QUALITY MANAGEMENT

TOOLS
 Quality Management Tools
(American Society for Quality)

1. Stratification
2. Fishbone diagram
3. Check Sheets
4. Histogram
5. Pareto chart
6. Scatter diagram
7. Control chart
 Stratification
• Stratification: The act of into
.
(American Society for Quality)

• Systematically segregated and sorted data can reveal the patterns for managerial decision making

100

99.8
Process 1 Process 2 Process 3
Input (x) Output (y) Input (x) Output (y) Input (x) Output (y) 99.6

1 0.45 98.95 1 0.6 98.85 1 0.5 98.25

99.4
2 0.27 99.05 2 0.63 98.95 2 0.4 98.35
3 0.26 99.3 3 0.21 99.45 3 0.45 98.6 99.2

4 0.1 99.3 4 0.29 99.55 4 0.34 98.65 99

5 0.24 99.4 5 0.34 99.55 5 0.29 98.75

98.8
6 0.4 99.55 6 0.24 99.65 6 0.21 98.75
7 0.22 99.55 7 0.25 99.75 7 0.31 98.85 98.6

8 8 8 0.22 98.85 98.4

9 9 9 0.13 98.95
98.2
1
0 10 10 0.11 99.05
98
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
 Stratification

■ Which data require stratification:

■ When to use stratification:
(American Society for Quality)

– Equipment
– When data come from several sources or
conditions, such as shifts, days of the week, – Shifts
suppliers, or population groups. – Departments
– Materials
– When data analysis may require separating – Suppliers
different sources or conditions – Day of the week
– Time of day
– Products
 Flow Charts

■ Flowchart demonstrates the

process flow by identifying the
each intermediate step where
operations are performed. Each
intermediate step has an input
and an output which are brought
into management attention for
review and analysis.
 Ishikawa Diagram (Cause-Effect Diagram)
• This is also known as Cause–and-effect Diagram, or Fishbone Diagram or Ishikawa diagram is used for
identifying the which are causing an undesired effect. causes may also vary.

The factors are generally classified under:

• Method, Manpower, Machinery and Material

• Policy, Procedure, People and Plant

• Surroundings, Suppliers, System, Skills

These factors can be subdivided further to

identify precisely the actual cause of the effect.
 Check Sheets
• A structured, prepared form for collecting and
analyzing data; a generic tool that can be adapted
for a wide variety of purposes.

• A check sheet or check list is a data gathering tool

where measurements (e.g., number of quality
errors or defects) are recorded for a given list of
items.

• Helps to make fact based decisions instead of

anecdotal evidence
Check Sheets

▪ When to use check sheets:

➢ When data can be observed and collected repeatedly by the same person or at the same
location
➢ When collecting data on the frequency or patterns of events, problems, defects, defect
location, defect causes, or similar issues
➢ When collecting data from a production process
 Histograms
Use Histograms when you need to know:
• Central Tendency of the distribution
• Variation in the distribution
• Compare the two different
distribution
• If the shape of the distribution is
normal What if there
• if the data is falling in a are two peaks
predetermined pattern
 Pareto Charts

• A Pareto chart is a bar graph.

• The lengths of the bars represent
frequency of the quality attribute, and
are arranged with longest bars on the
left and the shortest to the right. In
this way the chart visually depicts
which situations are more significant.
Pareto Charts

■ The assumption:
– Often referred to as the 80–20 rule, the Pareto concept states that
approximately 80 percent of the problems come from 20 percent of the items.
For instance, 80 percent of machine breakdowns come from 20 percent of the
machines, and 80 percent of the product defects come from 20 percent of the
causes of defects.
■ When to use pareto Chart:
– When analyzing data about the frequency of problems or causes in a process
– When there are many problems or causes and you want to focus on the most
significant
– When analyzing broad causes by looking at their specific components
 Pareto Charts

Type of Defect Frequency of % Total Cumulative %

Defect
Button Defect 23 39 39
Pocket Defect 16 27.1 66.1
Collar Defect 10 16.9 83
Cuff Defect 7 11.9 94.9
Sleeve Defect 3 5.1 100
Total 59 100
Scatter Diagram
• Scatter diagram depicts the relationship
between two variables X & Y.

