Polyarticulated measuring arms (AACMM) are three-dimensional measuring devices consisting of articulated arms comprising 6 or 7 joints. Their use is increasingly common in the metrological control of mechanical parts. They are portable and therefore operate outside metrology laboratories in unconditioned environments (mechanical workshops, workstations, etc...). This mobility exposes them to sources of measurement errors (temperature, vibration, operator effort, etc…) in addition to other sources such as angular encoders, sets and deformations in links, etc…). As a result, measurements made by these arms are less accurate than threedimensional measuring machines (CMM) working in well-conditioned laboratories. In this article, we propose a method to optimize the neural network model of a polyarticulated arm by acting on the network input data.
Volume 12 | 04-Special Issue
Pages: 1137-1146
DOI: 10.5373/JARDCS/V12SP4/20201587