Scribd
Upload
170 views20 pages

The Latest On Quantitative Sampling Theory: Liberation, Shape and Granulometric Factors

The document discusses factors that influence the relative variance of error in quantitative sampling theory. It covers the liberation factor l, which represents the proportion of liberated mineral and depends on liberation and comminution sizes. It also discusses the shape factor f, which accounts for differences in mineral and gangue shape, and the granulometric factor g, which typically takes a value of 0.25 when using the P95 comminution size. The document suggests formulas to properly account for cases where these factors differ, such as when mineral and gangue have different shape factors. It also examines the granulometric factor g' when sampling closely sieved material between two screens.

Uploaded by

Luis Katsumoto Huere Anaya
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
170 views20 pages

The Latest On Quantitative Sampling Theory: Liberation, Shape and Granulometric Factors

The document discusses factors that influence the relative variance of error in quantitative sampling theory. It covers the liberation factor l, which represents the proportion of liberated mineral and depends on liberation and comminution sizes. It also discusses the shape factor f, which accounts for differences in mineral and gangue shape, and the granulometric factor g, which typically takes a value of 0.25 when using the P95 comminution size. The document suggests formulas to properly account for cases where these factors differ, such as when mineral and gangue have different shape factors. It also examines the granulometric factor g' when sampling closely sieved material between two screens.

Uploaded by

Luis Katsumoto Huere Anaya
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 20

The Latest on Quantitative Sampling Theory:

Liberation, Shape and Granulometric Factors


Dominique FRANCOIS-BONGARCON
[email protected]

Rel.Var. = c l f g dN3 (1/MS - 1/ML)


D.Francois-Bongarcon, AICI, Vancouver, Canada
‘Gy’s Formula’ for predicting the relative
variance of error of a correct, random sample
not affected by segregation:

Rel.Var. = c l f g 3
dN (1/MS - 1/ML)

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Liberation Factor l
• l was initially introduced by Gy as a number in [0,1] to reduce the
variance for the non-liberated case. Indeed, the formula was
demonstrated only for the liberated case!
• Gy found l was correlated to the degree of metallurgical liberation of the
mineral or metal of interest, hence the name of that factor
• Gy had the intuition it was equal to the proportion of liberated mineral
but he did not publish any demonstration
• It was more recently shown to depend upon the liberation and
comminution sizes, plus mineralogical properties at microscopic scale.
• Models for l have been established and successfully used

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Liberation Factor l [cont’d]
• On the surface, there is no reason for l to be equal to the proportion
of liberated mineral.
• Indeed l was shown to be the ratio of two variances of actual,
special samples (taken at comminution and at liberation size),
making it difficult to accredit such a result, or make it obvious. Why
would that physical proportion be equal to a ratio of two variances?
• But in 2017, it was established l was indeed, at the very least, a very
good approximation to that proportion of liberated mineral.

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Liberation Factor l [conclusion]
• TODAY:

• l is very close to the proportion of liberated mineral and can


therefore be used as an estimate of it

• It is usefully modeled as l = ( dl / dN )b , where 0 < b < 3 and dl and dN


respectively are the liberation and comminution sizes

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Shape Factor f: Peeking into Future Results
• Unpublished research into the demonstration of Gy’s formula has
recently established the following:

Rel. Var. = (1/MS - 1/ML) c l 


vG
where c = [( 1 - tL) / tL] [( 1 - t L) k ρ M + tL ρ G] and k = vM / 
vG
is the ‘generalized mineralogical factor’, vM and  v G the average
mineral and gangue fragment volumes and t L , ρ M and ρG respectively
the grade of the lot and the densities of mineral and gangue

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Shape Factor f: Peeking into Future Results
[cont’d]
Rel. Var. = (1/MS - 1/ML) c l 
vG
• Interestingly, this formula shows an older result: a lot to be sampled
will respond identically to sampling a lot in which all fragments have
a same volume, equal to the average gangue fragment volume
(provided c is calculate properly, which we will see is not always
straightforward...)

