Method to operate an internal combustion engine - engine management system using adaptive ignition and fuel quantity optimization with minimal sensor requirements for standard and bio-fuels

Summary

Method with minimal sensor requirements to control ignition tinning for internal combustion engines or injection tinning for diesel engines, as well as for optimizing fuel injection quantities.

Most ignition systems are predictive, approximating the time from ignition to combustion ("ignition time") with stored test engine parameters. By contrast, this invention uses one sensor measuring directly or indirectly the "ignition time" in real time, where the measuring can also be used for fuel quantity optimization. Therefore, fewer sensors are needed in general and in particular when fuel types are mixed (petrol, ethanol, gas or diesel and bio-fuel). Cost to produce an automotive engine management system can be reduced, while the system reliability increases. Fuel consumption is reduced, since the engine runs optimally under more combinations of external parameters (humidity, air pressure, air and engine temperature, fuel quality and mix, wear and tear, etc.).

Description

This invention concerns a method to operate an internal combustion engine as well as an internal combustion engine, which operates in accordance with the method described in this invention.

When operating an internal combustion engine, it is often necessary to control the point where combustion commences. Since there is a delay between fuel ignition

(firing spark plugs or injecting diesel) and start of (full) combustion, an ignition control

system must start the ignition process a certain amount of time ahead of a reference point, where the amount of time depends on the time needed for the fuel to ignite.

To compute optimal timing and fuel dosage parameters, traditionally, a predictive approach is used, where in a test laboratory/ test field setting; engines are subjected to different environment and load situations. Such test field results are then stored in an appropriate form, mostly electronically (ROM) or mechanically (spring). During engine operation, the stored parameters together with proxy sensors are used to predict the ignition time or the injection time. However, most of these methods try to predict or estimate parameters, which are at least to some degree, based on the environment during the laboratory trials. Wear and tear or unusual combinations of circumstances can only partially be accommodated. Also, changes in fuel mixtures (addition of bio fuels, ethanol, gas, etc) and changes in fuel quality can only be accommodated for in approximate terms.

Here, an adaptive approach is superior, where the actual "real time" primary parameters (no proxies) of an engine are used to predict optimum ignition timing (petrol engine) or injection timing (diesel fuel) and fuel dosage, particularly when the elapsed time between the firing of spark plugs or injecting of diesel to start of combustion can be measured in real time.

It's an object of the present invention to provide a method for an engine which runs closer to its peak performance. With an approach according to the present invention, the engine can run closer to its peak performance, considering its unique characteristic and actual environment. Also, being able to run each production engine close to or at optimal performance under a far wider set of environmental parameters, fuel consumption can be further optimized. In addition, the described method can accommodate varying fuel qualities and fuel compositions/ mixtures (gasoline, alcohol/ethanol, gas, bio-fuel, etc.). Depending on the sensor arrangement and algorithm applied, adaptation to fuel changes can take place already after one ignition cycle. In addition, with this invention, the laboratory testing of new engines

can be simplified. Stored values from laboratory engines are only used to recover from errors or to provide seed values (start-up, specific operating conditions, etc.), to accelerate the approximation of the optimum values. Hence, the laboratory values do not have to be as accurate and in depth, as with a traditional approach.

The basis of this invention, called "adaptive ignition" (Al), is a method to detect with relatively simple means directly or indirectly the point (or range) at which the fuel mixture ignition phase has completed and combustion commenced, as per claim 1. It does this by analysing a sensor signal, where such a signal relates to the combustion activity. The method can either detect a relatively sharp signal point or band during the transition phase from the ignition phase to the combustion phase, or select such a point based on analysis of the signal slope (sharp rise or similar) or signal amplitude (set value, proportional value, or similar). Such point is usually referenced through time or position, where such a reference point is either fixed (for example UD, Upper Deadpoint) or variable.

If now the speed of the engine (time needed for one cycle) is measured, as well as the optimum position is defined (usually UD) where combustion must commence, then the point at which the next spark plug ignition must commence can be calculated. The same applies to a diesel engine, where analogous, the optimum injection point for the fuel can be calculated. If this process starts with an ignition/ injection at a reference point (for example: UD) during engine start-up, or while recovering from a sensor error, a linear or adaptive algorithm can be applied to continuously calculate in real time the point at which ignition (or diesel injection) must commence. For the next cycle, in its simplest form, the ignition point is a certain amount of time in advance of a reference point (usually UD). The advancement time is the delay time from start of ignition until the mixture is sufficiently ignited, as measured during the last or a previous cycle(s), whereas the time required to reach the next reference point (UD) is deduced from the engine speed (time needed for one cycle). The strength of such an algorithm lies in high accuracies and generally, the avoidance of proxy sensors (manifold pressure, air temperature, etc.). Further,

the engine can run closer to optimum parameters, even in many unforeseen circumstances (wear and tear, unusual climatic environment, varying fuel mixtures and qualities, etc.) or unusual combinations of such circumstances.

