Design of experiments, particularly with the Taguchi method, are a powerful approach to engineering and process optimization. Genichi Taguchia renowned Japanese statistician, developed this method with the aim of designing robust experiments, i.e. experiments that are less sensitive to undesirable variations in factors. The ultimate aim is to optimize the performance of a product or process, while minimizing the impact of sources of variation.
The Key Steps of the Taguchi Method :
Identification of Factors and Levels :
The first crucial step is to identify the factors that may influence the outcome of the experiment. These factors may be design variables, process settings or other relevant parameters. It is essential to select the most significant factors for the study objective.
Level definition :
Each factor must have different levels, representing the specific values the factor can take on. The variability of the levels makes it possible to explore the influence of each factor on the system's response.
Taguchi Array selection :
Using the "Taguchi Array" or other pre-established experimental designs, specific combinations of levels are selected for each factor. The Taguchi approach is designed to maximize experimental efficiency with a minimum number of trials.
Test replication :
To assess the variability of results, test replications are carried out. This makes it possible to distinguish the real effects of factors from random variations.
Results analysis :
Statistical techniques are applied to evaluate the influence of each factor on system response. The aim is to identify the most influential factors, and model the response as a function of these factors, while minimizing sensitivity to undesirable variations.
Optimization and Validation :
Once the results have been analyzed, a parameter setting is selected to optimize product or process performance. These results can be validated by further testing if necessary.
The array below summarizes the Taguchi experimental designs proposed by Ellistat in the module Data Analysis according to the number of factors studied :
Let's take the following case: in the bonding process, we want to study the impact of two factors, temperature and pressure, on bonding strength.
- The temperature varies between 20°C and 30°C
- Pressure ranges from 5bar to 10bar
The proposed experimental design is an L4 array, i.e. there will be 4 tests to be carried out, varying the two factors (temperature and pressure) simultaneously.
You'll notice that the two factors can only have minimum or maximum values, and that the 4 proposed tests present the 4 corners of the response surface in test space 4.