• E.g., X can be humidity and Y can be

number of errors per hour detected during
and operation.
Scatter Diagram
■ When to use scatter diagram?

– When you have paired numerical data which are related

– When your dependent variable may have multiple values for each value of your
independent variable
– When trying to identify potential root causes of problems
– After brainstorming causes and effects using a fishbone diagram to determine
objectively whether a particular cause and effect are related
– When determining whether two effects that appear to be related both occur
with the same cause
 Control Charts
• A control chart is used to study how a process changes over time.

• Comparing current data to historical control limits leads to conclusions about whether the process
variation is consistent (in control) or is unpredictable (out of control, affected by special causes of
variation).

• It can help detect the presence of correctable causes of variation.

Statistical Chart of a time ordered values of a sample statistic

QUALITY CONTROL
Reference Text

■ Chapter 10: Quality Control

– William Stevenson, Operations Management — 13th ed., McGraw Hill
Education, NY
 Introduction
■ Quality control is a process by which one can evaluate the output of a
process based on the pre-stablished standards and corrective actions are
taken when output doesn’t meet those standards.

■ When can we apply quality control ?

– When there is a way to measure quality in terms of specific parameters
– When we have preestablish quality standards
– When we have check points to monitor the quality

■ Preventive or corrective ?
 Quality Appraisal

■ Inspection:
– Inspection is an appraisal activity that compares goods or services to a
given standard.

■ Three important questions:

– What to inspect ?
– Where to inspect ?
– How much to inspect ?
Points of inspection

Process Control

Quality of Inputs Quality of Transformation Quality of Outputs

Pre-production Production Post-production

Acceptance Sampling
Statistical Sampling to determine where to
accept or reject the production lot.
Inspection Methods

Non-destructive test

Destructive testing
 How much inspection is adequate ?
■ How much effort one should make in inspection ?

– The amount of inspection can range 0% to 100%

– 100% inspection? Manufacturers typically

use to pull 1-10% of the
■ Inspection cost will be very high items for inspection
■ Inspection time will be long
■ Infeasible choice in destructive testing Ø Low cost high volume items requires little
inspection. e.g., paper clips, roofing nails,
and wooden pencils
– 0% inspection ?
■ Cost of passing the defectives will be very high Ø High-cost, low-volume items requires more
intensive inspection
 Statistical Process Control (SPC)
■ SPC evaluates a process output to decide if a process is “in control” or if corrective action is needed.
– It involves a set of techniques to ensure that the process conforms to the requirements.

■ Acceptance Sampling
– Used to infer the quality of a batch of products by sampling a specified number of items for testing.
– Collect Sample for measurement (one or multiple random samples)
– Size of sample and number of samples depend on how much risk is assiciated with the product

■ Process Control
– Define the type of measurements
– Variables: Measures on a continuous scale (e.g., 23.5mm, 0.35 gms, 36 seconds. )
– Attributes: Measured on a discrete scale (e.g., pass-fail, fraction of defectives, number of defects on
a surface, etc.)
 Process Variability

■ Random Variations
– Chance or common causes inherent in every process
– Deviation in the process output are natural in every
process
– Created by countless minor factors

■ Non-Random Variations
– Deviations in the process output can be clearly identified
and managed
– Systematic or assignable causes present in the preocess
– Process need a corrective actions
– In manufacturing set-ups, some of the causes are variation
primarily include - operational inputs such as raw materials,
personnel, equipment, measurement methods, etc. and the
environmental conditions.
 Statistical Process Control (SPC)
■ Periodic samples of process output are taken
■ Sample statistics, (e.g., sample means) are measured.
■ For example, if you have taken n = 50 samples, having a sample size of 100 units
– Their means may be denoted by �1, �2, �3, … �50,
– The variability in sampling statistics (��)can be described by its sampling distribution.

Sampling distribution is a (theoretical) probability distribution obtained through a large

number of samples that describes the random variability of sample statistics.
Sampling Distribution

Sampling distribution

Distribution of process output

Sampling Distribution:

§ The variability of sample statistic can be described by its sampling distribution,

§ It is a theoretical distribution that describes the random variability of sample statistic.

