Nanomaterial characterization through image treatment, 3D reconstruction and AI techniques.

摘要

Nanotechnology is not only the science of the future, but it is indeed the science of today. It is used in all sectors, from health to energy, including information technologies and transport.;For the present investigation, we have taken carbon black as a use case. This nanomaterial is mixed with a wide variety of materials to improve their properties, like abrasion resistance, tire and plastic wear or tinting strength in pigments.;Nowadays, indirect methods of analysis, like oil absorption or nitrogen adsorption are the most common techniques of the nanomaterial industry. These procedures measure the change in the physical state while adding oil and nitrogen. In this way, the superficial area is estimated and related with the properties of the material.;Nevertheless, we have chosen to improve the existent direct methods, which consist in analysing microscopy images of nanomaterials. We have made progress in the image processing treatments and in the extracted features. In fact, some of them have overcome the existing features in the literature.;In addition, we have applied, for the first time in the literature, machine learning to aggregate categorization. In this way, we identify automatically their morphology, which will determine the final properties of the material that is mixed with.;Finally, we have presented an aggregate reconstruction genetic algorithm that, with only two orthogonal images, provides more information than a tomography, which needs a lot of images.;To summarize, we have improved the state of the art in direct analysing techniques, allowing in the near future the replacement of the current indirect techniques.