3.1. Equivalence ratio at LBO

The equivalence ratio ( $\phi $ ) at LBO for the ten fuels under the above three air flow conditions is directly calculated based on the recorded readings from the mass flowmeter at the LBO point. The average value for each fuel and condition is plotted in Fig. 4, and the error bar is one standard deviation of the results. It can be observed overall that as the inlet air volume increases, all fuels are extinguished at a higher equivalence ratio. This shows that an increase in the amount of inlet air will negatively impact the stability of the lean flame in the combustion chamber. This phenomenon could be due to the fact that the entry of cold air and subsequently increase oxygen supply into the chamber from the dilution hole further enhances the uncertainty of the flow field and reaction rate in the combustion chamber. The mixing degree, secondary atomisation, local incidence of autoignition and combustion reaction rates would all change when flow rate changes.

Figure 4. LBO equivalence ratio results of the ten tested fuels at three air flow conditions.

It also is evident from the Fig. 4 that under all the operating conditions, C1, the fuel consisting of 100% Iso-paraffins with the smallest DCN, blows out at the highest equivalence ratio ( $\phi $ ). This may mean that a fuel composed of a single molecular compound performs poorly in terms of flame stability. On the other hand, this could be an impact of lower DCN value of this particular fuel. Due to lower DCN fuel does not ignites easily, and the autoignition propensity of fuel with such droplets severely limit its ability to continue burning when the fuel is lean. On the other hand, C9, which exhibits the highest DCN, has the lowest equivalence ratio (0.01944 at air flow 0.2kg/s condition). This phenomenon coincides with Won’s speculation [Reference Won, Rock, Lim, Nates, Carpenter, Emerson, Lieuwen, Edwards and Dryer23] that the LBO is highly relative to the autoignition propensity of the fuel as suggested by its DCN. Notably, except for C9, all the alternative fuels with a DCN ranging from 17 to 47 experience blowout at higher lean boundaries than that of A3. Therefore, the impact of other fuel property factors on the LBO performance cannot be ignored.

Since this experimental project is a joint research project, the experimental configuration of each participating research institution varies. In order to ensure that the results of this experiment have universal comparability and to facilitate meta-data interpretation [Reference Zheng, Ahmed, Ubogu, Zhang and Khandelwal35], the results of this experiment have been standardised using the A3 results as the base value. LBO results at different air flow rates presented in Fig. 5 have been normalised against LBO readings of reference fuel A3. Using 0% as the dividing line, a negative value means that its LBO performance is better than A3, and vice versa. It can be clearly observed that C9 fuel gave the best performance and C1 fuel gave the worst performance as compared to A3. This means that, in all air mass flow operating conditions, most alternative fuels showed worse blowout resistance as compared to existing fuels. At the same time, it also indicates that when alternative fuels are used as drop-in fuel, a large amount of theoretical and experimental data is required as a guiding basis to specify strict safety indicators to ensure the safety performance of the engine’s stable combustion.

Figure 5. LBO performance at three air flow conditions normalised against A3.

3.2. Fuel property impact on LBO performance

Regarding the impact of fuel characteristics on LBO, preliminary research data obtained by the four research institutions participating in the NJFCP project shows that fuel evaporation, mixing within the combustor, and chemical reactivity are the three most important factors [Reference Peiffer, Heyne and Colket38], and this relationship is shown in Equation (1).

(1) \begin{equation}\phi \left( {LBO} \right){\rm{\sim }}{\left( {{1 \over {{\tau _{chem}}}} + {1 \over {{\tau _{eva}}}} + {1 \over {{\tau _{mix}}}}} \right)^{ - 1}}\end{equation}

The chemical coefficient ( ${\tau _{chem}}$ ) describes the fuel burning reaction at a given temperature, pressure and local fuel/air ratio. The coefficient for evaporation ( ${\tau _{evap}}$ ) is related to the fuel properties of viscosity, density and surface tension and thus fuel temperature. The mixing coefficient ${\tau _{mix}}$ is the function of combustor geometry and relative convective velocities. Through this formula, it can be observed that there is a linear relationship between LBO and a single factor. For this reason, the linear correlation statistical method was used in following analysis.