Generally, the optimization process aims to complete the ignition phase and start combustion immediately after reaching the upper dead-point (UD) position of the piston, or another reference point. Al does not prescribe which particular reference point must be used, although using UD generally avoids harmful early ignition (shock on bearings), as well as a wasteful late ignition, or harmful very late ignition (overheating of valves, combustion in exhaust). However, it is also possible to advance or delay the timing, such us to optimize secondary parameters like knocking (premature ignition) or pollution (NOx, lambda sensor/ catalytic converter, etc.), i.e. a setup where the reference point is variable (result of an algorithm or similar). Thus, while UD is in most cases a natural reference point, it can be substituted with any other point. Al only requires that there is a reference point in order to optimize, but makes no demands as to whether UD or any other fixed or dynamic point is chosen for reference purposes. Similarly, the axis unit (time, angle, distance, etc.) along which the ignition point (transition from ignition to combustion) or a reference point is measured, is not vital for the functioning of Al, as long as it permits a reasonably accurate functioning of the proposed method.

The Al method allows an assessment at which piston position (or point in time) the combustion process actually started (for example: how much early or late in regards to UD measured in time or relative to distance/angle from UD), as per claim 1 or the following claims. In a stable situation (no acceleration), it is then possible to correct and reach the optimum position (usually UD) on the next cycle. Any further deviations are then instantly or adaptively/ iteratively removed. This is particularly relevant if there are changes to the environment of sudden (load change, etc.) or slow nature (air pressure, humidity, fuel quality, etc.) or changes over longer periods of time (wear and tear, etc.). Generally, even sudden changes can be compensated

to be < 1% error on the following cycle. Rapid acceleration can cause short time errors in the order of 5%, if only one reference (UD) sensor is used.

A key point of this invention is therefore the direct or indirect detection of the point or range where a mixture in an internal combustion engine transits from an initiation (ignition or injection) phase to the beginning of the combustion phase, where combustion has commenced, or combustion is about to commence. When such a point is compared to an engine reference point in time or space, then predictions can be made as to when to initiate firing of the spark plugs or injecting fuel during a following cycle or cycles. Such predictions can be made in a linear fashion, or using an iterative and/or adaptive algorithm. The aim of such an algorithm is that the point or range, where combustion begins, coincides with a fixed or variable engine reference point.

Further, the invention concerns itself with the means to detect, directly or indirectly, the point or band where a mixture in an internal combustion engine is about to commence combustion, or combustion has commenced and combines such detection with a reference point or points, in such a way as to optimise the timing of firing spark plugs or beginning of fuel injection. Such optimization can occur in the immediate next cycle, or subsequent cycles.

In a further variation, the invention may also contain means to directly or indirectly detect the intensity of the combustion process and combines such detection with means to optimize fuel quantities. Additional details regarding the method are outlined below in the claims section.

Compared to other recent inventions in this field, Al differentiates itself as follows: • DE 103 30 819 B4 deals with a method to obtain a homogenous combustion and minimising the amount of particles by measuring light emitted at very specific wavelengths.

• DE 103 07 367 A 1 deals with a method to control engines fuelled with gas, where the cylinder pressure is measured and analysed. The principal aim is to control the fuel (gas) quantity.

The present invention is specified with reference to the accompanying drawings, but is not limited to the shown embodiments:

Figure 1 shows a preferred operation of the present invention

Figure 2 shows a measurement of a load condition measured by a piezo sensor (CH1 : Sensor; CH2: ignition, primary coil)

Figure 3 shows a measurement of an idle/no load condition measured by a optical sensor (CH1 : Sensor; CH2: ignition, primary coil)

Figure 4 shows a diagram of a measurement of crank shaft speed/acceleration during one revolution (Example: early ignition with oscillations)

Figure 5 depicts an application of an optical sensor for a 2 stroke example where the glass surface remains clean despite oil deposits on cylinder head

Figure 1 shows the preferred operation of the described method: At the start point (1.), the spark plugs are fired or diesel is injected. Combustion is about to commence or has commenced when reaching point 3. This point may or may not be after a reference position (usually UD or T_zero) labelled 2. It is now possible to measure or deduce the following elapsed times: T_cycle, the time required for one cycle (engine speed), T_error, the delta of combustion commencing before or after a reference point (UD) and TJg n, the time required after firing spark plugs or injecting diesel until combustion is established. Using these measured or

deduced values, T start, the time at which the next ignition (petrol engine) or injection (diesel engine) is to be initiated (starting from the reference point, usually UD), can be calculated in a continuous fashion.

