1. Introduction

The ubiquitous utilization of internal combustion (IC) engines has required automotive engineers to continually improve engine efficiency while meeting strict emission standards. Exhaust-gas recirculation (EGR) is an important component of combustion strategies in modern engines. EGR essentially increases the amount of inert gas in the cylinder, either by returning gas from the exhaust duct to the intake (external EGR) or by retaining residual burnt gas in the cylinder at the end of the exhaust stroke (internal EGR). Depending on the type of engine, EGR has multiple effects on combustion. For example, the lowered peak temperature is beneficial in controlling engine-out nitric oxides, while for premixed combustion ignitability and flame speed are lowered, which can be detrimental. Therefore exact knowledge of the amount as well as the temporal and spatial distribution of the residual gas before ignition is necessary. The rate of steady-state external EGR can be measured with reasonable effort, but this time-average may not represent the residual-gas content in a given single cycle. For the quantification of internal EGR, no satisfactory solution exists yet. Currently available techniques like gas sampling and subsequent analysis with a gas chromatograph or mass spectrometer mostly yield cycle-averaged information [1,2]. In contrast, optical methods can be applied in situ with high temporal resolution.

Optical methods have been used previously for in situ measurements of the concentration of various species in the compression stroke of IC engines. Coherent anti-Stokes Raman scattering (CARS) [3] and laser-induced fluorescence (LIF) [4,5] enabled point and imaging measurements, respectively, while various absorption-based methods served for line-of-sight-integrated concentration measurements [6–8]. The most important EGR-relevant species are H₂O and CO₂. However, they cannot be observed directly by fluorescence except at temperatures that are unrealistically high for the compression stroke [9]. On the other hand both species have relatively high absorption cross-sections in the near-infrared spectral range, which makes absorption-based detection interesting.

Numerous absorption methods have been applied to the detection of H₂O in engines: Some were based on broadband absorption spectroscopy (BAS) [10], fixed-wavelength absorption spectroscopy (FWAS) [11], or scanned-wavelength techniques like tunable diode laser absorption spectroscopy (TDLAS) and wavelength modulation spectroscopy (WMS) [12,13]. Methods like BAS and FWAS record a single, for BAS spectrally-integrated signal to determine the absorber-species concentration. This allows the use of simple and inexpensive data acquisition (DAQ) systems. However, these fixed-wavelength or wavelength-integrated methods provide only limited selectivity, require rather complicated, time consuming calibration, while it is difficult to correct for transmission losses or luminous background emissions.

On the other hand, WMS, being a scanning wavelength method, provides selectivity, but uses lock-in-techniques which require additional equipment and in the majority of the cases calibration. Recently calibration-free WMS has been demonstrated, but this technique relies on complicated models for the calculation of the expected WMS signals [14], which are rather difficult to parameterize.

Direct TDLAS (dTDLAS) [15,16] combines several advantageous properties: dTDLAS captures DC-coupled signals, which allows for very effective correction of multiplicative transmission losses (e.g., laser attenuation on windows or due to particles) as well as of additive signal components, e.g. luminescence of flames and hot surfaces. dTDLAS also has been shown to allow calibration-free absolute concentration measurements [17]. Calibration-free optical methods are particularly valuable for species like water vapor, whose measurement by extractive sampling is susceptible to artifacts due to strong surface adsorption and possible condensation. Furthermore, traditional H₂O sensors are difficult to calibrate since stable calibration gases are not available. In addition, dTDLAS provides high-resolution spectral absorption line-profiles, which also contain information on gas pressure and temperature as well as on background emission. High-speed TDLAS measurements with a scanning rate of several kHz can be implemented with modern data acquisition systems, thus allowing for continuous H₂O measurements for extended time series, limited only by the available hard-disk capacity. dTDLAS has been successfully used for in situ measurements in many combustion applications [18], such as flame analysis [19], biomass gasifiers [20], and full-size coal power plants [21].

