Scribd Logo
Upload

Welcome to Scribd!

  • 17 views

    VII of If l0 N N Will: Deposits

    Uploaded by

    EndyMpeNk
    1. This document describes a method for simulating block grades by first conducting a non-conditional point simulation and then conditioning the simulation results on actual block data. 2. The method involves 8 steps, including simulating point values, averaging to get block values, transforming to a Gaussian distribution, simulating kriging errors, kriging block values, conditioning on data, and transforming back to get the final simulated block grades. 3. An alternative direct method is described where the block grades can be simulated directly using the structural model and univariate distribution of grades, without first simulating point values.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    VII of If l0 N N Will: Deposits

    Uploaded by

    EndyMpeNk
    17 views2 pages
    1. This document describes a method for simulating block grades by first conducting a non-conditional point simulation and then conditioning the simulation results on actual block data. 2. The method involves 8 steps, including simulating point values, averaging to get block values, transforming to a Gaussian distribution, simulating kriging errors, kriging block values, conditioning on data, and transforming back to get the final simulated block grades. 3. An alternative direct method is described where the block grades can be simulated directly using the structural model and univariate distribution of grades, without first simulating point values.

    Original Description:

    gfd

    Original Title

    23-24

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    1. This document describes a method for simulating block grades by first conducting a non-conditional point simulation and then conditioning the simulation results on actual block data. 2. The method involves 8 steps, including simulating point values, averaging to get block values, transforming to a Gaussian distribution, simulating kriging errors, kriging block values, conditioning on data, and transforming back to get the final simulated block grades. 3. An alternative direct method is described where the block grades can be simulated directly using the structural model and univariate distribution of grades, without first simulating point values.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    17 views2 pages

    VII of If l0 N N Will: Deposits

    Uploaded by

    EndyMpeNk
    1. This document describes a method for simulating block grades by first conducting a non-conditional point simulation and then conditioning the simulation results on actual block data. 2. The method involves 8 steps, including simulating point values, averaging to get block values, transforming to a Gaussian distribution, simulating kriging errors, kriging block values, conditioning on data, and transforming back to get the final simulated block grades. 3. An alternative direct method is described where the block grades can be simulated directly using the structural model and univariate distribution of grades, without first simulating point values.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 2

    VII SIMULATTON OF

    ulalinn will also be

    i: I to n), and the

    rr point

    grades

    nulatbn of blocks,
    ,ariarrce C"(h) or
    lriabitrity of support
    a regularization on
    h) b3z means of the
    :la (II-4 I ). In pracre d.iscrete approxiC6, corresponding
    ,se

    tozero,

    ea

    pbri

    the n the

    dispersion

    0)- As an example

    rangc a, a sufficient
    r/4 can be obtained

    regular square grid


    utated point values
    :s and lhe grades of
    :ntial for a study of
    amptrcs- It will be

    could arisc with storage and calcutation time. Cuts in computer time can be
    obtaincd, firsr by kriging only the N blocks, aod second by simulating and
    kriging directly these N blocks, as shown in the.following scction'

    Non-conditional simulation of points and conditioning on blocks

    This procedure consis(s in carrying out a non-conditional point simulation.


    and rhcn conditioning the simulation by kriging the blocks. More prccisely,
    i( consists of the foltowing eight successive stepsout a non-conditional simularion of Gaussian distributed point
    values a,(x); u,(x) is a realization of the Gaussian RF U(x)deduced
    from rhe initial point RF &(x ) by rhe transform f unction fitted to the

    A. Carry

    experimental histogram of the point data z,r(x.)' i'e', z6(r):


    qz-lulx)) and u(r) = 9)lzs(x)lB. Using the transfOrrn function gzo, transform this non-conditional
    simulation, i.e.,

    z,(r) = cr"lu,(r\-

    C. Average this non-conditionat simulation z,(r) within the required

    blocks, i.e., 2,.,(r ) = (l /n ) Ir eafu,Gt)\, where the unit u is made up


    of thc n interior Points r1.
    (fhesc first threc steps constitute thc standard procedurc for thc nonoaditiooal sirnulation of blocks by averaging non-conditionally simulated

    points.)

    D.

