E-Fuel Blend Operation of Small Industral SI-Engines with Carburetors

To consider the effects of blending gasoline by methanol the relevant specifications were calculated. Due to the quite similar density of gasoline and methanol varied the values for the blends by max. 5 %, so that for dimensioning the fuel tanks the mass-related lower heating value Dhu gives a good approximation for the volume-related energy content, too (Tab. 1). For combustion calculation the elementary analyses of all fuel-blends used, had to be calculated by mass-related superposition of both single fuels. For these calculations the fuel data for gasoline Super E5 were taken from [3]. The elementary analysis of pure methanol M100 was calculated by its molecular formula and the relevant molar masses of carbon, hydrogen and oxygen.

Actually, for measurement of air/fuel-ratio l the atomic fractions C : H : O for the blends used had to be known. As well as it is very time-consuming to change these values during the measurements in lambda-meter and in exhaust-gas analysis, it was decided to leave the mass-fraction of gasoline in both measuring device and calculate the exact air/fuel-ratio for lean and sub-stoichiometric combustions by methods, usable for fuels containing important rates of oxygen, while evaluating the experiments afterwards. Simultaneously, the knowledge of the composition of fuels is basic precondition for interpretation of exhaust-gas consistency.

Due to the half energy content of methanol related to gasoline the lower heating value Dhu decreases linearly by increasing its fraction in fuel blends (Fig. 2a, top). Furthermore, the low heating value limits its energetic shares in the blended fuels. Even with a volumetric rate of 60 % methanol its energetic portion in lower heating value is below 45 % (Fig. 2b, top). With perfect combustion the energy-specific CO2-emission decreases by only 13 %. If the methanol is partly produced by regenerative energy and/or the necessary CO2 for production is captured from industrial exhaust-gases or ambient air the effective CO2-emissions are more reduced. Dependent of the literature used the values for research octane-number of methanol vary between RON = 106 [4] … 108.6 [5] … 114 [6, p. 475].

Fig. 2a

Relevant fuel specifications for blends with 0 … 100 vol.% CH₃OH

Fig. 2b

Relevant fuel specifications for blends with 0 … 100 vol.% CH₃OH

From literature there is no save numeric relation for RON of fuel-blends known. An approximation with volume-share related superposition of the properties of all single components is given by [7]. To check this approximation, the resulting RON a gasoline ‘DEA-Super’ was calculated this way. In [8] its composition (butane 3.8 vol.%, reformate 48,4 vol.%, light crack gasoline 27.6 vol.%, heavy crack gasoline 6.9 vol.%, tertiary butyl alcohol 1.4 vol.%, methyl-tertiary butylether 11.9 vol.%), the RON of all single components with RON = 76 … 116 and the resulting RON = 97.7 of the mixture were published. The result of the volume related superposition is calculated to RON = 98.0, so that the use of this approximation for the fuel blends used seemed to be useful. Based on [7] the RON of the blends M30 and M60 were estimated and shown in Table 1.

With respect to the RON = 95 of gasoline Super E5 it can be proposed that the knocking resistance of all blends will be raised. Beside this assumption the knocking resistance is irrelevant for small industrial engines, as well as the indicated mean effective pressure is quite low and this way the thermal load, too. Occurred by the special demands for starting by hand (e.g. guarantees by Briggs&Stratton) and the worldwide operation with dubious fuel qualities, hot ambient temperatures / high altitudes and inadequate maintenance at site all important suppliers (Briggs&Stratton, Honda) work with low compression ratios of ec = 7.5 … 8.5. Briggs&Stratton requires at least RON = 91 for their engines [9], so that every blend of gasoline Super E5 and methanol fullfills this demand.

For calculative conversion of concentrations in exhaust-gas for components measured by physical techniques according to ISO8178 in dry gas, generally for CO, CO2, SO2, O2 and depending for NOx, into output-specific mass-flows the coefficient KExh as relation between concentrations fExh_dry in dry and fExh_wet in wet exhaust-gas is necessary. As well as the water-steam concentration in exhaust-gas is not measured, KExh must be determined by combustion calculations, separately for lean (perfect combustion) and sub-stoichiometric conditions.

