DE112009005254B4 - Control system for an internal combustion engine - Google Patents

Abstract

A control system for an internal combustion engine (1), the control system being characterized by comprising: control parameter calculating means (20) for calculating a control (IDTH) parameter of the engine (1) at Use of a stationary-engine-operating-state model and an unsteady-engine operating-state model, wherein the stationary-engine-operating-state model corresponds to stationary operation of the engine (1) and a predetermined operating parameter (THCMD) of the engine (1) using a neural network (SOMSS ), and the unsteady engine operating state model corresponds to a transient operation of the engine (1) and outputs the predetermined operating parameter (THCMD) using another neural network (SOMTS), the control parameter calculating means (20) comprising: Unsteady engine operating condition determining means (20) for determining whether the engine (1) is in the transient operating state or not; and selecting means (20) for selecting one of the stationary-engine-operation-state model and the unsteady-engine-operation-state model according to the determination result of the in-vehicle engine operation state determination means, the control parameter calculation means (20) determining the control parameter (IDTH ) according to the operating parameter (THCMD) output of the selected model, the transient engine operating state determining means (20) determining that the engine (1) is in the steady-state operating state when change amounts (DGAIRCMD, DPB, DPI, DNE) input parameters (GAIRCMD, PB, PI, NE) used for calculating the control parameter (IDTH) of the engine (1) are smaller than a predetermined amount of change (DGATH, DPBTH, DPITH, DNETH) during a period of time, ...

Description

TECHNICAL AREA

The present invention relates to a control system for an internal combustion engine, and more particularly to a system for controlling the engine using a neural network.

BACKGROUND

The JP H11-85719 A FIG. 12 shows a parameter estimating apparatus for estimating an air-fuel ratio using a neural network to which parameters indicating an operating state of the engine such as a throttle valve opening, an intake pressure, an engine rotational speed and an intake air temperature are input.

In this apparatus, a plurality of engine operating regions are set according to the input parameters, and the calculation path in the neural network used is changed according to the engine operating region.

OVERVIEW OF THE INVENTION

Problems to be solved by the invention

When an engine operating parameter, such as the air-fuel ratio, is estimated (calculated) based on other engine operating parameters, the estimation accuracy may vary greatly depending on whether the engine operating condition is a stationary engine operating condition or a transient engine operating condition. The apparatus of Patent Document 1 deals with the transient operating condition by inputting previous values of the engine operating parameter. However, there is a possibility that the device can not handle the transient operating state using only one neural network. Furthermore, the size of a neural network becomes very large, which increases the amount of computation required.

From the DE 10 2006 061 659 A1 For example, a single controller is known in which a physical model is implemented. The determination of manipulated variables by means of the control device takes place as a function of at least one operating variable, regardless of whether the internal combustion engine is in a stationary or in a transient operating state.

From the DE 10 2004 031 296 A1 a control method for an internal combustion engine is known. According to this method, determination of operating parameters of the internal combustion engine in a steady state is performed using a stationary state model of the engine. In the transient state, however, the operating parameters are determined without using a model.

In addition, the describes DE 10 2004 031 296 A1 in that a determination of operating parameters in the transient operating state should be performed without using a model, because if, in addition to a steady state model, a model describing an internal combustion engine in a transient state is used, the required computing power increases significantly Increased the hardware requirements and made the engine control unit considerably more expensive. Therefore, the beats DE 10 2004 031 296 A1 before, then, when in the operation of the internal combustion engine, a transient condition occurs to move from the model-based determination of the parameters of operating parameters to a firing synchronous determination of the parameters of operating parameters.

The object of the present invention is to provide a control system for an internal combustion engine which employs a neural network not only in a stationary engine operating state but also in a transient operating state such that it is possible to provide control accuracy in suppressing the engine Calculation load to improve during engine operation.