In our previous work, we successfully developed a dTDLAS spectrometer for in-cylinder H₂O concentration measurements based on a fiber-coupled distributed feedback (DFB) diode laser [22]. This system was able to determine H₂O concentration in the compression stroke of a motored IC engine with a time resolution of 250 µs. However, the DFBs used in that work show only a limited wavelength tuning range. This range was not sufficient to fully capture the significantly broadened absorption lines, at increased pressures in the second half of the compression stroke. The strong decrease of the tuning range with increasing modulation frequency further limited the spectrometer performance.

Vertical-cavity surface-emitting lasers (VCSELs) are quite promising light sources due to their much higher modulation bandwidth and significantly wider wavelength tuning range compared to DFB lasers. In the past, VCSELs have been used for methane concentration measurements at repetition rates up to 5 MHz [23] and for measurements of oxygen, carbon monoxide, and HCl at pressures up to 1.1 MPa [24–26]. However, they have not been used in engine applications since their low output power has been assumed to be insufficient for engine measurements. In this paper we describe what we believe to be the first VCSEL-based, high-speed dTDLAS-based in situ laser hygrometer for in-cylinder measurements in IC engines.

2. Vertical-cavity surface-emitting laser (VCSEL)

A VCSEL with a central wavelength of 7300.12 cm⁻¹ (1369.84 nm) at a laser temperature of 40 °C and an operating current of 6 mA was used for the studies presented here. Two aspheric lenses coupled the output of the VCSEL into a single-mode fiber. The maximum output power from the fiber was in the range from 0.17 mW (T_laser = 40 °C) to 0.25 mW (T_laser = 20 °C) which is quite low compared to the 30 mW provided by the fiber-coupled DFB laser used in previous studies. However, due to our efficient coupling and detection optics in combination with a sensitive data analysis the VCSEL output proved to be sufficient for the application reported here.

The well-characterized absorption line 000-101/110-211 at 7299.431 cm^–1 (1369.97 nm) was utilized here. The same line is also used in the DFB-laser-based version of the engine hygrometer [22] and in numerous other TDLAS applications [27–29]. This absorption line was chosen because of the availability of advanced optical components for this wavelength and experimentally verified line parameters [30]. The line selection also takes into account that due to significant pressure broadening in the engine application the line must be well isolated from other lines and that absorption needs to be strong enough over the whole temperature and pressure range. During the compression stroke of our research engine we expect gas temperatures ranging from nearly room temperature up to 700 K and total pressures from around 0.05 up to 1 MPa. The H₂O vapor concentration – depending on the humidity of the intake air and the recirculated gas concentration – is expected to vary between 0.5 and 5 vol.%. In order to monitor the H₂O concentration at typical engine speeds, i.e., at 1000–5000 rpm, a temporal resolution of the analytical system in the range of a few 100 µs is required. Spectral simulations based on HITRAN2008 data [31] were used to check for the compatibility of the VCSEL tuning range and the pressure and temperature dependent spectral width of the target line.

Figure 1 shows the wide tunability of the VCSEL with current modulation at three different laser temperatures. With very slow, quasi-static current tuning (modulation frequency f_mod < 0.01 Hz) the laser continuously scans over 16 cm⁻¹, which is 5–8 times wider than typical current-tuning ranges of a DFB laser (2–3 cm⁻¹, also indicated in Fig. 1 as a yellow bar). For water concentrations from 1 to 5 vol.% and an engine-typical absorption path-length in the order of 10 cm we expect peak absorptions from 6 to 25% at the beginning of the compression stroke, i.e. at 0.1 MPa gas pressure and 300 K gas temperature.

 

Fig. 1 Temperature and operation current dependence of the static wavelength tuning of the VCSEL for three different temperatures (red lines). Superimposed is the water-vapor absorption spectrum for T = 500 K, p = 0.5 MPa and c[H2O] = 1 vol.%). The orange bar indicates the significantly narrower tuning range of the previously-used DFB laser.

