METHOD FOR POSITIVE CRANKSHAFT VENTILATION DIAGNOSIS

Abstract

A method for positive crankcase ventilation diagnosis in an engine system with an engine, an intake duct connected to the engine, a PCV line connecting a crankcase of the engine with the intake duct and a sensor for determining an actual PCV pressure (p.sub.act) in the PCV line. The method comprises the steps of: performing a diagnosis measurement by repeatedly determining at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q), which is linked to the operation of the engine system, and the actual PCV pressure (p.sub.act), to obtain a plurality of data samples for a plurality of sample times (t.sub.s); using a prediction model (M) to determine a predicted PCV pressure (p.sub.pre); and comparing the actual PCV pressure (p.sub.act) with the predicted PCV pressure (p.sub.pre) to diagnose the PCV line, wherein a diagnosis of the PCV line is based on a model error (p.sub.err).

Claims

1. A method for positive crankcase ventilation diagnosis in an engine system with an internal combustion engine, an intake duct connected to the engine, a PCV line connecting a crankcase of the engine with the intake duct and a sensor for determining an actual PCV pressure (p.sub.act) in the PCV line, the method comprising the steps of: performing a diagnosis measurement during a measurement period by repeatedly determining at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q), which is linked to the operation of the engine system, and the actual PCV pressure (p.sub.act), to obtain a plurality of data samples for a plurality of sample times (t.sub.s); using a prediction model (M) to determine a predicted PCV pressure (p.sub.pre) based on the determined at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q); and comparing the actual PCV pressure (p.sub.act) with the predicted PCV pressure (p.sub.pre) to diagnose the PCV line, wherein a diagnosis of the PCV line is based on a model error (p.sub.err), which is the difference between the predicted PCV pressure (p.sub.pre) and the actual PCV pressure (p.sub.act), wherein the model error (p.sub.err) being below a lower threshold (p.sub.lt) is considered to indicate that the PCV line is intact, the model error (p.sub.err) being above an upper threshold (p.sub.ut) is considered to indicate that the PCV line is damaged, and the model error (p.sub.err) being between the lower threshold (p.sub.lt) and the upper threshold (p.sub.ut) is considered as an inconclusive result.

2. The method according to claim 1, wherein the diagnosis is based on a comparison between a first number of sample times (t.sub.s) with a model error (p.sub.err) indicating that the PCV line is intact and a second number of sample times (t.sub.s) with a model error (p.sub.err) indicating that the PCV line is damaged.

3. The method according to claim 1, comprising, before the diagnosis measurement, performing at least one set-up measurement by determining the actual PCV pressure (p.sub.act) and the at least one input parameter (p.sub.init, n.sub.e, p.sub.atm . . . , Q) in an intact engine system; and determining the prediction model (M) at least partially based on the at least one set-up measurement.

4. The method according to claim 1, wherein the at least one set-up measurement is performed during a set-up period by repeatedly determining the actual PCV pressure (p.sub.act) and the at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q) in an intact engine system, to obtain a plurality of set-up data samples for a plurality of set-up sample times.

5. The method according to claim 1, wherein at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q) changes during at least one set-up measurement.

6. The method according to claim 1, wherein the prediction model (M) is determined based on a neural network trained with set-up data samples corresponding to at least one set-up measurement, in which the actual PCV pressure (p.sub.act) and the at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q) in an intact engine system have been determined.

7. The method according to claim 1, wherein a calibration is performed for the actual PCV pressure (p.sub.act).

8. The method according to claim 1, wherein at least one input parameter is selected from among an engine speed (n.sub.e), an intake pressure (p.sub.int), an intake gas flow (Q), an atmospheric pressure (p.sub.atm), a coolant temperature of a cooling system of the engine, and a throttle position of a throttle in the intake duct.

9. The method according to claim 1, wherein an actual PCV pressure (p.sub.act) of a sample time (t.sub.s) in the measurement period is only used for the diagnosis if at least one enabling condition is fulfilled for this sample time (t.sub.s), and is disregarded otherwise.

