EP3090167B1

EP3090167B1 - Control system and control method for an internal combustion engine, and an internal combustion engine

Info

Publication number: EP3090167B1
Application number: EP14821238.4A
Authority: EP
Inventors: Tom Kaas; Fredrik ÖSTMAN
Original assignee: Wartsila Finland Oy
Current assignee: Wartsila Finland Oy
Priority date: 2014-01-03
Filing date: 2014-12-17
Publication date: 2019-02-06
Anticipated expiration: 2034-12-17
Also published as: KR20160104070A; CN105934575B; CN105934575A; WO2015101706A1; KR102067868B1; FI125058B; FI20145008A; EP3090167A1

Description

TECHNICAL FIELD

The invention concerns in general the technology of internal combustion engines, such as large diesel engines. In particular the invention concerns the way in which feedback control is utilized to control the values of dynamic quantities in the internal combustion engine during its operation. An example of background art is disclosed in US2011/0126807 A1 .

BACKGROUND OF THE INVENTION

Operating a modern internal combustion engine, such as a large diesel engine, involves setting up and running a number of feedback control loops to control the respective processes that take place as a part of the operation of the engine. Fig 1 is a schematic illustration of a process 101 and a controller 102 that applies feedback control. A sensor 103 monitors the state of the process 101 and produces a feedback value, which is an indicator of a measured dynamic quantity such as e.g. pressure, temperature, speed, frequency, flow rate, surface level, or the like. The controller 102 compares the feedback value to a setpoint value and produces an output on the basis of the comparison. The output constitutes a control signal to an actuator 104, with the aim of changing the state of the process 101 so that the difference between the feedback value and the setpoint value would become as small as possible. Known feedback control schemes may include for example proportional control, integral control, and/or derivative control. In these the intensity of corrective action depends on current difference to setpoint (proportional), weighted sum of current and previous differences (integral), or slope of the difference over time (derivative). Disturbances are factors that tend to change the state of the process 101. Measurable disturbances are those, the effects of which are known beforehand and/or can be measured online with reasonable accuracy. Additionally there are non-measurable disturbances, which may involve e.g. the mechanical wear of components in the process 101. The effect of non-measurable disturbances on the state of the process 101 are difficult, if not impossible, to predict.
Pure closed-loop feedback control such as in fig. 1 involves the inherent disadvantage that it only reacts to effects that have already taken place in the process, and thus involves certain latency and dynamics. Fig. 2 illustrates how this disadvantage can be at least partly dealt with by adding an element of feed-forward control. The control system shown in fig. 2 comprises, in addition to the elements explained above in association with fig. 1, a feed-forward controller 201. It is configured to receive one or more input values that are indicative of the currently actual measurable disturbances. The feed-forward controller 201 produces an output that is at least partly based on its input value(s). The outputs of both the feedback controller 102 and the feed-forward controller 201 are coupled to a combiner 202, which delivers their combination as a control signal to the actuator 104. The combination is not necessarily a straightforward sum, but it is intuitive to think that the way in which the actuator 104 should affect the state of the process 101 takes into account the outputs of both controllers.
As an example we may think that the process 101 is common rail fuel injection, and the sensor 103 monitors the pressure in the fuel delivery line delivering fuel to injectors for injecting into cylinders of the internal combustion engine. In this example the actuator 104 drives the flow control valve, which controls the fuel flow into the fuel delivery line (i.e. the rail). A deliberate increase in injection duration is a measurable disturbance. If only feedback control was applied according to fig. 1, the increased injection duration would cause a pressure drop. The sensor 103 would convey decreasing pressure values to the feedback controller 102, which would then try to compensate the measured pressure drop by using the actuator 104 to open more the flow control valve. Latency and dynamics in the feedback control loop would mean that a certain transient drop in the common rail pressure was inevitable.
If the control system of fig. 2 was in use, the feed-forward controller 201 would receive information about the increase in injection duration in real time. The feed-forward controller 201 can then react quickly by producing an output signal which, after going through the combiner 202, increases the fuel flow into the rail faster than in the simple feedback control case explained above.
The weakness of the combined control approach of fig. 2 is that the feed-forward controller 201 inevitably operates on the basis of assumptions about how the measurable disturbances will affect the process. Such assumptions may lose their accuracy over time, or they may fail to take into account unexpected changes. For example, in a new engine that receives clean fuel a command to increase injection duration by a certain fraction of crank angle will cause a certain increase in the injected amount of fuel per cycle. If 5000 hours of operation have passed since the last injector overhaul, and/or if the consistency of the fuel is not quite what it should be, the same command may cause a significantly different increase in the injected amount of fuel. Mechanical wear of injectors and varying consistency of fuel are examples of non-measurable disturbances.

