JP2022024224A - Control device for internal combustion engine - Google Patents

Abstract

To provide a control device for an internal combustion engine, which can calculate various control variables of the internal combustion engine while reducing the number of calculations as much as possible, which use a neural network representing a characteristic of torque or thermal efficiency for an operation state of the internal combustion engine including ignition timing.SOLUTION: A control device for an internal combustion engine: calculates N pieces of output information corresponding to a current operation state and each of N pieces of ignition timing by using a neural network in which a relationship between an operation state and ignition timing, and output information which is torque or a thermal efficiency of the internal combustion engine is learned in advance; calculates a coefficient of an approximate function approximating a relationship between the N pieces of ignition timing and the N pieces of output information; calculates output information corresponding to ignition timing set for control by using the coefficient; and calculates a control amount of the internal combustion engine based on the calculated output information.SELECTED DRAWING: Figure 5

Description

The present application relates to a control device for an internal combustion engine.

In recent years, the torque of the internal combustion engine, which is a physical quantity directly acting on the control of the vehicle, has been used as the required value of the internal combustion engine output received from the driver and each vehicle system (hybrid motor control, transmission control, brake control, traction control, etc.). With this as the target value of the output of the internal combustion engine, the amount of air, the amount of fuel, the ignition timing, etc., which are the control amounts of the internal combustion engine, are determined, and the actual torque is estimated from the actual operating state of the internal combustion engine for each vehicle system. A control device and a control method for an internal combustion engine have been proposed, which can realize cooperative control and obtain good running performance by transmitting to.

Such a control method is generally called torque-based control, but in this method of control method, it is important to be able to accurately calculate the actual torque based on the operating state of the internal combustion engine.

In addition, torque is basically used as a common control index, but in recent years, in order to give more freedom to control, the torque characteristics for ignition timing and the parameter group for calculating it (the torque that you want to finally obtain). Alternatively, a characteristic factor that constitutes thermal efficiency, which is the basis of torque calculation, may be used as a common control index at the same time.

Normally, in order to estimate the torque of an internal combustion engine, each parameter for calculating the torque characteristic with respect to the ignition timing described above is decomposed according to the operating state, and the one divided into elements that are different characteristic factors is controlled. It is stored in the device in advance. Then, using these stored elements, parameters are often calculated according to the operating state to estimate the torque, which is also shown in Patent Document 1.

By the way, as a method of estimating torque from an operating state of an internal combustion engine, a method applying a neural network technique as in Patent Document 2, for example, has been proposed in addition to the calculation method using map data as described above. ing. Here, the neural network is a mathematical model that aims to express some characteristics found in brain functions by simulation on a computer, and is applied to a feedforward neural network (FNN) in advance with respect to an input value. If the output value is trained as teacher data, it can be used as a general-purpose approximation function that simulates the relationship between the learned input value and the output value. As a learning method of a neural network, an error back propagation method (backpropagation method) is generally known.

Patent No. 322508 Japanese Unexamined Patent Publication No. 11-35145

In contrast to the mechanism for internal combustion engine control, which has become more complicated in recent years to improve fuel efficiency, the internal combustion engine control system has also become more complicated, and an increase in the number of man-hours required for adaptation has become a major problem. As an example of a mechanism for controlling an internal combustion engine that becomes complicated, an intake / exhaust VVT (Variable Valve Timing), a variable valve lift, a variable compression ratio, a turbocharger, a swirl control valve, a tumble control valve, and the like are known. In the case of the control method using map data as in Patent Document 1, if the mechanism for controlling the internal combustion engine becomes complicated, more map data is required, and there is a problem that the matching man-hours also increase accordingly. .. From the viewpoint of testing the internal combustion engine required for conformity, commercially available MBC (Model Based Calibration) tools have been enhanced in recent years. For example, a test plan for an internal combustion engine is made based on DOE (Design of Experiments), data is collected in conjunction with the test equipment of the internal combustion engine, and a statistical model of the internal combustion engine is created from the results. Map data used for control can be created based on this model.

However, although map data can be created by the MBC tool, it takes a considerable amount of man-hours to create a lot of map data, and it is also a lot to manage the data for each internal combustion engine model. Man-hours are required. Furthermore, when map data for control is created from the statistical model of the MBC tool, it is considered that the accuracy will decrease because the number of parameters of the operating state of the internal combustion engine that can be considered decreases, so control using this map data is considered. A lot of man-hours are required for checking the accuracy and making fine adjustments. As described above, there is a problem that the conventional map control still requires a huge amount of man-hours for conforming even if the MBC tool or the like is introduced.

Further, regarding the method of estimating the torque from the operating state of the internal combustion engine using a feedforward neural network (FNN) as in Patent Document 2, in the conventional method in which there is only one intermediate layer, FNN is used as an approximate function. There was a problem that sufficient accuracy could not be obtained even if it was used. From the viewpoint of approximation accuracy, a method called deep learning has been known in recent years. For example, by increasing the number of layers (deepening) of the same neural network as in the past, the accuracy as an approximate function can be greatly improved. While the conventional learning method could not perform good learning due to the vanishing gradient problem, various learning techniques developed in recent years have made it possible to perform learning well. This deep learning is also known as one method of artificial intelligence (AI) and machine learning, which has been attracting attention in recent years.

Therefore, if the torque or thermal efficiency is estimated from the operating state of the internal combustion engine using FNN as an approximate function, the torque and thermal efficiency can be estimated by creating teacher data with the MBC tool and learning it. It is considered that it can be done well with the limited number of conforming man-hours. Furthermore, since a neural network may be used as one of the methods for creating a statistical model of an internal combustion engine in the MBC tool, the torque is estimated from the operating state of the internal combustion engine using the statistical model of the internal combustion engine created by the MBC tool itself. In this case, the man-hours can be further reduced.

However, as it is, only the torque is estimated, and the target ignition timing and the target intake air amount, which are the control amounts of the internal combustion engine, cannot be calculated. For example, if the ignition timing or the intake air amount is changed little by little and the torque is repeatedly calculated using a function such as FNN, the target ignition timing or the target intake air amount that realizes the target torque can be searched for. However, if the calculation using a function such as FNN is repeated, the calculation load increases. In particular, when the system configuration becomes complicated and the functions such as FNN become complicated, the arithmetic load of FNN increases significantly. Further, there are a plurality of types of control amounts of the internal combustion engine, and the calculation method thereof also varies depending on the control amount. Therefore, the number of FNN calculations increases in order to correspond to the type of control amount and the calculation method. Therefore, it is desired to reduce the number of operations using a function such as FNN as much as possible.

Therefore, the present application can calculate various control amounts of an internal combustion engine by reducing the number of calculations using a neural network representing the characteristics of torque or thermal efficiency with respect to the operating state of the internal combustion engine including ignition timing as much as possible. It is an object of the present invention to provide the control device of.

The control device for the internal combustion engine of the present application is
Set the ignition timing for calculating N coefficients (N is an integer of 2 or more),
Using a neural network in which the relationship between the operating state and ignition timing for calculating the output of the internal combustion engine and the output information which is the torque or thermal efficiency of the internal combustion engine is learned in advance, the current operating state for calculating the output and the N An ignition-compatible output calculation unit that calculates N of the output information corresponding to each of the ignition timings for calculating the coefficients, and an ignition-compatible output calculation unit.
A coefficient calculation unit that calculates the coefficient of an approximate function that approximates the relationship between the ignition timing for calculating the N coefficients and the output information of the N elements.
A control amount calculation unit that calculates the output information corresponding to the ignition timing set for control using the coefficient and calculates the control amount of the internal combustion engine based on the calculated output information.
An engine control unit that controls an internal combustion engine based on the amount of control of the internal combustion engine,
It is equipped with.