Taking the combustion state where the mass flow rate is 0.2kg/s as an example, the linear regression algorithm of fuel properties and LBO performance was used to determine their correlation. The LBO results for all the fuels have been normalised against the data of A3 ( ${\phi _{fuel}}/{\phi _{A3}}$ ), which are shown in the y-axis direction. Since each fuel property has its own unit and order of magnitude, the resulting linear trend line gradient (k) does not have the direct comparability. Therefore, it is necessary to standardise all the characteristic values, in order to limit them to the range of 0–1, where 0 is the minimum value of the same characteristic, 1 is the maximum value, and the rest are distributed in proportions between 0-1. The calculation method is shown in Equation (2).

(2) \begin{equation}{x_{{\rm{standardize}}\;}} = {{x - {x_{min}}} \over {{x_{max}} - {x_{min}}}}\end{equation}

Using the calibrated LBO value with the normalised fuel characteristic value ( ${\phi _{fuel}}/{\phi _{A3}}$ ), scatters diagrams for each fuel metric can be made. The normalised fuel property values represented by Equation (2) are plotted along the x-axis direction. The scatter diagram for each fuel property and normalised LBO performance at the three air flow velocity conditions (0.2kg/s in blue circle symbol, 0.26kg/s in orange cross symbol and 0.32kg/s in green triangle symbol) are shown in Fig. 6. A trend line is added to the scatter chart to show slope (k) and coefficient of determination ( ${R^2}$ ). In each diagram, the gradient of linear trend “k” indicates the influence trend of particular fuel property on the LBO. The downward trend line (k < 0) means that as the value of this characteristic increases, LBO occurs at a lower fuel/air ratio (improved LBO) and vice versa for upward trend line (k > 0). The trend line in each state is also in each plot and distinguished by different colors. The ${R^2}$ value represents the extent to which the established model can explain the proportion of variance of each variable within the sample data. The larger the value, the stronger the model’s ability to explain the sample data, and the better the model fits. To facilitate subsequent analysis based on these statistics, the values of k and ${R^2}$ for each variable is shown in Table 4.

Table 4. The data results of k and ${R^2}$ for linear correlation model

Figure 6. Linear relationship between fuels characteristics and LBO performance.

From an overall point of view, the relationship between most fuel characteristics and its LBO performance is not statistically significant, as only the value R ² for smoke point and DCN exceeds 0.6. This shows that most fuel characteristics when taken in isolation do not directly determine changes in LBO performance. There is interesting phenomenon observed previously that remains following normalisation. The yellow trend line representing 0.26kg/s is at the lowest position in each graph. That is because the value of the ordinate is presented in the form of ratio after normalisation of the A3 data, as previously discussed for the raw data in Fig. 5.

Regarding the degree of influence of each fuel property on performance, in addition to the growth rate, the degree of dispersion should also be taken into consideration. Even if it has a large k value is observed for a given variable, if the degree of data dispersion is very high (low R ²), this means that the reliability of its k value is low. Therefore, when measuring the importance, it is necessary to consider both growth rate (k) and coefficient of determination (R ²). These two factors have been integrated as effect index and expressed as shown in Equation (3). The effect index for each fuel variable at all the three conditions are plotted in Fig. 7.

(3) \begin{equation}Effect\ index = k \times R^2\end{equation}

Figure 7. Effect rank of fuel properties on the LBO performance based on the linear correlation model.

Following the normalisation technique applied using Equation (2), there are positive and negative values, the sign of which precedes the k value. When the effect index is negative, it means that as the value of the variable increases, the performance of LBO will be improved. When it is positive, the opposite is true. It can be observed from Fig. 7 that among all the considered fuel properties, DCN exerts the most significant impact on the LBO value. This is consistent with the conclusion obtained by most institutions participating in the NJFCP project using the random forest regression method [Reference Peiffer, Heyne and Colket38]. It is also worth noting that as the air flow increases, the role of DCN becomes more significant as indicated by the comparatively larger magnitude of the k value at the 0.32 vs 0.2kg/s and 0.26kg/s (−0.184 vs −0.124 and −0.135, respectively). This may be due to the fact that as the amount of intake air increases, the temperature in the combustion chamber decreases and the probability of leaner local equivalence ratio increases. Under these conditions, the fuel’s autoignition propensity plays an important in sustaining stable, steady state combustion. In other words, fuel characteristics related to ignition parameters play an important role in LBO performance. This observation agrees with the findings of Won et al. [Reference Won, Rock, Lim, Nates, Carpenter, Emerson, Lieuwen, Edwards and Dryer23].