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Shape Factor f: Peeking into Future Results
[cont’d]

• In the particular case where mineral and gangue comminute


together and have the same shape factors, k = 1 and this transforms
into the classical, well known formula:

Rel. Var. = c f g l dN3 (1/MS - 1/ML)

• and f is the common shape factor of the mineral and gangue

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Shape Factor f: Peeking into Future Results
[cont’d]
• There are other cases that are not handled properly by the classical
formula. For these cases, the following set of formulas will provide
the proper developments, as required:
Rel. Var. = (1/MS - 1/ML) c l  vG
c = [( 1 - tL ) / t L] [( 1 - tL) k ρ M + tL ρ G]
k=
vM / 
vG

vM = fM g dNM3 and 
vG = fG g dNG3 where d NM and dNG
are the mineral and gangue comminution sizes

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Shape Factor f: Peeking into Future Results
[cont’d]
• In the special case where mineral and gangue comminute together
but have different shape factors fM and fG (e.g. gold flakes), then k =
fM / fG , which modifies the expression of the mineralogical factor c
but keeps the shape factor of the gangue in the rest of the formula,
which is at odds with the simplistic choice of a general, common
shape factor f = 0.2:

Rel. Var. = (1/MS - 1/ML) c l fG g d NG3

D.Francois-Bongarcon, AICI, Vancouver, Canada


The Granulometric Factor g
• Some 14 years ago, progresses were made to understand why, as
suggested by some of Gy’s experiments, only a P95 should be used in
TOS formulas to represent the degree of comminution, along with a
value g=0.25 of the granulometric factor.
• [ this raises some concern about the use of P80 in metallurgical tests ]
• In more recent times, some research was expended on size distribution
models and on a curve published by Gy for modifying the use of that
factor when sampling material that has been closely sieved...
• ...with interesting ancillary results

D.Francois-Bongarcon, AICI, Vancouver, Canada


P95 as a Function of Other Definition Screens
• In 2005, relatively simple simulations of the
comminution process have elucidated the
pinching of Gy’s experimental g(x) curves at a
x=5% reject screen, where the value of g varies
quite closely around an average value of 0.25.

• Later, using those of Gy’s 114 experimental curves


that corresponded more to laboratory equipment,
an interesting, approximate progression was
empirically found as follows:

D.Francois-Bongarcon, AICI, Vancouver, Canada


P95 as a Function of Other Definition Screens

• Oftentimes, P95 is approximated reasonably well by:

1.25 x P95
1.50 x P85
1.75 x P80

• ...which suggests a very regular progression.


• More later, after speaking first of closely sieved material...

D.Francois-Bongarcon, AICI, Vancouver, Canada


Factor g’ for Closely Sieved Material
• In some cases (e.g. Iron ore sample preparation in Brazil), one may
be lead to sample size fractions between two screens.

• Gy studied this particular problem. Using the progression of AFNOR


sieve sizes, he produced a curve giving a substitute g’ to be used
instead of g in the sample variance formula and to be applied to
dMAX3 instead of dN3.

D.Francois-Bongarcon, AICI, Vancouver, Canada


Factor g’ for Closely Sieved Material [cont’d]

D.Francois-Bongarcon, AICI, Vancouver, Canada


Factor g’ for Closely Sieved Material [cont’d]
• One should remember quantity [ f g dN3 ] in the variance formula
represents the average fragment volume (the real variable
representing comminution size, at least for sampling).
• Given we do not really have access to the distribution of fragment
sizes within the size fraction of interest, an attempt was made at
figuring what using the midsize of the fraction to approximate the
average volume would do, and the result was stunning...
• The value of g’ does not depend on dMAX any longer, only on the
ratio r of screens, large or not. The calculation simplifies to:
g' = [ (1+r)/(2r) ]3
• ...and gives almost identical results!
D.Francois-Bongarcon, AICI, Vancouver, Canada
Factor g’ for Closely Sieved Material [cont’d]

D.Francois-Bongarcon, AICI, Vancouver, Canada


Factor g’ for Closely Sieved Material [cont’d]
• At very large sizes of r, the midsize
approximation does not hold true, nor do
some of the assumptions Gy had made.
• Extrapolating the meaningful part of these
curves, it seems for r infinite the asymptotic
value is 0.1
• As a consequence:
g(=0.25) dN3 must equal g’(large r, i.e =0.1) dMAX3
• This implies dN = 0.75 x dMAX
• This completes our previous progression.....

D.Francois-Bongarcon, AICI, Vancouver, Canada


P95 as a Function of Other Definition Screens
• P95 is approximated reasonably well by:

0.75 x P100
1.00 x P95
1.25 x P95
1.50 x P85
1.75 x P80
• This confirms this remarkable result that is being investigated. Already:
• Both the Gaudin-Schuhmann and the Rosin-Rammler size distributions
restitute this progression for certain values of their parameters.

D.Francois-Bongarcon, AICI, Vancouver, Canada


THE END
THANK YOU!
D.Francois-Bongarcon, AICI, Vancouver, Canada

You might also like