A crucial knowledge repository for Reliability professionals
One of the most valuable tools for a Reliability Engineering team is an asset life model repository or Library. This is also known as the Weibull Library. It contains the life models for all the critical assets in the organization. This information is crucial for failure prediction and ultimately decision making in the domain of asset management. It is the company’s own and true version of a “failure database”. It avoids the use other generic external databases that might not reflect the true behavior of the assets in the organization.
Contents
The qualifier “Weibull” is a bit of a misnomer. Not all life distributions follow a Weibull distribution. Lognormal, Exponential, Beta or Poisson could be some of the typical distributions making up a life model. The basic requirements of a Weibull Library setup are as follows:
- The asset class or asset type such as a centrifugal pump, power transformers or valves just to name a few. Sub-asset classes should be included. This depending on the variety and diversity of asset classes, . For example, if there are different types of transformers, sub asset classes would contain 168KV transformers versus 66KV transformers. The Library can also include component life models if sufficient records are available. Those relate to non repairable items making up the asset class. Equally called spare parts, componenets are typically discarded after failure. For example, if the asset class is a centrifugal pump, the components would be mechanical seals, bearings, wear rings etc.
- The life distribution model type and corresponding distribution parameters. This is essentially the most important element of a Weibull Library! For example, if the model is a two parameter Weibull distribution, the Library would include the shape (beta) and scale (eta) parameters.
- The source data of the model. This includes the Computer Maintenance Management System (CMMS) records used to build the model.
- The date of model construction. This indicates if the model needs updating.
- The number of data points in the model. This is an important criterion. The model accuracy depends on the number of data point it contains. The more, the better.
- The life variable used to build the model. This is typically calendar time, operating time, number of cycles etc.
- Any relevant documentation on how the model was built. This includes assumptions made.
The above list is the basic requirements to set up Weibull Library. However, the Library owner can include any other information they see fit such as reports or specific comments.
One other aspect of asset management is maintainability. This relates to resources used to maintain or repair assets when they fail. Those resources include cost to repair, labour and spares. The time to repair the asset will be the basis for the repair time statistical model. The Library should contain those above-mentioned repair models as they are crucial for other studies such as Reliability, Availability and Maintainability (RAM) models.
Maintaining and making use of the Weibull Library
The Weibull Library is a dynamic database. If additional data points or records are available, life models should be updated. New assets should also be added to the Library. In essence the Reliability Engineer will have to populate the Library with all the critical assets starting from the most critical and working their way down to the less critical. It can be a full-time job but it is in my experience, time extremely well spent.
The setup of a Weibull Library does not require specialized software . A simple Microsoft Excel spreadsheet can suffice. One tab can contain the life distributions whilst the other tab contains the repair distributions.
Information regarding an asset’s future performance come from Weibull Library life models. If a customer or decision maker needs a quick answer on a critical asset, it avoids doing lengthy studies and thus delaying the response. In addition to having good maintenance records, setting up a Weibull Library in the company’s asset information system is like transforming gold into value added jewelry. That is why every company should treat their records like gold – they are one of their most valuable assets!
Larry George says
Good idea. People need reliability database for their products and parts! How else are they going to manage: Warranty? Reserves? Inspections? PMs? Replacements? Spares stocks? Alternatives? Risk? Warranty extensions? Recalls? End-of-Life support?
Dmitri Kececioglu gave students a library of reliability parameters in the early 1980s. My boss took Kececioglu’s course. I suspect Kececioglu assumed Weibull. There is some justification for Weibull distribution for strength of materials in compression (extreme-value distribution for series system, weakest link failure . I can’t remember author!).
I used Kececioglu’s means and variances for strengths of materials for multivariate (lognormal_ stress-strength seismic risk of nuclear power plants.
Paul Barrington had a library of Weibull reliability parameters; I don’t know if it’s still available.
Has anyone generalized the Kaplan-Meier estimator for renewal processes? I happen to be doing that now, assuming Weibull renewal process for grouped (by cohort) renewal counts. (Until I figure out to do the nonparametric equivalent.) If somebody wants Weibull parameter estimates from cohort sizes and grouped renewal counts, send data to pstlarry@yahoo.com.
André-Michel Ferrari says
Thanks for the comments Larry. The Paul Barringer Weibull Library is not available online any more since he passed a few years ago. This was indeed a good attempt by Paul to showcase this concept. One question that may benefit myself and others – what do you mean by the term: “grouped (by cohort) renewal counts”? If it is in one of your previous articles please share the link.