According to the control device of the internal combustion engine, once the coefficient of the approximation function that approximates the relationship between the ignition timing and the output information under the current operating conditions is calculated, various calculations using the coefficient with a small processing load can be performed. Output information corresponding to the ignition timing for control can be calculated, and the control amount of various internal combustion engines can be calculated based on the output information. Therefore, it is not necessary to repeatedly perform the calculation using the neural network, which has a large processing load, and the calculation processing load can be reduced.

It is a schematic block diagram of the internal combustion engine and the control device which concerns on Embodiment 1. FIG. It is a schematic block diagram of the control device which concerns on Embodiment 1. FIG. It is a hardware block diagram of the control device which concerns on Embodiment 1. FIG. It is a figure which shows the neural network of the output characteristic which concerns on Embodiment 1. FIG. It is a block diagram of the torque interface part which concerns on Embodiment 1. FIG. It is a figure explaining the approximation function which concerns on Embodiment 1. FIG. It is a block diagram of the torque interface part which concerns on Embodiment 2. FIG. It is a figure explaining the approximation function which concerns on Embodiment 2.

Embodiment 1.
The internal combustion engine control device 30 (hereinafter, simply referred to as a control device 30) according to the first embodiment will be described with reference to the drawings. FIG. 1 is a schematic configuration diagram of an internal combustion engine 1 according to the present embodiment, and FIG. 2 is a schematic block diagram of a control device 30 according to the present embodiment. The internal combustion engine 1 and the control device 30 are mounted on the vehicle, and the internal combustion engine 1 serves as a driving force source for the vehicle (wheels).

1. 1. Configuration of Internal Combustion Engine 1 As shown in FIG. 1, the internal combustion engine 1 includes a combustion chamber 25 that burns a mixture of air and fuel. The internal combustion engine 1 includes an intake pipe 23 for supplying air to the combustion chamber 25 and an exhaust pipe 17 for discharging the exhaust gas burned in the combustion chamber 25. The combustion chamber 25 is composed of a cylinder (cylinder) and a piston. Hereinafter, the combustion chamber 25 is also referred to as a cylinder. The internal combustion engine 1 is a gasoline engine. The internal combustion engine 1 includes a throttle valve 6 that opens and closes the intake pipe 23. The throttle valve 6 is an electronically controlled throttle valve that is opened and closed by an electric motor controlled by the control device 30. The throttle valve 6 is provided with a throttle opening sensor 7 that outputs an electric signal according to the opening degree of the throttle valve 6.

The intake pipe 23 on the upstream side of the throttle valve 6 has an air flow sensor 3 that outputs an electric signal according to the flow rate of the intake air sucked into the intake pipe 23, and an intake air that outputs an electric signal according to the temperature of the intake air. A temperature sensor 4 is provided. The temperature of the intake air detected by the intake air temperature sensor 4 can be regarded as equal to the outside air temperature.

The internal combustion engine 1 has an EGR flow path 21 that recirculates exhaust gas from the exhaust pipe 17 to the intake manifold 12, and an EGR valve 22 that opens and closes the EGR flow path 21. The intake manifold 12 is a portion of the intake pipe 23 on the downstream side of the throttle valve 6. The EGR valve 22 is an electronically controlled EGR valve that is opened and closed by an electric motor controlled by the control device 30. The EGR valve 22 is provided with an EGR opening sensor 27 that outputs an electric signal according to the opening degree of the EGR valve 22. EGR is an acronym for Exhaust Gas Recirculation, that is, Exhaust Gas Recirculation. The EGR in which the exhaust gas recirculates through the EGR valve 22 is called an external EGR, and the EGR in which the exhaust gas remains in the combustion chamber due to the valve overlap of the intake / exhaust valves is called an internal EGR. Hereinafter, the external EGR is simply referred to as EGR.

The intake manifold 12 has a manifold pressure sensor 8 that outputs an electric signal corresponding to the manifold pressure, which is the pressure of the gas in the intake manifold 12, and an electric signal corresponding to the manifold temperature, which is the temperature of the gas in the intake manifold 12. A manifold temperature sensor 9 for output is provided.

The internal combustion engine 1 is provided with an injector 13 that supplies fuel to the combustion chamber 25. The injector 13 is provided so as to inject fuel directly into the combustion chamber 25. The injector 13 may be provided so as to inject fuel into a portion on the downstream side of the intake manifold 12. The internal combustion engine 1 is provided with an atmospheric pressure sensor 2 that outputs an electric signal corresponding to the atmospheric pressure.

The top of the combustion chamber 25 is provided with a spark plug that ignites a mixture of air and fuel, and an ignition coil 16 that supplies ignition energy to the spark plug. Further, at the top of the combustion chamber 25, an intake valve 14 for adjusting the amount of intake air sucked from the intake pipe 23 into the combustion chamber 25 and an exhaust gas amount discharged from the combustion chamber 25 to the exhaust pipe 17 are adjusted. An exhaust valve 15 is provided. The intake valve 14 is provided with an intake variable valve timing mechanism that changes the valve opening / closing timing. The exhaust valve 15 is provided with a variable exhaust valve timing mechanism that changes the opening / closing timing of the valve. The variable valve timing mechanisms 14 and 15 have an electric actuator. A mechanism such as a variable valve lift, a variable compression ratio, a turbocharger, a swirl control valve, and a tumble control valve may be provided. The crank shaft of the internal combustion engine 1 is provided with a crank angle sensor 20 that outputs an electric signal corresponding to the rotation angle thereof. A knock sensor 28 is fixed to the cylinder block.

The exhaust pipe 17 is provided with an air-fuel ratio sensor 18 that outputs an electric signal according to the air-fuel ratio AF (Air / Fuel), which is the ratio of air in the exhaust gas to fuel. Further, the exhaust pipe 17 is provided with a catalyst 19 for purifying the exhaust gas.

2. 2. Configuration of Control Device 30 Next, the control device 30 will be described. The control device 30 is a control device that controls the internal combustion engine 1. As shown in the block diagram of FIG. 2, the control device 30 includes control units such as a torque control unit 31, a torque interface unit 32, and an engine control unit 33. Each of the control units 31 to 33 and the like of the control device 30 is realized by a processing circuit provided in the control device 30. Specifically, as shown in FIG. 3, the control device 30 is a storage device 91 that exchanges data with an arithmetic processing unit 90 (computer) such as a CPU (Central Processing Unit) and an arithmetic processing unit 90 as a processing circuit. The arithmetic processing unit 90 includes an input circuit 92 for inputting an external signal, an output circuit 93 for outputting a signal from the arithmetic processing unit 90 to the outside, a communication circuit 94, and the like.

The arithmetic processing device 90 is provided with an ASIC (Application Specific Integrated Circuit), an IC (Integrated Circuit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), various logic circuits, various signal processing circuits, and the like. You may. Further, as the arithmetic processing unit 90, a plurality of the same type or different types may be provided, and each processing may be shared and executed. As the storage device 91, a RAM (Random Access Memory) configured to be able to read and write data from the arithmetic processing unit 90, a ROM (Read Only Memory) configured to be able to read data from the arithmetic processing unit 90, and the like are used. It is prepared. The input circuit 92 includes an A / D converter or the like to which various sensors and switches are connected and the output signals of these sensors and switches are input to the arithmetic processing device 90. The output circuit 93 includes a drive circuit or the like to which an electric load is connected and a control signal is output from the arithmetic processing unit 90 to the electric load.

