unisa ITA  unisa ENG


VIRPROFOOD

The White Page On Virtualization Of Processes In Food Engineering

by Francesco Marra, Ph.D., Department of Industrial Engineering, University of Salerno, Italy

fmarra@unisa.it


Virtualization and its aims in the process industry
Virtualization is a term very popular in IT and computer science especially when referred to "virtual computer" as a logical representation of a computer in software. By decoupling the physical hardware from the operating system, virtualization provides more operational flexibility and increases the utilization rate of the underlying physical hardware (IBM, 2007). So, basically, thanks to virtualization it is possible to have a machine (the virtual one) that is totally independent by the hardware. In a similar way, under "virtualization in food engineering" we should include a list of activities (that could be purely intellectual and/or mathematical and/or computational) such as "data modeling", "statistical analysis", "fundamental modeling", "dynamic simulation", "optimization" when they are able to provide the researcher with a virtual machine/plant/process/product that has all the characteristics of the "real case" it was inspired to, without needing any physical hardware (here hardware has a different meaning of course), so that it provides more process flexibility ("I can explore more phenomena, deeply, in the same time") and it increases the utilization rate of the underlying physical hardware (since with virtualization the final goal will be optimization and thus the increase of utilization of a given plant or process). Thus modelling, simulation, optimization and dynamic studies are a part of a wider scheme that is described as virtualization. While a number of manufacturing sectors (e.g., aerospace, defense, automotive) are benefiting for decades from modeling activities and virtualization of processes, the food industry (that represents more than 5% of the global GDP) is still lagging to utilize the wide spectrum potential offered by virtualization as an engineering design tool (Marra, 2014). Some years ago (2004), Singh and Erdogdu published a book on "Virtual Experiments in Food Processing", claiming that virtualization allows to conduct experiments using a computer or a mobile device. Undoubtedly, the use of virtual tools reduces the overall time needed for developing, designing and validating processes and equipment, circumvents the tailored made production of un-efficient numerous prototypes, reduces development costs and consequently reduces the time to market. Thus, virtualization addresses and fulfills the needs of new and sophisticated strategic tools for innovation in the food industry reducing the time needed for developing, designing and validating processes and equipment.

Modeling approach in the virtualization framework
Virtualization is based on virtual representation of reality. Virtual representation of reality needs a mathematical formulation, so that virtual framework can be seen as a mathematical analog of the physical reality, describing the properties and characteristics of a real system in terms of variables and mathematical operations. As well explained by Datta (2008), mathematical models can be based exclusively on observation of a certain phenomenon (empirical based models), or they can be based on physical laws that should describe the presumed physical phenomenon which is under investigation (physics-based models). Actually, purely physics-based models do not exist because any model - to be solved - requires the knowledge of physical properties which, in turn, need experimental observations to be determined (see below). At same time, empirical based models using observations uniquely to find equations fitting the experimental data (without any theoretical explanation) should not be included in any category of models. Unfortunately, traditionally the technical people in the food industry are often more used to observation-based models; to this adding also that the food industry hired for years food scientists who were in charge for what were the issues of primary interest in food processing, microbiological safety and quality, and who were trained very well in food chemistry and microbiology more than in physics based models (Datta, 2008). Beside others, this was also the cause of a scarce use of models in the food processing sector, while aerospace, defense and automotive used models as part of manufacturing and distribution. Physics based models can be differentiate according with the scale they look at: molecular dynamics models look at microscale; mesoscale is the domain of interest of models making use of Lattice-Boltzmann method; at so called macroscale continuum we find models based on momentum (fluid flow), heat and mass transfer, structural mechanics and kinetics. In the class of empirical based models, mathematical methods such as neural network, fractal analysis and fuzzy logic find their place.
The general steps allowing to build-up a model may be summarized in the following:
1) Problem definition (what is the phenomenon under investigation, what is the system in which the phenomenon occurs, what is the surrounding environment);
2) Application of the theory (the physics based laws) governing the phenomenon under investigation, defining the system (with its initial status and the boundary conditions) and the surrounding environment ;
3) Mathematical formulation of point 2);
4) Formulation of hypothesis to eventually reduce the mathematical complexity of the model;
5) Computation of the solution of the mathematical formulation, by means by analytical techniques (rare), or building-up an algorithm written on purpose, or by usage of general purpose computational software;
6) Analysis of results;
7) Post-processing of results;
8) Verification of model's results by means of experimental validation;
9) Eventually, correction of problem definition (back to step 1), or application of the theory (back to step 2), or mathematical formulation (back to step 3) or model formulation hypothesis (back to step 4).
This way to proceed generates one or more loops, sometime interconnected, which bring to the multi-loop endless pursuit.