30/10/2020

Statistical Process Control

• Statistics and Quality

• Data Collection and Graphical Summaries
• Control Chart
• SPC in Process

Introduction
 Statistical studies can be classified into
two types:
◦ Enumerative and
◦ Analytic
 Enumerative study is to describe a
population in a static sense.
 Analytic study considers the population in
a dynamic sense.
 Overall objective: to predict or improve a
process or product in the future

1
 30/10/2020

• A population can be defined as including all items

(e.g. people, food products, plants, animal, etc)
with the characteristic one wishes to understand.

• Sometimes the definition of a population is obvious.

For example, a manufacturer needs to decide
whether a batch of material from production is of
high enough quality to be released to the customer,
or should be sentenced for scrap or rework due to
poor quality. In this case, the batch is the population.

 Obtaining a portion that is representative

of the whole is referred to as sampling, and
the total quantity from which a sample is
obtained is called the population.
 Sampling is an important aspect of data
collection.

2
 30/10/2020

Note: Sampling from a

population of carrots.
Sample should be
homogenous and
representative of the
whole carrots
population.

Cont…
• Researchers rarely survey the entire
population for two reasons (1) the
cost is too high, and (2) the population
is dynamic in that the individuals
making up the population may change
over time.

• The three main advantages of sampling

are that (1) the cost is lower, (2) data
collection is faster, and (3) since the
data set is smaller it is possible to
ensure homogeneity and to improve
the accuracy and quality of the data.

3
 30/10/2020

Cont.
 The validity of the conclusion drawn from the
analysis of a food depends, among other
things, on the methods used in obtaining and
preserving the sample.
 Sampling and any subsequent sample
preparation may be greatest sources of error
in food analyses.

Cont.
 An ideal sample should be identical in all
of its intrinsic properties with the bulk of
material from which it is taken
(population).
 In practice, a sample is satisfactory if the
properties under investigation
correspond to (representative) those of
the bulk material within the limits set by
the nature of the test.

4
 30/10/2020

Process of sampling
The sampling process comprises several
stages:
1. Defining the population of concern.
2. Specifying a sampling plan.
3. Determining the sample size.
4. Implementing the sampling procedure.
5. Proceed with sample preparation or
sample storage.

Defining the population

 Populations may be finite or infinite.
 Therefore, it is important to clearly define
the population that is to be sampled.
◦ The population may vary in size from a
production lot, a day’s production, a week’s
production, or even a month’s production. So,
need to define it properly!
 Extrapolating information obtained from a
sample to the population can be done
satisfactorily , providing that the sample
analysis is reliable.

5
 30/10/2020

Sampling plan
 The International Union of Pure and Applied
Chemistry (IUPAC) defines a sampling plan as “a
document that predetermined procedure for the
selection, withdrawal, preservation, transportation and
preparation of the portions to be removed from a
population as samples”.
 A sampling plan should be a well-organized
document that establishes the required procedure
for accomplishing the primary objective of food
analysis.
 It should address the issues of who, what, where,
why and how.

Sample size
 The reliability of sampling is dependent on the
sample size.
 The larger the sample size the more reliable the
sampling.
 However, sample size is limited by:-
◦ Time
◦ Cost
◦ Sampling methods
◦ Logistics of sample handling
◦ Analysis
◦ Data processing.

6
 30/10/2020

Sample size…(cont.)

Sampling procedure
 Sampling procedure is subset to the Sampling Plan
and addressed the question “how”.
 Details for the sampling of specific food products
has been developed by organization such as
AOAC International, MPOB, Code of Federal
Regulation (CFR), etc. For instance:-
◦ The AOAC Method 925.08 described the method for
sampling flour from sacks.
◦ MPOB – for sampling and testing palm oil products.
◦ CFR 145.3 method for sampling canned fruits.

7
 30/10/2020

Types of sampling design

Several types of sampling designs:-
 Simple Random Sampling
 Systematic Sampling
 Stratified sampling
 Composite sampling

Simple random sampling (SRS)

 SRS requires that the number of units in the
population be known and each unit is assigned
a number.
 Then, a specific quantity of random numbers
between one and the total number of
population units is selected.
 Various sampling plan are used, including
random number tables and computer-
generated random numbers.
 Units corresponding to random numbers then
are analyzed as an estimate of the population.

8
 30/10/2020

Systematic sampling
 Is used when samples are distributed
evenly over time or space, such as on a
production line.
 The first sample is selected at random
and then every nth unit after that. i.e.
every 10 minutes, 20 minutes or 30
minutes.

