JP5488561B2 - Internal combustion engine learning device - Google Patents

Description

  The present invention relates to a learning device for an internal combustion engine that learns values used for correcting various calculations such as the rotational speed of a crankshaft and output torque.

  2. Description of the Related Art Conventionally, there is a technique for measuring an instantaneous rotational speed (instantaneous NE) of a crankshaft based on a crank angle signal output from a crank angle sensor and estimating an output torque of the crankshaft based on fluctuations in the measured instantaneous NE ( Patent Document 1).

  This type of crank angle sensor generally detects a plurality of teeth (detected portions) provided on a rotor that rotates integrally with the crankshaft, and each time the crankshaft rotates a predetermined angle. A tooth part will be detected. The crank angle signal is a pulse signal that switches on / off of the pulse every time it rotates by a predetermined angle.

  Therefore, if the position of the tooth portion with respect to the rotor is deviated from the normal position, an angle error of the crank angle signal is generated, and an instantaneous NE measurement error and an output torque estimation error are generated. As a countermeasure, Patent Document 2 discloses a technique for learning the angle error caused by the processing accuracy of the tooth portion and correcting the crank angle signal, the instantaneous NE or the output torque based on the learned value. ing.

  Specifically, the crank angle (360 CA) for one rotation of the crankshaft is divided into a plurality of sections, the average of instantaneous NE (section NE) in each section is calculated, and the average of sections NE in all sections (overall average NE) ) Is calculated. Since the overall average NE is less affected by the angle error than the section NE, an error of the section NE with respect to the overall average NE (section NE−overall average NE) is calculated using the overall average NE as a reference. Then, the NE error ratio (error / overall average NE), which is the ratio between the error and the overall average NE, is learned as a value representing the angle error of each section.

JP 2007-32540 A JP 2007-56713 A

  However, the NE error ratio becomes a different value depending on the engine speed NE at that time. The engine speed NE referred to here is an average of instantaneous NEs in a section where the crank angle is 360 CA or more. That is, the NE error ratio at the time of high NE and the NE error ratio at the time of low NE are different values. For this reason, the present inventor has obtained the knowledge that the NE error ratio learned as described above is not a numerical value that accurately represents the actual tooth position displacement (angle error of the crank angle signal).

  Hereinafter, the detail of the said knowledge is demonstrated using FIG.

  The instantaneous torque is mainly a sum of torque (gas torque) resulting from the compression and expansion of the in-cylinder gas and torque (inertia torque) resulting from the inertia of the piston. The solid line in FIG. 9A is a waveform showing the change in gas torque with respect to the crank angle, and the dotted line is a waveform showing the change in inertia torque. As shown in the figure, the waveform of the gas torque does not depend on the engine rotational speed NE at that time, whereas the waveform of the inertia torque has a different waveform depending on the engine rotational speed NE at that time.

  Therefore, the change in the instantaneous NE when NE is small is less influenced by the inertia torque and the gas torque is dominant. Therefore, for example, in the case of a four-cylinder engine in which an increase in the instantaneous NE due to the combustion stroke appears every 180 CA, the change in the instantaneous NE at the time of the small NE has a waveform shown in FIG. On the other hand, the change in the instantaneous NE when NE is large has the waveform shown in FIG. 9D because the influence of the inertia torque becomes large.

  Here, when the angle error does not exist, the change of the instantaneous NE becomes a waveform (ideal waveform) shown in (c) and (d). For example, when the angle error shown in (e) exists, The change in the instantaneous NE measured based on the crank angle signal becomes a waveform (measurement waveform) indicated by a solid line in (f) and (g). Therefore, the difference between the ideal waveform indicated by the dotted line in (f) and (g) and the measurement waveform indicated by the solid line corresponds to the actual angle error.

  However, in the method described in Patent Document 2, since learning is performed by calculating the error ratio (NE error ratio) between the overall average NE shown in the alternate long and short dash line and the measurement waveform, the value of the calculated NE error ratio is the value at that time. It becomes a different value depending on NE. For example, in the case of the measurement waveform at the time of small NE shown in (f), the NE error ratio becomes the waveform shown in the dotted line in (h), and in the case of the measurement waveform at the time of large NE shown in (g), NE The error ratio is a waveform indicated by a solid line in (h).

  Thus, in the conventional method of learning based on the overall average NE, the angle error (see (h)) calculated from the measurement waveform becomes a different value depending on NE, and the actual angle error (see (e)) is used. It is not a numerical value that is expressed accurately.

  The present invention has been made in order to solve the above-mentioned problems, and its purpose is to accurately calculate a numerical value representing the deviation amount with respect to the deviation amount from the normal position of the detected portion detected by the crank angle sensor. An object of the present invention is to provide a learning device for an internal combustion engine that can be learned by the above.

  Hereinafter, means for solving the above-described problems and the operation and effects thereof will be described.

  According to the first aspect of the present invention, a detected rotating body in which a plurality of detected portions are provided on a rotor that rotates in synchronization with rotation of a crankshaft, and a crank that detects the detected portions and outputs a crank angle signal. It is assumed that the present invention is applied to an internal combustion engine including an angle sensor and a measuring unit that measures an instantaneous rotation speed of the crankshaft based on the detected crank angle signal.

  A waveform representing a change in the instantaneous rotational speed with respect to the crank angle is called an NE waveform, and the NE waveform when a plurality of detected parts are assumed to be in normal positions with respect to the rotor is called an ideal waveform. The waveform representing the change in the instantaneous rotational speed measured by the measuring means is called a measurement waveform, the mathematical parameter derived from the ideal waveform is called an ideal parameter, and the mathematical parameter derived from the measurement waveform is a measurement parameter. In the case of calling, the following means are provided.

