METHOD FOR CONTROLLING A TEST BENCH ARRANGEMENT

Description

[0001] The present invention relates to a method for controlling a test rig arrangement in which a rotating test specimen is connected to a rotating load machine via a mechanical shaft connection and in which at least one angular velocity prevailing in the test rig arrangement is measured, wherein the rotational behavior of the test rig arrangement and thus the dynamic behavior of at least the measured angular velocity and a test specimen angular velocity prevailing in the test specimen is modeled with a first system of differential equations.

[0002] For a long time now, test benches have been an indispensable tool in a wide variety of technological applications. This is especially true in the automotive industry, where trends such as the electrification of the powertrain, coupled with ever-shortening innovation cycles, continuously lead to new testing challenges. In this context, the demands for lower costs, greater flexibility, shorter setup times, and increased dynamics are particularly noteworthy, as they are the dominant factors in many of these new testing challenges.

[0003] Several parallels can be drawn between the aforementioned factors. First, costs can be reduced primarily by decreasing the amount of hardware installed in a test bench. This applies particularly to measurement and sensor devices, which in many cases are not only expensive components but can also significantly increase the implementation, commissioning, and maintenance effort of a test bench. Furthermore, in the vast majority of applications, a reduced number of sensors can considerably increase the flexibility of a test bench. This is primarily due to the reduced cabling effort typically associated with fewer sensors, which can be particularly advantageous during modifications and / or adaptations of test bench hardware. For the same reasons, a smaller number of sensors often also allows for shorter setup times when changing test specimens.

[0004] The fourth factor mentioned, the requirement for higher dynamics, often leads to a different conclusion regarding the number of sensors to be installed. "Dynamics" here refers to changes in non-steady-state operating conditions, i.e., changes in certain parameters of the test bench setup, such as rotational speed or torque, at short intervals, for example, in the millisecond or even microsecond range. These changes can also be large. For highly dynamic operation of a test bench, it is often crucial to have the most precise information possible about the state of the test bench in order to ultimately enable highly dynamic control of the test bench setup. In this context, every additional measured parameter can be helpful or even necessary.This tension regarding the number of sensors to be provided arises in a wide variety of testing applications.

[0005] In powertrain test benches, the powertrain under test is positioned as the test specimen on the test bench and typically connected to at least one load machine (dynamometer) via several mechanical connecting shafts. The drive unit of the powertrain, e.g., an internal combustion engine and / or an electric motor, then operates against the at least one load machine on the test bench to test various load conditions expected in real-world operation. Such a powertrain test bench with multiple load machines is known, for example, from DE 10 2008 041 883 A1. In such cases, it is often helpful to know precisely the torques delivered by the load machines, among other things to avoid cross-interference between the load machines.

[0006] In contrast, engine test benches typically use only a single drive unit, such as an internal combustion engine and / or an electric motor, as the test specimen and are connected to a load machine (dynamometer) via a shaft connection. In such cases, a particularly noteworthy example of an operating mode where additional measurement information can be crucial is the so-called damping control, as described, for example, in AT 519092 A4. A common objective here is to determine a damping component of a control signal used for test bench control from the current rotational speed of the test specimen, hereinafter also referred to as "test specimen speed," and / or the test specimen torque generated by the test specimen. Measured values ​​of test specimen speed and / or test specimen torque significantly simplify the implementation of the control concept in this case.

[0007] Besides the desire to reduce expensive and complex sensor technology for cost and efficiency reasons, it often occurs that it is simply impossible to measure certain parameters, such as the rotational angle of a test specimen, its rotational speed, or the torque it generates, with sufficient accuracy, even though this would allow for clear improvements in the operational behavior of a test bench. Either the necessary sensors are unavailable, or the significant effort involved in installing and using existing sensors precludes their use. Furthermore, it is often necessary to test specimens that do not allow for the acquisition of measurement data for the parameters of interest, or in which such measurement is not provided for.An example of this is the internal torque generated by an internal combustion engine or electric motor; that is, not the torque actually delivered at the output shaft, but the torque actually generated. This internal torque often cannot be measured directly.

[0008] For the reasons mentioned above, several approaches exist in the prior art in which sensors not present in real test bench environments are replaced by so-called virtual sensors. This is usually achieved computationally, often using the observer technique well-known in control engineering, to infer desired, but not directly available, measured quantities from existing measured variables. Measurement signals generated in this way are often referred to as virtual measurement signals.