The second most important feature, according to the effect index, is the smoke point, which is also one of the main indicators for evaluating aviation fuels’ propensity to produce carbonaceous particulate matter emissions. The higher the smoke point value, the lower the tendency of combustion to generate carbonaceous nvPM, and such fuel is generally observed as burning ‘cleaner’ than a fuel with a low smoke point. Therefore, in order to reduce the propensity of aviation fuel to generate carbonaceous nvPM, aviation fuel product specification standards strictly limit the smoke point value. However, the results of this experiment indicate that a high smoke point will have a negative impact on the performance of LBO, as shown in the corresponding variable in Fig. 7. Therefore, in terms of potential optimisation, it is challenging to obtain lower particulate matter emission performance and improved LBO performance at the same time by adjusting the value of the smoke point. The selection of potential optimised blends must adopt to these conflicting behaviours. Fuel composition could be changed to gain the desired performance, and a balanced position must be taken based on the objectives of the application in question. Studies in the literature show a general trend of smoke point decrease with an increase of aromatics and cycloalkanes content. Studies in the literature also show that the value of smoke point is significantly influenced by the content of mono, cyclo and polycyclic hydrocarbons, although this is a significant field of interest as to what property of aromatic content determines this behaviour [Reference Menon, Eggenspeiler and Porumbel9]. As the content of these two molecules increases, the smoke point decreases. This may also be the reason why the effect value of the proportion of C-paraffins is in the fourth place and potentially improves LBO, opposite to the influence of smoke point. In other words, high aromatics and high C-paraffins content have potential benefits for LBO performance. This can be observed from the Fig. 6, which shows a decreasing LBO trend with increased aromatic content. This is possibly due to the fact that the molecular structure of aromatics and C-paraffin are relatively stable, and it takes longer to burn, causing a certain delay in flameout. The longer burning time helps to ignite subsequent droplets, thus sustaining combustion and improving the performance of LBO. The same theory can also explain why an increase in I-paraffin content (fifth importance by effect value) is detrimental to LBO. The molecules of I-paraffin are easier to decompose during combustion, so that the combustion time is shortened, and steady state combustion is more difficult to sustain as a consequence.

Another key point of discussion is the physical property of fuel such as surface tension, which is the third most important property controlling LBO according to the effect value. Surface tension is an important factor that impacts the degree of atomisation. A fuel with a larger surface tension value will cause poor atomisation, making the droplet particles larger and more difficult to fully react with oxygen. From an LBO point of view, the greater the surface tension, the larger the droplet particles, and they could be better resist being blown out, with-in certain limits of surface tension. This phenomenon is also consistent with the experimental results observed by Grohmann et al. [Reference Grohmann, Rauch, Kathrotia, Meier and Aigner22]. As the macromolecular structure of the fuel analysed previously increases reaction rates and sustains combustion, so too do the longer times to atomise and combust larger droplets, which helps the subsequent ignition of additional droplets. However, we have also observed that with an increase of air intake, the value of surface tension has decreasing influence on the improvement of the LBO value. This is potentially due to the increase in air intake, which promotes the degree of secondary atomisation of droplets, weakening the effect of surface tension, and reduces the ability to resist blowout.