Means for solving the problem

In order to achieve the above object, the present invention provides a control system for an internal combustion engine, comprising control parameter calculating means for calculating a control parameter (IDTH) of the engine using a stationary-engine running-state model and an unsteady-state engine operating condition models. The steady-state engine operating mode model corresponds to stationary operation of the engine and outputs a predetermined operating parameter (THCMD) of the engine using a neural network. (SOMSS) off. The unsteady engine operating mode model corresponds to transient operation of the engine and outputs the predetermined operating parameter (THCMD) using another neural network (SOMTS). The control parameter calculating means includes transient engine operating state determining means and selecting means. The unsteady engine operating state determining means determines whether or not the engine is in the transient operating state. The selection means selects one of the stationary-engine-operation-state model and the unsteady-engine-operation-state model according to the determination result of the in-station-engine-operation-state determination means. The control parameter calculating means calculates the control / control parameter (IDTH) according to the operation parameter (THCMD) output of the selected model. The on-state engine operating condition determining means determines that the engine is in the steady-state operating condition when change amounts (DGAIRCMD, DPB, DPI, DNE) of input parameters (GAIRCMD, PB, PI, NE) used to calculate the control parameter (FIG. IDTH) of the motor during a period of time are smaller than a predetermined amount of change (DGATH, DPBTH, DPITH, DNETH). The on-state engine operating condition determining means determines that the engine is in the transient operating condition when at least one of change amounts (DGAIRCMD, DPB, DPI, DNE) of the input parameters (GAIRCMD, PB, PI, NE) which is in the steady-state engine operating condition Model is input during a period of time greater than a predetermined amount of change (DGATH, DPBTH, DPITH, DNETH). Further, a first set of input parameters (GAIRCMD, PB, PI, NE) is input to the steady state engine operating mode model, and a second set of input parameters (DGAIRCMD, DPB, DPI, DNE) is input to the transient engine operating state model the second set of input parameters (DGAIRCMD, DPB, DPI, DNE) corresponds to amounts of change from the first set of input parameters (GAIRCMD, PB, PI, NE) for a constant time period (TC).

With this configuration, the control / regulation parameter of the engine is calculated by using the stationary-engine-operation-state model, which corresponds to steady-state operation of the engine, and outputs a predetermined operation parameter of the engine using a neural network, and the non-stationary-engine-operation-state model , which corresponds to a transient state of the motor, and which outputs the predetermined operating parameter using another neural network. Specifically, it is determined whether or not the engine is in the transient operating state, and one of the steady state engine operating state model and the inoperative engine operating state model is selected according to the determination result, and the control parameter is selected according to the output of the selected one Models calculated. Thus, the control parameter suitable for both the steady-state and non-steady-state operation of the engine can be obtained, thereby increasing the control accuracy with the control parameter calculated by the operation parameter.

Further, the first and second sets of input parameters, which are different from each other, are input to the stationary-engine-operation-state model and the in-stationary-engine-operation-state model, respectively, and the calculation of the model output is performed only with respect to a model selected by the selection means. Accordingly, it is not necessary to perform the calculations corresponding to the neural networks in the two models, thereby suppressing the computation load.

Further, it is determined that the engine is in the transient operating state when at least one of the change amounts of the input parameters (GAIRCMD, PB, PI, NE) input to the in-vehicle engine running state model is larger than the predetermined change amount. Accordingly, the transient operating state in which the transient engine running state model is to be used can be appropriately determined.

BRIEF DESCRIPTION OF THE FIGURES

1 shows a configuration of an internal combustion engine and a control system thereof according to an embodiment of the present invention.

2 is a diagram showing a self-organizing map.

3 is a diagram showing another self-organizing map.

4 shows a relationship between an intake air flow rate (GAIR) and a throttle valve opening (TH).

5 FIG. 10 is a flowchart of a method of calculating a target opening (THCMD) of the throttle valve.

6 FIG. 10 is a flowchart of a self-organizing map (SOM) calculation method used in the method of FIG 5 is performed.

Means for carrying out the invention

Preferred embodiments of the present invention will now be explained with reference to the figures.

1 FIG. 12 is a schematic diagram showing a configuration of an internal combustion engine and a control system thereof according to an embodiment of the present invention. FIG. The internal combustion engine 1 (hereinafter referred to as "engine") is a diesel engine wherein fuel is injected directly into the cylinders. Each cylinder is equipped with a fuel injection valve 9 provided, which with an electronic control unit 20 (hereinafter referred to as "ECU") is electrically connected. The ECU 20 controls a valve opening timing and a valve opening period of each fuel injection valve 9 , Ie. the fuel injection period and the fuel injection amount are determined by the ECU 20 regulated / controlled.