Download Full Size | PDF

The comparison with the absorption spectrum of water for T = 500 K, p = 0.5 MPa and c[H2O] = 1 vol.% (blue line) shows the advantage of the wider tuning range of the VCSEL (red lines in Fig. 1). At T = 300 K and p = 0.1 MPa the FWHM (full width at half maximum) of the selected absorption line is 0.2 cm⁻¹. This value increases to 0.8 cm⁻¹ at T = 500 K and p = 0.5 MPa. Such wide lines are well covered by a VCSEL but may usually only be incompletely captured by our previously used DFB laser. This makes fitting procedures for DFB lasers more complicated, less stable and may also result in systematic errors, which will also be investigated in this paper.

The problem of incomplete coverage of the absorption line becomes worse and even more important if laser modulation frequencies in the kHz range (or higher), are required in order to achieve high crank-angle resolution. This further complication is caused by the strong modulation frequency dependence of the tuning-range of most DFB laser, as shown in Fig. 2.

 

Fig. 2 Dynamic tuning range versus laser modulation frequency for DFB laser (green) and VCSEL (red). Note the break in the ordinate.

Download Full Size | PDF

Here the DFB tuning range decreases by 50%, i.e., 5% per kHz, when the modulation frequency is increased from 0.1 to 10 kHz. Because of this decrease, we performed our DFB laser measurements previously only with modulation frequencies of up to 4 kHz, which allowed a tuning range of 1.3 cm⁻¹. In comparison, the VCSEL tuning range decreases by only 10% over the same range in modulation frequency. Even with 30 kHz modulation the dynamic tuning range of the VCSEL remains greater than 8 cm⁻¹ and therefore covers the full absorption line in the relevant pressure range.

3. Experiment

The principle of dTDLAS is based on the analysis of the wavelength-dependent attenuation of quasi mono-chromatic light passing through a gas sample. To describe this behavior the Lambert-Beer law can be used [32]:

The spectrometer is schematically shown in Fig. 3. The VCSEL is temperature-stabilized with an internal Peltier element and coupled into a single-mode fiber. A function generator provides a periodic saw tooth signal modulating the laser current. At the fiber exit, the emitted laser light is collimated and passed directly through the probe volume, i.e., the engine’s combustion chamber. The transmitted light is detected by a 1 mm diameter InGaAs semiconductor photodiode (Hamamatsu), converted into a voltage signal by a low-noise transimpedance amplifier (Femto), subsequently digitized with a 14 bit 100 MS/s A/D converter (National Instruments) and evaluated in our proprietary LabView fitting code. For the complete laser wavelength ramp, also called scan, 2000 (DFB laser) and 3300 (VCSEL) data points or “pixels” (i.e. voltage values) were captured, respectively.

 

Fig. 3 Schematic drawing of laser hygrometer and engine. The function generator modulates the laser-diode current to generate a wavelength scan of the emitted light.

Download Full Size | PDF

To evaluate the H₂O absorption line, a polynomial baseline (containing all disturbance information) as well as a Voigt line-shape [34] was fitted to the measured H₂O spectrum using a non-linear Levenberg-Marquardt algorithm [35]. Significant overlap of the absorption lines from 7294 cm⁻¹ to 7306 cm⁻¹ occurs for pressures above 0.1 MPa. Therefore, a multi-line Voigt fit with 21 lines (with line strength larger than 10⁻²³ (cm⁻¹)/(molecule·cm⁻²)) in the fitted laser scan range was applied. Further 8 lines outside this area with line strength larger than 10⁻²² (cm⁻¹)/(molecule·cm⁻²) were considered during the fit process. The width of all lines was calculated from the measured gas pressure and the derived temperature using HITRAN data. For the target line at 7299,431 cm⁻¹, which is used to determine the H₂O concentration, the air broadening coefficient and the temperature broadening measured by Hunsmann et al. [30] was used.

4. Single-cylinder research engine

In-cylinder in situ TDLAS was performed in a single-cylinder research engine. The engine is designed to provide optical access through a transparent quartz cylinder and an extended slotted piston for viewing through a piston window. However, neither of these large-diameter optical ports was used here. For the DFB-laser-based version of the hygrometer we had developed two compact and minimally-invasive fiber-optical access ports denoted as FOAP and FOAP-2, respectively [22]. For measurements presented here we used the FOAP-2 port, which completely replaces the quartz cylinder liner with a metal liner and two wedged sapphire windows (16 mm diameter). The laser collimator and a photodiode are each positioned on opposite sides of the cylinder. Possible ambient humidity in the gaps between laser collimator and the window as well as between photodiode and window was minimized to near zero by purging with dry N₂ to suppress unwanted H₂O absorption.

