EP2529090B1

EP2529090B1 - Control arrangement of an electro-hydraulic gas exchange valve actuation system

Info

Publication number: EP2529090B1
Application number: EP10795030.5A
Authority: EP
Inventors: Tony Glader; Mika Herranen; Kalevi Huhtala; Tapio Virvalo
Original assignee: Wartsila Finland Oy
Current assignee: Wartsila Finland Oy
Priority date: 2010-01-29
Filing date: 2010-11-10
Publication date: 2013-07-03
Anticipated expiration: 2030-11-10
Also published as: WO2011092372A3; EP2529090A2; CN102713174B; CN102713174A; KR20120118022A; FI20105082A0; WO2011092372A2; KR101573110B1

Description

Field of Technology

This invention relates to an arrangement to control an electro-hydraulic gas exchange valve actuation system (EHVA system). EHVA systems are used with internal combustion engines.

Background

An internal combustion engine comprises cylinders wherein combustion of fuel occurs. Combustion air is fed to the cylinders through intakes valves, and exhaust gases are routed out of the cylinders through exhaust valves. The valves are driven by actuators that are controlled by hydraulic valves. The hydraulic valves are directed electrically in turn. This kind of arrangement is called as an electro-hydraulic gas exchange valve actuation system i.e. EHVA system.
The nonlinear dynamics and the delay of the EHVA system provide difficulties in controlling the system. A simple feedback controller cannot adjust EHVA in a required accuracy. Therefore more complex and advanced controllers are used. However, tuning of parameters of such a controller is very time consuming, which make the use of the advanced controller impractical in many cases.

Short Description of Invention

The aim of the invention is to alleviate the said control problems. The aim is achieved in a way described in the independent claim. The inventive control arrangement comprises a feedback controller, in an iterative learning controller, and a delay controller. The feedback controller drives EHVA system. The EHVA systems means in this context the system needed to drive one valve or the system driving several valves. For simplicity, it is easier and more understandable to think about the driving of one valve. It is clear that the drive of one valve is expandable to the system driving several valves.
The iterative learning controller drives the feedback controller by adjusting reference signal, and the delay compensator drives the iterative learning controller by modifying the reference signal to be adjusted. Thus, the iterative learning controller provides new reference signal to the feedback controller. The arrangement comprises also a valve lift profile element to provide a lift profile to the iterative learning controller and to the delay compensator.
The iterative learning controller comprises an input interface to receive an actual valve lift data of the valve actuation system, an error memory unit to keep valve position tracking errors in memory, an output command memory unit to keep output commands of the iterative learning controller in memory, and an iteration actuating arrangement for switching on and off an iteration process of the iterative learning controller.
The delay compensator comprises a second input interface to receive the actual valve lift data, a third interface to receive crank angle data of an engine, and a comparator. The comparator compares the actual valve lift and crank angle data pair to a corresponding data pair of the lift profile and performs shift signal as response to the comparison, and transmits the shift signal to the iterative learning controller to drive the iteration actuation arrangement. The iteration process is arranged to stop in case of the shifting the lift profile and to restart when a next cycle of the valve actuation system starts.
Three controllers of the arrangement are arranged in such a way that they do not disturb each other when running.

Drawings

In the next, the invention is described in more detail with help of the figures of the attached drawings in which

Figure 1: shows an example of an inventive arrangement partly,
Figure 2: shows an example of an inventive arrangement,
Figure 3: shows an example of delay compensation according to the invention,
Figure 4: shows an example of the command memory unit of the invention, and
Figure 5: shows an example of the error memory unit of the invention.