The motor 1 includes an intake pipe 2 , an exhaust pipe 4 and a turbocharger 8th , The turbocharger 8th includes a turbine 11 and a compressor 16 , The turbine 11 has a turbine wheel 10 which is rotatably driven by the kinetic energy of the exhaust gases. The compressor 16 has a compressor wheel 15 on which with the turbine wheel 10 by means of a wave 14 connected is. The compressor wheel 15 sets the intake air of the engine 1 under pressure (compresses the intake air).

The turbine 11 has a plurality of movable blades 12 (Only two are shown) and an actuator (not shown) for actuating the movable blades 12 to open and close. The majority of moving blades 12 are actuated to open and close to allow throughput into the turbine wheel 10 to change injected exhaust gases. The turbine 11 is configured such that the flow rate of the exhaust gases entering the turbine wheel 10 be injected by changing an opening of the movable blade 12 (hereinafter called "blade opening") θvgt is changed to the rotational speed of the turbine wheel 10 to change. The actuator, which is the moving blades 12 pressed, is at the ECU 20 connected and the vane opening θvgt is through the ECU 20 regulated / controlled. In particular, the ECU supplies 20 a control signal of a variable duty ratio on the actuator and controls the blade opening θvgt by the control signal. The configuration of the turbocharger with moving blades is well known, such as in the JP H01-208501 A disclosed.

The intake pipe 2 is with a the compressor 16 downstream intercooler 18 and with an intercooler 18 downstream throttle valve 3 intended. The throttle valve 3 is configured to by an actuator 19 to be operated for opening and closing, and the actuator 19 is with the ECU 20 connected. The ECU 20 performs a control / regulation of the opening of the throttle valve 3 by means of the actuator 19 by.

An exhaust gas recirculation passage 5 for recirculating exhaust gases to the intake pipe 2 is between the exhaust pipe 4 and the intake pipe 2 provided. The exhaust gas recirculation passage 5 is with an exhaust gas recirculation passage control valve 6 (hereinafter referred to as "EGR valve 6 , Which controls the amount (EGR amount) of the exhaust gases that are recycled. The EGR Valve 6 is an electromagnetic valve having a solenoid. A valve opening of the EGR valve 6 is through the ECU 20 regulated / controlled. The EGR valve 6 is with a stroke sensor 7 for detecting a valve opening (a valve lift amount) LACT, and the detection signal becomes the ECU 20 fed. The exhaust gas recirculation passage 5 and the EGR valve 6 form an exhaust gas recirculation device.

An intake air flow sensor 21 , a boost pressure sensor 22 , an intake air temperature sensor 23 and a suction pressure sensor 24 are in the intake pipe 2 appropriate. The intake air flow rate sensor 21 detects an intake air flow GA. The boost pressure sensor 22 detects an intake pressure (boost pressure) PB at a portion of the intake pipe 2 downstream of the compressor 16 , The intake air temperature sensor 23 detects a temperature TI of the intake air. The suction pressure sensor 24 detects an intake pressure PI.

These sensors 21 to 24 are with the ECU 20 connected, and the detection signals from the sensors 21 to 24 become the ECU 20 fed.

A lean NOx catalyst 31 and a particle filter 32 are downstream of the turbine 11 in the exhaust pipe 4 appropriate. The lean NOx catalyst 31 is a NOx removing device for removing NOx contained in the exhaust gases. The particle filter 32 captures particulate matter contained in the exhaust gases (which mainly consists of soot). The lean NOx catalyst 31 is configured so that NOx is trapped in a state in which an oxygen concentration in the exhaust gases is relatively high, that is, a concentration of the reduction components (HC, CO) is relatively low, and the trapped NOx is reduced by the reduction components and discharged in one state becomes, in which the concentration of the reduction components in the exhaust gases is relatively high.

An acceleration sensor 27 and an engine speed sensor 28 are with the ECU 20 connected. The acceleration sensor 27 detects an amount of operation AP of an accelerator pedal (not shown) by the engine 1 driven vehicle (hereinafter referred to as "accelerator operation amount AP"). The engine speed sensor 28 detects a motor rotation speed NE. The detection signals of these sensors become the ECU 20 fed. The engine speed sensor 28 leads the ECU 20 a crank angle pulse and a TDC pulse too. The crank angle pulse is generated at each predetermined crank angle (eg, 6 degrees). The TDC pulse becomes synchronous at a time when a piston is in each cylinder of the engine 1 at top dead center is generated.