In theory, T_start = T_cycle - T_ign

Where T_zero (start point) is a reference point (UD, etc.)

In practice, it may be beneficial to measure the deviation (T_err) from a reference point (usually UD) and apply an adaptive, iterative algorithm, to compensate for T_err in steps.

T_start = T_cycle - (Tjgn + T_err)

A cycle usually refers to a power stroke in a four stroke or two stroke engine.

The major difference in how Al operates compared to conventional ignitions is that T_ign and/or T_err can be measured Jn situ", in real time during or after each cycle. With Al there is generally no need to estimate T_ign using stored values from a test engine in combination with proxy sensor values.

For practical purposes, an algorithm may be applied where T_err / 2 or similar is applied, to avoid sudden jumps and a positive delta is added to force ignition slightly after UD. There may also be a plausibility check to confirm that the calculated parameters are within expected limits. Should the firing not have taken place when reaching UD, the firing/ injection should be immediately initiated at UD (generally this applies during the start-up period, when reliable T_ign and T_cycle data are not yet available). There may also be scenarios where Tjgn is not measured on every cycle, requiring a modified algorithm. Since Al allows much faster measurements than conventional ignition control systems, where sensors only react with considerable delay, not every cylinder must be monitored with an Al sensor,

although measuring each cylinder will further improve results and equalize differences across cylinders. To further improve the reliability, an arrangement with several sensors can be used to provide redundancy, i.e. the values from sensors (in one cylinder) can be used instead of a failed sensor (in another cylinder). With such measures, the already excellent reliability of Al sensors can be even further improved. These practical considerations have no impact on the underlying functioning of Al. The illustrations of practical cases only serve to strengthen the algorithm, where such algorithm may vary considerably, depending on the actual application.

When starting an engine, cycle time compared to ignition time is low, hence firing could take place at a reference point (usually UD) and the linear or adaptive algorithm only needs to start when stable values for ignition time and cycle time are available. This may lead to late ignition for the first few cycles, but since the error is generally < 5%, this has no further consequences, particularly when considering that engine start-up is usually not on full load and of very short duration. Firing on UD during start-up is a simple way to start an engine, but there are of course also other algorithms feasible (amongst others: using of seed values, etc.).

In short, this invention concerns itself with a means to detect the transition between the ignition phase and the combustion phase of an internal combustion engine and uses this means to predict when the next firing of spark plugs (or injection of diesel) should take place. To further improve the invention, an adaptive/ iterative algorithm can be applied. For an ignition control system, one primary sensor is required to detect the threshold from ignition to combustion, assisted by a simple secondary sensor (UD position or similar reference point) or other means for referencing purposes. Complex proxy sensors like air flow, air temperature, manifold pressure or throttle position are generally not needed.

Apart from ignition (injection) timing control, Al can additionally, or separately, be used for fuel quantity control or optimisation. With most Al sensors, the intensity of

the combustion process can also be measured directly or indirectly to provide feedback as to how changes in fuel quantities relate to corresponding changes in engine performance. Unlike the ignition timing measurement, measurements of the combustion intensity does not generally allow an absolute or direct assessment whether a parameter was met or missed by how much. Instead, for fuel quantities, only relative measurements can generally be made to provide direct or indirect feedback in regards to the relation to the impact of fuel quantity change on the combustion process. Hence, it may take several cycles as well as a deliberate, periodic oscillation or other variations, to find the optimum fuel quantity.

To also control fuel injection (fuel quantity) with the Al method, an iterative process is proposed, although other processes may also be feasible. One example of such a process is the injection/using of an initial (seed) quantity of fuel. Subsequently, the quantity is altered to iteratively find the optimum fuel quantity by comparing combustion intensities with different fuel quantities. An example of such an algorithm is the injection of an initial fuel quantity, where this quantity is then slightly increased during the next cycle or over a period of time and a combustion intensity comparison is made to see whether the additional quantity has let to an improvement of the combustion. If yes, the quantity is further increased. If not, the fuel quantity is slightly decreased, to the point where the fuel reduction leads to a reduction of the combustion activity. At this point, the quantity is increased again and the cycle starts again. With this approach, the fuel quantity oscillates around the optimum for a given air supply (throttle position), being at all times close to the optimum. Again, the optimum position can be found by essentially needing only one sensor. For this purpose, the signal from the Al sensor needs further analysis, where the signal amplitude, the signal curve and/ or the integral of the signal amplitude over part of the combustion cycle is evaluated. For fuel injection quantity optimisation, one additional sensor may be required to detect load changes, such as a throttle position detector, manifold pressure or airflow sensor. However, this additional sensor does not demand high accuracy. Since only the approximate magnitude of load changes must be detected, such additional sensor can be of a low cost type. The load change

could then be used to approximate the step change required for the fuel quantity. Optimisation of the fuel quantity thereafter could occur iteratively, using for example the adaptive/ oscillation algorithm. For fuel injection purposes, a simple engine temperature sensor may also be beneficial, to differentiate a warm start from a cold start, when turning on an engine.

