EP1595066A1

EP1595066A1 - Method for detecting combustion failure by filtering

Info

Publication number: EP1595066A1
Application number: EP04709994A
Authority: EP
Inventors: Alain Rossignol
Original assignee: Siemens VDO Automotive SAS
Current assignee: Continental Automotive France SAS
Priority date: 2003-02-20
Filing date: 2004-02-11
Publication date: 2005-11-16
Also published as: WO2004074662A1; FR2851612B1; KR20050103949A; JP2006518435A; US20060074543A1; FR2851612A1; JP4340284B2; US7072760B2

Abstract

A method for detecting combustion failure in a 2-stroke or 4-stroke internal combustion engine, wherein an index ln (t) representing engine irregularities is calculated at a moment (t) during certain combustion phases in one of the cylinders; b) an observation function G (t), which is a function between index ln (t) and an estimated reference value In +/- i (t) of said index is calculated at a different moment during the operation of the engine and on the basis of said index ln (t); c) G (t) is compared to a threshold (S), which is a function of the load (C) and/or rpm (N) of the engine and d) a combustion failure is diagnosed when the value of G (t) exceeds the threshold (S) or: the observation function G (t) is such that G (t) = In (t) - ln (t), wherein In (t) is a function of ln (t) and In +/- i (t) represents the index ln (t) after weighting of the values of said index outside an acceptable predefined differential (BC), prior to the calculation of the function G(t), calculation occurs in terms of (t) of the differential between said value at moment t of index ln (t) and the reference value In +/-i (t) of said index at a different moment.

Description

METHOD OF DETECTING COMBUSTION RATES BY FILTERING

The present invention relates to a method of detecting misfire (or partial combustion) of an air / fuel mixture in the cylinders of an internal combustion engine, in particular an engine propelling a motor vehicle, this method providing that, preferably during each combustion phase of this mixture in any one of the engine cylinders, an index representative of the engine operating irregularities is calculated.

Combustion irregularities in an internal combustion engine are useful to diagnose because they reveal various disorders that can affect engine operation.

The FR-A-2 patent is known. 689 934 such a method, intended more particularly for such a diagnosis during an operation of the engine at medium and high revs. According to this method, during each combustion phase in any one of the cylinders, sets of time differences Δtj separating predetermined angular positions of the crankshaft are then calculated, from a linear combination of time differences of the whole, a "critical parameter", or "index", representative of the variations in the engine rotation speed (crankshaft) during this combustion phase and this index is compared to a fault threshold to diagnose the possible occurrence of 'a misfire.

To measure the aforementioned time differences which separate different angular positions, a rotating part (or target) integral with the crankshaft is used

Also known from FR-A-2 718 489 is a method of the type described in FR-A-2 689 934, but improved so as to eliminate disturbances in the diagnosis by variations in the geometry of the target.

Processes of the same type are also known, modified so as to prevent a "recovery" of forces, resulting from the movement of a vehicle powered by the engine on a surface of irregular geometry, from disturbing the diagnosis, and this without using a sensor dedicated to the detection of such efforts. The solutions hitherto proposed nevertheless have drawbacks which constitute a limitation which is difficult to accept in the context of a function of discrimination of poor combustion of an engine.

First of all, if the use of an index representative of the variations in the engine speed during the combustion phases is indeed appropriate, the parameter best suited to the detection of misfires of combustion at high engine speed s 'proves sensitive to high amplitude oscillations, noises and jolts typically transmitted to the crankshaft by the transmission chain, during the current use of a vehicle. These disturbances can typically be the consequence of clutch shocks or traffic on a bad road. This can in any case lead to inappropriate discrimination of misfires, depending on the intensity of the disturbances, in particular at low engine speed.

Conversely, a parameter proving to be well suited to detection at low engine speed is not very sensitive to these high amplitude oscillations, but itself has a low sensitivity to variations in speed characteristic of an absence of combustion at high engine speed.

Faced with this, we have already considered the selection of one parameter among several, according to the conditions of use of the engine, that is to say in "current" use. However, in this case, on each modification of these conditions of use, a reset sequence of the diagnostic process is necessary and may cause a temporary inhibition of the latter. The multiplication of these operating / reset sequences around an engine speed or load threshold, during an oscillation or a transient load situation, can also interrupt the diagnosis for a relatively long time interval.

