GB2465981A

GB2465981A - Vibration Control for a Vehicle

Info

Publication number: GB2465981A
Application number: GB0822037A
Authority: GB
Inventors: Michael Lindenstruth
Original assignee: GM Global Technology Operations LLC
Current assignee: GM Global Technology Operations LLC
Priority date: 2008-12-03
Filing date: 2008-12-03
Publication date: 2010-06-09
Also published as: GB0822037D0

Abstract

A vibration control system for controlling an active engine mount 34 to reduce vibrations of a car engine. The system comprises a force actuator acting on an engine mount controlled based on a force control signal; At least one vibration sensor 30, 32 for detecting a vibration; A controller 28 characterised by: Determining the force control signal for controlling said actuator based on a force reference signal, a stored force correction signal and said vibration data; An inverting unit for computing a first force correction signal based on the vibration data and the force reference signal; A filtering unit for computing a second force correction signal based on the said first force correction signal and the stored force correction signal; An iteration memory for storing said second signal as a stored force correction signal. Preferably, the control system comprises a learning control device for the suppression of periodic vibrations of the car engine. This learning device is referred to as POISON (periodic on-line iterative signal optimum navigation) controller. By iterative learning of a corrected reference signal the POISON controller is able to compensate for control errors which occur in conventional closed loop control systems. The POISON controller is capable of permanent online operation. Therefore it is able to compensate for certain changes in the controlled system. The POISON controller can easily be added to existing vibration control loops.

Description

TITLE

Vibration controller for attenuating engine generated vibra-tion in an automobile

FIELD OF TECRNOLOGY

The application relates to a vibration controller for attenu-ating engine-generated vibrations in an automobile.

BACKGROUND

An engine of an automobile can cause vibration to the automo- bile. If the engine is an internal combustion engine, the vi-bration can be produced by engine torque pulsation and mass imbalance of the engine. Periodic vibrations which are caused by mass imbalances occur at the revolution frequency of the motor shaft. In a four stroke engine there are also periodic vibrations at half the revolution frequency due to the four stroke combustion process which has a period of two revolu-tioris. Further periodic motor vibrations occur at multiples of the aforementioned frequencies and also at specific vibra-tion frequencies of car components. The engine vibrations can be reduced by an active engine mount. The vibration reduction is accomplished by measuring the motor vibrations, generating a compensating signal by a vibration controller and sending the compensating signal to actuators which are in mechanical connection with the engine mount.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

Fig. 1 illustrates a schematic view of a vibration control system, Fig. 2 illustrates a schematic view of an active engine mount, Fig. 3 illustrates vibration patterns of an engine, Fig. 4 illustrates a digital controller, Fig. 5 illustrates components of a POISON controller, Fig. 6 illustrates further details of the components of the POISON controller, Fig. 7 illustrates the components of a proportional-integral-derivative lag (PIDL) controller, Fig. 8 illustrates a PID controller in the controller of Fig. 5, Fig. 9 illustrates the determination of initial values for a stored correction signal in the controller of Fig. 5, Fig. 10 illustrates the computation of control signals, Fig. 11 illustrates a further embodiment of a controller comprising an analog PID controller, Fig. 12 illustrates a further embodiment of a digital con-troller, Fig. 13 illustrates a further embodiment of a digital con-troller having a parallel arrangement of the POISON controller, and Fig. 14 illustrates a further embodiment of a digital con-troller having a parallel arrangement of the POISON controller.

DETAILED DESCRIPTION

In the following description, details are provided to de- scribe the embodiments of the application. It shall be appar-ent to one skilled in the art, however, that the embodiments may be practised without such details.

Fig. 1 illustrates schematically an automobile 10 equipped with a vibration control system 1 which includes a vibration controller 28. The automobile 10 comprises a body 12, an in- ternal combustion engine 14, and a transmission 16. The in- ternal combustion engine 14 is a four-cylinder and four-stroke engine. The engine 14 and the transmission 16 are both supported on a body 12 of the automobile 10 by a right mount-ing unit 18 and a left mounting unit 20 for driving a front right wheel 22 and a front left wheel 24 of the automobile 10 respectively. The engine 14 includes a rotation sensor 26, which is in a form of a variable reluctance sensor for de-tecting the rotation angle of a crankshaft (not shown) of the engine 14 via a toothed flywheel which can be seen in Fig. 2.

The rotation sensor 26 generates a rotational angle signal, denoted as SPEED, of the crankshaft having pulses correspond-ing to passages of wheel teeth. The SPEED signal is sent via the connection 33 to the vibration controller 28.

The vibration controller 28 is a dual channel vibration con-troller with two vibration sensors 30, 32 and two actuators 34, 36. The acceleration sensors 30, 32 are mounted to the body 12 at opposites sides of the automobile 10 close to the mounting units 18, 20 for monitoring the vibration of body 12. The acceleration sensors 30, 32 are accelerometers. The actuators 34, 36 are also mounted to the automobile body 12 at opposite sides of the automobile 10 proximate the mounting units 18, 20, and are used to inversely vibrate the automo-bile body with respect to vibrational frequency components transferred to the body 12 from the engine 14. The actuators 34, 36 are electromechanical inertial mass shakers.

The rotation sensor 26 can also be realized in the form of a magneto-resistive sensor. The acceleration sensors 30, 32 can alternatively be in the form of displacement or vibration measuring devices. Other known types of actuators can also be used as alternatives to the shakers 34, 36 for mechanically vibrating the automobile body 12 in response to electrical signals. For example, piezoelectric or electromechanical ac-tuators connected in parallel with the mounting units 18, 20 between the body 12 and engine 14 can be used instead of the shakers 34, 36.

In an electric automobile or a hybrid car that employs an electric motor for providing a driving torque, a rotatable shaft of the electrical motor replaces the crankshaft of the internal combustion engine 14. Both the rotatable shaft of the electrical motor and the crankshaft of the internal com-bustion engine 14 are forms of a motor shaft.

During operation, the engine 14 generates mechanical vibra-tions containing sinusoidal components having amplitudes and frequencies that vary in relation to the rotational angle of the crankshaft (i.e. rotation of the engine). The four cyl-inder four-stroke engine 14 generates four combustion strokes during two revolutions of the crankshaft.

A uniform timing of the cylinders thus leads to a basic vi- bration period of 4rt/4 = 1 ii in addition to the basic vibra-tion periods of 2 ri and 4 n, corresponding to one crankshaft revolution and to one combustion cycle. Motor vibrations oc- cur at multiples of the corresponding frequencies. These en- gine-generated vibrations are transferred through the mount-ing units 18, 20 to vibrate the automobile body 12. There are also motor vibrations other than the abovementioned types, for example vibrations caused by the opening and closing of the valves or by vibrations of the car which are transmitted to the engine.

In an operation of attenuating the vibration, the dual chan-nel vibration controller 28 receives the angle signal from the rotation sensor 26, and generates an OUTPUT1 signal to a first channel 29 to drive the actuator 34, and an OUTPUT2 signal to a second channel 31 to drive the actuator 36. The acceleration sensors 30, 32 generate actual value signals ER-ROR1 and ERROR2 which are indicative of the vibration of the body 12 at their respective locations. The ERROR1 and ERROR2 signals are sent via two channels 25, 27 respectively, of the vibration controller 28.

Fig. 2 illustrates in more detail a portion of the automobile body 12 showing the location of sensor, actuator components and mounting units 18, 20 that support the engine 14 and transmission 16 on a right panel 12a and a left panel 12b of the automobile body.

The mounting units 18, 20 are vibration isolating type engine mounts that include body and engine mounting brackets sepa-rated by rubber, which is strong and resilient. The right body panel 12a and left body panel 12b represent portions of the automobile body 12 for separating engine compartment and wheel wells on opposite sides of the automobile 10. A right body rail l2c is also shown attached to the body panel l2a to increase body stiffness. There is a corresponding body rail on the left side for attaching to the body panel 12b.