While combustion calculations for oxygen-rich mixtures are quite easy, the calculation of sub-stoichiometric exhaust-gas components becomes more complex (Fig. 3a). In these investigations the possible components in exhaust-gas were defined by only 5 components (reaction products CO2, CO, H2O, H2 and inert N2). The available content of oxygen was defined by fuel composition CxHyOz and air/fuel-ration l. With 3 atomic balances for carbon, hydrogen and oxygen and a nominal oxygen-share of 21 vol.% in dry air the relevant reaction equations were formed.

Fig. 3a

Calculation of sub-stoichiometric combustion

Watergas-shift reaction with balance factor KC = 3.4 for frozen conditions [10] delivered the proportion of reaction products in wet exhaust-gas. A system of 6 linear equations for 6 unknown variables is the result (Fig. 3a, down), which was solved analytically. Related operations were carried out with different air/fuel-ratios l for all fuels and blends used and diagramed in Fig. 3b.

Fig. 3b

Conversion of exhaust-gas concentration

To avoid these complex calculations for each operational point measured, a regression for operation with normal gasoline Super E5 was developed. As carried out equivalently in former investigations with gasoline/ethanol-mixtures [11] the influence of rising methanol fractions in fuels were considered by an additive regression-correction DKExh = f(volume-fraction fCH₃OH, l) to KExh for combustion of gasoline Super E5 according to Eq. 1 … Eq. 3.

For post-processing of the experimental data the l-formular by Brettschneider [12] was used, which is also already available in the exhaust-gas analysis equipment.

For engine operation with gasoline Super E5: 1KExh−E5={ +0.3837λ3−1.2195λ2+1.2797λ1+0.6796for0.6≤λ<1.0−0.1364λ3+0.5788λ2−0.8866λ1+1.5675for1.0≤λ≤1.4{K_{Ex{h_ - }E5}} = \left\{ {\matrix{ { + 0.3837{\lambda ^3} - 1.2195{\lambda ^2} + 1.2797{\lambda ^1} + 0.6796{\rm{ }}for{\rm{ }}0.6 \le \lambda < 1.0} \hfill \cr { - 0.1364{\lambda ^3} + 0.5788{\lambda ^2} - 0.8866{\lambda ^1} + 1.5675{\rm{ }}for{\rm{ }}1.0 \le \lambda \le 1.4} \hfill \cr } } \right.

Additive regression-correction for operation with methanolshares: 2KExhMxx=KExhE5+0.12192·(ϕCH3OH[vol. %]/100)2+0.0547·(ϕCH3OH[vol. %]/100)1+0.008+ΔKExh(λ){K_{Ex{h_{Mxx}}}} = {K_{Ex{h_{E5}}}} + 0.12192\cdot\;{\left( {{\phi _{CH3OH}}[{\rm{ }}vol{\rm{. }}\% ]/100} \right)^2} + 0.0547\cdot\;{\left( {{\phi _{CH3OH}}[{\rm{ }}vol{\rm{. }}\% ]/100} \right)^1} + 0.008 + \Delta {{\rm{K}}_{Exh}}(\lambda ) with l-correction: 3ΔKExh(λ)={ 0for0.6≤λ<1.0(λ−1)0.7·(ϕCH3OH[ vol. %]/100)2for1.0≤λ≤1.5\Delta {{\rm{K}}_{Exh}}(\lambda ) = \left\{ {\matrix{ 0 \hfill & {{\rm{ }}for{\rm{ }}0.6 \le \lambda < 1.0} \hfill \cr {{{(\lambda - 1)}^{0.7}}\;\cdot\;{{\left( {{\phi _{{\rm{CH}}3{\rm{OH}}}}[{\rm{ vol}}{\rm{. }}\% ]{/_{100}}} \right)}^2}} \hfill & {{\rm{ }}for{\rm{ }}1.0 \le \lambda \le 1.5} \hfill \cr } } \right.

Testbed was originally concepted for output tests of two-stroke engines [13]. Continuous problems with stability of combustion processes occurred by membrane-type carburetor while temporary operation, vibration damages at the ridged mounted engine housing and hazard for emission measurement devices avoided reasonable operation. Further the clutch supplier did not ensure the operational safety of its chosen model at engine speeds of 9300 rpm. A cage for containment safety in case of clutch failure had to be installed.

