In lean-burn, pilot-ignited natural-gas/diesel dual-fuel marine engines, combustion phasing can be controlled through the spatial distribution of the diesel pilot (high-reactivity fuel), but the operating window in which combustion stability, avoidance of abnormal combustion, and NOx legislation limits are jointly satisfied is only a few crank-angle degrees wide, and its location and width depend on engine load, gas fraction, pilot mass, and effective compression ratio. The 2017 experimental study that established this on a medium-speed single-cylinder Wärtsilä 20DF engine concluded that cycle-to-cycle control strategies must be implemented for robustness and that the reported windows should be interpreted qualitatively. Closed-loop combustion control has been demonstrated on automotive-scale dual-fuel engines, and a control-oriented modeling lineage (multizone thermo-kinetic models, real-time predictive models, and a 2026 model-in-the-loop adaptive control study) now targets reactivity-controlled marine engines; no published work demonstrates cycle-to-cycle combustion-phasing control through pilot-fuel distribution on a medium-speed marine dual-fuel engine. The problem of achieving this control remains open.

Background

Marine emission legislation (the International Maritime Organization's IMO Tier III limits) imposes strict limits on nitrogen oxides (NOx), which has made lean-burn, pilot-ignited dual-fuel engines burning natural gas with a small diesel pilot one of the leading marine propulsion platforms, since they can form little NOx at the source and reduce aftertreatment requirements. In such engines the cylinder is filled with a lean premixed charge of natural gas — the low-reactivity fuel (LRF), resistant to autoignition — which is ignited by a small directly injected diesel quantity, the high-reactivity fuel (HRF). How that pilot diesel is distributed in space and time when ignition occurs (its stratification) determines the start of combustion, the local heat-release rate, and how the flame propagates through the lean gas mixture. Combustion phasing — usually quantified by the crank angle at which 50% of the heat is released (CA50) — must sit in a narrow range: too early risks knock (abnormal pressure oscillations caused by autoignition of the unburned charge), other abnormal combustion, and high NOx from near-stoichiometric HRF pockets; too late degrades efficiency, raises unburned hydrocarbons, and threatens misfire. Ignition of the pilot itself is governed by low-temperature diesel ignition chemistry with an ignition delay that varies strongly with local temperature, pressure, and pilot/charge equivalence ratio, so the mapping from injection settings to phasing is highly condition-dependent. The related reactivity controlled compression ignition (RCCI) concept generalizes this idea: the reactivity contrast between two fuels is exploited to shape in-cylinder heat release. The source study (García Valladolid et al. 2017, Energy Conversion and Management 149:192–205) mapped the effect of pilot injection timing and injection pressure on ignition, emissions, and stability at medium and high load on a single-cylinder medium-speed Wärtsilä 20DF research engine (fixed 1000 rpm), complemented by one-dimensional spray-distribution simulations. Its key findings were that the HRF spatial distribution controls phasing and hence NOx, instability, and knock; that at high load the injection-timing window in which all constraints (stability, abnormal combustion, NOx) are jointly satisfied is only about 3 crank-angle degrees wide; that the location and width of this window depend on operating conditions such as HRF mass, engine load, and effective compression ratio, so the reported windows are qualitative; and, as its central conclusion, that such narrow windows require cycle-to-cycle control strategies to be implemented for robustness. By cycle-to-cycle control is meant closed-loop control acting every engine cycle: sensing the outcome of the previous cycle(s) (e.g. from in-cylinder pressure) and adapting the injection settings for the next cycle so that phasing stays inside the admissible window despite disturbances. Subsequent work has built control-oriented models — fast thermo-kinetic simulations suitable for controller design — and, more broadly, closed-loop combustion controllers for low-temperature dual-fuel combustion.

