EP2143921A1

EP2143921A1 - Fuel control system for internal combustion engine

Info

Publication number: EP2143921A1
Application number: EP08752044A
Authority: EP
Inventors: Satoshi Kodou
Original assignee: Honda Motor Co Ltd
Current assignee: Honda Motor Co Ltd
Priority date: 2007-05-01
Filing date: 2008-04-24
Publication date: 2010-01-13
Anticipated expiration: 2028-04-24
Also published as: JP2008274876A; EP2143921A4; EP2143921B1; JP4804413B2; WO2008136357A1

Abstract

Intended is to smoothen the dispersion of combustions between a plurality of cylinders satisfactorily. In an internal combustion engine having a plurality of cylinders, a fuel control system for smoothing the dispersion of the combustions between the cylinders detects an output signal (dωact) indicating the running states of the internal combustion engine, sets a reference signal (Fcr#i) for each of the cylinders, takes a correlation between an output signal and the corresponding reference signal for each of the cylinders, and calculates the value of a correlation function (Cr#i). The fuel control system calculates a target value (Cr_trg) for the cylinders, and calculates a correction value (Kcom#i) for correcting a fuel injection rate (Kin#i) to the cylinders so that the value of the correlation function may converge to the target value for each cylinder. Here, one of the cylinders is set as a reference cylinder, and the reference signals set for the individual cylinders are generated on the basis of the signals obtained by extracting the portion corresponding to the reference cylinder, from the output signals.

Description

Technical Field

The present invention relates to a controller for smoothing variations in combustion between a plurality of cylinders of an internal combustion engine.

Background Art

In an internal combustion engine having a plurality of cylinders, if there are variations in combustion between the cylinders, variations occur in not only the torque but also the exhaust gas components between the cylinders, which may deteriorate the emission. Therefore, there has been proposed a manner for compensating for such variations between cylinders.
According to a manner disclosed in Japanese Patent Application Laid-Open No. 2006-161577 , a value of a correlation function that represents, for each cylinder, a correlation between a sensor detection value indicating an air-fuel ratio and a predetermined reference signal is calculated. A value obtained by smoothing the correlation function values for all the cylinders is used as a target value. For each cylinder, a control input into each cylinder is calculated such that the correlation function value of the cylinder converges to the target value.

