# What are the Taguchi Methods

A. An approach to design optimisation for quality: a system of cost-driven quality engineering that emphasises the effective application of engineering strategies rather than advanced statistical techniques.

B. A new philosophy, using advanced statistical tools must be employed to design and produce reliable, high quality space systems at low cost.

* These methods provide an efficient and systematic approach to optimise designs for performance, quality, and cost.

* The objective is to identify the settings of design parameters that optimise the performance characteristic and reduce the sensitivity of engineering designs to the sources of variation.

* Using the Taguchi method, information on variable trends and interactions can be identified efficiently leading to optimum solutions and considerable resource savings.

2.Application of Taguchi Methods

A. Using Taguchi Methods for Problem Solving will:

* Provide a strategy for dealing with multiple and interrelated problems

* Give you a process that will provide a better understanding of your products and processes

* Give you a more efficient way of designing experiments for industrial problem solving using Orthogonal Arrays, while focusing on cost as a key consideration

* Provide techniques for rational decision-making for prioritising problems, allowing you to better focus your engineering resources

* Provide a tool for optimising manufacturing processes

B. This application of Taguchi Methods focuses on how to cost-effectively conduct process control activities during the manufacturing stage. These methods help:

* Diagnose the health of a process

* Minimise production of defects

* Determine the optimal process checking intervals and control limits

* Establish optimal preventive maintenance systems

* Design the best process connection systems toward the goal of full automation in the future

* Rationalise inspection systems

* Achieve an equilibrium between being “quality conscious” and being “cost conscious”

3.Interactions, Optimisation, and Measurement Strategies

* Many people are under the misconception that interactions are not considered in Taguchi Methods; however, the opposite is true.

* In fact, Taguchi’s method considers interactions one of the most important issues in his approach.

* The Signal/noise is an index for the robustness of quality, and it shows the magnitude of the interaction between “control factors” and “noise factors.” Control and noise factors must be assigned to different groups for the study of robustness, where there are no distinctions between control and noise factors.

Two-step optimisation:

* Step1: reduce the variation

* Step2: adjust the mean target

* A key difference of Taguchi Methods is the emphasis on measuring the right things for data collection. Instead of measuring symptoms caused by variability of the function, such as defects or failure rate, we measure an energy-related response.

* Any system uses energy transformations to accomplish the intended function. Reducing variability of the energy transformations will minimise or eliminate the symptoms.

4.Fundamental Principles

A. Randomisation:

* Runs should be conducted in a random order.

* Random number tables

* Draw lots

* Dice

* If not randomly assigned, then changes with time may distort the analysis.

B. Replication:

* If excessive variability present, estimates may be far from the true average value.

* Benefits.

* Average values have less variability than individual measurements, so calculated averages will tend to be closer to the true factor effects.

* Without replication, a single erroneous or unusual sample value can distort the whole analysis.

* Data from replicated experiments can be used to estimate the amount of variability in the process.

* Data from replicated experiments can be used to determine which factors affect the mean level of the process and which factors affect the variability of the process.

C. Factor interaction:

* Previous example assumed that effects of factors were additive, but factors can interact – interaction effects

* Orthogonal designs protect against one factor causing an artificially high or low value for the estimated effect of another factor, but do not always identify interactions between factors

5.Difference between Taguchi’s methods and Western Statistical Methods

* Little real difference – Taguchi presents designs in engineering rather than statistical terms.

* Tagushi design been around for long time but couched in engineering rather than statistical language

* We have looked at 2n tables :

* 23 equates with Taguchi L8

* 24 equate with L16

* Differences in notation and ordering.

* Each design has eight runs and seven orthogonal columns.

* Each design has columns for their three effects, three two-factor interactions and one three-factor interaction.

* Interchange columns 1 & 3 and 4 & 6 get the Western 23 design.

* Replace – 1 by 2, interchange columns 3 & 4 and reverse order of runs – Taguchi L8.

6.Benefis of Tagushi’s experimental designs

* Developed by engineers rather than statisticians – removes language problems.

* Importance of noise variables which disrupt production must be considered in addition to the control variables introduced. Optimising a product involves getting quality characteristics on target and minimising variability away from target – link with SPC.

The Taguchi method is the ability to successfully determine the existence of interaction effects between pairs of variables and to quantify those effects. Detection of parameter interactions would not be possible with the traditional varying one parameter at a time approach to design optimisation.

Overall, results suggest that the Taguchi method is a powerful tool which offers simultaneous improvements in quality, performance, cost, and engineering productivity.

Using the Taguchi method, information on variable trends and interactions can be identified efficiently leading to optimum solutions and considerable time and resource savings.

* The three factors are balanced. But when your are adding them together, there are some variations.

* In fact, the temperature (B) affects much because the line where there is no B (line AC), is small.

* The last line (ABC) indicate strong interaction thus the good combination is the combination of Pressure, Temperature and Humidity.