It can also be noted that the "critical parameters" or "indices" of the processes of the prior art mentioned above, as well as the detection algorithms which make use of them, prove to be effective only under certain load and operating conditions of the engine and / or only for certain types of combustion faults. We can in fact distinguish in this regard erratic misfires, systematic misfires (affecting for example the same cylinder) and various sequences of misfires affecting different cylinders, for example. Note also that, for example in FR-A-2 689 934, the parameter used can under certain conditions be brought to carry out an inappropriate discrimination of real and false misfires of combustion if a defect of adaptation to the target is revealed, as for example the case of an uncontrolled drift of the sensor associated with the crankshaft.

One could be tempted to overcome these drawbacks by multiplying the diagnostic processes, each one being adapted to a particular area of engine operation, or to filtering out a cause of particular diagnostic disturbance. This approach is however cumbersome and costly, both in terms of the development of these diagnostic processes and in computation time, during the execution of these complex processes.

It is in this context that the invention aims to provide a solution to the above drawbacks: - by significantly reducing the development times necessary for carrying out a reliable detection process,

- by securing the detection diagnosis in the event of an uncontrolled component drift,

- by closely matching the real average conditions of the engine, and thus authorizing a diagnosis for more than 90% of the time of use of this engine,

- avoiding the addition of a sensor which would otherwise be necessary to describe the condition of the roadway,

- by suspending the diagnosis only during the time strictly necessary for the disappearance of the disturbing phenomenon,

- by reacting quickly, before the appearance of the critical fault, and by not requiring retroactive erasure of the faults detected and stored in memory,

- by securing the diagnosis of detection of combustion irregularities for the most part at least of the borderline cases not listed, by intrinsic inhibition taking into account elements of confidence index, confidence band and threshold, as indicated below .

This is how we achieve at least part of the goals mentioned above, by using a detection method in which: a) during at least some of the combustion phases in at least one of the cylinders, at an instant (t) of the engine operating cycles, a current value of an index l _n (t) representative of is calculated operating irregularities over time in the engine, b) while the engine is running, and from the current value at time t of the index l _n (t), the current value is calculated at the same time t d ' an observation function G (t) function of an instantaneous difference between said current value at time t of the index l _n (t) and an estimated reference value l _n ± ι (t) illustrative of this index at a different time, c) the current value of G (t) is compared to a predetermined threshold (S), depending on the load (C) and / or the engine speed (N), and d) a misfire is diagnosed combustion when the value of G (t) exceeds the threshold (S),

- the observation function G (t) being such that: G (t) = Ut) - l _n (t), with l _n (t) function of l _n (t) and l _n ± ι (t) and corresponding to a value representative of the index l _n (t) after exclusion and / or weighting of the values of this index outside a predetermined admissible deviation (BC).

- and before calculating the function G (t), a value E _n (t) is calculated, expression of the instantaneous difference between said current value at time t of the index l _n (t) and the value of corresponding reference l _n ± i (t), illustrative of this index at a different time,

In particular, the index l _n (t) could correspond to an estimated value of the engine torque, at a given moment in the cycle n. In particular, to secure the diagnosis in the event of an uncontrolled drift of a component and to authorize a diagnosis over a very wide operating range of the engine, a complementary characteristic of the invention recommends that l _n (t) be calculated from the formula: Mt) = ï _{± i} (t) + ax E _n (t) where a is a function correcting coefficient (a priori of decreasing monotonic type) of | E _n (t) |, with E _n (t) = l _n (t) - l _n ± ι (t), / being an operating cycle different from

U-

Preferably: - the coefficient a will also be a linear function of the engine ΔN / N dynamics, N being the engine speed, and l _n ± i (t) will be a filtered estimate, for example an average of previous events in which the extraordinary events (leading in particular to function values E _n

(t) or G (t) beyond the threshold S or outside BC) are eliminated or weighted. Furthermore, for the quality of the results, we recommend that n and / be two identical times, preferably two combustion times, of two successive cycles or separated from each other by at most five successive cycles (see on this point comments for engine cycles other than four-cylinder).

