Engine control system using a cascaded neural network

• US 20030217021A1
• Filed: 05/15/2002
• Published: 11/20/2003
• Est. Priority Date: 05/15/2002
• Status: Active Grant

• Current Assignee: Caterpillar Incorporated
• Original Assignee: Caterpillar Incorporated
• Inventors: Jacobson, Evan Earl
• Granted Patent: US 7,035,834 B2
• Time in Patent Office: Days
• US Class Current: 706/16
• CPC Class Codes: F02D 2041/1419 the control loops being cascaded, i.e. being placed in series or nested
• CPC Class Codes: F02D 2041/1433 using a model or simulation of the system
• CPC Class Codes: F02D 2250/12 Timing of calculation, i.e. specific timing aspects when calculation or updating of engine parameter is performed
• CPC Class Codes: F02D 41/0007 for control of turbo-charged or super-charged engines control of the pumps per se F02B37/12
• CPC Class Codes: F02D 41/1405 Neural network control
• CPC Class Codes: G05B 13/027 using neural networks only

Abstract

A method, system and machine-readable storage medium for monitoring an engine using a cascaded neural network that includes a plurality of neural networks is disclosed. In operation, the method, system and machine-readable storage medium store data corresponding to the cascaded neural network. Signals generated by a plurality of engine sensors are then inputted into the cascaded neural network. Next, a second neural network is updated at a first rate, with an output of a first neural network, wherein the output is based on the inputted signals. In response, the second neural network outputs at a second rate, at least one engine control signal, wherein the second rate is faster than the first rate.

Claims

1. A method for monitoring an engine using a cascaded neural network that includes a plurality of neural networks, the method comprising:

storing in a memory, data corresponding to the cascaded neural network;

inputting signals generated by a plurality of engine sensors into the cascaded neural network;

updating at a first rate, a second neural network with an output of a first neural network, wherein said output is based on the inputted signals; and

outputting at a second rate, at least one engine control signal from the second neural network;

wherein the second rate is faster than the first rate.

2. The method of claim 1, wherein the data corresponding to the cascaded neural network comprises a plurality of interconnected nodes and a plurality of weights corresponding to the nodes.

3. The method of claim 1, further comprising adjusting at least one engine response parameter based on the output from the second neural network.

4. The method of claim 1, wherein inputting signals further includes inputting signals generated by a plurality of engine sensors into one of a plurality of neural networks in the cascaded neural network.

5. The method of claim 1, wherein inputting signals further includes:

inputting signals generated by at least one engine sensor into a first of a plurality of neural networks in the cascaded neural network; and

inputting signals generated by at least one engine sensor into at least the second neural network.

6. The method of claim 1, wherein outputting further comprises:

outputting at least one engine control signal from the first of a plurality of neural networks in the cascaded neural network; and

outputting at least one engine control signal from the second neural network.

7. The method of claim 1, wherein the output of the first neural network is indicative of exhaust temperature and the engine control signal is one of fuel injection timing or fuel injection quantity.

8. The method of claim 1, wherein the output of the first neural network is indicative of a nitrogen oxide (NO_x) emission and the engine control signal is fuel injection timing.

9. A method for controlling an engine having an engine control monitor (ECM), the method comprising:

storing in a memory, data corresponding to a cascaded neural network including a plurality of neural networks;

inputting signals generated by a plurality of engine sensors into a first neural network;

transmitting from the first neural network, an output which characterizes performance of the engine as a function of predetermined parameters;

updating a cerebellar model articulation controller (CMAC) from an output of the first neural network which characterizes performance of the engine as a function of predetermined parameters; and

outputting a signal from the ECM based on an output from the CMAC.

10. The method of claim 9, wherein the predetermined parameters are selected from the group consisting of:

No_x production, torque, exhaust manifold temperature, and turbo speed.

11. The method of claim 9, further comprising adjusting at least one of the predetermined parameters based on the signal output from the ECM.

