December 2008

Columns

HP Reliability: To b(ias) or not to b(ias)? This is the question

How often would you bias an inferential model? Given the propensity of instruments to go off calibration, "never" is not a valid answer. "Every day upon receiving lab information" is an equally bad an..

Friedman, Y. Z., Petrocontrol

How often would you bias an inferential model? Given the propensity of instruments to go off calibration, "never" is not a valid answer. "Every day upon receiving lab information" is an equally bad answer, which means: I don't trust this inference at all, and every day I have to tell it what value it should be calculating. This last approach has been the prevailing one, although engineers who have some faith in their models have softened it with filtering techniques, for example: If that is a first-time deviation in that direction apply a bias of 10% of the deviation from lab data. For a second-time deviation bias 30% of the difference. For a third-time deviation bias 80% of the

Log in to view this article.

Not Yet A Subscriber? Here are Your Options.

1) Start a FREE TRIAL SUBSCRIPTION and gain access to all articles in the current issue of Hydrocarbon Processing magazine.

2) SUBSCRIBE to Hydrocarbon Processing magazine in print or digital format and gain ACCESS to the current issue as well as to 3 articles from the HP archives per month. $409 for an annual subscription*.

3) Start a FULL ACCESS PLAN SUBSCRIPTION and regain ACCESS to this article, the current issue, all past issues in the HP Archive, the HP Process Handbooks, HP Market Data, and more. $1,995 for an annual subscription.  For information about group rates or multi-year terms, contact email Peter Ramsay or call +44 20 3409 2240*.

*Access will be granted the next business day.

Related Articles

From the Archive

Comments

Comments

{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}