Engine control system and method

1. A system comprising:

an engine including a piston mounted for reciprocating within a cylinder including a combustion chamber, the combustion chamber including an intake valve associated with an intake passage, an exhaust valve associated with an exhaust passage for receiving an exhaust gas from the combustion chamber, and a fuel injector positioned to inject fuel into air received into the combustion chamber through the intake passage for combustion in the combustion chamber;

one or more sensors for sensing one or more parameters of the engine;

one or more processors in communication with the one or more sensors;

one or more computer-readable media storing instructions which, when executed by the one or more processors, program the one or more processors to perform operations comprising:

receiving sensor data from the one or more sensors;

based on the sensor data, accessing a data structure for determining a homogeneity index indicative of a homogeneity of an air-fuel mixture within the combustion chamber at least during a transient engine operating condition, the data structure including one or more homogeneity index distributions having been determined in advance using a computer simulation of the engine;

determining an estimated amount of NOx in the exhaust gas due at least to the transient engine operating condition based at least on inputting the homogeneity index into a machine learning model trained to output an indication of the estimated amount of NOx in the exhaust gas based at least on the homogeneity index; and

controlling an engine component based at least partially on the estimated amount of NOx in the exhaust gas.

2. The system as recited in claim 1 , wherein:

the sensor data includes at least one of: an indicated in-cylinder pressure, an indicated intake pressure, an indicated exhaust pressure, an indicated valve timing, an indicated fuel injection timing, or an indicated fuel injection quantity; and

the homogeneity index is determined by accessing the data structure based at least partially on the at least one of the indicated in-cylinder pressure, the indicated intake pressure, the indicated exhaust pressure, the indicated valve timing, the indicated fuel injection timing, or the indicated fuel injection quantity.

3. The system as recited in claim 1 , further comprising an NOx control device in communication with the exhaust passage, wherein following combustion, exhaust gas is expelled through the exhaust passage to the NOx control device, the operations further comprising:

comparing a total estimated amount of NOx since a last regeneration event for the NOx control device with a regeneration threshold for total NOx; and

based on the total estimated amount of NOx exceeding the regeneration threshold, controlling the engine component by sending the instruction to perform regeneration of the NOx control device.

4. The system as recited in claim 3 , wherein the trained machine learning model is an artificial neural network trained for the engine and the NOx control device using an external computing device.

5. The system as recited in claim 3 , wherein:

the NOx control device is a lean NOx trap; and

the regeneration includes operating the engine fora specified period of time with a rich air-fuel ratio.

6. The system as recited in claim 3 , wherein:

the NOx control device is a selective catalytic reduction device; and

the regeneration includes exposing the NOx control device to at least one of: urea, anhydrous ammonia, or aqueous ammonia.

7. The system as recited in claim 1 , wherein the fuel injector is positioned to one of:

inject fuel directly into the combustion chamber; or

inject fuel into the intake passage.

8. A method comprising:

receiving, by one or more processors, sensor data from one or more sensors indicating one or more engine parameters of an engine including a combustion chamber;

based on the sensor data, accessing a data structure for determining a homogeneity index indicative of a homogeneity of an air-fuel mixture within the combustion chamber at least during a transient engine operating condition, the data structure including one or more homogeneity index distributions having been determined in advance using a computer simulation of the ermine;

determining an estimated amount of NOx in an exhaust gas exiting the combustion chamber due at least to the transient engine operating condition based at least on inputting the homogeneity index into a machine learning model trained to output an indication of the estimated amount of NOx in the exhaust gas based at least on the homogeneity index; and

based at least in part on the estimated amount of NOx, sending an instruction to control an engine component.

9. The method as recited in claim 8 , wherein the sensor data includes at least one of: an indicated in-cylinder pressure, an indicated intake pressure, an indicated exhaust pressure, an indicated valve timing, an indicated fuel injection timing, or an indicated fuel injection quantity,

the method further comprising:

generating the data structure having one or more homogeneity index distributions determined in advance using a computer simulation of the engine, and

wherein the determining the homogeneity index includes accessing the data structure based at least partially on the at least one of: the indicated in-cylinder pressure, the indicated intake pressure, the indicated exhaust pressure, the indicated valve timing, the indicated fuel injection timing, or the indicated fuel injection quantity.

