Training of artificial intelligence for automated data analysis of phased array ultrasound data
R Rhéaume and C Catta
Training an artificial intelligence for automated data analysis of phased array ultrasound data is a complex task. Firstly, the source of the phased array ultrasound data is varied instrument-wise and coming from many different types of inspection: carbon steel welds, corrosion mapping, plastic welds. Secondly, the operator collecting the data and the conditions in which the data is collected influences the quality of the data. Thirdly, the knowledge and habits of the analysts analysing the data to be fed to the AI for training has a strong influence on the way the AI will recognise patterns. All of these aspects of the training must be well understood in order to train an unbiased artificial intelligence. In this paper, the authors will present the methodology and results in training Ondia’s own artificial intelligence dedicated to the automated analysis of phased array ultrasound data.
