Engine control system including feed-forward neural network controller

Patent metadata

Field	Value
Publication number	US-10196997-B2
Application number	US-201715467053-A
Country	US
Kind code	B2
Filing date	Mar 23, 2017
Priority date	Dec 22, 2016
Publication date	Feb 5, 2019
Grant date	Feb 5, 2019

Abstract

An automotive internal combustion engine combusts an air/fuel mixture to generate a drive torque. An engine control unit (ECU) determines a torque request to output a fuel request signal indicative of an amount of the fuel to inject into a cylinder of the engine. The ECU includes neural network controller installed with a neural network to generate a fuel setpoint signal based on the torque request and to define a combustion model of the engine. The neural network generates a start of injection (SOI) signal indicating a start time at which to inject the fuel based a crankshaft angle. The ECU further outputs the fuel request signal based on the fuel setpoint and the SOI. A fuel injector injects the fuel into the cylinder based on the fuel request signal to generate the drive torque indicated by the torque request.

First claim

What is claimed is:

1. An automotive internal combustion engine comprising:
a cylinder to combust a mixture of air and fuel;
a crankshaft rotatably coupled to the internal combustion engine and configured to generate a drive torque in response to combusting the mixture of air and fuel;
an engine fuel control system including a fuel injector configured to inject the fuel into the cylinder based on an injector setpoint signal generated by the internal combustion engine fuel control system; and
an engine control unit (ECU) configured to determine a torque request to output a fuel request signal that controls the injector setpoint signal thereby controlling an amount of the fuel to inject into the cylinder, the engine control unit including an active fuel demand controller having a neural network controller installed therein, the neural network controller configured to generate a fuel setpoint signal based on the torque request and a combustion model of the internal combustion engine that is generated by a neural network installed in the neural network controller, and to generate a start of injection (SOI) signal indicating a start time at which to inject the fuel based on a rotational angle of the crankshaft and the combustion model generated by the neural network,
wherein the neural network actively models each combustion cycle of the internal combustion engine based on a main fuel quantity needed to reach an indicated torque request, and a start of main injection angle needed to reach 50% of burnt fuel mass (MFB50), and actively generates the fuel setpoint signal and the start of injection (SOI) signal in response to each actively modeled combustion cycle;
wherein the engine fuel control system is located upstream from the neural network controller included in the active fuel demand controller, and wherein the active fuel demand controller outputs the fuel request signal to the engine fuel control system based on the fuel setpoint signal and the SOI signal actively generated by the neural network controller, and
wherein the fuel injector injects the fuel into the cylinder based on the fuel request signal to generate the drive torque indicated by the torque request.

2. The internal combustion engine of claim 1 , wherein the neural network includes at least one neuron that defines the combustion model.

3. The internal combustion engine of claim 2 , wherein the at least one neuron includes a plurality of neurons ranging from 10 neurons to 20 neurons.

4. The internal combustion engine of claim 3 , wherein the torque request is based on a brake torque request.

5. The internal combustion engine of claim 4 , wherein the brake torque request is based on at least one of an axle torque, frictional loss realized by the internal combustion engine, and a pumping energy loss realized by the internal combustion engine.

6. The internal combustion engine of claim 2 , further comprising an engine air control system including a throttle valve configured to adjust the air delivered into the cylinder.

7. The internal combustion engine of claim 6 , wherein the engine air control system adjusts the throttle valve to vary the air delivered into the cylinder based on the fuel request signal.

8. An engine control system configured to control an internal combustion engine, the engine control system comprising:
at least one air/fuel sensor configured to measure at least one of an amount air delivered to a cylinder of the internal combustion engine and an amount of fuel delivered to the cylinder of the internal combustion engine;
a crankshaft sensor configured to measure at least one of a rotational angle of a crankshaft and a rotational speed of the crankshaft rotatably coupled to the internal combustion engine;
an engine fuel control system including a fuel injector configured to inject the fuel into the cylinder; and
an engine control unit (ECU) in signal communication with the at least one air/fuel sensor, the crankshaft sensor, and the engine fuel control system, the ECU configured to determine a torque request to output a fuel request signal that controls an injector setpoint signal thereby controlling the amount of the fuel to inject into the cylinder, the ECU including an active fuel demand controller having a neural network controller installed therein, the neural network controller configured to generate a fuel setpoint signal based on the torque request and a neural network that defines a combustion model of the internal combustion engine, and to generate a start of injection (SOI) signal indicating a start time at which to inject the fuel based on the rotational angle of the crankshaft and the combustion model generated by the neural network,
wherein the neural network actively models each combustion cycle of the internal combustion engine based on a main fuel quantity needed to reach an indicated torque request, and a start of main injection angle needed to reach 50% of burnt fuel mass (MFB50), and actively generates the fuel setpoint signal and the start of injection (SOI) signal in response to each actively modeled combustion cycle,
wherein the engine fuel control system is located upstream from the neural network controller included in the active fuel demand controller, and wherein the active fuel demand controller outputs the fuel request signal to the engine fuel control system based on the fuel setpoint signal and the SOT signal actively generated by the neural network controller, and
wherein the fuel injector injects the fuel into the cylinder based on the fuel request signal to generate the drive torque indicated by the torque request.

9. The engine control system of claim 8 , wherein the neural network includes at least one neuron that defines the combustion model.

10. The engine control system of claim 9 , wherein the at least one neuron includes a plurality of neurons ranging from 10 neurons to 20 neurons.

11. The engine control system of claim 10 , wherein the torque request is based on a brake torque request input to the ECU.

12. The engine control system of claim 11 , wherein the brake torque request is based on at least one of an axle torque applied to the crankshaft, frictional loss realized by the internal combustion engine, and a pumping energy loss realized by the internal combustion engine.

13. The engine control system of claim 9 , further comprising an engine air control system including a throttle valve configured to adjust the air delivered into the cylinder.

14. The engine control system of claim 13 , wherein the engine air control system adjusts the throttle valve to vary the air delivered into the cylinder based on the fuel request signal.

15. A method of controlling an internal combustion engine in an automotive vehicle, the method comprising:
delivering a mixture of air and fuel into a cylinder of the internal combustion engine;
combusting the mixture of air and fuel in the cylinder;
receiving an indicated torque request;
generating, in response to the indicated torque request, a drive torque; via a crankshaft rotatably coupled to the internal combustion engine, in response to combusting the mixture of air and fuel;
actively modeling, via a neural network controller including a neural network, each combustion cycle of the internal combustion engine based on a main fuel quantity needed to reach the indicated torque request, and a start of main injection angle needed to reach 50% of burnt fuel mass (MFB50);
actively generating, via the neural network controller, a fuel setpoint signal based on a torque request input to the internal combustion engine;
defining, via the neural network controller, a

Assignees

• Gm Global Tech Operations Llc

Inventors

• Merlino Gennaro
• Conte Giuseppe
• Spessa Ezio
• Finesso Roberto

Classifications

• F02D2200/602

  Pedal position · CPC title

• F02D41/30

  Controlling fuel injection {(F02D41/182, F02D41/24 take precedence)} · CPC title

• F02D41/0002

  Controlling intake air · CPC title

• F02D2200/101

  Engine speed · CPC title

• F02D11/105

  characterised by the function converting demand to actuation, e.g. a map indicating relations between an accelerator pedal position and throttle valve opening or target engine torque · CPC title