Current Projects

1. On-road Tests of Advanced Engine Control Systems.

In the Engine Control Group (ECG) at DTU, advanced model based engine control systems have been under development since 1990. These control systems contain nonlinear compensators and nonlinear observers of different types. Such systems have been tested with a great deal of success on dynamometer mounted engines in the test cells the Laboratory for Energetics at DTU but this work has only shown the promise of such systems. At the beginning of 1997 work was started on the building of a dedicated research vehicle for the engine control strategies which have been developed. This will make possible a road test of the control systems developed and make it necessary to operate over a much larger temperature range with the complication of a broad range of engine loading scenarios.

The experimental vehicle has been built specifically to accommodate the experimental engine which has been used since 1990 for dynamometer testing. The vehicle is a Morris Mascot Van. The engine is a 1275 cc British Leyland engine with a 7 port cross flow head. With this head, the engine can deliver a maximum of about 85 hp at 6000 rpm (with the standard cam) which makes it a more or less modern engine. A port injection manifold, associated electronics and PC controller/datalogging system have also been built for the engine for vehicle use. The head was built to order by Å.K.M. Racing of Thisted, Denmark, specfically for the ECG at DTU (the address is given below).

The target date for the completion of the experimental engine is the end of 1997. The testing is to start in the beginning of 1998 on the chassis dynamometer at the Laboratory for Energetics, DTU. It is hoped that road testing can begin in the spring of 1998.

7 port 1275cc British Leyland Å.K.M. Racing Cross Flow Head

c/o Georg Mikkelsen
Sofievej 14
DK-7700 Thisted
Denmark
Tlf: (+45) 97920182
Fax: (+45) 97922932

2. A Port Air Mass Flow (MAFp) Sensor.

On the basis of work done earlier in connection with another project (on gas flow measurement and control in 1985), the workers in the ECG, DTU, constructed a prototype port air mass flow sensor. The sensor was about 2.5 cm thick and had a diameter of 38 mm.

The physical principle on which the sensor operates was analyzed and a model constructed for the sensor. It is this model which has been used to calculate the response characteristic of the sensor and to linearize its response: the model does not have adjustable constants. It is only this method which has been used to calibrate the sensor.

In order to test the sensor, it was placed in an experimental setup where adjustable pumping pulsations, superimposed on an adjustable stationary mean air mass flow, could be applied to the sensor. The equipment for these tests was obtained from the Department of Fluid Mechanics. It was an electrically driven single cylinder motor without intake valves (only used to simulate the pulsating flow) and a suction blower. In this way it was possible measure the sensitivity of the sensor as well as its response time over a large operating range, both for steady state and pulsating air flow. These initial experiments showed that the sensor had the desired accuracy and operating characteristics. Some development work was also carried out to improve the accuracy and frequency response of the sensor during these initial experiments.

After these experiments, the two sensors were constructed and installed on the experimental dynamometer mounted engine at LfE and tested against the laboratory standard laminar flow meter normally used. Because of the method of construction used, these two sensors had nearly identical response characteristics and accuracy. They were also insensitive to temperature changes and engine vibrations. The theoretically calculated calibration curve was used to linearize the sensors' outputs. As the sensors' response is nondynamic (extremely fast) up to very high frequencies, it is possible to include this calibration in a simple way in a microprocessor. The sensor accuracy is on the order of 1 % (standard deviation for the measurement error) for all engine speeds between 1000 and 4000 rpm.

To test the transient characteristics of the prototype sensors, fast throttle angle transients were applied to the test engine at different loads. During the dynamometer tests the engine was controlled with a conventional speed-density control system. This system of control was convenient because disabling the transient fuel compensation algorithm made it possible to obtain very large flow pulsations in the throttle and port air mass flows. The volume of the intake manifold is small compared to the engine displacement volume, the ratio being on the order of 1/2: this gives a very rapid air exchange and is thus a good test of the transient responses of the sensors. The sensor response time was on the order of 0.4 msec up to 4000 rpm and thus is sufficiently good for very fast response engines.

The transient tests conducted above suggest the possibility to use the sensor to map the internal EGR in an engine and to measure and evaluate the effects of external EGR in laboratory and production applications. In diagnostic applications, the sensor sould be useful in detecting the gradual reduction of the volumetric efficiency due to valve seat deterioration and piston ring wear.

