Turn raw run-life and failure records, including units still in service, into a defensible reliability picture and a forward maintenance forecast.
This tool fits a two-parameter Weibull distribution to field failure data by maximum likelihood estimation (MLE), estimates a non-parametric Kaplan-Meier survival curve, and projects future failure demand across a configurable horizon using Monte Carlo simulation. It is built for reliability and maintenance engineers working with messy operational data rather than clean laboratory test sets.
Most field datasets are right-censored: a meaningful share of units are still running when you pull the data, so their true life is unknown but not zero. Median-rank regression on a probability plot handles this poorly once censoring is heavy or the failure count is small. Maximum likelihood estimation incorporates each suspension explicitly through the survival term in the likelihood, which is why it is the preferred estimator for sparse, censored fleet data. The fitted shape parameter beta is the diagnostic that matters: beta below 1 indicates infant mortality, beta near 1 indicates random (constant-hazard) failures, and beta above 1 indicates wear-out, each of which points to a different maintenance response.
A fleet of 40 downhole pump assemblies yields 23 recorded failures; the remaining 17 are still operating and enter the fit as suspensions. MLE returns beta = 2.3 and a characteristic life eta = 540 days. Beta above 2 confirms a wear-out mechanism rather than random failure, so condition-based replacement is justified over run-to-failure. The B10 life (the age by which 10 percent are expected to fail) lands near 215 days, and the Monte Carlo module forecasts roughly 6 to 9 failures over the next 180 days, sizing the spares pool and the workover schedule directly from the data.
For: reliability engineers, maintenance planners, and asset managers converting historical run-life records into replacement strategy and spares forecasting.