Description

Figure 1 shows an example of the iterative learning controller 1 (ILC) of the invention and how it can be connected to the other parts of the invention. In this example, ILC controller drives P-controller 2, i.e. provides new reference to P-controller, but it should be noted that the controller being driven can be any position or tracking controller, for example, PI-, PD-, PID-, or a state controller and feed-forward loops. The P-controller comprises a gain element 6 and an element 7 for adjusting a reference signal transmitted from ILC controller 1 by a feedback signal from the valve actuation system 3 i.e. EHVA system. The feedback signal contains actual valve lift data 5. This data is also transmitted to ILC controller 1 and to the delay controller 16, i.e. the delay compensator (Fig. 2). P-controller sends a control signal 4 for lifting the valve in EHVA.
ILC controller 1 has an interface 24 for receiving said actual valve lift data. As can be noticed, this description does not refer to all interfaces of the elements of the invention. For example, there exist interfaces between the lift profile element 13 and ILC 1, and between the delay compensator 16 and ILC1 for transmitting signals among them, which are not specifically referred. Further, since a number of the interfaces depend on the structure of a real embodiment, it should be noted that one interface can actually take care of several signals to be sent or received. So, connections illustrated between the elements in the figures indicate that there exists an interface in practice through which the connection or connections can be created.
The ILC tracking control method is a method where the control system is learning from previous repeated control trials, weather succeed or failed. The observed error of previous repetition is used to adjust the command in current run so that command actually produces the desired trajectory. This is realized through memory based learning. The simplest ILC scheme is presented in Eq. 1, where u is the control input, i iteration index, q constant learning gain, and Δy tracking error. $u_{1 + 1} (t) = u_{i} (t) + q Δ y_{i} (t)$
Thus, the tracking error is measured at every data point, and saved in the memory as a curve similar than the valve lift curve. In next repetitive round, 'the error curve' of the previous round is added to the previously used reference valve lift curve and the modified curve is also saved. When the process goes on, the valve lift curve gradually transforms to the shape, which results the minimum tracking error of the system.
So, the lift profile element 13 provides a desired valve lift data to ILC controller 1 and an error calculator element 8 calculates the tracking error from the actual valve lift and the desired valve lift, and transmits the position error to the error memory unit 9. The error memory unit 9 keeps the tracking error in memory. A gain element 10 receives the position error from the error memory unit, multiplies the position error with a gain factor. The gain is adjustable by the iteration actuation arrangement. The gained tracking error is received and summed in the summing element 11 with the latest control output signal to be transmitted to the P-controller 2 and to the command memory.
The control output signals, i.e. reference valve lift for the P-controller 2, are kept in memory in the command memory unit 12 from which the latest command (the ILC control output) is delivered to the summing element 11.
The nature of the ILC is, that at some point, the tracking error stops decreasing, and may start to grow. The error cannot go to zero during the whole event due the dynamics of the system, but the learning process continues to modify the reference curve. At some point, the system cannot follow increasing reference value properly, and the error is also increasing. In the invention that problem is solved by stopping the learning process when the error is inside the predefined tolerance. The learning process can be stopped and restarted, for example, by change of learning gain 10 value. If the error is too large, the engine needs to be protected. Therefore the inventive arrangement may contain a tracking error indication unit 28 for following the error. If the error is too large, the alarm can be made in the alarm monitoring unit 29 and the valve actuation system 3 can be stopped or run to a certain state.
Figure 2 shows the inventive arrangement having the delay compensator 16. It was found that delay of the valve actuation system may cause a large part of the tracking error. Therefore the controlling arrangement is continuously watching position difference (lift difference) as a crank angle function between desired and actual position. When the compensator detects difference larger than, for example, 1 CA degree, the arrangement shifts valve lift curve, i.e. the lift profile fed to the controller to the desired direction. This is helpful feature during the start and stop sequences of the diesel engine, when rotational speed is changing rapidly, and other occasional variation of the engine RPM during the run.
So, the delay compensator 16 comprises a second input interface 25 to receive the actual valve lift data, a third interface 30 to receive crank angle data 20 of an engine, a comparator 17 to compare the actual valve lift and crank angle data pair to a corresponding data pair of the lift profile and to perform shift signal 15, 23 as response to the comparison, and to transmit 15, 23 the shift signal to the valve lift profile element 13 to keep the used compensation value in a memory, and to transmit the shift signal to the iterative learning controller 1 to drive the iteration actuation arrangement 21,23'. Memorized delay value is used when profile is reseted or new profile is changed. Value of the delay compensation is then send to the command memory unit 12 with curve initialization. For initialize the profile comparison element 17 or in case of profile change the lift profile element 13 is arranged to transmit the relevant lift profile 26 to the comparison element 17. The iteration process is arranged to stop in case of the shifting the lift profile and to restart when shifting the lift profile is completed.
As can be seen in Figure 1, the iterative learning controller 1 comprises an error memory unit 9 to keep valve position tracking errors in memory, an output command memory unit 12 to keep output commands of the iterative learning controller in memory, and also an iteration actuating arrangement (21, 14', 18' 19', 22', 23') for switching on and off an iteration process of the iterative learning controller. In the figures the arrangement is illustrated in a centralized manner, i.e. having an element 21 to receive the shift signal 23, the change info of the lift profile 14, and the position error 19. The element is arranged to switch on and off the iteration process by guiding the command memory 12 and the elements (9, 10) handling the position errors. The arrangement can also be realized in a distributed fashion in which case there is no central element 21 but similar functionality is located in the elements to be actuated. The signals needed can be delivered directly to these elements, not transmitted through the lift profile element 13 and/or central element 21.
A reference number of a signal indicates a certain signal. For example, reference number 23 indicates the profile shifting signal from the delay compensator 16 to the ILC controller 1. The received shifting signal starts the switching actions of the elements in ILC. These actions and corresponding actuation signals are referred with the same number having ' -mark, like 23'. So, number 18 means that the iterative learning process can started manually', in which case the learning gain 10 is actuated 18' to have a gain value above zero.
