| Περίληψη |
Laser-assisted selective removal of altered and unwanted crusts and coatings
from heritage surfaces is a particularly delicate procedure which urges for refined
and reliable monitoring protocols. This gets particularly important in the case of
encrustations that show similar physicochemical properties to the underlying
authentic surface. Thus, self-limiting laser ablation cannot be guaranteed, as for
example the removal of aged varnish films from painted surfaces [1]. Most of
the time conservators make the determination about the stopping point of the
cleaning based on personal judgment and visual experience. However, this
approach can compromise the cleaning outcome, as micro-damage may occur.
Among the different analytical methodologies to follow in real-time the ablation
process, the monitoring of acoustic signals produced upon laser-assisted material
removal, has been found to be remarkably straightforward and promising [1, 2,
3].
This thesis reflects feasibility studies to follow online laser cleaning through the
recording of the intrinsically generated acoustic waves during the process and
the use of artificial intelligence (AI) algorithms to predict the probability of the
next laser pulse being the cleaning pulse, justifying the decisions taken on the
continuation or the suspension of the ablation process. The intrinsically
generated photoacoustic (PA) signals combined with photoacoustic waves
generated provide the opportunity to follow the cleaning process accurately and
in real time. Laser cleaning was undertaken using infrared (1064 nm) ns pulses
emitted from a QS Nd:YAG laser on model plates of marble covered with black
graffiti films of varying thickness. Irradiation tests with various parameters
related to over-, under- and optimum cleaning outcomes are studied on the basis
of acoustic monitoring in order to determine the critical AI-indicated thresholds.
|
AI-informed acoustic monitoring of laser cleaning interventions