For the current work we deliberately worked with the motored engine only: First, because the gas temperature T during compression can be calculated more accurately from the measured cylinder pressure and the measured intake-air temperature. As in previous work [22], we assumed adiabatic compression to calculate the temporal evolution of the gas temperature based on the measure in-cylinder pressure. Second, in the absence of combustion the TDLAS-measured in-cylinder H₂O concentration should correspond to the independently measured humidity of the intake air (relative humidity: 48%, T = 298 K). This facilitates a simple check on the validity of the TDLAS results.

Engine parameters and operating conditions are presented in Table 1.

Table 1. Engine Specifications; Compression TDC Equals 0 CAD

View Table | View all tables in this article

The position of the piston is measured in crank angle degrees (CAD). One engine cycle (four strokes) consists of two complete crankshaft turns i.e. 720 CAD. The crank angle convention used here is that this cycle corresponds to –360 < CAD < + 360, where 0 CAD is called top dead center (TDC), which corresponds to the upper-most piston position in the compression stroke, i.e. the smallest volume in the combustion chamber.

TDLAS measurements were possible until −43 CAD, at which point the piston blocked (in our engine) the laser beam and prevented further measurements in the compression stroke. To synchronize engine and spectrometer, we recorded with an additional 60 MS/s A/D converter (National Instrument) a trigger signal in parallel to the TDLAS signal (see section 3) at every TDC position as well as a trigger from the opto-electronic crank-shaft encoder every 0.1 CAD. In order to determine the in-cylinder H₂O concentration from the laser absorption, gas pressure and temperature must be known (see Eq. (4). The in-cylinder pressure was measured with a piezo-capacitive transducer, which was pegged to the absolute pressure recorded by a piezo-resistive pressure transducer in the intake manifold. For the motored measurements a pressure trace averaged over 300 engine cycles was used. The in-cylinder temperature during compression was determined from the measured intake-air temperature and the averaged cylinder-pressure trace.

5. Results

All measurements shown below were performed in a motored single-cylinder engine at a speed of 2000 rpm. Figure 4 shows two typical single-shot water-vapor absorption profiles (at cylinder pressures of 0.054 MPa and 0.312 MPa) measured with the DFB laser during the compression stroke with a temporal resolution of 250 µs (f_mod = 4 kHz). All signals are filtered with a moving-average filter over 5 pixels = 0.0065 cm⁻¹ (DFB laser) or 0.016 cm⁻¹ (VCSEL) to minimize the effect of electrical noise, mainly caused by the large dynamometer of the engine test bed. The plots’ abscissae each cover a 70% subsection selected for data evaluation at the end of the total wavelength scan.

 

Fig. 4 In situ H₂O absorption profiles (measured data [circles] and fitted spectral profile [solid red lines]) measured with the DFB laser in the motored engine at −180 CAD (left) and −50 CAD (right) corresponding to cylinder pressures of 0.054 and 0.312 MPa, respectively.

Download Full Size | PDF

The strong increase in Lorentzian width with increasing pressure can be seen clearly. With increasing pressures it became more and more important to compute the line width using the measured pressure, instead of fitting with the width as a free parameter. With the line width computed, the scan-averaged optical residual (i.e. the standard deviation of the difference between measured and fitted data) could be reduced to 8.4 × 10⁻⁴ OD (1σ). This converts into an optical signal-to-noise ratio (SNR, 1σ) of 124. At 0.05 MPa this corresponds to a precision of 0.011 vol.% H₂O, which is equivalent to a length-normalized precision of 9 ppm·m. At 0.3 MPa (−50 CAD), the optical SNR decreased to 64, i.e. a precision of 0.023 vol.% or 19 ppm·m, respectively. Especially at −50 CAD it can be seen that the profile of the absorption line is significantly wider than the DFB laser’s tuning range. In the fit of the DFB laser signals this causes – above approximately 0.3 MPa – increasing ambiguity between line profile and baseline and hence increased uncertainty in the baseline determination. More robust fitting algorithms will be developed in the future in order to be able to use DFB lasers at higher pressures.

