Professional Documents
Culture Documents
Fault Tolerant Oxygen Control of A Diesel Engine A
Fault Tolerant Oxygen Control of A Diesel Engine A
net/publication/43236963
CITATIONS READS
3 2,922
4 authors, including:
Rainer Nitsche
Festo
30 PUBLICATIONS 616 CITATIONS
SEE PROFILE
All content following this page was uploaded by Rainer Nitsche on 07 September 2016.
Abstract: This paper is devoted to the fault tolerant control problem of a Diesel engine air
system having a jammed Exhaust Gas Recirculation (EGR) valve. The fault tolerant control is
based on replaning the trajectory in order to track a new controlled variable which is the oxygen
concentration in the intake manifold instead of the fresh air mass flow. The trajectory planning
is based on an inverse model approach, utilizing the fundamental thermodynamic relations of
the air system.
Keywords: Fault tolerant control, Diesel engine, Air system, Model-based trajectory planning
1. INTRODUCTION
MBC VNT
Figure 2. Schematic representation of the air path.
The known time variant parameters required for (3) and Therefore, the desired EGR mass flow ṁdEGR needed to
(16) are collected in the vectors ζ(t) and ξ(t). Note that compute (3) can be estimated from (4) and (5) by
the diagnosis task and the estimation of the jam positition pd
21 V̇eng
(Scheu, 2008) are not treated in this paper. Therefore, this cp,M ix RM ix − cp,F A ṁdF A T21
ˆ dEGR
ṁ = . (6)
jam position is assumed here to be perfectly known. cp,EGR TEGR
Moreover, with the overall mass balance and the oxygen
3. NEW CONTROLLED VARIABLE mass balance in the engine, the desired oxygen ratio
d
rO 2 ,EGR
in the EGR gas, also needed to compute (3), can
The new controlled variable, i.e., the oxygen concentration be obtained by:
rO2 ,22 in the intake manifold, is the ratio of the oxygen
mass flow ṁO2 ,eng over the total gas mass flow ṁeng , going d ṁdF A − rstoic ṁf uel
rO 2 ,EGR
= r O ,F A , (7)
in the engine:
2
ṁdF A + ṁf uel
ṁO2 ,eng where rstoic is the stoichiometric air to fuel coefficient and
rO2 ,22 = , (1)
ṁeng ṁf uel is the amount of injected fuel into the engine.
or, as the gas entering the engine is exclusively composed
of fresh air and EGR gas: 4. MODEL-BASED TRAJECTORY PLANNING OF
ṁF A rO2 ,F A + ṁEGR rO2 ,EGR THE DESIRED FRESH AIR MASS FLOW
rO2 ,22 = , (2)
ṁF A + ṁEGR
where rO2 ,F A and rO2 ,EGR are the oxygen ratios in the According to Figure 4, the desired oxygen concentration in
fresh air and EGR gas, respectively. the intake manifold can now be used to calculate the new
desired fresh air mass flow ṁd∗
F A in the faulty case. In other
This gives the desired oxygen concentration in the intake words, a new trajectory for the desired fresh air mass flow
manifold: has to be planned, depending on the known jam-position
d
ṁdF A rO2 ,F A + ṁdEGR rO
d
2 ,EGR
of the EGR valve and the desired oxygen concentration
rO 2 ,22
= , (3) in the intake manifold. For this, a model-based approach
ṁdF A + ṁdEGR
is used, where the inversion of the intake manifold model
where ṁdF A is the desired fresh air mass flow (given by the is based on a flatness analysis in the sense of Fliess et al.
Electronic Control Unit, ECU), and ṁdEGR is the desired (1992) and Rothfuß et al. (1997).
d
EGR mass flow and rO 2 ,EGR
the desired oxygen ratio in
the EGR gas, which can be calculated in the following way. 4.1 Inversion of the intake manifold model
The enthalpy balance in the intake manifold can be written
as For a model-based trajectory planning, the dynamics of
the intake manifold can be described with the following
cp,M ix ṁ22 T22 = cp,EGR ṁdEGR TEGR + cp,F A ṁdF A T21 , state, control input, and output definition:
(4)
where cp,M ix , cp,EGR and cp,F A are the heating capacities x = [mO2 ,22 , mO2 ,22 ]T , u = ṁF A , y = rO2 ,22 , (8)
of the mixture (EGR gas + fresh air), EGR gas and fresh where mO2 ,22 and mO2 ,22 are respectively the masses of
air, respectively, T22 is the intake manifold temperature, oxygen and all other gases in the intake manifold, with
TEGR the EGR gas temperature, and T21 the temperature the external time-varying parameters
of the fresh air over the throttle valve.
q = [ṁEGR , ṁeng , rO2 ,F A , rO2 ,EGR , neng ]T , (9)
Assuming that the upstream pressure p21 of the throttle with neng the engine speed.
valve is the same as the pressure p22 in the intake manifold
(Scheu, 2008), the ideal gas law gives: State Equations. The first state equation is obtained by
p22 V̇eng pd V̇eng considering the mass flow balance in the intake manifold:
ṁ22 T22 = = 21 , (5) ṁO2 ,22 = ṁO2 ,EGR + ṁO2 ,F A − ṁO2 ,eng ,
RM ix RM ix
where V̇eng is the variation of the engine volume, and RM ix which leads to
is the gas constant of the mixture (EGR gas + fresh air). ṁO2 ,22 = ṁEGR rO2 ,EGR + ṁF A rO2 ,F A − ṁeng rO2 ,22 ,
and finally to where
ẋ1 = q1 q4 + u q3 − y q2 . (10) p3 EGR
Π=
, Tcooler,out = T3 − ηHE (T3 − Tcoolingf luid ),
The second state equation is obtained as follows: p22
ṁO2 ,22 = ṁ22 − ṁO2 ,22 = ṁEGR + ṁF A − ṁeng − ẋ1 , and A∗EGR is the effective area of the EGR valve estimated
by a Kalman filter of a diagnosis algorithm developed by
leading to Scheu (2008).