  That is, an average speed of an instantaneous rotational speed in a predetermined period, a plurality of different average speeds set as reference values, a storage unit that stores the ideal parameter corresponding to each of the plurality of reference values, and the measurement unit Measurement parameter calculation means for calculating the measurement parameter at the current time based on the measurement result of the above, ideal parameter acquisition means for acquiring the ideal parameter corresponding to the average speed at the current time from the storage means, and acquisition of the ideal parameter Error learning means for learning an error between the ideal parameter acquired by the means and the measurement parameter calculated by the measurement parameter calculation means.

  In short, in the present invention, an ideal parameter is stored for each reference value (hereinafter referred to as a reference NE), and an error between the ideal parameter of the reference NE corresponding to the current average speed and the measurement parameter is learned. Therefore, it is possible to suppress the learning value from becoming a different value depending on the average speed at the present time (corresponding to the NE described above), and a numerical value (ideal parameter and measurement parameter) representing the deviation amount from the normal position of the detected portion. Error) with high accuracy. Therefore, using this learning value, the instantaneous rotational speed and the output torque of the crankshaft can be calculated with high accuracy.

  In the invention according to claim 2, the storage unit stores, as the ideal parameter, a parameter of a mathematical expression representing a waveform of a specific frequency component included in the ideal waveform, and the measurement parameter calculation unit includes the measurement parameter calculation unit, A parameter of a mathematical expression representing the waveform of the specific frequency component included in the measurement waveform is calculated as the measurement parameter.

  Here, the instantaneous rotation speed increases due to the torque generated by the combustion. Therefore, since this increase amount has a high correlation with the output torque due to combustion, if the increase amount can be measured with high accuracy, the output torque can be estimated with high accuracy. This rise appears in the NE waveform at a specific period due to combustion. Therefore, for this purpose, it is only necessary to focus on a specific period of the NE waveform, and the angular error needs only to consider this frequency component.

  In the above-mentioned invention in view of this point, only the parameter relating to the specific frequency component included in the ideal waveform is stored as an ideal parameter, and only the parameter relating to the specific frequency component included in the measurement waveform is calculated as the measurement parameter. . As the parameter, for example, the amplitude and phase of a specific frequency component are set.

  As a result, when learning is performed for each reference NE for all the frequency component parameters, that is, compared with the case of learning for each reference NE for the entire measurement waveform indicated by the solid lines in FIGS. 9 (f) and (g). The ideal parameters and the number of learning points can be reduced, and the calculation load for calculating the measurement parameters can be reduced. Nevertheless, the influence of the angle error can be eliminated with high accuracy for the purpose, and the fuel injection amount in the case of the self-ignition combustion engine, the intake air amount in the case of the ignition type combustion engine, and the like can be controlled with high accuracy.

  Furthermore, according to the above-described invention, it is possible to reduce the influence of road noise that the learning value receives as described below. In other words, road noise is superimposed on the measurement waveform due to the influence of road surface unevenness during vehicle travel. The superimposed road noise component has a waveform that fluctuates with a very long period compared to the period of the measurement waveform (period synchronized with the combustion cycle). In the above-described invention in view of this point, learning is performed using the ideal parameter and the measurement parameter for the specific frequency component, and thus the road noise frequency component is removed from these parameters. Therefore, the influence of road noise received by the learning value can be reduced, and the learning accuracy can be improved.

  According to a third aspect of the present invention, the storage unit stores the amplitude and phase of a waveform of a specific frequency component included in the ideal waveform as the ideal parameter, and the measurement parameter calculation unit includes the measurement waveform. The amplitude and phase obtained by performing discrete Fourier transform on the specific frequency are calculated as the measurement parameters.

  According to the present invention, since the amplitude and phase of the waveform of a specific frequency component are set as the parameters, as described above, the fuel injection amount, intake air amount, etc. can be controlled with a small ideal parameter storage amount and measurement parameter calculation amount. It becomes like this. Since the measurement waveform is subjected to discrete Fourier transform at a specific frequency to calculate the amplitude and phase as measurement parameters, the amplitude and phase of the specific frequency can be easily calculated, which is preferable.

  In the invention of claim 4, among the waveform components included in the ideal waveform, a waveform component resulting from the piston inertia of the internal combustion engine is referred to as an inertia-induced waveform, and is caused by the compression and expansion of the cylinder gas of the internal combustion engine. In the case where the waveform component is called a gas-induced waveform, the configuration is as follows.

  That is, the storage means stores the ideal parameter separately into an inertia-caused parameter that is a parameter that represents the inertia-caused waveform and a gas-caused parameter that is a parameter that represents the gas-caused waveform. Gas correction means for correcting the gas-derived parameter based on the amount of gas charged into the cylinder or a parameter corresponding to the gas amount is provided.

  Here, the inertia torque shown in FIG. 9A has different values depending on NE as described above, but the gas torque has different values depending on the amount of gas charged in the cylinder. .

  In the above-mentioned invention in view of this point, the ideal parameter is stored separately for the inertia-induced parameter and the gas-derived parameter, and the gas-derived parameter is corrected based on the amount of gas charged in the cylinder at the present time. The value of the ideal parameter used for calculation can be made highly accurate, and learning accuracy can be improved.

  In the invention according to claim 5, the storage means stores the ideal parameter for the ideal waveform when the fuel injection of the internal combustion engine is cut, and the error learning means is the internal combustion engine. The learning is performed on condition that the fuel injection of the engine is cut.

  According to the above invention, since the measurement parameter is calculated and learned in a state where the error between the command value relating to the injection such as the fuel injection amount and the injection timing and the actual value does not affect the measurement waveform, the learning accuracy is obtained. Can be improved.

  According to a sixth aspect of the present invention, the storage means separates the ideal parameter when the torque transmission path from the crankshaft to the vehicle drive wheel is connected by a mechanical part and the ideal parameter when not connected. The ideal parameter acquisition means acquires the ideal parameter according to the current connection state of the torque transmission path from the storage means.

  Specific examples of coupling of the torque transmission paths include connection of a manual transmission (MT) clutch and connection of an automatic transmission (AT) lockup mechanism. Then, the actual ideal parameter has a different value between when connected (linked) and when not connected.

