Wavelet-based artificial neural net combustion sensing

Abstract

A method and apparatus for real-time measurement of combustion characteristics of each combustion event in each individual cylinder coupled with an ability to control the engine based upon the combustion characteristics are shown. The invention includes using selective sampling techniques and wavelet transforms to extract a critical signal feature from an ionization signal that is generated by an in-cylinder ion sensor, and then feeds that critical signal feature into an artificial neural network to determine a desired combustion characteristic of the combustion event. The desired combustion characteristic of the combustion event includes a location of peak pressure, an air/fuel ratio, or a percentage of mass-fraction burned, among others. The control system of the engine is then operable to control the engine based upon the combustion characteristic.

Description

TECHNICAL FIELD This invention pertains generally to internal combustion engine control systems, and more specifically to real-time digital signal processing for engine control and diagnostics. BACKGROUND OF THE INVENTION There is a need to be able to effectively collect and analyze data related to combustion characteristics of an internal combustion engine and to control the engine based upon that data. Current engine control systems use exhaust gas sensors, primarily oxygen sensors, to provide feedback about the overall combustion operation of the engine. Other feedback devices that have been proposed for engine control systems include in-cylinder pressure sensors and in-cylinder temperature sensors A combustion quality measurement technique utilizing flame ionization detection wherein a spark plug is also used as a sensor has been in production for some time. The ionization signal is a measure of changes in electrical conductivity of a combustion flame front that is created in...

AI Technical Summary

Benefits of technology

The present invention provides an improvement over conventional engine controls in that it provides a method and apparatus for real-time measurement of combustion characteristics of each combustion event in each individual cylinder coupled with an ability to control the engine based upon the combustion characteristics. The invention includes using selective sampling techniques and wavelet transforms to extract a critical signal feature from an ionization signal that is generated by an in-cylinder ion sensor, and then feeds that critical signal feature into an artificial neural network to determine a desired combustion characteristic of the combustion event. The desired combustion characteristic of the combustion event includes a location of peak pressure, an air / fuel ratio, or a percentage of mass-fraction burned, among others. The control system of the engine is then operable to control the engine based upon the combustion characteristic. This includes control of engine torque, and more specifically fuel injection, exhaust gas recirculation, cam timing and phasing, as well as other engine control elements. This also includes spark timing and dwell when the engine is a spark-ignition engine.

Problems solved by technology

The ability to use an ionization signal for engine control is limited by the ability to glean critical signal features from the signal.

Fluctuations in the ionization signal caused by variations within an engine, engine to engine variations, and external factors have made more complete interpretation and utilization of the ionization signal difficult.

These factors, among others, make it difficult to perform a straightforward interpretation of an ionization signal created as an output of the combustion process.

The prior art has been unable to accomplish demonstrable advanced engine control and engine diagnostic capability using information from an ionization signal.

The prior art has been unable to provide real-time signal processing that leads to information related to critical signal features such as the location of peak pressure, air to fuel ratio, or percentage of mass fraction burned, when measured over a wide range of engine operating conditions.

The prior art has not been robust to changes in conditions that affect measured engine operating conditions, including external conditions such as fuel quality and ambient temperature.

The prior art also has not been robust to changes in operating conditions such as engine operating temperature and variations in in-cylinder temperatures.

Conventional analytical methods have not provided a level of robustness necessary for mass production application of an ionization system.

A limitation of artificial intelligence is that an ANN device only knows what it was taught; it can not extrapolate beyond the range of its training, nor can it perform any better than it was taught during training.

Training of the ANN also consumes time both to collect appropriate data sets for training, and to train so that it can acquire effective coefficients and biases for internal equations.

In practice, this might be impossible or at least extremely time and resource consuming.

Dedicated ANN DSP chips can be costly, and are generally dedicated to a specific application, which limits the flexibility of the device, and makes the operating characteristics of the ANN difficult to change.

The PCA method requires acquisition of a large quantity of data (vector array of 123 input elements in one case), and takes an extended amount of time to reduce to a useful signal.

This limits the throughput of the controller, and therefore the dynamic range over which the method is used to control a system.

Images