That’s why the testbed was converted to performance-tests of small 4-stroke SI-engines as used for snow-throwers, portable generators, cleaning equipment or water-pumps. With the engine the exchange of clutch was necessary, too. With respect of output, spare part supply and price a Briggs&Stratton engine Intek Pro 206 was chosen (Fig. 4a). Main engine data were collected from the website of a supplier. Length of con-rod was measured at an engine grabbed out. Compression ratio was investigated while installing the sensor for cylinder-pressure indication [14].

Fig. 4a

Test-engine Briggs & Stratton Intek Pro 206

Valve timing was measured with inductive lift-sensors several times while turning the engine by hand and averaged (Fig. 4b, down). Remarkably is the second opening of exhaust-valve during compression-stroke for de-compression in hand-start operation. After engine-start this device is disabled automatically by a mechanical ball-head-switch.

Fig. 4b

Performance at site and valve lift of test-engine

As well as the engines performance was given by manufacturer as gross-output (= brutto) according to SAE standard J1940 without attached auxiliary devices (silencer, cooling fan, ignition magnet and generator) no reliable rating value for practical use was known. Due to that fact own performance tests were carried out at testbed (Fig. 4b, top).

Engine performance was considered at site (approx. 22 deg C/1000 mbar) with all auxiliary devices necessary [15]. Output measured at testbed is not necessarily performance-maximum of this engine type as well as the supply voltage for the electric retarder was limited to BES = 29 V. Torque setting of the retarder was interrupted at 80 % of maximum because with settings of the speed-governor lower than 70 % no stable engine operation was possible anymore.

Remarkable is the strong speed droop, the mechanical speed-governor provides. While a speed drop of Dn/nN = +3 % between full and zero load is standard for large diesel-gensets this SI-engine does not reach its nominal speed nN = 3600 rpm already at torque settings of 40 %.

Even without any experimental investigations in methanol operation and with the experiences in ethanol-operation [11] it became clear, that mechanical carburetors must be adapted for the use of oxygen-rich fuels, as well as the air/fuel-ratio l would reach the lean limit for misfiring. Already in earlier projects concepts for adjusting the air/fuel-ratio for SI-engines with mechanical carburetor were investigated.

Based on the basic layout four different strategies were identified. Without changing the sub-stoichiometric sizing of the fuel’s main-jet (= calibrated main fuel metering orifice [16]) the easiest way seem to be systematic addition of secondary air (“fault-air”) after throttle of the carburetor (Fig. 5a, top).

Fig. 5a

Strategies for λ-adjustment with sub-stoichiometric basic layout

Pressure in intake manifold of natural aspirated SI-engines is below ambient pressure, so that no flow-supporting device would be necessary. With an easy metering-valve the amount of secondary air can be adjusted from basic air/fuel-ratio l = 0.7 to the lean limit of ignition. Experimental considerations showed that the pressure pIntake in intake manifold is not sufficiently negative for acceptable range of l-variation [14].

A second way for in-operation variation of air/fuel-ratio with sub-stoichiometric basic layout would be the use of a pulsed solenoid valve in fuel line before carburetor (Fig. 5a, down). Due to the non-available solenoid valve and its electronic amplifier for pulsing this option was refused. An alternative way was found while continued literature studies [17] with injecting air bubbles into the fuel line before carburetor [18]. This idea was rejected due to apprehension regarding save fuel supply for engines with fuel delivery by gravity as normally used for small SI-engines.

With lean basic layout the idea of a booster fuel-pump was tested. Even with a booster-pressure of 730 mbar no relevant change in air/fuel-ratio was measured (Fig. 5b, top). The addition of this fuel pump did not influence the operation of the float chamber.

Fig. 5b.

Strategies for λ-adjustment with lean basic layout

With these preliminary studies [14] [17] [19] [20] the old idea of combining lean basic layout by diameter-reduced fuel-jets and adjustable choking device for combustion air [21] [22] was reanimated (Fig. 5b, down). Based on the diameter in standard main jet, several jets were manufactured with smaller diameters and tested.

