1 Introduction

To increase the efficiency of internal combustion engines, physical and chemical processes during operation need to be understood. Besides the flow field, the temperature and the species concentration in the gas phase, the investigations of near-wall phenomena are of main interest. To fully achieve knowledge of those heat transfer processes in internal combustion engines, wall temperature distributions have to be known.

For this purpose several measurement techniques have been applied in the past. A short overview of wall temperature-measurements in engines can be found in [1]. The most promising candidate of all methods applied is the phosphor thermometry. The main advantages of this technique are its semi-invasiveness and its robustness against soot radiation and chemiluminescence, which both occur in internal combustion engines. Furthermore, it offers high spatial and temporal resolution over a broad range of temperatures.

For thermometry purposes, a rare-earth or transition metal doped ceramic phosphor-material is coated on the object of interest and excited with an ultraviolet laser pulse. The emitted, time-resolved luminescence decay is a function of surface temperature and can therefore be employed for thermometry purposes [2–4]. A second popular approach relies on the temperature dependent ratio of two spectrally separated emission peaks of certain phosphor materials [5, 6]. The main advantage of the first method is a higher precision and accuracy. Additionally, it also offers a higher temperature sensitivity than two-line thermometry [7].

The first application of phosphor thermometry at internal combustion engines can be found in [8]. They used a photomultiplier tube for capturing the luminescence decay and performing point-wise temperature measurements on the piston crown and intake valves. Another study used the intensity-ratio approach to resolve piston and valve temperatures two-dimensionally using an intensified charge-coupled device camera (ICCD) and an image doubler in an optical direct injection stratified-charge engine [1]. Results were shown for different times after initial firing of the engine. The same group published measurements of the intake and exhaust valves applying the lifetime-method using a photomultiplier [9]. Husberg et al. performed crank angle resolved phosphor thermometry on the piston of a heavy-duty diesel engine by temporally resolving the luminescence decay with a photomultiplier [10]. Kissel et al. employed a complementary metal oxide semiconductor (CMOS) high-speed camera to capture the luminescence decay behaviour two-dimensionally [11]. A first attempt to apply this detection system for wall temperature measurements in internal combustion engines was performed in [12]. The authors showed two-dimensional temperature maps on the cylinder wall. Data were acquired for two different degrees of crank angle during the exhaust stroke with two different temporal discretizations. Knappe et al. discussed the impact of phosphor layer thickness in internal combustion systems showing measurements with two photomultipliers [13]. The phosphor was coated on the inner side of the quartz ring and luminescence was acquired on the gas phase side and the glass side respectively. They proposed keeping the coating thickness below 30 μm to prevent systematic errors of up to 30 K.

In the present study, temperature measurements in an optically accessible direct injection internal combustion engine using thermographic phosphors are shown. Measurements are performed on the exhaust valve for two reasons:

  1. 1.

    The exhaust valves are critical components of an engine due to high-temperature exposure and high temporal and spatial temperature gradients.

  2. 2.

    The demonstration of the feasibility of accurate and precise thermometry at a moving object inside an engine is challenging.

Therefore, a CMOS high-speed camera is employed to capture the luminescence decay two-dimensionally. Crank angle resolved temperatures are shown for motored and fired operation, respectively.

2 Experimental setup

2.1 Optically accessible internal combustion engine

The experiments have been performed in a single-cylinder engine equipped with a quartz glass ring and piston. The direct injection engine had an 84 mm bore and a stroke of 90 mm. This results in an engine displacement of 500 cm3 and a compression ratio of 10:1. The intake valves opened at −350 CAD (crank angle degree) for a duration of 190 CAD, whilst the exhaust valves opened at 190 CAD after top dead centre and closed at −320 CAD. During operation, the temperature of the four-valve pent-roof cylinder head was kept constant at 60∘C whilst the intake air had a temperature of 35∘C.

Data were acquired in two operating conditions: motored, and in the fired part of a work-rest cycle, as detailed below. In motored operation, measurements have been performed after a warm up procedure, in which the gas temperature in the exhaust gas system reached a constant level. In doing so, the engine worked as a compression system, without injection and combustion at 2000 rpm.

Experiments in fired operation have been performed at 2,000 rpm using spark ignition at −39 CAD after injection of isooctane at −300 CAD for a duration of . The indicated mean effective pressure was 2 bar and the combustion-air ratio was λ=1.5. To ensure reproducibility of the in-cylinder cycle-to-cycle temperature gradients and to prevent the glass parts from damage, the engine has been operated in work-rest-mode, in which 275 cycles of fired operation were followed by 1,000 motored ones. Just as under motored conditions, the engine has been subject to a warm-up procedure, whose end was indicated by the convergence of both, the minimum and maximum exhaust gas temperature within all 275 fired cycles after a certain number of work-rest-sets.