The communication circuit 94 is an external control such as a transmission control device 95 for controlling a transmission, a motor control device 96 for controlling a motor provided in a hybrid vehicle, and a brake / traction control device 97 for performing brake control and traction control. It is connected to the device via a communication line and performs wired communication based on a communication protocol such as CAN (Controller Area Network).

Then, in each function of the control units 31 to 33 and the like included in the control device 30, the arithmetic processing unit 90 executes software (program) stored in the storage device 91 such as a ROM, and the storage device 91 and the input circuit 92. , And by cooperating with other hardware of the control device 30 such as the output circuit 93. The setting data such as the neural network, each function, and the constant used by the control units 31 to 33 and the like are stored in the storage device 91 such as ROM as a part of the software (program).

In the present embodiment, the input circuit 92 includes an atmospheric pressure sensor 2, an air flow sensor 3, an intake air temperature sensor 4, a throttle opening sensor 7, a manifold pressure sensor 8, a manifold temperature sensor 9, an air fuel ratio sensor 18, and a crank angle. A sensor 20, an accelerator opening sensor 26, an EGR opening sensor 27, a knock sensor 28, and the like are connected. A throttle valve 6 (electric motor), an injector 13, an intake variable valve timing mechanism 14, an exhaust variable valve timing mechanism 15, an ignition coil 16, an EGR valve 22 (electric actuator) and the like are connected to the output circuit 93. Various sensors, switches, actuators and the like (not shown) are connected to the control device 30.

2-1. The torque-based control control device 30 executes torque-based control for controlling the internal combustion engine 1 based on the target torque. As described above, the control device 30 generally includes a torque control unit 31, a torque interface unit 32, and an engine control unit 33. The torque control unit 31 calculates the target torque. The torque interface unit 32 calculates a target value of the controlled amount of the internal combustion engine based on the target torque and the like. The engine control unit 33 drives and controls various electric loads based on the target value of the controlled amount.

<Torque control unit 31>
The torque control unit 31 calculates the driver-required torque, which is the torque required by the driver for the internal combustion engine 1, based on the actual accelerator opening detected by the accelerator opening sensor 26. Further, the torque control unit 31 calculates the idling torque, which is the torque required to maintain the rotational speed during the idling operation. Further, the torque control unit 31 calculates an externally required torque which is a torque required from an external control device such as a transmission control device 95, a motor control device 96, and a brake traction control device 97. Then, the torque control unit 31 determines the priority order of the driver required torque, the idling torque, and the external required torque, and calculates the target torque (such calculation is also referred to as torque arbitration).

Here, the target torque includes a low response target torque Trqts and a high response target torque Trqtf. The low response target torque Trqts is the torque required for the internal combustion engine without the ignition timing retard, and the high response target torque Trqtf includes retarding the ignition timing and is required for the internal combustion engine. The torque is. Normally, the low response target torque Trqts and the high response target torque Trqtf match, but when there is a torque down request due to the retard angle of the ignition timing, the high response target torque Trqtf becomes higher than the low response target torque Trqts. Will also be low.

The torque control unit 31 mainly calculates the low response target torque Trqts based on the large one of the driver required torque and the idling torque in the steady state, and is based on the external required torque and the idling torque when the load changes. The high response target torque Trqtf is calculated.

<Torque interface unit 32>
The torque interface unit 32 performs mutual conversion between the target torque and the filling efficiency and mutual conversion between the target torque and the ignition timing based on the operating state of the internal combustion engine, and calculates the target filling efficiency Ect and the target ignition timing θight. , Is transmitted to the engine control unit 33. Further, the torque interface unit 32 calculates the actual torque Trqr based on the operating state of the internal combustion engine and transmits it to the torque control unit 31. The detailed processing of the torque interface unit 32 will be described later.

<Operating state detection unit 330>
The engine control unit 33 includes an operation state detection unit 330 that detects the operation state of the internal combustion engine. The operation state detection unit 330 detects various operation states based on output signals of various sensors and the like. Specifically, the operating state detection unit 330 detects the actual atmospheric pressure based on the output signal of the atmospheric pressure sensor 2, detects the actual intake air flow rate based on the output signal of the air flow sensor 3, and determines the intake air temperature sensor. The actual outside temperature is detected based on the output signal of 4, the actual throttle opening is detected based on the output signal of the throttle opening sensor 7, and the actual manifold pressure is detected based on the output signal of the manifold pressure sensor 8. The actual manifold temperature, which is the temperature of the gas in the intake manifold 12, is detected based on the output signal of the manifold temperature sensor 9, and the actual air-fuel ratio AFr of the exhaust gas is detected based on the output signal of the air-fuel ratio sensor 18. The actual accelerator opening degree is detected based on the output signal of the accelerator opening degree sensor 26, and the actual EGR opening degree is detected based on the output signal of the EGR opening degree sensor 27.

The operating state detection unit 330 detects the crank angle and the actual rotation speed Ne based on the output signal of the crank angle sensor 20. The operating state detection unit 330 includes the actual phase angle IVTr of the intake variable valve timing mechanism 14 (hereinafter referred to as intake VVT 14) and the exhaust variable based on the phase difference between the edge of the cam angle sensor (not shown) and the crank angle. The actual phase angle EVTr of the valve timing mechanism 15 (hereinafter referred to as exhaust VVT15) is detected.

The operating state detection unit 330 detects in-cylinder intake amount information, which is information on the amount of air sucked into the combustion chamber 25. The operating state detection unit 330 uses the actual intake air flow rate, the actual rotation speed Ner, and the like as the in-cylinder intake amount information, that is, the actual intake air amount Qcr [g / stroke] sucked into the combustion chamber 25, and the actual filling. Calculate the efficiency Ecr [%]. For example, the operating state detection unit 330 actually multiplies the actual intake air flow rate [g / s] by the stroke cycle corresponding to the rotation speed Ne, and performs a filter process simulating the delay of the intake manifold. Calculated as the intake air amount Qcr [g / stroke]. Alternatively, the operating state detection unit 330 may calculate the actual intake air amount Qcr [g / stroke] and the actual filling efficiency Ecr [%] based on the actual manifold pressure, the actual rotation speed Ner, and the like.

The operating state detection unit 330 calculates the actual EGR amount [g / stroke], which is the actual exhaust gas recirculation amount sucked into the combustion chamber 25, based on the EGR opening degree and the like. For example, the operating state detection unit 330 calculates the actual EGR flow rate [g / s] passing through the EGR valve 22 based on the EGR opening degree, the manifold pressure, and the like, and multiplies the actual EGR flow rate by the stroke cycle to obtain a value. The filtered value is calculated as the actual EGR amount [g / stroke]. The operating state detection unit 330 calculates the actual EGR rate Regrr [%], which is the ratio of the actual EGR amount to the actual intake air amount.

<Intake amount control unit 331>
The engine control unit 33 includes an intake air amount control unit 331 that controls the intake air amount. The intake air amount control unit 331 calculates the target intake air amount from the target filling efficiency Ect, and calculates the target intake air flow rate from the target intake air amount. The engine control unit 33 calculates the target throttle opening degree based on the actual intake air flow rate and the actual manifold pressure so as to achieve the target intake air flow rate, and drives and controls the electric motor of the throttle valve 6.

