WO2006042351A2

WO2006042351A2 - Active vibration reduction in rail vehicles

Info

Publication number: WO2006042351A2
Application number: PCT/AT2005/000412
Authority: WO
Inventors: Anton Stribersky; Martin Kollmann; Gerhard Schandl; Martin Kozek
Original assignee: Siemens Transportation Systems Gmbh & Co Kg
Priority date: 2004-10-21
Filing date: 2005-10-17
Publication date: 2006-04-27
Also published as: WO2006042351A3; AT502141A1

Abstract

The invention relates to a method for active vibration reduction in a rail vehicle (UBW) using a control system that comprises a number of sensors (AS1-AS6) for measuring deformations of the carriage body of the rail vehicle and a number of actuators (AS1-AS6), which are attached exclusively to the carriage body and generate actuating forces and/or actuating torques on the carriage body. A control device is configured to receive measurement signals from the sensors, derive characteristic variables from the signals, said variables relating to elastic deformations in terms of the vibrational eigenmode of the carriage body, calculate a set of actuating variables using these characteristic variables and supply actuating signals that correspond to the actuating variables to the actuators.

Description

ACTIVE VIBRATION REDUCTION ON RAIL VEHICLES

The invention relates to the active vibration reduction on a rail vehicle, starting from a control system for active vibration reduction on a Schienenfahr¬ convincing, with a number of sensors for measuring deformations of Schienenfahr¬ zeugs and a number of actuators for generating restoring forces and / or Stellmomen¬ th on the rail vehicle and a control device which is adapted to a) receive measurement signals from the sensors, b) from the sensor measurement signals to calculate a set of manipulated variables, and c) the manipulated variables corresponding control signals to the actuators.

Due to the consistent lightweight constructions in rail vehicle construction Eigen¬ the frequencies of a modern car body in the fully developed state are very low and close to the excitation frequency of the chassis, in most cases, the range of about 5 - 10 Hz, ie that range is relevant for the comfort perception of the passengers. The situation is often aggravated by the fact that the carbody construction becomes very complex with respect to the vibration behavior due to a large number of doors and windows, which act as structural weakening.

To increase the natural frequencies of the car body (eg to more than 10 Hz), so Anre¬ supply from the chassis and natural frequency does not resonate, local stiffeners are mounted in the structure in the classic rail vehicle, which causes an increase in vehicle mass and costly production and their effect to Eigen¬ frequency is often only slightly.

In order to improve the vibration comfort of rail vehicles with the lowest possible mass increase, the use of active vibration reduction measures offers itself. One possibility here is the reduction of the excitation acting on the car body by the use of active elements in the secondary spring stage, ie between Wagen¬ box and bogie frame, or in the primary spring stage. Such approaches are, for example, E. Foo, RM Goodall: Active Suspension Control of flexible-bodied railway vehicles using electro-hydraulic and electro-magnetic actuators, Control Engineering Practices 8 (2000) 507-518, and RM Goodall, W. Kortüm, 'Mechatronic developments for railway vehicles of the future', Control Engineering Practice 10 (2002) 887-898. However, the interventions concerning the chassis are to be regarded as critical from the safety point of view, since in the case of a system failure the suspension derailment safety is directly and abruptly - -

reduced. This approach should therefore be classified as safety-relevant and should be pursued with the extensive measures provided for this purpose.

A vibration suppression on the car body is described by J. Hanson et al., "Vibration Suppression of Railway Car Body with Piezoelectric Elements", International Symposium on Speed-up and Service Technology for Railway and Maglev Systems 2003 (STECH '03), 19-22.8.2003 Tokyo (JAPAN). By attaching piezoelectric elements damping of individual bending modes could be achieved. However, this is a passive vibration damping, which differs in principle from the considered here active vibration damping and completely different requirements for interpretation and placement of the damping elements.

Numerous contributions to active vibration damping for noise reduction are also known in the field of passenger cars, see, for example, M. Strassberger, H. Wallner, "Active Noise Reduction by Structural Control Using Piezoelectric Actuators", Mechatronics 10 (2000) 851-868 ; and C.K. Song et al., Active Vibration Control for Structural-acoustic Coupling System of a 3-D Vehicle Cabin Model, Journal of Sound and Vibration 267 (4) (2003) 851-865. However, due to the situation in passenger cars, those approaches to vibration reduction are not readily transferable to rail vehicles, as there is a completely different dimension and geometry of the vehicle and a different vibration characteristic (plate vibrations), here the number of possible ones Vibration types are much less manageable; In addition, the emphasis in the passenger car is on the reduction of Geräuschschwin¬ conditions (acoustic type) or a single specific fundamental vibration (damping the torsion in a convertible).