Figure 5 depicts typical single-shot H₂O line spectra measured with the VCSEL under the same engine operating conditions as in Fig. 4. Here a 10 kHz laser modulation was used, corresponding to a temporal resolution of 1.2 CAD and a laser scan time of 100 µsec.

 

Fig. 5 Spectral H₂O absorption profiles measured with the VCSEL at −180, −130, −90, −65, −50, and −45 CAD corresponding to pressures between 0.074 and 0.42 MPa. The organization of each plot is the same as in Fig. 4.

Download Full Size | PDF

With this temporal resolution and the same digital signal filters as used for the DFB laser signals, the absorption resolution achieved with in a single scan ranged from 2.5 × 10⁻³ (SNR = 28) to 3.4 × 10⁻³ (SNR = 16) at −180 CAD and −50 CAD, respectively. At −180 CAD and an in-cylinder pressure of 0.074 MPa, this results in a precision of 0.046 vol.% H₂O or a normalized sensitivity of 39 ppm·m. For −50 CAD (0.42 MPa), the precision is 0.081 vol.% or 69 ppm·m. This is only slightly worse than in DFB laser scans and still fully sufficient for the intended application of measuring EGR rates. In motored engine operation, from which the current results are, the H₂O concentration is about a factor of 5 lower than in fired operation, where parts of the exhaust gas with typical H₂O concentrations around 130 000 ppm are mixed with the intake air [36]. For measurements in the compression stroke until ignition the pressure and pressure rise rates are very similar only the temperature calculation becomes more complex. The presented measurements can thus be seen as a test case for the spectrometer sensitivity. The advantage of the VCSEL becomes visible in the later part of the compression stroke (e.g. at −50 and −45 CAD), where – despite pressures of up to approximately 0.5 MPa – the entire absorption-line shape still can be captured due to the VCSEL’s wider tuning range.

In motored operation, the in-cylinder water-vapor concentration is solely due to the outside air humidity, which allows assessing the accuracy of the TDLAS measurement under the strong pressure and temperature variations of the compression cycle. Such an assessment can be made on the basis of Fig. 6, which shows the water concentration measured over a single cycle (top) as well as pressure and temperature for the same cycle (bottom). As for Fig. 5 the laser modulation frequency was 10 kHz. The humidity of the intake air is defined by the humidity of the room air (approximately (14000 ± 1000) ppm, measured in the engine test stand with a simple room humidity sensor) and assumed to be constant over the few milliseconds that are represented in Fig. 6; therefore, the measured cycle-resolved water vapor concentration is also expected to be constant. Analyzing Fig. 6, the H₂O concentration over the entire engine cycle was quite constant – as expected – with an average of 13220 ppm and a statistical standard deviation (SD) of 264 ppm (i.e. 2% relative deviation) over the cycle. The slight decrease in the concentration in the beginning can be caused by an uncertainty in pressure ( ± 2 kPa) or temperature ( ± 5 K).

 

Fig. 6 H₂O concentration measured within a single engine cycle from −190 to −43 CAD using the VCSEL spectrometer (top). Measured pressure and calculated temperature (bottom).

Download Full Size | PDF

This compares favorably with previous measurements with the DFB laser at f_mod = 4 kHz, which showed up to 30 times larger fluctuations in the measured H₂O concentration, particularly at increased pressures beginning at −65 CAD [22]. This increased scatter was assumed to be caused A) by too rapid changes in pressure and temperature during the time of one laser scan, i.e. non-stationary conditions relative to the time needed for a scan, or B) by the baseline fitting problems caused by an insufficient scan width. As scan width and non-stationarity are both influenced by the laser modulation frequency, we investigated the modulation frequency dependence of the spectrometer response. Figure 7 compares the H₂O concentration measured (sequentially) with the DFB laser spectrometer at f_mod = 4 kHz (A) and with the VCSEL spectrometer at f_mod = 4 kHz (B) and at 10 kHz (C). The pressure and temperature traces are again shown at the bottom. DFB-laser-measurements at modulations frequencies above 4 kHz were not evaluated because the decrease in the tuning range caused clearly visible problems in the evaluation algorithm.