The intake amount control unit 331 calculates the target phase angle IVTt (hereinafter referred to as the target intake phase angle IVTt) of the intake VVT14 based on the actual rotation speed Ner and the target filling efficiency Ect, and drives and controls the electric actuator of the intake VVT14. do. The intake amount control unit 331 calculates the target phase angle EVTt (hereinafter referred to as the target exhaust phase angle EVTt) of the exhaust VVT15 based on the actual rotation speed Ner and the target filling efficiency Ect, and drives and controls the electric actuator of the exhaust VVT15. do. The intake amount control unit 331 calculates the target EGR opening degree based on the actual rotation speed Ner and the target filling efficiency Ect, and drives and controls the electric actuator of the EGR valve 22.

<Fuel control unit 332>
The engine control unit 33 includes a fuel control unit 332 that controls the fuel injection amount. The fuel control unit 332 calculates the fuel injection amount for achieving the target air-fuel ratio AFt based on the actual filling efficiency Ecr, and drives and controls the injector 13. The fuel control unit 332 calculates the target air-fuel ratio AFt based on the actual rotation speed Ner and the target filling efficiency Ect.

<Ignition control unit 333>
The engine control unit 33 includes an ignition control unit 333 that energizes the ignition coil. The ignition control unit 333 determines the final ignition timing θsa based on the target ignition timing θight transmitted from the torque interface unit 32. When the knock is detected by the knock sensor 28, the ignition control unit 333 corrects the retard angle with respect to the target ignition timing θight so that the knock does not occur, and calculates the final ignition timing θsa. Further, the ignition control unit 333 sets the ignition timing on the retard side by the retard upper limit ignition timing θrtd so that the final ignition timing θsa is not set on the retard side of the retard upper limit ignition timing θrtd to prevent misfire. Limit the retard angle. Then, the ignition control unit 333 controls the energization of the ignition coil 16 based on the final ignition timing θsa. This final ignition timing θsa is the actual ignition timing θsa. In this embodiment, the ignition timing is set by the crank angle.

2-2. Detailed configuration of the torque interface unit 32 As described above, the torque interface unit 32 performs mutual conversion between torque and filling efficiency and mutual conversion between torque and ignition timing based on the operating state of the internal combustion engine, and achieves the target filling efficiency. Calculate the Ect and the target ignition timing θight. As shown in FIG. 5, the torque interface unit 32 includes an ignition-compatible output calculation unit 321, a coefficient calculation unit 322, and a control amount calculation unit 323.

2-2-1. Ignition compatible output calculation unit 321
<Neural network of output characteristics>
The torque interface unit 32 is a neural network in which the relationship between the operating state and ignition timing for calculating the output of the internal combustion engine and the output information (thermal efficiency η in this example) which is the torque or thermal efficiency of the internal combustion engine is learned in advance (hereinafter referred to as a neural network). , Called a neural network of output characteristics) is stored.

In the present embodiment, as shown in FIG. 4, the neural network having output characteristics is composed of a feedforward neural network (FNN). FNN is a network in which units (also referred to as nodes and neurons) arranged on a hierarchy have a structure in which they are connected between adjacent layers, and information is propagated from an input side to an output side. In the operation performed by a unit, the value input from each unit in the previous layer is weighted, and the bias is added to the total input to that unit, and after this total input is passed through the activation function. Is the output of the unit.

In order to use the FNN composed of such units as an approximate function, it is necessary to adjust the weight and bias of each unit so that the input value to the FNN and the output value thereof have a desired relationship. This adjustment is performed by preparing a large number of input value and output value data sets called teacher data in advance and applying a method called an error backpropagation method (backpropagation method). Adjusting the weight and bias in this way is called neural network learning, and if it can be learned well, FNN can be used as a general-purpose function that stores the characteristics of the teacher data. The composition of FNN after adjustment and learning, and the weight and bias of each unit (neuron) are stored in the storage device 91 in advance.

It is thought that the larger the number of layers of FNN and the larger the number of units contained in the layers, the better the approximation accuracy. In such cases (this is called overfitting or overfitting), the necessary approximation accuracy can be obtained by stopping the learning in the middle to suppress overfitting, increasing the number of teacher data, and so on. Adjustments need to be made. The above is the outline of FNN, but since FNN and its learning method are known techniques, FNN will be described here as known.

In the example shown in FIG. 4, as the configuration of FNN, the operation state for output calculation and the ignition timing θig are input to the input layer, there are three intermediate layers having five units, and the thermal efficiency η is output in the output layer. It is configured to be.

In the present embodiment, the operating states for calculating the output are the rotation speed Ne, the filling efficiency Ec, the EGR rate Regr, and the air-fuel ratio AF. Environmental conditions such as water temperature, intake air temperature, atmospheric pressure, and manifold temperature may be used as the operating state for output calculation, and other internal combustion engines such as intake phase angle IVT, exhaust phase angle EVT, and EGR valve opening. It may be configured so that the operating state of the engine is input. Further, when the system configuration of the internal combustion engine is different, the operating state of the system configuration (for example, variable valve lift, variable compression ratio, etc.) may be input. Further, with respect to the intermediate layer, the number of units of each layer and the number of layers themselves may be increased or decreased. These are parameters that should be adjusted according to the approximation accuracy at the time of learning FNN performed in advance.

In the present embodiment, an example in which the thermal efficiency η is calculated by FNN is shown as output information, but the torque Trq may be calculated by FNN as output information. Regardless of whether the thermal efficiency η or the torque Trq is calculated, as shown in the equation (3) or the like, the thermal efficiency η and the torque Trq are mutually used by using the intake air amount, the air-fuel ratio, the heat amount of the fuel per unit mass, and the like. Can be converted to. The thermal efficiency η is a ratio of the amount of heat of the fuel supplied to the combustion chamber that is converted into torque.

<Ignition timing setting for calculating N coefficients>
The ignition-corresponding output calculation unit 321 sets ignition timings (N is an integer of 2 or more) for calculating N coefficients. In the present embodiment, the ignition-corresponding output calculation unit 321 has ignition timings for calculating N coefficients, which are larger than the maximum torque ignition timing θMBT, which is the ignition timing at which the torque of the internal combustion engine is maximized, and the maximum torque ignition timing. The ignition timing for calculating the two coefficients of the retard side ignition timing θspl, which is the retard side ignition timing, is set.

The ignition-compatible output calculation unit 321 uses an MBT setting function in which the relationship between the operating state for MBT calculation of the internal combustion engine and the maximum torque ignition timing θMBT is preset, and the maximum corresponding to the current operating state for MBT calculation. Calculate the torque ignition timing θMBT. The operating state for calculating the MBT is set to the operating state such as the rotation speed Ne, the filling efficiency Ec, the EGR rate Regr, and the air-fuel ratio AF. In the present embodiment, the operating state for MBT calculation is set to be the same as the operating state for output calculation.

The MBT setting function may be composed of a neural network or a plurality of map data. When composed of a plurality of map data, the ignition corresponding output calculation unit 321 uses the map data in which the relationship between the rotation speed and the filling efficiency and the basic value of the maximum torque ignition timing is preset, and the current rotation speed is used. And the map data in which the basic value of the maximum torque ignition timing corresponding to the filling efficiency is calculated and the relationship between the EGR rate and the correction value is preset, and the map data in which the relationship between the air-fuel ratio and the correction value is preset. The correction value corresponding to the current EGR rate and the air-fuel ratio is calculated, and the basic value of the maximum torque ignition timing is corrected by the correction value to calculate the maximum torque ignition timing θMBT.

For example, the ignition corresponding output calculation unit 321 sets an angle obtained by adding a preset retard angle Δθrtd (for example, 10 degrees) to the maximum torque ignition timing θMBT as the retard side ignition timing θspl. Alternatively, the ignition corresponding output calculation unit 321 may set the retard angle upper limit ignition timing θrtd, which will be described later, to the retard angle side ignition timing θspl.