  In the above invention in view of this point, the ideal parameter at the time of connection and the ideal parameter at the time of non-connection are stored separately, and learning is performed using the ideal parameter according to the current connection state. Can be improved.

  According to a seventh aspect of the present invention, the storage means stores the ideal parameter for each gear stage of a transmission that changes and transmits the rotation speed of the crankshaft, and the ideal parameter acquisition means The ideal parameter corresponding to the shift speed is acquired from the storage means.

  Here, if the gear position is different, the actual ideal parameter has a different value. In the above-mentioned invention in view of this point, the ideal parameter is stored for each shift speed, and learning is performed using the ideal parameter corresponding to the current shift speed, so that the learning accuracy can be improved. Note that storing ideal parameter correction values for each gear position and correcting the ideal parameter according to the current gear position is also included in the “store ideal parameter for each gear position” in the above invention.

  The invention according to claim 8 is characterized by comprising load correcting means for correcting the ideal parameter based on the current driving load of the auxiliary machine driven by the rotational force of the crankshaft.

  Specific examples of the auxiliary machine include a high-pressure pump that discharges high-pressure fuel to a common rail (a pressure vessel that accumulates fuel and distributes the fuel to a fuel injection valve), and a refrigerant compressor (a compressor that compresses refrigerant and circulates it in a refrigeration cycle). ), And generators. The actual ideal parameter varies depending on the driving load of these auxiliary machines. In the above invention in view of this point, the ideal parameter used for learning is corrected based on the current driving load, so that the learning accuracy can be improved.

  In the invention according to claim 9, the storage means stores, as the ideal parameter, a parameter of a mathematical expression representing a waveform obtained by combining a plurality of frequency components included in the ideal waveform, and the measurement parameter calculation means includes A parameter of a mathematical expression representing a waveform obtained by combining the plurality of frequency components included in the measurement waveform is calculated as the measurement parameter.

  As described above, in order to improve the estimation accuracy of the output torque due to combustion, it is sufficient to learn with the parameter for the specific frequency component. However, when it is desired to obtain the change (angle error waveform) of the angle error of the crank angle signal corresponding to the change of the crank angle, it is not sufficient to use only one frequency component parameter.

  In the above-described invention in view of this point, learning is performed on a parameter of a waveform obtained by combining a plurality of frequency components, so that an angle error waveform can be obtained with higher accuracy than when learning is performed on one frequency component. According to this, since the angle error itself can be learned with high accuracy, for example, an injection timing command can be calibrated with high accuracy.

  The invention according to claim 10 is characterized in that the error learning means performs the learning on condition that torque transmission from the crankshaft to the transmission is cut off.

  Here, if torque transmission is not interrupted, fluctuations in the load that the crankshaft receives from the transmission and vehicle drive wheels will affect the measurement waveform. On the other hand, in the above invention, since the measurement parameter is calculated and learned in a state where the measurement waveform is not affected by such load fluctuation, the learning accuracy can be improved.

  The invention according to claim 11 is characterized in that the error learning means performs the learning on condition that the cooling water temperature of the internal combustion engine is equal to or higher than a predetermined value.

  Here, when the cooling water temperature is lower than the predetermined value, the viscosity of the piston lubricating oil of the internal combustion engine is high and the piston friction is large, and this friction affects the measurement waveform. On the other hand, in the above invention, since the measurement parameters are calculated and learned in a state where the measurement waveform is less affected by the friction, the learning accuracy can be improved.

1 is an overall configuration diagram of an engine system to which a learning device according to an embodiment of the present invention is applied. The figure explaining the problem accompanying the structural error of the pulsar provided in a crankshaft. The figure explaining the method of calculating an angle error ratio in case the angle error of a pulsar exists in the said one Embodiment. The figure explaining that angle error waveform Wdeltadeg can be expressed combining the waveform of a plurality of frequency components. It is a figure which shows the test result which this inventor implemented, and is a figure corresponded to the map used by the learning process of FIG. The figure explaining the method of calculating the error Ce of the complex Fourier series Cmot by the ideal waveforms Wa and Wb and the complex Fourier series Crow by the measurement waveforms Va and Vb in the one embodiment. 7 is a flowchart showing a procedure for learning an error between ideal parameters Amot, θmot and measurement parameters Arrow, θrow in the embodiment. The figure which shows the computing equation used with the flowchart of FIG. The figure explaining the problem of the conventional learning apparatus.

  Hereinafter, an embodiment in which a learning device for an internal combustion engine according to the present invention is applied to a self-ignition combustion type four-cylinder diesel engine will be described with reference to the drawings.

  FIG. 1 shows the overall configuration of the engine system.

  The illustrated diesel engine 1 is configured as a 4-stroke multi-cylinder internal combustion engine (assuming four cylinders here), and includes an actuator such as a fuel injection valve 2 for each cylinder. The piston 3 of each cylinder is connected to the crankshaft 5 via a connecting rod 4. The crankshaft 5 can be connected to drive wheels via a manual transmission (MT10). On the other hand, the shift operation unit 12 is a portion where the shift position is operated by the user, and the shift position of the MT 10 is changed by the operation of the shift operation unit 12. The shift operation unit 12 includes a shift position sensor 14 that detects a shift operation position.

  The crankshaft 5 is provided with a rotor 20 as shown enlarged on the left side in the drawing. The rotor 20 is formed with a plurality of detected portions (tooth portions 22). Specifically, the rotor 20 is basically formed with tooth portions 22 at equal intervals (here, “30CA” is illustrated), and a missing tooth portion 24 is provided at one location of the rotor 20. . The tooth portion 22 is detected by a crank angle sensor 40. The rotor 20 and the tooth portion 22 constitute a detected rotating body (pulsar) and rotate together with the crankshaft 5.

  The crank angle sensor 40 outputs a crank angle signal that changes every 30 CA by detecting the tooth portion 22 every time the crankshaft 5 rotates by a predetermined angle (30 CA). The ECU 50 described later generates a pulse signal that switches on / off every 30 CA based on the crank angle signal. The ECU 50 (measurement means) calculates the rotational speed NE of the crankshaft 5 by measuring the time of this pulse interval.