The ECU 20 includes an input circuit, a central processing unit (hereinafter referred to as "CPU"), a memory circuit, and an output circuit. The input circuit performs various functions, including shaping the waveform of input signals from various sensors, correcting the voltage levels of the input signals to a predetermined level, and converting analog signal values to digital values. The memory circuit preliminarily stores various operation programs to be executed by the CPU, and stores the results of calculations or the like by the CPU. The output circuit provides control signals to the actuator for actuating the movable blades 12 the turbine 11 , the fuel injection valves 9 , the EGR valve 6 , the actuator 19 for actuating the throttle valve 3 and the same.

The ECU 20 performs fuel injection control with the fuel injection valve 9 , an exhaust gas recirculation control with the EGR valve 6 , a boost pressure control with the moving blades 12 and the like, according to an engine operating condition (which is substantially indicated by the engine rotational speed NE and an engine load target value Pmecmd). The engine load target value Pmecmd is calculated according to the accelerator operation amount AP and set to increase with an increase in the accelerator operation amount AP.

The ECU 20 calculates a target throttle valve opening (THCMD) according to a target intake air amount GAIRCMD [g / sec] using a neural network to which a self-organizing map algorithm is applied. This neural network is hereinafter simply referred to as a "self-organizing map". The ECU 20 drives the actuator 19 so that the determined throttle valve opening coincides with the target throttle valve opening THCMD.

In this embodiment, the target throttle valve opening THCMD is calculated by means of a self-organizing map SOMSS of a steady-state engine operating state model and a self-organizing map SOMTS of an unsteady engine operating state model. The self-organizing map SOMSS of a stationary-engine-operating-state model corresponds to iron stationary (steady state) operating state of the motor 1 and the self-organizing map SOMTS of an unsteady engine running state model corresponds to a transient operating state of the engine 1 ,

The self-organizing map is described in detail below.

An input data vector xj consisting of "N" elements is defined by the following equation (1), and a weighting vector wi of each neuron constituting the self-organizing map is defined by the following equation (2). A number of neurons are designated by "M". Ie. a parameter "i" assumes values from "1" to "M". An initial value of the weighting vector wi is given using a random number. xj = (xj1, xj2, ..., xjN) (1) wi = (wi1, wi2, ..., wiN) (2)

For each of the "M" neurons, an Euclidean distance DWX (= | wi-xj |) between the input data vector xj and the weighting vector wi of the corresponding neuron is calculated. A neuron whose distance DWX takes a minimum value is defined as the winner neuron. The Euclidean distance DWX is calculated by the following equation (3). [Eq. 1]

Next, the weighting vectors wi of the winner neuron and the neurons contained in a neuron set Nc in the neighborhood of the winner neuron are updated by the following equation (4). In the equation (4), "α (t)" denotes a training coefficient and "t" denotes a number of the training units (hereinafter referred to simply as "training number"). The training coefficient α (t) is set to "0.8" as an initial value, for example, and is set to decrease with an increase in the training number "t". wi (t + 1) = wi (t) + α (t) (xj - wi (t)) (4)

The weighting vectors wi of the neurons not included in the neuron amount Nc retain a previous value as shown by the following equation (5). wi (t + 1) = wi (t) (5)

It should be noted that the neuron amount Nc is also a function of the training number "t", and is set so that an area of the vicinity of the winning neuron becomes narrower as the training number "t" increases. The weighting vectors of the winner neuron and the neurons in the neighborhood of the winner neuron are changed by updating with the equation (4) to approximate the input data vector.

If the calculation is performed according to the training rule described above for many input data vectors, the distribution of the "M" neurons reflects the distribution of the input data vectors. For example, when the input data vectors to two-dimensional vectors are simplified and the distribution of the input data vectors is represented on a two-dimensional plane, the neurons spread uniformly across the plane as the input data vectors spread uniformly across the plane. Finally, if there is nonuniformity in the distribution of the input data vectors (if there are changes in the distribution density), then the distribution of the neurons eventually becomes a distribution having similar nonuniformity.

The self-organizing map obtained as described above may be further modified by application of the Learner Vector Quantization (LVQ) algorithm, thereby obtaining a more appropriate distribution of the neurons.

