TU Berlin

Collaborative Research Centre 1029SFB1029: A05


Page Content

to Navigation

A05: Modelling, monitoring and control of the pulsating combustion

Principal investigators:
Prof. R. King ()

Scientist:   Florian Arnold, M.Sc. ()
phone        (030) 314 23074




The focus of project A05 is the support of the A-projects "pulsed combustion" as well as the C-projects "interfaces" in aspects of control theory, aiming for reliable SEC and PDC processes under intermediate pressure level conditions. Due to the resulting shortened time scales in the case of SEC, an efficient "0-1" controller for switching valves will be developed. Furthermore, the control of the reactive set-ups is supported by state estimation methods. Approaches for start-up and partial load of the SEC process will be developed in simulation studies.  In the case of PDC, a multi-tube configuration allows the pressure distribution in the turbine plenum to be controlled by firing synchronization. The goal here is to guarantee the refilling of the tubes as well as damping of large pressure fluctuations, which are critical for the performance and life cycle of the turbine.

1st Funding period 2012 - 2016


Fig. 1: Experimental set-up (mixing tube)

One of the main tasks of the collaborative research center SFB 1029 is the investigation of progressive combustion concepts to further increase a gas turbines efficiency.  

Due to the increased thermodynamical efficiency of a constant-volume combustion compared to a constant-pressure regime, the former one represents an attractive alternative for future gas turbines. This efficiency of the isochoric combustion led to many investigations on adopting this combustion style for gas turbines. A method to achieve a nearly isochoric combustion despite the open ended construction of a gas turbine is to use a detonation combustion. The detonation is finished before the reactive mixture can expand. Hence, an almost isochoric combustion results. One problem that goes along with a detonation is that the high pressure peaks created in the system may cause entropy generation as well as high stresses on the components of the gas turbine.

A promising idea to avoid these pressure peaks is the so-called Shockless Explosion Combustion (SEC). Here, by a layered filling of the tube with fuel and air, resulting in an adequate adjustment of the ignition delay, a simultaneous autoignition of the fuel is achieved. This means that the objective during the filling process of the tube is to obtain a suitable profile of the ignition delay distribution of the gas along the tube. The ignition delay mainly depends on the pressure, temperature and fuel concentration of the used gas. From these only the concentration of the injected fuel can be regulated. The temperature and the pressure are treated as disturbances which make it necessary to use closed-loop control. If filled correctly, the temperature rise of the simultaneously ignited gas results in a smooth pressure increase along the tube. This pressure wave travels to the open end of the tube leading into an annulus. It is reflected as a suction wave. This suction wave allows the tube to be refilled due to a lower pressure level on the compressor side.  The filling process consists of an air buffer followed by the controlled charging of the fuel concentration. 

In the Collaborative Research Centre 1029 we investigate the detonative combustion in a pulsed detonation engine (PDE) as well as the SEC. 

Fuel mixture layering

Fig. 2: Experimental set-up (mixing tube)

As a first step, the filling process is investigated in a non-reactive surrogate set-up, where water is used instead of air. This allows examining the injection process at the same Reynolds numbers but at smaller velocities. The experimental set-up consists of a circular water flow which streams a constant amount of water through a tube. At the inlet of the tube fluorescent dye is injected which represents the fuel in the gas turbine. The local distribution of the dye in the tube is then measured using a laser and five photo diodes. This set-up allows the investigation of the distribution of the dye for single SEC filling processes with identical starting conditions.

Fig. 3: Hot experimental set-up, A01 and A03

The local distribution of the fuel is adjusted by two adaptive controllers, an iterative learning controller (ILC) and an extremum seeking controller (ESC). The ILC is able to converge to the desired fuel distribution within 10 time steps. Contrary to the ILC the ESC uses no model and therefore needs slightly more iterations to converge to the solution. The applied ESC algorithm is restricted to an input signal consisting of piecewise linear functions. Therefore, the obtained fuel distribution does not fit the reference as well as in the case of the ILC.