Disclosure of the Invention

According to the above-mentioned manner, it is assumed that the reference signal is predetermined such that the sensor detection value in the normal operating state has a high similarity to the reference signal. By making the correlation with such a reference signal, the sensor detection value can be controlled to be close to the reference value.
However, the above assumption may not be satisfied depending on the type of the internal combustion engine or the type of the operating state parameter to be detected by the sensor. For example, there may be a case where a magnitude of the waveform is similar between the sensor detection value and the reference signal whereas a shape of the waveform is different between the sensor detection value and the reference signal. In such a case, the correlation function value is lowered and hence the accuracy of determining a control input for compensating for the variations between cylinders may be deteriorated. Here, as described later, a degree of similarity in the magnitude of the waveform indicates a degree of similarity in an amplitude direction, and more specifically indicates how close the area of one waveform and the area of the other waveform are. As described later, a degree of similarity in the shape of the waveform indicates a degree of similarity in a time direction, and more specifically indicates how close in time the magnitude of the value of one waveform and the magnitude of the value of the other waveform in each time step are.
In a case where the internal combustion engine does not implement various combustion forms, or a case where a sensor detection value of the operating state parameter, such as an air-fuel ratio (LAF) sensor, that is hard to be influenced by the combustion form or external forces applied to the crank shaft is based, deviation in the shape of the waveform of the sensor detection value is small. Therefore, the reference signal may be fixed. However, in a case where the internal combustion engine implements various combustion forms such as a diesel engine or an engine that implements a multiple-injection, or a case where a sensor detection value of the operating state parameter, such as a crank angle sensor, that is easy to be influenced by the combustion form or external forces applied to the crank shaft is based, the above problems may occur because the shape of the waveform of the sensor detection value largely varies depending on the operating state. In order to avoid such problems, it is required to set a reference signal for each combustion form, or set a reference signal for each kind of the sensor to be used, which is difficult and may complicate the calculations.
Therefore, a controller for an internal combustion engine is desired that is capable of smoothing variations in combustion between a plurality of cylinders even in the internal combustion engine that implements various combustion forms and operating state. Furthermore, a controller for an internal combustion engine is also desired that is capable of smoothing variations in combustion between a plurality of cylinders independently of the type of the operating state parameter to be detected by a sensor.
According to one aspect of the invention, in an internal combustion engine having a plurality of cylinders, a fuel control apparatus for smoothing variations in combustion between the cylinders comprises a detection means for detecting an output signal (dωact) representing an operating state of the internal combustion engine, a means for setting a reference signal (Fcr#i) for each of the plurality of cylinders, a means for, for each of the plurality of cylinders, making a correlation between the output signal and the corresponding reference signal to calculate a value of a correlation function (Cr#i), a means for determining a target value (Cr_trg) for the plurality of cylinders, and a means for, for each of the plurality of cylinders, calculating a correction value (Kcom#i) for correcting a fuel injection amount (Kin#i) into the cylinder such that the value of the correlation function converges to the target value. Here, one of the plurality of cylinders is set as a reference cylinder. The reference signal that is set for each of the plurality of cylinders is generated based on a signal obtained by extracting from the output signal a portion corresponding to the reference cylinder.
According to the invention, a reference signal for each cylinder is created based on a signal obtained by extracting from the output signal a portion corresponding to a reference cylinder, and then the reference signal is used for calculating a value of the correlation function for each cylinder. Therefore, even if the combustion state of the internal combustion engine varies or the type of the operating state parameter that is detected as the output signal varies, mismatch components, which are caused by differences in shape of the waveform of the output signal with respect to the reference signal, can be removed from the correlation function value. As a result, the correlation function value represents a similarity regarding the magnitude of the waveform between the output signal and the reference signal. The similarity regarding the magnitude of the waveform represents variations in combustion between the cylinders. Therefore, variations in combustion between the cylinders can be smoothed with a better accuracy by correcting the fuel injection amount based on the correlation function value. Such smoothing of combustion variations enables variations in the output characteristics such as an output torque and exhaust gas components to be smoothed between the cylinders. Furthermore, such setting of the reference signal avoids necessities for setting a reference signal for each combustion state and setting a reference signal for each type of the operating state to be detected.
According to one embodiment of the invention, the reference signal is generated in each combustion cycle. Thus, the reference signal can be updated in such a manner as to immediately correspond to changes in the combustion state and operating state in each combustion cycle.
According to one embodiment of the invention, the output signal is a signal representing only changing components in the output (dωraw) of the internal combustion engine. Variations between the cylinders are caused by changing components in the output of the internal combustion engine. Therefore, by using a signal representing only the changing components, the calculation load can be reduced while the accuracy of calculating the correction value can be improved.
According to one embodiment of the invention, the target value (Cr_trg) is the correlation function value that is calculated for the reference cylinder. In doing so, the output characteristics such as an output torque and exhaust gas components for each cylinder can converge to the output characteristics for the reference cylinder.
According to one embodiment of the invention, the correlation function (Cr#i) is normalized by the square of the standard deviation (S_FCR#i) of the reference signal (Fcr#i) to calculate a normalized correlation function (CrH#i). By virtue of such normalization, the correlation function value thus calculated can be quantitatively obtained, and hence variations between the cylinders can be quantitatively handled. Furthermore, because the normalization is implemented with the standard deviation of the reference signal, a difference of the magnitude of the correlation function value for each cylinder with respect to the correlation function value for the reference cylinder can be extracted with a better accuracy.
According to one embodiment of the invention, the calculation of the correlation function value and the correction value is stopped for a predetermined time period when at least one predetermined condition is met. In one embodiment, the predetermined condition includes a case when an engine rotational speed becomes greater than or equal to a predetermined value. The output signal is a signal indicating an angular speed of the engine rotation. This is because a changing amount in the angular speed becomes smaller when the engine rotational speed is high, and hence the accuracy of the output signal may be deteriorated. In another embodiment, the predetermined condition includes a case when the fuel injection amount becomes less than or equal to a predetermined value. When the fuel injection amount is small, an influence on the output of the internal combustion engine is small, and hence the accuracy of detecting variations between the cylinders may be deteriorated. Further, the predetermined time period for the stopping can be set based on a cycle in which the reference signal is determined. Thus, calculating the correlation function value and the correction value is stopped for a time period during which an appropriate value of the correlation function cannot be calculated.
According to one embodiment of the invention, when a second correlation function (CrS#i) calculated by normalizing the correlation fuction (Cr#i) with the standard deviation (S_Fcr#i) of the reference signal (Fcr#i) and the standard deviation (S_dωdiv#i) of the output signal is less than or equal to a predetermined value, calculating the correction value is stopped. When the reference signal and the output signal do not match in the shape of the waveform, the correlation function value includes an influence of such a mismatch, which may deteriorate the accuracy of calculating the correction value. Therefore, in such a case, calculating the correction value is stopped.
According to one embodiment of the invention, for each of the plurality of cylinders, a value in the crank angle region corresponding to the cylinder is cut-out from the output signal (dωact), and a signal where a value in the crank angle region corresponding to the other cylinders is set to zero is generated as a cylinder-based output signal (dωdiv#i). The standard deviation of the output signal used for calculating the above second correlation function is the standard deviation of the cylinder-based output signal. Thus, an influence of the other cylinders is eliminated, thereby improving the accuracy of calculating the second correlation function.
According to one embodiment of the invention, the internal combustion engine is a diesel engine, and the output signal represents an angular speed of the engine rotation. Thus, the correlation function value can be appropriately and timely calculated for a diesel engine having various combustion forms, thereby smoothing variations between the cylinders at a better accuracy.

Brief Description of the Drawings

Fig. 1 is a diagram schematically illustrating an internal combustion engine and its controller according to one embodiment of the present invention.
Fig. 2 is a diagram for explaining problems of determining a correlation function in a conventional art.
Fig. 3 is a diagram for explaining a manner for determining a correlation function according to one embodiment of the present invention.
Fig. 4 is a diagram for explaining a manner for determining a correlation function according to one embodiment of the present invention.
Fig. 5 is a diagram for explaining a manner for determining a correlation function according to one embodiment of the present invention.
Fig. 6 is a block diagram for a controller according to one embodiment of the present invention.
Fig. 7 is a diagram for explaining a manner for extracting an angular speed signal according to one embodiment of the present invention.
Fig. 8 is a diagram for explaining a manner for cutting and dividing the angular speed signal into cylinder-based angular speed signals according to one embodiment of the present invention.
Fig. 9 is a diagram for explaining a manner for generating a reference signal according to one embodiment of the present invention.
Fig. 10 is a flowchart of a control process according to one embodiment of the present invention.