<Calculation of N output information>
The ignition-corresponding output calculation unit 321 uses the above-mentioned neural network of output characteristics to calculate N output information corresponding to each of the current operation state for output calculation and the ignition timing for N coefficient calculation. .. In the present embodiment, the ignition-corresponding output calculation unit 321 uses a neural network of output characteristics, and corresponds to the current operating state for output calculation and the ignition timings θMBT and θspl for calculating two coefficients, respectively. Calculate the thermal efficiencies ηMBT and ηspl. The thermal efficiency ηMBT corresponding to the maximum torque ignition timing θMBT is referred to as MBT thermal efficiency ηMBT, and the thermal efficiency ηspl corresponding to the retarded angle side ignition timing θspl is referred to as a retarded angle side thermal efficiency ηspl.

Specifically, the ignition-corresponding output calculation unit 321 is in operation for calculating the current output (in this example, the actual rotation speed Ne, the actual filling efficiency Ecr, the actual EGR rate Regrr, and the target air-fuel ratio AFt or the actual air-fuel ratio. AFr) and the maximum torque ignition timing θMBT are input to the neural network of the output characteristics, each unit of the neural network is calculated, and the thermal efficiency η output from the neural network is calculated as the MBT thermal efficiency ηMBT. Further, the ignition-compatible output calculation unit 321 inputs the current operation state for output calculation and the retard side ignition timing θspl to the neural network of the output characteristics, performs calculation processing for each unit of the neural network, and performs calculation processing from the neural network. The output thermal efficiency η is calculated as the retard side thermal efficiency ηspl.

The ignition-compatible output calculation unit 321 uses map data in which the relationship between the current output calculation operating state and the MBT thermal efficiency ηMBT is preset, and obtains the MBT thermal efficiency ηMBT corresponding to the current output calculation operating state. It may be configured to calculate. In this case, the ignition-corresponding output calculation unit 321 uses the map data in which the relationship between the rotation speed and the filling efficiency and the basic value of the MBT thermal efficiency is set in advance to obtain the MBT thermal efficiency corresponding to the current rotation speed and the filling efficiency. Corresponds to the current EGR rate and air-fuel ratio by calculating the basic value and using the map data in which the relationship between the EGR rate and the correction value is preset and the map data in which the relationship between the air-fuel ratio and the correction value is preset. The MBT thermal efficiency ηMBT is calculated by calculating the correction value to be performed and correcting the basic value of the MBT thermal efficiency with the correction value.

2-2-2. Coefficient calculation unit 322
The coefficient calculation unit 322 calculates the coefficient (parameter) of the approximate function that approximates the relationship between the ignition timing for calculating the N coefficients and the output information of the N elements. In the present embodiment, the coefficient calculation unit 322 sets the maximum torque ignition timing θMBT and the MBT thermal efficiency ηMBT corresponding to the maximum torque ignition timing θMBT to the extreme values of the approximate function as a quadratic function, and ignites on the retard side. The coefficient is calculated based on the timing θsp and the retard side thermal efficiency ηspl corresponding to the retard side ignition timing θspl.

The approximation function is set to the following equation and the quadratic function shown in FIG.
η = ηMBT × {1-Ka × (θig－θMBT) ² } ・ ・ ・ (1)
Here, Ka is one of the coefficients of the quadratic function, and is hereinafter referred to as the sensitivity coefficient Ka. The maximum torque ignition timing θMBT and the MBT thermal efficiency ηMBT are also coefficients of the quadratic function. That is, the maximum torque ignition timing θMBT, the MBT thermal efficiency ηMBT, and the sensitivity coefficient Ka are the coefficients of the quadratic function of the equation (1).

Then, the coefficient calculation unit 322 calculates the sensitivity coefficient Ka using the following equation. Equation (2) is an equation in which the sensitivity coefficient Ka is arranged by substituting the retard angle side ignition timing θsp and the retard angle side thermal efficiency ηspl into the equation (1).
Ka = (1-ηsp / ηMBT) / (θspl-θMBT) ² ... (2)

2-2-3. Control quantity calculation unit 323
The control amount calculation unit 323 calculates the output information corresponding to the ignition timing set for control using the coefficient, and calculates the control amount of the internal combustion engine based on the calculated output information. The control amount calculation unit 323 determines one or more of the target intake air amount of the internal combustion engine (target filling efficiency Ect in this example), the target ignition timing θight of the internal combustion engine, and the torque of the internal combustion engine as the control amount of the internal combustion engine. Is calculated.

According to this configuration, once the coefficient of the approximation function that approximates the relationship between the ignition timing and the output information under the current operating conditions is calculated, various control operations can be performed using the coefficient with a small processing load. The output information corresponding to the ignition timing of the above can be calculated, and the control amount of various internal combustion engines can be calculated based on the output information. Therefore, it is not necessary to repeatedly perform an operation using a neural network having output characteristics, which has a large processing load, and the operation processing load can be reduced.

2-2-3-1. Calculation of target ignition timing θight using a coefficient The control amount calculation unit 323 calculates the reference torque Trqb corresponding to the reference ignition timing θigb set for control using the coefficient, and torques to the reference torque Trqb. The target torque Trqt is calculated by adding the change amount ΔTrq of the above, and the target ignition timing θight corresponding to the target torque Trqt is calculated using a coefficient.

According to this configuration, the reference torque Trqb corresponding to the reference ignition timing θigb can be calculated by a small calculation of the processing load using the coefficient. Further, the target ignition timing θight corresponding to the target torque Trqt can be calculated by a small calculation of the processing load using the coefficient. In particular, the calculation of the target ignition timing θight corresponding to this target torque Trqt is the calculation of the inverse characteristic of the neural network of the output characteristic, so if the calculation is performed using the neural network of the output characteristic, the ignition timing changes little by little. Then, it is necessary to repeatedly perform the arithmetic processing of the neural network of the output characteristic to search for the ignition timing corresponding to the target torque Trqt, and the processing load becomes particularly large. Therefore, the effect of reducing the arithmetic processing load becomes large.

<Torque down request due to retardation timing>
The target ignition timing θight when there is a torque down request due to the retard angle of the ignition timing will be described. If the high response target torque Trqtf matches the low response target torque Trqts and there is no torque down request due to the retard angle of the ignition timing, the control amount calculation unit 323 sets the reference ignition timing θigb to the target ignition timing θight. Set to. When the high response target torque Trqtf is lower than the low response target torque Trqts and there is a torque down request due to the retard angle of the ignition timing, the control amount calculation unit 323 calculates the target ignition timing by the calculation using the coefficient described below. Set θight.

As shown in the following equation, the control amount calculation unit 323 calculates the reference thermal efficiency ηb corresponding to the reference ignition timing θigb using the coefficients Ka, θMBT, and ηMBT, and calculates the reference thermal efficiency ηb and the actual intake air amount Qcr. , And the reference torque Trqb is calculated based on the target air-fuel ratio AFt.
ηb = ηMBT × {1-Ka × (θigb-θMBT) ² }
Trqb = Qcr / AFt × HT × ηb ・ ・ ・ (3)
Here, HT is the calorific value of the fuel per unit mass, and is preset according to the type of fuel.