2 Fig. 13 shows the self-organizing map SOMSS of the stationary engine running state model for calculating the target throttle valve opening TH-CMD in this embodiment as a two-dimensional map. This two-dimensional map is defined by the target intake air amount GAIRCMD and the boost pressure PB, which represent two input parameters as the most dominant factors. An input data vector x TH is defined by the following equation (10). That is, the input parameters are the target intake air amount GAIRCMD, the boost pressure PB, the intake pressure PI, and the engine rotation speed NE. xTH = (GAIRCMD, PB, PI, NE) (10)

In the 2 The map shown is divided into a plurality of areas RNRi (i = 1 to M, M = 36). Each area includes a neuron NRi (represented by "*"). By performing the training (learning) with many input data vectors xTH in advance, the position (weighting vector wi) of each neuron NRi is determined. Each region RNRi is defined by drawing boundary lines taking into account the positional relationships with the neighboring neurons. By making the distribution of the input data vector xTH applied to the training in accordance with an actual distribution occurring in the actual engine operation, the distribution density of the neuron NRi becomes relatively high in the regions that have an engine operating condition with a high occurrence frequency during the engine actual engine operation. With this feature of the self-organizing map, accuracy of the target throttle valve opening THCMD in the high frequency operation states can be improved. In the 2 The map shown is obtained by performing the training corresponding to the reference engine (which is new and has an average performance). In 2 The input data that is applied to the training is displayed with black dots.

In training the self-organizing map, a weighting coefficient vector Ci (i = 1 to M) expressed by the following equation (11) is calculated and stored using the input data vector xTH and the actual throttle valve opening TH corresponding to the input data vector xTH. The weighting coefficient vector Ci is calculated and stored according to each neuron NRi. Ci = (C0i, C1i, C2i, C3i, C4i) (11)

In the actual control operation, the area RNRi comprising the current operating point on the map is selected first. The operating point is defined by the target intake air amount GAIRCMD and the boost pressure PB, which are elements of the input data vector xTH. Next, the weighting coefficient vector Ci corresponding to the neuron NRi representing the range RNRi and the input data vector xTH are applied to the following equation (12) to calculate the target throttle valve opening THCMD. THCMD = C1i × GAIRCMD + C2i × PB + C3i × PI + C4i × NE + C0i (12)

On the other hand, the change amounts corresponding to the input parameters of the self-organizing map SOMSS of the stationary engine running state model are applied to the self-organizing map SOMTS of the in-state engine running state model as input parameters. Specifically, a target intake air flow rate change amount DGAIRCMD, a boost pressure change amount DPB, an intake pressure change amount DPI, and a. Rotational speed change amount DNE is calculated by the following equations (21) - (24) and applied as input parameters to the self-organizing map SOMTS of the in-state engine running state model. In these equations, "k" is a discrete time digitized with a calculation period TC of the target throttle valve opening THCMD. DGAIRCMD = GAIRCMD (k) - GAIRCMD (k - 1) (21) DPB = PB (k) -PB (k-1) (22) DPI = PI (k) -PI (k-1) (23) DNE = NE (k) - NE (k-1) (24)

3 FIG. 13 shows the self-organizing map SOMTS of the in-state engine operating state model for calculating the target throttle valve opening THCMD in this embodiment as a two-dimensional map. This two-dimensional map is represented by the target intake air flow rate change amount DGAIRCMD and defines the boost pressure change amount DPB. The input data vector xTHD is defined by the following equation (25). xTHD = (DGAIRCMD, DPB, DPI, DNE) (25)

A weighting coefficient vector CDi (i = 1-M) shown by the following equation (26) is calculated by training by the same method as that used for training of the self-organizing map SOMSS of the stationary engine running state model. CDi = (CD0i, CD1i, CD2i, CD3i, CD4i) (26)

When the self-organizing map SOMTS of the unsteady engine running state model is used, the target throttle valve opening TH-CMD is calculated by the following equation (27). This equation (27) corresponds to the equation that defines the unsteady motor running mode model in this embodiment. THCMD = CD1i × DGAIRCMD + CD2i × DPB + CD3i × DPI + CD4i × DNE + CD0i (27)

4 FIG. 12 is a graph showing a relationship between the intake air flow rate GAIR [g / sec] and the throttle valve opening TH. The curves L1-L5 in 3 respectively correspond to the states in which the engine rotational speed NE 1000, 1500, 2000, 2500 and 3000 rpm.