Apart from the new applications above as a port air mass flow sensor, it is clear that the sensor can be used for other purposes. In particular it is clear that it could be used instead of a conventional hot wire MAFt sensor but without the main defects of this sensor: dynamic nonlinearity in detecting engine pumping fluctuations, temperature drift and lack of bi- directionality.

Summary of Sensor Characteristics.

The basic characteristics of the prototype port air mass flow sensors are summarized below. These specifications should be compared to those of the common conventional sensors.

Port Mass Air Flow (MAFp) Sensor Characteristics:

a.	Basic accuracy:	+/- 1 %.
b.	Linearity (steady state):	+/- 0.5 %, physical linearization.
c.	Linearity (dynamic):	+/- 1.0 %.
d.	Response direction	bi-directional, symmetric.
e.	Response time:	< 1 msec.
f.	Vibration sensitivity:	very small.
g.	Temperature error:	+/- 1.5 %, -40/+100^oC,(projected).
h.	Light off time:	< 10 msec.
i.	Size:	small.
j.	Mounting sensitivity:	non-critical.
k.	Cost:	medium, (projected).

It is thought that better results could be obtained using direct measurement for calibration. Incidentally the physical model of the sensor suggests how the sensor can be improved with regard to accuracy and bandwidth. While it is obvious that the sensors which have been constructed can be improved, it is thought that the results which have been obtained clearly demonstrate that the basic principle is sound and that the method can be developed for both laboratory and production flow measurement applications. This work is currently in progress.

3. Overall Air/Fuel Ratio Control of an SI Engine.

One of the main practical problems in SI engine air/fuel ratio (AFR) control is that connected with Transient Fuel Compensation (TFC). This is because the physical characteristics of the fuel film dynamic system are not known in detail, because of the time delay between fuel injection and engine exhaust and because the frequency characteristics of the lambda sensor are poorly known. The purpose of this project is to study in detail the characteristics of these subsystems and to obtain an overall closed loop systems solution to the TFC problem. The primary aim is to eliminate or reduce significantly the effort necessary to calibrate the TFC algorithm.

In earlier work at DTU, EH, SCS and co-workers have shown that the air mass flow to an engine can be estimated accurately using nonlinear observers and any given air mass flow related (AMFR) sensor (MAP, MAF, or port air mass flow sensors). The problem of controlling the fuel flow to an engine has also been investigated and the result is a nonlinear transient fuel compensation algorithm which has shown good first order compensation accuracy (+/- 3 %). Recent experiments at DTU have shown however that the existing models for the fueling dynamics of an SI engine are only approximate and there are higher order lags in the system which must be taken into account in order to obtain very accurate compensation.

For this reason and others a new controller for lambda control has been designed on the basis of recent identification experiments and tested on an experimental engine. This controller is a lambda observer and is used with a corresponding multivariable feedback control loop which takes into account higher order lags in the fueling dynamics, the injection/exhaust time delay and the lambda sensor time lag. This makes it posssible to obtain accurate closed loop AFR control with a bandwidth which is between 2 and 5 times that which is possible using current self-oscillating systems. The negative feedback level is on the order of 35 dB which gives the overall system a bandwidth of about 5 Hz. The lambda control accuracy obtainable is on the order of +/- 2.5 % without using adaptive control algorithms or electronic throttle control. Within certain limits, this can be accomplished with the use of either HEGO or UEGO sensors for large throttle angle steps. Such a large control bandwidth and gain eliminates the necessity of doing more than a cursory calibration of the fueling dynamics TFC (Transient Fuel Compensation) tables. It is thought that it is possible to improve the performance of this observer even further using nonlinear linearization control techniques.

In order to make this approach to lambda control loop design practical for production use, further investigation of the dynamics of the fueling sub-system and common lambda sensors (HEGO and UEGO) are being carried out. In order to identify the dynamics of the fueling system, it is first necessary to find the dynamics of the lambda sensor itself. This can be accomplished using spectrum analysis, correlation analysis and finally standard identification techniques. For such testing, the lambda sensor is to be driven by a specially constructed signal generator, a Pulse Width Modulated (PWM) minature switching valve. Such a device has been used earlier at IAU for gas flow system identification. Given a knowledge of the dynamics of the lambda sensor, the problem of extracting the dynamics of the fueling subsystem from engine measurements is fairly straight forward, using the same signal analysis techniques as above. It is also relatively easy to find the influence of valve deposits and lean engine operation on the fueling system parameters to an approximation sufficient for closed loop control. Preliminary results suggest that it is possible to make a physical model or the fueling dynamics and identify its parameters as physical constants.

Back to Homepage