Search

This research aims to predict knock occurrences by deep learning using in-cylinder pressure history from experiments and to elucidate the period in pressure history that is most important for knock prediction. Supervised deep learning was conducted using in-cylinder pressure history as an input and the presence or absence of knock in each cycle as a label. The learning process was conducted with and without cost-sensitive approaches to examine the influence of an imbalance in the numbers of knock and non-knock cycles. Without the cost-sensitive approach, the prediction accuracy exceeded 90% and both the precision and the recall were about 70%. In addition, the trade-off between precision and recall could be controlled by adjusting the weights of knock and non-knock cycles in the cost-sensitive approach. Meanwhile, it was found that including the pressure history of the previous cycle did not influence the classification accuracy, suggesting little relationship between the combustion behavior of the previous cycle and knock occurrence in the following cycle. Moreover, learning the pressure history up to 10° CA before a knock improved the classification accuracy, whereas learning it within 10° CA before a knock did not noticeably affect the accuracy. Finally, deep learning was conducted using data, including multiple operating conditions. The present study revealed that deep learning can effectively predict knock occurrences using in-cylinder pressure history.

This study investigated experimentally a response of premixed Bunsen flame to transient change in a fuel composition and a method to control the variation. Fuel mixtures were composed of methane, ethane and propane. The volumetric fraction of methane was varied from 92 % to 76 % by 4 % and the ethane/propane ratio was 5. Transition time of the fuel composition was 1, 5, 10 and 20 s. Flow velocity and the equivalence ratio were 0.8 m/s and 0.85 respectively before the transition. Channel length L between the merging point of fuels and the point of air was 0, 500, 1000, 1500 mm. The motion of flame was recorded by a high-speed video camera at 60 fps. The experiment was conducted under the fuel flow rate constant and the equivalence ratio constant conditions. In the former condition, the flame height varied almost linearly not depending on the conditions. In the latter condition, variation of the steady flame height was smaller than the former condition. However, the unsteady flame height exceeded or fell below temporarily the steady flame height. This is due to the time lag between the transmission of fuel flow rate variation and fuel compositions variation. Therefore, considering this time lag, the experiment was conducted with a system which had a density meter to detect the fuel composition variation. The system controlled automatically the fuel flow rate. At long L and short transition time, there were some peaks in the flame height variation. It is considered that this is caused by the difference between the actual density and the output of the density meter due to Taylor dispersion and response time of the density meter. However, the flame height variation was smaller than the variation without control, and particularly at long transition time, the variation was suppressed sufficiently.

The characteristics of a medical heat and moisture exchanger with filter (HMEF) that contained polyurethane sponges and CaCl2 were investigated experimentally and model analysis was performed. Variations in the temperature and humidity were measured. Saturated air at 37 °C simulating exhalation and dry air at 26 °C simulating inhalation were fed alternatively through the HMEF in opposite directions. The period was set as 256 s, 64 s, 16 s and 4 s. When the period was long (256 and 64 s), the polyurethane sponges and CaCl2 played an important role in condensation and evaporation of water and the amplitudes of the temperature variation was so large that it overshot the input condition. When the period was shortened (4 s), the amplitudes of the temperature and humidity variations decreased and the relative humidity of the patient side of the HMEF gradually reached that of the exhalation conditions. This suggests that the characteristic times of condensation and evaporation were longer than that of the respiration (τb = 4 s). In other words, the unsteady effect is important for the condensation and vaporization mechanism of HMEF. The mechanism was supported by the model analysis. In addition, comparing the results of the experiments and model analysis elucidated the effect of the connecting pipe, which corresponds to the dead space effect in clinical use.