The mounting units 18, 20 support the weight of the engine 14 and transmission 16 on the body 12. Consequently, the vibra-tional frequency components generated by the engine 14 are primarily transferred to the body 12 through the mounting units 18, 20 which vibrate or cause displacement of the body 12 in the vertical direction. In particular, the vibration force on the right mounting unit 18 is exerted on the right body rail 12c. For these reasons, each of the inertial mass shakers 34, 36 are located proximate to mounting units 18, 20, respectively, and are oriented to drive their respective inertial masses (not shown) in the vertical direction, so the body 12 can be inversely vibrated with respect to vibrational frequency components transferred from the engine 14 through the mounting units 18, 20. For the same reasons, each of the accelerometers 30, 32 are also located proximate to the mounting units 18 and 20, respectively, and are oriented to measure the acceleration or vibration of the body in the ver-tical direction.

The inertial mass shakers 34, 36, and the accelerometers 30, 32 are located as far as practical from any nodes associated with low frequency bending or beaming of the body 12.

In the automobile 10, the inertial mass shaker 34 is mounted to the body panel 12a approximately 20 cm in front of mount-ing unit 18, while inertial mass shaker 36 is mounted to the body panel 12b directly under the mounting unit 20. Acceler-ometer 30 is mounted to the body panel 12a via the body rail 12c, and accelerometer 32 is located on body panel 12b ap-proximately 10 cm above the mounting unit 20. The inertial mass shaker 34 and accelerometer 30 are then operatively as-sociated with mounting unit 18 for reducing the vibrational components transferred through it to the body 12, and like-wise for inertial mass shaker 36, accelerometer 32, and mounting unit 20.

The positioning of the components as illustrated in Fig. 2, avoids a body bending node that is located approximately 50 cm to the rear of the mounting units 18, 20. The mounting in-ertial mass shaker 34, which is approximately 20 cm in front of mounting unit 18, provides improved cancellation of body vibrations resulting from engine torque pulses transferred to the body 12 in different directions and from different loca-tions. For example, torque pulses are transferred to the body 12 through torque struts (not shown) that are connected hori-zontally between the body 12 and forward and aft positions on the engine 14 to reduce heaving, pitching and rolling move-ments of the engine 14 relative to the body 12.

The comfort of the passengers is improved by utilizing the vibration controller for compensating the vibration caused by the engine 10. Weight reduction of the automobile is also made possible because there is less dependency on mechanical low pass filtering with the aid of large masses or vibration dampers.

The POISON controller enables the automobile to generate in-verse vibration to compensate engine-generated vibrations.

This enables the vibration controller to compensate aging components that are mechanically coupled to the engine. Per- formance of the aging components is compensated by the shak-ers under the control of the vibration controller. The aging parts are worn or torn over the time due to friction, thermal shocks or corrosion during usage.

Fig. 3 illustrates vibration force patterns 41, 42, 43 of the engine 14 with respect to crankshaft angle. The vibration force patterns 41, 42, 43 are depicted in a two dimensional Cartesian coordinate system. A vertical axis 45 of the coor-dinate indicates force value in Newton, while a horizontal axis 46 of the coordinate indicates rotational angle position of the crankshaft in radian.

A first vibration force pattern 41 shows a vibration force pulsation that is measured at the right mounting unit 18 on the right panel 12a. The vibration force has substantially sinusoidal pulsating manner and corresponds to a basic exci- tation frequency of the combustion engine 14. The electrome- chanical inertial mass shakers 34, 36 do not exert any inver-sion vibrating force to the body 12 in producing the first vibration force pattern 41. The first vibration pattern 41 shows that the first vibration force pulsates in a cyclically repeated pattern according to the rotational angle of the crankshaft. Each complete cycle of rotation of the crankshaft is characterized by a 2ri period. Vibration force of the vi- bration force pattern 41 repeats itself by a in period of ro-tation about �65 N. A second engine vibration force pattern 42 shows the vibration force of the engine 14 on the right mounting unit 18 that is produced by the combustion engine 14 and the right actuator 34 (i.e. right electromechanical iner-tial mass shaker) . In the ideal case, as shown in Fig. 3, there is almost no vibration force found in the second vibra-tion force pattern 42.

A third engine vibration force pattern 43 shows a vibration force pulsation which is measured at the left mounting unit at the left panel 12b.

Fig. 4 shows the vibration controller having a POISON con-troller 50. The acronym "POISON" stands for "Periodic Online Iterative Signal Optimum Navigation". The provision of the POISON controller 50 leads to an improved suppression of pe-riodic disturbances and to an improved quality of control for periodic reference signals. The POISON controller 50 is also referred to as learning controller 50.

In the following, the term "reference signal" is also re- ferred to as a reference signal for an actuator or a refer-ence signal for a quantity. Such a quantity is given by, for example, a desired torque value, or a desired acceleration.

Signals will be named after the physical quantity they represent, e.g. as "torque reference signal" or "stored force correction signal".

Depending on the context, the word "cycle" can be used in the term "rotational cycle" where it refers to a repetition of combustion cycles which lead to periodic disturbances and hence to cycles of a periodic disturbance signal. On the oth-er hand, depending on the context, the word "cycle" can also be used in the term "computational cycle" where it refers to a repetition of computational steps.

The vibration controller 28 further comprises a digital to analog converter (DAC) 61, an analog to digital converter (ADC) 62 and further computation units that are described be-low. The DAC 61 has several input channels for reading in digitized control signals and also several output channels for sending analog control signals to a controlled system 35, 37. In the same way, the ADC 62 has several input channels for reading in analog actual value signals from the con-trolled systems 35, 37 and several output channels for output of digitized actual signals. The input channels of the ADC 62 are referred to as input channels of the vibration controller 28 and the output channels of the DAC 61 are referred to as the output channels of the vibration controller 28.

An input 51 of the POISON controller 50 is connected to an output channel of the ADC 62. The input channels of the ADC 62 are the inputs of the vibration controller 28. An output 52 of the POISON controller 50 is connected to an input 53 of an adder 54. A second input 55 of the adder 54 is connected to an output channel of the ADC 62. The output channels of the DAC 61 are the outputs of the vibration controller 28.

The POISON controller 50 further comprises memory sections containing -among others -correction signals and parameter settings for a controlled system. The controlled systems 35, 37 comprise the accelerometers 30, 32, respectively, the shakers 34, 36, respectively, the combustion engine 14 and other components which are mechanically coupled to the com-bustion engine 14 as, for example, the mounting unit 18. The controlled systems 35, 37 receive signals from the output of the vibration controller 28. The controlled systems 35, 37 send signals to the input of the vibration controller 28.

An output 56 of the adder 54 is connected to an input 57 of a system controller 58. The system controller 58 is a propor-tional-integral--derivative (PID) controller. An output 59 of the system controller 58 is connected to an input channel 60 of the DAC 61.

Outputs of the vibration controller 28 are connected to in- puts of the controlled systems 35, 37. Inputs of the vibra-tion controller 28 are connected to outputs of the controlled systems 35, 37.

For each channel that is connected to one of the controlled systems 35, 37 of Fig. 1 there is a separate POISON con-troller 50. They are connected to the DAC 61 and the ADC 62 in the same way as shown in Fig. 3. For reasons of clarity, in Fig. 4 this detail is shown for one controlled system only. A set of two parallel lines at the connections 25, 27 indicates that there are provided two output channels which connect the DAC 61 to the controlled systems 35, 37. Another set of two parallel lines at the connections 29, 31 indicates the provision of two input channels which connect the con-trolled systems 35, 37 to the ADC 62.

Figure 4 also illustrates functions of a POISON controller 50 as part of the vibration controller 28 in a control loop. The POISON controller 50 receives a digitized actual value signal from an output channel of the ADC 62. The POISON controller uses the stored correction signal, the reference signal and the digitized actual value signal to compute a new cor-rection signal. The stored correction signal is overwritten with the new correction signal. The POISON controller 50 gen-erates a corrected reference signal from the sum of the new correction signal and the reference signal and sends the cor- rected reference signal to the output 52 of the POISON con-troller 50.

The adder 54 receives the corrected reference signal from the POISON controller 50 and also a digitized actual value signal from one of the output channels of the ADC 62. The adder 54 then generates a control error signal by subtracting the digitized actual value signal from the corrected reference signal. The adder 54 sends the error signal to the system controller 58. The system controller 58 uses the control er-ror signal from the adder 54 to compute a control signal. The system controller 58 sends the control signal to an input channel of the DAC 61.

