Sistem Prediksi Nilai Resistansi Heater Pada Ladle Casting Area Plate Process Menggunakan Metode Support Vector Regression (SVR)

Basyari, Ahmad Adib (2026) Sistem Prediksi Nilai Resistansi Heater Pada Ladle Casting Area Plate Process Menggunakan Metode Support Vector Regression (SVR). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221016-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Heater pada ladle merupakan komponen krusial pada proses Casting di PT GS Battery Semarang Plant, dan kegagalannya yang tidak terencana dapat menghentikan operasional mesin Casting. Data operasional mencatat sepuluh kali kerusakan heater dalam setahun terakhir, dengan rata-rata downtime 50 menit per insiden akibat strategi maintenance yang masih bersifat reaktif. Penelitian ini mengembangkan sistem prediksi nilai resistansi heater yang memperkirakan degradasi heater ladle menggunakan metode Support Vector Regression (SVR) berdasarkan data historis arus dan resistansi yang diperoleh melalui sensor PZEM-016 dan ESP32. Pengujian sensor menunjukkan rata-rata error sebesar 0,58% untuk tegangan dan 1,13% untuk arus terhadap pengukuran alat ukur, yang menunjukkan akurasi pengukuran yang memadai. Data resistansi harian diolah melalui deteksi siklus penggantian heater, penentuan First Prediction Time (FPT) dan End Of Life (EOL), serta pembentukan fitur lag (lag1, lag2, lag3). Pengujian Hyperparameter dilakukan secara sistematis menggunakan skema cross validation berbasis siklus pada kernel Linear, Polinomial, dan RBF. Berdasarkan nilai MAPE terkecil, kernel Polinomial dengan konfigurasi epsilon = 0,0001, complexity = 10, degree = 1, dan gamma = 10 menghasilkan MAPE 0,74%, diikuti kernel Linear (0,75%) dan RBF (0,95%). Namun, pada validasi prediksi terhadap heater baru pasca-EOL, kernel RBF memberikan APE terkecil sebesar 0,53%, jauh lebih unggul dibandingkan kernel Polinomial (2,61%) dan Linear (3,13%). Sistem ini juga dilengkapi dengan Web Monitoring yang menampilkan visualisasi tren resistansi, status kondisi heater, serta hasil prediksi satu hari ke depan sebagai dasar perencanaan maintenance yang lebih terstruktur.
================================================================================================================================
The ladle heater is a critical component in the Casting process at PT GS Battery, Semarang Plant, and its unplanned failure halts production on the Casting machine. Historical data recorded ten Breakdowns within a year, each causing an average of 50 minutes of downtime under a purely reactive maintenance strategy. This study develops a predictive resistance value system that estimates ladle heater degradation using Support Vector Regression (SVR) based on historical current and resistance data acquired through a PZEM-016 sensor and an ESP32 microcontroller. Voltage and current sensor readings showed average errors of 0.58% and 1.13% against multimeter measurements, confirming acceptable measurement accuracy. Daily resistance data was processed through heater-replacement cycle detection, First Prediction Time (FPT) and End Of Life (EOL) determination, and lag-feature engineering (lag1, lag2, lag3). Hyperparameter tuning was conducted systematically using a cycle-based cross-validation scheme across Linear, Polinomial, and RBF kernels. Based on the lowest MAPE value, the Polinomial kernel with a configuration of epsilon = 0.0001, complexity = 10, degree = 1, and gamma = 10 achieved a MAPE of 0.74%, followed by the Linear kernel (0.75%) and the RBF kernel (0.95%). However, in the predictive validation on a brand-new post-EOL heater, the RBF kernel yielded the lowest APE of 0.53%, substantially outperforming the Polinomial (2.61%) and Linear (3.13%) kernels. The system is also equipped with a Web Monitoring interface that displays resistance trend visualizations, heater condition status, and next-day predictions as a basis for more structured maintenance planning.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Prediksi, Resistansi, Heater, Casting, Support Vector Regression (SVR)
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T174 Technological forecasting T Technology > TJ Mechanical engineering and machinery > TJ217.6 Predictive Control T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Ahmad Adib Basyari
Date Deposited:	11 Aug 2026 03:43
Last Modified:	11 Aug 2026 03:43
URI:	http://repository.its.ac.id/id/eprint/144297

Actions (login required)

View Item