DESCRIPTION
ENGINE MISFIRE DETECTION AND ENGINE EXHAUST SYSTEMS
The present invention relates to a method of detecting misfire in a multi-cylinder internal
combustion engine and to an apparatus for protecting an exhaust system of a multi-cylinder internal
combustion engine, which exhaust system includes a catalyser.
It is known that exhaust gas catalysers and even the exhaust manifolds of internal combustion engines can be damaged or destroyed due to overheating. The danger of overheating arises in the event of a misfire, that is to say, the fuel/air mixture in one or more of the combustion chambers fails to ignite through faulty ignition. The uncombusted mixture when it reaches the catalyser whose working temperature is around 550°C is immediately ignited, rapidly heating the catalyser to a temperature at which the catalyst is destroyed. There is even the danger of the motor vehicle in which the engine is installed being set alight.
It is an object of the invention to provide a method and apparatus whereby faulty combustion or misfire can be immediately identified so that steps can be taken to prevent consequential overheating of the exhaust system, particularly of the catalyser.
Advantages of the Invention
These disadvantages are avoided by the method claimed in claim 1 and the apparatus claimed in claim 5. Not only is a misfire promptly detected through the oxygen content of the uncombusted gases leaving the engine, but measures can be taken to enable the faultily operating cylinder or cylinder group to be identified. In the case of petrol injection individual to the cylinder or the cylinder groups, the feeding of
further fuel to the misfiring combustion chamber can be prevented.
Further advantageous features of the invention are mentioned in the dependant claims.
Drawings
The invention is further described, by way of example, with reference to the accompanying drawings, in which:-
Fig. 1 is a diagrammatic representation of a four stroke internal combustion engine whose exhaust system contains a catalyser, fitted with lambda sensors in accordance with one embodiment of the invention;
Fig. 2 is a graph showing the output voltages of the lambda sensors;
Fig. 3 is a schematic circuit diagram of an electronic fuel injection system operable with lambda control in accordance with another embodiment of the invention;
Fig. 4 is a series of graphs relating to the embodiment of Fig. 3;
Fig. 5 is a logic diagram relating to one mode of operation of the embodiment of Fig. 3;
Fig. 6 is a series of graphs relating to the mode of operation of Fig. 5;
Fig. 7 is a logic diagram relating to an
alternative mode of operation;
Fig. 8 is a graph showing the use of window discriminators; and
Fig. 9 is a diagram explaining the expected gas travel time.
Description of the Preferred Embodiment
Referring first to Fig. 1, a four-stroke petrol engine 10 for a vehicle has an exhaust system 12 containing at least one silencer 14 and an exhaust gas catalyser 16. The catalyser 16 is so disposed in the exhaust system that it runs at about 550°C, this being
the optimum temperature for catalysing the reduction of nitrogen oxide and the simultaneous oxidation of carbon monoxide and hydrocarbons, using up any
residual oxygen in the exhaust gases.
Should there be a misfire in one of the engine cylinders, the uncombusted gases exhausted from that cylinder are caused to ignite in the catalyser. This causes the catalyser to heat up rapidly to a temperature at which the expensive catalyst is destroyed and
there is a danger of the vehicle itself being set on fire, unless appropriate counter measures are taken.
According to the embodiment of Fig. 1, lambda sensors 18 and 20 are fitted in the exhaust system 12 upstream and downstream of the catalyser 16. The lambda sensors comprise a solid electrolyte and, as is well known, they deliver an output voltage which is dependant upon the residual oxygen content of the exhaust gases. The sensor voltages are shown very diagrammatically in Fig. 2. The output voltage u1 of the upstream sensor 18 fluctuates slightly in
synchronism with the exhaust strokes of the engine cylinder. On the other hand, residual oxygen is
consumed in the catalyser 16 so that the output
voltage u2 of the downstream sensor 20 is substantially constant. Assuming the fuel/air mixture fed to the engine is in stoichiometric relationship or slightly lean, the average of the slightly rippling voltage u1 is about the same as the voltage u2.
In Fig. 2, it is assumed that one cylinder
commences to misfire so that uncombusted fuel and air reach the upstream sensor 18 and the oxygen in the air causes a sudden drop in the voltage u1. On the other hand, the fuel and oxygen are combusted in th e c a tal ys e r 16 s o th a t th e ou tpu t voltage u2 of the downstream sensor 20 does not fall and it may even rise slightly.
To detect the occurrence of the sudden change in u2-u1, the outputs of the sensors are connected to a comparator 22 which delivers an output signal S when the voltage u2-u1 exceeds a predetermined value Δu.