10. The method according to claim 9, wherein at least one enabling condition is that the engine speed (n.sub.e) is above a predefined minimum speed, the difference between the intake pressure (p.sub.int) and atmospheric pressure is above a predefined minimum pressure, the intake gas flow (Q) is above a predefined minimum flow and/or the atmospheric pressure (p.sub.atm) is above a predefined minimum pressure.

11. The method according to claim 9, wherein one enabling condition is that a predefined first delay time has passed since a transient period of at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q).

12. The method according to claim 9, wherein the actual PCV pressure (p.sub.act) is only used if the at least one enabling condition has been fulfilled for a predefined second delay time before the sample time (t.sub.s).

13. The method according to claim 1, wherein a low-pass filter is applied to at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q).

14. The method according to claim 1, wherein a low-pass filter is applied to the model error (p.sub.err) before the diagnosis.

15. A diagnosis unit for positive crankcase ventilation diagnosis in an engine system with an engine, an intake duct connected to the engine, a PCV line connecting a crankcase of the engine with the intake duct and a sensor for determining an actual PCV pressure (p.sub.act) in the PCV line, the diagnosis unit being configured to: perform a diagnosis measurement during a measurement period by repeatedly determining at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q), which is linked to the operation of the engine system, and the actual PCV pressure (p.sub.act), to obtain a plurality of data samples for a plurality of sample times (t.sub.s); use a prediction model (M) to determine a predicted PCV pressure (p.sub.pre) based on the determined at least one input parameter (p.sub.int, n.sub.e, p.sub.atm, Q); and compare the actual PCV pressure (p.sub.act) with the predicted PCV pressure (p.sub.pre) to diagnose PCV line, wherein a diagnosis of the PCV line is based on a model error (p.sub.err), which is the difference between the predicted PCV pressure (p.sub.pre) and the actual PCV pressure (p.sub.act), wherein the model error (p.sub.err) being below a lower threshold (p.sub.lt) is considered to indicate that the PCV line is intact, the model error (p.sub.err) being above an upper threshold (p.sub.ut) is considered to indicate that the PCV line is damaged, and the model error (p.sub.err) being between the lower threshold (p.sub.lt) and the upper threshold (p.sub.ut) is considered as an inconclusive result.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Preferred embodiments of the invention will now be described, by way of example, with reference to the accompanying drawings, in which:

[0042] FIG. 1 is a principle diagram of an engine system and an inventive diagnosis unit;

[0043] FIG. 2 is a flowchart of an inventive method;

[0044] FIG. 3 is a first diagram showing a model error versus an intake pressure; and

[0045] FIG. 4 is a second diagram showing the model error versus the intake pressure.

DESCRIPTION OF PREFERRED EMBODIMENTS

[0046] FIG. 1 shows a principle diagram of an engine system 1 of a vehicle and an inventive diagnosis unit 20. The engine system 1 comprises an internal combustion engine 2, e.g., a turbocharged gasoline engine. It will be understood that the engine 2 normally has a plurality of cylinders, of which only one is shown in this representation. A combustion chamber 3 is connected to an intake duct 5 and an exhaust duct 6, each of which can be temporarily isolated from the combustion chamber 3 by an intake valve or exhaust valve, respectively. During the combustion cycle, air is taken in through the intake duct 5 into the combustion chamber 3 and fuel, e.g., gasoline, is injected. Afterwards, the combustion is initiated, and the exhaust gases are expelled through the exhaust duct 6. However, a portion of the exhaust gases escaped from the combustion chamber 3 past the cylinder and into a crankcase 4 of the engine 2. These blow-by gases need to be vented from the crankcase 4 via a PCV line 10, which comprises a non-return valve 11 to prevent backflow into the crankcase 4. The PCV line 10 is connected to the crankcase 4 and to the intake duct 5 upstream of a compressor 7 that is part of a turbocharger. Downstream of the compressor 7, a charge-air cooler 8 and a throttle 9 are disposed on the intake duct 5. A first pressure sensor 12 is disposed in the PCV line 10 to measure an actual PCV pressure pa, while a second pressure sensor 13 is disposed on the intake duct 5 to measure an intake pressure p.sub.int, which may also be referred to as a boost pressure. An intake gas flow Q inside the intake duct 5 can also be measured by a suitable sensor (like a moving vane meter or hot wire sensor) that is not shown for sake of simplicity.