The DAC 61 converts the control signal into an analogue con-trol signal and sends the analogue control signal to an input of the controlled system. The controlled system generates a feedback signal. The controlled system sends the feedback signal back to the input 32 of the ADC 62.

Fig. 5 shows the components between the input 51 and the out- put 52 of the POISON controller 50 in further detail. Compo-nents of the POISON controller 50 comprise an inverting unit 80, an iteration filter 81 and an iteration memory 82.

An input 87 to the inverting unit 80 is connected to an out-put 84 of an adder 83. The adder 83 comprises input 51 and input 85. The input 85 of the adder 83 is connected to a ref-erence signal 86. The reference signal 86 is provided by the vibration controller 28. An output 88 of the inverting unit 80 is connected to an input 89 of an adder 90. A second input 91 of the adder 90 is connected to an output 98 of the memory 82. The memory 82 contains a stored correction signal.

An output 92 of the adder 90 is connected to an input 94 of the iteration filter 81. An output 95 of the iteration filter 81 is connected to an input 96 of the memory 82. The output 98 of the memory 82 is connected to a first input 99 of an adder 100. A second input 102 of the adder 100 is connected to the reference signal 86. The output 52 of the adder 100 is connected to an input 53 of the adder 54.

In the following, the operation of the POISON controller is explained in further detail.

POISON controller 50 also comprises a transformation unit which is not shown in Fig. 5. The transformation unit is lo-cated between an output channel of the ADC 62 and the input 51 of the adder 83. It transforms the time domain actual val- ue signal of ADC 62 into an angle-periodic angle domain sig- nal. The transformation unit also receives an additional ac-tual value signal of the current crankshaft angle position in order to do the transformation. The controller of Fig. 5 fur-ther comprises an input for reading in an actual angle value.

The transformation unit and the input for the actual angle value are not shown in Fig. 5. The terms "time domain signal" and "angle domain signal" refer to time dependent and angle dependent signals, respectively. Such signals can be rea-lized, for example, as time-force or angle-force value pairs or by attributing the meaning of time or angle, respectively, to a position in a data stream.

In the following, k' refers to a discrete angle index, k re-fers to a discrete time index and t refers to a continuous time. The symbols w[k], e[k], v[k], y[k] in Fig. 5 denote digital reference, error, correction and actual value sig-nals. The symbol y[t] denotes an analog actual value signal.

Symbols f[k'] and f[t] refer to the value of a function at discrete angle k' and continuous time t, respectively. For indices and parameters which refer to the angle domain, primed symbols are used, like for example k', in', P', I', D'.

For indices and parameters which refer to the time domain, the respective unprimed symbols are used.

For reasons of simplicity, all figures show the simplified time domain diagrams and the belonging unprimed time domain symbols. The transformation device and the belonging primed symbols are not shown.

Differences k'-l', k'-m' and k'-n' refer to an angle index within the reference signal "w" or in the cycle of the car-rection signal v. If a difference results in an angle index before the beginning of a cycle, the length of the cycle in discrete angle units is added to the difference.

POISON controller 50 also comprises a supervision layer which is not shown in Fig. 4. The supervision layer takes into the account the effects of under-and oversampling which may oc-cur due to a sampling with a constant sampling frequency in the time domain.

The effect of undersampling may appear at high crankshaft ro-tation speeds. Undersampling in this context means that in a given computational cycle the actual value signal skips in- dices k' in the angle domain. At a binary level, those indic- es k' correspond to discrete ADC quantization values. The su-pervision layer inside the POISON controller 50 makes sure, that no discrete angle index of the signals y[k'], w[k'], e[k'] and v[k'] can be skipped during undersampling. When un-dersarnpling, the supervision layer makes the POISON controler perform additional interleaving computational steps for each skipped index k'.

Oversampling, on the other hand, may occur at slow crankshaft rotation speeds. The above mentioned supervision layer inside the POISON controller 50 also makes sure, that all values in-side the POISON controller 50 remain constant as long as the actual value signal of the crankshaft angle remains at the same discrete value. Regarding over-and undersampling in this way ensures, that every single iteration memory position v[k'] is exactly updated once per cycle of a periodic distur- bance. Other methods to account for the over-and undersam-pling issues which are known, for example, in the context of order tracking analysis for rotating machinery may also be employed.

During the operation of the engine, the adder 100 receives an angle-shifted correction signal v[k'-l'J from the memory 82 and a reference signal w[k']. The reference signal w[k'] is generated by the output of a stored reference signal. The ad-der 100 generates a corrected reference signal by adding the signal v[k'-l'] to the reference signal w[k'] and sends the corrected reference signal to the output 52 of the POISON controller 50.

During the operation of the engine, the POISON controller 50 further receives a digitized actual value signal y[k] from one of the output channels of the ADC 62. Next, the transfor- mation unit uses an actual angle signal from a position sen-sor at a crankshaft to transform the time signal y[k] into an angle signal y[k'J. The position sensor at the crankshaft is not shown in the figures of the application. The adder 83 ge-nerates an error signal e[k'] by subtracting the digitized actual value signal y[k'] from the reference value signal w[k'].

The inverting unit 80 receives the error signal e[k'] and calculates a first output signal. The adder 90 receives the first output signal at the input 89 and a angle shifted cor-rection signal v[k'-m'] from the memory 82 at the input 91 and generates a second output signal by adding the first out-put signal to the signal v[k'-m']. The adder 90 sends the second output signal to the input 94 to the iteration filter 81. The iteration filter 81 calculates a corrected reference signal v[k']. The signal v[k'] is angle shifted by "-n'" in-dex positions and the resulting signal v[k'-n'] is stored in the memory 82.

During a subsequent cycle of a periodic disturbance signal, the adder 100 uses the stored signal v[k'-l'] in memory 82 to calculate a corrected reference value in the way described above.

The angle shift "1'" is used to compensate for an angle lag in the response of the controlled systems 35, 37 whereas the angle shifts "rn'" and "n'" compensate angle shifts that are introduced by the inverting unit 80 and the iteration filter 81. The inverting unit 80 and the iteration filter 81 use past signal values in their computations. Consequently, the result of the computations corresponds to an earlier angle (i.e. crankshaft rotation angle position) In a further embodiment, which is not shown in Fig. 4, an iteration memory and a buffer memory are provided for storing correction signals. During a cycle of a periodic dis-turbance signal, the iteration memory is overwritten with the correction signal v of the current cycle of the periodic dis- turbance signal and the other memory buffer holds the correc-tion signal v-1 of the last cycle of the periodic disturbance signal. After the iteration memory has been completely up-dated, the correction signal v is copied to the buffer memory and the iteration memory is overwritten again. In the case of an abnormal termination of a computational cycle, the last cycle v of the correction signal may be corrupted, whereas the previous cycle v can be recovered.

Fig. 6 illustrates the inverting unit 80 and the iteration filter 81 in further detail. The inverting unit 80 comprises an inverse system controller 130, which is also referred to as proportional-derivative lag (i.e. PDL or PD lag) control- ler, and a first moving average filter 134. The PD lag con- troller 130 is a special case of a PIDL (proportional-integral-derivative-lag) controller. The iteration filter 81 comprises a second moving average filter.

The output 84 of the adder 83 is connected to an input 87 of the PDL controller 130. An output 131 of the PDL controller is connected to an input 132 of the first moving average filter 134. An output 88 of the first moving average filter 134 is connected to a first input 89 of the adder 90. A sec-ond input 91 of the adder 90 is connected to an output 135 of a first back-shift element 136, which is in turn connected to the output 98 of the memory 82.

The output 92 of the adder 90 is connected to an input 94 of the second moving average filter 81. An output 95 of the sec-ond moving average filter 81 is connected to an input 138 to the second back-shift element 139. An output 140 of the sec-ond back-shift element 139 is connected to the input 96 of the memory 82.

Fig. 6 also illustrates the signal processing between the output 84 of the adder 83 and the input 99 of the adder 100 in further detail. The PDL controller 130, which will be ex-plained later, receives the error signal e[kJ from the adder 83 and generates an output signal. The output signal of the PDL controller 130 is made smooth by the moving average f li- ter 134. As mentioned in the description previously, the ad-der 90 adds the shifted correction signal v[k'-m'] of the last cycle of a periodic disturbance signal to the output signal of the moving average filter 134. The angle shift by -m' steps is symbolized by the shift element 136.

