Data pre-processing/model initialisation in neurofuzzy modelling of structure-property relationships in Al-Zn-Mg-Cu alloys

The paper deals with the application of multiple linear regression and neur ofuzzy modelling approaches to 7xxx series based aluminium alloys. 36 compo sitional and ageing time variants and subsequent proof strength and electri cal conductivity measurements have been studied. The input datasets have be en transformed in two ways: to reveal more explicit microstructural informa tion and to reflect some empirical findings in the literature. Neurofuzzy m odelling exhibited improved performance in modelling proof strength and ele ctrical conductivity cf. the multiple linear regression approach. Electrica l conductivity is best modelled using the explicit microstructural input da taset, whilst proof strength is best modelled by a further modification of this dataset, decided upon after inspection of the subnetwork structures pr oduced by neurofuzzy modelling. Neurofuzzy modelling offers a transparent e mpirically based data-driven approach that can be combined with pre-process ing of the data and initialising of the model structure based upon physical understanding. An iterative modelling approach is defined whereby data-dri ven empirical modelling approaches are first used to assess underlying data structures and are validated against physically based understanding, these then inform subsequent initialised neurofuzzy models and input data transf ormations to provide both optimal subset and feature representation.