  The electronic control unit (ECU 50) includes a central processing unit (CPU 52), a nonvolatile read-only memory (ROM 54), a rewritable nonvolatile memory 56 (memory 56), and the like. And the detection value of the sensor which detects the various driving | running states of diesel engines, such as the crank angle sensor 40, and the detection value about the request | requirement from a user, such as the said shift position sensor 14, are taken in. And ECU50 controls the output of the diesel engine 1 by operating various actuators, such as the fuel injection valve 2, based on these detection results.

  Various programs are stored in the ROM 54 in order to appropriately perform the output control. As this program, for example, a fuel injection learning program 60 for calculating a learning value that compensates for variations in the injection characteristics of the fuel injection valve 2 of each cylinder, or a structural error (pulsar error) between the tooth portions 22. There is a pulsar error learning program 62 that compensates for).

  The learning value is calculated by the fuel injection learning program 60 in the following procedure. (A) The operation amount of the fuel injection valve 2 in each cylinder is set so that the difference in the amount of increase in the rotational speed of the crankshaft 5 associated with fuel injection in each cylinder is zero. (A) The difference between the reference operation amount and the operation amount obtained by the procedure (A) is set as a learning value for each cylinder.

  By using the learning value calculated by the above procedure, the amount of increase in the rotational speed calculated based on the output of the crank angle sensor 40 can be made uniform. In short, the output torque from each cylinder is reflected in the amount of increase in the rotational speed, so fluctuations in the output torque are suppressed and the drivability is improved by equalizing the amount of increase without variation among the cylinders.

  However, when there is a pulsar error, the rotation angle and rotation speed calculated based on the output of the crank angle sensor 40 are different from the actual rotation angle and rotation speed. When the deviation occurs, the variation in the injection characteristic of the fuel injection valve 2 of each cylinder cannot be compensated depending on the learning value obtained by the procedure (a).

  Therefore, in the present embodiment, first, a learning value for compensating the pulsar error is calculated using the pulsar error learning program 62. This will be described in detail below. Here, first, the structural error (pulsar error) of the interval between the tooth portions 22 will be described with reference to FIG.

  In FIG. 2 (a), for the rotor 20 provided on the crankshaft 5, a section defined by the tooth portions 22 on both sides of the missing tooth portion 24 is defined as a section A 0, and a section A 1 is sequentially set every “60 CA” in the clockwise direction. ~ A5 is defined. In FIG. 2A, since the tooth portion 22 is not displaced, the sections A0 to A5 are all equal to each other. On the other hand, the sections A2 ′ and A3 ′ show a case in which a shift occurs in the sections A2 and A3 because a shift has occurred in the tooth portion 22 indicated by the symbol P.

  FIG. 2B shows the rotational speed detected in the sections A0 to A5 during the fuel cut control for stopping the fuel injection that contributes to the generation of the torque of the crankshaft 5 of the diesel engine 1. FIG. 2C shows the time (elapsed time) required for the rotation of the sections A0 to A5 during the fuel cut control. 2 (b) and 2 (c) schematically show a solid line when there is no structural shift in the interval between the tooth portions 22. FIG. As shown in the drawing, since the fuel cut control is performed, the rotation speed is gradually decreased, and the elapsed time is gradually increased.

  On the other hand, the rotational speed detected when the above-described structural deviation occurs in the interval between the tooth portions 22 is shown by a one-dot chain line in FIG. As shown in the figure, the rotation speed once increases in the section A2 ′ and becomes smaller than the actual value in the section A3 ′. Moreover, the elapsed time detected when the above-mentioned structural shift | offset | difference arises in the space | interval between the tooth parts 22 is shown with a dashed-dotted line in FIG.2 (c). As shown in the figure, the elapsed time once decreases in the section A2 ′ and becomes larger than the actual value in the section A3 ′.

  FIGS. 9C and 9D show fluctuations in the rotational speed (instantaneous NE) of the crankshaft 5 during fuel cut control, that is, NE waveforms showing the relationship between the crank angle and the instantaneous NE. In FIG. 9B, the first to fourth cylinders are indicated by “# 1 to # 4”, respectively, and the relationship between the NE waveform shown in (c) and (d) and the combustion cycle of each cylinder is shown. ing. Incidentally, since the fuel cut control is assumed, the “combustion stroke” is not shown. Here, the decrease in the rotational speed due to the fuel cut control is shown neglected for convenience.

  (C) The NE waveform shown in (d) is a waveform (ideal waveform) when the tooth portion 22 is not displaced, but this ideal waveform varies depending on the engine rotational speed NE at that time. . The engine speed NE referred to here is an average of instantaneous NEs in a section where the crank angle is 360 CA or more. That is, the ideal waveform Wa at the time of low NE shown in (c) is different from the ideal waveform Wb at the time of high NE shown in (d).

  This is because, as described above with reference to FIG. 9A, the waveform of the gas torque does not depend on the rotational speed NE, whereas the waveform of the inertia torque varies depending on the rotational speed NE. To do. That is, the NE waveform at the time of low NE is dominated by the gas torque, and the NE waveform at the time of high NE has a greater degree of cancellation of the gas torque by the inertia torque.

  On the other hand, when the tooth portion 22 is displaced and the angle error shown in FIG. 3 (a) exists, the instantaneous NE calculated by the ECU 50 based on the crank angle signal is shown in FIGS. 3 (b) and 3 (c). This is the NE waveform (measurement waveform) indicated by the solid line. The broken lines in (b) and (c) are the ideal waveforms Wa and Wb in FIGS. 9 (c) and 9 (d). Moreover, the vertical axis | shaft of Fig.3 (a) has shown the deviation | shift amount (angle) with respect to the regular position (position of FIG. 2) of the tooth | gear part 22, and a horizontal axis shows a crank angle. Since the rotor 20 is directly attached to the crankshaft 5, the angle error waveform WΔdeg shown in FIG. 3A changes at a cycle of 360 degrees.