As a promising approach to vibration reduction, methods using modal models for controller design are known; see the articles D. HaHm, SO Reza Moheimani, Spatial Resonant Control of Flexible Structures Applica- tion to a Piezoelectric Laminate Beam, IEEE Trans. Control Syst. Techn. 9 (2001) 37-53 and S. Leleu et al., 'Piezoelectric Actuators and Sensors Location for Active Control of Flexible Structures', IEEE Trans. Instrum. Meas. 50 (2001) 1577-1582; The latter article also describes an analytical method for the placement of sensors and actuators, which, however, remains manageable only with simple structures. In these applications, however, simple, one- or two-dimensional systems have always been investigated, so that it was by no means certain how such methods can be successfully used in rail vehicles in which the structures are three-dimensional and, moreover, very complex. - -

In general, it should be pointed out at this point that the mentioned known approaches, in particular those for passenger cars, specifically target specific, simple vibration modes. Complex vibrations, even more so a superimposition of different types of vibration (for example, bending and torsional vibrations), have not been sufficiently considered.

The vibration reduction considered here on a specific rail vehicle, in particular a railway carriage for passenger transport, results in a considerably more difficult problem than in the various solutions of the state of the art mentioned above. The treated structure is three-dimensional and has a very inhomogeneous cross-section, which also has a large number of "holes" (doors, windows), which in particular means that the eigenmodes are considerably more complex than in previous applications the placement of sensors and actuators can not be calculated by conventional analytical (including analytic-numerical) methods due to the complicated eigenmodes and because of the structural limitations, not least in previous applications substantially smaller and more flexible structures have been damped; namely, in a rail vehicle, the forces and ways on the actuators are comparatively much higher.

It is therefore the object of the present invention to find a way of reducing vibration in rail vehicles by which, while maintaining the desired comfort improvement for the passengers, the safety problem associated with the chassis can be avoided.

This object is achieved by a control system of the type mentioned, in which according to the invention the actuators are mounted exclusively on attachment points of the car body of the rail vehicle and the control device is adapted to derive in step b) from the sensor measurement signals characteristic quantities related to elastic deformations refer to the vibration eigenmodes of the car body, and to calculate the manipulated variables based on these characteristic variables.

It should be noted here that the term "deformation" includes not only strains but also curvatures and torsional deformations (twisting).

This solution according to the invention comprises several aspects. On the one hand, an active vibration damping is completely shifted to the car body structure, thus independently of the bogie of the rail vehicle. This approach to mechatronic comfort improvement directly on the elastic vehicle structure avoids the safety - -

problematic, which occurs in connection with a vibration reduction on the chassis, since in a system failure of the vibration reduction, the derailment safety of the chassis is not affected - only the vibration comfort for the passengers deteriorates.

The displacement of the vibration reduction on the car body, instead of a Schwin¬ tion damping of the chassis or between the chassis and car body, brings problems of their own kind. This is in particular the fact that (as already explained above) now the car body as a vibratory structure, especially a Continuous system, must be considered and thus the excited via the chassis oscillations in the car body can be very diverse. In particular, now different types of modes - bending vibrations, torsions, Querschnittsverziehung - to detect that occur in superposition.

On the other hand, the invention therefore provides for a modal reduction, namely for calculating the manipulated variables from the sensor measurement signals by expressing these variables by means of characteristic quantities which relate respectively to vibration eigenmodes (modes) of the car body, namely only those modes of the lower frequencies must be taken into account expediently become.

Overall, the control system according to the invention manages the regulation of an elastic continuum in the dimensions of a rail vehicle, i. with a length of 20m or above. As a result, the oscillation reduction also succeeds for complex excitations which contain superimpositions of different types of modes, and also the consideration of the large changes in the model parameters of local rail vehicles in the design of the controller caused by different loading states is possible.

Here it has been shown that it - in addition to the associated simplification of Fahr¬ factory construction - usually sufficient, even if the sensors are arranged exclusively on the car body. Furthermore, if at least some of the actuators are additionally used as sensors, this double utilization results in a considerable simplification of the expenditure of actuator and sensor elements.

As actuators, different types of actuators can be used, eg hydraulic actuators; Various types are also conceivable for the sensors, eg strain gauges. In a preferred embodiment of the invention, however, the actuators can be realized as piezoelectric actuators, so that as a result of their short reaction times and low dead times - -

(in a few ms) vibrations in a wider frequency range can be treated. In particular, stacked piezoelectric actuators are suitable for coping with the large forces and paths required. For the same reasons, it is advantageous if the sensors are implemented as piezo sensors.

