Quality Assurance-Quality

Control
 Introduction

Mineral
Resource Model
Geostatistics

Chemical Analysis Geological Model

Sample Preparation

Sampling Geological Obsevation

 Introduction
• The goals of a good QC-QA program are:
– Prevent the entry of large errors into the
database used for resource modelling
– Demostrate that sampling and analytical
variances are small
– Provide assurance that the accuracy of data
underlying the resource model can be
confirmed.
 Introduction
• A well designed QA-QC program will:
– Reduce risk in a project
– Possibly reduce required drill spacing
– Reduce geostatitical nugget effect
– Reduce waste mining
– Maximise profit
 Definitions
• Quality Assurance
– Information collected to demostrate and
quantify the reliability of assay data
• Quality Control
– Procedures used to maintain a desired level of
quality in the assay database
 Definitions
• QA-QC programs are intended to monitor routine
sampling and laboratory performance in an effort to
control (minimise) the total possible error in the
sampling-splitting-analysis sequence.

• This is achieved by monitoring and controlling four

eseential comnponents of the sequence:
– Field/underground collection and splitting
– Laboratory sample preparation and splitting
– Analytical accuracy and precision
– Reporting (clerical) accuracy
 Laboratory sample preparation and
splitting
• Audits
• Planned Job Observations
• Standards
• Optimal Sample protocol
• Technology
• External measurement
 Field/underground collection and
splitting
• Planned Job Obeservation
• Standards
• Technology
• Training
• Motivation
• Management attention
 Analytical accuracy and precision

• Routine insertion of quality control

materials, at several critical stages of
sample collection, preparation and analysis
stage.
• Monitoring results on an ongoing basis
• Liason with lab
 Reporting (clerical) accuracy

• Standard checking of results

• Graphical review
• Check outliers
• Digital delivery of results
• Standardised procedure
• Computerised systems
 QC-QA : Reporting accuracy

• Most Common errors:

• Shifting results by one drill hole sampling interval
• Shifting the location of the decimal point to left or
right
• Entering data in wrong field
• Transposing digits
• Misreading / guessing results off illegible faxes,
copies or notes
• Handeling data below detection limits
• Miss co-ordinating samples
 QC-QA : Reporting accuracy

• Quality Assurance
– Checking a random 5% of samples
• Error should be less than 0.5% ie five errors per
thousand samples
– Check highest grade 3% of samples
• If error greater than 0.5% the check next 3%
– For feasibility study:
• Data should be entered by two persons and then
compared
 Types of QA-QC Samples
• Often termed controls
• Three basis types:

– Standard Reference Material (Standards)

