3D convolutional neural networks to estimate assembly process parameters using 3D point-clouds

Abstract

Please refer to published version for the most recent bibliographic citation information. If a published version is known of, the repository item page linked to above, will contain details on accessing it.

Towards Real-Time Physics-Based Variation Simulation of Assembly Systems with Compliant Sheet-Metal Parts Based on Reduced-Order Models

“…VS-based synthesis covers topics related to process optimisation, parametric and sensitivity analyses [1][2][3][4][5][6][7]. The common denominator is the desire to generate accurate results in a reasonable time (ideally in real-time), which is sometimes not achievable due to the complexity of the problem even with powerful computational systems (High Performance Computing -HPC, cloud computing, etc.).…”

“…Aderiani et al [5] applied a new method to optimise the fixture layout to two simple single-station cases, elapsing 110 hours and 160 hours, respectively. Sinha et al [6] proposed a Deep Learning-based methodology to aid multiple root causes analysis for an assembly process. To train their Deep Neural Network (DNN), they conducted 9 runs of 10,000 FEM simulations by varying the positions of only 5 clamps.…”

Keyhole mapping to enable closed-loop weld penetration depth control for remote laser welding of aluminum components using optical coherence tomography

“…However, at the present moment, only data-driven approaches are insufficient to handle all possible variations in process and welding setup, such as part-topart gap or seam misalignments. [11] [12] In-process quality control formed the central focus of a study by You et al [13] in which authors used five different types of sensors for laser welding process monitoring and x-ray imaging was used for keyhole depth monitoring. Bautze et al [7] compared multiple approaches of laser welding monitoring and control and concluded that due to intense process emissions and extreme temperature gradients, as well as highly unstable process states in the case of laser welding of aluminium, the only monitoring technique that can provide direct and in-process measurement of the keyhole depth is optical coherence tomography (OCT).…”

Object Shape Error Response using Bayesian 3D Convolutional Neural Networks for Assembly Systems with Compliant Parts

“…Past methods used to diagnose manufacturing dimensional quality faults are based on [2]: (i) statistical estimation; and, (ii) pattern matching based approaches. These approaches have been shown to have limitations in their applicability to complex, high dimensional and nonlinear systems [3] as these used linear models between process parameters and measurements of product dimensional quality [4][5] [6][7]. This significantly limits the application of the methods for 3D object shape error modelling and diagnosis in manufacturing.…”