Why is the preparation of life data important in a Weibull analysis?

It’s suggested that we should put 90% of the effort towards a Weibull analysis towards the preparation of the data – this is because it has such a profound and direct impact on the credibility of the result.

Poor quality data could jeopardise your entire analysis.

There are 4 steps to prepare life data:

1. Determine the asset(s) to be analysed.
2. Determine the component failure mode for that asset(s).
3. Obtain as much relevant life data as practical.
4. Classify life data.

Step 1: Determine the asset(s) to be analysed.

Based on your needs of the assessment, you can perform a Weibull analysis in a single asset or multiple assets. When using the Weibull analysis on multiple assets, make sure they are similar in design and function.

In other words -? their failure modes and failure rates should be alike.

Step 2: Determine the component failure mode for that asset(s).

Different components failure modes have different failure rates. After choosing the asset to be analysed, you need to treat the component failure modes separately to ensure an accurate and representative result.

Step 3: Obtain as much relevant life data as practical.

The term “Life data” refers to measurements of a product’s life. A product lifetime can be calculated in hours, kilometres, cycles or any other metric that applies to the period of successful operation of a product. Time is the most common measure of a product’s lifetime; therefore, life data are often called “times-to-failure” (TTF) data. In order to make accurate predictions about the life of all products in the population, you need to gather as much relevant life data as practical.

The better the data (along with the appropriate model choice), the better your predictions will be.

Step 4: Classify life data.

In Step 3, you collected as much available data sets as you can; however, not every data set has complete information.

Therefore, life data are categorised into 2 types: complete data (all information is available) or censored data (some of the information is missing). Within censored data, it has 3 sub-types: right censored (suspended), Internal censored, and left censored data.

Different data type requires different analysis methods.

Complete Data?

The exact time-to-failure (TTF) for the unit is known (e.g., the unit failed at 300 hours of operation). Usually from highly structured lab testing or fully accessible field data with high failure rates.

Right Censored Data (aka Suspended Data).?

The unit operated successfully for a known period of time and then continued (or could have continued) to operate for an additional unknown period of time (e.g., the unit was still operating at 300 hours of operation).

Interval Censored Data.?

The unit’s exact TTF is unknown, but it failed at some point within an interval. (e.g. the unit failed between 300 hours and 400 hours).

Rule of Thumb Regarding Interval Censored Data

Treat data as interval data if the granularity of the data is coarser than the desired results.

• If the desired result is in months and data points are in months as well, then consider the set to be complete data
• If the desired result is in days but data points are in months, then consider the set to be interval data

Left Censored Data

The unit’s exact TTF is only known to be before a certain time. (e.g., the unit failed between 0 hours and 300 hours).

You can read more about Weibull analysis and data preparation here - https://foxly.top/Weibull

#Reliability #Maintenance #MaintenanceManagement