 Sampling Distribution Acceptable Region (± 3σ)

Acceptable Region (± 2σ)

■ A sampling distribution serves as the theoretical Acceptable

Region (± σ)
basis for distinguishing between random and
non-random values of a sampling statistic.

■ Limits are selected within which most values of a

sample statistic should fall if its variations are
random (e.g. ± 3σ, ± 2σ etc.)

■ If the value of a sample statistic falls outside

those limits, there are high probability that
assignable causes might be present
– or, in other words, there is only a small
probability that the value reflects randomness. - 3σ - 2σ -σ -σ + 2σ - 3σ
Sampling Distribution
Using Sampling Distribution for Control Charts

Non random
Variation
LCL: Lower Control Limit

Zone of random/Acceptable variation

*
*

*
(for ± 3σ)
values

*
*

*
*

*
*
LCL: Lower Control Limit
Non random Sample ID (sample number/time order)
Variation

Control Limits are based on

sampling distribution
Control Chart
■ A time-ordered plot of sample statistic.
■ It is used to distinguish between random variability and non-random variability.
■ Elements of a control chart:
– Centerline (generally the the meanvalue of the sampling distribution)
– Control Limits (UCL & LCL)

Sample no. 12 is out of

control Variation due to assignable causes

Variation due to
random/chance causes

Variation due to assignable causes

 Type 1 and Type 2 Errors
■ Type 1 Error:
– Probability that a value will fall outside the limits even
though only random variations are present
There is always a small probability that a value will
fall outside the limits even though only random
– Concluding that variations are present.

■ non-random variations are present even though only

random variations are present
■ a process is not in control when it actually is. Non-Random Non-Random

– Also known as
■ producer’s risk Normal
■ false rejection (Random)

■ false fail α: Probability of making type 1 error

■ false positive = sum of probabilities in the two tails
 Type 1 and Type 2 Errors

■ There is always a probability that a value will fall inside the limits even though non-random variations are
present.

■ Type 2 Error:
– Concluding
■ non-random variations are not present, when they are
■ a process is in control when it is really out of control
– Also known as
■ consumer’s risk
■ false pass
§ The costs of making each error should be balanced by their
■ false acceptance probabilities
■ false negative
§ In practice, two-sigma limits and three-sigma limits are commonly
used as reference.
 The Control Process
■ Define: The set of characteristics which are suppose to be controlled
■ Measure: Establish the measurement method
■ Compare: With the quality standard (e.g., upper and lower control limits)
■ Evaluate: Establish a definition of ‘out of control’
– Distinguish random from non-random variability
– Non-random variability indicates that the process is out of control
■ Correct: Eliminate the cause of non-random variability
■ Monitor: To ensure that corrective action has a sustained effect
 Statistical Stability of a process

Which process is stable over time ?

Process is Stable over time Process Unstable over time

What to do if the process is unstable?

Process variability Vs Customer Specifications

Voice of the Process Voice of the Customer

(VoP) (VoC)

■ Each process shows some random ■ Provides the acceptable level of

variability variability
■ One can define control limits on ■ One can define specification limits on
the basis of sampling distribution the basis of customer’s requirements
■ UCL is the upper control limit and ■ USL is the upper specification limit
LCL is the lower control limit and LSL is the lower specification limit.
 Type 1 and Type 2 Errors

■ There is always a small probability that a value will fall

outside the limits even though only random variations
are present. Non-Random Non-Random

■ Type 1 Error:
– Probability that a value will fall outside the limits Normal
even though only random variations are present (Random)
– Concluding that non-random variations are
present even though only random variations are
present
– Concluding a process is not in control when it α: Probability of making type 1 error
actually is. = sum of probabilities in the two tails
– Also known as producer’s risk
 Type 1 and Type 2 Errors

■ There is always a probability that a value will fall inside the limits even though non-random
variations are present.

■ Type 2 Error:
– Concluding non-random variations are not present, when they are
– Concluding a process is in control when it is really out of control
– Also known as consumer’s risk
Control Charts

■ Control chart detects variation in processing and warns if there is any deviation from the specified control
limit (Voice of the Process)

■ The purpose is to monitor and stabilize the process through detecting and controlling the non-random
causes of variations.

