Taguchi methods

The Taguchi method, named after its inventor Taguchi Gen'ichi (Anglicised: Genichi Taguchi ), is a method of testing, which focuses on minimizing the dispersion around the setpoint. The Taguchi method attempts to achieve this objective by products, processes and systems are designed as robust as possible. This means that they should be as insensitive to interference ( engl. noise), where they will face in practice.

Applies the Taguchi method in the quality management and Six Sigma.

The loss function

Instead required tolerances should be considered as limits to be observed in the manufacture, evaluates the Taguchi method, any deviation from the nominal value ( within the tolerance limits) as an error, causing a concrete financial loss. This financial loss is modeled as part of the Taguchi method based on the so-called loss function. Mathematically, the loss function is a parabola Represents the words, the model assumes that a twice as large deviation from the nominal value would cause four times as high financial loss. The financial loss is then minimized when the value achieved exactly equal to the target value.

Taguchi's loss concept is not limited to the financial loss incurred by the manufacturer if he produces a bad product. Rather, the loss of function models the loss caused to the company when a consumer uses a product whose characteristics do not meet their target values ​​. The idea to minimize the loss to society instead of the loss for the company, represents a break with the traditional way of thinking

= Quality loss, = actual measured value = setpoint = Estimated money value that is related to the deviation from the nominal value in connection

Signal-to- interference ratio and Robustheitsmaß

As a measure of dispersion around the setpoint used the so-called signal-to- interference ratio, the signal-to- noise ratio ( engl. signal to noise ratio) is, following the proposal Taguchi, the Telecommunications ( Taguchi once worked in this area ) is modeled. But you also writes short S / N ratio.

With = effect of the signal = effect of interference ( " noise"), = mean value of the target size and standard deviation =

Since the S / N ratio includes the mean of the target size, it is therefore closely linked to their average position. For parameters where the standard deviation increases in step with the average result of physical contexts, this makes sense. However, only interested in the standard deviation, independent of the mean value, it shall be sufficient instead of the S / N ratio of the Robustheitsmaß to assess the process or system. The robustness is then calculated according to the following formula:

This S / N formulas apply to features that have a specified target value. For characteristics that should take the largest possible or the lowest possible values ​​in the ideal case, uses Taguchi other S / N calculation formulas.

The Taguchi design philosophy

Taguchi divides the development process into three steps:

• System Design
• Parameter Design
• Tolerance Design

These three steps are referred to by him as an offline quality control (English offline quality control ). Each of these steps has its own function:

In system design, the designers decide what is to be built for one type of system, so for example the technology to be used, the components that should exist, the system etc.

In the parameter design is about to optimize all the parameters of the design ( control variables, factors ) so that the system is as insensitive as possible to interference. That is, it will be ideal target values ​​for the various parameters determined. For this purpose, statistical experimental design methods are used.

In the tolerance design, the tolerances for the system parameters are set. Here again, statistical experimental design methods are used. The aim is to define the tolerances according to the actual impact of the parameters on the function of the system. If a factor has been shown to have much influence on the function, a wide tolerance is determined. This saves production costs.

Taguchi experimental design

Experimental designs by Taguchi are essentially partial factorial designs ( engl. fractional factorial designs), that is, there are not all the possible combinations of factor levels through the game, but only a well- selected subset. To create the test plans called orthogonal arrays are used (English orthogonal arrays), which are tabulations in reference books.

Taguchi experimental design for parameter optimization often include an internal and an external field; inner panels are the control variables ( the engineer free customizable design parameters ), and in the external field, the disturbances ( environmental factors, etc., which are subject in practice unavoidable and can not be influenced fluctuations and thereby have an impact on the process result have ). The outer panel (English outer array) is substantially smaller than the inner ( intra- array Sheet ) is generally. If the external field, for example, includes four test runs, each of the test runs of the internal field must be driven four times, once for each provided in an external field combination of Störgrößenstufen.

The goal is to find the combinations of control factor levels at which the effects of the disturbances are minimized and simultaneously the desired setpoint is maintained.

Evaluation of the tests performed in three steps, reflecting the fact that When adjusting is much easier to reach the desired value in general as to minimize the scattering:

Reception

Taguchi's quality philosophy aiming at robustness, associated with the use of statistical experimental design methods as a means for achieving this goal has been found in the industry is very wide acclaim. The statistical methods used by him are, however, complicated in part as unnecessary, been judged inefficient, improperly or improvement. Targets for criticism include the formulas for the S / N ratio, the lack of statistical efficiency of the composite of an inner and an outer field test plans and resulting from the application of Part factor plans danger of confusions between main effects and interactions.