[0009] In this context, the article "Nonlinear Observers for Closed-Loop Control of a Combustion Engine Test Bench," by G. Reale et al., American Control Conference (2009) 4648-4653, describes the use of various state observers on a combustion engine test bench. Based on these state observers, estimates are determined for the internal torque generated by the combustion engine, the rotational speeds of the combustion engine and the load machine, and the angle difference between the combustion engine and the load machine. However, in the described case, the aforementioned rotational speeds are also measured, which essentially renders their estimation obsolete.

[0010] AT 519092 A4 discloses a method for damping control in an internal combustion engine test bench. Based on a measured shaft torque and a measured test specimen rotational speed, an estimated value of the internal test specimen torque is determined; however, no details are given regarding the models used for the estimation and the state observers derived from them, which significantly complicates the transfer of the presented concept to other test bench configurations, such as electric motor test benches.

[0011] AT 522354 A4 describes the estimation of a shaft torque acting in the connecting shaft of an engine test bench. Based on this estimate, a further estimate of a test specimen torque is determined, which prevails in a test specimen comprising at least two rotating masses. Again, the specific design of the estimator used is not described in detail.

[0012] In contrast, DE 102020104314 A1 describes a model predictive control system for a test bench used to test drive components. In particular, it estimates future values ​​of quantities occurring in the test bench setup, such as the rotational speed of the connecting shaft and the torques generated by the load machine and the test specimen. While this document discloses a concrete model of the test bench setup, the modeling of the test specimen torque is only very rudimentary, as the focus of the disclosure is on the control of a state variable of the connecting shaft.

[0013] The cited state of the art exhibits several commonalities. A first striking point is that none of the publications describe a simultaneous estimation of the test specimen's rotational speed and torque; instead, at least one of the two quantities is always measured. However, precise information about the test specimen's rotational speed and torque is often a prerequisite for monitoring the power dissipated in a test specimen, which can be crucial, especially for high-speed electric motors. Particularly with high-speed electric motors, the problem frequently arises on the test bench that the test bench control system cannot access either the test specimen's rotational speed or torque, yet the power dissipated by the test specimen must be taken into account during test bench operation, among other things to avoid thermal overload of often valuable test specimens ("golden samples").

[0014] Another characteristic of the cited state of the art is that no document describes an estimation of the test specimen's rotational speed where the estimated rotational speed is not already available as a measured variable. For safety reasons alone, the often very high rotational speeds of electric motors, for example, must be monitored, which, in the absence of measurement technology, can only be achieved through reliable estimates.

[0015] Moreover, none of the aforementioned publications demonstrates how, particularly in the case of electric motor test benches, an internal test specimen torque generated by a test specimen can be advantageously modeled for subsequent estimation. Furthermore, for the models of test specimen torques known from the prior art, it is largely unclear how these can be meaningfully extended, for example, to increase model accuracy. Moreover, conceivable extension approaches often entail a significantly increased model complexity, which can be particularly disadvantageous in practical applications.

[0016] For the reasons mentioned above, the practical application of the cited concepts sometimes presents considerable difficulties. Possible consequences of these difficulties can range from inadequate operational quality of a test bench to damage to the test bench and / or the test specimen.

[0017] It is therefore an object of the present invention to provide an improved method for the simultaneous estimation of a test specimen speed and a test specimen torque, which can be used flexibly, in particular, on automotive test benches in general and on electric motor test benches in particular.

[0018] Diese Aufgabe The invention solves this problem by the features of the independent claims. It is based on a test bench arrangement in which a rotating test object, preferably in the form of an electric motor, is connected to a rotating load machine via a mechanical shaft connection, and in which at least one rotational speed prevailing in the test bench arrangement, but preferably the load speed assumed by the load machine, is measured.

[0019] The first step of the method according to the invention involves modeling the rotational behavior of the entire test rig arrangement. Typically, the dynamic behavior of the test rig parameters relevant to the rotational behavior of the arrangement, such as load speed, test specimen speed, shaft torque, test specimen torque, etc., is described using a model. At a minimum, however, the dynamic behavior of the measured speed and the test specimen's assumed speed is represented. For this purpose, a first system of differential equations is specified, which is preferably implemented as a rotational multibody system and accordingly has at least two rotating masses or inertias connected to each other via a spring-damper element. The at least two rotating masses or inertias are then...Inertias here serve to represent the test specimen and the loading machine, while the at least one spring-damper element represents the mechanical shaft connection between them. For the method according to the invention, it is important at this point that both the load torque generated by the loading machine and the test specimen torque are input variables of the first system of differential equations. The load torque acts on the inertia representing the loading machine, whereas the test specimen torque acts on the inertia representing the test specimen.

