Control method and control device with anomaly detection

Abstract

The embodiment of the invention relates to a control method and control equipment with anomaly detection. In order to provide a method for controlling a technical system (1) by means of a two-degree-of-freedom controller (2), which allows an improved accuracy, higher robustness and better safety to be achieved even in the event of a change in the technical system (1), an anomaly detection signal (w) is derived at an anomaly detection time (tw) during the control of the technical system (1), at least one control parameter (kp, ki, Umax) for parameterizing the two-degree-of-freedom control unit (2) is adjusted as a function of the anomaly detection signal (w), and the technical system (1) is controlled starting from an anomaly detection time (tw) by means of the two-degree-of-freedom control unit (2) parameterized by the at least one adjusted control parameter (kp, ki, Umax).

Description

TECHNICAL FIELD

[0001] The present invention relates to a method of controlling an output of a technical system to a predetermined setpoint by means of a two-degree-of-freedom controller having a feedforward part providing a first control output and a feedback part providing a second control output, the two-degree-of-freedom controller being parameterized by a plurality of control parameters, the feedback part computing the second control output as a function of a control error representing a deviation between the output of the technical system and the predetermined setpoint, the total sum control variable acting on the technical system to control the output to the setpoint being determined as a function of the first control output and the second control output.

[0002] Furthermore, the present invention relates to a control system having a control unit on which a two-degree-of-freedom controller is implemented to control an output of a technical system to a predetermined setpoint. BACKGROUND

[0003] Many technical applications are known in which a technical system has to be controlled by means of a controller. Examples of technical systems that need to be controlled are industrial robots having a robot arm and needing to control the movement of the robot arm, long-stator linear motors needing to control the movement of a shuttle, or engine test benches needing to control the torque generated by a dynamometer or a device under test, such as an internal combustion engine, etc. Typically, the set goal of the control problem is to control the output of the technical system to a predetermined reference value, such as a desired torque, a desired speed / position profile, a desired pressure, etc.

[0004] In the case described above, when designing and / or parameterizing a controller, many aspects have to be considered, including the static and / or dynamic behavior of the system to be controlled, potential modifications of the system, disturbances acting on the system, etc. Only a properly tuned controller can control a technical system sufficiently and satisfactorily over the entire operating range of the technical system, especially with respect to the resulting dynamic or steady-state control errors, or with respect to the resulting dynamics of the closed control loop, with respect to stability, controllability or other control measures, etc. Furthermore, in practice, many systems exhibit nonlinearities and change over time, making controller design and / or parameterization a difficult task. Therefore, in a first step, when designing a controller for a given technical system, usually only the structure of the controller (controller structure) is provided, and in a second step, the structure is parameterized (values are assigned to the controller parameters used in the structure) by tuning the controller parameters to achieve the desired (closed-loop) control behavior. Tuning the parameter values in the second step allows reacting to previously unknown nonlinearities and changes of the system. However, even in the case of using a simple controller structure, such as a PI controller or a PID controller or a sliding mode controller, the tuning of the controller parameters often proves to be complex and difficult due to nonlinearities of the system, or due to time-varying behavior and due to disturbances, etc. Therefore, usually highly trained and experienced experts are required to successfully parameterize a controller.

[0005] The above aspects are especially true for two-degree-of-freedom (2DoF) controllers, such as the two-degree-of-freedom controller known from CN 114779648 A. A 2DoF controller usually comprises a feedforward part providing a first control output and a feedback part providing a second control output, the feedback part computing the second control output depending on a control error representing a deviation between an output of the technical system to be controlled and a predetermined setpoint. A 2DoF controller has good reference tracking performance and robust disturbance rejection and is well suited for many applications. In the case of a proper parameterization, especially of the feedforward path, the feedforward path performs most of the control work, leaving only little work to the feedback part.

[0006] However, one problem often encountered with 2DoF controllers is that due to the fact that this type of controller usually uses the inverse of the system to be controlled in the feedforward part, the desired setpoint is reversed into a control signal, i.e. into a first control output, ensuring that only a small contribution of the feedback part is utilized to achieve the setpoint. In case the system to be controlled changes, e.g. due to disturbances acting on the system, or modifications of the system due to wear or aging, or nonlinear effects, the inverse model used in the feedforward path can deviate from the actual inverse model, i.e. from the correct inverse of the system at a given point in time, leading to wrong control actions of the feedforward path, requiring the feedback part to take over more work than planned and intended. This can lead to a performance decrease of the technical system controlled by the 2DoF controller, potentially leading to a reduced control accuracy, a reduced robustness, and even endangering the system to be controlled. SUMMARY

[0007] It is therefore an object of the present invention to provide a method of controlling a technical system by means of a two-degree-of-freedom controller, which allows to achieve an improved accuracy, a higher robustness and a better safety, even in case the system to be controlled changes.