12. The method of claim 9, wherein outputting further comprises outputting at least one engine response parameter.

13. The method of claim 12, wherein adjusting further comprises adjusting at least one of the predetermined parameters based on the output from the first neural network.

14. The method of claim 9, wherein inputting signals further comprises:

inputting signals generated by at least one engine sensor into the first neural network; and

inputting signals generated by at least one engine sensor into the CMAC.

15. A method for monitoring an engine, comprising:

storing in a memory, data corresponding to at least two neural networks;

inputting signals generated by a plurality of engine sensors into a first neural network;

outputting at least one engine control signal from the first neural network at a first rate; and

outputting at least a second engine control signal from the second neural network at a second rate, wherein the at least second signal is dependent on a second output from the first neural network and the second rate is faster than the first rate.

16. The method of claim 15, further comprising:

adjusting at least one engine response parameter based on the outputted first value from the first neural network; and

adjusting the at least second engine response parameter based on the outputted second value from the second neural network.

17. A method for monitoring an engine, comprising:

storing in a memory, data corresponding to at least one neural network and data corresponding to a polynomial;

inputting signals generated by a plurality of engine sensors into a first neural network;

updating a polynomial with an output of the first neural network at a first rate; and

outputting at least one engine control signal from the polynomial at a second rate;

wherein the second rate is faster than the first rate.

18. A machine-readable storage medium having stored thereon machine executable instructions, the execution of said instructions adapted to implement a method for monitoring an engine using a cascaded neural network that includes a plurality of neural networks, the method comprising:

storing in a memory, data corresponding to the cascaded neural network;

inputting signals generated by a plurality of engine sensors into the cascaded neural network;

updating at a first rate, a second neural network with an output of a first neural network, wherein said output is based on the inputted signals; and

outputting at a second rate, at least one engine control signal from the second neural network;

wherein the second rate is faster than the first rate.

19. The method of claim 18, wherein the data corresponding to the cascaded neural network comprises a plurality of interconnected nodes and a plurality of weights corresponding to the nodes.

20. The machine-readable storage medium of claim 18, further comprising adjusting at least one engine response parameter based on the output from the second neural network.

21. The machine-readable storage medium of claim 18, wherein inputting further comprises inputting signals generated by a plurality of engine sensors into at least one of a plurality of neural networks in the cascaded neural network.

22. The machine-readable storage medium of claim 18, wherein inputting signals further comprises:

inputting signals generated by at least one engine sensor into a first of a plurality of neural networks in the cascaded neural network; and

inputting signals generated by at least one engine sensor into at least a second neural network.

23. The machine-readable storage medium of claim 18, wherein outputting further comprises:

outputting at least one engine control signal from a first of a plurality of neural networks in the cascaded neural network; and

outputting at least one engine control signal from a second neural network.

24. A machine-readable storage medium having stored thereon machine executable instructions, the execution of said instructions adapted to implement a method for monitoring an engine having an engine control monitor (ECM), the method comprising:

storing in a memory, data corresponding to a cascaded neural network including a plurality of neural networks;

inputting signals generated by a plurality of engine sensors into a first neural network;

transmitting from the first neural network, an output which characterizes performance of the engine as a function of predetermined parameters;

updating a cerebellar model articulation controller (CMAC) from an output of the first neural network which characterizes performance of the engine as a function of predetermined parameters; and

outputting a signal from the ECM based on an output from the CMAC.

25. The machine-readable storage medium of claim 24, wherein the predetermined parameters are selected from the group consisting of:

No_x production, torque, exhaust manifold temperature, and turbo speed.

26. The machine-readable storage medium of claim 24, further comprising adjusting at least one of the predetermined parameters based on the signal.

27. The machine-readable storage medium of claim 24, wherein outputting further comprises outputting at least one engine control signal from the first neural network.

28. The machine-readable storage medium of claim 27, wherein adjusting further comprises adjusting at least one of the predetermined parameters based on the output from the first neural network.

