Spray combustion in auxiliary marine boilers: an experimental and numerical study

Abstract

Emissions and pollution reduction have been topics of constant interest for any industrial processes in the past decades and the marine sector is no exception. The vast majority of the marine industry is constituted by shipping, which moves around 80% of the world trade and is therefore an essential part of the global economy and supply chain. As such, it also is responsible for a significant share of the anthropogenic emission of pollutants and greenhouse gasses (GHG). Indeed, there is an immediate necessity for solutions to improve the design of new large ships and their on-board systems. One of the systems affected by the push towards alternative fuels are the auxiliary boilers, which have also seen more and more use in recent years, due to advent of slow steaming. This study is part of the effort to cope with these new demands by achieving more efficient and flexible auxiliary boilers. As part of this effort, a necessary step is to gain further insight into behaviour of the system at hand. While both experimental and computational methods have been applied extensively in the past to the study of liquid fuel flames and sprays, each approach has its limitations. To obtain detailed data with experimental methods is lengthy and expensive, while Computational Fluid Dynamic (CFD) simulations need to be validated before being used for practical purposes and might become cumbersome if too many phenomena are included. Therefore, a combination of the two was used in this project. The experimental work consisted of two parts. The first was a measuring campaign on a full-scale boiler to obtain data, which could be used to validate the CFD model. The second part was a characterization study of the atomizer used in the full-scale boiler to define the fuel injection in the CFD simulations. In addition to the collection of validation data, the measuring campaign was also a chance to fully test the boiler operating capability and to observe the effects of changing operating conditions, such as oil type, load and excess air, on flame stability and emissions. During the project, a full-scale boiler was modified with ports and connections to insert probes into the furnace and to sample the gas at the exhaust. The experimental investigation encompassed both boiler operation change and flame mapping. For the first part, the boiler has been run at three different loads and the flow rate of combustion air was changed to obtain an oxygen concentration in the exhaust over a range between 1% and 6%. The second part of the campaign consisted of gathering data to validate future CFD simulations. The experiments were repeated for marine diesel fuel and Heavy Fuel Oil (HFO). Both parts of the experimental campaign yielded interesting results. It was observed that using HFO compared to diesel had an almost negligible impact on heat transfer and temperatures in the boiler, but increased the emission of particulate almost tenfold, and of NOX and CO from 3 to 5 times. It was also possible to conclude that working at higher loads had a negative effect on all specific emissions, with the sole exception of diesel particulate emissions, which were consistently lower for higher loads. The increase in CO emission with load was clear when increasing the load from 40% to 60% for both fuels. The relative changes for NOX and particulate did not show such clear trends. Moreover, an air-fuel equivalence ratio (λ) above 1.2 did not lead to a further decrease in CO emission. While particulate emissions consistently decreased with increased λ, NOX concentration showed a relatively flat behaviour. The second experimental investigation was a spray characterization study on a spill-back atomizer, which provided the data needed as an input for the CFD calculations. The implementation of these data avoided the necessity of directly simulating the spray break-up process in the CFD simulations, thus greatly reducing their computational costs. This study also resulted in novel spray diagnostic method and further insight into the influence of key nozzle operation parameters on spray characteristics. The atomization characteristics of the nozzle were investigated in terms of droplet size and velocity distributions. The spray seen in the full-scale boiler was replicated in a cold setup. A pulsating LED optical imaging system that employs a CCD camera was used to capture image pairs with a delay as short as 1μs. Positions, velocity, sizes and shapes of single droplet were obtained by analysing these images. Water-glycerol solutions with different concentrations were used as model fluids, reproducing the range of viscosities found by rheology studies for fuels used in the full-scale boiler. The nozzle was characterized by varying three parameters: liquid viscosity, supply pressure and flow rate through the nozzle. As also described in literature, the droplet size distribution shifted towards higher values with an increase in viscosity and lower values with an increase in supply pressure, while the changes in flow rate had a negligible impact on the Sauter Mean Diameter. In the same operating range, consistent trends were also quantified for other two macroscopic parameters: the spray cone angle and the mean droplet velocity. The spray cone angle increased for higher viscosities and lower flow rates, while the mean velocity increased with both flow rate and supply pressure. Overall, in addition to the data needed for the CFD simulations, the spray characterization study provided a good understanding of the effects on atomization characteristics that results from changes in boiler load, fuel pressure and viscosity. The data obtained from the spray characterization study was then used to model the fuel injection in CFD simulations of the full-scale boiler in the commercial software ANSYS CFX. The CFD model was validated against the data from the measuring campaign. The comparison with the mapped flame showed that the CO2 concentration profile is well predicted in the simulation, while there is a deviation in the near burner region for temperature and concentrations of CO and O2. This discrepancy was ascribed to the continuation of fuel combustion in the measuring probe. The CFD simulations reproduced quite well the trends for the exhaust conditions observed in the experiments at varying amount of excess air and load. The simulations also correctly predicted the lower limit of combustion air needed to obtain a stable operation of the boiler. For all simulations, a recirculationzone was observed downstream of the swirling plate of the burner, which was identified as a key region for flame stability. The size and shape of its boundaries in relation to the load of the boiler were compared with the observations done during the experimental campaign and showed good qualitative agreement. The achieved CFD model provides a tool to evaluate further improvements in design and operation of the boiler and it is step stone for further practical studies of full-scale spray combustion furnaces.