Using group method of data handling (GMDH) neural network to predict the maximum stress on elastomeric layers in spherical elastomeric bearings

KAYA Y., MAKARACI M., Bayraklilar S., KUNCAN M.

JOURNAL OF THE FACULTY OF ENGINEERING AND ARCHITECTURE OF GAZI UNIVERSITY, vol.36, no.3, pp.1332-1345, 2021 (SCI-Expanded) identifier identifier identifier


While many studies on planar elastomeric bearings have attracted attention in the international literature, there are very few studies on spherical elastomeric bearings due to their multi-layered and design difficulties. While elastomeric bearings are rigid against loads perpendicular to the layers, they are flexible against the loads parallel to the layers. In this way, spherical elastomeric bearings provide rigid against the central force caused by the rotation of the helicopter propeller, and flexibility against the blade's flapping and rotational movement. In this study, the Group Method of Data Handling (GMDH) model is used to estimate the maximum stresses in elastomeric layers in a spherical elastomeric bearing under compression and angular displacement loading. theta (angular displacement loading), P (pressure loading), a (radius of axis), beta 0 (first joint angle), cos (beta 0) (first joint angle cosine value), beta 1 (second joint angle), beta 2 (third joint angle), phi t (cone angle), phi p (angle between the direction of the pressure loading and the plane perpendicular to the elastomer layer), cos (phi p), D (elastomer layer outer diameter), ne (elastomer layer number), d (elastomer layer hole diameter) and H (elastomer layer thickness) were used as input features for GMDH model. The results obtained with GMDH were also compared with different machine learning methods such as ANN, SVM, RF. According to the results obtained, GMDH model was found to be more successful than other models in estimating spherical elastomeric bearing.