Best Mode for carrying out the Invention

Embodiments of the present invention will be described with reference to the drawings. Fig. 1 is a schematic block diagram of an internal combustion engine (hereinafter referred to as an engine) and its controller according to one embodiment of the present invention.
An electronic control unit (hereinafter referred to as an "ECU") 1 is a computer having a central processing unit (CPU) and a memory. The memory can store one or more computer programs for implementing various controls of a vehicle and data (including one or more maps) used for executing the programs. The ECU 1 receives and computes data transmitted from each part of the vehicle and generates one or more control signals for controlling each part of the vehicle.
In this embodiment, the engine 2 is a diesel engine. The engine 2 has, for example, four cylinders 3a through 3d. First through fourth (#1 - #4) identification numbers are allocated to these cylinders. In the figure, these cylinders are arranged in series. However, the present invention is not limited to such an arrangement.
The cylinders 3a-3d of the engine 2 are coupled to intake manifolds 4a-4d branched from an intake passage 4, respectively, and are coupled to exhaust manifolds 5a-5d, respectively. The exhaust manifolds 5a-5d are connected to an exhaust passage 5 at a collection portion. A fuel injection valve 7a-7d is attached to each of the cylinders 3a-3d so as to face a combustion chamber of the cylinder.
The fuel injection valves 7a-7d are connected to a high pressure pump and a fuel tank (not illustrated) via a common-rail (not illustrated). The high pressure pump raises the pressure of fuel in the fuel tank and feeds the fuel to the fuel injection valves 7a-7d via the common-rail. Each of the fuel injection valves 7a-7d injects the received fuel into the corresponding combustion chamber. The amount of fuel injected by each of the fuel injection valves 7a-7d is controlled according to a control signal from the ECU 1. In this embodiment, the fuel injection is performed in an order of #1, #3, #4, #2 for the four cylinders 3a-3d.
A crank angle sensor 9 is attached to a crankshaft (not shown) of the engine 2. The crank angle sensor 9 outputs a CRK signal and a TDC signal, which are pulse signals, to the ECU 2 in accordance with the rotation of the crankshaft. The CRK (crank) signal is a pulse signal that is output at every predetermined crank angle (for example, every 30 degrees). The ECU1 calculates a rotational speed NE of the engine 2 in accordance with the crank pulse signal. The TDC signal is output at a crank angle associated with the top dead center (TDC) position of the piston (not shown) of each cylinder. The TDC pulse is output at every 180 degrees of the crank angle in the example of the four-cylinder type.
In this embodiment, the ECU 1 determines the fuel injection amount for each cylinder using one or more programs and data (including one or more maps) stored in the memory in response to a signal of the crank angle sensor 9. A control signal following the determined fuel injection amount is sent to the fuel injection valve 7a-7d of each cylinder, to drive the fuel injection valve 7a-7d. In doing so, smoothing the combustion between the cylinders is made, which enables the output characteristics such as an output torque and exhaust gas components to be smoothed.
Here, referring to Figs. 2 through 5, an aim and principle of the present invention will be discussed so as to deepen the understanding of the invention. Fig. 2 shows, for four cylinders #1-#4, an example of (a) a behavior of a signal S based on an output (output signal) of a predetermined sensor (for example, a crank angle sensor) for, for each cylinder, detecting an operating state of the internal combustion engine, (b) a behavior of a predetermined reference signal F, and (c) a correlation function C of the output signal S and the reference signal F in accordance with a conventional manner as described above.
In this embodiment, because the combustion is performed in an order of #1, #3, #4, and #2 during one combustion cycle T1 (720 degrees of the crank angle), the waveform of the output signal S has a form where a peak is generated in an order of #1, #3, #4, and #2. A cycle T2 allocated to the combustion of each cylinder is 180 degrees of the crank angle because there are four cylinders in this embodiment. For example, if an angular speed of the engine rotation is calculated from the output of the crank angle sensor, a waveform as shown in (a) is obtained, whose behavior is falling after rising, in accordance with the combustion of each cylinder.
The reference signal F as shown in (b) is a predetermined signal, and is generated in advance in such a manner as to represent features of the output signal S obtained from the above sensor. In (c), the output signal S is shown over the reference signal F. The value of the correlation function C between the output signal S and the reference signal F is also shown. Here, the correlation function C is calculated in accordance with the following equation (1). The correlation function C is calculated by smoothing a value obtained by integrating the product of the reference signal F and the output signal S over a time period corresponding to one combustion cycle for each cylinder #i (i=1 to 4). N is the number of crank pulses per one combustion cycle, and is 24 in this example. k indicates a time step, and the value of the correlation function C is calculated at time intervals. In the figure, the values of the correlation function C for the four cylinders are shown over each other. $\begin{array}{l} [Number 1] \\ C#i (k) = \frac{1}{N} Σ_{j = k - N + 1}^{k} F#i (j) \cdot S#i (j) \end{array}$
In a first combustion cycle from time t0 to t1, the output signal S and the reference signal F match, and hence the similarity in the magnitude and the shape of the waveform is high. Here, the similarity in the magnitude of the waveform indicates a degree of similarity in an amplitude direction between both waveforms. The similarity in the shape of the waveform indicates a degree of similarity in a time direction between both waveforms. More specifically, the similarity in the magnitude of the waveform indicates how close the area of one waveform and the area of the other waveform are. The similarity in the shape of the waveform indicates how close in time the magnitude of the value of one waveform and the magnitude of the value of the other waveform in each time step are. The value of the correlation function C reflects not only the similarity in the magnitude of the waveform but also the similarity in the shape of the waveform. Therefore, the value of the correlation function C in the first combustion cycle has a relative high value c1.
In a second combustion cycle from time t1 to t2, in comparison with the first combustion cycle, the output signal S varies because of, for example, changes in the combustion form. The area of the waveform of the output signal S is almost the same as the area of the waveform of the reference signal, and hence the similarity in the magnitude of the waveform is high. However, as shown by reference numeral 11, a time delay has occurred in rising of the output signal S. As a result, the peak of the output signal S deviates in time from the peak of the reference signal F. As shown in (c), the value c2 of the correlation function C in the second combustion cycle declines in comparison with the value c1 of the correlation function C in the first combustion cycle.