SUMMARY OF THE INVENTION

The following presents a simplified summary in order to provide a basic understanding of some aspects of various invention embodiments. The summary is not an extensive overview of the invention. It is neither intended to identify key or critical elements of the invention nor to delineate the scope of the invention. The following summary merely presents some concepts of the invention in a simplified form as a prelude to a more detailed description of exemplifying embodiments of the invention.
A method, control system, and an internal combustion engine would be needed in which the control approach could take into account also non-measurable disturbances, despite them being non-measurable. The control approach should be versatile so that it could be applied to controlling various processes in the internal combustion engine.
Advantageous objectives of the invention are achieved by using a primary controller for feedback-type control and a secondary controller for feed-forward-type control, and by additionally making the secondary controller aware of trends in the output of the primary controller so that the operation of the secondary controller can be changed in an adaptive manner.
A desired kind of adaptation of the secondary controller can be implemented so that the aim is to maintain the output of the primary controller at a fixed value, which may be zero or other corresponding "neutral" value. A neutral output of the primary controller is defined as the output the primary controller produces when it does not try to actively affect the state of the controlled process. Filtering, such as taking a mean or median value over a predefined time window, can be applied in order to make the adaptation of the secondary controller concentrate on trends in the primary controller output rather than transients. The exemplary embodiments of the invention presented in this patent application are not to be interpreted to pose limitations to the applicability of the appended claims. The verb "to comprise" is used in this patent application as an open limitation that does not exclude the existence of also unrecited features.
The novel features which are considered as characteristic of the invention are set forth in particular in the appended claims. The invention itself, however, both as to its construction and its method of operation, together with additional objects and advantages thereof, will be best understood from the following description of specific embodiments when read in connection with the accompanying drawings.

BRIEF DESCRIPTION OF DRAWINGS

Fig. 1: illustrates a prior art feedback control scheme,
fig. 2: illustrates a known combination of feedback and feed-forward control,
fig. 3: illustrates an adaptive control system and method,
fig. 4: illustrates an example of adapting a transfer function,
fig. 5: illustrates another example adapting a transfer function,
fig. 6: illustrates another example of adapting a transfer function, and
fig. 7: illustrates the application of an adaptive control system for controlling fuel pressure in a common rail.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