■ What to monitor ?
– Variable data: Measured (usually on a continuous scale amount of time needed to complete a task,
length or width of a component)
– Attribute data: Counted (e.g., the number of defective parts in a sample, the number of calls per
day)

■ Sampling for variables:

– Each sample must be taken at random
– The size of sample is generally kept as 5 but 10 to 15 units can be taken for sensitive control
charts.
Types of Control Charts

Control
Charts

Control
Control Chart
Charts for
for Attributes
Variables

Fraction of Defects per

Mean chart Range chart
defectives unit
 Control charts for variables
§ “x-bar” chart: Control chart for mean; monitors the of the process)
§ ‘R’ -chart: Control chart for range; monitors the dispersion of the process)

Mean is shifting Range is shifting

 Control charts for attributes

• p-chart • c-chart
• np-chart • U-chart

• When items are classified in two categories: • When items can’t be classified in two categories:
• e.g.; good/bad; pass/fail, acceptable & unacceptable, • e.g., number of accidents per year, number of bacterea in a
conforming & non-conforming, defective/nondefective water sample, number of scrateches on a polished surface, etc.
etc.

§ p-chart: Percentage/Fraction of defects ---------for varied sample

§ np-chart: Percentage/Fraction of defects------- for constant sample
§ c-chart: Defects per unit----------------------------------- for constant sample
§ U-chart: Defects per unit----------------------------------- for varied sample
 Illustration:
A quality inspector took five samples, each with four observations, of the length of time for glue to
dry. The analyst computed the mean of each sample and then computed the grand mean. All values
are in minutes. Use this information to obtain three-sigma control limits for means of future times. It
is known from previous experience that the standard deviation of the process is .02 minute.

Is the Process in control ?

Sample Average n = sample size

Grand Average
 x-bar (mean) chart: When Process Deviation is not known

A2 = .73 for n = 4 (from the reference table).

Sample Average

Sample Range R
Results are similar to the earlier case when
Average of process deviation was known
Sample
Range
 Example: Range Charts
Control Limits for the Range Charts Control Limits for the Range Charts when
process Standard Deviation is known

Sample Range of
five samples are

The values are R are Within the Control Limits

 Using Mean and Range Charts Together

■ Mean charts are sensitive to shifts in the

process mean, it fails to detect the dispersion

■ Range charts are sensitive to changes in

process dispersion, it fails to detect the shift in
mean

■ Thus, use of both charts provides more

complete information than either chart alone
 Control Chat for Attributes: p-chart

■ Used to monitor the proportion of defectives

§ When the data consists of multiple samples of n observations each

§ Theoretical basis is Binomial Distribution (when sample size is small)

§ The centerline on a p-chart is the average fraction defective in the population, p.

§ Control limits are obtained on the basis of about 20-25 samples and standard deviation
from the samples is calculated for further production control.
p-chart continued
Control limits:
Where, ‘p’ is fraction defective in the population

If p is unknown:
§ it can be estimated from samples
§ p̅replaces p in the preceding formulas
 Example:
An inspector counted the number of defective monthly billing statements of a telephone company in
each of 20 samples. Using the following information, construct a control chart that will describe 99.74
percent of the chance variation in the process when the process is in control. Each sample contained
100 statements.

?
 Control Chat for Attributes: c-chart

§ In manufacturing, sometime it is required to control burns, cracks, voids, dents, scratches,

missing and wrong components, rust etc.

§ Used to monitor the occurrences of defectives; Non-concurrences cannot be counted

§ Works with count type data, e.g. total number of nonconformities per unit

§ The underlying sampling distribution is the Poisson distribution.

§ Assumes that defects occur over some continuous region and that the probability of more than
one defect at any particular point is negligible.
Control Chat for Attributes: c-chart
c = number of occurrences (e.g., defects) per unit

If the value of c is unknown, use ̅c

̅c is the ratio between the total number of defects

found in all samples and the total number of samples
inspected.
Example:

?
 Are all variations within the control limit random ?

Trend

Types of non-random patterns

Cycle

Bias

High Dispersion

Next, go for a ‘Run test’

Cluster
 Run Test
■ A run is defined as a sequence of observations with a certain characteristic,
followed by one or more observations with a different characteristic.
■ It analyzes the occurrence of similar events that are separated by events that are
different

How does it work ?

