Design of Experiments (DOE)

DOE is structured method and is not a hit-or-miss experimentation where input parameters are adjusted randomly hoping to achieve process improvement. Taguchi method uses the orthogonal array in order to express the relationship among the factors under investigation. For each level of any one factor, all levels of other factors occur and equal number of times. This way the experiments are balanced and permit the effect of one factor under study to be separated from the effects of other factors. The design of an orthogonal array does not require that all combinations of factors be tested. Thus, the experiment is cost effective.

Robust Design:

DOE indicates that varying the factors can affect the product and/or process adversely. Some of the factors may not be within our control, e.g. weather. The challenge before quality engineering is to see that the product performs consistently. This is called ‘Product Robustness’. The goal is to reduce the sensitivity to uncontrollable factors so that the external variation affects the product performance as minimally as possible. A toffee manufacturer noticed that their product solidified to a rock hard condition at temperatures below 400F and became almost soupy above 800F. Using DOE, and appropriately a selecting/designing the input process parameters led to a product that was satisfactory hard in this temperature range. If these kinds of actions towards a ‘Robust Design’ are not taken, there may be failures due to the external ‘noise’. For instance, the ‘O’ rings which became brittle and failed at low temperature on NASA’s Space Shuttle Challenger is an example of a non-robustly designed product.

DOE is one of the ways, albeit as very effective way, of optimizing a process and reducing the causes of variations and defects. An older method of cause identification or identification of important input parameters is the Cause and Effect Diagram or Ishikawa Diagram (after the inventor of the method). It is also called as Fishbone Diagram. It shows the possible causes for a quality deviation.

The diagram shows main factors, the sub factors causing the main factors, the further sub-sub factors causing the sub factors and so on. Thus, by such branching, one can get to the root of the quality problem that is, the deviation in a quality characteristic. It also provides a good understanding of the various cause effect dependencies which helps in controlling quality. It can provide useful input to the DOE though it fails to provide an exact diagnosis as in the Design of Experiments. While using Ishikawa diagrams, one has to address all the root causes with equal attention.

Whereas, the DOE will point to select one or two causes which attended to will produce a major improvement in quality. Ishikawa diagrams help in systematically listing the factors that could possibly cause the variations in the desired quality characteristics. In the fact, the factors (parameters) to be evaluated in a DOE can be obtained from Ishikawa diagrams.

Quality Function Deployment (QFD):

Cause effect diagrams and Taguchi methods of DOE are good for using the consistency of delivery of the quality as already decided upon by the organization. But, the question to be asked is: How do we decide as to what quality characteristics are to be incorporated into the product? How do we decide as to which component of the product /service should provide which part / aspect of the quality? That is, how do we plan to incorporate the quality characteristics into the product? Quality Functions Deployment (QFD) which is a customer driven planning process, answers these ‘What’ and ‘How’ questions. It does this by capturing the voice of the customer, breaking down quality into tangible, manageable, technical and operational actions which may be design and/or process actions so as to ensure that the customer’s wants, needs and expectations are met to the maximum extent possible. QFD is the deployment of the quality, functions / characteristics desired by the customer into the design of the product and the processes producing the product. While QFD is very useful for the design of new products, it can also be useful in improving the existing products.