Fig. 3 can be read as an illustration of a control system for an internal combustion engine, by associating the illustrated entities with functional blocks of the control system. Alternatively fig. 3 can be read as an illustration of a method for controlling a process in an internal combustion engine, by associating the illustrated entities with method steps. Both interpretations are explained in more detail below.
As illustrated in fig. 3, the control system comprises a primary controller 301 that is configured to compare a feedback value to a setpoint value and to produce a primary output. The primary output is formed on the basis of said comparison; as a very simple example any change in the primary output may be proportional to the difference between the feedback value and the setpoint value. More elaborate relations between the primary output and the result of the comparison are possible. For example, there may be a "dead zone" of the very smallest comparison results that cause no change in the primary output at all, and/or the proportionality (if any) between any change to the primary output and the difference between the feedback value and the setpoint value may be linear, squared (and signed), or exponential, or it may have some other form.
The feedback value is an indicator of a measured dynamic quantity in a process 101 of the internal combustion engine. For example, if the dynamic quantity to be measured is pressure, the sensor 103 is a pressure sensor that may convert the measured pressure to a corresponding voltage, current, or resistance value. The feedback value could be even a mechanical displacement, for example if the measurement of pressure was based on a reversible deformation caused by said pressure, but since the implementation of feedback control typically involves an electronic control system, feedback values in electric form are preferable.
A secondary controller 302 is configured to receive an input value and to produce a secondary output. The words "primary" and "secondary" are just names that are used for the sake of unambiguous literal reference, and they include no connotations about e.g. the mutual significance of the control functions, or the respective control functions taking place in some particular order. The input value is schematically shown as coming to the secondary controller 302 from the left, and it is an indicator of a measurable disturbance affecting the process 101.
The production of a secondary output in the secondary controller 302 takes place according to a transfer function. A simple example of a transfer function is a time-independent transfer function $s = s (i)$
according to which each input value i results in a corresponding output s. More elaborate transfer functions may be used: for example, if the output s should depend not only on the input i but also at the time t at which the input comes to the secondary controller 302, we may write the general expression $s = s (i, t)$
The output value s(t) to be produced at a particular time t may include a weighted sum of the current input value i(t) and some previous input values according to the general formula $s (t) \sim \sum_{n = 0}^{N} a_{n} i (t - n)$
where the a_n are summing weights and the i(t - n) are the input values at times t, (t-1), (t-2), ... (t-N). The present invention does not place any particular restrictions to the transfer function, but in graphical illustrations and examples it is most straightforward to use a time-independent one-to-one relationship that maps each input value to a corresponding output value.
A combiner 303 is coupled to receive the primary output from the primary controller 301 and the secondary output from the secondary controller 302. It is coupled to deliver a combination of them as a control signal to an actuator 104, which in turn is configured to affect the process 101. The word combination is used here in a wide sense. It may mean a simple sum of the primary and secondary outputs, or it may mean a weighted sum, a filtered sum, and/or some other result that takes into account the outputs of both controllers and has a range of possible values that is suited to drive the actuator 104 so that the desired effect on the process 101 is achieved.
The secondary controller 302 is coupled to receive the primary output as such and/or some derivative thereof. The word derivative as used here means "something that is derived from", and is thus not restricted to e.g. a time derivative. Examples of derivatives meant here are for example a mean or median value of the primary output over a predefined time window.
The secondary controller 302 is configured to adapt its transfer function based at least partly on an aim of maintaining the primary output at a fixed value. This fixed value is preferably a so-called neutral value; in other words, the act of adapting the transfer function in the secondary controller aims at achieving a situation in which the primary controller would not try to actively affect the state of the controlled process 101.
Adapting the transfer function is illustrated in the following with some examples, and with reference to figs. 4, 5, and 6. In fig. 4 the leftmost case illustrates a situation where the transfer function takes initially the form of a relatively smooth curve. At some moment the secondary controller becomes aware that the primary output (or a derivative thereof, as mentioned above) has the value Δ₁. The secondary controller is configured to respond by augmenting or scaling all outputs given by the transfer function with a constant that is equal or proportional to the value Δ₁ (being equal is a special case of being linearly proportional, with the linear proportionality constant 1). Augmenting all outputs of the transfer function by Δ₁ is shown in the middle part of fig. 4. If we assume that the transfer function was initially of the form s = s(i), it is now s = s(i) + Δ₁. As an alternative, if all outputs given by the transfer function would be scaled (rather than augmented) with a constant aΔ₁ proportional to the value Δ₁, the new transfer function would be of the form s = aΔ₁ s(i). A combination of scaling with a proportionality constant a and augmenting with proportionality constant b, the new transfer function could be expressed as s = aΔ₁ s(i) + bΔ₁.