Problem Statement

For lean-burn, pilot-ignited natural-gas/diesel dual-fuel medium-speed marine engines of the class studied by García Valladolid et al. (2017) — exemplified by their single-cylinder Wärtsilä 20DF research engine at fixed 1000 rpm and medium and high load — the injection-timing window in which combustion stability (e.g. coefficient of variation of gross indicated mean effective pressure), avoidance of abnormal combustion, and NOx legislation limits are jointly satisfied is only a few crank-angle degrees wide, and its location and width depend on operating conditions including high-reactivity-fuel (pilot diesel) mass, engine load, gas fraction, and effective compression ratio, so that experimentally mapped windows are engine- and condition-specific and 'should be interpreted qualitatively' in the source's words. The admissible window is the source's pilot-injection-timing window, measured on the injection-timing axis: it must be no wider than the source's reported window — roughly the few crank-angle degrees (about 7 CAD at medium load and about 3 CAD at high load) of injection timing over which the source's stability, abnormal-combustion, and NOx criteria are jointly met — and, because combustion phasing varies monotonically with injection timing over the mapped range, keeping injection timing inside this window keeps combustion phasing inside the corresponding acceptable CA50 band. The combustion-stability, abnormal-combustion, and NOx criteria and thresholds that define this window include at least the source's own — a limit on the coefficient of variation of gross indicated mean effective pressure, absence of abnormal combustion, and the applicable engine-out NOx limit.

The open problem is robust NOx-compliant lean natural-gas combustion through control of the diesel-pilot distribution in this engine class: devise a cycle-to-cycle closed-loop control strategy — sensing (directly or indirectly) the combustion outcome of previous cycles, predicting or accounting for ignition-delay and local-reactivity variations, and actuating the pilot injection settings (injection timing and injection pressure as varied in the source study, and fuel quantity/splitting where the injection system provides that authority) — that keeps the pilot injection timing inside the window in which combustion stability, avoidance of abnormal combustion, and NOx legislation limits are jointly satisfied — and thereby combustion phasing inside the corresponding acceptable CA50 band — despite cycle-to-cycle variability, disturbances, and changing operating conditions. This is the source's own concluding call: a window that narrow 'requires cycle-to-cycle control strategies to be implemented for increased robustness'. Because the source reports that window location and width depend on operating conditions such as HRF mass, engine load and effective compression ratio, the required robustness extends across changing operating conditions; the source states that condition-dependence as a limitation on how its mapped windows are to be interpreted ('results should be interpreted qualitatively'), and does not attach a separate cross-engine window-prediction or transfer methodology to its call for cycle-to-cycle control. The open problem is to devise the cycle-to-cycle closed-loop control strategy itself — a design comprising the sensing and actuation scheme and the control law — and to demonstrate on a control-oriented engine model that it keeps the pilot injection timing inside the admissible window, and combustion phasing inside the corresponding acceptable CA50 band, at the source's two tested operating points (medium load, about 10 bar IMEPg and 88% gas fraction, and high load, about 19 bar IMEPg and 95% gas fraction, at fixed 1000 rpm). The engine model must be calibrated so that at the source's two operating points it reproduces the source's reported behaviour — stable combustion (coefficient of variation of gross indicated mean effective pressure below 5%), no abnormal combustion, and engine-out NOx below the limit the source adopts (500 mg/Nm³ at 5% O₂), within the source's injection-timing windows (at medium load about 42–49 CAD bTDC, 7 CAD wide; at high load about 3 CAD wide), and so that those criteria fail just outside the windows, the window location and width shifting with engine load, gas fraction, high-reactivity-fuel mass, and effective compression ratio — before the strategy is assessed on it. With the controller disabled, the imposed cycle-to-cycle disturbance must drive the coefficient of variation of gross indicated mean effective pressure to at least 5%, and with the controller active it must stay below 5%, at both points; and the strategy's robustness must be argued as engine load and gas fraction vary over at least the source's tested spans (about 10 to 19 bar IMEPg and about 88% to 95%) and as high-reactivity-fuel mass and effective compression ratio vary over nondegenerate ranges, stated and justified, that measurably shift the window's location or width, that shift being shown rather than merely asserted. The problem thus comprises the cycle-to-cycle closed-loop strategy design, the calibrated engine model that reproduces the source's behaviour as specified above, and the closed-loop demonstration at the two operating points, including the disturbance test, together with the criteria/threshold justification and the cross-envelope robustness argument above; the sensing, ignition-delay/local-reactivity prediction, and pilot-injection actuation describe the scope such a strategy covers rather than separate sub-problems. Because no externally fixed reference plant is available, the demonstration is on a model calibrated to the source's reported behaviour as specified above, with the cycle-to-cycle disturbance imposed within the demonstration rather than supplied externally.