The moving average filter 81 receives the output signal of the adder 90 and generates an output signal at the output 95.

The output signal of the moving average filter 81 is shifted by -n' steps. This is symbolized by the shift element 139. The output signal of the shift element 139 is the correction sig-nal which will be used in a subsequent cycle of a periodic disturbance signal. As mentioned earlier, the memory 82 stores the correction signal v[k'-n'I.

Fig. 7 illustrates the composition of a proportional-integral-derivative lag (PIDL) controller which is used in the vibration controller 28. The system controller 58 shown in Fig. 3 is configured as PIDL controller. In the following, both expressions, PID and PIDL controller are used for the controllers 58. A PID lag controller without an integrator component will be referred to as a PDL controller. The in-verse system controller 130 of the inverting unit 80 of Fig. is configured as a PDL controller.

An input 146 to the PID lag controller is connected to an in- put 147 of a lag element 148. An output 149 of the lag ele-rnent 148 is connected to an input 150 of a multiplier 151.

The output 152 of the multiplier 151 is connected to a first input 153 of an adder 154, to an input 155 of a differenti-ator 156 and to an input 157 of an integrator 158. An output 159 of the differentiator 156 is connected to a second input of the adder 154. An output 161 of the integrator 158 is connected to a third input 162 of the adder 154. An output 163 of the adder 154 is connected to an input 164 of an out-put limiter 165. An output 166 of the output limiter 165 is connected to an output 167 of the PIDL controller.

A lag element 148 receives an error signal elk] via the input 147. The lag element 148 generates an averaged error signal ê[k] by computing a weighted sum from a current value e[kI and a previous value e(k-l] of the error signal e[kI. A weight factor L of the lag element 148 allows adjustment of the weighted sum.

The multiplier 151 receives the output signal ê[kI of the lag element 148 at the input 150 and multiplies the signal ê[k] by a factor P. The differentiator 156 receives the output signal of the multiplier 151, computes a time derivative of the signal ê[k] by a backward differentiation formula and multiplies the result by a parameter D. The integrator 158 receives the output signal of the multiplier 151, computes the integral over past values of its input signal by a nu-merical integration formula and multiplies the result by a factor I. The adder 154 generates an output signal at its output 163 by summing up the output signal of the multiplier 151, the out-put signal of the differentiator 156 and the output signal of the integrator 158. The output limiter 165 receives the out- put signal of the adder 154 at the input 164. The output lim- iter 165 limits the output signal of the adder 154 by an up-per limit and a lower limit and further sends the resulting signal u[k] to the output 167 of the PIDL controller. The output limiter 165 of the PIDL controller prevents numerical instability by integral windup. The parameters P, D and I al-low the adjustment of the relative contributions of the three input signals from the inputs 153, 160, 162 of the adder 154.

Fig. 7 shows a controller which works in the time domain.

Without loss of generality, the time-based operations inte-gration and differentiation can be directly transformed into angle-based operations when working in the angle domain. Es-pecially the inverting unit 80 of the POISON controller 50 may be implemented in the angle domain. This also may require transformed parameters I, D and L and also transformed moving average filter lengths.

Fig. B illustrates the signal processing units between the adder 100 and the DAC 61 of the POISON controller 50 in fur-ther detail. In addition to the signal processing units of Fig. 5, Fig. 8 shows two output limiters 170, 174 which are not shown in Fig. 5. The output 52 of the adder 100 of Fig. 5 is connected to an input 171 of the first output limiter 170.

The output 172 of the first output limiter 170 is connected to the input 53 of the adder 54. The output 59 of the system controller 58 is connected to an input 173 of the second out-put limiter 174. An output 175 of the second output limiter 174 is connected to an input channel of the DAC 61.

Fig. 8 also illustrates how a stored correction signal is used for generating a control signal for the controlled sys-tems 35, 37.

The adder 100 receives an angle shifted correction signal v[k'-l'] from the output 98 of the memory 82 of Fig. 5 at its first input 99. The adder 100 receives a reference signal w[k'] at its second input 102 and generates a corrected ref-erence signal at the output 52 by summing up the correction signal v[k'-l'] and the reference signal w[k']. The output limiter 170 limits the corrected reference signal by a lower limit and an upper limit and sends the output to the input 53 of the adder 54. The adder 54 receives an actual value signal y[k] at the input, sums up the input signals and sends the resulting signal to the system controller 58.

The system controller 58 computes a control signal and sends the result to the input 173 of the output limiter 174. The output limiter 174 limits the output signal of the system controller 58 to be with a predefined voltage range and sends the resulting signal to the input to the DAC 61. The DAC 61 converts the output signal of the output limiter 174 into an analog control signal and sends the converted analog control signal to one of the controlled systems 35, 37.

A whole series of other forces occur, however, which disturb the periodicity of the engine vibrations. For example mass forces which are caused by all types of acceleration e.g. by vehicle accelerations, by the condition of the road, by brak-ing processes or by steering movements. The POISON controller is especially suited to suppress disturbances which have pe-riod lengths of an integer fraction of a given period of time, angle or some other variable. It can be thus a useful option to use evaluation algorithms such as Fourier analysis with an adapted frequency filtration, inverse Fourier trans-formation, order analysis or cepstrum analysis, in order to identify the disturbances of the useful signal and to elimi-nate certain vibration signals from the input signal in order to speed up convergence of the POISON controller. This ap-plies especially to vibration signals which are non-periodic or which have a period length that is not an integer fraction of a given period of time or angle.

As mentioned previously, in addition to vibrations having frequencies which are multiples of a basic excitation fre-quency there are also other periodic vibrations present which are determined by certain resonance modes. Often, the period length of a basic excitation is not a multiple of the period length of the resonance modes. A Fourier analysis may be used to determine those resonance frequencies.

The signals may then be split up into a signal component which corresponds to the basic excitation frequency and other signal components which correspond to the resonance frequen- cies. The signal components are then transmitted over sepa-rate channels The POISON controller 50 then uses the separate channels in order to suppress the corresponding vibrations individually. Alternatively, the signal components corre-sponding to the resonances may be filtered out and suppressed by another type of controller which is used in combination with the POISON controller. For example, sine signals with appropriate phase shifts may be used.

Fig. 9 shows a flow diagram which illustrates a determination of an initial correction signal v[k].

After the start 228, in a first decision step 230, a decision is taken, if the initial correction signal v[kJ will be de-termined from prior data. If this is not the case, in step 231, the initial correction signal v[kJ is set to zero. In a second decision step 232, a decision is taken, if the stored correction signal v[k] will be computed. If this is not the case, in a step 233 the stored correction signal of an imme-diately preceding operation is initialized with a correction signal from a subsequent operation. If several correction signals from previous operations are available, the POISON controller 50 may use a correction signal from a previous op-eration with the best matching parameters.

If in the step 232 a decision has been taken to compute an initial correction signal, a third decision step 234 decides, if information from previous operations will be used. If this is not the case, in a step 235 an initial correction signal is computed which is based on parameters of the POISON con-troller 50. Otherwise, in a step 236, the POISON controller computes an initial correction signal, which is based on the controller parameters and on stored correction signals of previous operations. A step 237 symbolizes further steps that are useful for operating the vibration control system 1.

Fig. 10 shows the steps that are performed by the vibration controller during a computational cycle in further detail.

In a first step 240, the vibration controller 28 reads in the actual values for the controlled systems 35, 37 from the in-put channels of the ADC 62. In a next step 241, the vibration controller 28 reads in the reference value for the controlled systems 35, 37. In a computation step 242, the vibration con-troller 28 calculates a control signal u[k] according to the description of Fig. 6. In a next step 246, the vibration con-troller 28 calculates a new correction signal v[kJ, according to the description of Fig. 5. In a step 249, the output of the vibration controller 28 is clipped to be within a prede-fined voltage range. In step 250, the control values for the actuators 34, 36 are sent to the respective input channels of the DAC 61.

The moving average filters 81, 134 represent a special type of a finite impulse response (FIR) low pass filter. The band-widths of the moving average filters 81, 134 are adjustable parameters of the POISON controller 50.