  FIG. 4 is a diagram illustrating that the angle error waveform WΔdeg can be expressed by combining waveforms of a plurality of frequency components. A solid line WΔ1 in FIG. 4A is a constant component. For example, an average value of the angle error waveform WΔdeg may be set as the constant component WΔ1. A solid line W360 in (b) is a waveform indicating a frequency component of a wave that changes in a period of 360 CA in the angle error waveform WΔdeg. A solid line WΔ2 in (c) is a waveform obtained by adding a component W360 having a period of 360 CA to the constant component WΔ1.

  Similarly, solid lines W180 and W120 in (d) and (f) show the waveforms of the components of the 180 CA period and the 120 CA period, and the solid line WΔ3 in (e) shows WΔ2 + W180. A solid line WΔn in (g) indicates a waveform obtained by adding n component waveforms to WΔ1, and this waveform WΔn is substantially the same as the angle error waveform WΔdeg. Thus, the angle error waveform WΔdeg can be expressed by a combination of a plurality of frequency components (W360, W180, W120...). Each frequency component can be calculated by performing a known discrete Fourier transform on the angle error waveform WΔdeg.

  Although the angle error waveform WΔdeg has been described with reference to FIG. 4, the same applies to the measurement waveforms Va and Vb shown in FIGS. 3B and 3C. These measurement waveforms Va and Vb are combinations of waveforms of a plurality of frequency components. Can be expressed. In this embodiment, the measurement waveforms Va and Vb are subjected to discrete Fourier transform with a specific period (180CA) to calculate component waveforms Va180 and Vb180 having the period 180CA. 3D and 3E, the solid lines indicate the 180CA component waveforms Va180 and Vb180 for the measurement waveforms Va and Vb, and the broken lines indicate the 180CA component waveforms Wa180 and Wb180 for the ideal waveforms Wa and Wb.

  Here, the complex Fourier series C of these 180CA component waveforms Va180, Vb180, Wa180, and Wb180 is expressed as A · exp (jθ) using an imaginary number j where the amplitude is A and the phase is θ. In the following description, among the complex Fourier series C, those with the ideal waveforms Wa and Wb are expressed as Cmot, and those with the measured waveforms Va and Vb are expressed as Crow. The error Ce between Cmot and Crow represents an angle error due to an error in the tooth portion 22 interval (pulsar error).

  Therefore, in this embodiment, the values of Cmot amplitude Amot and phase θmot (ideal parameter) are stored in advance in ROM 54 (storage means), and Crow amplitude Arrow and phase θrow (measurement parameter) are measured. Then, the error between the ideal parameters Amot, θmot and the measurement parameters Arow, θrow is calculated as the amplitude Ae and the phase θe of the angle error, and stored and updated as a learning value. This learning value corresponds to a learning value obtained by the pulsar error learning program 62 described above.

  However, since the ideal parameters Amot and θmot have different values according to the average NE, the ideal parameters Amot and θmot are stored for each average NE, and the measurement parameters Arrow and θrow are measured according to the average NE. Using the ideal parameters Amot and θmot, the angle error amplitude Ae and phase θe (learned value) described above are calculated.

  FIGS. 5A and 5B are test results conducted by the present inventor, and show changes in ideal parameters Amot and θmot with respect to average NE. In this test, pulsars 20 and 22 having no angular error in the non-injection state in which the fuel injection from the fuel injection valve 2 is stopped are used. This is a test in which the amplitude and phase of the waveform obtained by computing the 180 CA periodic component of the detected NE waveform by discrete Fourier transform are acquired as ideal parameters Amot and θmot, and the test is performed for each different average NE. doing.

  Then, the horizontal axis (average NE) in FIG. 5 is divided by a plurality of reference values g1 to gn, and ideal parameters Amot and θmot for each of the reference values g1 to gn are stored in the ROM 54 in advance. However, as will be described in detail later, these Amot and θmot may be stored in the ROM 54 separately for the amplitude Ap and phase θp of the gas-derived component and the amplitude Am and phase θm of the inertia-derived component. Note that the average NE is an average speed of the instantaneous NE in a predetermined period, and the predetermined period is preferably set to a time required for rotating at a crank angle of 720 CA or more.

  Next, an outline of the calculation method for calculating the error Ce will be described with reference to FIG.

  Equation (a) in FIG. 6 is a mathematical expression that expresses the amplitude Amot and phase θmot of the 180 CA periodic component in the case of no angular error by a complex Fourier series. The real part Re, the imaginary part Im, the amplitude Amot, and the phase θmot in the expression (a) are expressed by the expression (b). Further, Nemot (n) in the equation (a) indicates an instantaneous NE at a crank angle corresponding to the nth tooth portion 22, and is represented by the equation (c).

  In the example of FIG. 2, since the missing tooth portion 24 exists, the number of the tooth portions 22 is 11. However, for the sake of convenience, it is assumed that the missing tooth portion 24 does not exist in the formula of FIG. It is assumed that the portions 22 exist at equal intervals over the entire 360CA circumference. And t in (c) type | formula shows the time required for the space | interval (30 deg) of the adjacent tooth | gear part 22 to pass.

  Equation (d) is an equation representing the instantaneous NE (Nero (n)) at the crank angle corresponding to the nth tooth portion 22 when there is an angular error. Δe in the equation (d) is a ratio of the angle error to the design value of the interval between the tooth portions 22 (30 CA in the example of FIG. 2) (see FIGS. 3F and 3G).

  Equation (e) is a mathematical expression that expresses the amplitude Arrow and the phase θrow of the 180 CA periodic component in the case where there is an angle error, in a complex Fourier series. In Equation (e), “Neve” is an average value of instantaneous NEs (corresponding to the above-described average NE) in a predetermined period. Further, Ce in the equation (e) is a complex Fourier series representing an error between Cmot and Crow as described above, and is an equation representing an angular error ratio component at a specific frequency (a frequency of 180 CA cycles). .