Of particular importance is of course the choice of bodies on the car body, where the actuators are located. In order to be able to act in a targeted manner on the vibration eigenmodes, the actuators can advantageously be attached to the vehicle body in the vicinity of the maximum deformation of a vibration eigenform. In this case, it is particularly favorable if the actuators are mounted on the car body in the vicinity of the maximum deviation in curvature of a vibration eigenform; Curvature deviation is understood to mean the difference in curvature in comparison to the corresponding oscillation eigenform of a standard car body (simple box shape without recesses for doors, windows, etc.). When selecting the actuator type, only those vibration eigenmodes that are located in the lower frequency range must be taken into account. Which forms of vibration these are can be determined from a simulation. The number of vibration eigenmodes considered can, for example, be determined in such a way that the lower N vibration eigenmodes are used when N actuators are used, or the vibration eigenfors below a cutoff frequency are selected; as cut-off frequency, here e.g. 8-10 Hz are used.

Another rule for selecting the type of actuator is that the actuators are mounted on the car body in the vicinity of a structure weakening relevant to the shape of a vibration eigenform. These are to be understood as structural weakenings such as doors or windows in the vehicle body, by means of which the shape of a vibration shape is changed (without being compensated for by compensating measures, such as a reinforcing element, such as a reinforcing frame). The environment is e.g. an area around the respectively considered structural weakening (or model deformation) which corresponds to half the extent of the structural weakening (modal deformation) (effective radius).

Preferably, the characteristic quantities represent a state vector relating to the deformation state with respect to the vehicle body's inherent vibration modes; such a state vector may include both displacements and their temporal changes (velocities) which can provide a complete description of the instantaneous state of deformation, at least concerns the lower vibrational forms. - -

A particularly favorable variant of the invention consists in that the control device calculates the characteristic variables exclusively from the sensor measurement signals and predefinable set values according to a predetermined algorithm. In another advantageous variant, the control device can derive the characteristic quantities by means of an observer estimate. In this case, the observer estimate can be realized as an LQ observer who is characterized by particular robustness.

The invention is particularly suitable for a rail vehicle with a number of sensors for measuring deformations of the rail vehicle and a number of actuators for generating restoring forces and / or adjusting torques on the rail vehicle and with a control system according to the invention for active vibration reduction.

The above object is likewise achieved by a control device for active vibration reduction in a rail vehicle, which is set up in cooperation with a number of sensors for measuring deformations of the rail vehicle and a number of actuators for generating restoring forces and / or Stell¬ moments on the rail vehicle a) receive measurement signals from the sensors, b) from the sensor measurement signals to calculate a set of manipulated variables, and c) the control variables corresponding control signals to the actuators, the control device, based on the assumption that the Actuators are aus¬ finally attached to attachment points of the car body of the rail vehicle, is also arranged to derive in step b) from the sensor measurement signals Sizes that relate to elastic deformations in terms of Schwingungseigen¬ the car body, and on the basis of this char akteristischen sizes to calculate the manipulated variables.

Advantageous developments of the control device according to the invention have been discussed in connection with the control system according to the invention.

The invention together with further advantages will be explained in more detail below with reference to a non-limiting embodiment, which is illustrated in the accompanying drawings. The drawings show

Big. 1 shows a metro car with a control system according to the invention;

Fig. 2 the attachment of actuators on the car body of the carriage of Fig. 1; 3 shows a mounting bracket of an actuator. - -

4 shows the first vertical bending shape as an example of a vibration eigenfig; and

5 shows the structure of the control loop.

The embodiment relates to the active vibration reduction on a subway car. Since the implementation of the vibration reduction, including the determination of the control parameters and the location of the actuators and sensors, based on an analysis of Schwin¬ own geodesic modes of the car using a simulation and the achievable by the invention improvement of the vibration characteristics was studied in detailed simulations, are in The following also describes the vibration simulation together with the essential underlying simulation models.

1. Overview

1 shows a trolley UBW of a subway ("Metro") with a control system according to the invention for active vibration reduction, wherein the positions of six actuator / sensor pairs AS1, AS2, AS3, AS4, AS5, AS6 are visible, Likewise, a control cabinet REK with a data processing system, which is a control device in the context of the invention, drivers and the associated power supply.

As already explained at the outset, in particular the lowest structural vibrations of the car body WGK are to be damped directly on the elastic structure and thereby increase the ride comfort for the passengers. The measurement and influencing of the oscillations occurs by means of the sensors and actuators AS1-AS6.

One of the challenges of the present system is that the controlled system is not classically defined as a clear interface between two rigid bodies, but is present as a continuum, namely as an elastic body structure. In principle, the attachment of actuators and sensors is possible at any location along the elastic structure; however, it is clear that efficiency is not the same everywhere. In principle, the number of sensors and actuators used can in principle be freely selected; an economically justifiable degree of effort and effect is found via the simulation.

According to the invention, the actuators and sensors are arranged on the car body WGK - in contrast to previous approaches, in which elements for vibration reduction on the chassis FGL or between the chassis and the car body are mounted. As shown in FIG. 2 (a section of FIG. 1) using the example of two actuators AK1, AK2, the actuators can be mounted on the underside of the car body, namely on the longitudinal member LGT of the carcass undercarriage. In the embodiment of Figs. 1 and 2 are - -

the actuators are all arranged parallel to the longitudinal direction of the carriage UBW; In addition, the actuators of a Aktorpaares are arranged opposite each other on the two side rails, while the actuators of a page in a row behind each other. Of course, other arrangements of actuators are possible, for example, all or individual actuators transverse to the longitudinal direction on the side member or Quer¬ striving be provided.