Oxygen content in exhaust-gas is measured by a separate lambda-sensor Bosch LSU4.9. Its signal is converted into an analogue voltage of 0 … 5 V and transmitted to a microcontroller, which compares the measured l with the given target value. According to this control deviation the necessary position of choke (Pos_Choke) is calculated by the self-developed PID-controller software of microcontroller and adjusted by a small servomotor, driving the shaft of the choke plate.

With the finally chosen setup of main jet and flow area of choke lambda-variation in the range of l = 0.7 … 1.2 became possible in gasoline-operation.

With a test of fuel-switch from normal gasoline Super E5 to fuelblend M60 with a content of 60 vol.% methanol the assumption of necessary in-operation adaptation of carburetor setting was confirmed experimentally (Fig. 6a). Air/fuel-ratio raised from l = 0.75 to l = 1.1. In combination of a carburetor layout of l ≈ 0.85 for gasolineuse and higher methanol additions a faultless operation was not ensured anymore.

Fig. 6a.

Fuel-switch without adaptations of carburetor

Fig. 6b shows a fuel-switch from gasoline Super E5 to methanol-blend M30 with enabled lambda-control. Air/fuel ratio is set to a value of l = 0.86, which is normal for these industrial SI-engines at part load. As shown in Fig. 6a there is a time delay of approx. 60 s of entering the blend into the float chamber in relevant quantity. Choke position increases immediately because of the lower stoichiometric air requirements of the methanol content. Air/fuel ratio is quite constant with tolerances lower than Dl = 0.05, occurred by instabilities in combustion. Because of higher evaporation enthalpy the exhaust gas temperature becomes with DtExh = -10 K slightly lower than in gasoline combustion. With constant sub-stoichiometric air/fuel ratio the lambda-effect of hotter combustion while converging to l = 1 is excluded.

Fig. 6b

Fuel-switch with enabled λ-control

Testbed is equipped with standard pressure sensors and temperature probes (NiCr-Ni, Pt100) for integral operation values of the engine and all necessary monitoring sensors for safe operation. Selection of relevant test equipment is displayed in Fig. 7.

Fig. 7.

Relevant test equipment

Engine itself is equipped with a mechanical speed-governor for engine speed nM acting the throttle of carburetor. Engine speed nM is detected by an optical sensor, observing a reflecting mark at the surface of the clutch. Position of throttle Pos_Throttle is detected by rotary potentiometer as well as the setting of choke Pos_Choke.

Additionally, the pressure inside intake manifold pIntake was taken to exclude influences of fluctuating ambient pressures. Ambient conditions (t_Amb, p_Amb, f_Amb) were measured inside an air-container, which can be supplied by an aircondition.

Engine tests were carried out on 3 different days within 4 weeks. Fuel-mixtures were prepared with pure methanol (99.9 vol. %) and commercial gasoline Super E5 (JET) with an ethanol content of max. 5 vol.%. Mixtures were blended according to calculated mass fractions and filled for the tests into tank 2 while pure gasoline was stored in tank 1. For all measurements a standard output Pe = 1.4 kW @ 3200 rpm (break torque Md = 4.12 Nm) was used. The engine was started in gasoline operation and driven up to normal lube-oil temperatures tLube = 91 °C. After the heating-up procedure tests with variations of fuel started. Not consumed fuel-blends were removed from tank 2, while the engine was re-switched to gasoline operation from tank 1 and replaced by the next blend for testing. On each testing day reference measurements in gasoline operation (Super E5, Ref. 1, Ref. 2 in Fig. 8, Fig. 9a–c) were carried out.

Fig. 8.

Fault of Lambda-Meter without adjusting fuel-parameters

Fig. 9a.

Test results – Operational values

Fig. 9b

Test results – Standard deviation of peak pressure

Fig. 9c

Test results – Specific exhaust gas emissions

Each testing-day was ended by long-term gasoline operation from tank 2 for purging whole fuel system, as compatibility with materials of tank 2, fuel-line, sealings, jets, needle-valve, float and its chamber was not clear and became subject of additional considerations.

