Faster Failure Prediction

O&M contributor Henry Canaday sent along the following after attending a conference sponsored by Teradata:

 

Teradata is implementing a novel technique for predicting failure rates for new parts with two aviation manufacturers. The technique, which uses the Weibull Distribution, was developed in work Teradata did for Ford Motor Co. for the launch of new models. Unlike other statistical methods, which require large sample sizes, the Weibull approach can use even the earliest performance of parts in actual operation to make realistic forecasts, according to VP Industry Consulting Jerry Hill. He believes the method will be especially useful to aircraft manufacturers, whose product runs are very small, compared with autos or other consumer industries. The Weibull calculations are done 'on the fly,' as each failure report is entered in Teradata's data warehouse, so they always capture the latest information instantly. And failures are attached to all the attributes of a part, for example operator, component manufacturer(s), date of production and so forth, so remedial action can be focused on potential problem areas. Although calculations are done in the data warehouse, business intelligence tools are needed to view the failure predictions. 

Teradata's data warehouses have been used extensively for airline revenue and operations by carriers such as Continental Airlines and Lufthansa. Teradata is now actively promoting these warehouses for aviation manufacture and supply-chain programs.

Tags: om99, failure, Teradata