In the exemplary embodiment shown, the actuators also serve as sensors. Via electrical supply lines VL they are connected to the control device, which at the same time supplies the control voltages for the actuators and receives the sensor measurement signals.

Fig. 3 shows the mounting bracket BEK for the actuator AKl in detail, namely Fig. 3a in side view and Fig. 3b in front view. The actuator AKl is a piezoelectric stacking axial actuator, whose two end pieces are mounted by means of bolts BLZ in the cheeks BEW console BEK. The actuator AK1 can apply tensile and compressive forces via this mounting and thus bring about a local bending moment via a local compression or expansion on the underframe.

In addition, of course, a variety of other actuators / sensor geometries can be used. An example among many is a bending actuator which, as a so-called patch actuator, extends flatly onto a support surface, e.g. a bottom side or side surface of a longitudinal member is glued to impress there controlled bending moments.

2. Simulation

For the implementation and simulation of the vibration reduction system shown here, a suitable multi-body dynamics system (MBS) program is used, for example the SIMPACK package (a commercial MKS programmer package sold by Intec). The calculation of the movement of a point of the flexible Wagen¬ box is carried out using the method of moving reference frame. Accordingly, the movement of the point is composed of an arbitrarily large movement of a reference coordinate system associated with the body and a small deformation, the movement of the point relative to this reference system. The deformation u (R, f) is represented by a Ritz's approach of the form u (R, f) = Φ (R) q (£), where eigenmodes of the car body are complemented by additional starting functions (so-called 'Frequency Response Modes'). FRMs) can be used. The multipliers q are called mod coordinates. The equation of motion of a flexible body (without rigid body degrees of freedom) is then given

M _ee q + D _ee q + IQ _e q = φF _fc . (2.1) The modal mass matrix M _ee and the modal stiffness matrix K ₆₆ are then obtained from the mass and stiffness matrix M and K of the component with the aid of the seed functions Φ by the transformation M «= Φ ⁷ MΦ and K ^ = φ ^ϊ KΦ. D _ee denotes the modally transformed attenuation matrix.

The controller design is done by transforming the structure to be controlled into the modal state space and by analyzing the system equations obtained with it. The system is doing in the form

x = Ax + Buy = Ax + Bu (2.2)

shown. Here, A is referred to as a system matrix, B as a control matrix, C as an observation matrix and D as a passage matrix, x is referred to as the state vector of the system. As state vector, the modal coordinates and their derivatives are chosen. It then holds x = (qq) ^τ . The system matrix then becomes too

A = 0 E (2.3)

L -M ^ - ¹ IQ ₆ -M _ee ⁴ D _β

The control matrix B describes the influence of the input vector u on the system. In the application case considered here, the input vector corresponds to the control voltages applied to the piezoactuators; Consequently, the control matrix is calculated from the characteristics of the piezo actuators. The observation matrix C describes the relationship between the state vector of the system and the sensor output signals. The pass-through matrix D describes a possible reaction of the input vector on the output vector y (for example, with the simultaneous use of a piezoelement as actuator and sensor). The resulting system description is then used for the controller design (see section 4).

In order to test the method, the oscillation reduction of simple beam structures and then a simplified model of a rail vehicle, whereby the flexible vehicle body is represented in a greatly simplified form as a beam structure, were examined first. Most recently, the active vibration reduction was tested on the complex simulation model of a real rail vehicle. In this case, the damping of the first vertical bending natural frequency, the transverse distortion and the first torsional natural frequency of the car body is increased by the active vibration reduction. These three natural frequencies are selected because their influence on ride comfort is relatively large. In order to sound out the principle improvement potentials of an active vibration reduction, the travel of this vehicle is simulated on a track with track position errors. To assess the ride comfort, the vertical accelerations that are rated comfort to ISO 2631 are used used individual measuring points and with a. Vehicle compared without active Schwingungs¬ reduction.

The result of these simulations shows that, in principle, there is a considerable potential for improvement due to the active vibration reduction. Thus, significant reductions in comfort-valued vertical accelerations are achieved throughout; at individual points, there is even a reduction of more than 30%. In a further step, the behavior of the model-based controllers used for model deviations is examined. It shows that a good robustness against model deviations exists.