The application provides the POISON controller 50 in control- ling inertial mass shakers to reduce motor vibrations by con- tinuous iterative learning of a pattern of a corrective vi-bration signal. The cycle or phase of the control signals of the POISON controller 50 are obtained by a sensor or multiple sensors. Correction signal for compensating vibration of the combustion engine 14 is a function of the sensor data, such as the crankshaft angle. The vibration pattern of the combus- tion engine 14 is a periodic function of the crankshaft an-gle.

The POISON controller 50 works online. Therefore, it is able to readjust itself constantly. The vibration controller 28 with the POISON controller 50 according to the present appli- cation can adapt to changes of controlled operations. There- fore, the improved vibration controller 28 is able to compen-sate vibrations, which are inherent to a combustion engine and are difficult to avoid.

As the POISON controller 50 adjusts itself during operations, there is no need for an iterative adjustment before the start of an operation. There are only a few parameters that users need to be adjusted in advance. Therefore, system identifica-tion is not required. However, the system identification or iterations to initially adjust the POISON controller 50 may be performed in advance, if desired.

Moreover, the signal processing algorithm of the POISON con-trailer 50 can be implemented by using only computations in the angle domain, thereby avoiding the Overhead for addi-tional transformations to the frequency domain and vice versa.

Furthermore, the POISON controller 50 acts as a feed forward controller during a cycle of a periodic input signal and is able to take corrective action before a control deviation oc-curs. This ability is due to the use of a stored correction signal from a previous cycle of a periodic input signal.

Moreover, the computation in the POISON controller 50 uses computationally efficient building blocks. This leads to a fast algorithm allowing for execution on a real time process-ing unit at high angle resolution.

As the POISON controller 50 is always online during the op- eration of the combustion engine, convergence is fast and ad-aptation to changed system conditions takes place from one cycle to the next.

The vibration controller 28 according to the application only needs a simple model of the controlled systems 35, 37 whose parameters remain fixed during the operation. The parameters may also be allowed to vary with time. The application avoids the difficulties of matching the parameters of an adaptive controller with a large number of degrees of freedom. These adaptive controllers cannot be applied easily.

A further advantage of the application is that no detailed knowledge of the controlled systems 35, 37 is required for adjusting the parameters of the controller, as it is the case with adaptive controllers. Once the parameters of the inverse system model have been determined, the vibration controller 28 will adjust itself during the operations. These filters, which are provided in an embodiment of the application, can easily be adapted with basic control theory knowledge for providing a convergent control strategy. Simple online tests can help to improve the function of these filters.

The active engine mount reduces the transferred vibrations from the engine. Thereby, the driving comfort is enhanced.

Also, the life expectancy of mechanical components is en-hanced are less affected by vibrations. This in turn enhances the safety of the passengers and the performance of the auto-mobile.

Due to the iterative improvement of the stored correction signal, an opposite vibration pattern pulsation with an ap-propriate phase shift develops automatically for all motor vibration that are multiples of the same basic frequency.

In the embodiment of Fig. 4, which uses a serial arrangement, the POISON controller 50 can be easily integrated into an ex-isting control loop of a vibration control system 1, simply by using the output signal of the POISON controller 50 as in- put signal to an existing system controller. The serial ar-rangernent of Fig. 4 has an additional advantage compared to a parallel arrangement of a learning controller as shown later in that it prevents the emergence of an undesired contribu- tion in the iteration memory which counteracts the integra-tion component of a PID system controller.

Correction signals of the POISON controller 50 are continu-ously checked so that changes in the vibration behavior of the engine 14 are detected. In case of a critical condition, which can be detected by exceptional amplitudes of the vibra-tion compensation, warning messages can be sent to the driver at an early stage. This enables the driver to send the auto-mobile to a repair shop before the automobile breaks down.

The POISON controller 50 enables the vibration control system 1 to compensate vibrations of the combustion engine 14. This makes it possible that an aged engine has a vibration behav-iour which is nearly as good as that of a new engine. In other words, the driver of the automobile avoids suffering from deteriorated performance of the combustion engine 14, which enhances the driving comfort.

In fact, regular maintenance can be reduced because the driver only needs to send the engine 14 for repair when the vibration controller 28 informs the driver that the vibra-tions of the combustion engine 14 have exceeded predetermined limits, which signifies a failure in the combustion engine 14. Therefore, the driver can avoid or at least reduce rou- tine, standard and costly maintenance by just following warn- ing signals generated by the vibration controller 28 for re-pair.

The POISON controller 50 according to the application com- prises a first learning controller input for receiving an ac-tual value signal. In the embodiment of Fig. 5, the first learning controller input corresponds to the input 51 of the adder 83. The actual value signal is derived from a con-trolled system, e.g. generated by a sensor of a controlled system 35, 37. A controlled system 35, 37 further comprises an actuator for applying a control signal and also all parts, which interact with the sensor and with the actuator.

The POISON controller 50 according to the application also comprises a second POISON controller input for receiving a reference signal. The second POISON controller input corre-sponds to the input 85 of the adder 83 of Fig. 5. The POISON controller 50 also comprises a POISON controller output. In the embodiment of Fig. 5, the output of the POISON controller corresponds to the output 52 of the adder 100.

The output signal of the POISON controller 50 is used as an input signal for a control unit. In the embodiment of Fig. 5, the control unit corresponds to the adder 54 and the control-ler 58. The control unit derives a second input signal from the actual value signal of the controlled systems 35, 37. In the embodiment of Fig. 5, the control unit reads in the sec-ond input signal from the input 55 of the adder 54.

An inverse system unit in the learning controller uses the deviation between the actual value signal and the reference signal to derive a first correction signal. In the embodiment of Fig. 5, the inverse system unit corresponds to the adder 83 and the inverting unit 80. The first correction signal corresponds to the output signal of the inverting unit 80.

A filtering unit uses a previously stored signal from an it-eration memory and the first correction signal for deriving a * 30 filtered correction signal. In the embodiment of Fig. 5, the filtering unit corresponds to the adder 90 and the iteration filter 81. The filtered correction signal is then stored in the iteration memory 82 for use in one of the subsequent cal-culation cycles.

The POISON controller 50 further comprises a correction sig-nal unit for deriving a correction output signal from the stored correction signal and from the reference signal. In the embodiment of Fig. 5, this correction signal unit corre-sponds to the adder 100. The correction output signal is the output signal of the POISON controller 50. It corresponds to the corrected reference signal at the output 52 of the adder in the embodiment of Fig. 5. The deriving of the correc-tion output signal can be done with analog means or with digital means. The expressions "computing" and "deriving" are not restricted to the calculation with a digital computer but they are also applicable to generation of the correction sig-nal with an analog circuit. Both options can be combined to use analog/digital means for computing the output correction signal.

A controller according to the application comprises one or more features of the aforementioned POISON controller 50 and of the aforementioned vibration controller 28. This is best seen in Fig. 4, which shows an embodiment of a vibration con- troller 28 that comprises a POISON controller 50. The con-troller derives an input signal from the actual value signal of the controlled system. The output signal of the controller is derived from the output signal of the control unit in the vibration controller 28.

In a broader sense, a control device according to the appli-cation can itself be made up of several control devices, each one performing a dedicated task. For example, the dual chan-nel vibration controller might be made out of two separate POISON controllers and of an additional cross coupling unit.

A cross coupling unit may be used at the input and/or at the output of the two separate POISON controllers. The cross cou- pling unit performs a linear transformation of the input sig-nals of the cross coupling unit the output signals of the cross coupling unit.

More than two vibration sensors and/or more than two inter-tial mass shakers may be used and they may be controlled by one ore more POISON controllers.

The POISON controller 50 is a learning controller, which may be designed in various ways. The arrangement of the POISON controller 50 according to Fig. 5 in which the system con-troller 58 derives its input from the output of the POISON controller 50 is called a serial arrangement. In the serial arrangement, the correction output signal of the POISON con-troller 50 is also referred to as first reference signal and the aforementioned reference signal is also referred to as second reference signal.

