Predictive modeling of mechanical properties for Fused Deposition Modeling parts: A focus on processing and environmental parameters

Fused Deposition Modelling (FDM) is a widely used additive manufacturing technique for prototyping and manufacturing functional parts due to its cost-effectiveness and low waste generation. The mechanical properties of FDM parts depend on multiple processing parameters and environmental conditions. Due to the complex physical phenomena of FDM, it is a challenge to accurately determine part properties. These properties are generally determined after performing experimental investigation which is time consuming and expensive. Alternatively, predictive models developed for determining part properties have proved to be a valuable tool. However, these models are typically developed using data from a single printer and do not consider environmental parameters as an input factor. To address these shortcomings, a data-driven multi-step scalable modeling technique that provide the ability to gather data from multiple sources is presented. This technique is used to develop models for predicting mechanical properties using experimental data from multiple sources resulting in models that are robust compared to models developed using data from one printer. Additionally, one environmental parameter has been introduced as the input factor. Regression and Artificial Neural Network models are developed to predict mechanical properties using experimental data from two printers. Five processing and environmental parameters namely, extrusion temperature, layer thickness, print bed temperature, print speed and humidity are used as the input factors. These models predict the toughness and tensile strength of the parts. The experimental values from the two printers are compared to the predicted values to demonstrate the superior predictive capabilities of the models. Finally, to demonstrate the scalability and utility of the proposed technique, data from one more printer is introduced and is used to develop regression models. (c) 2023 The Authors. Published by ELSEVIER Ltd. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)