[0020] The next process step constitutes the core of the present invention. In this step, a model of the test specimen torque is specified using a second, autonomous system of differential equations, which can have a high model order. A first subsystem represents the DC component, and m further oscillating subsystems represent m harmonics of the test specimen torque. The m oscillating subsystems for modeling the m harmonics are preferably implemented as damped and / or undamped oscillators. In control engineering, such systems for describing input variables are already known as so-called "exosystems." However, they have not yet been applied in test bench technology, although their use can lead to a number of surprisingly positive effects.

[0021] According to the invention, the first and second systems of differential equations are subsequently combined into a comprehensive model, on the basis of which a state observer is designed in a further process step to estimate the test specimen's rotational speed and torque. The measured rotational speed constitutes an input, and the test specimen's rotational speed and torque to be estimated constitute the outputs of the state observer. Subsequently, estimated values ​​for the test specimen's rotational speed and torque are determined using the designed state observer, and these are used in a final process step to control at least one rotational quantity prevalent on the test bench.

[0022] A primary benefit of using an autonomous exosystem to model the test specimen torque arises in this particular case from the fact that the test specimen torque, which appears as an unknown input in the first system of differential equations, can be replaced by the presented exosystem. The resulting model can be considered a new, comprehensive model encompassing both the rotational dynamics of the test rig setup and the dynamics of the test specimen torque. Due to the autonomous nature of the exosystem, the dependencies on unknown external quantities present in the first system of differential equations are resolved. Since unknown external input quantities can pose a significant challenge in many applications of observation technology, this effect alone often improves the quality of the resulting estimates.Furthermore, if the test object's rotational speed is estimated by reconstructing the state of the overall model on which the estimate is based, as is common practice in the observer technique well-known from control engineering, the test object's torque is estimated automatically in this specific case without any additional effort. The aforementioned observation of the power dissipated in the test object thus occurs essentially automatically.

[0023] A second, practically valuable advantage of the inventive approach arises from the structure of the second differential equation system or the structure of the exosystem for modeling the test specimen torque. Considering that the accuracy of a signal reconstruction can be arbitrarily increased by including harmonics, this directly opens up a possibility for extending the presented exosystem or the second differential equation system. Accordingly, if additional oscillating subsystems in the form of oscillators are added to account for further harmonics, the number of state variables in the exosystem, and thus in the resulting overall model, increases.Since the respective oscillating subsystems do not influence each other but are decoupled, neither the exosystem nor the resulting overall model becomes significantly more complex or difficult to handle. This aspect is particularly relevant when assessing the practical applicability of the method according to the invention, as in practice, one usually tries to avoid models with high order. However, since the aforementioned oscillating subsystems do not influence each other in any way, the only negative consequence of adding further oscillating subsystems—namely, the need to consider additional state variables—is outweighed by the significant advantage of increased model accuracy. This property naturally also applies to exosystems that inherently exhibit a high model order.

[0024] A further advantageous aspect of the method according to the invention can be attributed to the stability of the exosystem used and the resulting overall system. Since the oscillating subsystems for modeling the harmonics are preferably damped and / or undamped oscillators whose eigenvalues ​​lie either on the imaginary axis or in the left open half-plane of the complex plane, the stability characteristics of the resulting overall system cannot be fundamentally altered by adding further oscillators. Since there is no risk of a potential loss of stability in the models used for estimation, even when the model parameters are changed, the exosystem for modeling the test specimen torque can be adapted to changing rotational parameters of the test rig without concern during operation.These are tracked, preferably at a measured rotational speed such as the load speed assumed by the load machine.

[0025] It should be noted here that while the object of the present invention originates in the field of motor and, in particular, electric motor test benches, the method according to the invention can also be applied to a variety of other test benches, for example, powertrain test benches, roller test benches, or component test benches. The specific control task, which is solved according to the method according to the invention using the estimated test specimen torque and speed, is also broadly applicable. In addition to the damping control already mentioned, mass or inertia simulation and n-alpha operation are particularly noteworthy as further applications.

[0026] The present invention is described below with reference to the Figuren 1 bis 6 In more detail, the invention is explained, and exemplary, schematic, and non-restrictive embodiments are shown. This includes showing Fig.1 a schematic representation of a test bench setup Fig.2 a schematic representation of a model of a two-mass oscillator Fig.3 an exemplary curve of a test specimen torque Fig.4 a block diagram of an exemplary controller structure without the estimation of test specimen speed and test specimen torque according to the invention Fig.5 a block diagram of an exemplary controller structure with estimation of test specimen speed and test specimen torque according to the invention Fig.6 a test bench setup with two load machines

