MTBF is a reliability metric, especially for products or systems exhibiting failures in a shape of a bathtub curve. Information on MTBFs supports decision making for engineering development and service operations in management and planning.

A bathtub curve generally represents three phases of product failures, including an early phase, a normal service life phase, and a wear-out phase. For a real product, we may or may not observe all three phases. For an MTBF estimation purpose, we expect a constant failure rate during a product's service period.

1. Field return data fitting in a bathtub curve

This is an example of how a set of real field return data look like after being processed and presented in a failure rate plot. It can be observed that the data fit a bathtub curve showing an early and then a constant failure rate phase of the product in this example.

If you need to know whether your products, systems or assets in operation also exhibit a bathtub failure behavior, or if you simply need MTBF info to be extracted from field data for some reason, we can help, We provide a free service for a preliminary analysis. You can click here to contact us.

2. The original data

A Nevada chart and a Pareto histogram plot are used to present the original field return data of this example.

Some observations of the data include:
1-The data set represents a total of 245 units and 284 failure reports returned from field;
2-Data were collected across a time span from Mar.2014 to Jun.2015;
3-There appeared an early failure mode;
4-A major percentage of either no failure found (NFF) or failures yet to be determined (TBD) or isolated is included in the data set.

3. An identified early failure

An early failure is identified. Data are singled out for a cumulative failure percentage analysis and its result is presented in a Weibull plot, in which a cumulative Weibull distribution function appears linear.

4. MTBF

Data points of the identified early failures are removed from the original bathtub curve for the purpose of a better presentation of random failures and so to have a more persuasive estimation of MTBF.

The example product experienced continuous modifications for improvements after being delivered to customers. The data represents its three revisions of v2, v4 and v5. An estimated overall MTBF value as well as each for every one of those design revisions can be extracted from the data.

5. Reliability growth

Upon removal of the identified early failures, random failures can be viewed and MTBF estimations compared for different design revisions. Then, reliability improvements with revisions are quantitively evaluated.

A summary of accomplishments includes:
1-The identified early failure has been corrected.
2-Reliability improvements are confirmed and evaluated quantitatively per MTBF estimates.