The embodiments can be carried out with other means which are adapted to the needs of the person skilled in the art. For instance, the shakers 34, 36 can also be any other form of force actuators, such as a hydraulic actuators. The shakers 34, 36 may also be mounted on any other part of an engine mount, which is also referred to as "mounting part". The four stroke four cylinder machine can also be a different type of combustion engine, a hybrid engine or an electric motor. The vibration sensors may also be any type of sensor from which an acceleration or force can be derived. Within this applica-tion the output signal of a vibration sensor is referred to as "actual force value signal". The vibration sensors may be provided at any part which is rigidly connected with a body 12 of a car, hereafter also referred to as "body part" of a car.

The position sensor at the crankshaft of the engine may also be any type of sensor from which a position can derived, for example a velocity sensor. The position sensor may also be attached to any shaft which is driven by the crankshaft of an engine, hereafter referred to as "output shaft" of the en-gine.

The computation of a control signal may be carried out in parallel for two or more of the controlled systems 35, 37.

For parallel computation, a scheduler of the vibration con-troller 28 of Fig. 1 attributes time slices to each parallel process. If the vibration controller 28 has several proces- sors, the time slices may be attributed to different proces-sors. A global memory section is used for the exchange of data between the control loops.

In a further embodiment, the controller may combine the ac-tual value signals y[k] of several cycles into one signal in order to get a higher time resolution of the actual value signals. This can be accomplished by shifting the sampling time interval relative to a previous cycle.

The actual value signals from the acceleration sensors 30 and 32 are dependent on each other. This may be accounted for by using cross coupled PID controllers. In this case, the input of a cross coupled PID controller is derived from one channel whilst the output of the cross coupled PID controller is added to the other channel. This technique may be used in the inverse system unit 80 and the PID controller 58.

The POISON controller 50 may also comprise an adaptive proce-dure which readjusts the parameters during the operation of the vibration control system 1. Alternatively, the vibration controller 28 may further comprise an adaptive controller.

Many known controllers work in the time domain because the response of the controlled system depends on time. In a ro-tating system, there are perturbations which are periodic with respect to the rotation time. Therefore it is advanta- geous to design a controller that operates in the angle do- main. Consequently, in the aforementioned embodiments, a sam- pling index k' is used which corresponds to an angle varia-ble.

However, the response of the controlled system not only de- pends on the angle but also directly on time. As a conse-quence, the optimal control parameters depend on the rotation velocity. Such a dependence exists already in the time do-main, as the system response in the time domain will also change with respect to rotation velocity. A learning control- ler according to the application can -at least partially -compensate for a changing system response by its learning feature. The convergence of the learning process may be slow-er if the control parameters are not chosen in an optimal way. Even so, the simplest approach to leave the control pa-rameters constant will be sufficient in many cases.

A simple approach to compensate for the introduced dependence on a rotation velocity o consist in multiplying the control parameters by factors proportional to o or 1/co. If the rota-tion is fast enough, the rotation velocity is essentially constant within one rotation period and can therefore be re-placed by a mean value. Instead of transforming the control parameters it is equally possible to transform the angle in-dex k' into the time domain by dividing the angle index k' by co before carrying out the calculations.

According to this approach, the control parameters are trans-formed with respect to the rotation velocity co as explained hereafter.

In the following, T0 is the sampling time which is equal to the computational time period of the vibration controller (T0 = 1 / f CONTRoLLER_SAMPLING_FREQUENCY), T is the time period of one crankshaft rotation (T = 2ri I co), S is the angle sen-sor resolution per rotation period (i.e. the amount of angle sensor digits per revolution). P, I, D, L, k, 1, m, n, the filter width FWIT of the iteration filter 81 and the filter width FWMA of the moving average filter 134 are parameters of the time domain POISON controller. To do the transforma- tion of those parameters, we assume almost steady state con- ditions, i.e. a crankshaft rotation speed which remains al-most constant from one rotation to the other.

In every computational vibration controller cycle, all actual value signals are measured. Furthermore, in every computa- tiorial vibration controller cycle, the actual crankshaft ro-tation time period is measured or estimated. This allows to transform all time domain controller parameters into their angle domain equivalents: P' = P as the P-gain is a dimensionless constant I' = I * T / S as I is a reciprocal time 0' = D * S / T as 0 is the differentiation time L' =L*SIT asListhelagtime k' = round(k * * S / T) time index becomes angle index 1' = round(l * T0 * S / T) time shift becomes angle shift round(m * T0 * S / T) time shift becomes angle shift n' = round(ri * T0 S / T) time shift becomes angle shift FWIT' round(FWIT * T0 * S / T) FWMA' = round(FW_MA * To * S / T) wherein "*" denotes multiplication, "I" denotes division, UU denotes a parameter in the angle domain and "round" denotes a rounding operation.

The abovementioned transformations with respect to the rota-tion velocity can be combined with an adaptive adjustment of the control parameters according to a system model. The system model may also depend on system parameters other than the rotation velocity (0 such as parameters which depend on the load on the system, for example.

When the iteration array contains angle dependent values v[k'] there is also a angle-to-time transformation required to determine the discrete time index k of the correction sig-nal u[k] at the input of the DAC converter 61.

With respect to the location of the transformation unit there are further alternatives. For example, the time-to-angle transformation and the angle-to-time transformation may also be performed at the time of reading and writing to the itera-tion array v(k'] and at the time of reading from an array w[k'] of reference values. In this case, the remaining compu- tations can be carried out in the time domain and the above-mentioned parameter transformations can be avoided.

A further alternative to speed up the convergence of a learn-ing controller according to the application is the use of multiple iteration arrays wherein each iteration array cor- responds to a different working point of the controlled sys-tern. The working point of the controlled system is determined by suitable system parameters, for example by the rotational velocity of the crankshaft or the opening of a throttle valve. For a given cycle, a subset of the iteration arrays is selected for reading and also a subset of the iteration ar-rays is selected for writing. The selections of the subsets depend on the system parameters. The use of a multiplicity of iteration arrays provides an improved starting point for the learning process and can therefore reduce the number of ite-rations to reach convergence.

In an alternative embodiment, the sampling index k corres-ponds to a time instead of an angle. In this case, the length of the correction signal v[kJ in terms of the sample index k is inversely proportional to the angular velocity of the crankshaft. To accommodate for this fact, the reading posi-tion k of the iteration array v[kJ and the reference signal w[k] may be adjusted accordingly. A possible approach is to use the index k*(u1Io1) instead of the index k as a reading position, where co3 is the mean rotation velocity during the i-th rotation period.

From the above it is understood that the vibration controller 28 of Fig. 4 further comprises an input for reading in an ac- tual angle value or an actual value of the rotational veloci- ty, respectively. This input is not shown in Fig. 4. The ac-tual angle value is used by the vibration controller 28 in the various possible ways described above. Prior to further processing, the actual angle value may also be grouped to- gether with a corresponding actual force value which is meas-ured at the same sampling time.

The addition of the reference signal to the output of the it- eration memory 82, as shown in Fig. 5, is typical for the Se-rial arrangement of a POISON controller.

In the embodiment of Fig. 6, for reasons of numerical stabil-ity, the inverse system controller 130 of the inverting unit is realized as PDL controller without integration compo-nent. However, an integration component may be used in the inverse system controller 130, if desired.

Instead of moving average filters, general finite impulse re-sponse (FIR) low pass filters may be used for the iteration filter 81 in the filtering unit and the filter 134 in the in- verse system unit. The iteration filter 81 may also be ar-ranged between the output 98 of the memory 82 and the input 91 to the adder 90. There may also be an additional anti-aliasing filter between the input of the vibration controller 28 and the ADC converter 62. Further, the two adders 54 and of Fig. 5 may be combined into a single adder having three inputs.

The channels provided for communication between various corn-ponents of the vibration control system 1 can be implemented by electric cable runs, which is effective and reliable.

Additional system identification based on operation signals or initial iteration may be performed preceding to an opera-tion of the vibration control system 1.

In a further embodiment, a controller according to the appli-cation may also comprise a POISON controller and a control 3.7 unit in a parallel arrangement, as shown later. A controller with the parallel arrangement further comprises a correction signal unit. The correction signal unit derives an external control signal from the correction output signal of the PCI-SON controller and from an internal control signal which is derived from the control unit. In the later embodiment of the parallel arrangement, the correction signal unit corresponds to two adders 54'' and 54''', respectively.