[0027] Fig. 1 Figure 1 shows a schematic representation of the essential components of an engine test bench. A rotating test specimen 1, preferably in the form of an electric motor or internal combustion engine, is connected via a connecting shaft 3 to a load machine 2, which applies a load torque to the test specimen 1. The load machine is generally an electric machine. In the context of this description, the unit consisting of the test specimen 1, the load machine 2, and the connecting shaft 3 is also referred to as the test bench arrangement 4. In the situation shown, an automation system 5 determines control variables for the operation of the test bench arrangement 4, specifically for the individual active components, and specifies these variables to the test bench arrangement 4, which in particular specifies a load machine torque. T D includes the load machine 2. Besides the control variable for the load machine torque T D The automation system 5 can also output a control variable α for the test object 1. Depending on the test object 1, the control variable can be... α These parameters can have different meanings; in the case of an internal combustion engine, it can represent a pedal position, throttle position, or similar, while in the case of an electric motor, its meaning is often a percentage of the maximum torque that the test object 1 can generate. It should be noted here that the test object 1 does not necessarily have to be connected to the automation system 5. In such cases, the specification of a control variable occurs... α no longer by automation system 5, but must be done in another way, for example by a specification from a test bench operator or by a control unit provided on the test specimen 1.

[0028] The control variables are subsequently converted by an actuator 6 of the load machine 2 or an actuator 6' of the test specimen 1 into corresponding manipulated variables, which are used to control these. If the load machine 2 is designed as an electric machine, for example as a synchronous or asynchronous machine, as is often the case, a load machine torque transmitted as a control variable is used. T D in the form of an air gap moment T L realized. As a control variable for generating this air gap torque. T L In this case, the winding currents or winding voltages of the load machine 2 can then be used. The air gap torque T L is generated in a known manner by the load machine 2 and acts on the rotor of the load machine 2.

[0029] The actual values ​​of the controlled variables are determined using suitable sensors and transmitted to the automation system 5 via suitable signal lines, as described in Fig. 1 The actual value of the angular velocity of the load machine 2, which is subsequently also referred to as the "load angular velocity" ωD, is shown. The load angular velocity ωD can be measured, for example, using a rotary encoder 7. The test specimen angular velocity ω P is not available as a measured quantity in the method according to the invention, and neither is the shaft torque transmitted by the connecting shaft 3. T ST is often not available in the form of a measured quantity, which is indicated by the dashed signal lines Y in Fig. 1 As indicated. Angular velocities ω and rotational speeds. n of the test bench arrangement 4, for example test specimen angular velocity ω P and test specimen speed n P , depend on the well-known expression ω = π n 30 These terms are to be considered synonymous with regard to their effect on the method according to the invention. To ensure consistency in the following text and the mathematical description of the models mentioned, angular velocities will be used predominantly and without limitation of generality.

[0030] The first step of the method according to the invention involves specifying a first system of differential equations for modeling the rotational behavior, hereinafter also referred to as vibrational behavior, of the test rig arrangement 4. For this purpose, a rotational multibody system is preferably used, the simplest embodiment of which is a two-mass oscillator. Fig. 2 A two-mass oscillator of this type is shown. In addition to the test bench parameters relevant to the vibration behavior, shaft moment... T ST , Air gap torque generated by the load machine 2 T L , Test specimen torque T P , Loading machine angular velocity ω D and test specimen angular velocity ω P contains Fig. 2 the shaft stiffness c and the shaft damping d to describe the mechanical connecting shaft 3 between test specimen 1 and load machine 2, as well as the test specimen moment of inertia J P and the load machine moment of inertia J D These parameters can be assumed to be known. Based on the in Fig. 2 The two-mass oscillator shown can be described by a well-known first system of differential equations in the form of: d dt Δφ ω D ω P ︸ x = 0 1 − 1 − c J D − d J D d J D c J P d J P − d J P ︸ A Δφ ω D ω P + 0 − 1 J D 0 ︸ b L T L + 0 0 1 J P ︸ b P T P with the initial equation ω D = 0 1 0 ︸ c A Δφ ω D ω P be specified, whereby Δφ This represents the differential rotation angle between test specimen 1 and load machine 2. The variables x, A, b L , b P and c A Here, in the appropriate order, they represent the state vector, the dynamics matrix, the first and second input vectors, and the output vector of the first input.

[0031] The system of differential equations results from the modeling. In this case, the output equation describes the angular velocity ωD of the load machine 2, which is common practice since the corresponding load rotational speed is typically available as a measured quantity. Various other configurations of the output equation, and thus of the output quantity, are conceivable by choosing an alternative output vector.