  As shown in the equation (f) derived from the equation (e), the amplitude Ae and the phase θe of the angular error ratio component Ce can be represented by Cmot in the equation (a), Crow and Neve in the equation (e). . The expression (g) is a value obtained by multiplying the design value of the interval of the tooth portion 22 (30 CA in the example of FIG. 2) by Ce represented by the expression (f), and this value is shown in FIG. This corresponds to the angle error of the 180 CA periodic component shown in FIG.

  FIG. 6H is a diagram in which the distribution of the angular error calculated by the equation (g) with respect to the crank angle is converted into the distribution with respect to the crank angle of the instantaneous NE, and the ideal Nemot ( An embodiment of detection New (n) represented by the formulas n) and (d) is shown.

  Next, a procedure for calculating and learning the amplitude Ae and the phase θe of the angle error Ce described above will be described with reference to FIGS. 7 is repeatedly executed every time the instantaneous NE is calculated by the pulser error learning program 62 (for example, every 30 CA).

  First, in steps S10 and S11 shown in FIG. 7, it is determined whether or not it is time to execute angle error learning. Specifically, in step S10, the value of the angle counter CNT is added for every 30 CA in which the process of FIG. 7 is performed (see equation (1) in FIG. 8). When the value of the angle counter CNT reaches the calculation cycle CNT0 (for example, 360CA) of angle error learning (S11: YES), the value of the angle counter CNT is reset to zero in the subsequent step S12, and in the subsequent processing Perform angular error learning.

  If the angle counter CNT has not reached CNT0, the process proceeds to step S19 without performing steps S12 to S18 described later. The process in step S19 is performed every 30 CA rotations regardless of whether or not the angle counter CNT has reached CNT0 or whether or not angle error learning is permitted in step S14 described later. The

  In step S19, discrete Fourier transform is performed on the measurement waveform Va or Vb according to the equation (2) in FIG. That is, the real part Re and the imaginary part Im of the Fourier series Crow are calculated according to the equation (2) and added to the previous values of CrowRe and CrowIm. Note that T in the equation (2) is a specific period corresponding to the angle error detection frequency (specific frequency), and is set to 180 CA, for example.

  At the timing of executing the angle error learning, in step S13 (measurement parameter calculation means), the amplitude Arow and the phase θrow of the component waveform of the specific period T with respect to the waveform Va or Vb of the measurement value of the instantaneous NE are calculated. Specifically, CrowRe and CrowIm in the equation (2) calculated in step S19 are substituted into the equation (3) in FIG. 8 to calculate the instantaneous amplitude Arrow and the instantaneous phase θrow.

  In a succeeding step S14, it is determined whether or not angle error learning is permitted. For example, it is a non-injection state in which fuel injection from the fuel injection valve 2 is not performed, the engine coolant temperature is equal to or higher than a predetermined value, the clutch of the MT 10 is disconnected, and the crankshaft 5 is connected to the vehicle drive wheels. Angle error learning is permitted when all conditions such as the torque transmission path being interrupted are satisfied.

  When angle error learning is permitted (S14: YES), ideal amplitude Amot and phase θmot at the angle error detection frequency (specific period T) when there is no angle error are calculated in subsequent steps S15 and S16. .

  Specifically, first, in step S15 (ideal parameter acquisition means), among the instantaneous NE fluctuation at the time of no injection at the angular error detection frequency, the amplitude Ap and phase θp of the component due to the expansion and compression force of the engine cylinder gas, The amplitude Am and the phase θm of the component due to the inertial force of the piston 3 are obtained with reference to a map prepared in advance based on the average NE (Neave).

  In the example illustrated in FIG. 5, it has been described that the ideal parameters Amot and θmot are stored for each of the reference values g1 to gn, but in detail, the ideal waveforms Wa and Wb are represented by the waveforms of the gas-derived components (gas-induced waveforms). Assuming that the waveform of the inertia-caused component (inertia-caused waveform) is synthesized, the ideal parameters Amot and θmot are divided into the amplitude Ap and the phase θp of the gas-caused waveform and the amplitude Am and the phase θm of the inertia-caused waveform. Each of the upper reference values g1 to gn is stored.

  In the subsequent step S16 (gas correction means), the real part Re and the imaginary part Im of Cmot by the ideal waveforms Wa and Wb are calculated according to the equation (4) in FIG. Specifically, the real part Re and the imaginary part Im of Cmot are calculated by substituting the amplitudes Ap and Am and the phases θp and θm obtained in step S15 into the equations (4a) and (4b), and the CmotRe, Based on CmotIm, ideal parameters Amot and θmot are calculated according to equation (4).

  Here, in the equations (4a) and (4b), the gas-derived parameters Ap and θp are corrected and used based on the current pressure of the cylinder gas. Specifically, the correction is performed by multiplying the gas-derived parameters Ap and θp by pim / pim0. Pim0 in the equation is a pressure in the intake manifold (intake pipe) when a test for acquiring a parameter to be stored on the map (the test in FIG. 5) is performed. That is, as the intake pressure increases, the amount of gas charged into the cylinder increases and the pressure of the gas in the cylinder increases. Therefore, the gas-derived parameters Ap, θp are increased so as to increase the degree of influence of the gas-derived parameters Ap, θp. Is corrected.

  In the subsequent step S17 (error learning means), the ideal parameters Amot and θmot obtained in step S18 and the instantaneous amplitude Arrow and instantaneous phase θrow obtained in step S13 are substituted into the equation (5) in FIG. The angle error amplitude Ae and phase θe in the angle error detection period are calculated. These parameters Ae and θe are stored and updated as learning values.