In Fig. 14, the control unit corresponds to the adder 83''' and the controller 58'''. In Fig. 14, the inverse system unit corresponds to the adder 83''' and the inverting unit 80''', whereas in Fig. 13 the inverse system unit corresponds to the adder 83'', the ADC 62'' and the inverting unit 80''.

En the Fig. 15 the filtering unit corresponds to the adder 90'' and an iteration filter which is not shown. This itera-tion filter could be arranged between the output 98'' of the iteration memory 82'' and the input 91'' of the adder 90''.

It could also be arranged between the adder 90'' and the in-put 96'' of the iteration memory 82''.

A method for compensating a vibration of the combustion en- gine 14 according to the application is disclosed. During op-eration, a control signal is derived from a reference signal, from an actual value signal and from a stored correction sig-nal. The control signal is used for actuating the engine 14 according to the control signal. In the embodiment of Fig. 1, this is accomplished by sending a control signal to each ac-tuator 34, 36. The actuators act on the engine through the mounting units 18, 20.

In the serial arrangement of Fig. 5, the reference signal oc- curs at the input 102 of the adder 100, the actual value sig-nal occurs at the input 55 of the adder 54 and the stored correction signal occurs at the input 99 of the adder 100. In the parallel arrangement of Fig. 14, the reference signal oc-curs at the input of the adder 83''', the actual value signal occurs at a second input 51''' of the adder 83'''and the stored signal occurs at a second input 53''' of the adder 54',, The stored signal is in turn derived from a reference signal, an actual value signal and a previously stored correction signal. In the serial arrangement, as shown in Fig. 5, in an intermediate step, a corrected reference signal is derived from the stored correction signal and the reference signal and, in a further step, the control signal is derived from the corrected reference signal and the actual value signal.

The deriving of the stored correction signal comprises the deriving of a first correction signal. In the embodiment of Fig. 5, this is accomplished by the adder 83 and the invert-ing unit 80. A further part of the method of the application is the deriving of a second correction signal. In the embodi- ment of Fig. 5, this is accomplished by the adder 90. Filter-ing the first correction signal and storing the correction signal for later use as a stored correction signal is also part of the method of the application. In the embodiment of Fig. 5, this is accomplished by the filtering unit 81 and the iteration memory 82.

The deriving of the first correction signal may further com- prise deriving a difference signal from the actual value sig-nal and the reference signal. In the embodiment of Fig. 5, this is accomplished by the adder 83. The deriving of the first correction signal may also comprise the computation of a derivative of the difference signal and the computation of a weighted sum of the difference signal and the derivative of the difference signal. In the embodiment of Fig. 6, this is accomplished by the P component and the D component of the PDL controller 130. In Fig. 7, these steps are accomplished by the multiplier 151, the differentiator 156 and the adder 154.

The deriving of the control signal from the corrected refer-ence signal and the actual value signal may further comprise the computation of an integral and of a derivative of the corrected reference signal and the computation of a weighted sum from the derivative and the integral of the corrected reference signal and the corrected reference signal. In the embodiment of Fig. 8, this is accomplished by the PID con-troller 58. In Fig. 7, these steps are accomplished by the multiplier 151, the differentiator 156, the integrator 158 and the adder 154. In place of a PID controller a PD control-ler without integration component may also be used and the PD or PID controller may also comprise a lag component, as shown in Fig. 7.

The method may also comprise phase compensation steps. The phase compensation may be used in conjunction with any unit that uses past values of an input signal for the computation of an output signal and thereby introduces a phase lag. In the embodiment of Fig. 6, the phase compensation is accom- pushed by the backshift elements 139 and 136. The phase com- pensation of the backshift elements 139 and 136 is accom-plished by a cyclic backshift operation which is explained in connection with the operation.

During the first loop of the operation, the stored correction signal is taken from initial values. As shown in connection with Fig. 9, there are several possibilities to generate such initial values.

An update of the stored correction signal may takeplace each crankshaft angle position when a new sample of an actual value signal is generated. The correction signal may also be derived from multiple signal values at multiple angle posi-tions k', from signal values of one operation of one complete period 1 ri or even from signal values of multiple operations of multiple periods (e. g. fl*rI). In other words, instead of taking one value from the stored correction signal, a new correction signal value is computed from several values of the stored correction signal values of multiple cycles. In the latter case, a trend over multiple cycles of the correc-tion signal may be derived to speed up convergence. It is also possible to calculate several values of the control sig-nal in one computation step, in order to cope with high-speed requirements of the vibration control system 1.

The embodiment of Fig. 1 also discloses a vibration control system according to the application which comprises at least one controlled system and wherein the controlled system has at least one actuator and at least one sensor unit and has at least one controller. The at least one acceleration sensor unit is provided at the engine. This is understood to be any suitable place in the automobile for obtaining a motor vibra-tion value, which is the engine mount in the embodiment of Fig. 1. Likewise, an actuator for actuating on the engine is understood to act on a part of the engine or on a part of the automobile connected to the engine. For example in the em-bodirnent of Fig. 1, the at least one actuator and the at least one sensor unit may correspond to the right actuator 34 and the right accelerometer 30.

Another embodiment according to the application comprises at least two controlled systems wherein each of the controlled systems comprises a acceleration sensor and an actuator. In the embodiment of Fig. 1, the acceleration sensors correspond to the accelerometers 30, 32 and the acuators correspond to the shakers 34, 36.

In Figs. 11-14, components with similar functions have the same reference numbers as the components in the aforemen- tioned figures and prime symbols have been added to demon-strate this.

Fig. 11 shows another embodiment in which the POISON control-ler 50 of Fig. 4 is realized on a separate digital controller which controls a control loop with an analog adder 54' and an analog PID controller. Like parts have been given like refer- ence numbers. This embodiment can be combined with the vibra- tion control system 1 of Fig. 1 as well as with the other em-bodiments.

The parts inside the digital controller 50' of Fig. 11 are as shown in Fig. 5. With reference to Fig. 5, in the embodiment of Fig. 11 the output 52 of the adder 100 is connected to an input channel of a DAC 61'. The input 51 to the adder 83 is connected to an output channel of an ADC 62'. The adder 54' and the system controller 58' are not part of a POISON con-troller. They are realized as separate analog components 54', 58' In Fig. 11, a first input 53' to an analog adder 54' is con-nected to an output channel of the DAC 61'. A second input 55' to the analog adder 54' is connected to a controlled sys-tem 35'. An output 56' of the analog adder 54' is connected to an input 57' to an analog system controller 58'. An output of the analog system controller 58' is connected to an input of the controlled system 35'.

Similar to Fig. 5, there is one adder 54' and one system con-troller 58' for each of the controlled systems 35, 37. The components inside the vibration controller 28, which are shown in Fig. 5, are realized for each of the controlled sys-tems 35', 37' as in Figs. 4 and 5. The same applies to the corresponding connections.

Fig. 12 shows a further embodiment of a controller according to the application. The controller 58'' of Fig. 12 represents any type of analog controller. The parts between the adder 83' and the adder 100' of Fig. 12 are similar to the parts between the adder 83 and the adder 100 of Fig. 5. Unlike in the controller of Fig. 5, there is no iteration filter in Fig 12 and the inverting unit 80 of Fig. 5 is replaced by a gen-eral digital filter 80'. The output value of the general digital filter 80' is given by a sum of a linear combination of present and past values of the input values of the general digital filter 80' and a linear combination of present and past values of the output signal of the general digital fil-ter 80' Figs. 13 and 14 show two further embodiments of a controller according to the application. Unlike in the previously shown embodiments the POISON controller is used in a parallel ar-rangernent. In the parallel arrangement, the correction signal is added to the output signal of a system controller and not to the input signal. In a serial arrangement, as in Fig. 5, the correction signal is added to the input signal of a sys- tem controller 58. The parallel arrangement is easier to im- plement if the system controller is part of a digital con- troller, as in Fig. 14, because in this case it is not neces-sary to insert an analog adder between the system controller and the controlled system. The serial arrangement of Fig. 4 has the advantage over the parallel arrangement of Figs. 13 & 14 that it can be used with an existing controller 58 without the need to reconfigure the existing controller 58.

In the parallel arrangement of Fig. 14 an adder 54''' is pro-vided for adding the stored correction signal to the output of the controller 58'''.