 

Fig. 7 Scatter plot of individual H₂O measurements (colored dots) over 80 compression strokes (at 2000 rpm) from −190 to −43 CAD measured with the DFB laser (A), and the VCSEL for two modulation frequencies (B/C: 4/10 kHz). Bottom: Measured pressure trace and calculated temperature. Shown to the right of each scatter plot is a histogram of the concentration measured over all cycles at −46 CAD. Note the different scaling of the y-axis.

Download Full Size | PDF

To investigate the stability of the spectrometers, but also to show the sensors’ potential to study possible cycle-to-cycle variations of the engine, we captured, evaluated, and compared 80 consecutive compression cycles using both laser types and two modulation frequencies. The data from these 80 cycles are plotted in Fig. 7, with every colored point indicating an individual laser-based H₂O measurement. The temporally local, crank-angle dependent statistics of the H₂O concentration scatter was investigated by binning the data in narrow time bins of 3 CAD (f_mod = 4 kHz) or 1.2 CAD (f_mod = 10 kHz) width. For each bin we calculated the mean (white rectangles in Fig. 7) and the standard deviation (error bars, indicating 2σ standard deviation). On the right side of Fig. 7 a histogram of the H₂O concentration at −46 CAD is shown for each configuration. Since the engine was motored, the measured H₂O concentration is expected to be constant over the whole compression stroke. The evolution of laser-based H₂O concentration vs. detection time in Fig. 7 and in particular the deviations from the expected constant H₂O concentration therefore allow evaluating the suitability of the laser type and modulation frequency for high-speed H₂O measurements in engines.

At later detection times, i.e. higher pressure and higher pressure rise-rate, increased scatter in the H₂O concentration appears, in particular for the DFB laser: This is the case for the 4 kHz-measurement (Fig. 7: A and B) after about −65 CAD. Here, from −190 to −65 CAD (f_mod = 4 kHz) the average standard deviation within each bin is quite similar, i.e. 123 ppm for the DFB laser and 193 ppm for the VCSEL. After −65 CAD the measured concentration from both spectrometers with 4 kHz modulation frequency shows much greater fluctuations. At −43 CAD the standard deviation is 2970 ppm for the DFB and 1155 ppm for the VCSEL.

The fact that the scatter is a problem common to both spectrometers (i.e. VCSEL and DFB laser) and thus appears to be relatively independent of the individual width of the laser scan indicates that the scatter is mainly caused by the high rate of change in pressure and temperature and not by a baseline-fitting problem. Using the given engine and spectrometer arrangement we can deduce that in a single 4 kHz-laser-scan the maximum pressure change can be as great as 0.02 MPa, i.e., a relative change of 6%, by calculating the difference in the pressure at the beginning and end of the scan. Towards the later part of the compression stroke this effect, causing a varying absorption profile during the scan together with electrical noise can result in relative H₂O-concentration fluctuations of up to 50%.

Hence, faster modulation frequencies were studied to minimize variations in pressure and temperature during a single laser scan. Indeed, Fig. 7C shows that VCSEL measurements with f_mod = 10 kHz yield much less scatter. During the late compression stroke (−65 to −43 CAD) scatter amounts to only 13% of that incurred at 4 kHz, which supports the hypothesis that the scatter is caused by instationary pressure conditions.

Table 2 summarizes the statistics of measured H₂O concentrations over a range from −190 to −43 CAD. All three independent measurements show good agreement in average H₂O concentrations. This is a significant finding since none of the TDLAS spectrometers was ever calibrated against an external humidity reference, i.e. all measurements derive their absolute character solely from the underlying spectroscopic data evaluation. The agreement thus shows the ability of dTDLAS to measure absolute concentrations without the need for calibration even under engine conditions.