In the last step of these investigations, the effect of the active vibration reduction on the safety against derailment is examined. It will be the regular operation of the active vibration reduction system and an assumed accident, which leads to an excitation of torsional vibrations of the car body, examined. In the case of the scenarios examined here, only very small retroactive effects on the derailment safety can be found even in the event of a fault. The influence of active vibration reduction on driving safety and, above all, behavior in the event of a failure, thus appears to be considerably more favorable than direct intervention on the chassis. In summary, it can be stated that considerable improvement in ride comfort can be achieved by active vibration reduction of vehicle body structures.

3. Piezoelectricity

The use of piezoelectric elements as sensors and actuators for vibration reduction of structures is well known, making use of the indirect, for sensors the direct piezoelectric effect for the actuator. In connection with the active vibration reduction on rail vehicles, the use of piezoelectric components as sensors and / or actuators brings particular advantages for the realization of the invention.

By the (direct) piezoelectric effect is meant the property of certain crystals under the action of an external force to produce an electric charge. Conversely, materials with piezoelectric properties react to the action of an electric field with a deformation (indirect piezoelectric effect).

The forces are applied as follows: If a voltage (and thus an electric field) is applied to the actuator, it is subjected to an expansion. Becomes hinders this stretching, he exerts a corresponding force. The generated force thus decreases with increasing distance; If the actuator can move freely, no more force is generated.

A distinction piezo actuators in three different groups, namely axial actuators, transversal actuators and bending actuators. Axial and transverse actuators allow the application of large forces with small displacements, while bending actuators are suitable for applications in which large paths are required for smaller forces. As for the vibration reduction on large structures such. B. rail car bodies come only actuators in question, which can muster correspondingly large forces, come for the application in the first place so-called stack actuators in question.

In actual actuators not only a piezo element is used, since very large Bauteil¬ lengths and thus high electrical voltages would be required to achieve technically usable deformations. Therefore, in real running actuators several thin piezoceramic disks are used one behind the other. Such actuators are referred to as stack actuators. By constructing the actuator from thin piezoceramic disks (typically approx. 0.3 to 1 mm thick), it is possible to control voltages of the order of 100V. Actuators of this kind are offered, for example, by Morgan Electro Ceramics (http://inorganelectroceramics.com); another manufacturer of piezo elements is APC International Ltd. (http://americanpiezo.com). In addition, since piezoceramics should not be subjected to high tensile stresses, stack actuators are often provided with a mechanical bias. This biasing force is spielsweise - to allow a symmetrical effect of the actuator for train / pressure - up to 50% of the maximum force of the actuator.

4. Control concepts for active vibration reduction

Numerous control concepts are presented in the literature for active vibration reduction of flexible structures. In the following, concepts (known per se) such as the state feedback and the LQ control are discussed, since these methods - especially the state feedback with pole specification - allow a clear interpretation of the effect of the control on natural frequency and attenuation of the structure and at the same time favorable Represent methods for active vibration reduction.

The control concepts shown here can also be used to check the functionality and applicability. It can thus be estimated the potential for improvement by active vibration reduction and the magnitude of the restoring forces. - -

4.1 Status feedback with pole specification

As a starting point, a controlled system is considered in state space representation, but it is assumed that the sensors are separated from the actuators and therefore no direct influence of the manipulated variable is present on the sensor signal. The passage matrix D can therefore be in. Following are omitted.

x (i) = Ax (i) + Bu (f) (4.01) y (f) = Cx (O (4.02)

The goal of the state feedback with pole specification is a feedback matrix K for the calculation of the manipulated variable

u (i) = -Kx (t) (4.03)

• to be chosen such that the system matrix A ¹

A ¹ = A - BK (4.04)

of the closed loop

x = (A - BK) x (4.05)

has given eigenvalues. The advantage of this method is that by choosing the eigenvalues a precisely defined, physically interpretable influence on the system is taken.

When using this method for active vibration reduction so that both the attenuation at this natural frequency can be increased for a given natural frequency, as well as the position of the natural frequency are shifted. By shifting the natural frequency, the controlled system can be made "stiffer" than the unregulated one.

In order to realize a state feedback, the system must be controllable. If the system can not be controlled, the number of actuators must be increased or their position changed.

The disadvantage of this method is that the state vector x (i) of the system must be known. However, this is often not the case with active vibration reduction. In the application of the state feedback with Polvorgabe for active Schwingungsreduk¬ tion the modal coordinates q and their derivatives q are selected as state variables. However, since these quantities represent pure arithmetic variables that are physically not measurable, - -

they must be obtained for the application of a feedback from the available measurands. This is done by an observer (see section 4.3).

4.2 LQ regulation

Unlike the state feedback with Polvorgabe here the feedback matrix K is determined under specification of quality criteria for the course of manipulated variable and controlled variable. This is achieved by defining a quality function / as a measure of the quality of the control loop.

