The parameter estimates define the best-fitting parameter estimates for the distribution that you selected for each failure mode. All other parametric distribution analysis graphs and statistics are based on the distribution. Therefore, to ensure that the results are accurate, the distribution that you select must adequately fit the data.
You cannot determine from the estimated distribution parameters whether the distribution fits the data well. Use the distribution ID plot, probability plot, and goodness-of-fit measures to determine whether the distribution adequately fits the data.
For the dishwasher data, the engineers selected a Weibull distribution to model spray arm breaks, and a lognormal distribution to model spray arm obstructions. The following parameters define the best-fitting distributions for each failure mode:
Shape = 1.97672 and Scale = 891.929 for spray arm breaks
Location = 5.75328 and Scale = 1.95933 for spray arm obstructions
Multiple failure mode analysis – percentiles
The percentiles indicate the age by which a percentage of the population is expected to fail. Use the percentile values to determine whether your product meets reliability requirements, or to determine which failure modes impact the overall reliability.
Use these values only when the distribution fits the data adequately. If the distribution fits the data poorly, these estimates will be inaccurate. Use the distribution ID plot, probability plot, and goodness-of-fit measures to determine if the distribution adequately fits the data.
For the dishwasher data, based on the distributions fitted to each failure mode, the engineers conclude the following:
1% of the spray arms fail due to breakage by 87.0276 cycles
1% of the spray arms fail due to obstruction by 3.30424 cycles
Overall, by 3.30048 cycles, 1% of the spray arms will fail. For the greatest improvement in reliability, the engineers should focus improvement efforts on minimizing spray arm obstructions.