Stratified sampling
 Involves dividing the population
into subgroups so that each
subgroup is as homogenous as
possible.
 Random samples then are taken
from each subgroup.
 Followed by applying SRS or
systematic sampling
 The procedure provides a
representative sample because no
part of the population is
excluded.

9
 30/10/2020

Composite sampling
 Is used to obtain samples from bagged
products.
 Two or more samples are combined to
obtain one sample for analysis that
reduces differences between samples.
 i.e. curry powder, 3 in 1 beverages, mixed
coffee and etc.

Method of sampling
 Two methods of sampling commonly
employed in food industries:-
1. Manual sampling
2. Automatic sampling

10
 30/10/2020

Manual sampling
 Taken manually
 Homogenous materials such as single-phase liquids
should be mixed thoroughly before sampling.
• Milk must be thoroughly mixing before the fat rises
to the top and the composition changes on standing.
• Excessive mixing of cream is inadvisable. Formation
of bubbles, aeration, or oxidation may influence the
sampling and analysis.

Manual sampling (cont.)

 Granular or powdered solids (i.e.
cereals/grain) are generally sampled by
probes and triers.

11
 30/10/2020

Manual sampling…(cont.)

 For flowing materials

stored in bag or
container.
 Sample will then
combined and reduced
its size for lab analysis.

Manual sampling…(cont.)
 Some powders materials may
be obtained by coning and
quartering of the materials,
discarding two opposite
quarters, remixing the
remaining material and
repeating the process until
the sample is reduced to a
desired size.
 For reducing sample size,
until small enough sample is
generated prior to lab
analysis.
 Suitable for free flowing
powder/grain.
Coning & Quartering Technique

12
 30/10/2020

Cont.
 Partly or completely frozen, crystallized, or
solidified fluids must be liquefied completely and
mixed.
 Butter and hard cheese samples are generally
taken with a stainless steel borer/spatula.
 Soft cheeses are sampled by cutting out a
representative segment.

Scope for powdery or

cereal products.

Cont.

Microbiological sampling kit

13
 30/10/2020

Automatic sampling
 For sample that is in motion
 i.e. during production of wheat flour
 Several type of automatic sampler has
been developed and applied in the food
industry. For instance:-
◦ Snorkel type point sampler
◦ Auger sampler

Automatic sampling
Equipment for automatic sampling

14
 30/10/2020

Automatic sampling…(cont.)
 Factor that affect the amount of sample
taken by a sampler include:
Size of the cutter opening.
 The speed of the cutter as it travels
through the stream.
 Frequency of making a sample cut.

Automatic sampling…(cont.)
 The minimum width of a cutter opening
should be three times the size of the
largest particles to sampled.
 The cutter should move through the
stream at a speed that gives the smallest
possible sample without deflecting
particles that should enter the cutter

15
 30/10/2020

Sampling error
 Lack of randomness
 Human bias
 Changes in composition
 Non-homogeneity

Storage of sample
• A combination of suitable container, cold
storage and chemical preservation may be
required to maintain sample integrity in
the event of prolonged storage.

16
 30/10/2020

Sample storage
 Unreliable data can be obtained by poor sample storage
resulting in sample degradation.
 Samples should be stored in a container that protects the
sample from moisture and other environmental factors that
may affect the sample (e.g., heat, light, air).
◦ To protect against changes in moisture content, sample
should be stored in an airtight container.
◦ Light sensitive samples – in container made of opaque glass,
or wrapped with aluminium.
◦ Oxygen sensitive samples – under N2 or inert gas.
◦ Refrigeration or freezing may be necessary to protect
chemically unstable samples.

Statistical Process Control

An effective program of process control
aims to:
1. Manufacture a product that consistently
meets specifications
2. Ensure that only acceptable products
are released from each step of the overall
process
3. Favor improved stability
4. Reduce variability in the processing
steps and in the final product

17
 30/10/2020

Design and analysis of studies

 In statistical quality control and statistical
process control, in order to perform an
experiment, it is important to consider
observations under time or production
sequence.
 There is another approach in which time sequence is
not important in the analysis itself. This is considered
off-line experimentation and is known as
experimental design.
 Through this kind of experimentation, it is possible to
discover or confirm cause-and-effect
relationships.
 Experimental design is a better alternative to trial and
error methods because it provides an efficient and
effective means of knowledge acquisition and a higher
level of formality.

