Algorithms on local weld geometries
Rafat, Chowdhury Fazleh Rabbi (2026)
Kandidaatintyö
2026
School of Energy Systems, Konetekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2026060463892
https://urn.fi/URN:NBN:fi-fe2026060463892
Tiivistelmä
The weld toe radius and weld toe angle have a strong influence on the fatigue life of welded joints. No single standardized method for measuring them from scan data exists. Different algorithms produce different results, which makes findings difficult to compare. This thesis compares three such algorithms, the Curvature Method (CM), the Least Squares Method (LSM), and the Iterative Method (IM) taken from the open-source WeldAnalysis toolkit.
The three algorithms were implemented in MATLAB and run through a single controlled preprocessing and measurement pipeline. They were applied to a set of 2D weld profiles. The algorithms were compared on the agreement between each other and not against a reference geometry with a known radius.
CM and LSM agreed closely on profiles with a moderate toe radius. IM returned a radius several times larger on sharp toes, where its fixed tolerance cannot separate candidate circles. LSM returned a higher radius on wide toes, where its curvature threshold extends the fitted region into the flank. The reported angle was consistent between the methods except where CM used a different reference point. The choice of algorithm therefore had little effect on profiles with a moderate toe radius but a large effect on sharp and wide toe.
The three algorithms were implemented in MATLAB and run through a single controlled preprocessing and measurement pipeline. They were applied to a set of 2D weld profiles. The algorithms were compared on the agreement between each other and not against a reference geometry with a known radius.
CM and LSM agreed closely on profiles with a moderate toe radius. IM returned a radius several times larger on sharp toes, where its fixed tolerance cannot separate candidate circles. LSM returned a higher radius on wide toes, where its curvature threshold extends the fitted region into the flank. The reported angle was consistent between the methods except where CM used a different reference point. The choice of algorithm therefore had little effect on profiles with a moderate toe radius but a large effect on sharp and wide toe.