The concept of being proportional can be generalized to mean all cases where the magnitude of the augmenting or scaling constant increases unambiguously and monotonously with increasing values of Δ₁. Thus dependencies that qualify as proportional are e.g. linear proportionality, exponential proportionality, logarithmic proportionality, and piecewise defined proportionality.
Simultaneously the middle part of fig. 4 shows that a next value of the primary output (or a derivative thereof, as mentioned above) is received, and has the value -Δ₂. During the time period that is known to have affected the generation of this primary output, the input i to the secondary controller had some other, relatively large value, for which reason the circled-cross symbol of the newly received primary output (or derivative thereof) appears in the right-hand part of the input/output diagram. The following response of the secondary controller in adapting the transfer function is shown in the right part of fig. 4: this time the secondary controller responds by augmenting all outputs of the transfer function by -Δ₂. Thus the act of adapting the transfer function means in this case moving the transfer function curve up or down by the amount indicated by the primary output (or derivative thereof).
Since each step of adapting the transfer function according to the model shown in fig. 4 moves the whole transfer function curve, it is preferable that the value on which the adapting is based is some filtered version of the primary output value, like a mean or median value over a relatively long time window. Also, this kind of approach to adapting the transfer function is most suitable for cases in which we may be reasonably sure about the form of the transfer function, but non-measurable disturbances that are discrete by appearance and take place relatively seldom constitute a basis for the adaptation. An example of such a non-measurable disturbance could be a change in the exact constitution of fuel. When a nearly empty fuel tank is filled to the top from a different source than earlier, the exact constitution of fuel that is available to the engine may change relatively abruptly, but stays more or less the same after that, until the next fill-up.
In fig. 5 the leftmost part illustrates the same starting point as above in fig. 4: the initial form of the transfer function is a relatively smooth curve, and a primary output value (or derivative thereof) is found to have the magnitude Δ₁ during a period of time when a characteristic input to the second controller was i ₁. However, in this case the secondary controller does not start moving the whole transfer function curve. Rather, it associates said primary output (or derivative thereof) with a particular sub-range of input values Δi, which includes the input i ₁ that was characteristic for a period of time over which said primary output (or derivative thereof) was obtained. The secondary controller adapts locally the transfer function so that outputs that the previous form of the transfer function gives for inputs within said sub-range are augmented with values proportional to said primary output (or derivative thereof).
The middle part of fig. 5 shows one example of such local adapting. Within the sub-range of input values Δi, the transfer function curve is stretched so that it reaches the point that was above the original transfer function curve by Δ₁. In order not to create discontinuities in the transfer function curve, the effect of the adaptation is inversely proportional to the difference between the respective input and the characteristic input mentioned above. Another possibility would have been to cut a piece of the original transfer function curve within the sub-range of input values Δi, and to move that piece translationally upwards by Δ₁, but that would naturally result in a discontinuity in the transfer function curve at both ends of the sub-range Δi.
Mathematically the adapted transfer function could be expressed as $s = \begin{array}{l} s (i) + t (i), & when i \in Δ i \\ s (i), & elsewhere \end{array}$
where t(i) is an augmentation function that is defined within the sub-range of input values Δi. The middle part of fig. 5 also shows that the next received primary output (or derivative thereof) is associated with a significantly larger concurrent input value, and is below the (original!) transfer function curve by Δ₂. The rightmost part of fig. 5 shows how also in that case the transfer function has been adapted locally so that outputs that the previous form of the transfer function gives for inputs within the appropriate sub-range (not separately shown) are augmented with values proportional to said primary output (or derivative thereof). Again, graphically the result seems like stretching the transfer function curve so that one part of it reaches the point at which the circled-cross symbol appeared.
Repeated adaptations of the transfer function in this way will eventually adapt the transfer function curve so that in its adapted form it goes through all points in the input/output plane for which a corresponding primary output (or derivative thereof) has been received. Such an approach to adapting is well suited for cases in which it would be difficult to define exactly the most optimal transfer function on the basis of pre-existing information only.