In a third combustion cycle from time t2 to t3, the output signal S further varies. The area of the waveform of the output signal S is almost the same as the area of the waveform of the reference signal F. However, as shown by reference numeral 12, a time delay has further occurred in rising of the output signal S2. As a result, the peak of the output signal S deviates in time from the peak of the reference signal F. As shown in (c), because the similarity in the shape of the waveform of both signals further declines, the value c3 of the correlation function C in the third combustion cycle further declines in comparison with the value c2 of the correlation function C in the second combustion cycle.
Thus, even if the similarity in the magnitude of the waveform between the output signal S and the reference signal F is high, the value of the correlation function C decreases as the similarity in the shape of the waveform decreases. However, variations between cylinders should be mainly determined based on the similarity in the magnitude of the waveform. This is because the area of the waveform represents a magnitude of the output of the cylinder. Even if a time delay occurs in the waveform as shown by reference numerals 11 and 12, the area of the output signal S is the same between the cylinders, and hence it can be said that the combustion has been smoothed between the cylinders. Therefore, if variations between cylinders are determined based on the correlation function C having a value lowered mainly by the mismatch in the shape of the waveform, the determination accuracy may be deteriorated. Preferably, the value of the correlation function C is calculated in such a manner as to reflect mainly the similarity in the magnitude of the waveform, not the similarity in the shape of the waveform.
As described above, in a case where the internal combustion engine that does not have various combustion forms is used and a case where the output of a sensor that is difficult to be influenced by the combustion form or the external force applied to the crankshaft is used, the shape of the waveform of the output signal S is almost constant, and the reference signal may be fixed. However, in a case where the internal combustion engine having various combustion forms, such as a diesel engine and an engine that implements a multiple-injection, is used and a case where a sensor that is easy to be influenced by the combustion form or the external force applied to the crankshaft is used, the problems as described above may occur because the shape of the waveform largely changes dependently on the operating state.
Therefore, in the present invention, the reference signal is not generated as a predetermined signal. Instead, the reference signal is generated from the output of the sensor (that is, actual value). Referring to an example in Fig. 3, the cylinder #1 is set as a reference cylinder. The waveform of the output signal S of the cylinder #1 in each combustion cycle is used as the reference signal for all the cylinders. More specifically, in each combustion cycle, the reference signals F for the cylinders #3, #4 and #2 are generated by shifting in time the waveform of the cylinder #1 shown by reference numeral 13.
For each cylinder, the value of the correlation function C between the reference signal F thus generated and the output signal S is calculated. Because the reference signal F is generated using the actual value, the value c1 of the correlation function C is maintained and it is prevented that the correlation function value declines, even when the shape of the waveform of the output signal S changes over the first through the third combustion cycles.
In this embodiment, as shown in the equation (1), the correlation function C is based on a value obtained by integrating the product of the output signal S and the reference signal F over a time period corresponding to one combustion cycle. Therefore, as shown by reference numeral 15, at timing at which the shape of the waveform is switched (that is, timing at which the first combustion cycle is switched to the second combustion cycle, and timing at which the second combustion cycle is switched to the third combustion cycle), a temporary disorder appears in the value of the correlation function C (which is removable as described later). However, it should be noted that the value of the correlation function C is generally maintained constant. For example, it is assumed that the present time is a time point indicated by tx3. The value of the correlation function C for the cylinder #1, which is calculated at the time point tx1 one combustion cycle before, differs from the value of the correlation function C of the cylinder #1, which is calculated at the present time point tx3, because the shape of the waveform changes between the first and second combustion cycles. That is, the value of the correlation function C calculated at tx3 is less than the value of the correlation function C at tx1. However, because the area of the waveform 16 is the same as the area of the waveform 17, the value c1 of the correlation function C calculated at the time point tx4 becomes equal to the value c1 of the correlation function C calculated at the time point tx2. Thus, at the end of the cycle T2 of each cylinder, the value of the correlation function C has converged to the constant value c1.
Fig. 4(a) indicates a case where the magnitude of the waveform of the output signal S largely changes because of, for example, changes in the combustion form. The reference signal F is generated based on the output signal S of the reference cylinder #1 as described referring to Fig. 3. The output signal S and the reference signal F match, and both are shown over each other for the sake of simplicity.
As the area of the output signal S increases, the value of the correlation function C increases as shown by c1 to c3. Thus, if the value of the correlation function C increases, the correlation function has different values over the first through the third combustion cycles even if there is no variations between the cylinders. This may deteriorate the accuracy of determining variations between the cylinders. This is almost trivial when the magnitude of the output signal remains almost unchanged such as a LAF sensor. However, this becomes a problem when the magnitude of the output signal changes at any time dependently on the operating state such as a crank angle sensor. It is preferable that the value of the correlation function is quantitatively obtained.
Therefore, in a more preferable embodiment, the correlation function C is normalized with the square of the standard deviation of the reference signal F. Because the correlation function C can be considered as the covariance of the reference signal F and the output signal S as shown by the equation (1), the correlation function C can be quantitatively calculated by normalizing it with the square of the standard deviation of the reference signal F. Fig. 4(b) indicates the value of the correlation function C' thus normalized. Even when the magnitude of the output signal S changes, the value of the normalized correlation function C' is constant.