For other fuel properties (net heat combustion, aromatic content, density, hydrogen content, etc.) evaluated in this experiment, there is almost no direct linear influence on the LBO value, so limited conclusions can be drawn, if any, from the effect of altering these variables. But through the above analysis, we can observe that there are several key variables that are the dominant factors in the performance of LBO, namely DCN, smoke point and surface tension. From the above analysis, it appears that both the chemical properties (DCN) and the physical properties (smoke point) are dominant factors impacting the performance of the LBO. But the physical property of fuel’s surface tension is also directly caused by the chemical molecular composition. The ignition propensity (DCN) could be due to concentration of the CH₂ chemical functional group [Reference Won, Haas, Dooley, Edwards and Dryer39]. Therefore, this study predicts that at low temperature and low pressure, the chemical characteristics of the fuel play a decisive factor, and as the air supply in the combustion chamber increases, its decisiveness becomes more obvious.

DCN is used as an indicator to measure ignition performance and does not have a negative impact on emissions. The results above provide the opportunity to further optimise the LBO limit by changing the fuel properties by increasing its DCN. Therefore, there is potential to improve the DCN value of a given fuel and to measure to what extent the LBO could be optimised. These questions will be explored in the next section.

3.3. Cetane additive effect

There are several ways to adjust the DCN of a given fuel. One approach is to blend high DCN fuel and low DCN fuel, but this will inevitably cause a wide range of changes in other chemical and physical properties, most likely deleteriously. Another approach is to use a cetane additive that can potentially improve a given fuels DCN without significantly impacting any other fuel properties. For this work, a DCN additive was used as a means of potentially augmenting fuels DCN values for selected fuels in this experiment. The value of the DCN was systematically changed by addition of Di-tert-butyl peroxide (DTBP). Di-tert-butyl peroxide (DTBP)—(CH₃)₃COOC(CH₃)₃ is an DCN improver commonly used in diesel due to its low impact on fuel properties. The specific chemical reaction principle of this catalyst can be found here [Reference Daeyup, Shinichi, Hidekazu, Yoshitaka and Makihiko40]. This particular compound has not been approved for use in jet fuel to date. DTBP was chosen as compared to other potential cetane additives such as 2-ethylhexyl nitrate due to the fact that it is made up of completely organic molecule. This was important as the potential presence of nitrogen found in other cetane additives, if used, may have contributed to fuel bound NOx in other aspects of the experimental campaign outside of the scope of this work.

LBO experiments were conducted by mixing two selected fuels with different proportions of DTBP between 0% and 1% by volume. According to the research of Braun-Unkhoff [Reference Braun-Unkhoff, Kathrotia, Rauch and Riedel41], the contents of n-paraffins and i-paraffins are the dominant factors determining the DCN value. Hence, 100% n-paraffin fuel (C1) and over 99% i-paraffin fuel (Banner solvent, NP1014) were selected as the base fuels for adding the DTBP DCN additive.

The DCNs of C1 fuel was determined by ignition quality testing (IQT) analysis conducted at Southwest Research Institute USA using ASTM D6890. For evaluating the impact of DTBP addition on the DCN values of fuels, ASTM D6890 standard was used for DCN measurement at a testing laboratory named Intertek in the UK. In addition, the NJFCP fuels used previously in this study were also sent to the same facility for verification testing to ensure that all the DCN values are measured at the same location. The results showed that the difference is within a reasonable range of 3%, therefore, the DCN data measured by the commissioning agency is considered within acceptable bounds.

The value of the DCN obtained by addition of the additive to two different base fuels is shown in Fig. 8. Different fuel compositions gave different DCN responses to the additive. This is probably due to the fact that DTBP, as a strong oxide, can stimulate different levels of active free radicals when reacting with different molecules in different fuels. For both fuels, there exists a trend of weakly non-linear growth. With an increasing additive proportion, the corresponding growth trend of the DCN slightly decreases. The DCN value of the C1 fuel slowly increases from 16.1 to 23.42 with the addition of up to 1.0% of DTBP by volume. In contrast, the DCN response of the i-paraffin fuel is significantly higher to the additive content. For the Banner solvent fuel, the DCN increases by 26.7 and reaches the value of nearly 100 after the addition of only 0.35% volume of DTBP.

Figure 8. Value of the DCN in response to fuel blending with DTBP.