[0032] Like any model of a real physical system, the present first system of differential equations represents a simplified, model-based abstraction of the real behavior of the modeled test rig setup 4. Thus, in formulating the model, the mass inertia of the elastic connecting shaft can be attributed equally to the inertia JD and the inertia J P The values ​​are added together. Furthermore, it is assumed that all system components are linear elements. Therefore, possible nonlinear characteristics of the test rig setup 4, such as shaft play, nonlinear stiffness, or static friction, are not considered. The actuator dynamics introduced by the load machine 2, which are typically characterized by a combination of dead time and first-order lag behavior during the conversion of the control variable, are also not taken into account. T D into an air gap moment T L The manifested dynamics are neglected. In particular, the air gap torque corresponds to the actuator dynamics when the actuator dynamics are neglected. T L the control variable T D . By considering the features neglected here, further, more complex models of the test rig arrangement 4 could be directly specified in the form of systems of differential equations. However, the first system of differential equations shown is sufficient to describe the method according to the invention. The model parameters are assumed to be known for the following considerations. Based on the given model, the formulation of the general problem of the present invention can be specified for the illustrated embodiment and for better understanding, namely as the determination of a method for estimating the quantities. ω P and T P .

[0033] To solve this problem, as mentioned, a second system of differential equations is used to describe the test specimen torque. T P The generated test specimen torque is specified. T P is modeled with a second system of differential equations, whereby the DC component of the test specimen torque is determined using a first subsystem of the second system of differential equations. T P is represented and based on a number m of further oscillating subsystems of the second differential equation system in the form of damped and / or undamped oscillators m harmonics of the test specimen torque T P be depicted.

[0034] In this specific embodiment, three (m = 3) harmonics are taken into account, giving the second system of differential equations the form d dt z 0 z 1 , 1 z 1 , 2 z 2 , 1 z 2 , 2 z 3 , 1 z 3 , 2 ︸ z = 0 0 0 0 0 0 0 0 0 − ζ 1 0 0 0 0 0 ζ 1 0 0 0 0 0 0 0 0 0 − ζ 2 0 0 0 0 0 ζ 2 0 0 0 0 0 0 0 0 0 − ζ 3 0 0 0 0 0 ζ 3 0 ︸ S z 0 z 1 , 1 z 1 , 2 z 2 , 1 z 2 , 2 z 3 , 1 z 3 , 2 with the initial equation T P = 1 1 0 1 0 1 0 ︸ c S z 0 z 1 , 1 z 1 , 2 z 2 , 1 z 2 , 2 z 3 , 1 z 3 , 2 assumes. Here too, analogous to the first system of differential equations, the variables z, S and c s The following are specified, which, in the corresponding order, denote the state vector with a number of 2m + 1 state variables z, the dynamics matrix, and the output vector of the second system of differential equations. Based on the state variable z 0, which of the differential equation d dt z 0 = 0 When obeyed, the equal component of the test specimen torque is determined. T P described.

[0035] The remaining state variables z k ,1 , z k ,2 are always to be considered in pairs, each of which corresponds to the k -ten oscillating subsystems. The running variable kThis serves to index the oscillating subsystems and their associated variables. Specifically, oscillating subsystems are understood to be those parts of the differential equation system that are each formed by mutually influencing state variables of the state vector. Accordingly, for example, the following applies to the state variables of each of these oscillating subsystems: z k ,1 and z k,2 the vector-valued differential equation d dt z k , 1 z k , 2 = 0 − ζ k ζ k 0 z k , 1 z k , 2 , which is known to be a harmonic oscillator with the angular frequency ζ k This corresponds to the eigenvalues ​​of this subsystem. jζ k on the imaginary axis of the complex plane, and these can also occur with variations in angular frequency ζ k do not leave. One way to extend this oscillator is by taking damping into account using a suitable damping factor η. k Given. Such a modification would result in a damped oscillator with the differential equation d dt z k , 1 z k , 2 = − η k − ζ k ζ k − η k z k , 1 z k , 2 , lead to whose eigenvalues ​​in the case of a positive and real damping factor η k The eigenvalues ​​of an oscillator that are greater than or equal to zero generally lie in the left half of the complex plane. The property that the eigenvalues ​​of such an oscillator can never lie in the right half of the complex plane is particularly advantageous in cases where the second system of differential equations is used to model the test specimen torque. T P is adapted to a variable of the test bench arrangement 4 that changes during operation of the test bench, for example to the angular velocity ω D .

[0036] Starting from a non-zero initial state, the two variables z k ,1 and z k,2 Each produces a sinusoidal waveform that is phase-shifted by 90 degrees. By selecting one of these two state variables, a single harmonic of a test specimen's torque can subsequently be determined. T P and by summing several such determined harmonics, a total harmonic-laden test specimen torque T P be described.

