Adaptive fabrication of material extrusion-AM process using machine learning algorithms for print process optimization

Material extrusion (ME) is an extensively used technique in additive manufacturing for making parts and prototypes. Optimizing the relationship between build time and material usage in ME is challenging but essential for improving production performance. This study aims to find the ideal parameter s...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of intelligent manufacturing Jg. 36; H. 7; S. 5087 - 5111
Hauptverfasser: Sridhar, S., Venkatesh, K., Revathy, G., Venkatesan, M., Venkatraman, R.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York Springer US 01.10.2025
Springer Nature B.V
Schlagworte:
3-D printers
Accuracy
Algorithms
Business and Management
Consumption
Control
Datasets
Design of experiments
Evolutionary algorithms
Extrusion
Fabrication
Fused deposition
Genetic algorithms
Heuristic
Investigations
Machine learning
Machines
Manufacturing
Mechanical properties
Mechatronics
Neural networks
Optimization
Optimization techniques
Parameter identification
Process parameters
Processes
Production
Response surface methodology
Robotics
Sorting algorithms
Statistical analysis
Taguchi methods
Tensile strength
Variance analysis
Material extrusion
Optimization
Machine learning
Material consumption
Build time
ISSN:0956-5515, 1572-8145
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Material extrusion (ME) is an extensively used technique in additive manufacturing for making parts and prototypes. Optimizing the relationship between build time and material usage in ME is challenging but essential for improving production performance. This study aims to find the ideal parameter settings to minimize both time and material consumption. Machine learning (ML) models and a multidisciplinary evolution algorithm were used to predict and optimize the fused deposition process parameters. A definitive screening design (DSD) was conducted on 14 process parameters to identify those that significantly affect the output. Using the response surface method, a dataset of these influential parameters was created for the ML models. The prediction accuracy of the ML models was evaluated using regression metrics. It has been found that the random forest algorithm has a prediction accuracy greater than 90% over the other models after training and testing the dataset. To further optimize the process, the Non-dominated Sorting Genetic Algorithm was utilized to fine-tune the hyperparameters of the best-performing machine-learning model. The results showed that using the optimal parameter settings identified by NSGA-II, build time was reduced by 37% and material usage by 40% compared to standard print settings. A statistical two-way ANOVA test confirmed that these optimized settings significantly (p < 0.05) reduced both time and material consumption. Graphical abstract
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0956-5515
1572-8145
DOI:10.1007/s10845-024-02495-z