– Duplicates
– Blanks
 Standards
• Required characteristics of a good
standard:
– Matrix identical to samples being analysed
– Extremely low hetrogeneity
– Grade within ore grade range
– Reliable prepared
– Accuratly charaterised
 Types of Standards
• Commercial
– CCRMP : Canadian Certified Reference Materials
Project
– CANMET : Canada Centre for Mineral and Energy
Technology
– ROCKLABS : Australia
– GEOSTATS : Australia
• In House Standards
– Mine or company specific standards
 In House Standard
• A number of bulk standards are suggested:
– A standard with a mean near ore/waste cut-off
– A standard with gold grade near the 85Th
percentile of all ore samples
– Standards with grades near any other potential
decision points
– Standards covering all major rock types eg
oxide and sulphide
 Testing In House Standards
• 24 Randomised “grab” sample or representative split of pulp material
• Submit to high quality lab
• Use best analytical technique available given matrix and expected grade
• To be run as single batch so as to reduce sample mix up
• Test designed to test homogeneity
• If relative STD exceeds labs claims for method and grade range then do
additional work to reduce hetrogeneity
• If grades 15 to 20 times > than the detection limit 90% of values must be
within 5% of mean and all values within 10% of mean.
• More latitude for lower samples.
• Difficult for coarse gold deposits.
– Screen out coarse gold
 Testing In House Standards
• Once Homogeneity established
• Split standard and place in individual packets
• Establish best value:
– Submit to a series of accredited labs
– Submit as two varying sized separate lots with a total of 8 – 10 samples
– Hide nature of submittion by sending check pulp assays as well.
– Restrict mass so as to avoid repeated assays of same sample
– Calculaled mean and STD after excluding labs with highest and lowest values.
– Discard any values outside 3 STD’s and recalculate the STD.
• Beware drift
-Package and randomise
• Riffle or rotary splitters
 Blanks
• Samples containing negligible, insignificant or
zero gold
• Help monitor carry-over contamination
• Help monitor mis-sequencing of samples during
preparation and analysis
• Should be of similar matrix to samples
• Should undergo same sample preparation
• Should be stored away from samples and other
materials containing gold.
 Blanks
• Two types:
– Coarse Blanks:
• Inserted before sample preparation phase
• Should return values less than or equal to twice the detection limit
80% of the time
• Designed to test for contamiation from sample preparation
equipment
• Tests for cross contamination from dust
– Pulp Blanks
• Inserted after sample preparation
• Should return values less than or equal to twice the detection limit
90% of the time
• Tests for cross contamination in anlysis such as boil overs of gold in
bricks
• After correcting swap sample and replacing results from failed
batches the mean value should be less than the detection limit
 Duplicates
• There are two types of duplicate sample:
– Field Duplicates: These are samples taken twice
at the same location in the field.
– Laboratory Duplicates: This is when a laboratory
splits a sample at some stage during or after
sample preparation but prior to weighing i.e. a
single aliquot is taken from each split for
weighing.
 Replicates
• There are two kinds of replicate
– Field Replicate: This is when two portions or
aliquots are submitted to a laboratory from the
same sample bag, often with different sample
identification. The laboratory would then weigh
out and assay separate aliquots from each bag.
– Laboratory Replicate: These are repeat analyses
undertaken by a laboratory at the same time as
the original assays with the sample material
being derived from the same aliquot
Alternative View
 Duplicates
• There are three types of sample duplicates:
– Check Laboratory Duplicates
– Coarse-reject Duplicates
– Field(Drill rig) Duplicates
• Other samples refered to as “duplicates”
– Split Core
– Twinned holes
– Twinned samples
 Check Laboratory Duplicates
• Splits from final routine pulp
• Sent to umpire lab
• Can be re-labeled and resubmittet to
primary lab
• Measures accuracy of the initial
determination
• Allows an estimate of the analytical
variance + pulp sub-sampling variance
 Coarse-reject Duplicates
• Splits taken at point where the sample mass is
first reduced
• Often after first crush
• Best analysed by primary lab
• Checks on the adequacy of sample preparation
• Provides a measure of the variance included in
the check assays + variance introduced by sample
preparation procedures + sub-sampling variance
of the coarse reject
 Field(Drill rig) Duplicates
• Duplicate splits collect at drill site
• Can provide an estimate of the variances
measured by coarse rejects + sub-sampling
variance introduced by the drill site’s
sample splitter
• Results often difficult to interpret
 Twinned “Duplicates”
• Split core
– Spatially separated
– Adds geological variance
• Twinned holes
– Even greater spacial separation
– Difficult to evaluate
• Different drill techniques
– Compare on distribution basis
– QQ plot
Insertion of controls into batches
• Standards should be submitted to the lab in
sufficient number and frequency such that
each lab batch contains at least one ore
grade standard, one blank and one low
grade standard or pulp duplicate
• The lab internal; QC-QA program should
include at least one duplicate and/or one
control sample and one blank.
 Insertion of blanks
• Essential as a method of controlling lab error as
ore/waste cutoff approaches the detection limit
• Rule of thumb: if the ore-waste cuttoff is less
than 15 times the detection limit then insert a
blank in every batch of samples
• Coarse blanks sould be submitted such that one
sample is procesed per shift
• These blanks should be inserted after visuallt
identified ore material
 Insertion of standards
• Low grade standards should have grades close to
the anticpated cutoff grade ie 0.4 to 1 g/t
• High grade standards should be around the 85th
percentile of all ore grade samples previously
assayed
• A further high grade standard should have grades
at or above the routine linear range of the AA
• Other standards could be at the median grade of
the ore grade samples, at the median grades of
the oxide and sulphide samples if they vary
significantly
Suggested amounts of QA-QC materials