Defining the quality function of the form

/ = _o r (x (f) ^τ Qx (i) + u (i) ^τ Ru (f)) ät (4.06)

with the symmetric, positive-definite weighting matrices Q and R, this leads, under consideration of the governor law for the state feedback (4.03), to an optimization problem, namely according to the minimum value of the quality function, min / [K]. As a solution to this problem, the feedback matrix K for a controlled system described in (6.01) can be calculated in the following form

K = R-iβτp (4.09)

Where P represents the symmetric, positive definite solution of the matrix-Riccati equation:

Aτp + PA - PBR- ⁱ Bτp + Q = 0 (4.10)

Since the function / quadratic and the controlled system are linear, this form of control is called linear-quadratic control (LQ control). A special feature of the LQ control is known to be robust to model uncertainties.

4.3 Observer

In real technical systems, instead of the complete state vector i. a. only one output vector y (t) can be measured. In order nevertheless to be able to use control with state feedback, a so-called observer is used which reconstructs the state vector from the output vector. In order to be able to realize an observer, the observability of the system must be guaranteed. Otherwise, the number or positioning of the sensors for measuring y (f) must be changed.

4.4 Luenberger observers

The Luenberger observer is based on a parallel connection of the controlled system model to the controlled system with a feedback of the difference y (ε) -y (f) between the real measured variable - -

y (t) and the measured variable f (t) estimated from the observer's reconstruction on the model.

This feedback is used to equalize the state of the model with that of the controlled system, just as in a control circuit the deviation of the controlled variable from one setpoint is restricted. The relationship holds for the estimated value x of the state vector

St {t) = Ax (i) + Bu (f) + LC (x (t) -k (t)) (4.11)

If we introduce an observation error e (f),

e (t) = LC (x (i) -x (f)) (4.12)

Thus, the observation error sounds for any initial states of the system and the observer if and only if (limt → ∞ || e || = 0), if all eigenvalues of the matrix (A-LC) have a negative real part.

If a state feedback with the feedback matrix K is to be realized with the observer, the observer's eigenvalues are chosen such that the magnitude of the observer's real parts is approximately 2 to 6 times the magnitude of the real parts of the closed loop A-BK (ie that the observation error decays rapidly compared to the track behavior).

4.5 LQ observers

The LQ observer is based on the same structure as the Luenberger observer (4.11), but the return matrix L is determined in a different way. The design goal of the LQ observer is that the mean square observation error is as small as possible. This leads to the same task as with the controller design by means of LQ control, the feedback matrix results

L = PCTR-i (4.13)

where P is again the positive definite solution of the matrix Riccati equation (4.10). The matrices occurring therein are also in this case Q and R symmetric positive definite weighting matrices chosen before the design. The advantage of the LQ observer over the Luenberger observer lies again, as with the LQ controller, in the greater robustness against model uncertainties. - -

5. Active vibration reduction of a rail car body

After testing the active vibration reduction on simple structures, the principle is applied to a metro car body of a car UBW of the type shown in FIG. 1. The carbody WGK used has five door openings TRI, TR2, TR3, TR4, TR5 in each side wall. This results in a weakening of the structure at these points, which can be recognized by an "S-beat" of the mode shapes in this region.This is shown in Fig. 4 using the example of the first vertical bending shape of the carbody; particularly large in each case in the area of the doors TRI, TR2, TR4, TR5 (with the exception of the central door TR3, at whose location the oscillation maximum is located.) The actuator / sensor positions AS1-AS6 were based as shown in FIG.

5.1 vehicle model

The flexibility of the car body is taken into account through a modal approach. The first 17 eigenmodes from a finite element (FE) calculation and 12 FRMs for taking account of the local deformations by the actuators are used as the approach functions. For the simulation of the entire vehicle, this car body is connected to two bogies.

5.1.1 Actuator modeling

Actuators use twelve piezo stack actuators. In modeling, the portion of the acting force, which is dependent on the actor strain, is added or subtracted to the force resulting from the piezoelectric effect. For the force generated by the piezoelectric effect, a linear relationship between the applied electrical voltage and the generated force is again assumed.

The consideration of the local deformation as a result of the actuator effect takes place again by calculating FRMs for the twelve actuators.

5.1.2 Sensor modeling

As sensors twelve piezo elements are provided. They are positioned in the same places as the actuators. Their output signal is proportional to the auftre¬ at their installation point limiting strain.

5.1.3 Actuator and sensor placement

The selection of the actuator and sensor mounting locations is based on several criteria: First, the installation should be made in places where the eigenmodes to be controlled as large as possible - -

Strain to ensure maximum effectiveness. Furthermore, the structure at the installation locations must enable the initiation of the occurring forces and moments without occurrence of strength problems. Finally, it must also be checked whether there is a corresponding installation space at the selected positions. Within these limits, however, it is undoubtedly possible to further optimize the position of the control technology.

After examining these criteria, therefore, the edges of the door cutouts of the respective inner three doors TR2-TR4 on the lower longitudinal beam LGT were selected as installation locations (FIG. 1).