In order to improve thermal efficiency of automobile engine, it is necessary to reduce the cooling loss. A method for analyzing local cooling loss (heat flux) using coaxial type thin film temperature sensor has been developed. Recently, results of heat flux analysis using the sensor composed of thermocouple materials standardized based on ANSI and JIS have been reported, but structure and measurement accuracy of its sensor are not clear enough. In this study, the measurement accuracy of sensor A composed of K type materials and sensor B whose body is combustion chamber material (aluminum alloy) was compared using numerical analysis (two-dimensional coordinate heat conduction equation when the sensor is embedded in combustion chamber). As a result, it was important to use same material as combustion chamber for sensor body to measure wall surface instantaneous temperature with high accuracy. Additionally, results of one-dimensional (axial direction) heat flux analysis when the material of the combustion chamber and the material of the sensor body are different had error, it was considered to effect of heat flow (temperature difference) in radial direction due to different thermal conductivity. Therefore, it was confirmed that newly proposed two-dimensional heat flux analysis method using temperature distribution in axial and radial direction can analyze more accurately than one-dimension. These numerical analysis results showed same tendency as measured verifications results.

Flame height variations of a laminar stagnating premixed flame with gas component transition between fuel mixture A (methane:92%, ethane:6.7%, propane:1.3%) and B(methane:84%, ethane:13.3%, propane:2.7%), and between A and C (methane:76%, ethane:20.0%, propane:4.0%) are discussed. The estimation of the flame speed and the flame temperature of fuel mixtures A, B and C by the Chemkin-Pro show almost the same values for all fuel mixtures. In experiments, the transition between A and B and between A and C were performed. The transition time changes from 1 s to 10 s, keeping flow velocity and equivalence ratio constant at 0.8 m/s and 0.7 respectively. The flame position was measured by using images taken by a high speed camera with 125 fps. Two transition patterns are examined, keeping the total flow rate of premixed mixture constant. One is the constant flow rate transition, which changes the gas component, keeping the fuel mixture flow rate constant. The other is the constant equivalence ratio transition, keeping the equivalence ratio constant. In the case of the constant flow rate transition, the equivalence ratio changes which results in the variation in the flame speed, the flame temperature and the flame height. It means that the combustion state changes with the transition. On the other hand, in the case of the constant equivalence ratio transition, the flame speed, the flame temperature and the flame height keep almost the same state, which means that the gas component can be changed, keeping the combustion state the same.

The effect of feed flow rate of hydrogen mixture on the H2 permeation for a flat sheet Pd/Ag membrane with stagnating flow at the upstream side was investigated experimentally and theoretically. A 77wt.% Pd/23wt.% Ag flat sheet membrane with 25µm thickness and 0.02m diameter was used. The permeation rate of H2 was investigated under various feed flow rates (1.489 × 10-5-2.976 × 10-4mol/s), for pressures of 0.20-0.30MPa and reference membrane temperatures of 523-723K. Experimental results demonstrated that when the feed flow rate is decreased, the H2 permeation rate decreases. This is supposed to be due to the phenomena of hydrogen concentration decrease at the membrane surface of the upstream side, as a result of the effect of H2 permeation itself. When a theoretical equation that takes into account the effect of H2 permeation is used, the H2 permeation mole flux can be predicted quantitatively by using the concentration of H2 of the feed mixture. This shows that the diffusive transport effect plays an important role as well as the convective transport effect when determining H2 concentration at the membrane surface. In addition, the normalization of the theoretical results shows that the trend of the decrease in the H2 permeation mole flux with respect to the feed mole flux follows the first order lag function, regardless of the inlet H2 partial pressures.

