MIME-Version: 1.0 Content-Type: multipart/related; boundary="----=_NextPart_01CA2CB5.CF7B56B0" This document is a Single File Web Page, also known as a Web Archive file. If you are seeing this message, your browser or editor doesn't support Web Archive files. Please download a browser that supports Web Archive, such as Windows® Internet Explorer®. ------=_NextPart_01CA2CB5.CF7B56B0 Content-Location: file:///C:/1D29BA59/2007-12-19.htm Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset="us-ascii"
<=
/span>
Canadian Operational Research Society
http://www.corscalgary.org/<=
/span>
PROFESSIONAL DEVELOPMENT SEMINAR
Room 217
450 - 1 Street SW
(See attached map)
Tom Morrison, M.Sc.
TOPIC
Fredholm Integral Equations of the First Kind: A
Discussion of the Influence of Measurement Error
Abstract:
The de= signer and vendor of a measuring system is concerned with how accurately their system reports the measurements obtained during an inspection. Measurement error includes accurac= y, systematic error and repeatability.
Consid=
er the
situation where some data have been obtained and it is desired to fit a
distribution to the sizes of corrosion features, say on a pipeline. Some researchers will select a
reasonable distribution and fit it to the data. The same researchers will then make
comments about the effect of the measurement error on said distribution.
Other researchers may attempt to remove the measurement error from the reported distribution to obtain a “true” distribution. The problem involves the estimatio= n of f(y) inside the Fredholm integral equation of the first kind.
<=
/sub>
The function K(x,y) is the resolution function of the measurin=
g system. The function f(y) is the true
distribution of sizes and g(x) is the measured or reported distribution.
Fredholm integral equations of the first kind used to be notor= iously difficult to resolve. =
In this talk Tom will discuss one of the simplest methods of estimating f(y) based on a measured distribution. The summary of the talk is that a researcher who fits a distribution to the measured data may be mislead as to the true form of the distribution. Tom will review one situation in particular and make comments on the different interpretations of the data obtained during an inspection. In particular he will concentrate = on interpretation of the measured distribution and the consequences for estima= ting extreme sizes as part of a risk analysis study.
Tom will include a simple example in the presentation. Given an exponential distribution = of true sizes, what is the effect of a normal distribution of measurement erro= r on the true distribution? What a= re the consequences for estimating maximum sizes?= Given the measured distribution, Tom will show that operational rese= arch techniques can be used to back-calculate the true distribution f(y).=
About the Speaker:
Tom has 27 years experience in the oil and gas industry and ha= s also performed some statistical work for the Canadian nuclear industry.= p>
Tom’s expertise focuses on risk analysis from a mathemat= ical and statistical viewpoint. = span>
Early in his career he was a research scientist working on arc= tic and ice related issues for structure and pipeline design in the Canadian Beaufort Sea and East Coast. =
About 20 years ago Tom left the offshore exploration part of t= he oil and gas industry and created a company specializing in corrosion issues on pipelines. His major claim to= fame is the development of a procedure whereby corrosion pits from one smart pig inspection of a pipeline could be matched to data obtained during subsequent inspections. This procedure e= nabled transmission companies to obtain the best estimates of corrosion severity on gas and oil pipelines of internal and external corrosion features, complete with a full risk analysis of the current and future corrosion issues on pipelines. As part of this wo= rk he specialized in corrosion growth modelling and risk analyses for pipelines a= nd tubes in steam generators in nuclear power plants.
Tom is currently an independent contractor working again on of=
fshore
environmental design criteria and ice thickness modelling for the Canadian
Beaufort Sea and
Tom has a master’s degree in statistics and operations
research, awarded by the
There is no charge for atten=
ding the
meeting. The room is available
until 1:30 PM for those interested in staying afterwards to mingle and meet
other OR practitioners.
Sketch of Location for C=
ORS
Meetings at TransCanada
Tower in Conference Rooms
214 and 217
<=
o:p>