TERMS AND CONCEPTS

 Experiment. For the purpose of this section, a study to provide an action to be taken
on a cause-and-effect system.
 Response Variable. Observed or measured variable under given circumstances in the
experiment. It is often the measure of performance or a quality characteristic of the
process or product. It is also called the dependent variable.
 Factor. Independent, causal, or controllable variable. A variable intentionally changed in
the experiment so as to observe its influence on the response variable. Factors can be
quantitative, for example, a cooking time of 30 or 60 min; or qualitative, for example,
method X, Y, or Z. Sometimes, qualitative factors are called class or classification
variables.
 Experimental Unit. The smallest division of material in an experiment such that any
two units may receive different combinations of factors.
 Block. Groups of experimental units treated similarly during a designed experiment.
 Background Variable. Blocking or noise variable; a variable that might affect a
response variable in the experiment but is not of interest as a controllable variable.
Examples include: operator, shift, day, or supplier. Background variables can be
controlled during the experiment by holding them constant or by the use of blocks.
 Level. A specific setting or value of a factor included in the experiment.The levels may
be fixed at certain values of interest, or they may consist of a random sample from a set
of many possible values.The levels of a factor are sometimes called treatments.
 Effect. The change in the response variable caused when a factor is changed from one
level to another.

18
 30/10/2020

SINGLE FACTOR EXPERIMENTS

 The completely randomized design is the simplest
of the single factor experiments.
 Only one factor is manipulated during the
experiment.
 It is assumed that other factors are held
constant or controlled at some level.
 Randomization plays a critical role in
experimental designs.
 There are always a number of uncontrolled
variables; to average out their effect, it is
necessary to randomize the order of
experimentation in the corresponding design.

Example – Single factor experiment

 A manufacturer of sausages is interested in
comparing the smoke intensity obtained by three
different smoking methods: M1, M2, and M3.
 Fourteen samples of smoke intensity are obtained
for each method. A 7-point evaluation scale is
used where 1 denotes “not noticeable amount of
smoking” and 7 denotes “extremely strong
amount of smoking.”
 Table 5.1 shows the collected data and Figure 5.1
is boxplot obtained from analysis.

19
 30/10/2020

Data analysis

 Figure 5.1 shows that methods 1 and 2 lead to

practically the same smoking intensity level while
method 3 provides a lower level of that measured
quality characteristic.
 However, from a statistical point of view, there is
no difference among the three methods.
 Box plot shows:
◦ central tendency and dispersion,
◦ dot in the box represents mean of the dataset
◦ Q1, Q2 and Q3 on the box;
 Q1 – quartile 1, the median of the lower half of the data set
 Q2 – quartile 2, the median of the entire data set
 Q3 – quartile 3, the median of the upper half of the data set

20
 30/10/2020

Control Chart
 A control chart is a graphic display of the actual
quality performance judged against a
reference frame showing a central line
representing the average quality value (m) and
upper and lower lines called the upper control
limit (UCL) and lower control limit (LCL).
 Control charts, as the frequency histogram,
summarize the data but also take time into
consideration, indicate changes in the operation,
tell us within certain limits when the changes are
occurring so corrective action can be taken, and
are a tool for decision making.

Control Charts
 The control charts can be divided into variable and attribute
control charts.
 Variable control charts are used for important variables or
continuous quality data:
◦ Such as weight control, can seam dimensions, volumes, yield, and chemical,
physicochemical, nutritional, and microbiological characteristics.
◦ Numerical values give more information than necessary and the results are more
precise, with fewer samples than attributes; however, variable charts can only be
applied to a single characteristic or property.
 The attribute control charts are used for quality characteristics
data that cannot be expressed as a measurement and are
collected by counting.
◦ For example, in the visual examination of can seams, we can count a variety of
defects such as drops, lips, cut-over, skidding, dents or scratches, etc. and, on the
basis of the results, classify the seams into one of two classes, conforming or
nonconforming (pass/fail, good/bad), or we can express the nonconformities in
units according to their importance.

21
 30/10/2020

Means chart for net weight

22
 30/10/2020

Range chart for net weight

 x = mean value within each sample no.

R = Range difference within each
sample no.

 For = x1+x2+x3+….+xn/n – mean, n is

sample size
 For R = xlarger – x smaller

23
 30/10/2020

grand average or mean of means , X

 X = x1 + x 2 ….. + x n/m, where m is
the number of samples
 R = R1 + R2 …. + Rm/m

Example:
The net weight in ounces of
a product is to be monitored
by
–
X and R control charts using
a sample size of n = 4.
Twenty-five
samples were taken from a
given production line at
regular intervals. Table 5.11
shows the data and the
computed averages
–
x and R.

24
30/10/2020

25
 30/10/2020

Means chart for net weight

Range chart for net weight

26