Deep learning as a powerful tool in digital photoelasticity: Developments, challenges, and implementation

Abstract

Stress field evaluation, through fringe order maps, has always been of great importance in various engineering domains, providing essential insights into the mechanical response of loaded structures and enabling the prevention of potential failures, thus ensuring the safety and reliability of critical components. In recent years, the advent of new algorithms has ushered in a transformative era, particularly through computational techniques such as machine learning and, in particular, deep learning. These advanced approaches have proven instrumental in optimizing fringe demodulation processes, providing more accurate predictions, and uncovering complex patterns within structural behavior. This paper reports a careful review of papers and articles discussing the application of deep learning to photoelasticity studies. It explores the impact and applicability of deep learning in the field of digital photoelasticity, highlighting its benefits and advancements. In addition, the challenges associated with the implementation of this technique are examined to provide a critical insight. The paper also discusses experimental results and possible suggestions for future work in this new paradigm.