[0037] Regarding the structure of the initial vector c S It can be seen that the initial quantity is formed T P The state variable z 0 and exactly one associated state variable from each of the oscillating subsystems are summed. Fig. 3 This shows a possible progression of a test specimen torque generated in this way. T P .

[0038] To determine the angular frequencies ζ k to parameterize, for example to integer multiples of the load angular velocity ω D recourse is made to the connection ζ k = kω D leads to the angular frequencies. ζ k thus become functions of the load angular velocity ω D , which can be used to determine the angular frequencies ζ k during operation, the load angular velocity ω may change. D to adjust. Instead of the load angular velocity ω D However, other rotational quantities of test bench arrangement 4 can also be used to parameterize the angular frequencies. ζ k can be used. Furthermore, it is also possible to use the angular frequencies. ζ k Select the frequency once at the beginning of operation in the manner described, but then do not change it again and keep it constant. For electric motor test benches, it is often advantageous to select the frequencies in the same way. ζ k the integer number of pole pairs p to consider what to ζ k = kpω D leads to and can contribute to a reduction of the resulting model order.

[0039] It should be noted at this point that the choice of frequencies ζ k It can also be achieved in a completely different way, for example by using odd ratios between the frequencies. ζ k and the load angle velocity ω D . The frequencies ζ k However, they can also be selected by a specialist for specific applications.

[0040] An extension of the second system of differential equations mentioned earlier, aimed at increasing model accuracy, can easily be achieved by adding further oscillators, thereby increasing the number m. It should also be noted that consecutive harmonics are not necessarily required. For example, only odd harmonics, only even harmonics, or any selection of harmonics can be used.

[0041] To actually determine the estimated values ​​for ω P and T P The first and second systems of differential equations are combined into a single system. To describe this step in more detail, the first system of differential equations is first expressed in its compact form using the previously introduced abbreviations for the system parameters. d dt x = Ax + b L T L + b P T P ω D = c A z and the second system of differential equations in compact form d dt z = Sz T P = c S z angeschrieben. If you replace the entrance T P of the first system of differential equations through the output T P of the second system of differential equations, the overall system is obtained d dt x z = A b P c S 0 x S x z + b L 0 z T L ω D = c A 0 z x z , wherein 0 x and 0 z for zero matrices of suitable dimension. For such a linear system, various approaches exist to provide a state observer for estimating the state vector. x z to design.

[0042] However, one preferred approach is to use the well-known Luenberger observer, which should by no means be considered restrictive. As is well known, a Luenberger observer comprises a copy of the system of differential equations that describes the dynamics of the variable to be estimated, and a correction term, which in this specific case leads to the observer differential equation. d dt x ^ z ^ = A b P c S 0 x S x ^ z ^ + b L 0 z T L + K ω D − c A 0 z x ^ z ^ leads to x̂ and ẑ for the estimated values ​​of the state vectors x and to be determined z . The vector-valued correction gain K is usually chosen such that the dynamics matrix  the differential equation of the estimation error e = x z − x ^ z ^ It assumes predetermined eigenvalues, which is a well-known procedure in control engineering. The dynamic matrix of the differential equation for the estimation error e is directly derived from the equation above as A ^ = A b P c S 0 x S − K c A 0 z , in which the effect of the correction amplification K The parameters of the matrix, and thus its eigenvalues, are clearly evident. Besides directly specifying eigenvalues, other approaches to choosing them are also possible. K It is conceivable; in many cases, the well-known LQR approach is used for this purpose.

[0043] If the aforementioned observer differential equation is further modified by adding the vector-valued initial equation... ω ^ P T ^ P = 0 0 1 0 z 0 x c S x ^ z ^ Extended, the desired estimated values ​​for the test specimen angular velocity are obtained. ω̂ P and test specimen torque T̂ P as outputs of the completed state observer.

[0044] A particularly noteworthy advantage of the method according to the invention is that the presented initial equation of the state observer can be easily extended, for example by adding an additional estimate. T̂ ST for the wave moment T̂ ST to determine. In a preferred embodiment of the present invention, the initial equation of the state observer is supplemented for this purpose by a line in which the estimated value is determined. T̂ ST The known parameters of the wave stiffness c and the wave damping d are used, which leads to the initial equation. ω ^ P T ^ P T ^ ST = 0 0 1 0 z 0 x c S − c − d d 0 z x ^ z ^ This leads to further results. Other output variables can be generated in an analogous way, for example, for an estimated value of the load angular velocity ω̂. D Combining several existing output variables to create new output variables can also be advantageous in this context. The observer differential equation of the described embodiment shows that the air gap torque T L and the load angular velocity ω D The following quantities are entered into this system as input variables, which is usually not a problem, since both quantities are only unknown in extremely unusual cases. However, it should be noted here that both the air gap torque T L as well as the load angular velocity ω D in the implementation of the method according to the invention, other angular velocities or torques predominant in the test rig arrangement 4 can be replaced, for example by the shaft torque. T ST and an angular velocity of the wave link 3. In such cases, only the first system of differential equations needs to be adjusted accordingly; however, the basic procedure remains unchanged. By adjusting the first system of differential equations, the implementation of a control variable can also be achieved. T D into an air gap moment T L This must be taken into account. Typically, the relationship between the control variable is considered. T D and air gap moment T L modeled by a combination of dead time and PT1 behavior, where dead time can advantageously be described using a Padé approximation.

