By Nicholas Sheble
Knowing when the instruments and controls in an automated system will fail presents obvious opportunities for saving and making money along with improving quality and safety.
“We looked at failure rate data and the variance in the numbers was sometimes an order of magnitude off,” said Dr. William Goble, principle partner of exida during Wednesday’s company webinar: “Why Do Field Failure Studies Give Different Results?”
Goble is an expert in programmable electronic systems analysis, safety. He developed many of the techniques used for probabilistic evaluation of safety and high availability automation systems.
Why is there such variance in the data?
“There are a number of factors,” said Goble who has not only studied the issue academically and statistically but has spent time observing factory-floor machine and instrument maintenance practices.
“Sometimes data from old and new instruments and technology are inappropriately combined. Or perhaps the maintenance capability of one facility differs from another. That can skew data. Or even something as fundamental as the definition of what failure is comes into play. As well, there’s the issue of analysis and the technique the company uses to render the data to conclusion,” Goble said.
The other primary source of field failure data comes from manufacturers. “Manufacturers can be overly optimistic,” he said. They also have inconsistent methods of data analysis and definitions of defect. Another factor that bumps the data is the instrument’s warranty.
The key Goble said is record all events associated with the process and instruments, and to gather information automatically whenever possible. “Don’t tax the maintenance mechanics with forms to fill out. Collect data from the instruments,” advised Goble.
Traditionally the most common sources of data are a device’s historical data.
Many organizations maintain internal databases of failure information on the devices or systems they produce. One can use these to calculate failure rates for those devices or systems.
For new devices or systems, the historical data for similar devices or systems can serve as a useful estimate.
Handbooks of failure rate data for various components are available from government and commercial sources. Exida maintains predictive benchmarks.
The most accurate source of data is to test samples of the actual devices or systems themselves and generate failure data. This, however, is often quite expensive and or impractical.
This webinar, part of a Wednesday series of education events, shows how field failure studies are a useful source of failure rate data. However, when comparing various studies, the published data varies considerably for the same product category. It outlines:
• Reasons why field failure analysis can produce widely divergent results
• The benefits of good data collection and usage
• A method for cost effective data collection and analysis
Send your questions on failure rate data to Goble himself at firstname.lastname@example.org .
Nicholas Sheble (email@example.com) is an engineering writer and technical editor in Raleigh, NC.