Optimasi Matriks Risiko Dengan Menggunakan Algoritma Artificial Neural Network Pada Peralatan High Pressure Steam Drum (1-V-101) Di Pt. Abc

Dwianto, Sigit (2026) Optimasi Matriks Risiko Dengan Menggunakan Algoritma Artificial Neural Network Pada Peralatan High Pressure Steam Drum (1-V-101) Di Pt. Abc. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5011221013-Undergraduated_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

High Pressure Steam Drum merupakan peralatan kritis pada industri petrokimia yang beroperasi pada tekanan dan temperatur tinggi sehingga rentan mengalami degradasi material yang dapat meningkatkan risiko kegagalan. Penelitian ini bertujuan melakukan penilaian risiko berdasarkan standar API RP 581 serta mengembangkan model Artificial Neural Network (ANN) Multi-Output sebagai metode pre-screening untuk memprediksi kategori Probability of Failure, Consequence of Failure, dan tingkat risiko peralatan. Penelitian diawali dengan identifikasi mekanisme kegagalan menggunakan API 571 dan ASME PCC-3, kemudian dilanjutkan dengan perhitungan POF dan COF berdasarkan API 581. Hasil analisis menunjukkan bahwa mekanisme kegagalan dominan pada High Pressure Steam Drum adalah thinning, sehingga perhitungan damage factor difokuskan pada mekanisme tersebut. Berdasarkan evaluasi API 581, peralatan berada pada kategori risiko Medium. Model ANN yang dikembangkan menunjukkan proses pembelajaran yang stabil tanpa indikasi overfitting serta mampu memprediksi kategori POF dengan akurasi 73% dan weighted F1-score 0,72, sedangkan prediksi COF memperoleh akurasi 50% dengan weighted F1-score 0,46. Analisis distribusi prediksi melalui confusion matrix menunjukkan sebagian besar hasil prediksi berada pada kategori yang sama atau berdekatan dengan kelas aktual sehingga mampu mempertahankan konsistensi klasifikasi tingkat risiko. Analisis SHAP (Shapley Additive Explanation) menunjukkan bahwa Remaining Life, Current Thickness, dan Corrosion Rate merupakan parameter yang paling berpengaruh terhadap prediksi POF, sedangkan parameter kondisi operasi dan karakteristik material memberikan kontribusi terhadap prediksi COF. Hasil penelitian menunjukkan bahwa model ANN Multi-Output mampu mempelajari hubungan antara parameter operasi, inspeksi, dan material terhadap kategori risiko berdasarkan API 581 serta berpotensi digunakan sebagai metode pre-screening untuk mempercepat proses penilaian Risk Based Inspection tanpa menggantikan metode perhitungan standar.
==========================================================================================================================================
High-pressure steam drums are critical equipment in the petrochemical industry that operate at high pressures and temperatures, making them susceptible to material degradation that can increase the risk of failure. This study aims to conduct a risk assessment based on the API RP 581 standard and to develop a Multi-Output Artificial Neural Network (ANN) model as a pre-screening method to predict the POF, COF, and risk level of the equipment. The study began with the identification of failure mechanisms using API 571 and ASME PCC-3, followed by the calculation of POF and COF based on API 581. The analysis results indicate that the dominant failure mechanism in High-Pressure Steam Drums is thinning; therefore, the damage factor calculation focused on this mechanism. Based on the API 581 evaluation, the equipment falls into the Medium risk category. The developed ANN model demonstrated a stable learning process with no signs of overfitting and was able to predict the POF category with 73% accuracy and a weighted F1-score of 0.72, while COF predictions achieved 50% accuracy with a weighted F1-score of 0.46. Analysis of the prediction distribution via the confusion matrix shows that most prediction results fall into the same category or are close to the actual class, thereby maintaining consistency in risk-level classification. SHAP (Shapley Additive Explanation) analysis indicates that Remaining Life, Current Thickness, and Corrosion Rate are the parameters that most influence POF predictions, while operating condition and material characteristic parameters contribute to COF predictions. The research results show that the Multi-Output ANN model is capable of learning the relationships between operational, inspection, and material parameters and risk categories based on API 581 and has the potential to be used as a pre-screening method to accelerate the Risk-Based Inspection assessment process without replacing standard calculation methods.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	artificial neural network, high pressure steam drum, risk based inspection.
Subjects:	Q Science Q Science > Q Science (General) Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Industrial Technology > Material & Metallurgical Engineering > 28201-(S1) Undergraduate Thesis
Depositing User:	Sigit Dwianto
Date Deposited:	24 Jul 2026 02:12
Last Modified:	24 Jul 2026 02:12
URI:	http://repository.its.ac.id/id/eprint/136871

Actions (login required)

View Item