. Physical Interpretation of Shape Parameter (β)
The value of β dictates the governing physical failure mechanism along the classical "Bathtub Curve":
- β < 1.0 (Infant Mortality / Early Failure): Decreasing failure rate (DFR). Failures are caused by material defects, substandard assembly, installation misalignments, or commissioning errors. Preventive time-based replacement is strictly counter-productive here because replacing units introduces new infant mortality risks.
- β = 1.0 (Random / Exponential Distribution): Constant failure rate (CFR). Failures are induced purely by random external events, voltage spikes, operator errors, or foreign object ingestion. Mean Time Between Failures (MTBF) equals characteristic life η. Preventive replacement provides zero reliability benefit.
- β > 1.0 (Wear-Out / Aging): Increasing failure rate (IFR). Probability of failure increases monotonically with operating hours.
- β ≈ 3.44 (Normal-Like Wear-Out): Symmetrical Gaussian degradation curve typical of bearing rolling contact fatigue or conveyor roller wear.
- β ≥ 4.0 (Rapid Rapid Mechanical Wear): Severe thermal cracking, chemical corrosion perforation, or turbine blade erosion. Preventive component discard before B10 life yields massive economic returns.
3. Handling Right-Censored Suspension Data
In operational reliability, test populations rarely run until 100% of units fail. Operating units removed from service unfailed, or machines currently surviving in the field, are called right-censored suspensions.
Omitting surviving units artificially skews the calculated failure curve to the left, causing engineers to severely underestimate asset lifespan. Using Bernard's Median Rank Formula combined with Johnson's Order Increment Adjustment, AshToolbox mathematically accounts for operating hours accumulated by surviving equipment:
Execute In-Browser Weibull Analysis
Fit failure and suspension datasets, estimate B10/B50 lives, compute MTTF via exact Gamma functions, and generate probability plots client-side with zero data upload.