METHOD AND ASSOCIATED DEVICE FOR SENSING THE AIR/FUEL RATIO OF AN INTERNAL COMBUSTION ENGINE

• US 20090005953A1
• Filed: 09/02/2008
• Published: 01/01/2009
• Est. Priority Date: 03/04/2005
• Status: Active Grant

• Current Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Original Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Inventors: Taglialatela -Scafati, Ferdinando, Amato, Paolo, Pirozzi, Francesco, Carpentieri, Francesco, Di Meglio, Maurizio, Cesario, Nicola, Moselli, Giovanni
• Granted Patent: US 7,962,272 B2
• Time in Patent Office: Days
• US Class Current: 701/102
• CPC Class Codes: F02D 2041/141 using a feed-forward control element
• CPC Class Codes: F02D 35/023 by determining the cylinder pressure
• CPC Class Codes: F02D 41/1401 characterised by the control or regulation method F02D41/1473, F02D41/1477 take precedence
• CPC Class Codes: F02D 41/1404 Fuzzy logic control
• CPC Class Codes: F02D 41/1405 Neural network control
• CPC Class Codes: F02D 41/1458 with determination means using an estimation
• CPC Class Codes: F02D 41/187 using a hot wire flow sensor
• CPC Class Codes: F02D 41/2454 Learning of the air-fuel ratio control
• CPC Class Codes: G06F 18/23213 with fixed number of clusters, e.g. K-means clustering

Abstract

A method of sensing the air/fuel ratio in a combustion chamber of an internal combustion engine that may be easily implemented by a respective low-cost device includes a pressure sensor and a learning machine that generates a sensing signal representing the air/fuel ratio by processing the waveform of the pressure in at least one cylinder of the engine. In practice, the learning machine extracts characteristic parameters of the waveform of the pressure and as a function of a certain number of them generates the sensing signal.

Claims

1-25. -25. (canceled)

26. A method for sensing an air/fuel ratio in a combustion chamber of an internal combustion engine, the method comprising:

sensing a pressure in at least one cylinder of the internal combustion engine;

generating a pressure signal representative of the sensed pressure;

extracting characteristic parameters of the pressure signal in a pressure cycle of the internal combustion engine;

normalizing the characteristic parameters in a set range;

choosing a subset of the characteristic parameters by a clustering algorithm;

calculating moving averages of the chosen subset of parameters on a number of pressure cycles of the internal combustion engine; and

generating a signal representative of the air/fuel ratio of the internal combustion engine as a function of the chosen subset of parameters.

27. The method of claim 26 wherein the chosen subset of parameters is chosen during a preliminary test phase, the preliminary test phase comprising:

measuring a value of each characteristic parameter in each pressure cycle;

sensing the air/fuel ratio at each pressure cycle;

calculating a correlation coefficient between a sequence of air/fuel ratios and a sequence of values of each characteristic parameter; and

choosing a number of characteristic parameters that are correlated with the air/fuel ratio.

28. The method of claim 26 wherein the number of pressure cycles on which the chosen subset of parameters are averaged is determined by at least:

establishing a set of possible values of the chosen subset of parameters for the number of pressure cycles;

for each possible value of the chosen subset of parameters for each pressure cycle, generating a respective dataset comprising sequences of moving averages of the chosen subset of parameters calculated on the number of pressure cycles;

grouping each dataset in clusters by executing a clustering algorithm thereon;

calculating a performance index for each execution of the clustering algorithm;

estimating an ideal performance index of the clustering algorithm for each dataset corresponding to a variance of moving averages belonging to a same cluster being at a lower threshold value, and to a variance of moving averages belonging to different clusters being an upper threshold value;

calculating a clustering factor as a ratio between the performance index and the ideal performance index; and

choosing a desired value of the number of pressure cycles as a value corresponding to the execution of the clustering algorithm for which the clustering factor is at an upper threshold value.

29. A device for sensing an air/fuel ratio in a combustion chamber of an internal combustion engine, the device comprising:

a pressure sensor to generate a pressure signal of a pressure in at least one cylinder of the internal combustion engine; and

a learning machine to be input with the pressure signal, to extract from the pressure signal characteristic parameters thereof, and to generate as a function of the characteristic parameters a signal representative of the air/fuel ratio.

