Advantages of Simultaneous In Situ Multispecies Detection for Portable Emission Measurement Applications

In this work, an in situ multispecies portable emission measurement system (PEMS) is presented. The system is based on tunable diode laser absorption spectroscopy (TDLAS) and is capable of measuring tailpipe emissions without the necessity of online calibration. It is intended for application on passenger cars within the real drive emission (RDE) procedure of the Worldwide Harmonized Light Duty Test Procedure (WLTP). In contrast to the extractive measurement principles of commercially available PEMS, the introduced measurement system does not require gas sampling or preconditioning and thus does not suff er from the same low-pass ﬁ lter eff ects on the measurements. These diff erences are suspected to have an impact on certiﬁ cation-relevant measurement data. Measurements have been conducted on a 3-cylinder 1 liter EURO 6b gasoline engine test bench to investigate the diff erences between the presented measurement system and a commercially available PEMS. For the WLTP relevant investigation, water (H 2 O), carbon dioxide (CO 2 ), carbon monoxide (CO), nitrogen oxide (NO), ammonia (NH 3 )


Introduction
D evelopment goals in the automotive industry are lowering targeted fuel consumption and pollutant emissions [ 1 ]. Due to the harmful e ects of air pollutants on humans and the environment [ 2 ], vehicle manufacturers must comply with the increasingly strict legal requirements de ned by the legislation. e European Union de nes exhaust gas pollutant limits for new vehicles in the EU 715/2007 [ 3 ] regulation through the Euro 1 to Euro 6 standards. Within the certi cation process for passenger cars, the cumulative emissions of the regulated species are essential for approval. e regulated pollutants include nitrogen oxides (NOx), carbon monoxides (CO), hydrocarbons (THC), particle number (PN), and particle mass (PM). e current test procedure for passenger cars within the European Union is the "Worldwide harmonized Light vehicles Test Procedure" (WLTP). e WLTP includes a standardized driving cycle (WLTC) on a chassis dynamometer and the recently introduced RDE procedure (since EURO 6d-TEMP). RDE tests are carried out on public roads under xed boundary conditions and speci cations regarding driving characteristics and environmental conditions. e measurement of real driving emissions on the road requires the use of a PEMS [ 3 , 4 ].
Currently, commercially available PEMS are based on extractive measurement principles, where gas samples are collected and preconditioned (e.g., water separation) before the measurement. ey are installed on the trailer hitch and add weight and volume to the back of the vehicle, which has a negative impact on vehicle-and aerodynamics. An extensive preparation phase including heating and pre-and post-calibration is required for every test [ 4 , 5 ]. The sampling and preconditioning of exhaust gas before a measurement is suspected to in uence pollutant concentrations and induce unwanted low-pass lter e ects on the measurement data.
As shown in previous works [ 6 , 7 , 8 , 9 , 10 , 11 ], laser absorption spectroscopy is a well-proven principle for the measurement of gas parameters in harsh environments. e concept of using laser absorption for exhaust gas analysis is implemented already in a commercially available system, but based on extractive gas sampling [ 12 ]. In this work, a novel PEMS is presented, which is based on an in situ exhaust gas measurement principle. e applied measurement principle is tunable diode laser absorption spectroscopy. e system is capable of simultaneously measuring H 2 O, CO 2 , CO, NO, NH 3 , and CH 4 with a high e ective measurement rate. It consists of a measurement cell that is directly attached to the exhaust tailpipe of the vehicle and peripherals in the trunk, which are connected to the cell via ber optic cables.
Due to the noninvasive laser-optical measuring method, no sampling or preconditioning of the exhaust gas is necessary. As a result, there are no temporal lter e ects due to di usive processes in the sampling lines and the analyzers, as is suspected to be the case in conventional systems. Furthermore, the temporal matching to the separately measured exhaust gas mass ow is simpler than with other systems, since there are no varying dead times caused by di erent analyzers and individual probe line lengths.
Within the WLTP the legislator regulates the cumulative emission of pollutants per kilometer. In this article possible di erences in the calculated cumulative emissions between a high-resolution in situ measurement system and a commercially available system are investigated. For this purpose, dynamic measurements were performed on a full engine test bench.