Fig. 6 illustrates yet another example of adapting a transfer function. In this case the secondary controller has the nature of a self-organizing map or neural network, and it is coupled to receive two different and mutually independent types of input values. In this schematic illustration we assume that each possible pair of received values (INPUT 1, INPUT 2) makes the secondary controller produce a secondary output, the value of which is represented by the phase angle (angle in relation to the horizontal direction to the right) of the corresponding arrow in the drawing. The transfer function is equal to the unambiguous mapping from each possible pair of input values to the corresponding output value.
On the left in fig. 6 we assume that a primary output (or derivative thereof) is received in the secondary controller, and said primary output concerns a time period during which a particular characteristic pair of values (INPUT 1, INPUT 2) is received by the secondary controller as represented by point 601. We also assume that the secondary output value that was previously associated with this pair of input values was the one represented with a dashed line to the upper right from the point 601. The received primary output (or derivative thereof) defines the new secondary output value 603 in a way that is analogous to that applied above in figs. 4 and 5: it is assumed that if the secondary output had already had the value 603, the corresponding primary output would have had a neutral value.
A further assumption in the left-hand part of fig. 6 is that the concept "sub-range of input values" that was used in association with fig. 5 has a corresponding two-dimensional form in the self-organizing map or neural network. In other words, the effect of changing the output value associated with point 601 will "bleed" into its immediate surroundings, and cause similar (yet smaller) changes in the output values associated with neighboring points. The points that will be affected are those that fit in the elliptical region 604. The right-hand side of fig. 6 shows the self-organizing map or neural network after the whole adaptation round has been made. Dashed lines illustrate the previous output values associated with those points for which a new output value was defined as a part of adapting the transfer function (note that the previous value for the actual point 601 is not shown any more on the right, because it was already shown in the left-hand part).
The mapping from two inputs to one (secondary) output in fig. 6 can be generalized so that the secondary controller may have any number of mutually depending and/or mutually independent inputs, as long the transfer function is unequivocally defined as a mapping from each possible combination of input values to a corresponding secondary output value.
Fig. 7 illustrates one possible practical application of a control system described above in an internal combustion engine, such as a large diesel engine of the common rail type. On the lower right in the drawing are a fuel delivery line 701 and one or more injectors 702 for injecting fuel coming from the fuel delivery line 701 into cylinders (not shown) of the internal combustion engine. The dynamic quantity to be measured is the fuel pressure in the fuel delivery line 701. A sensor 103 is configured to measure the fuel pressure and to provide a feedback value to the primary controller 301, which feedback value is an indicator of the measured fuel pressure. The actuator 104 is a flow control apparatus that is configured to regulate the flow of fuel 703 into the fuel delivery line 701.
The input value to the secondary controller 302 is an indicator of the injection duration of one or more of the injectors 702. A deliberate increase in injection duration aims at increasing the output power of the engine, and requires a corresponding increase in the flow of fuel into the fuel delivery line 701. Thus when the secondary controller 302 receives an input that indicates an increase in injection duration, it produces a secondary output that goes through the combiner 303 to the actuator 104 and increases the fuel flow.
Non-measurable disturbances include all such factors that make this increase in fuel flow inaccurate for reasons that would be difficult or impossible to predict. For example if the flow control apparatus is worn, a particular movement of the actuator 104 may increase the fuel flow too much or too little. Feedback control through the loop including the sensor 103 and primary controller 301 corrects the fuel pressure, and the secondary controller 302 receives knowledge about the appeared need for correction in the form of the primary output that the primary controller 301 produced. If the initial increase in fuel flow was too small, the primary controller 301 produced a primary output that moved the actuator 104 a little bit further. The secondary controller 302 notices this, so it becomes aware that next time when a similar increase in injection duration is made, the secondary controller 302 should already in the first place move the actuator 104 a little more than previously.
Similar principles can be applied to the controlling of various processes in an internal combustion engine. In order to ensure the applicability of the description given above, it is advantageous that if a more proactive input signal and a more reactive input signal are available, the more proactive one is used as the input to the secondary controller while the more reactive one is used a feedback value fed to the primary controller. For example, if pilot fuel injection is used in a dual-fuel engine (or pilot gas injection in a solely gas-fuelled engine), the pilot fuel pressure (or pilot gas pressure) could be controlled so that information about pilot duration is used as an input to the secondary controller and a measured pressure in the pilot delivery line as a feedback value to the primary controller. Also in a dual-fuelled or gas-fuelled engine the main gas pressure control could come into question, so that the main gas duration is used as an input value to the secondary controller and the main gas pressure as a feedback value to the primary controller.

