Neural networks to provide guidance system for large-scale industrial laser cleaning system
Neural networks to provide guidance system for large-scale industrial laser cleaning system
Neural Computer Sciences has been selected to provide the laser control and guidance correction system for the European RESTOR project. The project aims to bring laser cleaning technology on to the city streets for restoring building façades to their original glory. Laser beam quality is affected in complex ways by local environmental conditions and other factors. Neural Computer Sciences (NCS) will develop embedded neural networks that combine various sensor parameters into a corrective input for the laser optics.
At a research institute,the French national laboratory for the conservation of monuments (Laboratoire de Recherche sur les Monuments Historiques) and French laser specialists B M Industries have perfected laser technology for cleaning statues. The process is easy to install and provides high quality surface cleaning with absolutely no damage to the underlying substrate. However, the current technology is slow and only suitable for fine sculptures; RESTOR aims to develop the process into a system suitable for cleaning some 10 square metres of building façade per hour.
Utilising pulsed lasers operating with 10 nanosecond bursts at a 20kHz repetition rate, surface dirt is removed leaving the underlying building undamaged. However, parameters such as temperature influence the power and quality of the laser beam and, most importantly, its pulse beam profile. NCS's neural network technology provides the ideal method for converting these parameters into suitable automatic beam correction factors, a process which does not lend itself to conventional computer control techniques.
Embedded neural networks are helping to bring laser cleaning technology on to the city streets for restoring building façades to their original glory, such as the partially cleaned church window in Lille, France
Intelligent software technologies such as neural computing are being used to solve real-world problems that are too complex, laborious or not sufficiently well understood to be addressed by conventional processing methods. Neural networks learn the relationships present in data through a training process and, once trained, will then respond accurately to new input data. During the experimental phase of RESTOR, the laser company will help NCS define the performance-related parameters and collect training data for use in setting up the controlling neural network.
NCS will lead the technical development of the neural net software, its associated embedded control board and supervisory system software. The resulting open loop correction system will be built into the laser to maintain its performance under the conditions found during building cleaning operations.
Clean urban areas are safer and more pleasant places, and hence cleaning the façades of public buildings will have a positive impact on the environment. Buildings are subject to discoloration by atmospheric and fossil fuel pollutants, and black encrustation from acid rain. Current cleaning technology relies on sand blasting which erodes detail from the structure or water jets, which can damage the building and require the waste water to be purified and recycled. Alternative chemical cleaners can discolour the building material rendering it permanently marked. Laser cleaning has none of these drawbacks and offers significantly increased productivity.
RESTOR is funded under the European Union EUREKA initiative, reference EU 1644, and managed in the UK by the Department of Trade and Industry. The project is led by B M Industries; in addition to NCS and LRMH, project partners include the Fraunhofer-Institute for Material and Beam Technology, monument restoration company Quelin S.A., cooling specialists Thermal Engineering Systems Limited and F.O.R.T.H., the Greek Institute of Electronic Structures and Lasers.
For further details please contact Isabelle Bret, Neural Computer Sciences, Unit 3 Lulworth Business Centre, Nutwood Way, Totton SO4 3WW, UK. Tel: +44 1703 667775, Fax: +44 1703 663730; Email:sales@ncs.co.uk Web site: www.ncs.co.uk
