4Sight: Predictive Analytics Platform for Critical Oil and Gas Transmitter Performance with Time-to-Failure Forecasting

Sinaga, Raihan Devano (2026) 4Sight: Predictive Analytics Platform for Critical Oil and Gas Transmitter Performance with Time-to-Failure Forecasting. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211198-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Peramalan transmitter kritis penting untuk mendukung intervensi tepat waktu dalam operasi industri, tetapi banyak platform data terkontekstualisasi belum menyediakan analitik prediktif yang mudah diakses dan dapat digunakan sesuai kebutuhan. Penelitian ini mengembangkan 4Sight, sebuah alat pendukung keputusan untuk peramalan dan estimasi waktu menuju kegagalan (TTF) pada lingkungan data berbasis digital twin. Data deret waktu sensor historis diambil, dibersihkan, diresampling pada interval yang teratur, serta diperkaya dengan analitik ambang sigma dan laju perubahan sebelum dimodelkan menggunakan LSTM, XGBoost, Random Forest, dan SVM. TTF diestimasi berdasarkan pelampauan ambang hasil peramalan. Dengan menggunakan tiga dataset transmitter, data per jam selama satu tahun, validasi kronologis, dan horizon peramalan dua bulan, LSTM menghasilkan akurasi terbaik, sedangkan Random Forest menjadi alternatif yang lebih cepat. 4Sight diterapkan melalui antarmuka Streamlit dengan mode Fast, Balanced, dan Performance untuk membantu pengguna memahami keseimbangan antara beban komputasi dan kedalaman analisis. Hasil penelitian menunjukkan bahwa integrasi peramalan, analitik TTF, dan penerapan berorientasi pengguna dapat mendukung pemantauan proaktif serta perencanaan operasional.
====================================================================================================================================
Critical transmitter forecasting is essential for timely intervention in industrial operations, yet many contextualized data platforms still lack accessible, on-demand predictive analytics. This study develops 4Sight, a decision-support tool for forecasting and time-to-failure (TTF) estimation on top of digital twin-based industrial data environments. Historical sensor time-series were retrieved, cleaned, resampled to regular intervals, and enriched with sigma-threshold and rate-of-change analytics before being modelled using LSTM, XGBoost, Random Forest, and SVM. TTF was estimated from forecasted threshold breaches. Using three representative transmitter datasets, one year of hourly data, chronological holdout validation, and a two-month forecast horizon, LSTM achieved the best overall accuracy, while Random Forest provided a faster alternative for lighter analysis. 4Sight was deployed through a Streamlit interface that exposes Fast, Balanced, and Performance analysis modes, helping users understand the trade-off between computational effort and analytical depth in simple terms to focus on actionable insights. Overall, the study shows that integrating forecasting, TTF analytics, and user-centred deployment can transform contextualized industrial time-series data into a transferable decision-support capability for proactive monitoring and operational planning across different organizations and data contextualization platforms.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Predictive analytics, Time-to-Failure, Machine learning, Predictive maintenance, Time-series forecasting, LSTM
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Raihan Devano Sinaga
Date Deposited:	25 Sep 2026 06:11
Last Modified:	25 Sep 2026 06:11
URI:	http://repository.its.ac.id/id/eprint/144361

Actions (login required)

View Item