ẋ2 = q1 + u − q2 − ẋ1 , (11)
or
ẋ2 = q1 + u − q2 − (q1 q4 + u q3 − y q2 ). 4.3 Calculation of the engine mass flow
Output Equation. The system output y is equal to To estimate the engine mass flow ṁeng , the following
mO2 ,22 equation can be used:
y = rO2 ,22 = ,
mO2 ,22 + mO2 ,22 pmeas
22 Veng
ṁeng = ASMod ψ (neng ) , (18)
which gives T22 R 2
x1
y= . (12) where pmeas
22 is the measured intake manifold pressure,
x1 + x2 ASMod
T22 is the ASMod estimation of the intake mani-
System Inversion. The input u can be expressed as a fold temperature, pmeas
22 is the measured intake manifold
ASMod
function of the output y and its successive derivatives. pressure, T22 is the ASMod estimation of the intake
Indeed, (12) gives manifold temperature, Veng is the engine volume, ψ (neng )
x1 = y(x1 + x2 ) = y m22 . is a correction term representing the engine volumetric
efficiency, R is the perfect gases constant, and cp the
As m22 = V22 heating capacity of the intake gas at constant pressure is
RM ix f (neng ) ṁeng , where the known function
V22
the measured intake manifold pressure.
RM ix f (neng ) ,
with V22 the intake manifold volume, con-
siders the amount of air consumption for the engine, the
first state becomes 4.4 Calculation of the oxygen rate in the exhaust gas
V22
x1 = q2 y. (13) To estimate the oxygen rate rO2 ,EGR in the exhaust gas,
RM ix f (neng )
(7) can be used by replacing the desired values by the real
Moreover, (12) gives values:
x1 ṁmeas − rstoic ṁf uel
x2 = − x1 rO2 ,EGR = rO2 ,F A F Ameas , (19)
y ṁF A + ṁf uel
and with rstoic = 14.5.
ẋ1 y − ẏ x1
ẋ2 = − ẋ1 . (14)
y2
Identifying (14) with (11) leads to
4.5 Test of the model-based trajectory planning in the
ẋ1 y − ẏ x1
q1 + u − q2 = . faultless case
y2
Then, using (10) and (13), simple calculations give finally Equation (16) has been implemented in the simulation
the output u as environment and tested using a special test cycle in the
q1 (q4 − y) − RM ixVf22(neng ) q2 ẏ faultless case. Since the faultless case is considered, the
u= . (15) old (given by the ECU) and the new desired fresh air
y − q3
mass flow should be equal. Figure 5 draws a comparison
New desired fresh air mass flow. This leads directly to of these two signals, which are shown most of the time
the expression of the new desired fresh air mass flow: equal. Therefore, it can be assessed that the model-based
d trajectory planning of the desired fresh air mass flow is
(rO2 ,EGR − rO 2 ,22
)ṁEGR correct.
ṁd∗
FA = d
rO 2 ,22
− rO2 ,F A 150
V22 d old desired
RM ix f(neng ) ṁeng ṙO2 ,22 new desired
− . (16)
fresh air mass flow [g/s]
rOd − rO2 ,F A
2 ,22 100
The next step (see Figure 4) is to check if the values of the 300
calculated new desired fresh air mass flow ṁd∗ F A (16) are
250
200
physically reachable (higher limit), and not below a certain
150
limit (lower limit). This reachability analysis is decribed
100
below.
50
0
Higher limit. The fresh air mass flow depends on sev- 0 50 100 150 200 250
as inputs, and the maximal reachable fresh air mass flow 0.2
as output. The output values of this table are the fresh
rO ,22 [−]
air mass flow values obtained in the faultless case from 0.18
2
appropriate simulation runs, where
0.16
120
lower limit Faulty case 50%. As in the previous case, the system
needs sometimes more fresh air than the maximal physi-
fresh air mass flow [g/s]
100
cally reachable value in order to reach the desired oxygen
80 concentration in the intake manifold (Figure 8, first plot).
60
However, in spite of the physical limitation of the system,
the model-based trajectory planning of the desired fresh
40
air mass flow makes the system reach the desired oxygen
20 concentration in the intake manifold in the most operating
0
regions (Figure 8, second plot), except in the full-load
0 50 100 150 200 250 300 350 400
time [s]
regions where the desired oxygen concentration requires
to have only fresh air in the intake manifold, and that is
Figure 6. The new desired fresh air mass flow ṁd∗
F A , with not possible when the EGR valve is blocked at a middle
the maximal reachable and minimal allowed values. position.
200
higher limit
lower limit
150
7. CONCLUSION
100
The model-based trajectory planning of the desired fresh
50
air mass flow has shown many advantages. It allows to
keep the same control structure, which already works
0
pretty well in the faultless case. Furthermore, regarding
0 50 100 150 200 250
the simulation results for different jam positions of the
EGR valve, it has come out that this fault-tolerant control
approach gives satisfactory results. Indeed, it makes the
0.24 system to get closer to the desired oxygen concentration
in the intake manifold than without any replanning of the
0.22
fresh air mass flow trajectory.
0.2
Another additional actuator could be used in parallel
rO ,22 [−]
0.22
0.2
rO ,22 [−]
0.18
2
0.16
desired
measured (with FTC)
0.14
measured (without FTC)
0.12
0 50 100 150 200 250
time [s]