Experimental Setup
e measurement technique used in this article is based on the principle of tunable diode laser absorption spectroscopy (TDLAS). e emitted wavelength of tuneable diode lasers corresponds to the operating current. By modulation of this current the emitted wavelength can be tuned over time. In this work a ramp modulation of the laser operating current is chosen. e resulting ramp-shaped laser modulation pro le sweeps through a narrow wavelength range of typically a few nm. Each timestep in the modulation ramp therefore corresponds to a speci c operating current and therefore wavelength. For every modulation ramp a spectrum of absorption lines is recorded. e molecule-speci c spectra recorded in this way are evaluated selectively for a single species and correspond to mole fraction. e modulation of the diode lasers and thus also the detection of absorption spectra of gases is performed typically in the kHz range [ 13 ].
According to the extended Lambert-Beer law ( Equation 1 ), an attenuation of the intensity in the area of the absorption lines of the exhaust gas is detected [ 14 ]. e detected intensity I at the spectral position v depends on the initial intensity I 0 , the line strength of the molecular transition ( T ), the number density N , the absorption length L , and the absorption line shape g , which is speci c for the corresponding molecule. Disturbances from the process can be compensated by the time-depended transmission correction ( v ) (i.e., broadband absorption from particles) and the emission correction E ( v ) (i.e., background radiation). e number density N of a species can be determined by a nonlinear t of a spectral model (HITRAN [ 15 ]) to extract the area of the line shape ( Equation 1 ). Taking the known temperature and pressure, that is simultaneously measured, the ideal gas law can then be applied to calculate the mole fraction (see Figure 1 ). Since all parameters to estimate the mole fraction are known or were measured, this measurement principle does not need any online calibration [ 16 , 17 ]. In technical systems, broadband absorption and emission, caused by particle load or contamination of the mirror optics, but also intensive heat radiation from hot components, can disturb the signal. erefore, background and o set compensation was implemented into the evaluation algorithm, which increases the robustness against broadband absorption and emission signi cantly.
In typical RDE applications, the probe volume is located directly inside the exhaust tailpipe. An advantage of this measurement method is that the light can be guided to the probe volume via glass ber. is means that the probe volume does not have to be in the immediate proximity of the measurement periphery. Hence, the sensor setup can be decoupled from the optical sensor head and placed in the trunk of the car. e sensor head contains the optical measurement cell were the exhaust gas volume can pass through without disturbance of the ow. is in situ design enables the determination of the gas mole fraction without sampling or gas conditioning. Following Equation 1 an important parameter in uencing the design of such a sensor setup is the absorption path length. With higher absorption path length, the sensitivity can be increased. For exhaust gas diagnostics especially species with low mole fraction are important and therefore a high sensitivity of the sensor is necessary to resolve the dynamic changes of the gas matrix. is only can be realized by using path lengths that are a multiple of the exhaust pipe diameter. To increase the absorption path length within the exhaust gas pipe, the multipath principle of the white cell [ 18 ] was used here in a modi ed assembly. is optical cell consists of a symmetrical arrangement of three spherical mirrors of the same focal length. e volume between this three mirrors then can be scanned by folding the laser beam several times along the central optical axis [ 13 ]. As can be seen in Diemel et al. [ 13 ], the number of species and therefore wavelength regions can be increased by merging two white cells and dual path cells together in one sensor head.
For each species to be measured, the use of an individual diode laser of a certain wavelength is necessary. For simultaneous detection of several species, several optical cells can be used, or several lasers can be combined in one optical cell by time multiplexing. For time multiplexing, lasers are time-shi ed modulated and their light is combined into one ber by means of ber combiners. e measurement system presented here can detect H 2 O, CO 2 , CO, CH 4 , NH 3 , and NO, integrated into two adjacent measuring cells. e measuring cells have three white cells and two dual-pass channels. e dual-pass channels are a simple re ection on the eld mirror of one white cell each [ 13 , 19 ]. e line selection for the di erent species is based on previous publications of di erent research groups [ 7 , 8 , 9 , 10 , 13 , 20 , 21 , 22 ]. e lasers were combined as shown in Figure 2 and distributed among the ve optical channels available. e only exceptions are CH 4 and NH 3 . e tuning of the laser used for this purpose o ers the possibility to measure both species on one laser.
The laser modulation frequency of the presented measuring system is 500 Hz (H 2 O, CO 2 ) or 5 kHz (NO, NO 2 , CO, NH 3 /CH 4 ), respectively. To improve the quality of the results, a phase average of the raw signals with an e ective time resolution of 10 Hz is performed. e structure of the measurement system is shown in Figure 2 . e system consists of the optical measurement cells, shown on the right, which is directly attached to the tailpipe end outside the vehicle and the measurement peripherals, shown on the le , which are positioned in the trunk. A function generator (Fgen) outputs the corresponding modulation scheme for each individual laser. In addition, the function generator triggers the data acquisition system (DAQ). Each laser is operated by a laser driver (LD Driver). e laser drivers convert the corresponding modulation scheme into the modulated laser operating current and control the operating temperature of the laser diode. A ber combiner (FC) is required for time multiplexing. A er passing through the measurement volumes, the laser signal is detected on a  photodiode (PD). e photocurrent is converted into an ampli ed voltage by means of transimpedance ampli ers (TIAs). In addition, pressure (P) and temperatures (T) are recorded at the measuring point. An overview of the speci cations of the measurement system can be found in Table 1 .
For low absorption strengths, the measurement range is limited by the detection limit. As discussed in Diemel et al. [ 13 ], a signal to noise ratio (SNR) of 2 o ers a reasonable estimation for the detection limit and is therefore applied in this study. e ratio of the maximum absorbance to the standard deviation of the residual is used to calculate the SNR. For high absorption strengths, the measuring range is physically limited by total absorption on the measurement signal. As a conservative estimation the upper limits of the measurement range, given in Table 1 , refer to the maximum concentrations measured in this study. As discussed earlier the detection limits are directly proportional to the absorption length. By adjusting this parameter, the measurement system can be adapted to future tailpipe emission regulations. e precision of each species i at each measuring point j can be estimated according to Equation 2 . For better comparability with the comparative measurement methods, the precision of the measurement system is related to the respective measurement ranges shown. e detection limit for H 2 O based on extrapolation as measured mole fractions did not fall below ambient conditions. e accuracy errors given in Table 1 are estimated by the uncertainty in the absorption line strength according to HITRAN [ 13 , 15 ]. ese are the dominant systematic uncertainties in the absorption line strengths. ese scale with the respective concentrations and have no inf luence on the precision.

