Investigation on the mechanical properties of polyamide 12 printed by multi jet fusion

• Main Author: Koh, Zhi Hui
• Other Authors: Du Hejun
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Multi jet fusion; Polyamide 12; Build position; Thermal history; Crystallization; Mechanical properties; Storage conditions; Long term ageing; Artificial neural network
• Online Access: https://hdl.handle.net/10356/174572
• Institution: Nanyang Technological University

Multi Jet Fusion (MJF) is a relatively new powder-based sintering additive manufacturing technique (AM) that exhibits great potential in high-volume manufacturing due to its rapid printing speed. The most common material used for this technique is Polyamide 12 (PA12). For AM techniques to be utilized for the mass production of the commercial product, reproducibility and reliability of the printed part are required. While many studies has been conducted to study the mechanical properties of MJF printed PA12 parts (MJF PA12), there were contradictory results reported across several works that had yet to be addressed. There is also lack of work performed to study the effect of post-printing factors such as long-term ageing and humidity on MJF PA12 parts. In addition, currently no model has been developed to predict the mechanical properties of MJF PA12 parts with consideration of the influencing factors that could affect the parts’ performance. Such model would be useful in providing some guidance and preliminary information on the material for the user before printing. Hence, this thesis focused on the investigation on the influence of factors such as build positions, build orientations, storage period, and moisture content on mechanical properties of the MJF PA12 printed parts to improve their reproducibility and reliability for the fabrication of commercial parts. The differential scanning calorimetry (DSC) test, dynamic mechanical analysis (DMA), X-ray diffraction (XRD) test, uniaxial tension test, and scanning electron microscopy (SEM) test were performed. Following, a machine learning workflow, consisting of a sequence of neural network models, was developed to analyse the complex relations between different influencing factors and to the performances of the printed parts. Firstly, an in-depth investigation of the effect of build position on the thermal history, crystallization, and mechanical properties of MJF PA12 was conducted. From the Thermal Prediction Engine developed by Hewlett Packard Labs (HP labs, Palo Alto, California), it was found that as compared to the parts printed in the middle region of the chamber, parts printed in the boundary regions experienced a faster cooling rate. A slightly lower tensile modulus was found for these parts printed in the boundary. However, a significantly larger elongation at break and strain energy density was found for these parts as compared to those printed in the middle regions. This work serves as a guide to selecting the build position of MJF PA12 to obtain the desirable mechanical properties. Next, influence of post-printing factors, such as humidity and long-term ageing, on the physical and mechanical properties of differently-orientated MJF PA12 specimens stored under ambient and dryer conditions for 474 days were investigated. The effect of moisture absorption was found to have insignificant effect on crystallinity and crystallite size of the MJP PA12 printed specimens. However, it was found that with higher moisture absorption rate of the parts, the change in mechanical properties, such as the tensile modulus and ultimate tensile strength, was more significant. This work serves to provide abetter reliability and safety assessment for MJF-printed products. A machine learning workflow consisting of two artificial neural networks, Enigma Box 1, and Enigma Box 2, was developed to predict the mechanical properties of MJF PA12. The Enigma Box 1 aims to predict the crystallinity of the MJF PA12 specimens from the features extracted from their cooling histories. The crystallinity predicted by Enigma Box 1 was used as input for Enigma Box 2. Hence, Enigma Box 2 aims to predict the mechanical properties of the MJF PA12 dog-bone specimens under the influence of build position, build orientation, storage period, and moisture content. The predicted results from the trained model were accurate, especially for the tensile modulus, yield strength, and ultimate tensile strength with a mean absolute percentage error (MAPE) of less than 5.0%.