Fig. 14 shows a further embodiment of a controller according to the application which is similar to the embodiment shown in Fig. 13 but in which the controller 58''' is a part of a digital controller 27'''.

The controllers 27'', 27''' of Fig. 13 or Fig. 14 may also comprise an iteration filter between the adder 90'', 90''' and the input 96'', 96''' of the iteration memory 82'', 82''' or between the output 98'', 98''' of the iteration memory 82'', 82''' and the adder 90'', 90'''.

Referring now back to Fig. 1, the action of the shaker 34 will also have an influence on the acceleration sensor 32 and the action of the shaker 36 will also have an influence on the acceleration sensor 30. To obtain a stable control system it is therefore desirable to decouple the two control loops which are defined by the shaker 34 and the acceleration sen-sor 30 and by the shaker 36 and the acceleration sensor 32, respectively. The parameters for the optimal decoupling of the two control loops will further depend on the operating condition of the motor.

For decoupling the control loops, in a further embodiment, the vibration controller 28 further comprises an input de-coupling unit and an output decoupling unit. Now referring back to Fig. 4, the input decoupling unit is placed between the two output channels of the ADC 62 and the inputs 51 and 55. As explained above, there are separate POISON controllers for each of the two control loops of which only one is shown in Fig. 4. Therefore there are also two inputs 51 and two in-puts 55. The input decoupling unit comprises two inputs and two outputs. The two inputs of the input decoupling unit are connected to the two of the output channels of the ADC 62 which correspond to the two controlled systems 35, 37. Each of the two outputs of the input decoupling unit is connected to the two inputs 51, 55 which correspond to a respective control loop.

The output decoupling unit is placed between the outputs 59 of the respective PID controllers 58 and the two input chan-nels of the DAC 61. The output decoupling unit also has two input and two outputs. The two outputs of the output decoup-ling unit are connected to the two of the input channels of the ADC 62 which correspond to the two controlled systems 35, 37. Each of the two inputs of the output decoupling unit is connected to one of the outputs 59 which correspond to a re-spective control loop.

During operation, the input decoupling unit performs a linear transformation from the input signals to the output signals.

The linear transformation can be defined bya multiplication with a 2x2 input matrix which has four input matrix parame-ters. The input decoupling unit receives further actual value signals from further output channels of the ADC 62. Depending on the further actual value signals, the input decoupling unit determines the four input matrix parameters based on a lookup table. The values of the lookup table may be preset by a previous measurement of the system response. In a more so-phisticated alternative, the degree of coupling between the two controlled systems 35, 37 is determined online, for exam-pie by computing a cross-correlation. Based on the degree of coupling, the input decoupling unit determines the input ma-trix parameters.

Likewise, the output decoupling unit performs a linear trans-formation from the input signals of the output decoupling unit to the output signals of the output decoupling unit.

Again, the linear transformation can be defined by a multi- plication with a 2x2 output matrix which has four output ma- trix parameters. The output decoupling unit adjusts the out- put matrix parameter according to one of the methods ex-plained in conjunction with the input decoupling unit.

Claims

CLAIMS1. Vibration control system for an engine, the vibration control system comprising: -a force actuator for acting upon a mounting part of the engine according to a force control signal, -at least one position sensor for deriving an actual motion value signal of the motion of an output shaft of the engine, -at least one vibration sensor for deriving an actual force value signal of a vibration of the engine, -a control device for receiving the actual motion value signal and the actual force value signal, for computing the force control signal, for outputting the force con-trol signal to the force actuator, the control device comprising -a control unit for computing the force control signal from a force reference signal, from an ac-tual force value signal of the vibration sensor, from an actual motion value signal of the position sensor and from a stored force correction signal, and -an inverting unit for computing a first force correction signal from the actual force value sig-nal and from the force reference signal, -a filtering unit for computing a second force correction signal from the first force correction signal and from the stored force correction signal, -an iteration memory for storing the second force correction signal as a stored force correction sig-nal for later use.
2. Vibration control system according to claim 1, wherein the control unit comprises a correction signal unit for computing a corrected force reference signal from the force reference signal and from the stored force correc-tion signal, and wherein the control unit comprises a controller for computing the force control signal from the corrected force reference signal and from the actual force value signal.
3. Vibration control system according to claim 1, wherein the control unit comprises a controller for computing an internal force control signal from the force reference value signal and from the actual force value signal, and wherein the control unit further comprises a control signal unit for computing the force control signal from the stored force correction signal and from the internal force control signal.
4. Vibration control system according to claim 1, wherein the filtering unit comprises a low pass filter.
5. Vehicle with a vibration control system, the vibration control system comprising -at least one force actuator for actuating upon a mounting part of an engine according to a force control signal, -at least one position sensor at an output shaft of the engine for deriving at least one actual motion value signal, -at least one vibration sensor for deriving an actual force value signal of a vibration of the engine, -a controller for receiving the actual motion value signal and the actual force value signal arid for deriv-ing the force control signal and outputting the force control signal to the force actuator, the controller comprising -a control unit for deriving the force control signal from a reference signal, from the actual mo- tion value signal, from the actual force value sig-nal and from a stored correction signal, -an inverting unit for deriving a first correction signal from the actual force value signal and from the reference signal, -a filtering unit for deriving a second correction signal from the first correction signal and from the stored correction signal, -an iteration memory for storing the second cor-rection signal as a stored correction signal for later use.
6. Vehicle according to claim 5, wherein the filtering unit comprises a low pass filter and the inverting unit com-prises at least a proportional component and at least a derivative component of a PDL controller.
7. Method comprising -providing a force actuator at a mounting part of an engine of a car, -computing a force control signal from a reference sig-nal, from an actual force value signal of a vibration sensor at a body part of the car, from an actual motion value signal of a position sensor at an output shaft of the combustion engine and from a stored correction sig-nal, -actuating the force actuator according to the force control signal, -computing a first correction signal as a difference of the actual force value signal and of the reference sig-nal, computing a second correction signal as a sum of the first correction signal and of the stored correction signal, -filtering the second correction signal, -storing the second correction signal as a stored cor-rection signal for later use.
8. Method according to claim 7, further comprising -deriving of a corrected reference signal by adding the reference signal to the stored correction signal, where-in the force control signal is derived from a corrected reference signal and from the actual force value signal.
9. Method according to claim 7, wherein the computing of the first correction signal comprises -computing a difference signal from the actual force value signal and from the reference signal, -computing a derivative of the difference signal, -computing a weighted sum from the derivative of the difference signal and from the difference signal.
10. Method according to claim 7, wherein the filtering of the second correction signal comprises filtering with a low pass filter.
11. Method according to claim 7, wherein -deriving of the force control signal comprises reading out an iteration memory and the reading out of the ite-ration memory comprises determining a read position by a cyclic backshift operation, -deriving of the second correction signal comprises reading out the iteration memory and the reading out of the iteration memory comprises determining a second read position by a second cyclic backshift operation, -storing of the second correction signal comprises writing to the iteration memory and the writing to the iteration memory comprises determining a write position by a third cyclic backshift operation.
12. Method comprising -providing a force actuator at a mounting part of an engine of a car, -deriving a force control signal from a reference sig-nal, from an actual force value signal of a vibration sensor at a body part of the car, -actuating the force actuator according to the force control signal, -deriving a first correction signal from the actual force value signal and from the reference signal, -deriving a second correction signal from the first correction signal and from the stored correction signal, -filtering the second correction signal, -storing the second correction signal as a stored cor-rection signal for later use.
13. Method according to claim 12, further comprising -deriving of a corrected reference signal from the ref-erence signal and from the stored correction signal, wherein the force control signal is derived from the corrected reference signal and from the actual force value signal.
14. Method according to claim 12, further comprising -transforming a time domain signal into an angle domain signal.
15. Method according to claim 12, wherein the computing of the force control signal comprises reading of the stored correction signal from one or more of a multiplicity of iteration arrays and wherein the storing of the second correction signal for later use comprises writing to one or more of the mul-tiplicity of iteration arrays and wherein the selection of the iteration arrays being read from is dependent on the actual motion value signal and the selection of the iteration arrays being written to is dependent on one or more parameters which character-ize an operating condition of the combustion engine.