[0008] For the method at the beginning of this text, the object is achieved in that at an anomaly detection point in time during the control of the output of the technical system, an anomaly detection signal is derived from a control error present at the anomaly detection point in time, at least one control parameter of a plurality of control parameters parameterizing the two-degree-of-freedom controller is adjusted in dependence on the anomaly detection signal, and starting from the anomaly detection point in time, the output of the technical system is controlled by means of the two-degree-of-freedom controller parameterized by the at least one adjusted control parameter. Preferably, as described above, the feedforward path employs an inverse model of the technical system to calculate the first control output from the setpoint. However, the present invention allows for flexibility in this respect, i.e. in case an inverse model of the technical system is used, such that the first control output can also be calculated outside the two-degree-of-freedom controller, most preferably by means of an inverse model of the technical system, and can be fed as an input to the two-degree-of-freedom controller and can simply be passed, i.e. fed forward, in the feedforward path. Preferably, the sum control variable is calculated from the first control output and the second control output by simply summing the first control output and the second control output. However, also in this respect, different approaches can be employed, e.g. scaling the first control output and / or the second control output with appropriately chosen scaling weights before summing them, or filtering the first control output and the second control output before calculating the sum control variable.

[0009] The core idea and / or the core insight of the present invention is that, in case of a technical system being controlled using a two-degree-of-freedom controller, the control error between the output of the technical system and the predetermined setpoint is indicative of a deviation between the assumed model behavior of the real technical system and the real behavior. In case of a (classical) controller not including a feedforward path or at least not including an initially correctly parameterized feedforward path, such a conclusion cannot be drawn. In case of a classical controller being used, the control error is even necessary to allow for generating any control action. Thus, in this case, the control error does not provide any insight into the presence of any kind of anomaly.

[0010] Due to the correlation between the control error and the anomaly, the information conveyed by the control error can be used to react to changes in the technical system to be controlled and, in turn, to improve the controller. In particular, in case of a deviation caused by an anomaly being detected by analyzing the control error, the controller and / or the parameters of the controller can be adjusted.

[0011] In a preferred embodiment of the present invention, the anomaly detection signal can be derived from the control error as an absolute value, preferably a scaled absolute value, or as a squared value, preferably a scaled squared value, or as an L-norm, preferably a scaled L-norm, of the control error. Depending on the specifics of a given use case, an appropriate norm can be used to overcome and optimally handle potential obstacles present in a particular use case, such as signal noise or dynamic disturbances or system disturbances.

[0012] For preparing the anomaly detection signal for further signal processing, the anomaly detection signal can also be filtered by a predetermined filter, preferably a low-pass filter, or a band-pass filter, or a high-pass filter, or a notch filter, before being used for adjusting at least one of the plurality of control parameters.

[0013] For adjusting the control parameter in response to identifying the anomaly, a preferably static linear function or a nonlinear function can be provided, describing a correlation between the anomaly detection signal and at least one of the plurality of control parameters to be adjusted. Based on such a linear function or a nonlinear function, the anomaly detection signal can be fed into the linear function or the nonlinear function, thereby calculating an adjusted control parameter value to adjust and, in turn, to update and improve the controller. The correlation between the anomaly detection signal and the control parameter can also be implemented in form of a dynamic function and, thus, a dynamic system, such as the filter mentioned above or the like. Furthermore, the correlation between the anomaly detection signal and the control parameter can also include a specific nonlinearity, particularly preferably a dead zone, suitable for the requirements of a given use case, to overcome, for example, signal noise, wherein, in case of the anomaly detection signal lying inside the dead zone, no adjustment or only a reduced adjustment is performed.

[0014] The invention also provides for flexibility in terms of the way the two-degree-of-freedom controller can be modified based on the anomaly detection signal. Specifically, in a further preferred embodiment, a limiting element parameterized by at least one limiting parameter, which is one of the plurality of control parameters used to parameterize the two-degree-of-freedom controller, can be provided in the two-degree-of-freedom controller to limit the first control output and / or to limit the second control output and / or to limit the sum control variable, and the at least one limiting parameter is adjusted in accordance with the anomaly detection signal. Limiting the output of the controller is particularly useful in cases where the root cause of the anomaly is not immediately apparent. In such cases, a more cautious control is often justified, and thus further limiting and reducing the available and thus applied control energy and / or control action to avoid causing damage. Particularly preferably, the feedback part can only comprise a proportional controller without an integral part. Omitting the integral part avoids potential problems caused by the well-known phenomenon of controller integral wind-up, especially in case the limiting parameter is modified, in particular reduced.