Claims

A fuel delivery control system for an internal combustion engine, said system comprising:
- a primary controller (301) configured to compare a feedback value to a setpoint value and to produce a primary output on the basis of said comparison, wherein said feedback value is an indicator of fuel pressure in a fuel delivery line (701) in said internal combustion engine,

- a secondary controller (302) configured to receive an input value and to produce a secondary output according to a transfer function from said input value, wherein said input value is an indicator of injection duration of one or more injectors (702) that inject fuel coming from said fuel delivery line (701) into cylinders of said internal combustion engine, and

- a combiner (303) coupled to receive said primary and secondary outputs and to deliver their combination as a control signal to a flow control apparatus configured to regulate the flow of fuel (703) into said fuel delivery line (701);
wherein said secondary controller (302) is coupled to receive a filtered primary output and configured to adapt said transfer function based at least partly on an aim of maintaining said filtered primary output at a fixed value, wherein said filtered primary output represents one of a mean value or a median value of said primary output over a predefined time window.
A control system according to claim 1, wherein said fixed value is a neutral value that does not actively affect the injection duration of the one or more injectors (702).
A control system according to any of the previous claims, wherein said secondary controller (302) is configured to respond to a received filtered primary output value by augmenting or scaling outputs given by said transfer function with a constant that is proportional to said filtered primary output value.
A control system according to claim 1 or 2, wherein said secondary controller (302) is configured to respond to said filtered primary output by associating said filtered primary output with a particular sub-range of input values that includes an input that was characteristic for a period of time over which said filtered primary output was obtained, and by locally adapting said transfer function so that outputs that said transfer function gives for inputs within said sub-range are augmented with values proportional to said filtered primary output.
A control system according to claim 4, wherein said secondary controller (302) is configured to augment outputs, which said transfer function gives for inputs within said sub-range, with values that are inversely proportional to a difference between the respective input and said characteristic input.
An internal combustion engine, comprising:
- a fuel delivery line (701),

- one or more injectors (702) for injecting fuel coming from said fuel delivery line (701) into cylinders of the internal combustion engine, and

- a control system according to claim 1.
A method for controlling fuel delivery in an internal combustion engine, the method comprising:
- measuring fuel pressure in a fuel delivery line (701) in said internal combustion engine and producing a feedback value indicative of the fuel pressure in said fuel delivery line (701),

- comparing (301) said feedback value to a setpoint value and producing a primary output on the basis of said comparison,

- producing (302) a secondary output according to a transfer function from an input value that is an indicator of injection duration of one or more injectors (702) that inject fuel coming from said fuel delivery line (701) into cylinders of said internal combustion engine,

- using a combination (303) of said primary and secondary outputs as a control signal to a flow control apparatus configured to regulate the flow of fuel (703) into said fuel delivery line (701), and

- using a filtered primary output to adapt (302) said transfer function based at least partly on an aim of maintaining said filtered primary output at a fixed value, wherein said filtered primary output represents one of a mean value or a median value of said primary output over a predefined time window.
A method according to claim 7, wherein said transfer function is adapted based on an aim of maintaining said filtered primary output at a neutral value that does not actively affect the injection duration of the one or more injectors (702).
A method according to claim 7 or 8, wherein said filtered primary output comprises a filtered primary output value, and the adapting of said transfer function involves augmenting or scaling outputs given by said transfer function with a constant that is proportional to said filtered primary output value.
A method according to claim 7 or 8, wherein said filtered primary output is associated with a particular sub-range of input values including an input that was characteristic for a period of time over which said filtered primary output was obtained, and said transfer function is adapted locally so that outputs that said transfer function gives for inputs within said sub-range are augmented with values proportional to said filtered primary output.
A method according to claim 10, wherein outputs, which said transfer function gives for inputs within said sub-range, are augmented with values that are inversely proportional to a difference between the respective input and said characteristic input.
The use of a method according to any of claims 7 to 11 to control a fuel pressure in a common rail (701) of a common rail type diesel engine.