2.2 Phosphor thermometry

As shown in a previous study [14] the temperature sensitive range of the phosphor Gd3Ga5O12:Cr covers the engine relevant temperatures between 350 and 750 K, as can be seen in Fig. 1. At 350 K the phosphor has a lifetime of about , which is in the order of magnitude of 1 CAD at 2,000 rpm (). For higher temperatures, the lifetime strongly decreases, resulting in an increase of the temporal resolution. For these reasons, the phosphor Gd3Ga5O12:Cr was used to capture the transient and two-dimensional temperature distribution at the exhaust valves. Evidence about precision and accuracy of this phosphor and the measurement technique in principle can be found in [7, 14]. Furthermore compositional and pressure effects of the surrounding gas phase on the decay behaviour have been investigated in [14].

Fig. 1
figure 1

Temperature lifetime characteristics of Gd3Ga5O12:Cr. Each dot corresponds to the average of 12,800 individually evaluated single-pixel and single-shot decay characteristics. A polynomial fit has been employed to guide the eyes

Full size image

For the coating of relevant surfaces by a thin phosphor layer, only chemical binders came into consideration. Although methods like chemical vapor deposition [15] or sol-gel-deposition [16] have the ability to produce very thin and homogeneous coatings, the complexity and limited size of the object of interest hinder the application at internal combustion engines. The coating process using a chemical binder involves the mixing of the phosphor powder with the binder and spraying it on the device to be tested via airbrush. The finished coating and therewith the properties of the chemical binder have to fulfill the following requirements:

  • Transmission of ultraviolet (excitation) and visible light (emission)

  • Homogeneous and thin () coating as proposed by Knappe et al. [13]

  • Temperature resistance (up to 750 K)

  • Mechanical adhesion

  • Insolubility in fuel

  • Solubility in a different substance for residue-free removal

Several binders have been investigated regarding the above mentioned criteria: SP-115 binder (VHT, AZ, USA), sodium silicate, Silikophen P40W and P50X (Evonik, Essen, Germany), Sauereisen cement no. 31 (Sauereisen, Pittsburgh, USA), Dupli-Color Supertherm (Motip Dupli, Hassmersheim, Germany) and Ulfalux Ofenfarbe (Ulfalux Lackfabrik GmbH, Wilhermsdorf, Germany). The binder which showed best performance was Ulfalux Ofenfarbe. After coating the surface, the dispersion was dried for several hours at room temperature. To ensure temperature resistance, the coating was cured at about 100∘C for half an hour. After this procedure the binder can be removed easily by using acetone.

All measurements shown in this work are based on coatings made from a dispersion of Gd3Ga5O12:Cr and Ulfalux Ofenfarbe, sprayed via airbrush.

Figure 2 shows a schematic of the optically accessible internal combustion engine together with the phosphor thermometry setup. The phosphor coating was excited by the fourth harmonic (266 nm) of a pulsed Nd:YAG laser (Quanta Ray, INDI, repetition rate: 10 Hz, pulse width: 5.8 ns), whose beam profile was smoothed by a beam homogenizer (Holo-Or, Israel, diffusing angle: 0.68∘). The laser beam was directed towards the coating via a broadband mirror inside the piston. The luminescence decay behaviour was captured by a 12 bit CMOS high-speed camera system (LaVision High Speed Star 6, maximum frame rate: 675 kHz) via the same mirror. The camera was operated at 360 kHz to gain a resolution of 64×64 pixels. The working distance of the camera was 80 cm which leads to a spatial resolution of approximately per pixel. The photo lens used was a Nikkor 85 mm f/1.4. Figure 3 shows a schematic of the cylinder head, observed from below. One of the exhaust valves was coated with a dispersion of phosphor and binder. The field of view was defined by the camera adjustment, mentioned above. The calculated depth of focus for f=1.4 was about 5 mm. During operation the exhaust valve (lift h=3.5 mm) is always focussed.