[0015] In the above-described embodiments, the adjustment of the at least one of the plurality of control parameters can only be performed in case the absolute value of the anomaly detection signal exceeds a predetermined anomaly threshold value to ensure that the adjustment, e.g. a reduction of the absolute value of the at least one of the plurality of control parameters of the two-degree-of-freedom controller, is only performed in case of a significant deviation and anomaly, thus avoiding constant and thus useless adjustments due to e.g. signal noise.

[0016] Furthermore, the invention allows for flexibility in terms of implementation details and the system to which the method is applied. Thus, a planar motor can be controlled as the technical system, or a long-stator linear motor can be controlled as the technical system, or an industrial robot can be controlled as the technical system, or a machine tool can be controlled as the technical system. In case an industrial robot is controlled as the technical system, the detected anomaly can particularly indicate a collision of the robot with a mechanical object.

[0017] In addition to the above-summarized method, the object mentioned at the outset is also achieved by a control system having a control unit on which a two-degree-of-freedom controller is implemented for controlling an output of a technical system to a predetermined setpoint, the control unit being capable of performing the above-described method according to the invention. BRIEF DESCRIPTION OF DRAWINGS

[0018] The invention is described in more detail below with reference to Figures 1 to 7 The invention is described in more detail below with reference to

[0019] Figure 1 A known structure of a 2DoF controller is shown,

[0020] Figure 2 The measured 2DoF control signal is shown.

[0021] Figures 3a to 3b The present invention illustrates the control error signal, anomaly threshold, and anomaly detection signal that can be obtained using this invention.

[0022] Figure 4 A first implementation of a 2DoF controller is shown.

[0023] Figure 5 A second implementation of the 2DoF controller is shown.

[0024] Figure 6 A first embodiment of the invention based on adjusted limit values ​​is shown.

[0025] Figure 7 A second embodiment of the invention based on adjusted limit values ​​is shown. Detailed Implementation

[0026] exist Figure 1 The diagram illustrates the general and well-known structure of a two-degree-of-freedom ("2DoF", 2DoF and "two degrees of freedom" are used synonymously hereinafter) controller 2 for controlling a control system 1. The two-degree-of-freedom controller 2 includes a feedforward section 3 and a feedback section 4. Since it is generally not possible to simultaneously achieve good setpoint tracking and fast disturbance suppression using ordinary feedback controllers with only one degree of freedom (such as classic P controllers, PI controllers, PID controllers, etc.), a more advanced control system can be used... Figure 1 The diagram shows a two-degree-of-freedom (2DoF) controller 2. The technical system 1 is, for example, a mechanical system driven by a total control variable u, providing the actual position as output y, which is preferably measured by an encoder, for example. More specifically, a planar motor PM, a long stator linear motor LLM, an industrial robot, a delta robot, or a machine tool can be controlled as technical system 1. Depending on the details of technical system 1, the total control variable u can be fed to an actuator, such as an amplifier, power amplifier, frequency converter, or servo drive, which ultimately acts on technical system 1 according to the total control variable u. Those skilled in the art familiar with the technical system 1 to be controlled will certainly know which actuator is best tuned for a particular use case. However, the total control variable u can also act directly on the technical system 1 to be controlled.

[0027] The objective of the two-degree-of-freedom controller shown is to control the output y of the illustrated technical system 1 to a predetermined setpoint y. set As is well known in control engineering, the feedforward section 3 and the feedback section 4 typically include parameters, such as the proportional gain k in the proportional path. por the integral gain ki of the integral path or the gain of an Anti-Windup scheme, e.g. the anti-windup scheme in feedback part 4 by Hanus (see Hanus, Raymond, Michel Kinnaert, and J-L. Henrotte. "Conditioning technique, a general anti-windup and bumpless transfer method." Automatica 23.6 (1987): 729-739) or the feedforward gain in feedforward part 3 or the limit value u limiting at least one, possibly more, control signals calculated in controller 2 max .

[0028] As is typical in control engineering practice, feedback part 4 is designed to calculate a second control output u y from a control error e FB , the control error e y denoting the deviation between the output y of technical system 1 and a predetermined setpoint y set , and two-degree-of-freedom controller 2 is designed to finally determine the sum control variable u acting on technical system 1 to control the output y to the setpoint y set from the first control output u FF and the second control output u FB . As can be seen from Figure 1 , the sum control variable u is calculated from the first control output u FF and the second control output u FB by summing the first control output u FF and the second control output u FB . In a mechanical system as technical system 1, the sum control variable u can correspond to a torque or force.

