Feasibility of Predictive Models for the Quality of Additive Manufactured Components Based on Artificial Neural Networks

Abstract

Additive manufacturing technologies present a series of advantages such as high flexibility, direct CAD to final product fabrication, and compact production techniques which make them an attractive option for fields ranging from medicine and aeronautics to rapid prototyping and Industry 4.0 concepts. However, additive manufacturing also presents a series of disadvantages, the most notable being low dimensional accuracy, low surface quality, and orthotropic mechanical behaviour. These characteristics are influenced by material properties and the process parameters used during manufacturing. Therefore, a predictive model for the characteristics of additive manufactured components is conceivable. This paper proposes a study on the feasibility of implementing Deep Neural Networks for predicting the dimensional accuracy and the mechanical characteristics of components obtained through the Fused Deposition Modelling method using empirical data acquired by high precision metrology. The study is performed on parts manufactured using PETG and PLA materials with known process parameters. Different Deep Neural Network architectures are trained using datasets acquired by high precision metrology, and their performance is tested by comparing the mean absolute error of predictions on training and validation data. Results show good model generalisation and convergence at high accuracy, indicating that a predictive model is feasible.

Artificial Neural Network-Based Predictive Model for Finite Element Analysis of Additive-Manufactured Components

“…The predictive model has at its core an artificial neural network which takes process parameters (speed, temperature, orientation) as inputs and empirical data in the form of tensile strength from tensile testing as outputs for the training process. The neural network architecture choice was informed by previous studies [2], where a number of neural network architectures were developed and tested for performance and compared to statistical techniques and models from other studies [1]. The presented neural network was tested against the best model from the study [2] as a reference point, and the parameters and architecture were adjusted to increase prediction accuracy.…”

“…The neural network architecture choice was informed by previous studies [2], where a number of neural network architectures were developed and tested for performance and compared to statistical techniques and models from other studies [1]. The presented neural network was tested against the best model from the study [2] as a reference point, and the parameters and architecture were adjusted to increase prediction accuracy.…”

“…Further studies by S.D. Grozav et al [2] tested, with good results, the feasibility of employing neural-network-based models for predicting the dimensional accuracy as well as the mechanical properties of components obtained by FDM additive manufacturing.…”

“…The fabricated components were subjected to tensile strength measurements, providing empirical data on the mechanical characteristics of different process parameter combinations. PLA was chosen because in a previous study [2] it was shown to have the highest variability of mechanical properties for different process parameter combinations and because it is one of the most popular 3D-printing materials. Another reason for choosing PLA is the availability of more data from previous studies on components manufactured using this material.…”

Editorial for Special Issue “Advances in Computer-Aided Technology”

“…The study by the authors Ryumin et al describes the development in the field of design of ship structures, focusing on the detailed description of an integrated system capable of performing both the design and modeling of the structure of merchant ships [10]. The paper by the authors Grozav et al proposes a study on the feasibility of implementing Deep Neural Networks for predicting the dimensional accuracy and the mechanical characteristics of components obtained through the FDM method using empirical data acquired by high-precision metrology [11]. The paper by the authors Solfronk et al presents an investigation of the computational strategy used in FEA on the springback prediction of a thin sandwich material made of micro-alloyed steel.…”