Furthermore, in particular to avoid a sensor or auxiliary information for describing the condition of the roadway, provision is made, according to a possible additional characteristic, for a running of a vehicle propelled by said engine on an irregular geometry surface, when the function | E _n (t) | takes several consecutive values between the maximum value at the instant considered of said predetermined admissible deviation (BC) and the threshold S, this diagnosis inhibiting that of misfire in step d). Preferably, the width of this "confidence band" BC will be a function of the variance of the function G (t).

Advantageously, the invention provides that the inhibition of misfire diagnosis is supplemented by the opening of a time window (fixed period of time initiated by the detection of taxiing on an irregular surface) during which no misfire recording combustion is not carried out, since then this event is due to the driving conditions and not to an engine malfunction. This time window continues as long as the engine control detects taxiing on an irregular surface.

In addition, always for the sake of reliability, for diagnostic security and for particularly rapid adaptation to real operating conditions, in the context of the most universal diagnosis possible, we recommend that the index l _n (t) be calculated from a linear combination of at least two elementary indices (h, l ₂ , I3, ...) each representative of the engine operating irregularities in an operating domain (ΔC, ΔN) predetermined engine (i.e. for a speed and a load obtained under substantially identical operating conditions on a reference engine), the elementary indices, (h, I2,, ... ) being weighted, in the combination, by predetermined relevance coefficients (α, β, γ, ...) obtained therefore also under substantially identical operating conditions on a reference engine.

It will also be noted that, in the present description, the term "cycle" corresponds to the time interval separating two top active dead centers from a cylinder, a "top active dead center" (PMHA) corresponding to the angular position of the crankshaft which precedes the expansion of the gases, i.e. the minimum distance between the piston and the top of the cylinder head of this engine.

In relation to this and the times when the readings prior to the index calculation are carried out, it is advised that the values of l _n (t) and l _n ± i (t) calculated correspond to current values, at instants of operation (typically combustion centers located a few degrees after PMHA, their exact position varying according to the type of engine) belonging to two consecutive engine operating cycles, or to cycles distant from each other by more than five consecutive cycles. Thus, the reliability of the detection will be increased. The value of five cycles is given for a four-cylinder engine and corresponds to 720 ° at the crankshaft. For another engine architecture, the optimum number of cycles will be established taking this remark into account. An even more detailed description of the invention will now be provided with reference to the appended drawings in which: FIG. 1 is a flow diagram showing steps of a method in accordance with that of the invention,

- Figure 2 is a graph useful for the description of the diagnosis that is established in accordance with the method of the invention.

In FIG. 1, it appears that the method according to the invention uses signals conventionally delivered by a position sensor 1, with magnetic reluctance for example, sensitive to the movement of the teeth of a toothed wheel 2, or target, fixed on the 'output shaft of an internal combustion engine (not shown). The angular position sensor 1 is thus subject to a fixed part of the engine opposite the wheel 2 which is therefore secured to the crankshaft. Typically, the wheel carries a series of identical regularly spaced teeth and a singular tooth (for example, of width equal to three common teeth) whose absolute position is perfectly known with respect to the crankshaft. This sensor delivers a signal containing, at each revolution of the crankshaft, data on the angular position corresponding to the passage of each current tooth and absolute position information corresponding to the passage of the singular tooth. This data can consist of a series of pulses, after shaping in electronic processing means associated with the sensor. The absolute position data make it possible to know the position of the active high dead centers. A counter receives the signal Se and delivers values of time difference Δt-i, Δt ₂ , ... by a measurement of the times thus separating the passage of predetermined teeth in front of the sensor. These time differences are stored in a memory of the computer provided on the vehicle and processed by the latter. These calculation means are programmed to group the time difference values into sets of samples Δt-i, Δt ₂ , ... from which the critical parameter is calculated.

Thus, after being conformed, the signals delivered by the sensor 1 are used in step 3 to calculate this critical parameter, or index, representative of the engine operating irregularities, which are therefore supposed to result, in particular, from misfire . In fact, this index is representative of the variations in the instantaneous speed of rotation of the engine at the instant of the chosen cycle (preferably a reference instant of the combustion phase), relative to the average speed between two top dead centers consecutive assets, for example.