[0029] Controller design is usually performed by a (mathematical) model of technical system 1, which also includes certain state variables x i (i denotes an index, as is well known in control engineering). The model of technical system 1 maps the input (control variable u) to the output y. There can of course also be more than one output variable y. The state variables x i are usually stacked in a state vector x, where the dimension nd≥ 1, n describes the dimension of system 1. The 2DoF controller 2 often also utilizes certain state variables x i of technical system 1, e.g. the engine speed n in the case of an engine as technical system 1 to be controlled. The state variables x iThe state variable x can be measured with appropriate sensors or encoders, or can be calculated with a simulation model, or can be estimated from other known (e.g. measured) variables of the technical system 1 (e.g. from the output y) by means of an observer i .

[0030] As is also well known from the prior art, the feedforward part 3 can comprise the inverse of the technical system 1 to be controlled, i.e. the inverse of a model of the technical system 1. By means of the inverse model of the technical system 1, the desired setpoint y set in the feedforward part 3 can be calculated from the desired setpoint y FF , which has allowed to achieve the control objective, i.e. to control the output y to the setpoint y set in the ideal case without disturbances d and without deviations between the real behavior of the system 1 and the assumed behavior, which can be reflected in the mathematical model of the technical system 1.

[0031] For these considerations, in the case shown in Figure 1 , it is assumed that a disturbance d acts on the technical system 1, which has to be compensated by the controller and which can lead to deviations between the real behavior of the technical system 1 and the assumed behavior, which can be reflected in the mathematical model of the technical system 1. In the ideal case without disturbances (d = 0), the model parameters are identical to the real system parameters, u = u FF itself already forms the control signal which leads to y = y set . In this ideal case (feedforward part 3 is inverse of the real, acting system), the feedback part 4 is irrelevant (open-loop operation).

[0032] The feedback part 4 can comprise only a proportional controller k p *e y without an integral part, which allows to avoid problems which can potentially be associated with integral saturation phenomena. As will be explained later, this aspect is particularly advantageous in the scope of the present invention, since this avoids potential problems associated with the well-known integral saturation phenomena, see above. However, for the feedback part 4, other control architectures can also be conceivable, such as a PI controller or a PID controller or a sliding mode controller, etc.

[0033] Figure 2A 2DoF control signal, i.e. a sum control variable u, measured on a shaft of an industrial robot, is shown in the case of a mechanical system as a technical system 1 to be controlled. In the case of a robot, the set torque and the set position of a robot arm are centrally calculated on a central computing unit, which can be implemented in the form of a PLC, and the set torque and the set position are transmitted to a specific control unit via a field bus system, for example Ethernet or Ethercat or Ethernet Powerlink or Profibus, etc., on which a control structure as shown in Figure 1 is implemented to perform the control of the specific arm. Within the scope of the present invention, the control unit running the 2DoF controller 2 as shown in Figure 1 can be implemented in the form of an FPGA, a microcontroller, etc.

[0034] As can be seen from Figure 2 , the first control output u FF generated by the feedforward part 3 is highly similar to the sum control variable u FF finally output by the controller 2 in the case of a proper selection of the controller parameters. As already mentioned, the deviation between the sum control variable u y and the first control output u FF generated by the feedforward part 3 can be explained as the proportion of the control signal still to be applied by the feedback part 4 in order to achieve the control goal with sufficient accuracy. Ideally, the deviation is close to zero and essentially consists of oscillations caused by control behavior or model deviations or disturbances d not taken into account when establishing the model. In the ideal case of a perfect model of the technical system 1, in turn using a perfect inverse model in the feedforward part 3, the feedback part 3 would already solve the control task itself. Thus, the deviation between the first control output u y corresponding to the control error e w and the sum control variable u can be explained as a measure of model inaccuracies and / or disturbances. Inaccuracies, disturbances and other influences leading to deviations from the ideal model behavior are collectively referred to as "anomalies" in the following. In the case of too high inaccuracies, the controller 2 can no longer continue to function properly. Therefore, the present invention provides a method allowing a reaction to said anomalies, which lead to a modification of the technical system 1 and thus to said inaccuracies.

[0035] To this end, within the scope of the present invention, it is proposed that at an anomaly detection point in time t w during the output y of the control technical system 1, a control error e w existing at the anomaly detection point in time t y is derived, and that an anomaly detection signal w is derived from the control error e y in order to adjust a plurality of control parameters k p , k i , u of the parameterized two-degree-of-freedom controller 2 in accordance with the anomaly detection signal w.max at least one control parameter and from the abnormality detection time point t w onwards, the output y of the technical system 1 is controlled by means of a two- degree-of-freedom controller 2 parameterized with the at least one adjusted control parameter k p , k i , u max . Thus, the present invention implements a relatively simple way of detecting abnormalities and reacting accordingly (positively or actively).