Table 2. Average H₂O Concentration and Standard Deviation in Each Bin during the Compression Stroke from −190 to −43 CAD for Both Laser Systems and Different Modulation Frequencies

View Table | View all tables in this article

The influence of the temporally local CAD-dependent pressure rate of change dp/dt can be investigated further by systematic variations of that parameter via the engine’s throttle valve. Throttling varies intake pressure p_intake and thus in-cylinder pressure. Figure 8 shows the pressure traces over an engine cycle for three different intake pressures.

 

Fig. 8 In-cylinder pressure traces for three throttle positions and hence different intake pressures.

Download Full Size | PDF

Varying intake pressure from 0.048 to 0.083 MPa yields maximum in-cylinder pressure-rise rates from 0.011 MPa/CAD (p_intake = 0.048 MPa) to 0.018 MPa/CAD (p_intake = 0.083 MPa). This increasing rate should result in larger measurement fluctuations, especially for lower modulation frequencies.

The results of this study are shown in Fig. 9. Shown on the on the left side are the H₂O concentrations traces measured with the DFB laser at f_mod = 4 kHz. As expected increasing intake pressure and therefore higher pressure-rise rate lead to significantly increased H₂O fluctuations at the end of the compression stroke. The situation is completely different for the VCSEL at f_mod = 10 kHz, for which results are on the right side of Fig. 9. Here, due to rapid and wide tunability, a nearly constant, absolute concentration of about 13200 ppm could be measured over the entire compression stroke without larger systematic variations or a rapid increase in H₂O scatter towards late crank angles.

 

Fig. 9 Intra-cycle H₂O concentration from −190 to −43 CAD measured with DFB laser (f_mod = 4 kHz, left) and VCSEL (f_mod = 10 kHz, right) at varying intake pressure.

Download Full Size | PDF

Table 3 compares the H₂O concentrations (from Fig. 9) measured with both laser systems for the three intake pressures. The H₂O concentration measured in all three experiments was about 13200 ppm (corresponding to a relative humidity = 46% at T = 296 K), i.e. we could demonstrate excellent absolute accuracy of both spectrometer types, without the need for sensor calibration. Independent of the laser type, modulation frequency, and throttle position, all absolute average values are in agreement within the current uncertainty range (8.5% relative). Table 3 also correlates the spectrometer’s temporally local precision and absolute accuracy in the early and late stage of the compression (i.e. −130 and −46 CAD) with the local average pressure p and the pressure rise-rate Δp/ΔCAD. The resulting numbers are graphically presented in Fig. 10.

Table 3. Average H₂O Concentration and Single-cycle Precision During the Compression Stroke from the Data Represented in Fig. 9, i.e., Measured with DFB Laser and VCSEL for Different Intake Pressures*

View Table | View all tables in this article

 

Fig. 10 Ensemble-averaged H₂O concentration at −130 (left) and −46 CAD (right) for different intake pressures and pressure-rise rates measured with DFB laser (f_mod = 4 kHz, red circles) and VCSEL (f_mod = 10 kHz, black rectangles). The error bars correspond to 2σ. The cross-hatched band corresponds to the averaged H₂O concentration (over all measurements) with its uncertainty. Noted at the bottom is the pressure-rise rate, which is low early in the compression (−130 CAD) but much higher later in the compression stroke (−46 CAD).

Download Full Size | PDF

From this comparison across different rates of change in pressure and hence also in temperature the advantages of the wider scan range and higher modulation frequency possible with the VCSEL can be seen clearly. As Table 3 and Fig. 10 show, the DFB-spectrometer at 4 kHz modulation frequency performs very well in the early compression stroke, where an average H₂O scatter of only 200 ppm at −46 CAD was demonstrated. But in the late compression stroke and with higher intake pressures, using the DFB laser resulted in pronounced decrease of precision, with a scatter band of 1230 to 6490 ppm (i.e. a maximum relative scatter of almost 50%). In contrast, the 10 kHz-VCSEL-system had nearly constant precision of 260 ppm at −130 CAD and 400 ppm at −46 CAD.

