Localised and Learnt Applications of Machine Learning for Robotic Incremental Sheet Forming

Abstract

While fabrication is becoming a well-established field for architectural robotics, new possibilities for modelling and control situate feedback, modelling methods and adaptation as key concerns. In this paper we detail two methods for implementing adaptation, in the context of Robotic Incremental Sheet Forming (ISF) and exemplified in the fabrication of a bridge structure. The methods we describe compensate for springback and improve forming tolerance by using localised in-process distance sensing to adapt tool-paths, and by using pre-process supervised machine learning to predict stringback and generate corrected fabrication models.

Investigation of Thermal-Related Effects in Hot SPIF of Ti–6Al–4V Alloy

“…Though not applied on hot formed titanium parts, most promising contributions in this sense refer to tool path correction based on machine learning predictions. Work done by Khan et al 21 , Fiorentino et al 22 and Zwierzycki et al 23 confirm all significant improvements in the accuracy.…”

Accuracy and Surface Quality Improvements in the Manufacturing of Ti-6Al-4V Parts Using Hot Single Point Incremental Forming

“…However, these aspects are not enough to produce accurate parts using hot SPIF and further progress is still needed in this field. Though not applied on hot formed parts yet, the most promising contributions toward increase part accuracy in SPIF refer to tool path correction or optimization solutions based on machine learning predictions [8][9][10][11][12][13]. Behera et al [8,9] proposed a toolpath compensation strategy based on multivariate adaptive regression splines (MARS) for the prediction of the formed shape, which has been validated on different aluminum alloys.…”

“…The model was successfully applied on two pyramidal geometries made of DC04 stamping steel, which showed improved accuracy. More recently, Zwierzycki et al [13] studied two methods to compensate for springback, using localized in-process distance sensing to adapt tool-paths on the one hand, and using pre-process supervised machine learning to predict springback and generate corrected fabrication models on the other hand.…”

Application of machine learning algorithm in the sheet metal industry: an exploratory case study

“…Kashid and Kumar 2012) performed a review of the applications of artificial neural networks to sheet metal work. Zwierzycki, Nicholas, and Thomsen (2018) applied pre-process supervised machine learning to predict and improve sheet-forming tolerance and generate corrected fabrication models. Lin and Chang 1995) proposed a model using machine learning from neural networks in an expert system of sheet metal bending tooling.…”