Further, as shown in the following equation, the control amount calculation unit 323 determines the torque TrqMBT when the ignition timing is the maximum torque ignition timing θMBT based on the MBT thermal efficiency ηMBT, the actual intake air amount Qcr, and the target air-fuel ratio AFt. (MBT torque TrqMBT) is calculated.
TrqMBT = Qcr / AFt × HT × ηMBT ・ ・ ・ (4)

The control amount calculation unit 323 calculates the torque down amount ΔTdwn by subtracting the high response target torque Trqtf from the low response target torque Trqts as shown in the following equation. Then, the control amount calculation unit 323 calculates the target ignition timing θight based on the sensitivity coefficient Ka, the reference torque Trqb, the MBT torque TrqMBT, and the torque down amount ΔTdwn.

The control amount calculation unit 323 uses a reference ignition timing setting function in which the relationship between the operating state for calculating the reference ignition timing of the internal combustion engine and the reference ignition timing θigb is preset, and is used for calculating the current reference ignition timing. The reference ignition timing θigb corresponding to the operating condition is calculated. The operating state for calculating the reference ignition timing is set to the operating state such as the rotation speed Ne, the filling efficiency Ec, the EGR rate Regr, and the air-fuel ratio AF. In the present embodiment, the operating state for calculating the reference ignition timing is set to be the same as the operating state for calculating the output.

The reference ignition timing setting function may be configured by a neural network or may be configured by a plurality of map data. When composed of a plurality of map data, the control amount calculation unit 323 uses the map data in which the relationship between the rotation speed and the filling efficiency and the basic value of the reference ignition timing is preset, and the current rotation speed and the filling are used. The basic value of the reference ignition timing corresponding to the efficiency is calculated, and the map data in which the relationship between the EGR rate and the correction value is preset and the map data in which the relationship between the air-fuel ratio and the correction value are preset are used. The correction value corresponding to the EGR rate and the air-fuel ratio of is calculated, the basic value of the reference ignition timing is corrected by the correction value, and the reference ignition timing θigb is calculated.

<Target ignition timing when idling>
Next, the setting of the target ignition timing θight during idling will be described. As shown in the following equation, the control amount calculation unit 323 calculates the reference thermal efficiency ηb corresponding to the reference ignition timing θigb_idl during idling using the coefficients Ka, θMBT, and ηMBT, and calculates the reference thermal efficiency ηb and the actual suction. The reference torque Trqb is calculated based on the air amount Qcr and the target air-fuel ratio AFt.
ηb = ηMBT × {1-Ka × (θigb_idl-θMBT) ² }
Trqb = Qcr / AFt × HT × ηb ・ ・ ・ (6)

Further, as shown in the following equation, the control amount calculation unit 323 calculates the torque TrqMBT (referred to as MBT torque TrqMBT) when the ignition timing is the maximum torque ignition timing θMBT using the MBT thermal efficiency ηMBT.
TrqMBT = Qcr / AFt × HT × ηMBT ・ ・ ・ (7)

The control amount calculation unit 323 calculates the additional torque ΔTidl from the reference torque Trqb at the time of idling based on the operating state of each load, the water temperature, and the like. The load includes the operating state of the air conditioner, the operating state of the generator, and the like.

Then, the control amount calculation unit 323 calculates the target ignition timing θight at the time of idling based on the sensitivity coefficient Ka, the reference torque Trqb, the MBT torque TrqMBT, and the added torque ΔTidl, as shown in the following equation.

The control amount calculation unit 323 uses the current reference ignition timing setting function during idle ring in which the relationship between the operating state for idle ring of the internal combustion engine and the reference ignition timing θigb_idl during idle ring is preset. The reference ignition timing θigb_idl at the time of idle ring corresponding to the operating state for idle ring is calculated. The operating state for the idle ring is set to an operating state such as a rotation speed Ne and a filling efficiency Ec. The reference ignition timing setting function at the time of idle ring may be configured by map data or by a neural network.

2-2-3-2. Calculation of Target Filling Efficiency Ect Using Coefficient The control amount calculation unit 323 calculates the reference thermal efficiency ηb corresponding to the reference ignition timing θigb set for control using the coefficient, and calculates the required torque. Based on the value divided by the reference thermal efficiency ηb, the target intake air amount of the internal combustion engine (in this example, the target filling efficiency Ect) is calculated.

According to this configuration, the reference thermal efficiency ηb corresponding to the reference ignition timing θigb can be calculated by a small calculation of the processing load using the coefficient.

<Setting of target filling efficiency Ect corresponding to low response target torque Trqts>
The control amount calculation unit 323 calculates the reference thermal efficiency ηb corresponding to the reference ignition timing θigb by using the coefficients Ka, θMBT, and ηMBT as shown in the following equation.
ηb = ηMBT × {1-Ka × (θigb-θMBT) ² } ・ ・ ・ (9)

Then, the control amount calculation unit 323 calculates the target filling efficiency Ect based on the low response target torque Trqts, the reference thermal efficiency ηb, and the target air-fuel ratio AFt, as shown in the following equation.
Ect = Trqts / ηb / HT × AFt / (ρ × Vcyl) × 100
... (10)
Here, ρ is the air density and Vcyl is the volume of the combustion chamber.

<Setting the target filling efficiency Ect when idling>
As shown in the following equation, the control amount calculation unit 323 calculates the reference thermal efficiency ηb corresponding to the reference ignition timing θigb_idl at the time of idling by using the coefficients Ka, θMBT, and ηMBT.
ηb = ηMBT × {1-Ka × (θigb_idl-θMBT) ² }
... (11)

Then, the control amount calculation unit 323 calculates the target filling efficiency Ect based on the required torque Trq_idl at idling, the drive system torque Trqm, the reference thermal efficiency ηb, and the target air-fuel ratio AFt, as shown in the following equation.
Ect = (Trq_idl + Trqm) / ηb / HT × AFt
/ (Ρ × Vcyl) × 100 ・ ・ ・ (12)

The control amount calculation unit 323 calculates the required torque Trq_idl at the time of idling based on the operating state such as the rotation speed Ne. Further, the control amount calculation unit 323 calculates the drive system torque Trqm based on the water temperature, the operating state of the generator, and the like.

<Setting of target filling efficiency Ect when the ignition timing is retarded>
The retard angle of the ignition timing may be set due to the temperature rise of the catalyst or the like, and the setting of the target filling efficiency Ect in that case will be described. The control amount calculation unit 323 calculates the reference torque Trqb corresponding to the reference ignition timing θigb set for control using a coefficient, and corresponds to the target ignition timing θight retarded from the reference ignition timing θigb. The thermal efficiency ηrtd after retardation is calculated using a coefficient, and the reference torque Trqb is divided by the thermal efficiency ηrtd after retardation, and the target intake air amount of the internal combustion engine (in this example, the target filling efficiency). Ect) is calculated.

According to this configuration, the reference torque Trqb corresponding to the reference ignition timing θigb and the thermal efficiency ηrtd after the retard angle corresponding to the target ignition timing θight after the retard angle are calculated by a small calculation of the processing load using the coefficient. can do.

As shown in the following equation, the control amount calculation unit 323 calculates the reference thermal efficiency ηb corresponding to the reference ignition timing θigb using the coefficients Ka, θMBT, and ηMBT, and calculates the reference thermal efficiency ηb and the actual intake air amount Qcr. , And the reference torque Trqb is calculated based on the target air-fuel ratio AFt.
ηb = ηMBT × {1-Ka × (θigb-θMBT) ² }
Trqb = Qcr / AFt × HT × ηb ・ ・ ・ (13)

As shown in the following equation, the control amount calculation unit 323 calculates the thermal efficiency ηrtd after the retard angle corresponding to the target ignition timing θight after the retard angle using the coefficients Ka, θMBT, and ηMBT.
θight = θigb + Δθrtd
ηrtd = ηMBT × {1-Ka × (θight-θMBT) ² } ・ ・ ・ (14)
Here, Δθrtd is the retard angle amount of the ignition timing, and is set based on the water temperature and the like.