In case of the shifting the lift profile due to the delay of the actuation system, the learning gain 10 is actuated to zero, which means that the learning stops. Also vectors in the command memory unit 12 and in the error memory unit are shifted, which is explained later. When ILC process is not active during the shift, it avoids possible negative interactions between delay compensation and ILC process.
Figure 3 illustrates an example of a lift profile and how it is shifted. A lift value is on Y-axes and a crank angle value on X-axes. So, in a certain crank angle value, the valve is lifted a certain amount. The solid curve 31 is a reference lift profile. If a real measurement from the valve actuation system shows a lift value Y1 in crank angle value X2, it can be noticed from curve 32, with dashed line, that the Y1 should be arisen in crank angle X1. This means that the reference lift profile should be shifted in order to matching the measurement and the reference. More than one measurement point can be compared before making a decision for shifting. So, the file comparison element makes that comparison and provides profile shifting info to the lift profile and the ILC 1.
So, for shifting the lift profile and vector elements of the command memory unit and the error memory unit can be indexed by crank angle values. If the comparison of the comparator 17 shows that the measured crank angle and the desired crank angle value with a measured lift position differs enough, the indexes of the lift profile and said vectors are shifted in such a way that the difference decreases.
Figure 4 shows the command memory unit 12 in more detail. The unit comprises a vector element 41 in which the ILC output commands are kept in memory in a vector form. If the shifting occurs, the unit 9 receives a shifting signal 23' to shift a crank angle index of the vector. In this way the values of the vectors are shifted, which corresponds the shift illustrated in Figure 3 By shifting the vector the error of the iterative learning is kept in a reasonable level in the next round. So, for taking care of the shift, the command memory unit comprises a shift element 42 to receive the shift signal and to shift the content of the first vector element as response to the received shift signal.
Figure 5 shows the error memory unit 9 in more detail. The unit comprises vector elements 51 in which the tracking errors are kept in memory in vector forms. If the shifting occurs, the unit 9 receives a shifting signal 23' to shift a crank angle index of the vectors. In this way the values of the vectors are shifted, which corresponds the shift illustrated in Figure 3. By shifting the vectors the error change of the iterative learning is kept in a reasonable level in the next round and synchronized with lift profile inside command memory unit 12. So, for taking care of the shift, the error memory unit comprises a shift element 52 to receive the shift signal and to shift the contents of the second vector elements as response to the received shift signal.
The error vectors can be made for errors calculated in different ways. The different errors can be uses for different purposes.
The crank angle data is received by the command memory unit and the error memory unit for get a real crank angle value for the shift actions.
The error can be observed as an average during the valve lift or certain part of the valve lift. When the average error is inside the tolerance range, i.e. in a certain limited range around a desired error value, ILC iteration process is stopped 22, 22'. If average error goes outside of the tolerance, i.e. outside a second range around the desired error value, ILC iteration process is restarted 22. 22'. If the valve position tracking error is outside a third range around the desired error value, the iteration process is reset. In that third case the error is considered to be so large, that something in ILC process has went wrong, and is better to return in initial valve lift curve and start the ILC process from the very beginning.
The error can be calculated in different ways as well. For example, a sum of the error can be measured during the whole valve stroke. Or instead of the sum, mean square error can be used. Error for one point of the lift profile can also be calculated. So, different type of error vectors can be calculated simultaneously for different purposes. Thus the iteration process for a certain error point can be stopped after the error has achieved a certain range, and the iteration process for another error type still continues. For example, if the one-point-error iteration is stopped, the error is in an acceptable level in this particular point, but the sum-error iteration continues, since it has not achieved an acceptable level.
The inventive embodiment can also comprise a feature which allows changing the lift curve during the engine run. Due the memory-based system of the ILC, right timed shifting between the curves is important. The arrangement allows numerous preloaded lift curves, which the user can bring into play at any time. The arrangement takes care that no overshoots between the curves happens, and change is made in a proper moment.
A command for the profile change on the fly is given 14 to the arrangement. After the command for the change is given, the controller waits until the gas exchange valve is certainly closed, and changes the profile after that. Because the ILC process uses error signal of previous work cycle, the change can be used only after one work cycle, i.e. one cycle of the valve. At the same time, correct target reference fed to tracking error system is delayed by one cycle. If the tracking error with new reference curve is keeping under the third error range mentioned above, the delay compensation is kept at the last known value, memorized in the valve lift profile element 13. This way the error is not necessarily grown very high even the new profile is a basic unlearned profile in the beginning of the change. Therefore the change of the valve timing causes no alert.
So, referring to Figure 1 the valve lift profile element 13 comprises a number of lift profiles, and change of the lift profile is arranged to be run when receiving a command 14 for the change, in which case the iteration process is arranged to stop when the valve position indicates the valve being closed, and to restart after the cycle of the valve actuation system, i.e. the cycle of the valve. For changing the profile the ILC comprises necessary elements for directing 14' the new profile the command memory unit. The same elements can take care of the restart of the ILC if the tracking error is too large 19, 19'.
The feedback controller 2, the iterative learning controller 1, the delay compensator 16, and the valve lift profile element 13 can be implemented in one entity 27, which entity comprises connections among the controllers, compensator and the profile element, and interfaces for the valve actuation system and the internal combustion engine. A grouping of the elements, components and functions of the invention can be made in many ways. An implementation of the invention can be made, for example, by printed circuits or software programs installed on a computer device.
It is clear from this description that the invention can be obtained in many different ways in the scope of the claims.

