Home > Research > Publications & Outputs > Mass optimisation of 3D-printed specimens using...

Electronic data

  • X_ULMEANU_272113 - for PURE

    Rights statement: © 2022 Inderscience Enterprises Ltd.

    Accepted author manuscript, 2.97 MB, PDF document

    Available under license: CC BY-NC: Creative Commons Attribution-NonCommercial 4.0 International License

  • IJRapidM MEX Regression

    Rights statement: © 2022 Inderscience Enterprises Ltd.

    Final published version, 981 KB, PDF document


Text available via DOI:

View graph of relations

Mass optimisation of 3D-printed specimens using multivariable regression analysis

Research output: Contribution to Journal/MagazineJournal articlepeer-review

  • Cristian-Vasile Doicin
  • Mihaela-Elena Ulmeanu
  • Allan Rennie
  • Elena Lupeanu
<mark>Journal publication date</mark>31/12/2021
<mark>Journal</mark>International Journal of Rapid Manufacturing
Issue number1
Number of pages22
Pages (from-to)1-22
Publication StatusPublished
<mark>Original language</mark>English


Fused deposition modelling popularity is attributed to equipment affordability, materials availability and open-source software. Given the variety of optimisation combinations, process parameters can be elaborate. This paper provides methods for optimisation of mass calculation using multivariable regression analysis. Layer thickness, extrusion temperature and speed were considered independent variables for a two-level factorial experiment. DOE was used for 12 sets of programs and analysis (two stages) undertaken using Design-Expert® V11 Software. In stage-1, four models were found to be significant. Stage-2 involved redesigning the remaining eight models, iteratively increasing the number of replicates and blocks. Adequacy of models was analysed, demonstrating that: model is significant, F-value is large, p < 0.05; lack of fit is insignificant; adequate precision >4.00; residuals are well behaved; R^2 is as close as possible to 1.00 or for models with multiple replicates, the adjusted R^2 and predicted R^2 differential <0.2. All models were validated through measured, calculated responses.