    E.

    ielectioo is always
    >f sample grades),
    iabilily or the uniThe cntire simulae., lhc information

    5rl

    The disadvantagc of the method is that a conditional'simulation of N


    blocks requires a non-conditional simulation of nN points and, more
    significantly, the conditioning process requires kriging thesc nN points- If
    there arc N= 100000 blocks and n = l0 points per block, then problems

    2.

    sity of the discrete


    'ior points, depends
    rould not mask the
    mber n of interior

    DEPOSITS

    F,

    From the hisrogram of the simulated values 2,.,(x ). deterrnine cxperimentally thc Caussian transform f unction for a support u, i'e" z'-'(x ):
    values can be
    ez.{u"',(x)l.Once 94 is known, the Caussian simulated

    deduccd f rom ,,., (, )l 9 | {2,., (r )i. Note that' since rhe function 'p il is'
    in general, non-linear, u,.,(r ) * (l /n ) t, u,(', )
    Using the Gaussian simulated values 1,.,(r ) and a,(xo )' simulate the
    kriging error Ia,., (x )- r, i.,* {r )1. The kriged value a i., x (r } is a linear
    combinatiort of the simulate<J values u, (x. ) at the data points r. e /' i'e''
    ui.,*(r)=I"4.(.t). u,(r"). ln the kriging systcnr. thc lirst memtrer
    matrix consisrs of the experimental covariances E{g,(r )u,(y )} and the
    sccond mcmber matrix consists of E{n,(r )r.,.,(y )}'
    Calculatc rhe krigcd valucsof the blocks which are linear combinations

    *i

    i
    I

    MININ(; CEOSTATISTICS

    )I]

    th,:

    ie.' r'i1r;=
    thc irrrtial rransformed valucs rr('r")= e)t"lzu$")l'

    oI
    I,.,r.(t,). rr(.r.). Notc

    rhar the kriging weights

    A"(r)

    are idcntical to

    thtrsc ttt thc Prcccding stcP E'


    (i. ('orrtlrttoti [rv thctc (]itttsslatt block valtrcs' I c''

    tr,,.{ r. I =' ril

    (.r

    ;+ [rr,.,('r)- rrl'*(.t )l

    H.Firrally.tlterequlrcdctlrrtlitionalsinlulationofblockgradcsrsobtaincd
    for thc
    de rcrnrincd (in srep D)function 92rhc
    by'rrtcattsof

    fruiously

    suPl'rorlr"ic.
    2,.,.(x ) = 97. ( rr,.,.(r

    -llris

    itr;tt

    simulalron
    musl i)c c(r
    in t1t pc n dt
    I

    ti nc rr>l
    ll()n

    e(ir

    1(,

    at (ire

    san'

    lhis. thc d
    gnd (r-, .r

    bc lrss lh
    data (x.,,
    locatrons

    ))

    '

    of nN point vtllues
    blocks' Howcver' the condrlioning

    a non-conclitional srmulation
    1>rt.lccdtrrc rcquires

    N
    rr,(x ). n valrtcs for each tlf thc
    u' rather than krigingall nN Pornts'
    N
    blocks
    ;;;;r, rcqurrcs krigingonly
    thc nN point valucs is still
    lf, howe-ver. rhc nonc<l"nditional simulation of
    blt>cks
    *"tfto<J of direct conditioning of thc N
    rct<t costly' thcn the ritl'a
    should bc ustd'

    i(r..

    -l-

    vulue rif t
    thc smool

    where

    rr:ai data

    zciLl exp

    3. Direct corttlitionttl sinwlation of blocks


    rechniques' lhe structural
    By means of the geostatistical regularization
    f rom the quasi-point
    deduced
    be
    model 7,(h )def rned on tf't support-u can
    lengths' Using
    sample
    core
    to
    structural model y"(Linttta"ftr example'
    VI'B'

    outlined in section
    vartous p"rrn"n"nt" of aistriUution techniques
    Z"(r) can be deduced
    RF
    rhe univariare oirt,iUution of the regularized
    for instance' to the
    fitted'
    RF
    point
    of the
    f rom thc univariate distribution
    of Z'(r)
    distribution
    This
    lenglhs'
    cxperimental hislogram of core sample
    canalsobederlucedfromproductiondataorfromcontroldataattheentry
    to the milling Process'
    and the distribution of 2.,
    once the structural function (C,(h) or 7,(h))
    the block grades can be
    of
    zo'(x)
    31g known, a non-conditional simulation
    process poses a
    conditioning
    carried out directly. However, the subsequent
    few problems.
    on a supPort diflerent from
    The conditional dara aiL-in general, defined
    core sample lengths for instancc' Now'
    u, such as the quasi-point suppx'rt of
    rcquircd'in the conciitioning formula (VIl'3):

    cL:nStruc(

    lrue Srad
    This latt<

    7,,(li ), cf .

    This r
    rrpproxin
    k

    rigcd vi

    rs only'c-l

    thc spati
    only rvht

    rrnit u

    clispe rsic

    Once

    conditio

    point

    ca:

    consicler the terms

    z',(r)- zi.'* (r )l'


    simulation 2"'(r) can be immediately

    2,.,.(r ) = z jx (r )+

    Thc non-conditional block

    ,:
    E
    I

    IL

    from the rcal <lala zo(r.. ) lo providc


    obtaincd. and each block can bc krige<!

    The !
    tiorring

    kriging

    You might also like