Pradipta, Ivan (2026) Investigating The Dimensional Accuracy For 3D Printed Part Of Kenaf/Pp Composite Materials. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5007211062_Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5007211062_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy |
Abstract
Akurasi dimensi merupakan aspek penting dalam fused deposition modelling karena memengaruhi fungsi dan kesesuaian perakitan komponen hasil fabrikasi. Penelitian ini menganalisis pengaruh printing speed, infill density, layer height, dan kandungan kenaf terhadap akurasi dimensi spesimen kenaf/PP hasil 3D printing. Sebanyak 54 spesimen dibuat dari 18 kombinasi parameter dengan tiga replikasi. Lebar diukur pada W1–W4, ketebalan pada T1–T4, dan panjang diukur satu kali pada setiap spesimen. Data dianalisis menggunakan General Linear Model, analysis of variance (ANOVA), dan metode Tukey. W1 dan W4 dikelompokkan sebagai outer width, sedangkan W2 dan W3 dikelompokkan sebagai inner width. Printing speed dan infill density berpengaruh signifikan terhadap panjang, infill density dan kandungan kenaf berpengaruh signifikan terhadap ketebalan, sedangkan seluruh parameter yang diteliti berpengaruh terhadap lebar. Model artificial neural network (ANN) dikembangkan menggunakan Bayesian regularisation dengan satu hidden layer dan fungsi linear pada output layer. Model hidden layer terpilih menggunakan logistic sigmoid dengan lima neuron untuk outer width, hyperbolic tangent dengan sebelas neuron untuk inner width, logistic sigmoid dengan tujuh neuron untuk ketebalan, dan rectified linear unit (ReLU) dengan enam neuron untuk panjang. Model ANN diintegrasikan dengan Genetic Algorithm (GA). Sepuluh kali pengulangan menghasilkan kombinasi optimum yang sama, yaitu printing speed 50 mm/s, infill density 52%, layer height 0,19 mm, dan kandungan kenaf 5 wt%, dengan nilai objective sebesar 0,2571. Prediksi dimensi yang dihasilkan adalah 63,0877 mm untuk panjang, 12,8371 mm untuk inner width, 12,9361 mm untuk outer width, dan 3,3123 mm untuk ketebalan. Validasi menggunakan lima spesimen menghasilkan mean percentage error masing-masing sebesar 0,272%, 1,074%, 1,655%, dan 2,634%. Hasil ini menunjukkan bahwa metode ANN-GA mampu menentukan kombinasi parameter yang stabil dan dapat diterapkan secara praktis untuk meningkatkan akurasi dimensi.
===================================================================================================================================
Dimensional accuracy is a critical aspect of fused deposition modelling because it affects the functionality and assembly compatibility of fabricated components. This study investigates the effects of printing speed, infill density, layer height, and kenaf content on the dimensional accuracy of 3D-printed kenaf/PP specimens. A total of 54 specimens were fabricated using 18 parameter combinations with three replications. Width was measured at W1–W4, thickness at T1–T4, and length was measured once for each specimen. The data were analysed using the General Linear Model (GLM), analysis of variance (ANOVA), and the Tukey method. W1 and W4 were grouped as outer width, whereas W2 and W3 were grouped as inner width. Printing speed and infill density significantly affected length, infill density and kenaf content significantly affected thickness, and all investigated parameters significantly affected width. Artificial neural network (ANN) models were developed using Bayesian regularisation with a single hidden layer and a linear activation function in the output layer. The selected hidden-layer models employed a logistic sigmoid function with five neurons for outer width, a hyperbolic tangent function with eleven neurons for inner width, a logistic sigmoid function with seven neurons for thickness, and a rectified linear unit (ReLU) function with six neurons for length. The ANN models were integrated with a Genetic Algorithm (GA). Ten repeated optimisation runs consistently produced the same optimum parameter combination: a printing speed of 50 mm/s, an infill density of 52%, a layer height of 0.19 mm, and a kenaf content of 5 wt%, with an objective value of 0.2571. The predicted dimensions were 63.0877 mm for length, 12.8371 mm for inner width, 12.9361 mm for outer width, and 3.3123 mm for thickness. Validation using five specimens produced mean percentage errors of 0.272%, 1.074%, 1.655%, and 2.634%, respectively. These results indicate that the ANN-GA approach can identify a stable and practically applicable parameter combination for improving dimensional accuracy.
Actions (login required)
 |
View Item |