Responses of conical premixed methane/air mixture flames with equivalence ratio oscillations were numerically investigated at three different oscillation frequency regimes; low, medium and high (10, 50 and 150 s-1) under a rich condition. One-step and two-step reaction mechanism have been investigated. The pre-exponential factors for the one-step and two-step reaction mechanism were calibrated at minimum and maximum equivalence ratio of this study, Ø= 1.1 and 1.5, by comparing the flame length of the numerical results with the experimental results. The two-step reaction mechanism predicted well the variation in the flame shape under a steady state condition with equivalence ratio variations while the one-step reaction mechanism failed to predict the variation in the flame shape especially under the near rich flammability limit condition. This is because of the CO formation inclusion in the two-step reaction mechanism. The effects of the equivalence ratio oscillation frequency towards flame dynamics were discussed. The amplitude of the flame tip movement attenuates following the attenuation of the equivalence ratio oscillation towards the downstream direction at high oscillation frequency. The dynamic response of the flame tip for low, medium and high oscillation frequency regimes shows interesting behavior. The quasi-steady manner of the flame tip movement was observed at the low frequency regime. In the medium and high frequency regime cases, we found that the attenuation of the flame tip motion is affected by the wrinkling of the flame surface in addition to the attenuation of the equivalence ratio oscillation amplitude. Moreover, an increase in the number of wrinkles as the equivalence ratio oscillation frequency is increased induces large attenuation of the flame tip oscillation amplitude. Furthermore, we found that the period of the equivalence ratio oscillation, T and the convective transport time, τ control the flame response in that T/τ, is a controlling parameter of the conical flame response in quasi-steady (T/τ ≥ 1.0) or unsteady (T/τ ≤ 1.0) manner.

Performance of self-sustaining methanol auto-thermal reforming (ATR) was investigated experimentally in order to elucidate a reforming reaction mechanism and a condition required for high purity H2 production for compact reformer. The reformer consists of vaporizing and reforming sections in a single unit. The exothermic oxidation and endothermic steam reforming (STR) take place simultaneously in the reforming section. The reforming section is surrounded by the vaporizing section and then the heat for vaporization is supplied from the reforming section. Two types of exothermic oxidation reaction were investigated as the heat source for STR; one is a partial oxidation (POX) and the other is a total oxidation (TOX). CuO/ZnO/Al2O3 catalyst and Pt/Al2O3 catalyst were used for STR and POX, respectively. While, only CuO/ZnO/Al2O3 catalyst was needed for TOX because TOX took place when fuel and oxygen were supplied to the CuO/ZnO/Al2O3 catalyst. Experiments were investigated in the range of oxygen/carbon ratio (O/C ratio) 0.1-1.5, steam/carbon ratio (S/C ratio) 1.0-3.0 and N2 mole ratio 79-50 % in oxidizer. The results showed that the H2 formation reached maximum at around O/C=0.4 in both STR/POX and STR/TOX cases in the present study. When O/C ratio is decreased from 0.4, heat formation by the oxidation reactions decreases and is insufficient to reform residual CH3OH by STR. As a result, H2 formation and the methanol conversion ratio decrease. When O/C ratio is increased from 0.4, the H2 formation decreases, because methanol is consumed with the excess O2 by TOX and CH3OH for STR decreases. After all, O/C=0.4 gives an appropriate balance of heat supply and methanol for H2 production. These results elucidate that the reaction rate of oxidation reactions, POX and TOX, is much faster than that of STR. In other words, methanol is first consumed by the oxidation reaction and the residual methanol is used for STR. For S/C ratio, H2 formation is decreased in the higher S/C ratio. N2 mole ratio in oxidizer has few influence over the reforming gas. The chemical equilibrium calculations support the experimental results.

The performance of high purity hydrogen production from methanol for a compact steam reformer with a hydrogen purification membrane was investigated experimentally. A 77 wt.% Pd/23 wt.% Ag membrane with 25µm thickness and CuO/ZnO/ Al2O3 catalyst were used. Heating was performed by a Bunsen type burner using City Gas 13A. The methanol reforming and purification of H2 were investigated at different reference catalyst zone temperatures (589-689K), pressures at the retentate side (0.2-0.5MPa), steam to methanol(S/C) ratios (0.8-1.6) and reactant flow rates (1.7 ×10-4 to 4.4×10-4 mol/s). The results show that at high reference temperature, high pressure and certain points of the reactant flow rate, the maximum hydrogen permeation rate is obtained when the S/C ratio is around 1. The modified Sieverts’ equation which considers the decrease in H2 concentration at the membrane surface, was proposed. The experimental result was lower than the permeation rate estimated by the modified Sieverts’ equation, which is probably caused by the adsorption of non-H2 species during permeation. It is further demonstrated that the modified Sieverts’ equation is able to estimate a more reasonable hydrogen permeation rate in comparison to the estimation by the ordinary Sieverts’ equation. In addition, it is shown that the compact methanol steam reformer with a Pd/Ag membrane is able to produce high purity hydrogen with very low CO concentration, which fulfills the Polymer Electrolyte Fuel Cell (PEFC) requirement (