[0047] FIG. 1 also shows a schematic representation of a diagnosis unit 20 which may be integrated in the vehicle. The diagnosis unit 20 is adapted to perform a diagnosis of the engine system 1 or, more specifically, of the PCV line 10. To this end, a prediction model M is implemented in the diagnosis unit 20. Based on this prediction model M, the diagnosis unit 20 can determine a predicted PCV pressure p.sub.pre based on several input parameters. In the present case, these input parameters include an engine speed n.sub.e, of the engine 2, the intake pressure p.sub.int, the intake gas flow Q, a coolant temperature, and a position of the throttle 9.

[0048] FIG. 2 is a flow diagram illustrating an embodiment of the inventive method. After the start, the prediction model M needs to be determined. To this end, at least one set-up measurement is performed at step 100. For each set-up measurement, the time evolution of the actual PCV pressure p.sub.act and all input parameters n.sub.e, p.sub.int, Q is measured using an intact engine system. This is identical to the engine system 1 shown in FIG. 1, except for that integrity of the PCV line 10 is known in the intact engine system. The at least one set-up measurement is performed during a set-up period that may have a length of several minutes or even hours. With the set-up data samples recorded during the set-up measurement for a plurality of set-up sample times, the prediction model M is determined (at 105) using a machine learning method, e.g., using a neural network like a 2-layer feedforward Perceptron. The prediction model M establishes a connection between the input parameters and the actual PCV pressure p.sub.act. As a rule, the set-up measurement(s) and the determination of the prediction model M are not performed by the diagnosis unit 20, but by a set-up unit (not shown) that is not part of the vehicle. The set-up unit may comprise a desktop computer or a big data processing computer and its computing power is normally considerably greater than the computing power of the diagnosis unit 20. When the prediction model M has been determined, it can be transferred to the diagnosis unit 20.

[0049] Since the diagnosis of the PCV line 10 depends on an accurate measurement of the actual PCV pressure p.sub.act, the first pressure sensor 12 is calibrated at 110 before the diagnosis starts. Calibration can be performed when the engine 2 is off so that the actual PCV pressure p.sub.act should correspond to an atmospheric pressure p.sub.atm in the vicinity of the engine system 1. Then, at 115, at least one diagnosis measurement is performed on the engine system 1 during a measurement period, which may take several minutes or even hours.

[0050] During the diagnosis measurement, the input parameters n.sub.e, p.sub.int, Q and the actual PCV pressure pa are determined repeatedly. Thus, a plurality of data samples is obtained, each of which is associated with a corresponding sample time t.sub.s. The data samples are stored in a memory device of the diagnosis unit 20. After completion of the diagnosis measurements, in step 120, a low-pass filter is applied to one or several input parameters n.sub.e, p.sub.int, Q. In particular, such a low-pass filter may be applied to the intake pressure p.sub.int, thereby eliminating high-frequency components. It should be noted that an equivalent low-pass filter is usually applied to one or several input parameters n.sub.e, p.sub.int, Q that are measured during the set-up measurement(s) at step 100 before the prediction model M is determined at step 105. At step 125, the diagnosis unit 20 uses the prediction model M to determine a predicted PCV pressure p.sub.pre based on the input parameters n.sub.e, p.sub.int, Q that have been measured during the diagnosis measurement.

[0051] Afterwards, the method proceeds with a comparison section 130, that is based on a comparison of the predicted PCV pressure p.sub.pre and the actual PCV pressure p.sub.act. In step 135, a model error p.sub.err is determined, which is defined as the difference between the predicted PCV pressure p.sub.pre and the actual PCV pressure p.sub.act. It will be noted that the model error has the dimension of a pressure and depends on the sample time t.sub.s, i.e. the model error p.sub.err generally has different values for different sample times t.sub.s. In step 140, a low-pass filter is applied to the model error p.sub.err to eliminate high-frequency components.

