Prediksi Kuat Tekan Semen PCC Berdasarkan Senyawa Kimia dan Sifat Fisika Dengan Teknik Machine Learning

Aswamedhika, Aswamedhika (2023) Prediksi Kuat Tekan Semen PCC Berdasarkan Senyawa Kimia dan Sifat Fisika Dengan Teknik Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032211024-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (3MB) | Request a copy

Abstract

Indonesia memiliki misi untuk menjadi negara maju pada tahun 2045. Demi mencapai misi tersebut, pemerintah gencar untuk melakukan pemerataan pembangunan, salah satunya melalui pembangunan infrastruktur. Semen sajamerupakan salah satu komponen yang penting dalam pembangunan infrastruktur. Kuat tekan merupakan salah satu persyaratan kualitas yang harus dipenuhi oleh produk semen. Pengujian kuat tekan di industri menggunakan alat laboratorium memakan waktu lama hingga 28 hari hingga seluruh rangkaian hasil uji selesai. Dengan adanya delay waktu tersebut, akan membuat pengambilan keputusan jika terjadi ketidaksesuaian pada kualitas akan menjadi terlambat. Pengambilan data dilakukan di PT XYZ dengan durasi selama lima tahun. Data diambil dari pengujian laboratorium dan data operasional. Data yang diambil antara lain kuat tekan tiga hari, tujuh hari dan 28 hari sebagai target. Sedangkan fitur berupa senyawa kimia berupa C3S, C2S C3A, C4AF, FCaO terak, CaO, SO3, SiO2, FCaO semen, Mgo, Al2O3, FeO2, serta sifat fisika semen berupa LOI, INSOL, Blaine dan R45. Data dilakukan preprocessing, agar menjadi data yang siap dilakukan permodelan. Dalam penelitian ini menggunakan algaritma machine learning berupa regresi linier, random forest, dan neural network. Hasil dari permodelan kemudian dibandingkan dan dipilih berdasarkan akurasi terbaik. Hasil evaluasi model didapatkan bahwa neural network merupakan model terbaik untuk memprediksi kuat tekan tiga hari dengan nilai MSE 92.434, RMSE 9.614, MAE 6.574, MAPE 0.035 dan R2 0.875. Model prediksi kuat tekan tujuh dan 28 hari diperoleh dari algoritma random forest. Akurasi pada prediksi kuat tekan tujuh hari MSE sebesar 250.730, RMSE 15.834, MAE 11.957, MAPE 0.047, dan R2 0.749. akurasi prediksi kuat tekan 28 hari. MSE sebesar 403.999, RMSE 20.100, MAE 15.149, MAPE 0.046, dan R2 sebesar 0.712
====================================================================================================================================
Indonesia has a mission to become a developed country by 2045. In order to achieve this mission, the government is aggressively carrying out equitable development, one of which is through infrastructure development. Cement is one of the important components in infrastructure development. Compressive strength is one of the quality requirements that must be met by cement products. Compressive strength testing in industry using laboratory equipment takes a long time of up to 28 days for the entire set of test results to be completed. With delay time in cement quality, it will be too late to make a decision if there is a discrepancy in quality. Data collection was taken at PT ABC with a duration 5 years. Data is taken from laboratory testing and operational data. Data is preprocessed in order to data become ready to be modeled. In this study, machine learning algorithms used is linear regression, random forest, and neural networks. The results of the modeling are then compared and selected based on the best accuracy.
The results of the model evaluation found that the neural network is the best model for predicting compressive strength at 3 days with MSE 92.434, RMSE 9.614, MAE 6.574, MAPE 0.35 and R2 0.875. The 7 and 28 day compressive strength prediction models were obtained from the random forest algorithm. The accuracy of the prediction of compressive strength for 7 days MSE is 250,730, RMSE is 15,834, MAE is 11,957, MAPE is 0.047, and R2 is 0.749. compressive strength prediction accuracy of 28 days. MSE is 403,999, RMSE is 20,100, MAE is 15,149, MAPE is 0.046, and R2 is 0.712

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Semen, kuat tekan, regresi linier, random forest, neural network; Cement, Compressive strength, linear regression, random forest, neural network
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Aswamedhika Aswamedhika
Date Deposited:	04 Sep 2023 07:31
Last Modified:	04 Sep 2023 07:31
URI:	http://repository.its.ac.id/id/eprint/103262

Actions (login required)

View Item