  When the angle error learning is performed in step S17, in subsequent step S18, the values of CrowRe and CrowIm calculated in step S19 are reset to zero. Then, using the values Ae and θe learned in step S17, the values of the procedures (a) and (b) by the fuel injection learning program 60 are corrected. For example, the rotational speed calculated based on the output of the crank angle sensor 40 and its increase amount are corrected. Alternatively, the calculated value of the output torque calculated from the peak value of the instantaneous NE is corrected using the learned values Ae and θe.

  According to the embodiment described in detail above, the following effects can be obtained.

  (1) The ideal parameters Ap, θp, Am, θm are stored for each of the reference values g1 to gn, and the ideal parameters corresponding to the current average NE (Neave) are acquired. Then, an error between the instantaneous amplitude Arrow and the instantaneous phase θrow (measurement parameter) and the ideal parameters Ap, θp, Am, θm, that is, the amplitude Ae and the phase θe of the angle error are learned. Therefore, since the learning value (Ae, θe) of the angle error is calculated using the ideal parameter corresponding to the average NE, the learning value representing the deviation amount from the normal position of the tooth portion 22 can be obtained with high accuracy. Therefore, using this learning value, the instantaneous NE and the instantaneous output torque can be calculated with high accuracy, and the accuracy of fuel injection control by the fuel injection learning program 60 can be improved.

  (2) For the amplitude and phase (parameter) applied to a specific frequency (specific period T) component, the error between the ideal parameters Amot, θmot and the instantaneous parameters Arow, θrow (measured values) is learned, so all frequency components The number of learning points can be reduced as compared with the case where learning is performed for each of the reference values g1 to gn. Moreover, since the period of the road noise component is extremely longer than the specific period T set to one combustion cycle period or less, according to the present embodiment in which the parameters of the specific period T component are learned, the learning value (Ae, The influence of road noise on θe) can be reduced, and the learning accuracy can be improved.

  (3) The inertia-caused parameters (Am, θm) and the gas-derived parameters (Ap, θp) are stored separately, and the gas-derived parameters are corrected in accordance with the gas filling amount in the cylinder. The value of the ideal parameter used for calculation can be made highly accurate, and learning accuracy can be improved.

  (4) The angular error learning is permitted on condition that the engine is in a non-injection state, the engine coolant temperature is equal to or higher than a predetermined value, and the MT10 clutch is disengaged. Therefore, the learning value is not affected by the influence of the fuel injection state, the viscosity of the engine oil, and the torque fluctuation transmitted from the MT 10, so that learning accuracy can be improved.

(Other embodiments)
The present invention is not limited to the description of the above embodiment, and may be modified as follows. Moreover, you may make it combine the characteristic structure of each embodiment arbitrarily, respectively.

  Here, the ideal parameters Ap, θp, Am, and θm have different values if the speed of the MT 10 is different. In view of this point, the ideal parameters Ap, θp, Am, and θm may be stored for each shift speed of the MT 10 and learned using the ideal parameters according to the current shift speed. According to this, the learning accuracy can be improved. Alternatively, a correction value for the ideal parameter may be stored for each shift speed, and the ideal parameter may be corrected according to the current shift speed. The same applies to an automatic transmission (AT) that automatically switches the gear position.

  The ideal parameters Ap, θp, Am, and θm have different values depending on whether or not the clutch of the MT10 is engaged. In view of this point, ideal parameters Ap, θp, Am, θm corresponding to the presence / absence of clutch engagement may be stored, and learning may be performed using ideal parameters according to the current clutch engagement state. According to this, the learning accuracy can be improved. Alternatively, a correction value for the ideal parameter according to whether or not the clutch is connected may be stored, and the ideal parameter may be corrected according to the current clutch connection state.

  When the automatic transmission (AT) that automatically switches the gear stage has a lockup mechanism, the ideal parameters Ap, θp, Am, and θm have different values depending on whether or not the lockup mechanism is connected. In view of this point, ideal parameters Ap, θp, Am, θm corresponding to the presence / absence of lockup connection may be stored, and learning may be performed using ideal parameters according to the current lockup connection state. According to this, the learning accuracy can be improved.

  Alternatively, a correction value for the ideal parameter according to the presence or absence of the lockup connection may be stored, and the ideal parameter may be corrected according to the current lockup connection state. For reference, the lock-up mechanism is a well-known mechanism that mechanically connects (locks up) an input shaft and an output shaft of a torque converter that transmits torque by fluid (oil).

  The ideal parameters Ap, θp, Am, and θm have different values depending on the driving load of the auxiliary machine that is driven by the rotational force of the crankshaft 5. In view of this point, ideal parameters Ap, θp, Am, θm corresponding to the driving load of the auxiliary machine are stored, and the ECU 50 (gas correction means) uses the ideal parameters corresponding to the auxiliary driving load at the present time. You may make it learn. According to this, the learning accuracy can be improved.

  Alternatively, a correction value for the ideal parameter according to the accessory driving load may be stored, and the ideal parameter may be corrected according to the current accessory driving load. Incidentally, specific examples of the above-mentioned auxiliary machine include a high-pressure pump that discharges high-pressure fuel to a common rail (a pressure vessel that accumulates fuel and distributes the fuel to a fuel injection valve), and a refrigerant compressor (compresses the refrigerant and circulates it in the refrigeration cycle). Compressor), a generator, and the like.

  In the above embodiment, the error between the ideal parameter and the measurement parameter is learned for the component waveform having a period of 180 CA, but the ideal parameter is stored for a composite waveform obtained by combining the component waveforms of a plurality of periods. You may make it measure the measurement parameter about the said synthetic | combination waveform, and may learn these errors. According to this, the angle error waveform can be obtained with higher accuracy than in the case of learning about one frequency component.

  In the above embodiment, the inertia-induced parameters (Am, θm) and the gas-derived parameters (Ap, θp) are stored separately, and the gas-derived parameters are corrected according to the gas filling amount in the cylinder. However, the phase and amplitude parameters obtained by combining these parameters may be stored, and the correction may be abolished.