• What is a batch – a group of samples that

undergoes processing as a group
• Minumum program recommended by MRDI:
– An ore grade standard and a below ore grade or blank
sample per batch. At least 5% of samples submitted
– One blank per batch with additional coarse blanks
inserted following visually identified high grade
samples.
– 5% coarse reject duplicates and as much as 5%
renumbered (same pulp) resubmitions
 What now?
• Analytical results of the standards will be
used to evaluate and to accept or reject lab
batches as a result
• Coarse rejects are used to evaluate sample
preparation process
• Coarse blanks are used to evaluate both
sample preparation and assaying
 Maintaining anonymity of QA-QC Material

• Difficult to disguise standards as they are submitted as

pulps as are resubmitions
• Could in extraordinary circumstances prepare samples in
separate labs
• Can also use pair groups of standards
• Supply lab with two boxes :
– One with low grade standards, blanks and pulp duplicates
– The other with high grade standards and duplicates with high
grades results
• One of each then inserted into each batch
• Ensure not run separately and analysed in duplicate or
triplicate
Quality control Procedures and Criteria

• Batch Acceptance Criteria

• Generally done by batch
• However most errors are “one of a kind”:
– Misreading a number on the balance
– Transposing digits
– Spillage
– Boil over of a single sample
– Transient partial blockage in fluid flow of the AA
– Switching two samples
• Batch errors are rare:
– Reading a rack of samples backwards
– Reading wrong rack of samples
– Heating a group of samples for the incorrect time
– Using contaminated reagents
 Batch Failure Criteria
• For Standard reference material:
– Accepted range should be the sought value +/- 2std
(nominally 95 % confidence interval)
– NB: This is not the standard error of estimate for the
certified value but the range that encompasses 95% or
the determinations obtained during certification
– This tolerence is generous even in a mine lab
– Less than 5% of samples must fall outside limits
 Batch Failure Criteria
• Pulp blanks should return values less than
or equal to twice the detection limit.
• Coarse balnks should not exceed three
times the detection limit
 Batch Failure Criteria
• Duplicate analyses (Lab A and Lab B) performed
on coarse duplicates should be within +/- 20
relative percent (pair difference divided by pair
mean)
|A-B/0.5*(A-B)| < 0.2
• Duplicate analyses (Lab A and Lab B) performed
on pulp duplicates should be within +/- 10
relative percent (pair difference divided by pair
mean)
|A-B/0.5*(A-B)| < 0.1
 Laboratory Drift
• This is measured by plotting the mean
value of each standard against a time
sequence
• Similarily the difference between the
averages in the original and cheack assay
must be plotted over time
 Graphical presentaion of data
BLANK VALUE vs SAMPLE ASSAY
0.45

0.40

0.35
ASSAY (g/t)

0.30

0.25

0.20

0.15

0.10

0.05

0.00

106
113
120
127
134
15
22
29
36
43
50
57
64
71
78
85
92
99
1
8

NUMBER
 Blanks
• Plot as a time series or number series plot
• Plot as a bar graph
• Calculate the effective detection limit –
– Mean of blank assays + 2 STD’s
– Ensure all suspect samples are removed before
calculating ie sample swops etc
– The effective detection limit for production labs will
be in the order of 0.2 g/t
– Must have a significant volume of data before
calculating – at least six months data
 Standards
C o m p a ris o n o f S ta n d a rd s

2.0

1.8

1.6

1.4
Repeat Value

1.2

1.0

0.8

0.6

0.4

0.2

0.0
0.0 0.5 1.0 1.5 2.0
S ta n d a rd va lu e
 Value
g/t

-0.10
0.10
0.20
0.30
0.40
0.50
0.60
0.70
1

13
25

37
49

61

73
85

97
109

121

133
145

157
169

181
193

Date 205
Standards

217
229

241
253

265
Navachab Lab Standard - 0.17 g/t

277
289

301
313

325

337
349
 Standards
D is t r ib u t io n o f v a lu e s d e r iv e d f o r 0 .1 7 g /t
S ta n d a r d