How out 4 As can be seen, as the throttle valve opening TH increases, under the condition that the engine rotational speed NE is constant, the intake air flow rate GAIR increases to reach saturation at a substantially constant value (saturation level). At the saturation level, the intake air flow rate GAIR does not change when the throttle valve opening TH changes. In other words, changes in the throttle valve opening TH will not affect the intake air flow rate GAIR when the intake air flow rate GAIR has reached the saturation level (hereinafter referred to as "maximum intake air flow rate GAIRMAX"). Thus, in this embodiment, when the target intake air flow rate GAIRCMD set according to the accelerator operation amount AP and the engine rotation speed NE becomes equal to or greater than a determination threshold GAIRTH, the target throttle valve opening TH-CMD is set to the maximum opening THMAX (for example, "90 Degrees "). The determination threshold GAIRTH is obtained by multiplying a predetermined threshold coefficient KTH (for example, "0.95") by the maximum intake air flow GAIRMAX. According to this setting, the calculation load on the CPU in the ECU 20 be reduced without the performance of the intake air flow rate control / control is deteriorated. On the other hand, when the target intake air flow rate GAIRCMD is smaller than the maximum intake air amount GAIRMAX, the target throttle valve opening THCMD is calculated using the above-described self-organizing map (hereinafter referred to as "SOM"). According to this method of setting the target throttle valve opening THCMD, the throttle valve opening can be set to the target intake air amount GAIRCMD optimally when the actual intake air flow rate GAIR is controlled.

5 FIG. 10 is a flowchart of the method of calculating the target throttle valve opening THCMD. This process is performed at predetermined time intervals by the CPU in the ECU 20 executed.

In step S11, a GAIRCMD map (not shown) is retrieved according to the accelerator operation amount AP and the engine rotation speed NE to calculate the target intake air flow rate GAIRCMD. The GAIRCMD map is set such that the target intake air flow rate GAIRCMD increases as the accelerator pedal operation amount AP increases, and the target intake air flow rate GAIRCMD increases as the engine rotation speed NE increases.

In step S12, a GAIRMAX map (not shown) is retrieved according to the engine rotational speed NE and the boost pressure PB to calculate the maximum intake air flow rate GAIRMAX. The GAIRMAX map is set so that the maximum intake air flow GAIRMAX increases with increase in the engine rotation speed NE, and the maximum intake air flow GAIRMAX increases with increase of the boost pressure PB.

In step S13, the determination threshold GAIRTH is calculated by multiplying the predetermined threshold coefficient KTH by the maximum intake air flow GAIRMAX. In step S14, it is determined whether the target intake air flow rate GAIRCMD is smaller than the determination threshold value GAIRTH is or not. If the answer in step S14 is affirmative (YES), the in 6 SOM calculation methods performed to calculate the target throttle valve opening THCMD using the above-described self-organizing map SOMSS or SOMTS (step S15).

If the target intake air flow rate GAIRCMD is equal to or greater than the determination threshold GAIRTH in step S14, the target throttle valve opening THCMD is set to the maximum opening THMAX.

In step S21 6 For example, the intake air flow rate change amount DGAIRCMD, the boost pressure change amount DPB, the intake pressure change amount DPI, and the rotational speed change amount DNE are calculated by the above-described equations (21) - (24).

In step S22, it is determined whether or not the intake air flow rate change amount DGAIRCMD is greater than a predetermined air flow rate change amount DGATH. If the answer to step S22 is negative (NO), it is further determined whether or not the boost pressure changing amount is larger than a predetermined boost pressure changing amount DPBTH (step S23). If the answer to step S23 is negative (NO), it is further determined whether or not the intake pressure change amount DPI is larger than a predetermined intake pressure change amount DPITH (step S24). If the answer to step S24 is negative (NO), it is further determined whether the rotational speed change amount DNE is greater than a predetermined rotational speed change amount DNETH (step S25).

If any of the answers to steps S22-S25 are affirmative (YES), then it is determined that the engine 1 is in the transient operating state, and the target throttle valve opening THCMD is calculated using the self-organizing map SOMTS of the transient engine running state model (step S27).