Fig. 2
figure 2

Experimental arrangement

Full size image
Fig. 3
figure 3

Bottom view of the pent-roof cylinder head showing the coated exhaust valve and the field of view

Full size image

3 Data evaluation

3.1 Lifetime evaluation

Before the reduction of the signal decay to a scalar value (lifetime τ), each recorded image has been corrected pixel wise for nonlinearity of the CMOS chip by the procedure outlined in [17]. Fitting of the decays has been performed using a Matlab® routine. First, the intensity offset was eliminated by computing the time averaged intensity prior to the laser pulse and subtracting it for each pixel. Second, the temporal intensity distribution I(t) has been fitted pixel wise by a linear regression using

$$\ln(I(t)) = \ln(I_0) - \frac{t}{\tau},$$
(1)

where I 0 is the initial amplitude. Since the majority of phosphors do not decay linearly in the logarithmic scale, an algorithm of Brübach et al. has been used to define a lifetime related fitting window iteratively [18]. This procedure enables the user to compute identical lifetimes at constant temperatures, independently of the settings of the acquisition system such as temporal discretization, observation length, etcetera. The iteratively defined fitting window starts at t 1=t 0+c 1⋅τ and ends at t 2=t 0+c 2⋅τ, where t 0 denotes the instant of laser excitation in time. The parameters c 1 and c 2 are to be chosen specifically for each phosphor material. For the phosphor used in this study single-shot precision and residuals could be optimized by use of the parameters c 1=0.5 and c 2=3.5.

3.2 Correction of valve position

Due to the shape of the cylinder head and the mounting orientation of the valves there is a displacement in the image plane when the exhaust valve lifts. For cycle-resolved temperature distributions it has to be ensured that each pixel represents the same position at the valve, independently from the valve position.

Derivable from the cylinder head geometry and the valve lift curve, it can be computed that the movement in the image plane within one degree of crank angle is smaller than the detected object size per pixel. According to this, the valve position has not to be corrected prior to fitting the lifetimes within one decay behaviour, but after temperature maps for each crank angle degree have been computed.

To correct for apparent valve shift, a procedure implemented in Matlab® has been developed. The information of the position of the valves was derived from images, directly taken after excitation. There, the intensity of the coated valve was near saturation, whilst the uncoated parts showed only noise. Using a canny edge detection algorithm [19], the images were reduced to binary information, showing the edge of the valve. Subsequently a two-dimensional cross-correlation of each of these images showing the displaced valve edge was executed with a binary image of an undisplaced valve (e.g. at 180 CAD). The position of the maximum value of the cross-correlation indicates the valve displacement. Finally, the valve displacement was corrected by shifting the entire temperature information in the corresponding image.

4 Results and discussion

Two-dimensional, crank angle resolved temperature measurements at the exhaust valve have been performed. Surface temperature has been measured at 33 different crank angle positions with a higher discretization near top dead centre and during exhaust valve movement. In motored operation a set of 10 consecutive measurements has been recorded for each crank angle observed. In fired operation, every 5th of the 275 fired cycles has been observed, which results in 55 datasets for every crank angle observed.

4.1 Motored operation

Figure 4 shows the two-dimensional temperature distribution in the recorded field of view at the observed exhaust valve for selected crank angle degrees. Each pixel represents the mean value of 10 consecutive measurements. White areas in the images show uncoated regions or, as for the images near top dead centre, regions, where the piston truncates the laser beam. The temperatures are in the range of 320 to 400 K with overall highest temperatures near top dead centre, due to convection of the compressed air. Generally, the edge of the valves shows higher temperatures compared to the inner part, as observed in [1]. This can be explained by

  1. 1.

    Lower heat conduction towards the centre of the valve due to decreasing cross section at the edge and

  2. 2.

    Increasing gas velocities during valve opening and therewith high Nusselt numbers at the edge of the valve.

Fig. 4
figure 4

Temporally averaged exhaust valve temperature map in motored operation for selected crank angle degrees, where 0∘ CA denotes top dead centre. The blue and red rectangles in the upper left part represent areas to be averaged and illustrated in Fig. 5

Full size image

Figure 5 illustrates this more clearly. The red and blue curves show the temporally and spatially averaged temperatures, corresponding to the red and blue rectangles shown in the upper left part of Fig. 4. The error bars represent the temporal standard deviations of the spatial mean values which are about 1%. Additionally, the phase-averaged indicated mean effective pressure is shown. The temperatures at the edge of the valve are about 10 K higher than in the centre. The temperature variation is stronger for the central valve position (red line). After a slight cooling when the intake valves open, temperatures start increasing at about −90 CAD and reach their maximum at top dead centre. During expansion, temperatures first decrease and then peak again at about 70 CAD. A possible explanation for this effect could be the recirculation of hot air, caused by in-cylinder flow motion. When the exhaust valves open at 190 CAD, temperature starts increasing at the edge of the valve, whilst the red line remains at almost constant temperatures. This is implicated by high Nusselt numbers, caused by low pressure in the combustion chamber which starts expanding on the opening of the valve. When the exhaust valve is opened completely, temperatures drop at the edge of the valve due to pressure equalization, lower gas velocities and therewith lower Nusselt numbers.