The calculated index takes the form of a linear combination of time differences Δtj (acquired in step 3) separating predetermined angular positions of the crankshaft of the engine noted therefore thanks to the signal delivered by the sensor 1. Thus, in each window sampling, a series of measurements of time differences, for example angularly distributed on either side of a reference position, is carried out with a view to obtaining the above-mentioned set of samples Δti on said window, after what, for each set of samples a linear combination of these measured Δtj values is carried out, so as to obtain the index concerned (step 5).

As in the French patent FR-A-2 689 934 cited above, at the engine operating / cycle and the instant t of the cycle, this linear combination l _n (t) can be expressed by the relation: lι (t) = - Δti - 3 Δt ₂ - 2 Δt ₃ + 2Δt ₄ + 3Δt ₅ + Δt ₆ with possible weighting by a factor 1 / t ³ , where t is a parameter proportional to the duration of a rotation of the wheel 2. Thus , preferably for each cycle or for a selected series of cycles, a specific instant (for example the “combustion center” positioned a few degrees - depending on the type of engine concerned - after PMHA) is monitored by determining, by linear combinations simple, a parameter representative of the variation of the instantaneous speed of the motor, that is to say the acceleration, at this instant.

We can however consider that this parameter is only an elementary index, the "general" index (still) more representative (called l _n (t)) then being a linear combination of at least two elementary indices, such that (t), l ₂ (t), l ₃ (t), each adapted to a particular engine operating domain, defined by intervals of particular variations ΔC of the load C, and ΔN of the speed N of this engine, each elementary index therefore being obtained from measurement of time differences separating predetermined angular positions of the rotating part secured to the crankshaft.

Such calculations, like those mentioned elsewhere in the present description, are executed by the computer on board the vehicle which manages the operation of the engine and which is duly programmed for this purpose. Thus in step 4 of the flow diagram of FIG. 1, while the vehicle is in current use, the different indices (t), l ₂ (t), l ₃ (t) can be calculated. ., cited above.

In this hypothesis, in the following step 5, the current value l _n (t) of the index retained is calculated such that: ln (t) = α lι (t) ₊ β l ₂ (t) + γ l ₃ (t), where α, β, γ are "relevance" coefficients taken from a memory unit (table) where these coefficients were stored after bench tests as a function of the load C and of the engine speed N. These coefficients weight, in the expression of l _n (t), the associated indices according to the greater or lesser proximity of the instantaneous operating point (C, N) (current) of the motor, with respect to the operating range (ΔC, ΔN) associated with each elementary index. In a known manner, the charge can be measured by any conventional means

(depression in the intake manifold, air flow, ...), while the engine speed can be taken at each revolution of the engine by traditional means.

Whether the calculation of l _n (t) was made from a linear combination of one or more indices, such as elementary indices

'Ι. , I3 ..., we then have, in the computer memory, the values l _n (t), l _n -i (t), even l _{n + i} (t), n and / ^' being two different cycles of operation of the engine, t always being the characteristic instant of the measurement (the combustion center for example), the cycles concerned being preferably preferably consecutive or at least distant from each other by at most five consecutive cycles, the figure of five cycles having been chosen as illustrating the relative proximity in time between the values taken into account for a four-cylinder engine, so as in particular to avoid overloading the computer memory.

The value of five cycles, given for a four-cylinder engine, corresponds to 720 ° at the level of the crankshaft. For another engine architecture, the optimum number of cycles will be established taking this remark into account.

In what follows, we will consider that we have taken into account the indices l _n (t) and l _n- ι (t) at the combustion center (close to PMHA) of two successive cycles (n-1 and n) a four-cylinder engine.

In Step 6, we then calculate a function E _n (t) such that taking into account that in our example, we took into account the values of the indices l _n (t) and l _n- ι (t) on two successive cycles (n-1 and n).

I _n -ι (t) is a filtered value, after passing through a non-linear digital filter. I _n- ι (t) can for example be an average of the preceding samples, it being understood that the extraordinary values, that is to say those deviating (outside of a range of authorized values) from the normally expected value, are either excluded from the calculation, or weighted so as to reduce their effect on l _n-1 (t).