In addition to the above laboratory blended fuels, three more cetane changed fuels provided under NJFCP program were used in this study. The production method is confidential, but the main properties of the fuels are listed in Table 3. Fuels 31-FP, 44-FP, and 54-FP are named after their respective DCN values, as the values of the DCN for each fuel are 31, 44 and 54, respectively. The fuels have been processed in way that their properties, including density, smoke point, and net heat of combustion have similar values. The main difference is between them is their respective proportion of i-paraffins, n-paraffins and aromatics. The values of the ratio of i-paraffins to n-paraffins and their corresponding DCN show a linear relationship with a high coefficient of determination (R² = 0.98), as shown in Fig. 9. This indicates that a higher ratio of branched alkanes compared to straight chained alkanes results in a high DCN. Braun-Unkhoff et al. [Reference Braun-Unkhoff, Riedel and Wahl42] previously proposed the hypothesis that the DCN is directly affected by the content ratio of i-paraffins and n-paraffins thereby affecting a given fuels’ capacity to resist blowout.

Figure 9. Relationship between the ratio of i-paraffins/n-paraffins and DCN for certain NJFCP fuels.

Combining the previous LBO data of the fuels tested in Section 3.2 of this work with the cetane additive blends (BannerSol and C1 blends) from Section 3.3 and the NJFCP FP fuels, the LBO results were integrated into one graph to investigate the impact of DCN on lean blow off. The Banner Solvent based DCN changed fuels are denoted as BS blends in Fig. 10, while DCN changed C1 are denoted as C1 blends. FP fuel group are the DCN specific formulated fuels according to the NJFCP program with more details given in Table 5. The data of k and ${R^2}$ for each fuel and the degree of response to DCN is show in Table 6.

Figure 10. LBO fuel/air ratio results for all tested fuels with different DCN.

Table 5. The main properties of the NJFCP program DCN changed fuels

Table 6. The degree of response of different fuels to the DCN value

It can be clearly observed that all the different blends tested in this study shows varying degrees of improvement in LBO performance with increased DCN. The FP fuel has the highest slope (k) and Correlation Coefficients (R2) among all the fuels, and there is almost no linear dispersion. According to this sub-group, DCN almost completely independently describes the ability of fuel to resist blowout once significant variables such as density and net heat of combustion have been standardised, irrespective of the group type molecular composition of that fuel. However, as dataset is only for three fuels, it may be premature to generalise that the improvement in the blowout resistance capacity of this group of fuels is the solely due to the influence of DCN alone.

The LBO behaviour of the C1 group potentially shows a similar pattern to the behaviour of FP group fuels. For the C1 fuel group, its correlation and coefficient of determination are the weakest of all fuel groups, including in comparison to the blends previously tested in this work. By considering the lack of n-paraffinic content in C1 fuels, this behaviour is consistent with trends observed and discussed previously. First of all, it can be seen during fuel preparation that the DCN values for C1 do not increases significantly with increasing proportions of the additive. Additionally, after analysing Fig. 9, we know that the increase in DCN value is potentially determined by the ratio of I-paraffins and N-paraffins, with an increased proportion of I-paraffins suggesting a higher DCN. C1, however, is composed of 100% vol/vol of highly branched I-paraffins. From this analysis, there are two possibilities. One is that it is difficult to effectively convert N-paraffins and I-paraffins through a limited amount of catalyst. Second, when the N-paraffins content in the fuel is large, the base number of the denominator is too large, and the ratio change is not obvious. Regardless of the reason, the hypothesis of this work is that it may be impossible to significantly improve and effectively optimise the DCN value and subsequent LBO performance of a given fuel with high I-paraffin content using cetane improving additives alone.

As the control group bs fuel (100% I-paraffins), its LBO performance is second only to FP fuel in response to DCN and far better than pre-tested fuel. It can be considered that the catalyst has successfully optimised the ability of LBO to prevent blow-out. This may be due to the molecular structure of I-paraffins, it is easier to separate the CH^* group into N-paraffins under the strong oxidising action of the catalyst. Or because the N-paraffins base is small, it is easier to increase the ratio. The speculation here are only drawn from the appearance seen in the current experiment, and the specific real reason needs to be obtained through a deeper chemical mechanism analysis.