References

  1. Pablo García Valladolid, Per Tunestål, Javier Monsalve-Serrano, Antonio García, Jari Hyvönen (2017), "Impact of diesel pilot distribution on the ignition process of a dual fuel medium speed marine engine", Energy Conversion and Management 149:192–205. DOI: 10.1016/j.enconman.2017.07.023, https://doi.org/10.1016/j.enconman.2017.07.023
  2. Carlos Guardiola, Benjamín Pla, Pau Bares, Alvin Barbier (2020), "Closed-loop control of a dual-fuel engine working with different combustion modes using in-cylinder pressure feedback", International Journal of Engine Research 21(3):484–496 (first published online March 2019). DOI: 10.1177/1468087419835327, https://doi.org/10.1177/1468087419835327
  3. Vittorio Ravaglioli, Filippo Carra, Davide Moro, Matteo De Cesare, Federico Stola (2018), "Remote Sensing Methodology for the Closed-Loop Control of RCCI Dual Fuel Combustion", SAE Technical Paper 2018-01-0253. DOI: 10.4271/2018-01-0253, https://doi.org/10.4271/2018-01-0253
  4. Sokratis Stoumpos, Gerasimos Theotokatos, Christoforos Mavrelos, Evangelos Boulougouris (2020), "Towards Marine Dual Fuel Engines Digital Twins—Integrated Modelling of Thermodynamic Processes and Control System Functions", Journal of Marine Science and Engineering 8(3):200. DOI: 10.3390/jmse8030200, https://doi.org/10.3390/jmse8030200
  5. Aneesh Vasudev, Alberto Cafari, Martin Axelsson, Maciej Mikulski, Jari Hyvönen (2022), "Towards Next Generation Control-Oriented Thermo-Kinetic Model for Reactivity Controlled Compression Ignition Marine Engines", SAE Technical Paper 2022-01-1033. DOI: 10.4271/2022-01-1033, https://doi.org/10.4271/2022-01-1033
  6. Xiaoguo Storm, Aneesh Vasudev, Amir-Mohammad Shamekhi, Amin Modabberian, Kai Zenger, Jari Hyvönen, Maciej Mikulski (2023), "Real-time predictive model for reactivity controlled compression ignition marine engines", Control Engineering Practice 141:105724. DOI: 10.1016/j.conengprac.2023.105724, https://doi.org/10.1016/j.conengprac.2023.105724
  7. Xiaoguo Storm, Amir-Mohammad Shamekhi, Mohammad Raisi Esfarjani, Amin Modabberian, Aneesh Vasudev, Kai Zenger, Jari Hyvönen, Maciej Mikulski (2026), "Predictive adaptive reactivity-controlled compression ignition for a dual-fuel marine engine: A model-in-the-loop study", Control Engineering Practice 174:107033. DOI: 10.1016/j.conengprac.2026.107033, https://doi.org/10.1016/j.conengprac.2026.107033
  8. Hyunchun Park, Yuri M. Wright, Omar Seddik, Ales Srna, Panagiotis Kyrtatos, Konstantinos Boulouchos (2021), "Phenomenological micro-pilot ignition model for medium-speed dual-fuel engines", Fuel 285:118955. DOI: 10.1016/j.fuel.2020.118955, https://doi.org/10.1016/j.fuel.2020.118955
  9. Menno Merts, Saeed Derafshzan, Jari Hyvönen, Mattias Richter, Marcus Lundgren, Sebastian Verhelst (2021), "An optical investigation of dual fuel and RCCI pilot ignition in a medium speed engine", Fuel Communications 9:100037. DOI: 10.1016/j.jfueco.2021.100037, https://doi.org/10.1016/j.jfueco.2021.100037
  10. Yujun Tang, Xunhu Zhao, Jinfeng Feng, Shuzhan Bai, Guoxiang Li, Huiquan Duan, Sipeng Zhu (2025), "Challenges and opportunities of natural gas dual-fuel engines for marine applications", Fuel 400:135808. DOI: 10.1016/j.fuel.2025.135808, https://doi.org/10.1016/j.fuel.2025.135808