Then, the control amount calculation unit 323 calculates the target filling efficiency Ect based on the reference torque Trqb, the thermal efficiency ηrtd after the retard angle, and the target air-fuel ratio AFt, as shown in the following equation.
Ect = Trqb / ηrtd / HT × AFt / (ρ × Vcyl) × 100
... (15)

The target filling efficiency Ect at the time of retard may be set at the time of idling. In this case, the reference ignition timing θigb of the equation (13) is set to the reference ignition timing θigb_idl at the time of idling.

2-2-3-3. Calculation of the operable range of torque by ignition timing using coefficient The control amount calculation unit 323 calculates the reference torque Trqb corresponding to the reference ignition timing θigb set for control by using the coefficient, and is used for control. The torque Trqrtd of the upper limit of the retard angle side corresponding to the ignition timing of the upper limit of the retard angle side set to (in this example, the above-mentioned retard angle upper limit ignition timing θrtd) is calculated, and the torque Trqrtd of the retard angle side is calculated from the reference torque Trqb. Up to the upper limit torque Trqrtd is calculated as the operable range of the torque due to the retard angle of the ignition timing.

According to this configuration, the reference torque Trqb corresponding to the reference ignition timing θigb and the upper limit torque Trqrtd on the retard side corresponding to the retard angle upper limit ignition timing θrtd are calculated by a small calculation of the processing load using the coefficient. Then, the operable range of the torque due to the retard angle of the ignition timing can be calculated.

As shown in the following equation, the control amount calculation unit 323 calculates the reference thermal efficiency ηb corresponding to the reference ignition timing θigb using the coefficients Ka, θMBT, and ηMBT, and calculates the reference thermal efficiency ηb and the actual intake air amount Qcr. , And the reference torque Trqb is calculated based on the target air-fuel ratio AFt.
ηb = ηMBT × {1-Ka × (θigb-θMBT) ² }
Trqb = Qcr / AFt × HT × ηb ・ ・ ・ (16)

Further, as shown in the following equation, the control amount calculation unit 323 calculates the thermal efficiency ηrtd of the upper limit of the retard angle side corresponding to the retard angle upper limit ignition timing θrtd by using the coefficients Ka, θMBT, and ηMBT, and the retard angle side. The upper limit torque Trqrtd on the retard side is calculated based on the upper limit thermal efficiency ηrtd, the actual intake air amount Qcr, and the target air-fuel ratio AFt.
ηrtd = ηMBT × {1-Ka × (θrtd-θMBT) ² }
Trqrtd = Qcr / AFt × HT × ηrtd ・ ・ ・ (17)

The control amount calculation unit 323 uses the operable range from the calculated reference torque Trqb to the upper limit torque Trqrtd on the retard side for setting the target torque or transmitting it to an external control device.

2-2-3-4. Calculation of actual torque Trqr using a coefficient The control amount calculation unit 323 calculates the actual torque Trqr corresponding to the final ignition timing θsa using the coefficient. If the ignition timing retard correction is not performed by detecting knock, the target ignition timing θight may be used instead of the final ignition timing θsa.

According to this configuration, the actual torque Trqr corresponding to the final ignition timing θsa can be calculated by a small calculation of the processing load using the coefficient.

As shown in the following equation, the control amount calculation unit 323 calculates the actual thermal efficiency ηr corresponding to the final ignition timing θsa using the coefficients Ka, θMBT, and ηMBT, and the actual thermal efficiency ηr, the actual intake air amount Qcr, and the target. The actual torque Trqr is calculated based on the air-fuel ratio AFt.
ηr = ηMBT × {1-Ka × (θsa-θMBT) ² }
Trqr = Qcr / AFt × HT × ηr ・ ・ ・ (18)

2. 2. Embodiment 2
Next, the control device 30 according to the second embodiment will be described. The description of the same components as those in the first embodiment will be omitted. The basic configuration of the control device 30 according to the present embodiment is the same as that of the first embodiment, but the method of setting the ignition timing for calculating the N coefficients and the method of calculating the coefficients are the same as those of the first embodiment. different. FIG. 7 shows a block diagram of the torque interface unit 32 according to the present embodiment.

The ignition-corresponding output calculation unit 321 sets ignition timings θspl1, θspl2, and θspl3 for calculating three different coefficients as ignition timings for calculating N coefficients.

For example, as shown in the following equation, the ignition-corresponding output calculation unit 321 sets the ignition timing θspl1 for the first coefficient calculation to the reference ignition timing θigb, and sets the ignition timing θspl2 for the second coefficient calculation. The ignition timing θspl3 for calculating the third coefficient is set to the retard side by the first angle (for example, 5 degrees) from the reference ignition timing θigb, and the ignition timing θspl3 for calculating the third coefficient is set to the second angle (for example, 10) from the reference ignition timing θigb. Set only the degree) to the retard side.
θspl1 = θigb, θspl2 = θigb + 5, θspl3 = θigb +10
... (19)

The ignition timing θspl3 for calculating the third coefficient is set to the retard upper limit ignition timing θrtd, and the ignition timing θspl2 for calculating the second coefficient is the ignition timing θspl1 for calculating the first coefficient and the third coefficient. It may be set to an intermediate angle with the ignition timing θspl3 for calculation.

Then, the ignition-corresponding output calculation unit 321 uses a neural network of output characteristics, and has three ignition timings θspl1, θspl2, and θspl3 for calculating the current output and three coefficients. The thermal efficiencies ηspl1, ηspl2, and ηspl3 are calculated.

Specifically, the ignition-corresponding output calculation unit 321 is in operation for calculating the current output (in this example, the actual rotation speed Ne, the actual filling efficiency Ecr, the actual EGR rate Regrr, and the target air-fuel ratio AFt or the actual air-fuel ratio. AFr) and the ignition timing θspl1 for calculating the first coefficient are input to the neural network of the output characteristics, the neural network is calculated, and the thermal efficiency η output from the neural network is calculated as the first thermal efficiency ηspl1. .. Further, the ignition-compatible output calculation unit 321 inputs the current operating state for output calculation and the ignition timing θspl2 for the second coefficient calculation to the neural network of the output characteristics, performs the calculation process of the neural network, and performs the neural network. The thermal efficiency η output from is calculated as the second thermal efficiency ηspl2. The ignition-compatible output calculation unit 321 inputs the current operating state for output calculation and the ignition timing θspl3 for calculating the third coefficient into the neural network of the output characteristics, performs the calculation process of the neural network, and outputs it from the neural network. The calculated thermal efficiency η is calculated as a third thermal efficiency ηspl3.

Then, the coefficient calculation unit 322 calculates the coefficient of the quadratic function as an approximate function that approximates the relationship between the ignition timings θspl1, θspl2, and θspl3 for calculating the three coefficients and the three thermal efficiencies ηspl1, ηspl2, and ηspl3. calculate.

The approximation function is set to the following equation and the quadratic function shown in FIG.
η = A × θig ² + B × θig + C ・ ・ ・ (20)
Here, A is a coefficient of a second-order term, B is a coefficient of a first-order term, and C is a coefficient of a zero-order term.

The coefficient calculation unit 322 calculates three coefficients A, B, and C using the following equation.

The relationship between the three coefficients A, B, and C of the second embodiment and the three coefficients θMBT, ηMBT, and Ka of the first embodiment is as follows.

