Automated steel bridge coating inspection using neural network trained image processing

Ahmed Elbeheri, Tarek Zayed

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

Abstract

Steel bridges deterioration has been one of the problem in North America for the last years. Steel bridges deterioration mainly attributed to the difficult weather conditions however some solutions have been presented in the last years like weathered steel but there are still a large number of existing bridges that were built in other steel types other than weathered steel. Steel bridges suffer fatigue cracks and Corrosion, which necessitate immediate & regular inspection. Visual inspection is the common and traditional technique for steel bridges inspection but it depends on the inspector experience, conditions and work environment conditions. So many NDE models that have been developed use Non-destructive technologies to overcome these obstacles and be more reliable. Non-destructive techniques such as The Eddy Current Method, The Radiographic Method (RT), Ultra-Sonic Method (UT), Infra-red thermography and Laser technology have been used. Previous researchers had used Digital Image processing for rust detection as an Alternative for visual inspection. Different models had used grey-level and coloured digital image for processing. Coloured image Processing depend on the fact of using the colour of the rust, one of the rust unique characteristics to distinguish it from the different backgrounds. The detection of the rust is an important preliminary stage as it is the first warning for the corrosion and a sign of coating erosion. To decide which is the steel element to be repainted and how urgent it is the percentage of rust should be calculated and it can be always as indication of the corrosion level. In this paper, an image processing approach will be developed to detect corrosion and its severity. Two models were developed 1st to detect rust and 2nd to detect rust percentage. Machine learning was applied through neural network as it will be used to be trained to detect rust and calculate the percentage. The application of the neural network will ease the process of processing the inspection data. One of the great advantages of neural network that it can adapt to any new data it even helps as more data mean more training and better results.

Original languageEnglish
Title of host publicationMaintenance, Safety, Risk, Management and Life-Cycle Performance of Bridges - Proceedings of the 9th International Conference on Bridge Maintenance, Safety and Management, IABMAS 2018
EditorsNigel Powers, Dan M. Frangopol, Riadh Al-Mahaidi, Colin Caprani
PublisherCRC Press Balkema
Pages2545-2551
Number of pages7
ISBN (Print)9781138730458
Publication statusPublished - 2018
Event9th International Conference on Bridge Maintenance, Safety and Management, IABMAS 2018 - Melbourne, Australia
Duration: 9 Jul 201813 Jul 2018

Publication series

NameMaintenance, Safety, Risk, Management and Life-Cycle Performance of Bridges - Proceedings of the 9th International Conference on Bridge Maintenance, Safety and Management, IABMAS 2018

Conference

Conference9th International Conference on Bridge Maintenance, Safety and Management, IABMAS 2018
CountryAustralia
CityMelbourne
Period9/07/1813/07/18

Keywords

  • Bridge inspection
  • Image processing
  • Neural network
  • Steel bridge
  • Steel corrosion

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Safety, Risk, Reliability and Quality

Cite this