In short, this invention (Al) can also be used to optimise the fuel quantity which is to be injected/ measured into an engine. A further refinement is the application of an iterative/ oscillation approach, to find the optimum fuel quantity.

Primary sensors used for this purpose (Al) are all sensors which provide direct or indirect clues as to how well and fast combustion takes place during a combustion cycle, or how well combustion took place, when measuring at the end of a cycle or after completion of a cycle. Particularly interesting are optical sensors and torque sensors. It is also possible to implicitly assess the combustion parameters through pressure sensors or timing measurements as well as acoustic sensors. Common to all of these sensors is that they must produce a distinct signal (sharp rise, certain amplitude, or similar) during or at the end of the ignition phase or when transiting into the combustion phase.

In some environments, the primary sensor can be a torque sensor. Torque would ideally be measured between piston and the crank connection. However, this may not always be practical and an arrangement where torque is measured in the engine mounting or other suitable mountings may suffice, provided vibration levels do not mask out the main signal. Figure 2 shows the signal of a piezo sensor installed in the engine mounting. To measure torque in the engine mounting, piezo sensors are particularly adequate. In a similar arrangement, such piezos or other suitable sensors can be used to acoustically measure the progress of a combustion process. Such acoustic sensor must be mounted in a manner to receive mostly the combustion noise, (reasonable S/N factor). Ideal locations for acoustic sensors are

the spark plugs or the cylinder head. Again, this is only practical where vibration noise does not mask out the main signal.

A much simpler approach is to use a crank shaft sensor to measure the crank shaft speed at small intervals. Correct ignition timing will lead to an acceleration after the UD, whereas advance timing will lead to a decrease. See Figure 4 for the results of such a measurement. However, this approach is subject to distortions due to oscillations and influences from the drive shaft/ load and may not be suitable for all applications.

Very accurate results can be achieved using an optical sensor. Figure 3 shows the signal of such an optical sensor. For practical purposes, a high temperature resistant optical fibre (quartz glass or similar) could be used as a "conductor" (glass rod of 1- 3mm diameter for example) and fixed in the cylinder head or integrated into a spark plug (or the injection valve in a diesel application). This "conductor" should protrude into the cylinder/cylinder head space sufficiently (generally in the order of 1-2 cm) to allow the continuous burning off of combustion residues (4 stroke/ diesel) or oil (2 stroke environment). An optical sensor or "conductor" should also protrude sufficiently to be mostly "blind" to the light generated by a spark plug. See Figure 5 for an example. Outside the cylinder head, at a sufficient distance to avoid overheating (generally in the order of 1-2 cm), an optical receiver (example: full spectrum PIN Diode or similar) can be installed. Such an arrangement produces an electrical signal when the combustion process starts. The initial slope of this signal is quite steep, allowing a fairly accurate measurement of the combustion point/band. The amplitude and/ or the integral of amplitude over time during the combustion cycle allows for a simple approximation of the combustion energy. This in turn can be used in an adaptive algorithm to calculate the optimum fuel quantity.

The above description of an optical sensor is only one of many possible options for optical or other sensors. Depending on the environment, different arrangements to place sensors and accessories (for example a glass rod) are feasible, where a

suitable sensor is placed wherever a reliable signal can be obtained. Examples are items connected to the engine (engine mounting, etc.), engine block, cylinder head, parts which are added to the engine (spark plugs, injection valves, pre-heater, etc.).

Some of the benefits of such an arrangement are a reduction in production cost, higher reliability and reduced engine testing in a laboratory for new engines, as well as lower fuel consumption for production engines. The adaptive nature of this method also allows the use of bio-fuels and mixtures thereof (fuel/ethanol, bio-diesel, gas, etc). Compared to conventional sensors, there is also a reduced time lag between detecting input changes (load, environment, fuel, etc.) and being able to adjust ignition timing as well as fuel quantity. A reduction of production costs is possible since fewer (proxy) sensors are required, which also leads to a corresponding saving in interface electronics. Fewer sensors also lead to higher reliability, as measured in mean time between failures. The cost of the Al sensor(s) are marginal (low cost sensors).

This invention (Al) is equally applicable for the design of new engines, as well as the retrofit market. All or some of the Al sensors can be permanently connected to the engine or engine parts. Alternatively, some of the sensors can be placed in consumable items (such as spark plugs) to be replaced at periodic intervals (generating ongoing revenue).

An external factor on the engine is, for example the environment, internal factors are, for example engine status or wear and tear and fuel factors are, for example compositions and quality, mixtures of gasoline and ethanol, bio-fuels or similar.