Claims

A control arrangement of an electro-hydraulic gas exchange valve actuation system, which arrangement comprises a feedback controller (2) to drive the valve actuation system, characterized in that the arrangement further comprises
- an iterative learning controller (1) to drive the feedback controller,

- a delay compensator (16) to drive the iterative learning controller, and

- a valve lift profile element (13) to provide a lift profile to the iterative learning controller and to the delay compensator,
the iterative learning controller (1) comprising
- an input interface (24) to receive an actual valve lift data (5) of the valve actuation system,

- an error memory unit (9) to keep valve position tracking errors in memory,

- an output command memory unit (12) to keep output commands of the iterative learning controller in memory, and

- an iteration actuating arrangement (21, 14', 18' 19', 22', 23') for switching on and off an iteration process of the iterative learning controller,
the delay compensator (16) comprising
- a second input interface (25) to receive the actual valve lift data,

- a third interface (30) to receive crank angle data (20) of an engine, and

- a comparator (17) to compare the actual valve lift data (5) and crank angle data (20) pair to a corresponding data pair of the lift profile, to provide shift signal as response to the comparison, and to transmit the shift signal to the iterative learning controller (1) to drive the iteration actuation arrangement (21,23'), which iteration process is arranged to stop in case of the shifting the lift profile and to restart iteration process after completion of the shifting the lift profile.
A control arrangement according to claim 1, characterized in that the command memory unit (12) comprises a first vector element (41) to keep the output commands, and a first shifting element (42) to receive the shift signal and to shift the content of the first vector element as response to the received shift signal, and the error memory unit (9) comprises second vector elements (51) to keep the tracking errors, and a second shifting element (52) to receive the shift signal and to shift the contents of the second vector elements as response to the received shift signal.
A control arrangement according to claim 2, characterized in that the lift profile and said first and second vector element are indexed by crank angle values and if the comparison of the comparator (17) shows that the measured crank angle and the desired crank angle value with a measured lift position differs enough, said vectors are shifted in such a way that the difference decreases.
A control arrangement according to claim 3, characterized in that if the valve position tracking error is in a certain limited range around a desired error value, the iteration process is discontinued, and
if the valve position tracking error is outside a second range around the desired error value, the iteration process is restarted, and
if the valve position tracking error is outside a third range around the desired error value, the iteration process is reset.
A control arrangement according to claim 4, characterized in that the valve lift profile element (13) comprises a number of lift profiles, and change of the lift profile is arranged to be run when receiving a command for the change, in which case
the iteration process is arranged to stop when the valve position indicates the valve being closed, and to restart after the cycle of the valve actuation system.
A control arrangement according to claim 4 or 5, characterized in that the iterative learning controller (1) comprises
an error calculator element (8) to calculate the valve position tracking error from the actual valve lift data and the lift profile in use and to transmit the position error to the error memory (9), and
a gain element (10) to receive the position error from the error memory, to multiply the position error with a gain factor, which gain is adjustable by the iteration actuation arrangement and
a second summing element (11) to receive the multiplied position error and the latest output command from the command memory, to sum the received position error and the output command, and to transmit the sum to the feedback controller (2) and to the command memory.
A control arrangement according to claim 6, characterized in that the iteration actuating arrangement (21, 22) comprises an element (21) to receive the shift signal, the change of the lift profile, and the position error, which element is arranged to switch on and off the iteration process by guiding the command memory (12 and the elements (8, 9, 10, 11) handling the position errors.
A control arrangement according to claim 7, characterized in that the feedback controller (2), the iterative learning controller (1), the delay compensator (16), and the valve lift profile element (13) are implemented in one entity, which entity comprises connections among the controllers, compensator and the profile element, and interfaces for the valve actuation system and the internal combustion engine.
A control arrangement according to claim 7 or 8, characterized in that the shift of the lift profile occurs when the difference between a desired crank angle position and an actual position is larger than 0,5 degree.
A control arrangement according to claim 9, characterized in that the feedback controller (2) is P-, PI-, PD-, PID-, or a state controller.