Analisis Dinamika Proses dan Perancangan Pengendalian Level Rerata Berbasis SIMC pada Proses Tangki Penyangga Umpan

Farrly Sheva Rahmawan, Dr.-Ing. Awang Noor Indra Wardana, S.T., M.T., M.Sc., IPM.

2026 | Skripsi | FISIKA TEKNIK

Tangki penyangga umpan kilang PT Pertamina RU IV Cilacap beroperasi se bagai penyangga hidrolik aliran minyak gas berat. Kontroler PI eksisting menerapkan kendali level ketat, bertentangan dengan desain penyangga yang menuntut kendali le vel rerata. Larangan uji langkah kalang terbuka demi keselamatan menjadi kendala utama pembaruan penalaan. Guna mengatasi limitasi data, pendekatan hibrida diterapkan. Identifikasi ko tak abu-abu via L-BFGS-B dieksekusi pada data kalang tertutup untuk mendiagnosis bias identifikasi. Validasi silang dilakukan melalui metode kemiringan, uji Ljung Box, dan interval kepercayaan Bootstrap. Sebagai solusi fisis, parameter dinamika diturunkan dari pemodelan kotak putih berbasis prinsip dasar fisika. Kontroler ken dali level rerata dirancang menggunakan metode kendali model internal Skogestad (SIMC) dan dievaluasi melalui diagram Bode serta simulasi gangguan empiris. Identifikasi kotak abu-abu mengonfirmasi bias kalang tertutup melalui pengu atan proses negatif (Kp,grey = ?0,001096), menyimpang 235,98% (79?) dari fisika aktualnya (Kp,fisika = +0,000806). Berlandaskan pemodelan kotak putih, kontroler disintesis (Kc = 3,447, Ti = 1440 s) dan terbukti stabil secara frekuensi (margin fasa 66,5?, margin penguatan 19,3 dB). Dibandingkan DCS eksisting, kendali usulan mereduksi keausan aktuator (TV) sebesar 24,8?n menyuplai aliran hilir yang jauh lebih halus

Feed Surge Drum at PT Pertamina RU IV Cilacap operates as a hydraulic buffer for the Heavy Gas Oil stream. The existing PI controller utilizes Tight Le vel Control, contradicting the buffer design philosophy that requires Averaging Level Control (ALC). Safety prohibitions on open-loop step tests pose a primary challenge in updating controller parameters. To overcome data limitations, a hybrid approach is employed. Grey-Box iden tification via L-BFGS-B is executed on closed-loop data to diagnose closed-loop iden tification bias. Cross-validation is performed via Slope Method, Ljung-Box test, and Bootstrap Confidence Interval. As a physical resolution, dynamic parameters are derived from P&ID-based White-Box first-principles modeling. The proposed ALC controller is designed using the Skogestad Internal Model Control (SIMC) method and evaluated via Bode plot and empirical stress-testing. Grey-Box identification confirms closed-loop bias through an anomalous ne gative process gain (Kp,grey = ?0,001096), deviating 235.98% (79?) from physical reality (Kp,fisika = +0,000806). Based on the White-Box model, the synthesized ALC controller (Kc = 3,447, Ti = 1440 s) demonstrates superior frequency stability (Pha se Margin 66.5?, Gain Margin 19.3 dB). Compared to the existing DCS, the ALC reduces actuator wear (Total Variation) by 24.8% and provides significantly smoo ther downstream flow.

Kata Kunci : tangki penyangga umpan, kendali level rerata, SIMC, bias identifikasi kalang tertutup,