160

140

120

100
Frequency

80

60

40

20

0
0 0 .1 0 .2 0 .3 0 .4 0 .5 0 .6 0 .7 0 .8
V a lu e M id p o in t
g /t
 Standards
• Note:
– Ensure you use correct STD for calculation of
limits
– Assay aceptable if 95% of values inside 2
STD’s
– However still look for bias in results
– CUSUM tool is useful in determining major
inflection points in time analysis
 CUSUM

S ta n d a r d A ssa y (2 . 5 3 g / t)

3 .2 0
3 .0 0
2 .8 0
2 .6 0
grade (g/t)

2 .4 0
2 .2 0
2 .0 0
1 .8 0
1 .6 0
1 27 50 81 101 121 142 163 183 203 229
d ate

as s ay ul ll me a n 1 0 p o in t
 CUSUM

Cusum

1.50
1.00
0.50
cusum value

0.00
-0.50 0 50 100 150 200 250

-1.00
-1.50 Problem
-2.00
day
 CUSUM
• The CUSUM is the cummulative sum of
all (samples less the mean) plotted over
time
• It illustrates majot inflectionn points over
time and can be used to identify the time of
major change in lab procedure/effeciency
 Check Assay
Navachab - Mine versus AARL
4

0
0.0 5.0 10.0 15.0 20.0
Difference

-2

-4

-6 y = -0.2317x + 0.2098
2
R = 0.5701
-8

-10
AARL Gold g/t
 Check Assay
• Difference = Original assay value – check
assay value
• Not a robust techniques
 Check Assay
Ratio of AARL to Mine

300

250

200
AARL/Mine

150

100

50

0
0.000 0.500 1.000 1.500
AARL Grade

Data 20 Point Moving Average

 Check Assays
• Ratio = Original sample value / Check
sample value
• Can be used to test for lab drift
• Plot time series plot of ratio versus time
– Use same techniques as with standards
 Check Assay
Regression analysis - Navachab

20.000

18.000

16.000 y = 1.2181x - 0.1777

2
R = 0.9358
14.000

12.000

10.000

8.000

6.000

4.000

2.000

0.000
- 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 20.00
-2.000

Data Ideal Linear (Data) Linear (Ideal)

 Check Assay
y = 0.3503x +
Regression analysis - Navachab 2
R = 0.341

0.600

0.500

0.400

0.300

0.200

0.100

0.000
- 0.10 0.20 0.30 0.40 0.50 0

Data Ideal Linear (Data)

 Check Assay
• Regression analysis is a powerful tool
• BUT:
– Assumes normality in a skew distribution
– Need to look at grade ranges independantly
– Look for outliers and investigate
– Remove clear sample swops and errors
– Beware lollypop effect
– Treat low grade noise near the detection limit with
caution
 Check Assay
Histogram of Gold Deviations
Deviation of Mine from AARL

300

250

200

150

100

50

0
-5 -4 -3 -2 -1 0 1 2 3
 Check assays
• Calculate % Deviation (pair diff/pair
mean)
• Plot shows dierection of bias
 Check Assay
AARL versus Mine assays - Navachab
All Samples

2
1.8
Abs. Relative % Difference

1.6
1.4
1.2
1
0.8
0.6
0.4 Recommended for Coarse Duplicates
0.2 20%

0
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Percentile Rank
 Check Assay
• The relative difference versus percentile rank plot is
extremly useful
• Plot Absolute value of (pair diff/pair mean) versus
percentile rank
• Allows for measure of overall accuracy.
• For Coarse duplicates expect less than 20% error for
90% of samples
• For pulp duplicates expect less than 10% error for
90% of samples
• These limits will be higher for coarse gold deposits
• Values less than 2 x detection limit should be
excluded from calculation