Relevant results of experimental investigations with the gasoline Super E5, the blends M30 and M60 as well as pure Methanol M100 are shown in Fig. 9a … c. Diagrammed in Fig. 9a are the operational values fuel consumption m’_Fuel, the specific fuel consumption b_e42.7 related to standard lower heat value of Dh_u = 42.7 MJ/kg and that way reciprocal to the efficiency and the exhaust-gas temperature t_Exh versus the air/fuel-ratio l.

Air/fuel-ratio was measured with the lambda-meter by ETAS. Values of fuel composition H/C and O/C were set constant to the values for Super E5 and not changed during engine operation. Afterwards the results of lambda-meter were counterchecked by values calculated from complete exhaust-gas emissions by Brettschneider-regression, usable also for oxygen-rich fuels [12], with real fuel-composition. For single measured points additionally analytical calculation with all components in sub-stoichiometric combustion were carried out according to watergas-shift reaction with the balance constant K_C = 3.4. The failure of ETAS-device was approx. 1 %, related to Brettschneider-regression (Fig. 8) and 2 % related to combustion calculation within the interval of l = 0.7 … 1.0. Diagramming versus ETAS display without changing fuel-parameters seemed to be acceptable.

Fuel consumption m’_Fuel rises by using e-fuels due to poor energy content compared to standard hydrocarbon-fuels. Tank volume capacity must be enlarged by approx. 75 % to ensure same operating range (Fig. 9a). Surprising is the energy content-related specific fuel consumption b_e42.7. As long as the methanol volumeshare is below 30 % (M30) the efficiency is the same as for gasoline operation. This result can be explained by cylinder-indication (Fig. 10). Up to methanol volume-share to 30 % (M30) the heat release is nearly unchanged compared to standard operation with gasoline Super E5.

Fig. 10.

Test results – Pressure indications at λ ≈ 0.85

With higher methanol contents efficiency increases by three effects. First, even at same global air/fuel-ratio the oxygen content in fuel blends leads to more complete combustion due to local O₂-concentration, provided by the methanol. Concentration of unburnt components CO and THC in exhaust-gas decreases significantly (Fig. 9b). Second, due to higher local O₂-content in cylinder gas the combustion gets faster, what will be demonstrated by the cylinder-indications in Fig. 10, too. Oxygen contents in fuels enlarge the velocity of the flames during combustion [23]. The combustion process is significantly advanced by approx. 4 °CA in operation with M60 and M100. The same effect was already observed by the tests with ethanol-shares in gasoline [11].

Third, the higher evaporation enthalpie of methanol leads to lower gas temperatures inside the cylinder, what lowers the specific isochoric heat capacity of the cylinder-gases. The influence of this third effect is quite small. Cycle calculations with completly constant heat capacities at a large dual-fuel engine S.E.M.T. Pielstick 18PC2-5 DFC showed efficiency improvements lower than 5 % [24]. With pure methanol M100 the advantage in energy consumption is approx. 20 %, what does not compensate the lower content of chemical energy.

Exhaust-gas temperature becomes lower with raising methanol shares according to the standard evaporation enthalpy of Dh_{V_CH3OH} = 1100 kJ/kg @ 1013 mbar [25]. The cylinder gas mass is cooled in inlet manifold and inside the cylinder during intake stroke. Additionally, the fuel-mass burnt per cycle is approx. 75 % higher in methanol-operation, so that the temperature at the start of compression is lower than with gasoline (Dh_{V_Gasoline} = 380 … 500 kJ/kg @ 1013 mbar [25]). The whole temperature level of the engine’s process is dominated by the gas temperature at the start of compression stroke. This way exhaust-gas temperature decreases in e-fuel operation.

With lower gas temperatures inside the cylinder instable combustions were suspected. Fig. 9b shows the standard deviation of indicated peak pressures inside the cylinder. Up to volumic shares in fuel of 100 % methanol the standard deviations in peak pressures are within the tolerance of operation with pure gasoline, measured at the 3 different test-days (Super E5, Ref. 1 and Ref. 2).

