TY - JOUR
T1 - Predictability of mechanical behavior of additively manufactured particulate composites using machine learning and data-driven approaches
AU - Malley, Steven
AU - Reina, Crystal
AU - Nacy, Somer
AU - Gilles, Jérôme
AU - Koohbor, Behrad
AU - Youssef, George
N1 - Publisher Copyright:
© 2022 The Authors
PY - 2022/11
Y1 - 2022/11
N2 - Additive manufacturing and data analytics are independently flourishing research areas, where the latter can be leveraged to gain a great insight into the former. In this paper, the mechanical responses of additively manufactured samples using vat polymerization process with different weight ratios of magnetic microparticles were used to develop, train, and validate a neural network model. Samples with six different compositions, ranging from neat photopolymer to a composite of photopolymer with 4 wt.% of magnetic particles, were manufactured and mechanically tested at quasi-static strain rate and ambient environmental conditions. The experimental data were also synthesized using a data-driven approach based on shape-preserving piecewise interpolations while leveraging the concept of simple micromechanics rule of mixture. The overarching objective is to forecast the mechanical behavior of new compositions to eliminate or reduce the need for exhaustive post-manufacturing testing, resulting in an accelerated product development cycle. The ML model predictions were found to be in excellent agreement with the experimental data for prognostication of the mechanical behavior of physically tested samples with near-unity correlation coefficients. Furthermore, the ML model performed reasonably well in predicting the mechanical response of untested, newly formulated compositions of photopolymers and magnetic particles. On the other hand, the data-driven approach predictions suffered from processing artifacts, demonstrating the superiority of ML algorithms in handling this type of data. Overall, this analysis approach holds great potential in advancing the prospects of additive manufacturing and model-less mechanics of material analyses. A byproduct of the ML approach is using the results for quality assurance, accelerating the acceptance of additively manufactured parts into industrial deployments.
AB - Additive manufacturing and data analytics are independently flourishing research areas, where the latter can be leveraged to gain a great insight into the former. In this paper, the mechanical responses of additively manufactured samples using vat polymerization process with different weight ratios of magnetic microparticles were used to develop, train, and validate a neural network model. Samples with six different compositions, ranging from neat photopolymer to a composite of photopolymer with 4 wt.% of magnetic particles, were manufactured and mechanically tested at quasi-static strain rate and ambient environmental conditions. The experimental data were also synthesized using a data-driven approach based on shape-preserving piecewise interpolations while leveraging the concept of simple micromechanics rule of mixture. The overarching objective is to forecast the mechanical behavior of new compositions to eliminate or reduce the need for exhaustive post-manufacturing testing, resulting in an accelerated product development cycle. The ML model predictions were found to be in excellent agreement with the experimental data for prognostication of the mechanical behavior of physically tested samples with near-unity correlation coefficients. Furthermore, the ML model performed reasonably well in predicting the mechanical response of untested, newly formulated compositions of photopolymers and magnetic particles. On the other hand, the data-driven approach predictions suffered from processing artifacts, demonstrating the superiority of ML algorithms in handling this type of data. Overall, this analysis approach holds great potential in advancing the prospects of additive manufacturing and model-less mechanics of material analyses. A byproduct of the ML approach is using the results for quality assurance, accelerating the acceptance of additively manufactured parts into industrial deployments.
UR - http://www.scopus.com/inward/record.url?scp=85133865465&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85133865465&partnerID=8YFLogxK
U2 - 10.1016/j.compind.2022.103739
DO - 10.1016/j.compind.2022.103739
M3 - Article
AN - SCOPUS:85133865465
SN - 0166-3615
VL - 142
JO - Computers in Industry
JF - Computers in Industry
M1 - 103739
ER -