Scribd
Upload
163 views

Taguchi Loss Funcation

This document provides an overview of Taguchi Loss Function and related quality concepts: - Taguchi introduced the loss function concept which combines cost, target, and variation into a single metric, with specifications being secondary. He developed concepts of robustness and nominal-the-best loss functions. - Loss functions quantify societal losses from products failing to meet customer needs. Taguchi developed over 68 loss functions but most situations use the quadratic nominal-the-best function. - Orthogonal arrays provide a simplified method for designing experiments involving multiple factors and interactions. The degrees of freedom determine the appropriate array size. Linear graphs aid in assigning factors and interactions to array columns. - Signal-to-noise ratios

Uploaded by

Jyoti Arvind Pathak
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
163 views

Taguchi Loss Funcation

This document provides an overview of Taguchi Loss Function and related quality concepts: - Taguchi introduced the loss function concept which combines cost, target, and variation into a single metric, with specifications being secondary. He developed concepts of robustness and nominal-the-best loss functions. - Loss functions quantify societal losses from products failing to meet customer needs. Taguchi developed over 68 loss functions but most situations use the quadratic nominal-the-best function. - Orthogonal arrays provide a simplified method for designing experiments involving multiple factors and interactions. The degrees of freedom determine the appropriate array size. Linear graphs aid in assigning factors and interactions to array columns. - Signal-to-noise ratios

Uploaded by

Jyoti Arvind Pathak
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 25

Taguchi Loss Function

Introduction
• Significant body of knowledge associated with quality
sciences was developed in the UK as design of experiment
(DOE) and in the US as Statistical Quality Control (SQC)
and in Japan as Total Quality Control (TQC)
• Dr. Genichi Taguchi, Mechanical Engineer, winner of 4
Deming Prizes has contributed significantly
• Introduced loss function concept which combines cost,
target, and, variation into one metric with specifications
being of secondary importance.
• Developed the concept of Robustness, which
means noise factors are taken into account to
ensure that the system functions correctly.
• Noise factors are uncontrollable variable that can
cause significant variability in the process or the
product
Loss Function
• Quality – ‘Loss imparted to the society from the time
a product is shipped’…
• Societal losses (Failure to meet customer requirement,
ideal performance, harmful side effects … some
practitioners include losses due to production such as
raw material, energy, and, labor consumed on
unusable products or toxic by-products)
• Example: Production of Large Vinyl Covers to protect
materials from elements
• Meeting target vs. Meeting Specification
Nominal-the-Best
• Although Taguchi developed more than 68 loss
functions, many situations are approximated by
the quadratic function which is called the nominal-
the-best type
• Given by equation, L=k(y- ʈ)2
Where L= cost incurred as quality deviates from target
y = Performance Characteristic
ʈ = Target
k = Quality Loss Coefficient
An Example:

If the specifications are 10 +/- 3 for a particular


quality characteristic and the average repair cost is
$230, determine the loss function. Determine the
loss at y = 12.
Average Loss

• Loss described earlier assumes the quality characteristic is


static…in reality one cannot hit the target…due to noise..
• Noise may be classified as external and internal, with
internal being further classified as unit-to-unit and
deterioration
• Refrigerator : External, due to action of the user say, no. of
times door is opened, amount of food, temp. of food…
Unit-to-unit is variation because of production viz. seal
tightness, control sensor variation etc., deterioration say
leakage of refrigerant, mechanical wear of compressor etc.
Other Loss Functions

Two other common loss functions are


• Smaller the better: Target is 0, no negative values,
ex. Radiation leakage from Microwave appliance,
response time from computer, pollution from
automobile, etc.
• Larger the better: Target is ∞, no negative values,
worst case at y=0 …Reciprocal of earlier…ex.
Bond strength of adhesives, welding strength,
automobile mileage, etc.
• These three loss functions cover most of the
situations
• After selecting a loss function, one point on the
curve needs to be determined in order to obtain the
coefficient
• Easier to use the loss function that to develop an
elaborate quality cost system.
Orthogonal Arrays
• Simplified method of putting together an experiment
• Concept originally developed by Sir R. A. Fischer of
England in 1930s, Taguchi added 3 in 1956, and,
National Institute of Science and Technology (NIST),
USA, added 3
• 8 in OA8 represents no. of rows, also the no. of
Treatment Conditions (TC) and the degrees of
freedom…Top of the factor represents maximum no. of
factors, in this case 7…Levels are designated by 1 and
2…if more levels are there 3, 4, 5 etc. can be used…
other schemes such as -, 0, + can also be used
To determine the appropriate orthogonal array, use
the following procedure:
1. Define the number of factors and their levels
2. Determine the degrees of freedom
3. Select an orthogonal array
4. Consider any interaction
Degrees of Freedom
• The number of degrees of freedom is an important
value because it determines the minimum number of
treatment conditions..
• It is equal to the sum of
(Number of levels – 1) for each factor +
(Number of levels - 1)(Number of levels -1) for each
interaction + One for the Average
Degrees of Freedom - Example
• Given four two-level factors, A, B, C, D, and, two
suspected interactions, BC and CD, determine the
degrees of freedom, df. What is the answer if the
factors are three-level?
Selecting the Orthogonal Array
• Maximum no. of df = lf
• The number of treatment conditions is equal to
the number of rows in the OA and must be greater
than or equal to the dof.
• 3rd Column-Maximum no. of factors
• Last four columns- Maximum no. of columns
available at each level
Interaction Table
• Confounding is the inability to distinguish among the
effects of one factor from another factor and/or interaction.
To prevents this , one must know which columns to use for
factors.
• A-C1, B-C2, AB-C3, C-C4, AC-C5 i.e. columns are
reserved for interactions to find strong interaction…if there
are no interactions, then all columns can be used for factors
• The actual experiment is conducted using columns
designated for the factors…these columns are referred to as
‘design matrix’. All columns are referred as ‘design space’.
Linear Graphs
• Taguchi developed a simpler method of working with
interactions using linear graphs…make it easier to assign
factors and interactions to various columns of an array.
• Three level interactions use two columns, one column for
linear interaction and one for quadratic.
• The linear graphs and interaction tables are not designed
for three or more factor interactions, which rarely occur
• Linear Graphs can be modified.
• Trial-and-error approach & a number of solutions may be
possible.
Signal-to-Noise Ration (S/N)
• S/N = Amount of energy for intended function/ Amount of
energy wasted = Signal/ Noise
• Brake …. Slowdown Car (Signal)….Squeal,Wear, Heat,
Etc. (Noise)
• Signal factors are set by designer or operator and is
denoted by y (bar)
• Noise factors (s2) are not controlled or are very expensive
and difficult to control
• S/ NN = log10 (y(bar)2/s2)
Thank You!!!

You might also like