Penelitian ini bertujuan menganalisis tiga insiden kerja terkait kelelahan pada operator Heavy Dump Truck (HD) berdasarkan data Driving Monitoring System (DMS) di perusahaan pertambangan batubara Kalimantan Selatan tahun 2025. Desain penelitian adalah studi kasus dengan pendekatan triangulasi sumber, menggabungkan data sekunder berupa log event DMS dan dokumen LPI, data DMS sebagai pembanding, serta data primer dari wawancara mendalam dan Focus Group Discussion (FGD). Ketiga insiden fatigue HD yang tercatat sepanjang tahun 2025 menjadi unit analisis utama. Data yang dianalisis meliputi log event DMS dan dokumen LPI dari tiga insiden. Hasil penelitian menunjukkan bahwa kelelahan bukan kejadian sporadis melainkan kondisi sistemik yang berulang; unit HD 78185 mencatat baseline rata-rata 2,28 event Eyes Closed per hari (dihitung dari hari operasional dengan data tersedia, di luar hari lonjakan) dengan lonjakan episodik hingga 31 event (13,6× baseline) pada 4 Juni 2025 — tujuh hari sebelum insiden. Pola temporal LPI-01 (00.10 WITA) dan LPI-02 (03.18 WITA) konsisten dengan nadir sirkadian two-process model, sementara LPI-03 (08.00 WITA) dijelaskan melalui akumulasi tekanan tidur (Proses S) menjelang akhir shift. Terdapat keterkaitan yang dapat diidentifikasi antara pola event DMS dan kejadian insiden, meskipun pola peningkatan tidak sepenuhnya konsisten dengan prediksi awal. Temuan terpenting adalah pola kegagalan rantai respons yang konsisten: sistem DMS berhasil mendeteksi sinyal kelelahan, namun gagal direspons karena kegagalan teknis notifikasi (LPI-02) dan penghindaran sensor oleh operator (LPI- 03). Fragmentasi data DMS lintas insiden merupakan temuan tersendiri yang mencerminkan celah sistemik dalam tata kelola data monitoring kelelahan. Penelitian ini merekomendasikan implementasi monitoring tren DMS harian, standarisasi retensi data pasca-insiden, sistem pelaporan kelelahan yang tidak menghukum, serta redundansi arsitektur notifikasi platform DMS.

This study aims to analyze three work-related fatigue incidents among Heavy Dump Truck (HD) operators based on Driving Monitoring System (DMS) data at a coal mining company in South Kalimantan in 2025. The research design is a case study with a source triangulation approach, combining secondary data in the form of DMS event logs and LPI documents, DMS data as a comparison, and primary data from in-depth interviews and Focus Group Discussions (FGD). The three HD fatigue incidents recorded throughout 2025 serve as the main unit of analysis. The data analyzed include DMS event logs and LPI documents from the three incidents. The results show that fatigue is not a sporadic occurrence but a recurring systemic condition; HD unit 78185 recorded an average baseline of 2.28 Eyes Closed events per day (calculated from operational days with available data, excluding surge days) with an episodic spike of up to 31 events (13.6× baseline) on June 4, 2025—seven days before the incident. The temporal patterns of LPI-01 (00:10 WITA) and LPI- 02 (03:18 WITA) are consistent with the nadir circadian two-process model, while LPI-03 (08:00 WITA) is explained by the accumulation of sleep pressure (Process S) towards the end of the shift. There is an identifiable link between DMS event patterns and incident occurrence, although the pattern of increases is not entirely consistent with initial predictions. The most important finding is a consistent pattern of response chain failures: the DMS system successfully detected fatigue signals, but failed to respond due to technical notification failures (LPI-02) and operator sensor evasion (LPI-03). The fragmentation of DMS data across incidents is a separate finding that reflects systemic gaps in fatigue monitoring data governance. This study recommends the implementation of daily DMS trend monitoring, standardization of post-incident data retention, a non-punitive fatigue reporting system, and redundancy of the DMS platform notification architecture.