To reduce the likelihood of the fluctuations in the voltage u1 triggering a false alarm, a smoothing circuit 24 can be arranged between the sensor 18 and the
comparator 22.
The alarm signal s can be used to trigger an audible and/or visible alarm or to stop or restrict the fuel supply to the engine. If the misfiring cylinder can be identified, in the case of an internal combustion engine with petrol injection, the fuel supply to the faulty cylinder can be cut off by holding the respective inj e ction v alve clo sed . One way o f dete ctin g the faulty cylinder is to provide, instead of a single upstream sensor 18, several separate sensors in
respective branch tracts of the exhaust system leading from the engine exhaust valves. The output of voltage of each such sensor is then compared with that of the downstream sensor.
In the case of an engine fitted with lambda
control, in which the injected fuel quantity is
adjusted in response to the air number λ of the exhaust gases as measured by the upstream sensor 18, the
unconsumed oxygen in the event of a misfire and the consequent lowering of the output voltage u1 cause the lambda control to regulate the engine to a
leaner mixture so that the undulating output voltage u 2 then climbs back substantially to its previous value, as shown, in the event that the engine is
not immediately stopped.
Fig. 3 shows an electronic fuel injection system for a four-stroke, four cylinder internal combustion engine having electronically controlled fuel injection and ignition systems. The four injection valves 26,28,30,32 (shown
diagrammatically) are opened and closed in timed
relation to the crankshaft rotation under the control of a computer 34 and via end stage amplifiers 36.
The injection valves 26-32 are individual to the
four cylinders of the engine in that they inject the fuel into the engine inlet manifold branches adjacent the respective engine inlet valves. The computer
34 also controls the four spark plugs 38, 40, 42 and 44 (shown diagrammatically) via an ignition coil and distributor 46.
The computer 34 controls the injection valves
26-32 and the ignition trigger voltages suppled to the ignition coil 46 in accordance with operating parameters to provide the optimum injected fuel quantities, injection timing and ignition timing in conventional manner. The operating parameters include a reference pulse BM1 inlet pressure p (vacuum), load L (pedal position), engine speed n, the air number λ, as
measured by a lambda sensor in the engine exhaust system, and engine cooling water temperature T. The reference pulses BM are obtained in timed relation to rotation of the crankshaft and are for synchronisation purposes.
During normal running, the engine operates witn lambda control and the lambda sensor (like the sensor 18 of Fig. 1) is arranged upstream of a catalyser in the exhaust gas system of the engine. Lambda
control is control in accordance with the composition of the exhaust gases, and in particular, upon the oxygen content of the exhaust gases, whereby the
fuel/air ratio is that which provides for optimum combustion, i.e., close to stoichiometric and
minimum pollutants in the exhaust gases reaching the catalyser.
Fig. 4 shows at A the output of the lambda
sensor and at B, the instants of ignition of the individual cylinders so that the timed relationship between the changes in the measured air number λ and the individual combustion operations can be seen.
In normal operation as seen at the left of Fig. 4, the air number λ fluctuates between two extremes in accordance with on-off control by the servo loop forming part of the electronic control. In other words, a high air number λ>1 denotes a lean mixture whereby the fuel injection system operates to increase the injected fuel quantity until the air number λ<1 whereupon the injected fuel quantity is decreased.
Each on-off fluctuation lasts for several, e.g., six, revolutions of the crankshaft.
At the instant T1, one cylinder starts to misfire and in Fig. 4 continues to do so at every working stroke. The result is that the lambda sensor or oxygen sensor erroneously measures that the mixture is, on average, too lean so that the computer 34 .
operates to increase the injected fuel quantity until the lambda sensor again indicates an average air number λ=1, as shown in Fig. 4 by curve C. However, bec ause of th e now over - ric h mi xture an d the unc o nver ted oxygen from the misfiring cylinder, the output of the lambda sensor exhibits a sudden dip once for every two revolutions of the crankshaft as shown at the right-hand side of Fig. 4. To detect this situation, the output of the lambda probe is evaluated in a misfire detector 48 as will now be described.
In the simplest version, the output voltage uλ of the lambda sensor is fed to a window discriminator to ascertain whether the voltage uλ both rises above an upper threshold umax and falls below a lower threshold wmin within 720° of crankshaft rotation. If it does,
a pulse is sent to a misfire counter, whose count is shown at curve D in Fig. 4. When the counter has counted, say eight misfires an alarm signal s is
delivered.