Filling of tubes

Fig. 4: Cold test-rig

The combustion cycle in a single combustion tube is described by the following pattern:

filling with fresh air and fuel → ignition → combustion → purging with fresh air

Thus, the question arises how to get air and fuel, which are needed for the cycle states purging and filling, from the turbine’s compressor into the combustion chamber, i.e. from a lower to a higher pressure level. One method to overcome this challenge is to use the acoustics induced by the detonations itself. For explanation, we first consider a single detonation tube. By igniting a detonation in the tube an acoustic pressure wave is generated. It is then reflected as a negative pressure wave when reaching the open end of the tube. This end is coupled to a plenum. The pressure of the reflected wave is below the mean pressure level of the combustion chamber, and, even more important, below the pressure level of the turbine’s compressor. When this wave reaches the closed, upstream, end of the tube, which is connected to the compressor, it can be refilled with air from the compressor by opening the closed end at the correct time. A new detonation wave can be initiated, and the process is sustained. Hence, the crucial factor to keep the system running is the acoustic pressure wave generated by the detonation. If the detonation fails (e.g. wrong mixture of fuel and oxidizer for the given conditions in the tube), the tube cannot be refilled because of the missing negative pressure wave. Therefore, the combustion chamber design is composed of a circular array of detonation tubes which all fire into a common plenum. For a restart of a faulty tube, we propose to use the remaining working ones. This could be achieved by a proper firing synchronization producing a pressure minimum in the plenum at the open end of the faulty tube. We plan to use a modified model predictive control algorithm to calculate suitable firing patterns. To initiate such a firing sequence, it is necessary to first detect the failed detonation. As a detonation wave generates high temperatures and pressures, the application of sensors inside or near the detonation zone is not feasible. Moreover, for deliberately rotating the pressure profile inside the plenum in case of a misfiring, pressure sensor information from the plenum is needed in any case. For this reason only pressure sensors in the annular plenum are considered for fault detection. Since the acoustic waves generated by the detonations excite acoustic modes in the annular plenum, the tubes can be considered as actuators of a dynamic system from a control engineering point of view. Misfiring, hence, can be viewed as an actuator fault.

Fig. 5: Detonation detection

The investigated configuration consists of twelve ‘detonation’ tubes which are connected to a downstream annular plenum. A non-reacting surrogate set-up is used where the detonation tubes are replaced by loudspeakers mimicking the acoustic excitation of the plenum by the tubes during detonation events (see Fig. 4). Due to high temperatures in real detonation tubes no pressure readings from inside the tubes are assumed. Pressure measurements from the downstream annular plenum are utilized instead. There, lower temperatures and less harsh conditions are expected. For the mockup, these pressure readings are obtained by microphones. Based on these measurements, the system state is estimated by a Kalman filter. The innovation sequence calculated by the Kalman filter, i.e., the difference between the predicted system outputs and the measured ones, is used in an algorithm for misfiring detection. In Fig. 5, results from a fault detection experiment are presented. The tubes are ignited subsequently with a given firing pattern. Furthermore, some of the tubes are determined to misfire (marked by the symbol x) and thus, produce no acoustic excitation. The variable on the y-axis is an output of the fault detection algorithm and is used to decide between faulty and working tube. If this quantity is above a given threshold, the detonation is detected as successful. The variable is only computed at time instants when the next tube is to be ignited (marked by the symbol o). Thus, it indicates if the tube which was ignited previously failed to detonate. Thus, the mentioned algorithm is capable to detect all faulty tubes correctly.


Schäpel, J., R. King, B.C. Bobusch, J. Moeck und C. Paschereit: Adaptive control of mixture profiles for a combustion tube. In: Proceedings of ASME Turbo Expo, Nr. GT2015-42027, 2015.

Wolff, S. und R. King: Model-Based Detection of Misfirings in an Annular Burner Mockup. In: King, R. (Hrsg.): Active Flow and Combustion Control III. NNFM, Bd. 127, S. 229–244. Springer, Heidelberg (2015), 2014.

Wolff, S. und R. King: An annular pulsed detonation combustor mockup: system identification and misfiring detection. J. Eng. Gas Turbines Power, 138(4):041603–8, 2015.

Wolff, S., J. Schäpel, P. Berndt und R. King: State estimation for the homogeneous 1-D Euler equation by Unscented Kalman Filtering.
In: Proceedings of the ASME Dynamic Systems and Control Conference, 2015.


Quick Access

Schnellnavigation zur Seite über Nummerneingabe