On the other hand, if the answer of step S25 is negative (NO), then it is determined that the engine 1 is in the steady-state operation state, and the target throttle valve opening THCMD is calculated by using the self-organizing map SOMSS of the steady state engine operating state model (step S26).

The CPU in the ECU 20 calculates a drive parameter IDTH for driving the actuator 19 and performs a throttle valve opening control / control (intake air flow rate control / control), so that the detected throttle valve opening TH with the by the method of 5 and 6 calculated target throttle valve opening THCMD.

As described above in this embodiment, the drive parameter IDTH of the actuator 19 calculated by means of the stationary engine operating state model and the unsteady engine operating state model. The steady-state engine operating mode model corresponds to the steady-state operation of the engine 1 and outputs the target throttle valve opening THCMD using the self-organizing map SOMSS of the stationary engine running state model. The unsteady engine operating state model corresponds to the transient operation of the engine 1 and outputs the target throttle valve opening THCMD using the self-organizing map SOMTS of the on-state engine running state model. In particular, it is determined if the engine 1 in the transient operating state in steps S22-S25 6 is or not, one of the stationary-engine-operation-state model and the unsteady-engine-operation-state model is selected according to the determination result, and the drive parameter IDTH is calculated according to the target throttle valve opening THCMD which is the output of the selected model. Thus, the target throttle valve opening THCMD, which for each of the stationary operating state and the transient operating state of the engine 1 is adapted to be obtained, whereby a control accuracy is improved with the drive parameter IDTH calculated by means of the target throttle valve opening THCMD.

The target intake air flow rate GAIRCMD, the supercharging pressure PB, the intake pressure PI, and the engine rotation speed NE are input to the self-organizing map SOMSS of the steady-state engine operating state model, while the amounts of change from those input parameters to the self-organizing map SOMTS of the transient engine operating state model be entered. Further, in the stationary operating state of the engine, only the calculation is made with respect to the self-organizing map SOMSS of the stationary-engine running-state model, and in the transient operating state of the engine, only the calculation with respect to the self-organizing map SOMTS of the on-state engine running state -Models performed. Thus, it is not necessary to perform the calculations corresponding to the neural networks in the two models, thereby suppressing the computation load.

Further, it is determined that the engine 1 is in the transient operating state when at least one of the change amounts corresponding to the target intake air flow rate GAIRCMD, the boost pressure PB, the intake pressure PI, and the engine rotation speed NE, which are the input parameters of the self-organizing map SOMSS of the steady state engine operation state model, is greater than is the predetermined amount of change. Accordingly, the transient operating state in which the self-organizing map SOMTS of the model is to be used for a transient state can be appropriately determined.

In this embodiment, the ECU forms 20 the target intake air flow rate calculating means, the on-state engine operating state determining means and the selecting means. In particular, the target throttle valve opening THCMD corresponds to the predetermined operating parameter, the drive parameter IDTH of the actuator 19 corresponds to the motor control / control parameter, the method off 5 corresponds to a part of the control parameter calculating means and the method of FIG 6 corresponds to the unsteady engine operating state determining means and the selection means.

The present invention is not limited to the above-described embodiment, and various modifications can be made. For example, in the above-described embodiment, an example is shown that the target throttle valve opening THCMD is the predetermined operating parameter. Alternatively, as the predetermined operating parameter, one from the engine 1 NOx amount, exhaust gas recirculation flow rate (or exhaust gas recirculation amount, or target exhaust gas recirculation amount) or intake air flow rate can be calculated, and the fuel injection amount (control parameter) can be calculated according to the calculated predetermined operation parameter.

When the NOx amount is calculated, the engine rotation speed NE, the fuel supply amount (fuel injection amount), the air-fuel ratio, a temperature of the turbine 11 flowing exhaust gases, the supercharging pressure PB, the intake pressure PI and the intake air flow GAIR applied as input parameters of the self-organizing map of the stationary engine operating state model for calculating the amount of NOx and the amounts of change, the input parameters of the self-organizing map of the stationary engine operating state Models are used as input parameters of the self-organizing map of the transient engine running state model.