29. The machine-readable storage medium of claim 24, wherein inputting further comprises:

inputting signals generated by at least one engine sensor into the first neural network; and

inputting signals generated by at least one engine sensor into the CMAC.

30. A machine-readable storage medium having stored thereon machine executable instructions, the execution of said instructions adapted to implement a method for monitoring an engine, the method comprising:

storing in a memory, data corresponding to at least two neural networks;

inputting signals generated by a plurality of engine sensors into a first neural network;

outputting at least a first engine control signal from the first neural network at a first rate; and

outputting at least a second engine control signal from the second neural network at a second rate, wherein the at least second signal is dependent on a second output from the first neural network and the second rate is faster than the first rate.

31. The machine-readable storage medium of claim 30, further comprising:

adjusting at least a first engine response parameter based on the outputted first signal from the first neural network; and

adjusting at least a second engine response parameter based on the outputted second signal from the second neural network.

32. The machine-readable storage medium of claim 30, wherein the output from the first neural network occurs at a first time;

and the output from the second neural network occurs at a second time.

33. A machine-readable storage medium for monitoring an engine, the method comprising:

storing in a memory, data corresponding to at least one neural network and data corresponding to a polynomial;

inputting signals generated by a plurality of engine sensors into a first neural network;

updating a polynomial with an output of the first neural network at a first rate; and

outputting at least one engine control signal from the polynomial at a second rate;

wherein the second rate is faster than the first rate.

34. An apparatus for monitoring an engine using a cascaded neural network that includes a plurality of neural networks, the apparatus comprising:

a microprocessor that includes data corresponding to the cascaded neural network;

a module configured to receive signals generated by a plurality of engine sensors into the cascaded neural network;

a module configured to update at a first rate, a second neural network with an output of a first neural network, wherein said output is based on the inputted signals; and

a module configured to output at a second rate, at least one engine control signal from the second neural network;

wherein the second rate is faster than the first rate.

35. The apparatus of claim 34, wherein the plurality of modules comprise functionally related computer program code and data.

36. An apparatus for monitoring an engine having an engine control monitor (ECM), the apparatus comprising:

a microprocessor that includes data corresponding to a cascaded neural network including a plurality of neural networks;

a module configured to receive signals generated by a plurality of engine sensors into a first neural network;

a module configured to transmit from the first neural network, an output which characterizes performance as a function of predetermined parameters;

a module configured to update a cerebellar model articulation controller (CMAC) from an output of the first neural network which characterizes performance of the engine as a function of predetermined parameters; and

a module configured to output a signal from the ECM based on an output from the CMAC.

37. The apparatus of claim 36, wherein the predetermined parameters are selected from the group consisting of:

No_x production, torque, exhaust manifold temperature, and turbo speed.

38. The apparatus of claim 36, wherein the plurality of modules comprise functionally related computer program code and data.

39. An apparatus for monitoring an engine, the apparatus comprising:

a microprocessor that includes data corresponding to at least two neural networks;

a module configured to receive signals generated by a plurality of engine sensors into a first neural network;

a module configured to output at least a first engine control signal from the first neural network at a first rate; and

a module configured to output at least a second engine control signal from the second neural network at a second rate, wherein the at least second signal is dependent on a second output from the first neural network and the second rate is faster than the first rate.

40. The apparatus of claim 39, wherein the plurality of modules comprise functionally related computer program code and data.

41. An apparatus for monitoring an engine, the apparatus comprising:

a microprocessor that includes data corresponding to at least one neural network and data corresponding to a polynomial;

a module configured to store in a memory, data corresponding to at least one neural network and data corresponding to a polynomial;

a module configured to receive signals generated by a plurality of engine sensors into a first neural network;

a module configured to update a polynomial with an output of the first neural network at a first rate; and

a module configured to output at least one engine control signal from the polynomial at a second rate;

wherein the second rate is faster than the first rate.

42. The apparatus of claim 41, wherein the plurality of modules comprise functionally related computer program code and data.