Furthermore, Fig. 5 indicates a case where the magnitude and the shape of the waveform of the output signal simultaneously change. (a) indicates the value of the un-normalized correlation function C. (b) indicates the value of the normalized correlation function C'. The reference signal F is generated from the output signal S of the cylinder #1 in each combustion cycle as described referring to Fig. 3. The output signal S and the reference signal F match, and hence both are shown over each other for the sake of simplicity.
As shown in (a), when the correlation function is not normalized, a disorder appears in the correlation function C as described referring to Fig. 3 at timing at which the shape of the waveform of the output signal S is switched. However, such a disorder is not based on the mismatch in the shape of the waveform between the output signal S and the reference signal F. The waveform of the output signal S and the waveform of the reference signal F match. Thus, such a disorder can be removed by normalizing the correlation function as shown in (b).
Next, a specific embodiment for implementing the above discussion referring to Figs. 3 through 5 will be described. Fig. 6 shows a block diagram of a fuel control apparatus according to one embodiment of the invention. Each block is implemented in the ECU 1.
A crank pulse from the crank angle sensor 9 is received by a changing amount extracting part 51 at every 30 degrees of the crank angle. The changing amount extracting part 51 calculates an angular speed dωraw (rad/sec) of the engine based on the received crank pulse. The angular speed dωraw (rad/sec) can be calculated from the time intervals at which the crank pulse is issued from the crank angle sensor 9.
Here, referring to Fig. 7(a), an example of the angular speed signal dωraw is shown. k indicates a time step, which is typically represented in terms of crank angle. In this embodiment, the combustion is performed in the order of #1, #3. #4. #2 in one combustion cycle (720 degrees of the crank angle) as described above. The angular speed dωraw rises and then falls due to the air-fuel mixture combustion in each cylinder.
In the angular speed signal dωraw as shown in Fig. 7(a), the changing amount extracting part 51 sets the angular speed at the beginning of the combustion in each cylinder (which corresponds to the fuel injection timing plus the compression ignition delay for each cylinder, in this embodiment) to zero as shown by reference numeral 61, and extracts a difference (changing amount) of the angular speed signal dωraw with respect to zero. That is, the changing amount of the angular speed dωraw is extracted with the zero line to obtain an angular speed signal dωact as shown in Fig. 7(b).
Alternatively, the changing amount extracting part 51 may calculate an average value of the angular speed dωraw for all the cylinders in each combustion cycle, as shown by reference numeral 62. The average value thus calculated may be set to zero. By extracting the changing components of the angular speed dωraw with respect to zero, the angular speed signal dωact as shown in Fig. 7(c) can be obtained.
The purpose of the changing amount extracting part 51 is to obtain a signal representing, for each cylinder, the operating state of the engine, more specifically the output of the engine (this indicates an output in correlation with the fuel amount, and hence includes, for example, torque, air-fuel ratio, in-cylinder pressure). As described above, because variations between the cylinders appear in the changing components in the output, the steady components are not required. By extracting only the changing components, the amount of variations between the cylinders forms a large proportion of the absolute value of the correlation function as described later, which can improve the calculation accuracy.
Returning back to Fig. 6, a cylinder-based signal extracting part 52 cuts and divides the angular speed signal dωact into cylinder-based angular speed signals. In this embodiment, because four cylinders are provided, the combustion is performed in the order of #1, #3, #4, and #2 every 180 degrees of the crank angle. Therefore, the angular speed signal dωact is cut-out at every 180 degrees of the crank angle.
Referring to Fig. 8, the angular speed signal dωdiv#1 for the cylinder #1, the angular speed signal dcωdiv#3 for the cylinder #3, the angular speed signal dωdiv#4 for the cylinder #4, and the angular speed signal dωdi#2 for the cylinder #2 are cut-out from the angular speed signal dωact. As shown in the figure, the angular speed signal dωdiv# for each cylinder is generated such that it has zero value in the crank angle region corresponding to the other cylinders. For example, the angular speed signal dωdiv#1 has a value only in the crank angle region from zero to 180 degrees, and is zero in the crank angle region from 180 to 720 degrees.
Returning back to Fig. 6, a reference signal generating part 53 sets one of the cylinders (in this embodiment, cylinder #1) as a reference cylinder. Here, referring to Fig. 9, the angular speed signal dωdiv#1 of the cylinder #1 in each combustion cycle is set as a reference signal Fcr#1 corresponding to the cylinder #1.
By retarding the reference signal Fcr#1 by 540 degrees of the crank angle through a delay circuit 53b (Fig. 6), a reference signal Fcr#2 corresponding to the cylinder #2 is generated (Z^-18 of the delay circuit 53b indicates shifting 18 times in synchronization with the crank pulse that is obtained at every 30 degrees of the crank angle).
Furthermore, by retarding the reference signal Fcr#1 by 180 degrees of the crank angle through a delay circuit 53c, a reference signal Fcr#3 corresponding to the cylinder #3 is generated. By retarding the reference signal Fcr#1 by 360 degrees through a delay circuit 53d, a reference signal For#4 corresponding to the cylinder #4 is generated. Thus, the reference signal for each cylinder is generated based on the angular speed signal dωdiv#1 of the reference cylinder in each combustion cycle.
Fcr#2 through Fcr#4 may be generated in another delay method. For example, Fcr#2 may be generated by retarding Fcr#3 by 360 degrees.
Returning back to Fig. 6, a correlation function calculating part 54 calculates, for each cylinder, a correlation function Cr#i between the angular speed signal dωdiv#i obtained for the cylinder and the reference signal Fcr#i generated for the cylinder, in accordance with the equation (2). Here, "i" indicates a cylinder number, and takes values of 1 through 4 in this embodiment. N indicates the number of crank pulses per one combustion cycle (720 degrees of the crank angle). In this embodiment, because the crank pulse is acquired at every 30 degrees, N is equal to 24. k indicates a time step as described above. $\begin{array}{l} [Number 2] \\ Cr#i (k) = \frac{1}{N} Σ_{j = k - N + 1}^{k} Fcr#i (j) \cdot d ω div#i (j) \end{array}$
Thus, the correlation function Cr#i for each cylinder represents an average of a value obtained by integrating, over a time period having a length of one combustion cycle, the product of the reference signal Fcr#i generated for the cylinder and the angular speed signal dωdiv#i extracted for the cylinder. As the similarity between the waveform of the reference signal Fcx#i and the waveform of the angular speed signal dωdiv#i is higher, the correlation function Cr#i takes a higher value.