The calculation of the function E _n (t), therefore illustrative of the operation at time t of the cycle n, is then used for the calculation in step 7 of a coefficient of weighting a, linear function of | E „(t) |, and of ΔN / N, that is to say of the dynamics of the motor at the instant t concerned.

In particular, this coefficient a could be such that: a = a ₀ x [1 - min (| E _n (t) | / (nx Variance (Combustion)), 1)] with n = 4 and a ₀ = 0, 2 for example

The weighting coefficient a is then used to calculate in step 8 the value, always at time j of the cycle n in question, of

Tn (t) = Tn_ι (t) + a x E „(t).

And this value is itself used to calculate in step 9 the observation function G (t) such that;

G (t) = r _n (t) - l _n (t)

In step 10, the value (possibly the absolute value) of this observation function is compared with a threshold value S, a function of the speed N and of the load C of the engine at the instant concerned of the cycle in question. The value of the threshold S is typically established by bench tests, so that an overshoot of the threshold by the function G (t) certainly corresponds, at the instant of this overshoot, to the occurrence of a misfire (RC), thus diagnosed (in step 11).

The fault threshold S to which the critical parameter G (t) is compared could be calculated from the engine speed and load. Another solution consists in carrying out and memorizing, during the preliminary bench tests, a mapping of fault thresholds, providing a fault threshold value for each engine speed and each average load applied. At each cycle, while the vehicle is moving normally (and in particular, by a classic measurement at each top dead center), the current engine speed and the average current load applied are then detected and extracted from the memory unit the corresponding fault threshold in order to carry out the comparison mentioned above.

It will also be noted that the mapping of fault thresholds may have been defined beforehand on a bench by artificially causing a combustion fault at predefined speeds and loads, by carrying out the aforementioned measurements of time differences (Δtj) and by calculating the index or indices corresponding reviews, and also memorizing part of these indices as fault threshold (s). A threshold overshoot, at a particular cycle i, is illustrated by the reference point RC in FIG. 2 where a possible graph of evolution of the threshold S has been represented, over time (or according to the succession of cycles), in function of changes in engine speed N and engine load C. FIG. 2 also shows a "confidence band" BC corresponding to admissible deviations in the variation of | E _n (t) | which it is unlikely that | E _n (t) | so, in the absence of extraordinary accidents likely to affect the regularity of the operation of the engine (misfires, rolling on irregular surfaces, ...). This confidence band is therefore typically a function of the normal variance of | E _n (t) |.

If, in cycle n (see FIG. 2, point referenced Rl _n and test step 12 in FIG. 1), the function of | E _n (t) | takes a value between the confidence band BC (more exactly the maximum value of the confidence band at the instant concerned) and the threshold S, a step irregularity Rl _n is diagnosed in step 13. A check is then carried out on cycle n-1 (step 14) in order to check whether an operating irregularity Rl _n -ι meeting the same conditions (BC <| E _n- ι (t) | <S) had been detected.

If during consecutive cycles n-1 and n the values of | E _n- ι (t) | and of | E _n (t) | have generated operating irregularities, then the final RI diagnosis (step 15) is issued. The anomaly detected is then due to the rolling of the vehicle powered by this engine on an irregular surface, rolling which causes a rise in forces on the crankshaft of the engine, disturbing the regularity of rotation of the latter.

Simultaneously with the final diagnosis RI of step 15, an index FR is activated (memory passing from the value 0 to the value 1 for example). This index allows the computer to trigger a time window of a fixed duration, for example 300 ms, during which no RC diagnosis (combustion fault) will be carried out and recorded (step 11), since the computer has highlighted that the reason for the extraordinary event is rolling on an uneven surface. Thus, the point RC _{n +} ι (FIG. 2) of the cycle n + 1 immediately following the cycle n will not be recorded as a misfire.

If taxiing on an irregular surface continues, a succession of RI diagnostics is generated, which inhibits step 11 and initializes as many times the FR index, thus extending the duration of the time window during which the RC diagnosis is not recorded.

During the entire time of this inhibition, the computer will continue its operations and checks, and as soon as two consecutive values are no longer indicative of a run on an irregular surface, the misfire diagnosis can be put back in place, once elapsed the time window generated by the index FR.

For the proper functioning of the diagnostic process presented above, a reset (reset) of the RC and RI and FR alert emissions will be carried out before each iteration of calculations.