[0045] As previously described, the angular frequencies ζ k as functions of the load angular velocity ω D or as functions of another rotational quantity predominant in the test rig arrangement 4, and the correction gain K is, as is often the case, a function of the other parameters of the dynamics matrix  If this is assumed, a dependence of the resulting state observer on the load angular velocity ω also arises directly. D or from other rotary test bench parameters, of which the angular frequencies ζ k depend. In such cases, the state observer is automatically adapted to a potentially changing load angular velocity ω. D or adapted to a potentially changing, different rotational size of the test bench arrangement 4, which can improve the operating behavior of the test bench arrangement 4 to a significant degree.

[0046] How the estimated values ω P and T̂ P The following will demonstrate how the following can be used to regulate test bench arrangement 4: Figuren 4 and 5 shown. This shows Fig. 4 A block diagram of a control loop, which in this form is used particularly in electric motor test benches for so-called highly dynamic shaft torque control. This control loop is implemented, for example, in Automation System 5, preferably as software on microprocessor-based hardware of Automation System 5. However, other implementations are also conceivable, for example in the form of an integrated circuit (e.g., FPGA, ASIC, etc.).

[0047] The goal of such a highly dynamic shaft torque control is usually to reduce the shaft torque transmitted by the shaft connection 3. T ST predefined time profiles of a target wave moment T ST ∗ to track, whereby such target shaft torques can have highly dynamic components and can change fundamentally within a few milliseconds. In this context, the torques and angular velocities to be controlled are often also referred to as "control torque" and "control angular velocity." To solve the aforementioned control task, the following are typically started from the test specimen torque. T P and the current target wave moment T ST ∗ a target angular velocity ω* is determined, which is in the Fig. 4 The implementation shown in the Mass Simulation block MS is carried out. This target angular velocity ω* is subsequently regulated by the controller ωR and is usually chosen such that the resulting wave torque T ST the specified target value T ST ∗ assumes (with a specific, predetermined deviation from the standard). Since the test specimen torque T P Since the wave moment is usually unknown for the reasons mentioned above, a Kalman filter KF is often used to determine it, which includes, among other things, the measured wave moment. T ST acts as an entrance. The one in Fig. 4 The conventional implementation of this highly dynamic shaft torque control shown thus requires that the load angular velocity be determined on the test bench. ω D , the test specimen angular velocity ω P as well as the wave moment T ST The data are measured and made available to the test bench control system.

[0048] For the reasons mentioned above, however, the test specimen angular velocity is often not taken into account, especially in electric motor test benches. ω P still the wave moment T ST measured, which is why this approach is often not possible. Furthermore, for safety reasons, the power delivered by test specimen 1 often needs to be monitored, which also includes the torque generated by test specimen 1. T P This should be known. Furthermore, rotary encoders used on electric motors, where present, often deliver the desired measurement signal at very low sampling rates, such as 100 Hz. Since modern test bench controllers are typically designed with sampling rates of at least 5 kHz, this fact alone can significantly impair the dynamics of the test bench controller used.

[0049] This allows the concept of shaft torque control to function even without measuring the angular velocity of the test specimen. ω P To be used, the existing control concept is extended by the state observer according to the invention. This allows the in Fig. 5 The block diagram shown can be implemented, thus making the use of highly dynamic shaft torque control possible even in cases where neither the test specimen angular velocity is known. ω P still wave moment T ST as measured variables. A further advantage of the method according to the invention often arises from the fact that the previously described Kalman filter KF can be used for the initial estimation of the test specimen torque. T P can be dispensed with.

[0050] In addition to the one based on the Figuren 4 and 5 In the application example described, the inventive method can be used to solve a multitude of other test-related control problems, such as mass simulation, damping control, etc. Besides these control-related problems, further areas of application include, in particular, other test bench topologies, especially multi-axis test benches. Such a setup is described in Fig. 6 As shown, the rotating test specimen 1, which can now, for example, represent a gearbox, is connected via at least one further mechanical shaft connection 3' to at least one further rotating load machine 2'. The at least one further rotating load machine 2' can be used as an actuator in the control of the test rig arrangement 4, in addition to the already existing rotating load machine 2, without limitation to the test specimen angular velocity estimates according to the invention. ω P and test specimen torque T P can be used.