Unit under Test
e unit under test (UUT) used for the presented investigations is a series production gasoline engine. Based on the speci c engine parameters listed in Table 2 , the UUT can be regarded as a typical representative for mid-size passenger car engines. e described UUT is installed at an active engine testbed coupled with a vehicle simulation that enables virtual driving tests where the engine is integrated as real hardware. us, the combustion emissions resulting from the engine operation under real driving conditions can be investigated on this testbed.
The exhaust gas aftertreatment system consists of a combination of a standard three-way catalyst (TWC) closecoupled to the engine, that is directly attached at the outlet of the turbocharger. In combination, the whole system reaches a certi cation level of EU-6b in passenger car applications. e engine is operated with standard unleaded gasoline fuel, according to DIN EN 228 with at least 95% octane content and up to 5% bioethanol, fuel station grade.
Both exhaust gas measurement systems are positioned at the tailpipe end.

Reference Measurement System
Based on the availability and state of the art in gaseous measurement equipment, current legislation defines the measurement principle to be used in vehicle emission testing for approval procedures [ 25 ]. As this marks the nal requirement to be met within the development process, the underlying analyzation principles can be found in most measurement systems for engine vehicle testing. ese principles apply for mobile emissions measurement systems as well. e commercially available PEMS system from AVL (AVL 492 Gas PEMS IS) is chosen as reference system for the test procedures.
To analyze the probe, a raw exhaust gas ow of approx. 2 l/min is discharged through a heated extraction line to the device. A heated lter is integrated at the end of the heated sampling line to remove particles from the raw exhaust gas as a protective measure for the gas analyzers. Subsequently, the gas ow is precooled by ambient temperature and further transferred into a two-stage cooler. A er the rst cooler stage the gas ow passes the NDUV analyzer, where NO and NO 2 are measured separately. Following this, the gas ow is divided into two partial ows. One of the two partial ows passes the second cooler and then enters the NDIR analyzer for CO and CO 2 while oxygen is measured in a separate O 2 sensor. e second partial ow is led as a bypass. Both gas ows are controlled by the critical ori ce plates in the limited ori ce block. At the two thermoelectric coolers, the accumulated condensate and some sample gas is sucked o by pumps. Downstream of the ori ce block, the two partial ows are recombined, pass the sample gas pump, are mixed with the condensate ows from the condensate pumps, and leave the device via a drain output. e NDUV and NDIR analyzer and the O 2 sensor measure the pressure compensated concentrations of the exhaust gas components in ppm or vol% [ 26 ]. e exhaust gas mass ow is measured by the onboard exhaust ow meter (EFM) of the AVL PEMS.