Fig. 5
figure 5

Crank angle resolved temperature and pressure distribution in motored operation. The red and blue lines correspond to temporally and spatially averaged temperatures at the inner and outer part of the exhaust valve, respectively. The error bars represent the temporal standard deviations, the indicated in-cylinder mean pressure is shown in black. Additionally intake and exhaust valves opening (IO, EO) and closing (IC, EC) are marked

Full size image

4.2 Fired operation

Figure 6 demonstrates the transient temperature behaviour of valve surface temperatures in fired engine operation. Spatially averaged temperatures for two positions at the valve are shown over the number of fired cycles for −180 CAD. For temperatures near the edge of the valve, a steep increase during the first fired cycles can be seen, which then turns into a nearly constant increase in temperatures, whilst the red line shows no significant change in temperature gradients. Though gradients become smoother with increasing temperatures, steady state could not be reached after 275 cycles of fired operation. Therefore, for fired operation, only data from the last (275th) fired cycle for each crank angle locked measurement is shown.

Fig. 6
figure 6

Exhaust valve temperature for −180∘ CA versus the number of consecutive fired cycles. The red and blue lines correspond to regions at the inner and outer part of the exhaust valve, respectively, as shown in Fig. 4

Full size image

Figure 7 shows the two-dimensional temperature information in the same manner as Fig. 4. The temperature gradients from the centre to the edge of the valve are much steeper than under motored conditions. Temperatures can be measured during combustion (e.g. at 10 CAD) without the use of a band-pass filter, because the luminescence intensity of the phosphor is several orders of magnitude higher than the detectable flame luminosity. Hot spots in the temperature distributions as can be seen at the left border of the image for −360 CAD could possibly be explained by coating inhomogeneity or inclusion of air within the coating.

Fig. 7
figure 7

Exhaust valve temperature map in fired operation for selected crank angles, where 0∘ CA denotes top dead centre. The blue and red rectangles in the upper left part indicate the areas in which temperatures were averaged for Fig. 8

Full size image

Figure 8 shows the cycle-resolved exhaust valve temperatures in fired operation in analogy to Fig. 5. In contrast to motored operation, temperatures are only spatially averaged, representing a sequence of temporally single measurements. As observed under motored conditions, the temperature for the 275th fired cycle of the exhaust valve decreases during intake of cold fresh air. Additionally, fuel-injection leads to a decrease in air- and thereby valve temperature. The minimum temperature is reached at bottom dead centre. Due to compression, the temperature first increases slightly and but rises considerably due to combustion, with a phase lag of about 20 CAD with respect to ignition. After its maximum at 10 CAD the temperature decreases during the expansion stroke until the opening of the exhaust valves. A constant temperature can be observed between 80 and 180 CAD at the edge of the exhaust valve. At the beginning of the exhaust stroke, the outer valve temperature (blue line) shows a decrease in temperature, whilst the inner part of the valve remains nearly unaffected. This characteristic could be observed in every fired cycle recorded. A possible explanation could be a pressure impulse due to resonances within the exhaust system, transporting cooled air from the previous cycle to the valve, which leads to a cooling effect. This hypothesis is supported by a slight increase in in-cylinder pressure at about 230 CAD, shown in Fig. 8.

Fig. 8
figure 8

Crank angle resolved temperature and pressure distribution in fired operation. The red and blue lines correspond to spatially averaged temperatures at the inner and outer part of the exhaust valve, respectively. The indicated in-cylinder mean pressure is shown in black. Additionally intake and exhaust valves opening (IO, EO) and closing (IC, EC) are marked

Full size image

5 Summary

Thermographic phosphor thermometry has been applied to measure temperatures on the exhaust valves of an optically accessible direct injection internal combustion engine. Upon excitation by a UV laser, the luminescence decay of the phosphor Gd3Ga5O12:Cr has been captured by a CMOS high-speed camera. Two-dimensional temperature distributions as well as crank angle resolved temperatures were shown under motored and fired conditions. For fired operation, measurements could be performed during combustion and in the exhaust stroke.