General project outline and background
The structure of granules or agglomerates can be defined as the spatial arrangement of its basic components [1]. The basic components are primary particles, binder or liquid components and intra particle porosity. The quantification of granule structure is crucial for setting up processing maps and input data for simulations such as DEM and computational modelling [2]. Spatial statistics, also called stereology, provide well-defined measures to quantify the different aspects of powder structure. For infinite systems the covariance function (CVF) defines the distance relations of the particles and pores. In agglomerates a radial distribution function (RDF) quantifies shell or boundary structures. These statistical measures are well suited as a basis for stochastical property modelling. Several image analysis procedures provide the above mentioned data, e.g. morphological sieves, point sampling techniques or chord length measurements. The characteristic curves can also be correlated with stochastical processes e.g. Poisson processes. While structure measures have been determined in a number of applications, the fundamental combination of physical properties and structure is still immature. The parameter of the above mentioned stochastic models need to be linked to the key parameter of the structure generation process e.g. primary particle size, binder content, stress and growth history. Results of other simulation techniques like VOF (Volume of fluid) or Population Balance models can serve as input to generate the correlation between statistical parameter and process. The added value of a structure model would be its statistical reliability and ease of use for parameter variations. Some key powder properties are well understood physically [3-6] but lack systematic incorporation of structure measures above phase volume and particle size. Mathematical techniques to calculate structure dependent properties include cellular automata and convolution algorithms. The additional challenge is to convert local behavioural probabilities into bulk properties.