Test Program
The chosen operational tests for the benchmark of the measurement system at the UUT includes stationary engine operations for the fundamental investigation of exhaust emissions, where dynamic e ects in uencing the derived results can be neglected. As real driving operation is characterized by highly transient engine operation, the investigation is further extended to the WLTC as it is the current test cycle for certi cation.

Results and Discussion
In Figure 3 the mole fraction curves of all measured species of the TDL spectrometer are plotted for a section of the WLTC. An engine cold start and the urban part of the WLTC are shown.  Each component of the exhaust gas results from individual engine and catalytic formation processes. Some can be grouped according to their qualitative trend. As shown in Figure 3 , the qualitative mole fraction trend of some species such as CO 2 and H 2 O or CH 4 and CO show similar behavior. Other channels, such as NO or NH 3 , show di erent qualitative mole fraction characteristics due to their combustion engine formation processes and catalytic e ects in the exhaust gas a ertreatment.
Even apparently prominent events in the mole fraction pattern, such as at second 712, do not overlap exactly at closer examination (see Figure 4 ). e results of the TDL spectrometer are based on a common reference clock on the DAQ board and were measured at the same position in the sensor head.
Commercially available measurement procedures use different measurement methods for different species. This results in temporal shi s or nonlinear dri s between the results of the measurement channels. ese deviations must be compensated separately. e reference system used in this work synchronizes the species characteristics of NO with the characteristics of CO and CO 2 and the values of the exhaust gas mass ow.
Based on the di erent temporal behavior of the concentration pattern, it is not trivial to synchronize the individual mole fraction curves retrospectively based on individual events. e entire system must be considered. e synchronization of concentration curves of individual measuring systems to each other, for example by means of cross-correlations, can be subject to signi cant errors.
In the TDL spectrometer presented here, synchronization is not necessary as the data acquisition of all channels takes place simultaneously and is measured directly in the process. To calculate the cumulated emissions, both measurement systems require the synchronization of the mole fraction pattern to the exhaust gas mass ow.

