Design of Experiments - Multiple complex response predictive models..

Design of Experiments (DoE):

We make extensive use of Design of Experiment methods to investigate and optimise clients products. Using Taguchi matrices we have developed a technique and a suite of software to:

Automatically create experiment recipes (CAE or physical tests)
Create recipe permutations of CAE models.
Automatically mine analysis data and calculate main and interaction effects.

Through extensive automation, complex studies can be tackled with ease.

A lot of our work has been focused on responses in the frequency domain such as Noise Transfer Functions (NTF), Vibration Transfer Functions (VTF) or dynamic stiffness. Most 'off the shelf' software struggles with dealing with the shear volume of data that our software can process - this is what led us to develop our own tools. We are unaware of another toolset that allows for frequency or time domain predictive models. Purely understanding singular or a very limited number of responses drastically reduces the effectiveness and power of the Design of Experiments methodology - we avoid this limitation.

Our unique approach takes the automatically mined data and embeds it within a standard MS Office worksheet. Within this worksheet, a predictive model for any number of complex responses can be explored to find a design space solution. Using standard, readily available software such as Excel, the predictive model can be used with complete portability.

Using only a very a small number of analyses, or physical tests, our approach allows the user to explore a massive range of design permutations whilst understanding the likely effect on numerous responses. This predictive tool can be 'driven' live and used as either a concept design tool for resolving complex, multi-faceted issues with existing designs. The benefit of this approach over classical optimisation methods is that the new predictions can be made without the expense of resolving a CAE model or performing another physical test.

We have successfully applied our methods and tools to finding a solution to improve vehicle road noise, idle & driveline induced vibration, physical test and virtual test NTF issues and also for exploring concept designs for subframes, chassis', body structures through to components such as powertrain cross members.

We have also conducted studies where we have embedded cost models in to the attribute predictive model to allow for optimisation of a design for minimum cost. Being able to consider multi-attributes such as NVH, safety, durability, cost, mass etc. provides a unique opportunity to truly understand what makes a product tick and consequently design the best product for least cost.

"These methods and toolsets can be applied to a broad spectrum of engineering issues where an objective measure can be made - their power and cost effectiveness make them ideal for exploring product and process solutions alike. Please feel free to contact us concerning your own project needs."