  In the above embodiment, the amplitude and phase of the waveform obtained by performing discrete Fourier transform on the NE waveform are used as parameters of the mathematical formula derived from the NE waveform. However, the present invention is not limited to these parameters. Instead, for example, the amplitude, phase, frequency, and the like of the waveform extracted from the NE waveform by the filter circuit that extracts a predetermined frequency component may be used as the parameter.

  In the above embodiment, the amplitude and phase of the waveform of a specific frequency component obtained by the discrete Fourier transform are used as ideal parameters and measurement parameters, but the discrete Fourier transform is abolished and the measurement waveforms Va, Vb, The amplitude and phase of the ideal waveforms Wa and Wb may be used as ideal parameters and measurement parameters.

  DESCRIPTION OF SYMBOLS 5 ... Crankshaft, 20 ... Rotor, 22 ... Tooth part (detected part), 20, 22 ... Pulser (detected rotating body), 40 ... Crank angle sensor, 50 ... ECU (measuring means, load correcting means), 54 ... ROM (storage means), Amot, θmot ... ideal parameters, Arow, θrow ... measurement parameters, Am, θm ... inertia-induced parameters, Ap, θp ... gas-derived parameters, g1 to gn ... reference values, Va, Vb ... measurement waveforms , Wa, Wb ... ideal waveform, S13 ... measurement parameter calculation means, S15 ... ideal parameter acquisition means, S16 ... gas correction means, S17 ... error learning means.

Claims (11)

A detected rotating body provided with a plurality of detected portions on a rotor that rotates in synchronization with rotation of the crankshaft;
A crank angle sensor that detects the detected portion and outputs a crank angle signal;
Measuring means for measuring the instantaneous rotational speed of the crankshaft based on the detected crank angle signal;
Applied to an internal combustion engine with
A waveform representing a change in the instantaneous rotational speed with respect to the crank angle is referred to as an NE waveform, and the NE waveform when a plurality of detected parts are assumed to be in a normal position with respect to the rotor is referred to as an ideal waveform. A waveform representing a change in the instantaneous rotational speed measured by the measuring means is called a measurement waveform, a mathematical parameter derived from the ideal waveform is called an ideal parameter, and a mathematical parameter derived from the measurement waveform is called a measurement parameter. In case
Storage means for storing the ideal parameter corresponding to each of the plurality of reference values, which is an average speed of the instantaneous rotational speed in a predetermined period, and sets a plurality of different average speeds as a reference value;
Measurement parameter calculation means for calculating the measurement parameter at the present time based on the measurement result by the measurement means;
Ideal parameter acquisition means for acquiring the ideal parameter corresponding to the current average speed from the storage means;
Error learning means for learning an error between the ideal parameter acquired by the ideal parameter acquisition means and the measurement parameter calculated by the measurement parameter calculation means;
A learning apparatus for an internal combustion engine, comprising: The storage means stores, as the ideal parameter, a parameter of a mathematical expression representing a waveform of a specific frequency component included in the ideal waveform,
2. The learning device for an internal combustion engine according to claim 1, wherein the measurement parameter calculation unit calculates a parameter of a mathematical expression representing a waveform of the specific frequency component included in the measurement waveform as the measurement parameter. . The storage means stores the amplitude and phase of a specific frequency component waveform included in the ideal waveform as the ideal parameter,
The learning of the internal combustion engine according to claim 2, wherein the measurement parameter calculation unit calculates an amplitude and a phase obtained by performing a discrete Fourier transform on the measurement waveform at the specific frequency as the measurement parameter. apparatus. Of the waveform components included in the ideal waveform, the waveform component resulting from the piston inertia of the internal combustion engine is referred to as the inertia-induced waveform, and the waveform component resulting from the compression and expansion of the gas in the cylinder of the internal combustion engine is referred to as the gas-induced waveform. In case
The storage means stores the ideal parameter separately into an inertia-caused parameter that is a parameter that represents the inertia-caused waveform and a gas-caused parameter that is a parameter that represents the gas-caused waveform,
4. The gas correction device according to claim 1, further comprising a gas correction unit configured to correct the gas-derived parameter based on a gas amount charged into the cylinder at a current time or a parameter corresponding to the gas amount. A learning apparatus for an internal combustion engine as described. The storage means stores the ideal parameter for the ideal waveform when the fuel injection of the internal combustion engine is cut,
5. The learning device for an internal combustion engine according to claim 1, wherein the error learning unit performs the learning on condition that fuel injection of the internal combustion engine is cut. The storage means separately stores the ideal parameter when the torque transmission path from the crankshaft to the vehicle driving wheel is connected by a mechanical part and the ideal parameter when not connected,
The internal combustion engine according to any one of claims 1 to 5, wherein the ideal parameter acquisition unit acquires the ideal parameter according to a current connection state of the torque transmission path from the storage unit. Learning device. The storage means stores the ideal parameter for each shift stage of a transmission that shifts and transmits the rotation speed of the crankshaft,
The learning apparatus for an internal combustion engine according to any one of claims 1 to 6, wherein the ideal parameter acquisition unit acquires the ideal parameter corresponding to the current gear position from the storage unit.   The load correction means which correct | amends the said ideal parameter based on the driving load at the present time of the auxiliary machine driven with the rotational force of the said crankshaft is provided. A learning device for an internal combustion engine. The storage means stores, as the ideal parameter, a parameter of a mathematical expression representing a waveform obtained by combining a plurality of frequency components included in the ideal waveform,
The measurement parameter calculation unit calculates, as the measurement parameter, a parameter of a mathematical expression representing a waveform obtained by combining the plurality of frequency components included in the measurement waveform. A learning apparatus for an internal combustion engine according to claim 1.   The internal combustion engine according to any one of claims 1 to 9, wherein the error learning means performs the learning on the condition that torque transmission from the crankshaft to a vehicle driving wheel is interrupted. Institutional learning device.   The learning of the internal combustion engine according to any one of claims 1 to 10, wherein the error learning means performs the learning on the condition that a cooling water temperature of the internal combustion engine is equal to or higher than a predetermined value. apparatus.