Field Data Analysis First Look

What Can Field Failure Data Reveal?

Field data analysis starts with the collection of data. In a previous article, we used a Nevada chart to gather the counts per month of field failure data. The chart also provides the necessary data to account for how many units have not failed as of yet.

The Nevada chart on its own is just a table of numbers and does not reveal patterns of the changing nature of failure rates over time. Are we experiencing early life failures or wear out related failures?

We need to conduct some data analysis to learn what message the data contains.

Basic Weibull Field Failure Analysis

For non-repairable time to failure data I fit the data to a Weibull distribution. It’s where I start the field data analysis. The Weibull distribution is versatile as it’s able to adequately describe a wide range of time to failure data patterns. Plus it is easy to interpret.

Given we have the Nevada chart of data, I opened a Warranty folio in Weibull++ (from Reliasoft). We need to enter the data into two tabs. The first is for the Sales or units shipped. We have monthly data, so entered the data as follows.

Next, enter the failure count data again by month and by month shipped.

I left everyone concerning the fit at defaults for a two parameter Weibull using a rank regression fitting method. Tapping “Calculate” we find the Weibull parameters of Beta of 1.31 and Eta of 123 months.

The Analysis Step

The beta is informative yet, let’s first take a look at the plot to see how well the fit described the data.

The fitted line follows the data well. Of course there are statistical tests to determine the goodness-of-fit, yet for now this looks pretty good. This is just a first look at the data.

Back to the meaning of the beta value. A beta of 1.3 indicates an increasing failure rate over time. As the units remaining service longer, the chance of failure increases.

The issue is will the tally of failures or risk of failure for those using the product be acceptable?

The Weibull plot is a cumulative distribution function (CDF) plot. It uses a log-log plot with time, here in months, on the x-axis and the probability of failure (unreliability) on the y-axis.

At about 3 months of use there is approximately a 1% chance of failure. At one year (12 months) there is approximately a 15% chance of failure. Nearly all units will fail by 10 years of use.

These values represent the actual data as collected using the Nevada chart out to 6 month. Beyond that time is an extrapolation and provides a forecast of future performance is the same basic pattern of failures remains consistent with the failures seen within the data collection period.

For our situation, compare the field data to the expected reliability performance. Plus, work to understand if the current rate of failure is acceptable for your customers. The goal and/or the current performance may or may not meet the customer expectations, thus check.

Forecasting the Number of Failures

If you are staffing a repair center or ordering replacement units to honor warranty claims you may want to know how many failures to expect going forward.

The Weibull++ software package within the Warranty folio can calculate an extension to the Nevada chart. You need to enter how many months into the future you want to forecast and the future months of production/shipping that will likely occur.

Here I forecast out 6 months with 5,000 units forecast to ship each month. The result is an extension of the Nevada chart.

The number of returns each month is made up of returns from each month of production. Thus as units continue to ship the number of units available to failure increases. Also, if the rate of failure remains steady (not constant) as unit continue to age they have an increasing chance of failure.

The tally by month along the bottom of the chart provides an estimate for future expected returns. Plan accordingly.

Fred Schenkelberg is an experienced reliability engineering and management consultant with his firm FMS Reliability. His passion is working with teams to create cost-effective reliability programs that solve problems, create durable and reliable products, increase customer satisfaction, and reduce warranty costs. If you enjoyed this article consider subscribing to the ongoing series at Accendo Reliability.