Prediksi Kuat Tekan Semen OPC Berdasarkan Parameter Kimia Fisika Semen Menggunakan Teknik Machine Learning

Ramadhan, Syahrial (2025) Prediksi Kuat Tekan Semen OPC Berdasarkan Parameter Kimia Fisika Semen Menggunakan Teknik Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 2025-07-29 - Tesis - Syahrial R.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Kondisi industri semen yang ultra kompetitif membuat perusahaan persemenan harus berinovasi untuk memenangkan persaingan. Salah satu kendala dalam perumusan produk baru semen adalah proses trial yang memakan waktu setidaknya 1 bulan. Semen hasil produksi harus disimpan selama 28 hari sebelum dapat diuji kuat tekannya. Penggunaan teknik machine learning diusulkan untuk mempercepat proses ini. Parameter yang digunakan sebagai atribut prediktor adalah parameter kimia (C3S, C2S, C4AF, SiO2, dll.) dan parameter fisika (Blaine, Residu, LOI, dll.) dari semen OPC. Parameter prediktor dimodelkan dengan teknik machine learning yaitu random forest, gradient boosting, dan artificial neural network. Hasil pemodelan diuji berdasarkan nilai MAE, RMSE dan nilai R squared untuk menentukan model machine learning yang paling optimal dalam memprediksi kuat tekan semen OPC. Dari hasil penelitian diperoleh metode random forest memberikan koefisien determinasi yang cukup tinggi sebesar 0,856 dengan RMSE sebesar 13,086 kg/cm2 dan MAE sebesar 10,784 kg/cm2. Dengan atribut yang memiliki dampak signifikan adalah CaO, Insol, SiO2, MgO, Al2O3, dan SO3. Performa machine learning random forest dapat ditingkatkan melalui hyperparameter tuning dengan metode grid search. Melalui optimasi hyperparameter diperoleh koefisien determinasi sebesar 0,976 dengan RMSE sebesar 6,118 kg/cm2 dan MAE sebesar 5,198 kg/cm2. =================================================================================================================================
The ultra-competitive nature of the cement industry requires cement companies to continuously innovate to stay ahead in the market. Additionally, shifting trends in the construction materials market—the primary segment for cement companies—further drive the need for innovation. One of the main challenges in developing new cement products is the trial process, which takes at least one month. Freshly produced cement must be stored for 28 days before its compressive strength can be tested. This delay makes customization of Ordinary Portland Cement (OPC) products inefficient. To accelerate this process, machine learning techniques are being developed. The predictive parameters used include chemical parameters (C3S, C2S, C4AF, SiO2, etc.) and physical parameters (Blaine, Residue, LOI, etc.) of the cement. The predictive modeling is performed using machine learning techniques such as random forest, gradient boosting, and artificial neural networks. The model’s performance will be evaluated based on Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared values to determine the most optimal machine learning model for predicting the compressive strength of OPC cement. From the research results obtained that random forest method provides a fairly high coefficient of determination of 0.856 with RMSE of 13.086 kg/cm2 and MAE of 10.784 kg/cm2. With attributes that have a significant impact are CaO, Insol, SiO2, MgO, Al2O3, and SO3. Machine learning performance can further optimize using hyperparameter tuning with grid search method. The result are squared correlation is 0.976, RMSE is 6,118 kg/cm2, and MAE is 5,198 kg/cm2.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	machine learning, kuat tekan, prediktor, machine learning, compressive strength, predictor
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:	Syahrial Ramadhan
Date Deposited:	30 Jul 2025 04:32
Last Modified:	30 Jul 2025 04:32
URI:	http://repository.its.ac.id/id/eprint/123734

Actions (login required)

View Item