5.1.4 Used control concept

5 shows a representation of the structure of the control loop, wherein the upper part represents the Wagen¬ box or the controlled system 51, while the lower part is realized by the Regel¬ device 52 which is located in the control cabinet REK of the invention Regelsys¬ system ,

The controlled system 51 is formed by the car body WGK in the example considered here. For simulation purposes, this part can also be modeled; then, in order to enable a controller design, the relationship between input quantities u (here the forces exerted by the actuators on the structure or the electrical voltages with which the actuators are driven) and output values y (output signals of the sensors) on this controlled system described a mathematical model. The matrix Bw represents the effect of the actuators and indicates the relationship between the input variables and the state variables, the matrix Cw the relationship between the output variables y and the state variables. As state vector x, the modal coordinates and their derivatives are chosen. The matrix Aw represents the system matrix (2.3) of the car body or of the simulated physical system.

From these output signals, the state variables are reconstructed with the aid of the observer 521. The use of the matrices A, B, C in the observer with respect to Aw, Bw, Cw in the controlled system is intended to indicate that the model with which the observer operates has a smaller dimension than the controlled system. The reason for this is that the approach which describes the deformation as a linear combination of the individual eigenmodes is exact only if all (infinitely many) eigenmodes are used; in the observer, however, only a finite number of eigenmodes can be used, whereby it is endeavored to keep this number as small as possible.

The controller 522 itself performs only a matrix multiplication of the state variables x (estimated by the observer) with a feedback matrix K. When returning to the In addition, control variables w (setpoints) can be deducted if necessary.

A state feedback with pole specification is used. Since above all the first verti¬ kale natural bending frequency, the first torsion natural frequency and the roof transverse vibration have a great influence on the ride comfort, the damping should be increased at these three natural frequencies. In addition, the regulation also increases the damping of the FRMs in order to prevent vibration excitation by the forces introduced by the actuators. As a result of the regulation, the attenuation in the regulated frequencies is increased from the structural damping of 2% to 30%.

An observer is used to reconstruct the modal coordinates selected as state variables. Since the Luenberger observer does not have the required robustness, an LQ observer is used.

For comparison, a state feedback without observer was realized. This can only be done in a simulation since it is possible to directly access the physically non-measurable, modal coordinates selected as state variables.

To further approximate a realistic control system, the sensor signals are detected by a low-pass filter with a cut-off frequency of 190 Hz (filtering of high-frequency interferences) and the reconstructed state variables with a high-pass filter with a cut-off frequency of 6.7 Hz (filtering of the static deformation). changed.

It should be noted that a check whether a control device contains a modal model according to the invention can also be carried out without intervention in the control device. For example, the control device 52 can be taken as a closed component from the control loop (i.e., the connections to the actuators and sensors are temporarily disconnected), and the transmission behavior is checked by known methods of system identification. In this way, the behavior of the controller can be determined; If the controller behavior thus found specifically attenuates the natural frequencies of the structure, the design of the controller must be based on a modal model.

5.2 Simulation results

To test the performance of the invention, two different scenarios were simulated: In the first case, the vibration excitation of the car body was carried out solely by introducing forces at the secondary spring connection points. In the second case, the car body was connected to the associated bogies and simulated the ride on a predetermined route. The force was therefore not only to the - -

Sekundä ^ Federarüenkpunkten, but also at the attachment points of Querdämp¬ fern, roll stabilizer and Längsmitnahme.

The excited modes corresponded to (i) the first bending mode and (ii) the first torsional form. The oscillation excitation was carried out by a stochastic force F (t), which was initiated at the four pivot points of the secondary springs under the vehicle floor (two under the rear edge of the foremost doors TRI, two just before the rearmost doors TR5). In order to approximate the filtering of trackside interference by the primary and secondary stages of the bogies, a stochastic signal ("white noise") was filtered through a low pass filter with a cutoff frequency of 10 Hz before being applied to the structure as a force the time course of the forces applied to the second bogie equal to the time course of the forces applied to the first bogie, but with a delay by the time period U • In this way, the delay occurring in a journey of the vehicle on a route was between the passage of a fault the front and the passing through the rear bogie replicated.

In both cases, an evaluation is carried out by calculating the occurring vertical acceleration at a plurality of measuring points on the vehicle floor. For evaluation, the rms values of the occurring vertical accelerations at the measuring points were calculated. A comparison was drawn between a car body without active vibration reduction and a realizable vibration reduction with LQ observer.

In the case of bending excitation, there was a significant reduction (up to about 70%) in the vertical accelerations, which was (as expected) the greatest improvement in the mid-point points between the bogies. Significant reductions (about 50%) at the anterior and posterior measuring points also resulted for the torsion stimulus; only at the measuring points in the middle of the vehicle was the oscillation reduction only slight, which was due to the excitation of higher natural frequencies. Overall, there was a significant improvement in the vibration behavior by the inventive method.