10. The method as recited in claim 8 , wherein the engine includes an NOx control device in communication with an exhaust passage from the combustion chamber, wherein following combustion, exhaust gas is expelled through the exhaust passage to the NOx control device, the method further comprising:

comparing a total estimated amount of NOx since a last regeneration event for the NOx control device with a regeneration threshold for total NOx; and

based on the total estimated amount of NOx exceeding the regeneration threshold, controlling the engine component by sending the instruction to perform regeneration of the NOx control device.

11. The method as recited in claim 10 , wherein the trained machine learning model is an artificial neural network trained for the engine and the NOx control device using an external computing device.

12. The method as recited in claim 10 , wherein the NOx control device is a lean NOx trap, and sending the instruction to control the engine component comprises sending one or more control signals to operate the engine fora specified period of time with a rich air-fuel ratio for regenerating the lean NOx trap.

13. The method as recited in claim 10 , wherein the NOx control device is a selective catalytic reduction device, and sending the instruction to control the engine component comprises sending one or more control signals for regenerating the selective catalytic reduction device by causing the selective catalytic reduction device to be exposed to at least one of: urea, anhydrous ammonia, or aqueous ammonia.

14. The method as recited in claim 8 , wherein the engine includes a fuel injector positioned to one of:

inject fuel directly into the combustion chamber; or

inject fuel into an intake passage.

15. A system comprising:

an engine including a combustion chamber and one or more sensors;

one or more processors in communication with the one or more sensors; and

one or more computer-readable media storing instructions which, when executed by the one or more processors, program the one or more processors to perform operations comprising:

receiving sensor data from the one or more sensors indicating one or more engine parameters of the engine;

based on the sensor data, accessing a data structure for determining a homogeneity index for an air-fuel mixture within the combustion chamber at least during a transient engine operating condition, the data structure including one or more homogeneity index distributions having been determined in advance using a computer simulation of the engine;

determining an estimated amount of NOx in an exhaust gas exiting the combustion chamber due at least to the transient engine operating condition based at least on inputting the homogeneity index into a machine learning model trained to output an indication of the estimated amount of NOx in the exhaust gas based at least on the homogeneity index; and

based at least in part on the estimated amount of NOx, sending an instruction to control an engine component.

16. The system as recited in claim 15 , wherein the sensor data includes at least one of: an indicated in-cylinder pressure, an indicated intake pressure, an indicated exhaust pressure, an indicated valve timing, an indicated fuel injection timing, or an indicated fuel injection quantity,

the operations further comprising:

generating the data structure having one or more homogeneity index distributions determined in advance using a computer simulation of the engine; and

wherein the determining the homogeneity index includes accessing the data structure based at least partially on the at least one of: the indicated in-cylinder pressure, the indicated intake pressure, the indicated exhaust pressure, the indicated valve timing, the indicated fuel injection timing, or the indicated fuel injection quantity.

17. The system as recited in claim 15 , further comprising an NOx control device positioned to receive the exhaust gas exiting the combustion chamber, the operations further comprising:

comparing a total estimated amount of NOx since a last regeneration event for the NOx control device with a regeneration threshold for total NOx; and

based on the total estimated amount of NOx exceeding the regeneration threshold, controlling the engine component by sending the instruction to perform regeneration of the NOx control device.

18. The system as recited in claim 17 , wherein the trained machine learning model is an artificial neural network trained for the engine and the NOx control device using an external computing device.

19. The system as recited in claim 17 , wherein the NOx control device is a lean NOx trap, and sending the instruction to control the engine component comprises sending one or more control signals to operate the engine fora specified period of time with a rich air-fuel ratio for regenerating the lean NOx trap.

20. The system as recited in claim 17 , wherein the NOx control device is a selective catalytic reduction device, and sending the instruction to control the engine component comprises sending one or more control signals for regenerating the selective catalytic reduction device by causing the selective catalytic reduction device to be exposed to at least one of: urea, anhydrous ammonia, or aqueous ammonia.