Direct Comparison of Measurement Principles
To explain the di erences between the two measuring systems with regard to the effective measuring frequency and 10,000 FIGURE 4 Overview of all TDL spectrometers mole fraction curves-detail second 700-720.
occurring lter e ects, Figure 5 shows the measured CO concentration curves of the TDL spectrometer and the reference measuring system from a section of the WLTC as an example. e measurement data of both instruments show similar mole fraction curves, though ner structures and higher peaks are visible in the results of the TDL spectrometer. is observation corresponds to the expected system behavior of both measuring systems due to the di erent measuring methods. e extractive sampling and preconditioning of the test gas in the reference measuring system lead to various low-pass filter effects. The subsequent low-pass filter results in a smoothing of the data. Since the TDL spectrometer measures in situ, it does not exhibit any of these lter e ects and can therefore resolve more transient mole fraction events. Additionally, it can be concluded that the e ective temporal resolution of the TDL spectrometer is higher than that of the reference system. In areas of near stationary operation ( Figure  5 , A), however, the measuring systems also match quantitatively. e certi cation-relevant cumulative emissions are calculated by integrating the pollutant mass for each individual time step, which is calculated according to Equation 3 : For each species i the mole fraction c i is multiplied with the according density i divided by the density of the exhaust gas exh and nally multiplied by the exhaust gas mass m exh of the according timestep. For both measurement systems the exhaust gas mass ow of the EFM of the PEMS system is used. e TDL spectrometer mole fraction curves and the exhaust gas mass ow were synchronized to the CO 2 results of the certi ed AVL PEMS based on cross-correlation. As there are di erences in the measured mole fraction curves between the two systems (as shown exemplary in Figure 5 ), individual emission events might be considered di erently between the systems when multiplied with the same exhaust mass ow.
is leads to di erent results in the cumulated emissions of the same species and thus to different gravimetric total emissions.
To illustrate this interaction, Figure 6 shows the mole fraction trends, the cumulative emissions of CO and NO, and the exhaust gas mass ow in a range of high emission transients. As can be seen, the low-pass lter of the reference system results in a smoothing of the data, whereby individual emission events are attened in their maximum and broadened in their width.
Due to the low-pass lter e ect of the reference PEMS the concentration maxima are perceived to be signi cantly lower than those of the TDL spectrometer. is leads to higher calculated cumulative emissions of the TDL spectrometer when overlapped with the exhaust gas volume ow peaks. is e ect is exempli ed by mark B in Figure 6 for CO. On the other hand, the broadening e ect of the low-pass lter on the concentration curve of the reference PEMS leads to higher calculated cumulative emissions if the exhaust gas volume ow peaks are slightly shi ed in front or behind the peak of the emission event. is e ect can be seen in Figure 6 in area C for NO. While the NO emission event at second ~546 according to the TDL spectrometer has already almost subsided when the maximum exhaust gas volume ow (second 548) is reached, the reference PEMS still measures about 50% of the maximum value in this range. is leads to a stronger increase of the calculated cumulated emissions of the reference PEMS in this range and thus to a deviation from the TDL spectrometer. NO accumulated (g) FIGURE 7 Cumulative CO and NO emissions. e in uences of the discussed e ects of the low-pass lter on the calculation of the cumulated emissions of longer experimental measurements are systematic and primarily dependent on the characteristic engine pollutant formation mechanisms and the analyzer behavior of the di erent species. Resulting in a systematic distinction between the cumulative CO and NO results. Figure 7 shows the measured cumulative CO and NO emissions of the TDL spectrometer and the reference system during the urban part of the WLTC and the mole fraction curves.
Due to the engine pollutant formation and analyzer behavior of the reference PEMS, the e ect of the maxima superposition of exhaust gas mass ow and mole fraction shown on the le in Figure 6 dominates for CO over the entire period of the measurement shown. is leads to a 12% higher cumulative CO emission determined by the TDL spectrometer over the urban part of the WLTC (see Figure 7 ).
e broadening e ects of the low-pass lter shown on the right in Figure 6 dominates for NO, which leads to higher cumulative NO emissions of the reference PEMS ( Figure 7 ). In total, the calculated cumulative NO emissions of reference PEMS in the range shown are about 17% higher than those of TDL spectrometer. e deviations in the cumulative emissions also scale with the emission dynamics. e more dynamic the emission event, the greater the e ect of the low-pass lter on the di erences in the cumulative emissions of the two measuring systems. For more dynamic measurements, such as RDE, larger deviations are expected.

Conclusion
is article presents a novel measurement system suitable for calibration-free, multispecies, in situ emission measurement for RDE applications based on tuneable diode laser absorption spectroscopy. e TDLAS-system was adapted to a dynamic engine test bench and compared with established, portable emission measurement technology. Due to the di erent measuring methods, di erences in the measured mole fractions of the engine exhaust gases were observed. In a direct comparison, the principle-related low-pass lter e ects of the conventional measurement technology compared to the TDLAS system become apparent. e lter e ects of the extractive methods lead to a attening of mole fraction peaks and a broadening of emission events. In conjunction with the measured exhaust gas mass flow, these deviations are also reflected in the resulting cumulative emission values. Deviations of up to 17% were observed in the cumulative emission patterns in the urban part of the WLTP. ese deviations depend on the engine pollutant formation behavior and the measurement methods of the compared devices. In operating ranges of low emission transients, however, the two systems provide consistent results.
With increasing dynamics in engine operation and hence emission formation, such as those observed under real driving conditions, higher deviations of cumulated emissions are expected.