The embodiment described above uses an observer to determine the sizes of the state vector. If the state vector can be derived from the measured variables with sufficient reliability (for example, by suitable linear combinations), then the observer can be dispensed with. This is the case, for example, if the inverse O ^{1 of} the observation matrix is known with sufficient accuracy. In this case, the observer 521 in FIG. 1 can be replaced by this inverse C- ¹ , thus the manipulated variables u in the control device 52 are calculated according to u = -Kx = -KC ^ y. - -

Incidentally, whether an observer is implemented in a controller concept can be checked on the basis of the number and arrangement of the sensors and the number of attenuated modes. If the number of attenuated modes is large, the use of an observer can be inferred, since only with the use of an observer can all states of interest be estimated from a smaller number of measurements, and thus the intrinsic modes of interest can be attenuated.

Claims

- PERSONAL CLAIMS

1. Regelsy stone for active vibration reduction on a rail vehicle (JJBW), with a number of sensors (AS1-AS6) for measuring deformations of Schienenfahr¬ zeugs and a number of actuators (ASl-AS6) for generating actuating forces and / or actuating torques on the rail vehicle and a control device (52) which is equipped to a) receive measurement signals from the sensors, b) to calculate a set of control values from the sensor measurement signals, and c) to supply actuating signals to the actuators corresponding to the control variables, characterized in that the actuators (AK1, AK2) are mounted exclusively on attachment points (BEW) of the railcar of the rail vehicle and the control device (52) is arranged to derive characteristic quantities (x) from the sensor measurement signals (y) in step b) relate elastic deformations with respect to the vibration eigenmodes of the car body, and on the basis of this characteristic size to calculate the manipulated variables s (u).

2. Control system according to claim 1, characterized in that the sensors (AS1-AS6) are arranged on the car body.

3. Control system according to claim 1 or 2, characterized in that at least a part of the actuators are additionally used as sensors.

4. Control system according to one of claims 1 to 3, characterized in that the actuators are implemented as piezoelectric actuators, in particular stacked piezoelectric actuators.

5. Control system according to one of claims 1 to 4, characterized in that the sensors are implemented as piezoelectric sensors.

6. Control system according to one of claims 1 to 5, characterized in that the actuators are mounted on the car body in the vicinity of the maximum deformation of a Schwin¬ gungseigenform.

7. Control system according to claim 6, characterized in that the actuators are mounted on Wagen¬ box in the vicinity of the maximum curvature deviation of a Schwingungseigen¬ form.

8. Control system according to one of claims 1 to 7, characterized in that the actuators are mounted on the car body in the vicinity of a relevant for the shape of a vibration characteristic structure weakening.

9. Control system according to one of claims 1 to 8, characterized in that the characteristic variables represent a state vector (x), which refers to the state of deformation in terms of natural vibration forms of the car body.

10. Control system according to one of claims 1 to 9, characterized in that the Regel¬ device calculates the characteristic quantities exclusively from the sensor measurement signals (y) and predetermined desired values (w) according to a predetermined algorithm.

11. Control system according to one of claims 1 to 10, characterized in that the control device derives the characteristic quantities by means of an observer estimate (521).

12. Control system according to claim 11, characterized in that the observer estimate is realized as an LQ observer.

13. Rail vehicle having a number of sensors for measuring deformations of the rail vehicle and a number of actuators for generating restoring forces and / or adjusting torques on the rail vehicle, and with a control system for active vibration reduction according to one of claims 1 to 12.

14. A control device (REK) for active vibration reduction in a rail vehicle, which is arranged, in cooperation with a number of sensors for measuring deformations of the rail vehicle and a number of actuators for generating restoring forces and / or control torques on the rail vehicle a) measuring signals from to receive the sensors, b) to calculate a set of manipulated variables from the sensor measurement signals, and c) to supply actuating signals corresponding to the manipulated variables to the actuators, characterized in that - -

Starting from the assumption that the actuators (AK1, AK2) are mounted exclusively on fastening points (BEW) of the car body of the rail vehicle, the control device is set up in step b) from the sensor measurement signals (y) characteristic variables (x) derive, which relate to elastic deformations in terms of Schwin¬ donated own forms of the car body, and to calculate the manipulated variables (u) on the basis of these characteristic variables.

15. Control device according to claim 14, characterized in that the characteristic quantities represent a state vector (x), which relates to the state of deformation with regard to natural vibration forms of the car body.

16. Control device according to claim 14 or 15, characterized in that the characti ristic quantities exclusively from the sensor measurement signals (y) and predetermined Sollgrö¬ Shen (w) are calculated according to a predetermined algorithm.

17. Control device according to one of claims 14 to 16, characterized in that the characteristic quantities are derived by means of an observer estimate (521).

18. Control device according to claim 17, characterized in that the observer estimate is realized as an LQ observer.