| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Again, before we embark on large quantities of less than well-informed debate in this area, there are large bodies of knowledge and information modelling in respect of engineering tolerances.
Cheers,
Tim.
> -----Original Message-----
> From: Adam Pease [mailto:apease@ks.teknowledge.com]
> Sent: 21 August 2001 05:14
> To: Ian Niles; Standard-Upper-Ontology (E-mail)
> Subject: Re: SUO: Representation for Precision
>
>
>
> Folks,
> Another issue to consider is how to map the abstract
> notion of numeric
> precision to real world issues like engineering tolerances.
> I discussed
> with Ian an example of an engineering blueprint for a peg and
> a hole. Both
> would be dimensioned and the 'key' to the drawing might state
> that all
> dimensions have 0.01" tolerances.
> This issue is quite relevant in the real world if we want
> a reasoner to
> determine whether an error in the design would allow a
> delivery which fits
> the spec to result in a peg which doesn't fit in the hole.
> Another example
> application would be to determine whether an unnecessarily
> high precision
> is called for in a design, possibly adding to manufacturing expense.
>
> Adam
>
>
>
> At 04:10 PM 8/20/2001 -0700, Ian Niles wrote:
>
> >Hi All,
> >
> > I've been thinking about the issue of the representation of
> >precision of numbers and wanted to offer a particular
> proposal in terms of
> >the SUMO. First of all, the idea would be that all of the
> numbers in the
> >SUMO taxonomy (i.e. every subclass of 'Number') would be
> assumed to have
> >full precision. We could then define the function
> 'NumericPrecisionFn' as
> >follows to reduce the precision for particular applications.
> >
> >(instance NumericPrecisionFn BinaryFunction)
> >(domain NumericPrecisionFn 1 RealNumber)
> >(domain NumericPrecisionFn 2 NonnegativeInteger)
> >(range NumericPrecisionFn RealNumber)
> >(documentation NumericPrecisionFn "A general binary function
> for assigning a
> >specific precision to a quantity (either a number or a
> physical quantity).
> >(NumericPrecisionFn ?QUANT ?INT) means that the value of
> ?QUANT is accurate
> >to at least the decimal place ?INT.")
> >
> >For example, '(NumericPrecisionFn Pi 0)' would give the result 3, and
> >'(NumericPrecisionFn Pi 4)' would give the result 3.1415. I
> think this
> >proposal allows us to have fully precise numbers, but it
> also allows us to
> >manipulate these numbers for real world applications. In
> any case, let me
> >know what you think.
> >
> >-Ian
>
> Adam Pease
> Teknowledge
> (650) 424-0500 x571
>