Responses of the premixed methane/air mixture flames under equivalence ratio oscillations were numerically investigated assuming axi-symmetric stagnation flow fields. The flame motion was numerically investigated at three different oscillation frequencies (10, 20 and 50 Hz) and with three oscillation cases namely: lean case, rich case and lean rich crossover case. Methane/air mixture with equivalence ratio oscillation was issued from the burner exit with 1.0 m/s uniform velocity profile. The effects of frequency and amplitude of the equivalence ratio oscillation were discussed. The amplitude of the equivalence ratio oscillation is attenuated between the burner exit and the upstream edge of the preheat zone. The attenuation is much significant for higher frequency. The amplitude of the flame temperature oscillation attenuates following the attenuation of the equivalence ratio oscillation. The flame location makes the closed cycle around the flame location of correspond equivalence ratio in the steady state condition. The formation of the cycle can be explained by the back support effect. It was further demonstrated that, the back support effect influences the dynamic response of the flame location, in that, the direction of the cycles of the dynamic response in the lean case and the rich case are different. Furthermore, the time variation of the flame location plays a significant effect to the flame displacement speed.

Responses of stagnating laminar methane/air premixed flames under fuel concentration oscillation, i.c., equivalence ratio oscillation, were numerically investigated using a one-step overall reaction mechanism and a four-step reaction mechanism that included CO and H2 formation. The flame motion was numerically investigated for three different oscillation cases namely: lean, rich and lean-rich crossover case. Methane/air mixtures with sinusoidal equivalence ratio oscillations were issued from the burner exit with uniform 1.0 m/s velocity profiles. In the steady state condition, the one-step overall reaction mechanism and the four-step reaction mechanism had nearly the same characteristics in the lean region, while in the rich region variations in characteristics such as flame location and flame displacement speed for the four-step model were much more significant than those for one-step model. When the equivalence ratio was oscillated, the flame location oscillated. The amplitude of the flame location oscillation did not change with the equivalence ratio oscillation when the frequency of equivalence ratio oscillation was less than 8Hz, while it decreased monotonically when it exceeded 8 Hz. Here 8 Hz corresponds to a Strouhal number (St) of unity. Thus, this result indicates that the flame was in a quasi-steady state when St1.0. The variation in flame location and the flame displacement speed did not follow those for the steady state condition and made a limit cycles. This was due to the back support effect. The cycles were significantly inclined at higher frequencies. In the lean condition, the limit cycle was inclined similarly for both the one-step and four-step reaction mechanisms. In the rich condition, however, the limit cycle for the four-step reaction was more inclined than that for the one-step reaction. These results show that the formation of CO and H2 played an important role in the rich condition.

Reducing the cycle-to-cycle variation in an engine is a key to expanding the operational limits of high-efficiency engines such as lean-burn engines. The cycle-to-cycle variation of the engine output is caused by unstable combustion and is especially induced by the cycle-to-cycle variation of the in-cylinder flow. This investigation aims to clarify how the cycleto- cycle variation of in-cylinder flow develops during intake and compression strokes. To achieve this aim, 12 kHz timeresolved particle image velocimetry measurements were conducted on the symmetrical vertical plane in a four-valve, four-stroke optically accessible engine. This study focuses on the relationship between the flow at the plug position at the ignition timing and the flow distribution in the intake and compression stroke. Correlation map analysis reveals this relationship, and some flow distributions are extracted. The fluctuations in the flow distribution induce the cycle-to-cycle variation of the flow direction at the plug position at the ignition timing, the position of the stagnation plane between the intake flow from the intake valves and the upper side of the tumble vortex, the strength of the blowback flow into the intake port generated by the Miller cycle, and the position of the tumble vortex in the late compression stroke. The flow distributions extracted by the correlation map analysis clearly change between cycles. These results also suggest that correlation map analysis is an effective way to elucidate the factors that induce the cycle-to-cycle variation of the incylinder flow. [ABSTRACT FROM AUTHOR]