When calculating the ratio of the exhaust recirculation, the supercharging pressure PB, the intake pressure PI, the EGR valve opening, the intake air flow rate GAIR, the air-fuel ratio, the engine rotation speed NE, the blade opening θvgt of the turbine become 11 and the temperature of the recirculated exhaust gases used as the input parameters of the self-organizing map of the steady-state engine operating state model to calculate the ratio of the exhaust recirculation, and the amounts of change corresponding to the input parameters of the self-organizing map of the steady state engine operating state model is applied as the input parameters of the self-organizing map of the transient engine running state model.

When the intake air flow rate is calculated, the throttle valve opening TH, the boost pressure PB, the intake pressure PI, and the engine rotation speed NE are used as input parameters of the self-organizing map of the steady state engine operation state model to calculate the intake air flow rate and the change amounts corresponding to the input parameters The self-organizing map of the stationary engine running state model is used as the input parameters of the self-organizing map of the in-state engine running state model.

Further, in the above-described embodiment, the self-organizing map is used as the neural network. Alternatively, the neural network known as the so-called "perceptron" may be used.

The present invention can also be used to control a marine engine drive motor, such as an outboard motor having a vertically extending crankshaft.

LIST OF REFERENCE NUMBERS

1

internal combustion engine

2

intake

19

actuator

20

electronic control unit (control parameter calculating means, unsteady engine operating condition determining means, selecting means)

22

Boost pressure sensor

24

intake pressure

27

accelerometer

28

Motor rotation speed sensor

Claims (2)

Control system for an internal combustion engine ( 1 ), the control system characterized in that it comprises: control parameter calculating means ( 20 ) for calculating a control parameter (IDTH) of the engine ( 1 ) using a stationary engine operating state model and an in-stationary engine operating state model, wherein the stationary engine operating state model is for stationary operation of the engine ( 1 ) and a predetermined operating parameter (THCMD) of the engine ( 1 ) using a neural network (SOMSS), and the non-stationary engine mode model outputs a transient operation of the engine ( 1 ) and outputs the predetermined operating parameter (THCMD) using another neural network (SOMTS), wherein the control parameter calculating means ( 20 ) include: unsteady engine operating condition determining means ( 20 ) for determining whether the engine ( 1 ) is in the transient operating state or not; and selection means ( 20 ) for selecting one of the stationary-motor-operation-state model and the unsteady-motor-operation-state model according to the determination result of the on-state engine operating-condition determining means, wherein the control-parameter calculating means ( 20 ) calculate the control / control parameter (IDTH) in accordance with the operating parameter (THCMD) output of the selected model, the non-stationary engine operating condition determining means (17) 20 ) determine that the engine ( 1 ) is in the steady-state operating state when change amounts (DGAIRCMD, DPB, DPI, DNE) of input parameters (GAIRCMD, PB, PI, NE) used to calculate the control / parameter (IDTH) of the engine ( 1 ) are less than a predetermined amount of change (DGATH, DPBTH, DPITH, DNETH) during a period of time, the non-stationary engine operating condition determining means determining that the engine is ( 1 ) is in the transient operating state when at least one of change amounts (DGAIRCMD, DPB, DPI, DNE) of the input parameters (GAIRCMD, PB, PI, NE) input to the steady-state engine operating mode model is greater than one during a period of time predetermined amount of change (DGATH, DPBTH, DPITH, DNETH), and wherein a first set of input parameters (GAIRCMD, PB, PI, NE) is input to the stationary engine mode model and a second set of input parameters (DGAIRCMD, DPB, DPI , DNE) is input to the unsteady engine mode model, the second set of input parameters (DGAIRCMD, DPB, DPI, DNE) changing amounts of the first set of input parameters (GAIRCMD, PB, PI, NE) for a constant time period (TC ) corresponds. Control system according to claim 1, wherein an intake pipe ( 2 ) of the motor ( 1 ) with a throttle valve ( 3 ) and the predetermined operating parameter (THCMD) is a target opening of the throttle valve ( 3 ), wherein the control parameter calculating means ( 20 ) calculates the target opening (THCMD) using the steady state engine operating state model or the unsteady engine operating state model when a target intake air amount (GAIRCMD) of the engine (FIG. 1 ) is smaller than a determination threshold (GAIRTH), and the control parameter calculating means (20) sets the target opening (THCMD) to a maximum opening (THMAX) when the target intake air amount (GAIRCMD) is equal to or greater than the determination threshold value (GAIRTH). is.

Images