Furthermore, the correlation function calculating part 54 calculates, for each cylinder, the standard deviation S_FCR#i of the reference signal Fcr#i in accordance with the equation (3), and normalizes the correlation function Cr#i with the square of the standard deviation S_FCR#i in accordance with the equation (4) to calculate a normalized correlation function CrH#i. This normalization enables the value of the correlation function to be quantitatively obtained independently of the magnitude of the output signal (the angular speed signal dωact in this embodiment), as described referring to Figs. 4 and 5. $\begin{matrix} S_{FCR} #i (k) = \sqrt{\frac{1}{N} Σ_{j = k - N + 1}^{k} Fcr#i (j) \cdot Fcr#i (j) ()} & (3) \\ CrH#i (k) = \frac{Cr#i (k)}{S_{FCR} #i (k) \cdot S_{FCR} #i (k)} & (4) \end{matrix}$
A control part 55 sets the normalized correlation function CrH#1 for the reference cylinder #1 in a target value Cr_trg as shown in the equation (5). Because the waveform of the angular speed signal and the waveform of the reference signal match for the reference cylinder, an appropriate correlation function value is determined. Therefore, by setting this correlation function value in the target value, a fuel injection amount can be appropriately corrected such that the output characteristics of the other cylinders converge to the output characteristics of the reference cylinder. $\begin{array}{l} [Number 4] \\ {Cr_trg} (k) = CrH# 1 (k) \end{array}$
A predetermined control method is used to calculate, for each cylinder, a correction coefficient Kcom#1 for causing the normalized correlation function CrH#i to converge to the target value Cr_trg. In this embodiment, a response assignment control is used. According to the response assignment control, the convergence speed can be specified. Preferably, a 2-degree-freedom response assignment control is used. The 2-degree-freedom response assignment control is a control capable of individually specifying the convergence speed of the controlled variable with respect to a target value and the convergence speed of a difference caused when disturbance is applied. In order to perform this control, the target value Cr_trg is filtered in accordance with the following equation. Here, R is a parameter that represents the convergence speed of the controlled variable with respect to the target value as described above, and is preferably set to satisfy -1<R<0. This filtering converts the waveform of the target value into a waveform having asymptotic characteristics, which enables the controlled variable to smoothly converge to the target value. $\begin{array}{l} [Number 5] \\ {Cr_trg_f} (k) = - R \cdot Cr_trg_f (k - 1) + (1 + R) \cdot Cr_trg (k) \end{array}$
The control part 55 further defines, for each cylinder, a switching function σ#i as shown by the equation (7). Here, S is a parameter representing the convergence speed of the difference when disturbance is applied as described above, and is preferably set to satisfy -1<S<0. E#i is the difference between the normalized correlation function CrH#i and the filtered target value Cr_trg_f, as indicated by the equation (8). $\begin{array}{l} σ #i (k) = E#i (k) + S \cdot E#i (k - 1) & (7) \\ E#i (k) = CrH#i (k) - Cr_trg_f (k) & (8) \end{array}$
An equation of the switching function σ#i=0 is called an equivalent input system, which defines the convergence characteristics of the controlled variable (here, the difference E#i). If σ#i(k)=0, the equation (7) is represented as follows: $E#i (k) = - S \cdot E#i (k - 1)$
The above equation (9) can be represented with a straight line (referred to as a switching line) on a phase plane with E#i(k) on the vertical axis and E#i(k-1) on the horizontal axis. The response assignment control acts to place a state quantity (E#i(k-1), E#i(k)), which is a combination of E#i(k-1) and E#i(k), on the switching line. This enables the state quantity to converge to the origin 0 of the phase plane without being influenced by disturbance and in an extremely stable manner. As the absolute value of the parameter S is smaller, the convergence speed is higher.
The control input or the correction coefficient Kcom#i for causing the difference E#i to converge is calculated in accordance with the equation (10). Here, the first term is a reaching law input that is represented by a proportional term of the switching function σ, which is an input for placing the state quantity on the switching line. The second term is an adaptive law input that is represented by an integral term of the switching function σ, which is an input for placing the state quantity on the switching line while suppressing the steady error. Krch and Kadp are feedback gains, which are predetermined by, for example, simulations. By adding the reaching law input and the adaptive law input, the correction coefficient Kcom#i is calculated. As shown in Fig. 6, because the difference E does not occur for the cylinder #1, the correction coefficient Kcom for the cylinder #1 is not calculated. $\begin{array}{l} [Number 7] \\ Kcom#i (k) = - Krch# \cdot σ# i (k) - Kadp \cdot Σ_{j = 0}^{k} σ# i (j) \end{array}$
A correction part 56 corrects, for each cylinder, a reference value of the fuel injection amount with the correction coefficient Kcom#i thus calculated. The reference value is a value determined in accordance with the operating state, and has been already determined to achieve a desired engine output. For example, the reference value can be determined by referring to a predetermined map (which can be stored in a memory of the ECU 1), based on the requested torque (which is, for example, determined based on the opening degree of the accelerator pedal) and the engine rotational speed. The correction part 56 corrects, for each cylinder the reference value by multiplying the reference value with the correction coefficient Kcom#i calculated for the cylinder to determine the fuel injection amount Kin#i. Because the correction coefficient Kcom is not calculated for the reference cylinder #1, the correction of the reference value is not made and hence the reference value itself is used for the reference cylinder. $Kin#i (k) = Reference value (k) \times Kcom#i (k)$
For each cylinder, a control signal for driving the fuel injection valve is sent to the engine 2 such that the fuel injection amount Kin##i calculated for the cylinder is injected from the fuel injection valve. Thus, the fuel injection amount is controlled so that the combustion between the cylinders is smoothed.
Preferably, an execution conditions determining part 57 is provided. The execution conditions determining part 57 determines one or more execution conditions for the calculation of the correlation function by the correlation function calculating part 54 and one or more execution conditions for the calculation of the correction coefficient by the control part 55.
Prior to the calculation of the correlation function for each cylinder, the execution conditions determining part 57 stops, for a predetermined time period, both the calculation of the correlation function and the calculation of the correction coefficient for the cylinder when at least one of the following conditions 1) through 3) is met. The predetermined time period is preferably a cycle in which the reference signal is determined (updated). In this embodiment, the predetermined time period is a time period corresponding to one combustion cycle (720 degrees of the crank angle in this embodiment). While the calculations are stopped, the value of the correction coefficient Kcom for the cylinder is maintained unchanged (not updated).