In order to improve thermal efficiency of spark ignition engine under super-lean burn conditions (excess air ratio l'2:0), thermal stratification technique by in-cylinder water injection toward piston surface (stratified water insulated combustion architecture), in which low temperature water vapor layer is formed on the surface, is proposed. From the water spray visualization using the optically accessible engine, injected water is distributed near the piston top surface at ignition timing and thermal stratification can be achieved in the case that water injection timing (SOIw) is set at earlier stage of a compression stroke. In addition, heat flux on the piston surface measured at the same time is reduced by water injection. The 0.5-L single-cylinder engine test at l = 2.0 at a constant ignition timing also shows that water injection at earlier stage of a compression stroke makes it possible to mitigate knock without significant increase in combustion instability. On the other hand, part of water is distributed near the spark plug at ignition timing with the water injection at SOIw =260 °ATDC, resulting in unstable combustion. In addition, the engine test at l = 2.0 and water/fuel ratio (W/F) = 18% shows that knock mitigation by water injection enables spark advance and following combustion enhancement. As a result, combustion period becomes short and cooling loss decreases, followed by the 1.0-pt improvement of gross indicated thermal efficiency. Moreover, the engine can be operated at minimum spark advance for best torque by increasing W/F up to 35%. Finally, stratified water insulated combustion architecture concept is applied at l = 1.9 with a higher compression ratio of 17, showing that water injection at SOIw=2120 °ATDC and W/F = 50% enables minimum spark advance for best torque operation, and remarkably high gross indicated that thermal efficiency of 52.63% can be achieved with a sufficiently low knock level and coefficient of variation of indicated mean effective pressure. [ABSTRACT FROM AUTHOR]

This experimental study explored the influence of global/local flame behaviors on direct combustion noise produced by a lean-premixed gaseous H2/air low-swirl turbulent jet flame, with a focus on the mechanism resulting in characteristic peaks in combustion noise spectra. Ten kilohertz OH∗ chemiluminescence and OH planar laser-induced fluorescence (OH-PLIF) imaging were used to study the spatiotemporal evolution of heat release and flame structure fluctuations, respectively, whereas 2-D particle image velocimetry (2-D PIV) measurements at 4 Hz were applied to study the mean-based relation between the velocity/vorticity fields and flame structures. Pressure and global heat release fluctuation measurements carried out alongside these optical diagnostics revealed pronounced double peaks in both combustion noise and global heat release fluctuation spectra at global equivalence ratios of ϕ≥0.45. Spectral proper orthogonal decomposition of the OH∗ chemiluminescence and OH-PLIF images revealed that the first peak in the noise spectra to be caused by flame oscillations over nearly the entire flame region, whereas the secondary peak was attributed to periodically generated vortical flame structures near the downstream side of the flame boundary. The 2-D PIV results suggest that vortical flame structures are likely generated by the interaction between the flame and the inner/outer shear layers. [ABSTRACT FROM AUTHOR]

A staged injector developed by JAXA and fueled with kerosene is studied in a high-pressure combustion experiment. With a stable pilot fuel flow rate, the fuel flow rate in the main stage is progressively increased. A high-speed OH-planar laser-induced fluorescence (PLIF) system is used to record the flame motion at 10,000 fps. In the beginning of the recording, the flame behavior is dominated by relatively low-frequency rotation due to the swirling motion of the flow. These rotational motions then coexist with a thermo-acoustic instability around 475 Hz which increases the amplitude of the pressure fluctuations inside the chamber. Dynamic mode decomposition (DMD) analyses indicate that this instability is associated with a widening of the flame occurring when the pressure fluctuations are the highest, giving the instability a positive feedback. The instability frequency then abruptly switches to 500 Hz, while the mode shape remains the same. This frequency change is studied using time-frequency analysis to highlight a change in the feedback mechanism characterized by a modification of the time delay between pressure and heat release fluctuations. [ABSTRACT FROM AUTHOR]