[0052] At step 145, a first sample time t.sub.s is selected, which normally corresponds to the beginning of the measurement period. At step 150, several enabling conditions are checked for this sample time t.sub.s. A first enabling condition is whether the difference between the intake pressure p.sub.int and the atmospheric pressure p.sub.atm is above a predefined minimum pressure, e.g., 0.15 bar. A second enabling condition is whether the engine speed n.sub.e is above a predefined minimum speed, e.g., 2500 rpm. A third enabling condition is whether the intake gas flow Q is above a predefined minimum flow. A fourth enabling condition is whether the atmospheric pressure p.sub.atm is above a predefined minimum pressure. Also, if a transient period is identified for one of the input parameters n.sub.e, p.sub.int, Q, in particular for the intake pressure p.sub.int, another enabling condition may be that a first delay time (of e.g. 700 ms) has passed since the end of the transient period. Furthermore, any of the above-mentioned enabling conditions may not only be checked for the respective sample time t.sub.s, but also within a second delay time, of e.g. 800 ms, before this sample time t.sub.s. In this case, the enabling conditions have to be fulfilled for every sample time t.sub.s, starting from 800 ms before the sample time t.sub.s that is currently checked.

[0053] If any of the enabling conditions is not fulfilled, the actual PCV pressure p.sub.act and the model error p.sub.err for this sample time t.sub.s are ignored and the method continues at step 180, were this checked whether the current sample time t.sub.s was the last sample time t.sub.s. If not, the method selects the next sample time t.sub.s at step 185 and returns to step 150. If all enabling conditions are fulfilled, the model error p.sub.err is compared in step 155 with a lower threshold p.sub.lt. If the model error p.sub.err is below this lower threshold p.sub.lt, this is used in step 160 as an indication that the PCV line 10 is intact. However, this is not the final diagnosis result, but only an indication gained from the data for this specific sample time t.sub.s. If the model error p.sub.err is not below the lower threshold p.sub.lt, it is checked at step 165 if the model error p.sub.err is above an upper threshold p.sub.lt. If so, this is used at step 170 as an indication that the PCV line 10 is damaged. If not, i.e., if the model error p.sub.err is between the two thresholds p.sub.lt, p.sub.ut, this is interpreted at step 175 as an inconclusive result. After each of the steps 160, 170 and 175, the method continues with step 180. Accordingly, all sample times t.sub.s are processed sequentially. Afterwards, at step 190, a diagnosis result is determined.

[0054] FIGS. 3 and 4 illustrate, by way of example, the model error p.sub.err versus the intake pressure p.sub.int. FIG. 3 represents an engine system 1 with a damaged PCV line 10. A vertical dashed line represents the enabling condition according to which the intake pressure p.sub.int has to be above a predefined minimum pressure p.sub.min. Any samples not fulfilling this enabling condition, shown as full circles, are disregarded. A portion of the remaining samples, shown as empty circles, represent a model error between the two thresholds p.sub.lt, p.sub.ut, while another portion, shown as x 's, represent a model error above the upper threshold p.sub.ut. In this case, a large number of samples that fulfil the enabling condition are above the upper threshold p.sub.ut, while no samples are below the lower threshold p.sub.lt. In this case, the diagnosis result at step 190 is that the PCV line is damaged, i.e., that it either has a small leak or is completely disconnected.

[0055] In the diagram of FIG. 4, which represents an intact PCV line, there is also a considerable portion of data samples which do not fulfill the enabling condition. A large number of data samples represent a model error p.sub.err that is between the two thresholds p.sub.lt, p.sub.ut and therefore does not allow for any conclusion. A very small first number of data samples is above the upper threshold p.sub.ut, while a considerably larger second number of data samples, represented by crosses, is below the lower threshold p.sub.lt. The diagnosis is then based on a comparison of the first number with the second number. In this case, although there are some data symbols above the upper threshold p.sub.ut, indicating a damaged PCV line, it will be concluded at step 190 that the PCV line is intact, due to the much higher (second) number of data samples below the lower threshold pr.

LEGEND OF REFERENCE NUMBERS

[0056] 1 engine system [0057] 2 engine [0058] 3 combustion chamber [0059] 4 crankcase [0060] 5 intake duct [0061] 6 exhaust duct [0062] 7 compressor [0063] 8 cooler [0064] 9 throttle [0065] 10 PCV line [0066] 11 non-return valve [0067] 12,13 pressure sensor [0068] 20 diagnosis unit