Diagrammed in Fig. 9c are the mass-related specific emissions for carbon-monoxide CO, residual hydrocarbons THC from unburnt fuel and Nitrous oxides NOx. As well as the concentrations for NOx and CO were measured in dry exhaust gas, the measured values had to be corrected to wet conditions in un-cooled exhaust-gas by factor KExh according to Fig. 3b. Additionally, the specific NOx emissions must be corrected depending on the water-content inside the combustion air at site according to ISO 8178.

As mentioned above, the local availability of O2 close to the flame during combustion decreases the content of unburned components CO and THC in exhaust-gas. Significantly is the influence on specific NOx emissions. For the formation of NOx are high local gas-temperatures, sufficient local concentration of O₂ and time are necessary.

At the same operating point of the engine the influence of time is excluded. With constant rating and engine speed in that operating point a shift in ignition timing can be excluded, too. While higher O2-concentration leads to higher NOx-level the formation of thermal NOx would be decreased by lower gas-temperatures. It becomes clear that the low gas-temperatures in blend-operation and with pure methanol dominate this effect as well as the level for NOx-emissions is decreased significantly.

Cylinder indication was carried out with 100 cycles per measuring point. While post-processing it became clear, that especially at higher air/fuel-ratios with unstable combustions the recording of more cycle would have improved the results.

Based on the pressure indications carried out a slight influence on the cylinder peak pressure pmax can be detected (Fig. 10). Due to advanced combustion detected by the crank-angle MBF50 of half chemical energy released the cylinder peak pressure pmax was enlarged by less than 1 bar. Compared to the cycle-to-cycle deviation of more than 6 bar this small increase is irrelevant for the stability of crank-drive and cylinder unit.

Heat rating of the sparkplug was not changed for operation with methanol/gasoline-blends. Despite concerns regarding the lower gas-temperature inside the combustion chamber and that way forming of deposits at the electrodes the original sparkplug was not replaced. Even in operation with pure methanol no anomalies in combustion cycles were detected.

Despite the lower gas temperatures inside the cylinder at compression start due to the high enthalpy for evaporation of methanol no abnormal deposits were found during the optical inspection at the electrodes of the sparkplug (Fig. 11a, photo). No changes of sparkplug configuration seemed to be necessary.

Fig. 11a

Sparkplug after methanol operation

Combustion chamber was inspected after the tests by endoscope (Fig. 11b). Deposits of lube-oil coke were detected at piston head. No cleaning effect of methanol could be observed. The upper flank of first piston ring seemed to be corroded, what is a possible effect of sub-stoichiometric combustion of alcohols, forming formic acid. In consequence to avoid further damage, sparkplug was removed after every day’s test-run, cylinder was filled by fresh lubeoil and engine was turned by hand to distribute the lube-oil at the liners surface.

Fig. 11b

Endoscopy of combustion chamber [26]

Fresh and used lube-oil after operation with blends and pure methanol were analyzed by ATR-spectroscopy (Fig. 12). Both spectra seemed to be quite equal. Methanol could not be detected in the ATR-spectrum. The characteristic band of the C-O stretching vibration and the OH deformation vibration at 1021 cm⁻¹ is not present in the spectrum of the used lube oil. But the engine was operated after all tests with gasoline over a time-period of approx. 1 hr. for purging all fuel equipment. Possibly in this way the contamination of lube-oil with methanol could be eliminated by evaporation.

Fig. 12.

Results of analysis of lubrication-oil

As a hint for carbonyl groups differences in spectrum were detected at the band of 1739 cm-1. The carbonyl groups can be occurred by formaldehyde or formic acid. Formaldehyde seems to be implausible due to its evaporation behavior, so that probably formic acid, produced in sub-stoichiometric combustion of methanol, contaminated the lube-oil.

Additionally X-ray fluorescence examinations showed contamination with the element iron, what could be occurred by the influence of acid at cylinder liner and piston rings as described above. This observation supports the theory of acid input into crankcase. No TBN were investigated.

Chemical resistance of different materials used in engine components were tested with pure methanol and its blends with gasoline. Besides optical evaluations mechanical tests were carried out, too. In fact, all materials tested and used in components of the engine were chemically attacked by these fuels (Fig. 13).

Fig. 13.

Results of material tests [26]