1) the angular speed signal for the reference cylinder has not been obtained,
2) the engine rotational speed NE is greater than a predetermined value, and
3) the reference value of the fuel injection amount is less than a predetermined value.
The above 1) is defined because the reference signal cannot be generated when the angular speed signal for the reference signal is not obtained. The above 2) is defined because a changing amount in the angular speed is small when the engine rotational speed NE is high, which may deteriorate the accuracy of detecting the changing amount in the angular speed. The above 3) is defined because influence of the fuel injection amount on changes in the engine rotational speed is small when the reference value of the fuel injection amount is small, which may deteriorate the accuracy of detecting the changing amount in the angular speed.
Preferably, prior to the calculation of the correction coefficient for each cylinder, the execution conditions determining part 57 stops the calculation of the correction coefficient for each cylinder for a predetermined time period (for example, 10 seconds) when at least one of the following conditions 4) and 5) is met, and waits until the operating state becomes stable. Alternatively, the predetermined time period may be a time period corresponding to one combustion cycle. While the calculation of the correction coefficient is stopped, the correction coefficient Kcom for the cylinder is not updated.
4) a changing amount in the normalized correlation function CrH#i(k) is greater than a predetermined value,
5) CrS#i(k) is less than a predetermined value.

Here, the changing amount in the normalized correlation function CrH#i(k) is, for example, represented by a difference between the current value CrH#i(k) of the normalized correlation function and the previous value CrH#i(k-1) of the normalized correlation function.
The above 4) is defined because the probability that a mismatch occurs in the shape of the waveform between the reference signal Fcr#i and the angular speed signal dωdiv#i due to changes in the operating state or the combustion state is high. As described above referring to Fig. 4, the normalization of the correlation function enables the value of the correlation function to be constant even if the magnitude of the waveform of the angular speed signal varies. However, for example, if the operating state or the combustion state rapidly changes between the combustion in the reference cylinder and the combustion in another cylinder, a mismatch may occur between the shape of the waveform of the angular speed signal and the shape of the waveform of the reference signal for the cylinders other than the reference cylinder. As the degree of such a mismatch is larger, the value of the normalized correlation function declines. In such a case, because variations between the cylinders may not be correctly determined as described above referring to Fig. 2, it is preferable that the correction coefficient is not updated.
The above 5) is defined because it indicates that there is a mismatch between the reference signal Fcr#i and the angular speed signal dωdiv#i in the shape of the waveform. Here, CrS is a value obtained by normalizing the correlation function Cr with the reference signal Fcr and the angular speed signal dωdiv in accordance with the equation (12) (Crs is called a second correlation function). $\begin{array}{l} CrS#i (k) = \frac{Cr#i (k)}{S_{d ω div} #i (k) \cdot S_{FCR} #i (k)} & (12) \\ S_{d ω div} #i (k) = \sqrt{\frac{1}{N} Σ_{j = k - N + 1}^{k} d ω div#i (j) \cdot d ω div#i (j)} & (13) \end{array}$
As described above, because the correlation function Cr can be considered as representing the covariance of the reference signal Fcr and the angular speed signal dωdiv, the value of the second correlation function CrS#i is a correlation function value that is not affected by the magnitude of the reference signal Fcr#i or the magnitude of the angular speed signal dωdiv#i. However, if a mismatch occurs between the reference signal and the angular speed signal in the ,shape of the waveform, the value of the correlation function CrS#i is decreased as described above referring to Fig. 2. Therefore, if the value of the second correlation function CrS#i is less than a predetermined value in any one of the cylinders, it is preferable that updating the correction coefficients for all the cylinders is stopped.
Fig. 10 is a flowchart of a fuel control process, which is executed by the CPU of the ECU1, in accordance with one embodiment of the present invention. More specifically, the process is performed by the control apparatus shown in Fig. 6. In this embodiment, the process is performed at every predetermined crank angle (for example, every 30 degrees).
In step S1, a crank pulse is received from the crank angle sensor 9 to generate the angular speed signal dωraw of the engine rotation. Further, as described above referring to Figs. 7(a) and 7(b), th value of the angular speed signal dωraw at the beginning time point of the combustion in each cylinder is set to zero, to cut-out the angular speed signal dωraw with the zero line. Alternatively, as shown in Fig. 7(c), the average of the angular speed signal over the combustion cycle may be set to zero. Thus, the angular speed signal dωact as shown in Figs. 7(b) or 7(c) is extracted.
In step S2, as described above referring to Fig. 8, the angular speed signal dωadiv#i for each cylinder is cut out from the angular speed signal dωact. In step S3, the reference signal Fcr#i is generated for each cylinder.
In step S4, it is determined whether the above conditions 1) through 3) are met. Any one of the conditions 1) through 3) is not met, a first flag is set in step S5. The first flag is associated with a counter (not shown). The counter counts a predetermined time period (for example, 720 degrees of the crank angle) every time the first flag is set. When the predetermined time period has elapsed, the counter is reset to zero.
If all the conditions 1) through 3) are met, it is determined whether the first flag has been reset to zero in step S6. If the determination result of this step is No, it indicates that the above predetermined time period has not elapsed. The process proceeds to step S13. Thus, if any one of the conditions 1) through 3) is not met, the update of the correlation function and the correction coefficient is stopped for the predetermined time period.
If the determination result of step S6 is Yes, it indicates that the predetermined time period has elapsed. In step S7, the correlation function Cr#i is calculated for each cylinder in accordance with the equation (2). In step S8, the normalized correlation function CrH#i is calculated for each cylinder in accordance with the equation (4).
In step S9, it is determined whether the above conditions 4) and 5) are met. If any one of the conditions 4) and 5) is not met, a second flag is set in step S10. The second flag is associated with a counter (not shown). The counter counts a predetermined time period (for example, 10 seconds) every time the second flag is set. When the predetermined time period has elapsed, the second counter is reset to zero.
If both the conditions 4) and 5) are met, it is determined whether the second flag has been reset in step S11. If the determination result of the step S11 is No, it indicates that the predetermined time period has not elapsed. The process proceeds to step S13. Thus, if any one of the conditions 4) and 5) is not met, the update of the correction coefficient is stopped for the predetermined time period.
If the determination result of the step S9 is Yes, in indicates that the predetermined time period has elapsed. In step S12, the response assignment control is performed to calculate the fuel correction coefficient Kcom#i for each cylinder in accordance with the equation (10).
In step S13, the fuel injection amount reference value is corrected with the fuel correction coefficient Kcom#i thus calculated to determine the corrected fuel injection amount Kin#i. Thus, when a fuel injection timing for each cylinder arrives, the fuel injection valve of the cylinder is driven such that fuel is injected in accordance with the corrected fuel injection amount Kin#i. When a cylinder into which fuel is to be injected is the reference cylinder, the fuel injection valve of the reference cylinder is driven such that fuel corresponding to the fuel injection amount reference value is injected.
In the above embodiments, the cylinder-based angular speed signal dωdiv#i is used to calculate the correlation function Cr#i. By determining the cylinder-based angular speed signal dωdiv#i, the accuracy of calculating the second correlation function is improved. Alternatively, the angular speed signal dωact may be used to calculate the correlation function Cr#i. The correlation between the angular speed signal dωact and the reference signal Fcr#5 is made. In this case, dωdiv#i(j) in the equation (2) is replaced with dωact(j). In this embodiment, the value of the correlation function Cr#i is also calculated in such a manner as to reflect, for each cylinder, the similarity between the value in the crank angle region of the cylinder in the angular speed signal dωact and the reference signal Fcr#i of the cylinder.
In the above embodiments, the crank angle sensor is used as a sensor for calculating the correlation function. However, the present invention is not limited to this sensor. The present invention can be applied to various sensors (for example, air-fuel ratio sensor) indicating the operation state of the internal combustion engine (more specifically, indicating the output having a correlation with the fuel amount, as described above). Furthermore, in the above embodiments, a fuel control apparatus is described for a diesel engine. However, the present invention can be applied to another engine (general gasoline engine). In the above embodiments, the internal combustion engine having four cylinders is used as an example. However, the present invention can be applied to the internal combustion engine having any number of cylinders.
The above embodiments are applicable to an internal combustion engine for general purpose use (such as an outboard engine).