Therefore, by substituting the equation (22) into the equations (3) to (18) of the first embodiment, the control amount is calculated using the three coefficients A, B, and C according to the present embodiment. be able to. Alternatively, the equations (3) to (18) may be transformed in advance according to the three coefficients A, B, and C.

Note that N may be set to a number of 4 or more. In this case, the relationship between the ignition timing θig and the torque Trq is calculated at 4 points or more, and 3 by a regression analysis method such as the least squares method. Coefficients A, B, C may be calculated.

Although the present application describes various exemplary embodiments and examples, the various features, embodiments, and functions described in one or more embodiments are applications of a particular embodiment. It is not limited to, but can be applied to embodiments alone or in various combinations. Therefore, innumerable variations not exemplified are envisioned within the scope of the techniques disclosed herein. For example, it is assumed that at least one component is modified, added or omitted, and further, at least one component is extracted and combined with the components of other embodiments.

1 internal combustion engine, 30 internal combustion engine control device, 321 ignition compatible output calculation unit, 322 coefficient calculation unit, 323 control amount calculation unit, 33 engine control unit

The control device for the internal combustion engine of the present application is
Set the ignition timing for calculating N coefficients (N is an integer of 2 or more),
Using a neural network in which the relationship between the operating state and ignition timing for calculating the output of the internal combustion engine and the output information which is the torque or thermal efficiency of the internal combustion engine is learned in advance, the current operating state for calculating the output and the N An ignition-compatible output calculation unit that calculates N of the output information corresponding to each of the ignition timings for calculating the coefficients, and an ignition-compatible output calculation unit.
A coefficient calculation unit that calculates the coefficient of an approximate function that approximates the relationship between the ignition timing for calculating the N coefficients and the output information of the N elements.
A control amount calculation unit that calculates the output information corresponding to the ignition timing set for control using the coefficient and calculates the control amount of the internal combustion engine based on the calculated output information.
An engine control unit that controls an internal combustion engine based on the amount of control of the internal combustion engine,
Equipped with
The control amount calculation unit calculates the reference thermal efficiency corresponding to the reference ignition timing set for control using the coefficient, and is based on the value obtained by dividing the required torque by the reference thermal efficiency. , The target intake air amount of the internal combustion engine is calculated.

Claims (11)

Set the ignition timing for calculating N coefficients (N is an integer of 2 or more),
Using a neural network in which the relationship between the operating state and ignition timing for calculating the output of the internal combustion engine and the output information which is the torque or thermal efficiency of the internal combustion engine is learned in advance, the current operating state for calculating the output and the N An ignition-compatible output calculation unit that calculates N of the output information corresponding to each of the ignition timings for calculating the coefficients, and an ignition-compatible output calculation unit.
A coefficient calculation unit that calculates the coefficient of an approximate function that approximates the relationship between the ignition timing for calculating the N coefficients and the output information of the N elements.
A control amount calculation unit that calculates the output information corresponding to the ignition timing set for control using the coefficient and calculates the control amount of the internal combustion engine based on the calculated output information.
An engine control unit that controls an internal combustion engine based on the amount of control of the internal combustion engine,
Internal combustion engine control device equipped with. The first aspect of claim 1, wherein the control amount calculation unit calculates one or more of the target intake air amount of the internal combustion engine, the target ignition timing of the internal combustion engine, and the torque of the internal combustion engine as the control amount of the internal combustion engine. Internal combustion engine control device. The control amount calculation unit calculates the reference torque corresponding to the reference ignition timing set for control using the coefficient, adds the amount of change in torque to the reference torque, and obtains the target torque. The control device for an internal combustion engine according to claim 1 or 2, wherein the target ignition timing corresponding to the target torque is calculated using the coefficient. The control amount calculation unit calculates the reference thermal efficiency corresponding to the reference ignition timing set for control using the coefficient, and divides the required torque by the reference thermal efficiency based on the value. The control device for an internal combustion engine according to any one of claims 1 to 3, wherein the target intake air amount of the internal combustion engine is calculated. The control amount calculation unit calculates the reference torque corresponding to the reference ignition timing set for control using the coefficient, and delays corresponding to the target ignition timing retarded from the reference ignition timing. The target intake air amount of the internal combustion engine is calculated based on the value obtained by calculating the thermal efficiency after the angle using the coefficient and dividing the reference torque by the thermal efficiency after the retard angle. The control device for an internal combustion engine according to any one of the following items. The control amount calculation unit calculates the reference torque corresponding to the reference ignition timing set for control using the coefficient, and corresponds to the upper limit ignition timing on the retard side set for control. Any one of claims 1 to 5 for calculating the upper limit torque on the retard side and calculating from the reference torque to the upper limit torque on the retard side as the operable range of the torque due to the retard of the ignition timing. The control device for the internal combustion engine according to the section. The control device for an internal combustion engine according to any one of claims 1 to 6, wherein the control amount calculation unit calculates a torque corresponding to the final ignition timing by using the coefficient. The ignition-compatible output calculation unit uses an MBT setting function in which the relationship between the operating state for calculating the MBT of the internal combustion engine and the maximum torque ignition timing, which is the ignition timing at which the torque of the internal combustion engine is maximized, is preset. The maximum torque ignition timing corresponding to the operating state for calculating the MBT of the above is calculated.
As the ignition timing for calculating the N coefficients, the ignition timing for calculating the two coefficients, the maximum torque ignition timing and the ignition timing on the retard side, which is the ignition timing on the retard side of the maximum torque ignition timing, are used. Set,
Using the neural network, the two output information corresponding to the current operating state for the output calculation and the ignition timing for the two coefficient calculation are calculated.
The coefficient calculation unit sets the maximum torque ignition timing and the output information corresponding to the maximum torque ignition timing to the extreme value of the approximate function as a quadratic function, and sets the retard side ignition timing and the retardation side ignition timing. The control device for an internal combustion engine according to any one of claims 1 to 7, wherein the coefficient is calculated based on the output information corresponding to the retard side ignition timing. The coefficient calculation unit sets the maximum torque ignition timing as θMBT, the output information corresponding to the maximum torque ignition timing as OIMBT, the ignition timing as θig, and the output information as OI, and the coefficient. The sensitivity coefficient, which is one of the above, is set to Ka.
The approximation function,
OI = OIMBT × {1-Ka × (θig-θMBT) ² }
Set to the quadratic function of
The retard angle side ignition timing is θspl, the output information corresponding to the retard angle side ignition timing is OISpl, and the sensitivity coefficient, which is one of the coefficients, is Ka.
Ka = (1-OIspl / OIMBT) / (θspl-θMBT) ²
Calculate the sensitivity coefficient using the calculation formula of
The control device for an internal combustion engine according to claim 8, wherein the output information corresponding to the maximum torque ignition timing and the maximum torque ignition timing are calculated as the remaining two coefficients. The ignition-corresponding output calculation unit does not use the neural network, but uses map data in which the relationship between the operation state for output calculation and the output information corresponding to the maximum torque ignition timing is preset. The control device for an internal combustion engine according to claim 8 or 9, which calculates the output information corresponding to the maximum torque ignition timing corresponding to the operating state for calculating the output. The ignition-corresponding output calculation unit sets ignition timings for calculating three different coefficients as ignition timings for calculating the N coefficients.
Using the neural network, the three output information corresponding to the current operating state for the output calculation and the ignition timing for the three coefficient calculation are calculated.
The coefficient calculation unit calculates the coefficient of the quadratic function as the approximate function that approximates the relationship between the ignition timing for calculating the three coefficients and the three output information timings. The control device for an internal combustion engine according to any one of the above items.

Images