What is MTBF?

MTBF: mean time between failure.

Purpose of testing: to measure the reliability indicator of a product (especially electrical products).

Unit: usually hours, for example 5,000,000 hours.

MTBF test methods

MTBF prediction method

The MTBF prediction method enters various influencing parameters into simulation software such as PCT. Taking a laptop (motherboard) as an example, the test conditions entered include the calculation standard, calculation method, temperature, environment and component quality; the software then models the product's parameters by querying a database and calculates an MTBF value. This is an estimated value.

Reference standards for reliability prediction:

MIL-HDBK-217, Reliability Prediction of Electronic Equipment

Note: proposed by the US Department of Defense Reliability Analysis Center and Rome Laboratory and now an industry standard.

GJB/Z 299B, Reliability Prediction of Electronic Equipment

Note: the Chinese military standard.

Bellcore SR-332, Reliability Prediction Procedure for Electronic Equipment

Note: proposed by AT&T Bell Laboratories and now the industry standard for calculating MTBF of commercial electronic products.

MTBF experimental methods

There are three MTBF experimental methods:

1. Full life test: requires all samples to eventually fail during the test, after which a simple arithmetic mean gives the MTBF;

2. Censored sequential test: during the test the equipment under test is monitored continuously or at short intervals, and the accumulated relevant test time and number of relevant failures are compared with the criteria for deciding whether to accept, reject or continue testing;

3. Time-censored or failure-censored test: during the test the equipment under test is monitored continuously or at short intervals, accumulating relevant test time until the predetermined relevant test time is reached or exceeded (acceptance) or the predetermined number of relevant failures occurs (rejection).

Note: in practice all samples under test will fail during testing, and for products with a high reliability level the test time could be years, decades or even centuries, so this method can only be used for products with a short service life.

Reference standards for reliability testing methods:

GB/T 5080.7-1986 Equipment reliability testing - Compliance test plans for validity of the constant failure rate assumption

GB/T 9813.1-2016 clause 5.9 General specification for computers - Part 1: Desktop microcomputers - Reliability testing

GB/T 9813.2-2016 clause 5.9 General specification for computers - Part 2: Portable microcomputers - Reliability testing

GB/T 14081-2010 clause 5.9 General specification for keyboards for information processing - Reliability testing

GB/T 18220-2012 clause 5.17 General specification for handheld information processing devices - Reliability and life testing

MTBF acceleration method

In manufacturing, reliability needs to be understood before products leave the factory in order to assess after-sales risk. The traditional MTBF calculation approach is clearly not suitable here, because by the time the actual MTBF test value is available the product is likely already obsolete or superseded, making the result meaningless. Other methods are therefore needed to simulate and calculate the MTBF value, such as the MTBF acceleration method.

The MTBF acceleration method uses accelerated life testing in the laboratory to evaluate a product's MTBF within an acceptable test time without destroying its original design characteristics. The acceleration factors affecting electronic products mainly include temperature, humidity, resistance and voltage.

Reference standard for accelerated testing:

GB/T 34986-2017, Accelerated testing methods for products

MTBF acceleration method case study

Taking a monitor from a medical device company in Shanghai as an example, the following describes the accelerated life test process.

Step 1. Determine the evaluation path according to product characteristics

The product has a single function, so whole-unit verification was chosen.

Step 2. Determine the number of samples

Three whole units.

Step 3. Specify the expected service life and confidence level

Expected service life 10 years, confidence level 99%. The plan is to use mean time between failure to simulate the expected service life, that is MTBF = 10 years = 87,600 h.

Step 4. Determine the clinical use modes

The monitor has three modes during use: storage mode, standby mode and operating mode.

Step 5. Determine the clinical use stress profile

Given that the monitor is used to monitor patients before surgery, the clinical use stresses are mainly temperature and humidity. The product is used at 25 degrees C and 60% RH.

Step 6. Determine the stress levels

The maximum temperature and humidity the product can withstand:

Storage mode: 60 degrees C, 93% RH

Standby mode: 55 degrees C, 85% RH

Operating mode: 45 degrees C, 85% RH

To avoid damaging the product at excessive temperature and to avoid introducing more variables through temperature changes, a constant 45 degrees C and 85% RH was chosen for the test.

Step 7. Select the failure model

Based on the clinical use stress profile, the Peck model proposed by D. S. Peck for temperature and humidity stresses was used.

Step 8. Determine the model coefficients from the literature

Consulting the Telcordia SR-332 (2016) reliability prediction standard for electronic products, the coefficients were set as Ea = 1.0 eV and n = 2.66.

Step 9. Calculate the acceleration factor

Acceleration model:

where n = 2.66 and Ea = 1.0 eV.

Calculated acceleration factor AF = 2.525 x 11.607 = 29.3.

Step 10. Determine the accelerated test time

With a confidence level of 99%, the normal distribution table gives Z = 2.58.

Cumulative test time Ttot = MTBF x Z / AF = 87,600 x 2.58 / 29.3 = 7,714 h.

Adding a 10% margin, the total test time Ttot = 8,485 h.

Three prototype units are provided for testing. Assuming no failures occur and each unit completes the whole test cycle, each unit needs a test time T = 8,485 / 3 = 2,828 h = 117.8 days.

Step 11. Apply the stresses

In actual use the product is used 7 days a week, 5 hours a day in operation, 3 hours in standby and the remainder in storage.

To simulate actual use, the working stress is applied in 24-hour cycles: 5 hours of continuous operation, then 3 hours of standby, then 16 hours of storage.

Step 12. Conclusion

If the three prototype units are tested at 45 degrees C and 85% RH with the working stress applied as specified in section 5 and run for 117.8 days without failure, the lower limit of the product's MTBF is considered to be 10 years (with a 10% margin).

In summary, although time-censored and sequential tests are very useful, the test time is too long, especially for products with a particularly high reliability level.

In actual production, in order to shorten time to market, reduce product cost and meet the demand for cost-effective accelerated test methods, accelerated testing solves the problem more effectively. It shortens test time by raising the stress level applied to the product or increasing the frequency of alternating stress, and identifies and reduces product failure modes, so that the product's reliability level can be assessed quickly and improved.