Claims

1. Method for controlling a test bench arrangement (4) in which a rotating specimen (1) is connected to a rotating loading machine (2) via a mechanical shaft connection (3) and in which at least one angular velocity prevailing in the test bench arrangement (4) is measured, the rotational behavior of the test bench arrangement (4) and thus the dynamic behavior at least of the measured angular velocity and of a specimen angular velocity (ωρ) prevailing in the specimen (1) being modeled using a first system of differential equations, characterized in that - the specimen torque (Tρ) generated by the specimen (1) is modeled using a second system of differential equations, the direct component of the specimen torque (Tρ) being modeled on the basis of a first subsystem of the second system of differential equations, and m harmonics of the specimen torque (Tρ) being modeled on the basis of a number m of further oscillating subsystems of the second system of differential equations in the form of damped and / or undamped oscillators, - a state observer for estimating the specimen angular velocity (ωρ) and the specimen torque (Tρ) is designed based on the first and the second system of differential equations, the measured angular velocity representing an input of the state observer, - estimates of the specimen angular velocity (ω̂P) and of the specimen torque (T̂ρ) are determined by means of the state observer, and - the determined estimates of the specimen angular velocity (ω̂P) and of the specimen torque (T̂ρ) are used to control at least one control angular velocity prevailing in the test bench arrangement (4) and / or at least one control torque prevailing in the test bench arrangement (4).

2. Method according to claim 1, characterized in that, by means of the state observer, in addition to the estimates of the specimen angular velocity (ω̂P)and of the specimen torque (T̂ρ) an estimate of a shaft torque (T̂ST) and / or an estimate of a load angular velocity (ω̂P) are determined, which are also used to control at least one control angular velocity prevailing in the test bench arrangement (4) and / or at least one control torque prevailing in the test bench arrangement (4).

3. Method according to claim 1 or 2, characterized in that the parameters of the second system of differential equations for modeling the specimen torque (Tρ) are selected as functions of at least one rotational variable which changes during the operation of the test bench arrangement (4) and prevails in the test bench arrangement (4), in order to adapt the parameters of the second system of differential equations, and thus the second system of differential equations, to changes in the at least one rotational variable prevailing in the test bench arrangement (4).

4. Method according to claim 3, characterized in that the parameters of the state observer designed on the basis of the first and second system of differential equations are selected as functions of at least one rotational variable which changes during the operation of the test bench arrangement (4) and prevails in the test bench arrangement (4), in order to adapt the parameters of the state observer, and thus the state observer, to changes in the at least one rotational variable prevailing in the test bench arrangement (4).

5. Method according to any of the preceding claims, characterized in that the rotating specimen (1) is connected via at least one further mechanical shaft connection (3') to at least one further rotating loading machine (2'), and that the at least one further rotating loading machine (2') is used in addition to the existing rotating loading machine (2) as an actuator in the control of at least one control angular velocity prevailing in the test bench arrangement (4) and / or of at least one control torque prevailing in the test bench arrangement (4).

6. Test bench comprising a test bench arrangement (4) in which a rotating specimen (1) is connected to a rotating loading machine (2) via a mechanical shaft connection (3) and which is designed to detect, by measurement, at least one angular velocity prevailing in the test bench arrangement (4), and an automation system (5) which is designed to model the rotational behavior of the test bench arrangement (4) and thus the dynamic behavior of at least the measured angular velocity and a specimen angular velocity (ωρ) prevailing in the specimen (1) using a first system of differential equations, characterized in that the automation system (5) is further designed - to model the specimen torque (Tρ) generated by the specimen (1) using a second system of differential equations and thereby model the direct component of the specimen torque (Tρ) on the basis of a first subsystem of the second system of differential equations, and model m harmonics of the specimen torque (Tρ) on the basis of a number m of further oscillating subsystems of the second system of differential equations in the form of damped and / or undamped oscillators, - on the basis of the first and the second differential equations, to design a state observer for estimating the specimen angular velocity (ωρ) and the specimen torque (Tρ), the measured angular velocity representing an input of the state observer, - to determine the specimen angular velocity (ω̂P) and the specimen torque (T̂ρ) by means of the state observer, and - to use the determined estimates of specimen angular velocity (ω̂P) and specimen torque (T̂ρ) for controlling at least one control angular velocity prevailing in the test bench arrangement (4) and / or at least one control torque prevailing in the test bench arrangement (4).