It now appears that the invention makes it possible to achieve the stated aims, namely in particular to provide a reliable diagnosis with regard to the occurrence of misfires in the combustion of an internal combustion engine, and this whatever the circumstances likely to affect the safety of this diagnosis.

This advantageous result is therefore obtained on the one hand thanks to the adaptation described above of the sampling of the index _ \ considered, taken into account in the calculation of [, to the various evolutions which can affect the operation of the engine. , whether these changes are normal (regime transitions), accidental (misfires, rolling on irregular surfaces) or systematic (wear, daggerboards, ...) and secondly, thanks to the adaptation of the index I himself to these same developments.

Claims

1.- Method for detecting misfires of a mixture between a fuel and an oxidizer, in any one of the cylinders of an internal combustion engine having operating cycles at several times, such as two times or four time, this process comprising the following steps: a) during at least some of the combustion phases in at least one of the cylinders, at an instant (t) of the engine operating cycles, a current value of one is calculated index l _n (t) representative of operating irregularities over time of the engine, b) while the engine is running, and from the current value at time t of the index l _n (t), we calculate the current value at the same time t of an observation function G (t) function of an instantaneous difference between said current value at time t of the index l _n (t) and an estimated reference value l _n ± ι (t) illustrative of this index at a different time, c) we compare l a current value of G (t) at a predetermined threshold (S), depending on the load (C) and / or the engine speed (N), and d) a misfire is diagnosed when the value of G (t ) exceeds the threshold (S), characterized in that:

- the observation function G (t) is such that: G (t) = Mt) - I _n (t), with Y _n {t) ≈ Tn ± ι (t) + ax E _n (t) l _n (t) corresponding to a value representative of the index l _n (t) after exclusion and or weighting of the values of this index outside a predetermined admissible deviation (BC), and where a is a correction coefficient function of | E _n ( t) | , and E _n (t) = l _n (t) - l _n ± i (t), / being an operating cycle different from n, E _n (t) being an expression of the instantaneous difference between said current value at the instant t of the index l _n (t) and the corresponding reference value l _n ± _\ (t), illustrative of this index at a different time.

2.- Method according to claim 1, characterized in that the coefficient a is also a linear function of the dynamic ΔN / N of the engine, N being the speed of this engine.

3.- Method according to any one of the preceding claims, characterized in that l _n ± ι (t) is an average of the values I (t) representative of the engine operating irregularities over time, this average eliminating or weighting the values that deviate from the expected normal value.

4.- Method according to any one of the preceding claims, characterized in that n and / are two identical engine times, preferably two combustion times, of two successive cycles or separated from each other by at most 720 ° rotation of the crankshaft.

5.- Method according to any one of the preceding claims, characterized in that a running of a vehicle propelled by said engine is diagnosed on a surface with irregular geometry, when the function | E _n (t) | takes at least two consecutive values between the maximum value at the instant considered of said predetermined admissible deviation (BC) and the threshold S, this diagnosis inhibiting that of misfire in step d).

6.- Method according to claim 5, characterized in that the predetermined admissible deviation (BC) has a width which is a function of the variance of the function | E _n (t) |.

7.- Method according to claim 5 or claim 6, characterized in that the running diagnosis of a vehicle powered by the engine on an irregular geometry surface generates a period of time during which no misfire diagnosis is is saved.

8.- Method according to claim 7, characterized in that the period of time during which no misfire diagnosis is recorded is extended as long as rolling on an irregular geometry surface continues and is detected.

9.- Method according to any one of the preceding claims, characterized in that the index l _n (t) corresponds to an estimate of the value of the torque of the motor, at a given moment of the cycle n.

10.- Method according to any one of the preceding claims, characterized in that the index l _n (t) is calculated from a linear combination of at least two elementary indices (li, l ₂ , l _3ι . ..) each representative of the engine operating irregularities in a predetermined operating range (ΔC, ΔN) of this engine as a function of the load (C) and / or of the speed (N) at a given time, the elementary indices, ( h, l ₂ , l ₃ , ...) being weighted, in the combination, by predetermined relevance coefficients (α, β, γ, ...) obtained under substantially identical operating conditions on a reference engine.