■ For example, a series of 20 coin tosses might produce the following sequence of
heads (H) and tails (T).
– HHTTHTHHHHTHHTTTTTHH
■ Check for patterns in a sequence of observations
– The number of runs for this series is = nine.
 Run Test

■ Two useful run tests involve examination of:

– The number of runs up (U) and down (D) (The first value does not receive either a U or a D
■ U/D Runs =3 because nothing precedes it.)

– The number of runs above (A) and

below (B) the median (or mean) Median = 36.5
■ A/B (Median) Runs =3
Illustration: Run chart (number of runs)

7 runs

8 runs
Run Test
■ Transform the data into both As and Bs and Us and Ds
■ Count the actual number of runs (r) in for N observations
■ Calculate the expected number of runs (Er)
■ Calculate the standard deviation of number of runs (σ)
■ Calculate the z-score
■ Decide on the basis of z-score (assuming acceptable limit lies between +2 to -2)

Expected Number of Median Runs Expected Number of Up/Down Runs

Run Test
■ Calculating Z –score

For Up/Down Runs For Median Runs

N = Total number of observations
r = Observed number of runs
 Example:
Twenty sample means have been taken from a process. The means are shown in the following
table. Use median and up/down run tests with z = 2 to determine if assignable causes of variation are
present. Assume the median is 11.0.

Number or observations ?

N =20

Number or runs ?
Process Capability
At any given instance, the output of a process may or may not conform to
specifications even though the process may be in statistical control

Voice of the Process Voice of the Customers

▪ A range of acceptable values established by

▪ Limits of manufacturing abilities engineering design or customer requirements
▪ Includes random process variability ▪ Translated down to target value and
Specification Limits (USL & LSL)

• Process capability refers to the inherent variability of process output relative to the variation allowed by
the design specifications

• If the random variations that a process is capable to produce lies within an acceptable specifications,
the process is said to be “capable.” If it is not, the manager must decide how to correct the situation.
Capability Analysis
Voice of the Process Voice of the Customers

Example: a) LSL = 8.5 & USL = 11.5

Process Av is = 10 & Process SD is =0.5

• 3σ control limits would be 10 ± 1.5) b) LSL = 7 & USL = 13

• UCL = 11.5
• LCL=8.5 c) LSL = 9 & USL = 11

Process capability index

Comparing
VoP & VoC
Process Width = UCL – LCL
= 6𝜎 (-3σ to + 3σ)
▪ Cp ≥ 1.33 is used as industry standard
Specification Width = USL − LCL
 Capability Analysis
Lower Upper
Lower Upper Specification Specification
Specification Specification

8.5 10 11.5
8.5 10 11.5
a) LSL = 8.5 & USL = 11.5 b) LSL = 7 & USL = 13
LCL= 8.5 & UCL = 11.5
LCL= 8.5 & UCL = 11.5

Lower Upper
Specification Specification
Process capability index a) Cp= 3/3 = 1
b) Cp= 6/3 = 2
c) Cp= 2/3 = 0.67

8.5 10 11.5
c) LSL = 9 & USL = 11
LCL= 8.5 & UCL = 11.5
 Only B is capable (Cp ≥ 1.33)

Process capability index

Process Width = 6𝜎
Specification Width = USL − LCL
= 0.80
 Three-sigma versus Six-Sigma capability

▪ The goal of 6σ 𝐶𝑎𝑝𝑎𝑏𝑖𝑙𝑖𝑡𝑦isachieving a process

variability so small that the design specifications
represent six standard deviations above and
belowthe process mean.

▪ Achieving a process capability index equal to 2.00

 If a process is not centred
■ Cp index computation does not involve the process mean
– The assumption is that, target value (i.e., process mean) is centred between the upper
and lower specifications

■ Cpk index is used instead of Cp

■ Cpk = Smaller of
,

i.e., Smaller of [Cpu , Cpl]

 Mean
Check

7.5
Process mean is centred between USL & LSL
4.6

6.1 Process mean is not centred between USL & LSL

▪ Calculate Cp index for processes 1 & 2

▪ Calculate Cpk index for the process 3

= Smaller of
capable

Not capable
Not capable
In order to be capable, Cp & Cp k must be at least 1.33