30. The device of claim 29 wherein said learning machine comprises an offline trained learning machine.

31. The device of claim 30 wherein said offline trained learning machine is based upon a fuzzy system.

32. The device of claim 30 wherein said offline trained learning machine is based upon a neural network.

33. The device of claim 29 wherein said learning machine comprises:

an identification subsystem to choose parameters of the pressure signal by a clustering algorithm, said offline trained learning machine to generate the signal representative of the air/fuel ratio as a function of the chosen parameters.

34. The device of claim 32 wherein said offline trained learning machine further comprises:

a pre-processor coupled in cascade to said identification subsystem and to generate a signal representing a moving average of the characteristic parameters of the pressure signal;

a neural network coupled in cascade to said pre-processor and to generate an intermediate signal indicative of the air/fuel ratio; and

a post-processor to be input with the intermediate signal and to generate a signal representative of estimated air/fuel ratio of the internal combustion engine as a moving average of the intermediate signal.

35. The device of claim 31 wherein said offline trained learning machine comprises:

a first fuzzy subsystem trained to generate a first intermediate signal indicative of the air/fuel ratio when the air/fuel ratio is less than 1;

a second fuzzy subsystem trained to generate a second intermediate signal indicative of the air/fuel ratio when the air/fuel ratio is greater than 1;

a third fuzzy subsystem trained to generate a third intermediate signal indicative of the air/fuel ratio when the air/fuel ratio is about 1;

three post-processors coupled in cascade to a respective fuzzy subsystem and to generate a first signal, a second signal, and a third signal, respectively, of an estimated air/fuel ratio of the internal combustion engine as a moving average of the respective intermediate signal inputted therein; and

a multiplexer to be input with the first signal, the second signal, and the third signal of the estimated air/fuel ratio, and to output the first signal of the estimated air/fuel ratio or the second signal of the estimated air/fuel ratio as a function of the third intermediate signal.

36. The device of claim 35 wherein the third intermediate signal comprises a power signal.

37. The device of claim 35 wherein each of said fuzzy subsystems includes three inputs corresponding to a pressure value for a crank angle of 40 degrees, a pressure value for a crank angle of 50 degrees, and an upper threshold pressure value.

38. A device for sensing an air/fuel ratio in a combustion chamber of an internal combustion engine, the device comprising:

a pressure sensor to generate a pressure signal of a pressure in at least one cylinder of the internal combustion engine; and

offline trained learning machine being based upon a fuzzy system and comprising an identification subsystem choosing parameters of the pressure signal by a clustering algorithm;

said offline trained learning machine to be input with the pressure signal, to extract from the pressure signal characteristic parameters thereof, and to generate the signal representative of the air/fuel ratio as a function of the chosen parameters.

39. The device of claim 38 wherein said offline trained learning machine further comprises:

a pre-processor coupled in cascade to said identification subsystem and to generate a signal representing a moving average of the characteristic parameters of the pressure signal;

a neural network coupled in cascade to said pre-processor and to generate an intermediate signal indicative of the air/fuel ratio; and

a post-processor to be input with the intermediate signal and to generate a signal representative of estimated air/fuel ratio of the internal combustion engine as a moving average of the intermediate signal.

40. The device of claim 38 wherein said offline trained learning machine comprises:

a first fuzzy subsystem trained to generate a first intermediate signal indicative of the air/fuel ratio when the ratio air/fuel is less than 1;

a second fuzzy subsystem trained to generate a second intermediate signal indicative of the air/fuel ratio when the air/fuel ratio is greater than 1;

a third fuzzy subsystem trained to generate a third intermediate signal indicative of the air/fuel ratio when the air/fuel ratio is about 1;

three post-processors coupled in cascade to a respective fuzzy subsystem and to generate a first signal, a second signal, and a third signal, respectively, of an estimated air/fuel ratio of the internal combustion engine as a moving average of the respective intermediate signal inputted therein; and

a multiplexer to be input with the first signal, the second signal, and the third signal of the estimated air/fuel ratio, and to output the first signal of the estimated air/fuel ratio or the second signal of the estimated air/fuel ratio as a function of the third intermediate signal.

41. A probabilistic neural network comprising:

a hidden layer of neurons, each neuron computing respective membership matrix elements for an input vector of the neural network based upon a respective radial basis function defined by a respective spread factor and according to a distance of the input vector from a respective constant vector;

said hidden layer comprising at least two neurons having different spread factors.

42. The probabilistic neural network of claim 41 wherein the radial basis function comprises a Gaussian function.