The Effect, Prediction, and Optimization of Fe Particles on Wear Behavior of Fe-ABS Composites Fabricated by Fused Deposition Modeling

Fused deposition modeling, one of the additive manufacturing techniques, has drawn the most attention because it can produce intricate parts quickly and cheaply. However, low wear properties can limit its application. To solve this, Fe particles with three different weights including 10, 20, and 30 percent were added to pure ABS to build Fe-ABS composite filament, after which parts were printed with three parameters, namely filling pattern, nozzle temperature, and layer thickness based on the face centered central composite design. The pin on disk test, then, was conducted to evaluate the impact of Fe-percentage, filling pattern, nozzle temperature, and layer thickness on the wear rate. It was also revealed that the Fe-percentage plays the most important role in wear of Fe-ABS composites and the amount of it decreased when the percentage of reinforcement was increased. The results have also shown that the filling pattern was the second important element and the nozzle temperature was the least effective component. After that, Response surface methodology-based model and the couple of genetic algorithm and artificial neural network were used for prediction and optimization. The performance of GA-ANN was slightly better than RSM with the total goodness function of 1.9964 and 1.981076 respectively. Finally, the result of optimization was in a perfect agreement with the experiments with the error of 0.25%.

Automated summary

Fused deposition modeling is a popular additive manufacturing technique known for its ability to quickly and inexpensively produce intricate parts. However, the low wear properties of the materials limit its application. To address this issue, Fe particles were added to ABS to create Fe-ABS composite filament, and parts were printed with varying parameters. The wear rate was evaluated through pin-on-disk tests, and it was found that the Fe percentage had the greatest impact on wear, followed by the filling pattern and nozzle temperature. Optimization was then carried out using a response surface methodology-based model and genetic algorithm with artificial neural network. The optimized result closely matched the experimental result with only a 0.25% error.