If it is desired to identify the misfiring cylinder the window discriminator extends over 180° crankshaft angle, once for each cylinder, the cylinders being identified by the ignition trigger pulses and it is checked whether the sensor uλ voltage first falls below the upper threshold umax and then below the lower
threshold umin within each 180° window. If it does, a counting pulse is delivered to a respective counter and when the counter is counted out, a fault signal s is delivered to indicate an alarm and to which
cylinder it relates. The fault signal s can be used in this case to inhibit the fuel injection signal to the respective injection valve by interrupting the corresponding output from the computer 34 of the
end stage ampl i f i er 36 .
S p ur i o us misfires are not harmful to the
catalyser and it is desirable to prevent the misfire detector 48 from responding to misfiring at spaced random intervals. For this purpose, a storage time counter is also used, as shown in Fig. 5. For each cylinder, it is checked whether within the respective window of 180° crankshaft angle uλ < umax and uλ < umin.
If it does, a stepping pulse is applied to the fault counter. If it does not, a stepping pulse is applied to the storage time counter. Once the fault counter is counted out, the alarm signal s is delivered.
However, should the storage time counter be counted out before the fault counter has been counted out, both counters are re-set. In other words if fewer than the number of misfires needed to count out the fault counter take place within the count-out period of the storage time counter, no fault is indicated.
The total count of each of the two counters can be adjusted to suit the operating conditions of the engine. Thus, for each counter, an empirically determined characteristic can be stored in the
computer 34 to determine in accordance with, say engine speed n and engine load L, the number of misfires to be counted in the fault counter to
trigger the alarm signal s and the number of crankshaft revolutions without misfire to be counted in the storage time counter before detected misfires are to be disregarded.
Fig. 6 illustrates this operation further. Curve A is the output voltage uλ of the lambda sensor. It can be seen that the output is not as regular as
Fig. 4(A) would suggest but the overall pattern of Fig. 4(A) is perceptible in Fig. 6(A). Curve B in Fig. 6 represents synchronising pulses obtained from the ignition system. Curve C represents the count of the fault counter. It can be seen that, following the start of misfiring at T1, the sensor voltage uλ falls below umax = 110mV and then below umin = 50mV to step the fault counter on 1. Over the following four working cycles (2x4=8 revolutions), the sensor voltage uλ does not rise above umax so that it cannot fall through the upper threshold to step the fault
counter further and the latter is then re-set by the storage time counter timing out (counting out). The fault counter is then counted to 2 before being re-set However, the effect of the lambda control then
becomes apparent and the sensor voltage uλ rises to significant values between the troughs and the fault counter is rapidly counted out to trigger the alarm at T2. Curve D shows that the catalyser temperature only rises a few tens of degrees before the alarm is triggered. Fig. 6 assumes that no counter-measures are taken after the alarm is triggered or otherwise
the sensor voltage uλ would not follow the course shown after T2.
Another possibility is for the amplitude of the fluctuations in the output voltage uλ of the lambda sensor to be measured in window discriminators
within a corresponding 180° crankshaft angle for each cylinder. Should the difference between the highest and lowest values of the sensor voltage uλ exceed a value predetermined in accordance with engine speed and/or load, a misfire signal is applied to a fault counter as described above. Also, as described above, the fault counter can be re-set in the absence of any further misfire signals within a predetermined,
possible speed and/or load dependant interval. This is shown further in Fig. 7.
Since it takes time for the gases exhausted from a misfiring cylinder to reach the lambda sensor, the gas travel time must be taken into account in order to identify a misfiring cylinder by reference to the sensor output voltage and the ignition trigger pulses. Fig. 8 shows how the uncombusted exhaust gases from a misfiring cylinder can be identified by the lambda sensor. For a four cylinder, four stroke engine, the working cycle of 720° crankshaft angle is divided into four windows of 180° each. The window in which the falling flank of the sensor voltage uλ crosses the upper and lower threshold corresponds to the
uncombusted gases. Therefore, to ascertain the cylinder from which these gases came, it is necessary to delay the ignition pulses by the time taken for the working stroke of a cylinder to take place and the time taken for the gases to travel through the exhaust tract as far as the lambda sensor. Fig. 9 shows the gas travel time measured from the TDC of the misfiring cylinder. It is typically 560° crankshaft angle but does vary
according to engine operating parameters. Thus, the signals identifying the cylinders must be delayed by this gas travel time before correlating them with the misfire signal obtained from the lambda probe.
Whilst mention is made above of using the ignition trigger signals as cylinder identification signals, inasmuch as the ignition trigger signal is close to TDC, this is not essential. It is possible to obtain cylinder identification signals in other ways, e.g., by dividing the interval between successive reference marks BM by the number of cylinders as shown in Fig. 9.
The expected gas travel time can be derived from an empirically determined characteristic field which is stored in the computer and which gives the expected gas travel time in dependence on engine operating characteristics, such as engine speed n and/or engine load L.