Prediksi Kualitas Klinker Semen Melalui Kandungan F-Cao Dengan Teknik Machine Learning Berdasarkan Parameter Operasi Pembakaran Kiln

Hantoro, Arthur Hajar (2023) Prediksi Kualitas Klinker Semen Melalui Kandungan F-Cao Dengan Teknik Machine Learning Berdasarkan Parameter Operasi Pembakaran Kiln. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032211023-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (4MB) | Request a copy

Abstract

PT. Semen XYZ memerlukan pembakaran kiln yang tepat untuk menghasilkan klinker semen yang baik. Kandungan F-CaO dalam klinker merupakan sisa batu kapur tidak terbakar sempurna dalam indeks persen. Indeks F-CaO dianggap penting karena mempengaruhi kualitas klinker dan menurunkan kualitas semen. Kandungan f-CaO dalam klinker semen tidak dapat diukur secara langsung dan diprediksi komposisinya secara cepat serta akurat. Prediksi f-CaO sulit dilakukan karena hubungan antar variabel dan ketidakpastian proses kalsinasi klinker. Diperlukan model prediksi hasil klinker berdasarkan parameter operasi sehingga mudah dikendalikan oleh operator CCR maupun manager operasi. Prediksi kandungan f-CaO dapat mempengaruhi kualitas semen dan dapat digunakan dalam optimasi konsumsi energi dalam proses produksi klinker.
F-CaO klinker semen diprediksi berdasarkan data operasi yang bersifat real-time tiap jam yang dikumpulkan selama satu tahun. Data yang diperlukan meliputi parameter bahan baku umpan dan parameter operasi yang diatur pada Central Control Room (CCR). Data yang terkumpul perlu diproses dahulu menggunakan pembersihan outlier, pembersihan baris dengan nilai kosong dan normalisasi. Setelah data yang terkumpul siap, data diolah dengan menggunakan beberapa teknik, diantaranya Support Vector Regression (SVR), Neural Network (NN) dan Gradient Boosting (GB). Diperlukan pengaturan yang tepat pada ketiga teknik tersebut agar menghasilkan prediksi yang paling dekat dengan data aktualnya.
Faktor utama yang mempengaruhi hasil prediksi meliputi: temperatur material pada stage 4 SLC (0.536), total batu bara burner (0.536), temperatur exit preheater SLC (0.531), total kilnfeed yang masuk pada preheater (0.526), kecepatan ID fan EP (0.518), speed kiln (-0.517), temperatur material ILC (0.513), draft udara pada inlet kiln (0.490), LSF pada klinker (0.481), vakum pada kiln hood draft (0.412). Algoritma Gradient Boosting (GB) terpilih sebagai algoritma yang terbaik dengan nilai evaluasi MSE 0.107, RMSE 0.328, MAE 0.216 dan R-squared 0.887.

====================================================================================================================================

PT. XYZ cement requires proper kiln combustion to produce good cement clinker. The F-CaO content in clinker is the rest of limestone not completely burned in the percent index. The F-CaO index is considered important because it affects clinker quality and decreases cement quality. The f-CaO content in cement clinker cannot be measured directly and its composition is predicted quickly and accurately. F-CaO prediction is difficult because of the relationship between variables and the uncertainty of the clinker calcination process. A clinker result prediction model is needed based on operating parameters so that it is easily controlled by CCR operators and operations managers. Prediction of f-CaO content can affect cement quality and can be used in optimizing energy consumption in clinker production processes.
F-CaO cement clinker is predicted based on real-time hourly operating data collected over the course of a year. The required data includes feed raw material parameters and operating parameters set in the Central Control Room (CCR). The collected data needs to be processed first using outlier cleanup, row cleaning with blank values and normalization. After the collected data is ready, the data is processed using several techniques, including Support Vector Regression (SVR), Neural Network (NN) and Gradient Boosting (GB). It is necessary to find the right settings in order to produce predictions that are closest to the actual data.
The main factors influencing the prediction results include: material temperature at stage 4 SLC (0.536), total coal burner (0.536), exit preheater SLC temperature (0.531), total kilnfeed entering the preheater (0.526), ID fan EP speed (0.518), kiln speed (-0.517), ILC material temperature (0.513), air draft at the kiln inlet (0.490), LSF at clinker (0.481), vacuum at the kiln hood draft (0.412). The Gradient Boosting (GB) algorithm was chosen as the best algorithm with an evaluation value of MSE 0.107, RMSE 0.328, MAE 0.216 and R-squared 0.887.

Keywords: Free lime, Gradient Boosting, Machine Learning, Neural Network, Support Vector Regression.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Free lime, Freelime, Gradient Boosting, Machine Learning, Neural Network, Support Vector Regression
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. T Technology > TP Chemical technology > TP883 Portland cement.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Arthur Hajar Hantoro
Date Deposited:	06 Aug 2023 14:04
Last Modified:	06 Aug 2023 14:04
URI:	http://repository.its.ac.id/id/eprint/104093

Actions (login required)

View Item