Claims

In an internal combustion engine having a plurality of cylinders, a fuel control apparatus for smoothing variations in combustion between the cylinders comprising:
a detection means for detecting an output signal representing an operating state of the internal combustion engine;

a reference signal setting means for setting a reference signal for each of the plurality of cylinders;

a correlation function calculating means for, for each of the plurality of cylinders, making a correlation between the output signal and the corresponding reference signal to calculate a value of a correlation function;

a means for determining a target value for the plurality of cylinders; and

a correction value calculating means for, for each of the plurality of cylinders, calculating a correction value for correcting a fuel injection amount into the cylinder such that the value of the correlation function converges to the target value,
wherein the reference signal setting means sets one of the plurality of cylinders as a reference cylinder,
wherein the reference signal that is set for each of the plurality of cylinders is generated based on a signal obtained by extracting from the output signal a portion corresponding to the reference cylinder.
The fuel control apparatus of claim 1,
wherein the reference signal setting means generates the reference signal in each combustion cycle.
The fuel control apparatus of claim 1 or 2,
wherein the output signal is a signal representing only changing components in the output of the internal combustion engine.
The fuel control apparatus of any one of claims 1 through 3,
wherein the target value is the correlation function value that is calculated for the reference cylinder.
The fuel control apparatus of any one of claims 1 through 4,
wherein the correlation function calculating means further normalizes the correlation function with the square of the standard deviation of the reference signal to calculate a normalized correlation function,
wherein the correction value calculating means calculates, for each of the plurality of cylinder, the correction value for each cylinder used for correcting the fuel injection amount into the cylinder such that the normalized correlation function value converges to the target value.
The fuel control apparatus of any one of claims 1 through 5,
wherein the calculation of the correlation function value and the correction value is stopped for a predetermined time period when at least one predetermined condition is met.
The fuel control apparatus of claim 6,
wherein the predetermined condition includes a case when an engine rotational speed becomes greater than or equal to a predetermined value,
wherein the output signal is a signal indicating an angular speed of the engine rotation.
The fuel control apparatus of claim 6 or 7,
wherein the predetermined condition includes a case when the fuel injection amount becomes less than or equal to a predetermined value.
The fuel control apparatus of any one of claims 6 through 8,
wherein the predetermined time period for the stopping is set based on a cycle in which the reference signal is determined.
The fuel control apparatus of any one of claims 1 through 9,
wherein, for each of the plurality of cylinders, when a second correlation function calculated by normalizing the correlation function with the standard deviation of the reference signal and the standard deviation of the output signal is less than or equal to a predetermined value, calculating the correction value is stopped.
The fuel control apparatus of claim 10, further comprising:
a means for, for each of the plurality of cylinders, cutting out a value in the crank angle region corresponding to the cylinder from the output signal, and generating a signal where a value in the crank angle region corresponding to the other cylinders is set to zero as a cylinder-based output signal,
wherein the standard deviation of the output signal used for calculating the second correlation function is the standard deviation of the cylinder-based output signal.
The fuel control apparatus of any one of claims 1 through 11,
wherein the internal combustion engine is a diesel engine, and
wherein the output signal is a signal based on the output of a sensor attached to a crankshaft of the engine.