The absolute accuracy of the in-cylinder H₂O concentration measurements is affected by numerous contributions, the most important of which we list in Table 4. The table shows all measurands that were used for the uncertainty calculation, as well as their value, their estimated uncertainty and the corresponding influence on the total uncertainty of the derived H₂O concentration for a measurement at −50°CAD. The greatest influence in the total uncertainty has the uncertainty of the calculated in-cylinder gas temperature with over 74%.

Table 4. Contributions to the Total Uncertainty in H₂O Concentration at −50 CAD

View Table | View all tables in this article

The uncertainty in temperature is low at the beginning of compression, because the charge temperature is assumed to be between the intake air and the engine inlet temperature, leading to an total uncertainty in the H₂O concentration of 7.6%. The temperature uncertainty increases towards the end of the compression, because it becomes more difficult to account for the increasing heat transfer, which influences the temperature derived from the in-cylinder pressure. This leads to an increased uncertainty in the concentration at the end of the compression stroke of 8.5%. The uncertainty in the averaged concentrations was also estimated to be 8.5%. Clearly, most beneficial for improving uncertainty in future measurements would be more accurate temperature information. This could come from a more sophisticated engine simulation with thermodynamic model or from simultaneous TDLAS-based temperature measurements.

6. Conclusions and outlook

A new high-speed laser hygrometer for sampling- and calibration-free, in situ gas analysis in the compression cycle in internal combustion engines has been developed, successfully demonstrated and validated. This fully fiber-coupled hygrometer is based on direct in situ TDLAS and uses for the first time a 1.4 µm VCSEL with a rather low ex-fiber optical output power of only 0.17 mW. The laser was fiber-coupled, which simplified the hygrometer optics, in particular the connection to the optical interface developed for single-cylinder research engines with exchangeable quartz/steel cylinders. By scanning the laser at up 10 kHz across an isolated H₂O absorption line at 7299.4 cm⁻¹ we were able to extract absolute H₂O concentrations with temporal resolutions of up to 100 µs, i.e. a crank angle resolution of 1.2 CAD at 2000 rpm. The spectrometer was successfully used to measure H₂O concentrations within the compression stroke of a single-cylinder research engine motored at 2000 rpm, thereby covering an in-cylinder temperature and pressure range from 300 to 600 K and 0.05–0.5 MPa, respectively. Motored operation of the engine with room air of constant humidity allowed validating the spectrometer under engine conditions.

In the best spectrometer configuration, at H₂O concentrations around 1.3 vol.%, we could demonstrate a measurement precision of 0.046 vol.% H₂O, which indicates a sensitivity and dynamic range well suited for engine applications. Additionally, we also quantified the influence of laser scan speed (i.e. laser modulation frequency) and of the laser scan range on the accuracy and precision of the spectrometer. Furthermore, we compared the suitability of VCSEL and DFB lasers. For this engine experiment, the VCSEL-based spectrometer, modulated at 10 kHz, proved to be the best choice and significantly superior to a DFB-laser-based spectrometer, despite the more than 100 times higher DFB output power. Both systems (DFB laser and VCSEL) showed equal absolute accuracy during the early compression phase, i.e. at rather low in-cylinder pressures (less than 0.2 MPa). However, later in the compression, at higher pressures and more rapid pressure changes, the VCSEL hygrometer showed much better H₂O stability and much less systematic errors over the course of the compression stroke.

In conclusion we could demonstrate (to our knowledge) the first VCSEL- and direct TDLAS-based hygrometer for sampling- and calibration-free, high-speed H₂O concentration measurements inside an IC engine. This opens up new possibilities for measuring EGR, for which water vapor can serve as a proxy. Due to its “single-scan, single-cycle” capability the new spectrometer is well suited to study intra- and inter-cycle variations of EGR in engines, which is of great practical interest. In the future we intend to further validate the spectrometer in other engines and operating conditions. We also aim to expand its measurement capabilities to other species like CO or CO₂ and to simultaneous measurements of gas temperature and gas concentrations. This would be particularly useful in fired engine operation.