[0036] Within the scope of the present invention, it is found that, in case of using a two-degree-of-freedom controller, the control error indicates a deviation between the assumed model behavior and the behavior of the real technical system. In case of a (classical) controller not including a feedforward path, such a conclusion cannot be drawn. In case of a classical controller, the control error is necessary to allow for generating any control input. Thus, in this case, the control error does not provide any insight into the existence of any kind of abnormality.

[0037] As is typical in modern control engineering, the two-degree-of-freedom controller 2 can be implemented as a discrete-time two-degree-of-freedom controller 2, the control output u FB , u FF being continuously calculated at equidistantly spaced discrete control time points t k , the discrete control time points typically being spaced apart by a constant predetermined sampling time T d , e.g. T d = 10 μs or T d = 100 μs or T d = 1 ms or T d = 10 ms etc., the abnormality detection signal w being continuously calculated at equidistantly spaced discrete detection time points t w . Advantageously, the discrete control time points t k and the detection time points t w may coincide.

[0038] For discussing the present invention in more detail, Figures 3a to 3b the control error signal e y , the abnormality threshold W and the abnormality detection signal w are shown, respectively, which can be obtained by the present invention. In particular, the abnormality threshold W can be provided to perform the adjustment of at least one control parameter k y , k p , u i only in case the absolute value of the control error e max exceeds the predetermined abnormality threshold W. As can be seen from Figure 3a , the abnormality detection signal w is set equal to the control error e yHowever, there are also other design options for deriving the anomaly detection signal w, which will be explained later. At the anomaly detection point in time t w , the detection signal w exceeds a predetermined anomaly threshold W, which is additionally emphasized with the status signal s.

[0039] It can also be seen from Figure 3a that the control parameter indicated by the parameters k Figure 3a and u p is reduced in max as soon as the anomaly threshold W is exceeded. It is particularly preferred that the control parameter is reduced. An important reason is that the anomaly indicates some deviation from the assumed model behavior. Therefore, in order to reduce the risk of problems, for example potential losses of stability, that can be caused by such a deviation, the control parameter is reduced to allow an increase of, for example, the stability margin, and thus to make the operation safer again. Reducing the control parameter allows to reject the controller as a whole.

[0040] Again in summary, for detecting the anomaly, a non-linear function can be provided to obtain the anomaly detection signal w:

[0041]

[0042] In the simplest case, the deviation of the actual output y from the target setpoint y set can be directly used for this purpose:

[0043] w := e = y set - y

[0044] If the signal quality allows, also higher order derivatives of the mentioned variables can be included:

[0045]

[0046] As shown in the above equation, the control error e y = y set - y, and its derivatives can of course also be scaled by appropriate scaling weights. Thus, the anomaly detection signal w can be derived from the control error e y as the absolute value |e| of the control error e y , preferably a scaled absolute value, or as the squared value e^2 of the control error e y , preferably a scaled squared value, or as the L-norm L(e) of the control error e y , preferably a scaled L-norm, thereby allowing high flexibility, especially when the invention needs to be adapted for a specific practical use case.

[0047] With regard to the correlation between the anomaly value and at least one control parameter, a linear or non-linear function is provided to describe the anomaly detection signal w in dependence on a plurality of control parameters k p , ki u max The correlation between at least one control parameter in the system. However, different approaches can be considered in this regard. For example, if the abnormal threshold W is exceeded, the summation control variable u can be frozen to a predetermined value, thereby effectively disabling the 2DoF controller 2.

[0048] In the above context, differentiation behavior can improve sensitivity and lead to a faster response to the state information described below. In the case of noisy signals, the f(...) function can also have an additional filtering effect. Specific deviations can also be used to create state signals s, as already... Figure 3a As shown in the image:

[0049]

[0050] Another approach is to make the threshold W asymmetric (using two values ​​(W1, W2)) or time-varying (W1(t), W2(t)), thereby enabling more independent adjustment of the detection.

[0051] As mentioned at the beginning of this document, a planar motor (PM) or a long stator linear motor (LLM) or an industrial robot, especially a delta robot, can be controlled as a technical system 1. This application can be used anywhere the electric drive shaft comes into contact with other machine parts, such as in packaging machines, robotics, planar motors (PM), long stator linear motors (LLM), etc. When an industrial robot is controlled as technical system 1, detected anomalies can specifically indicate a collision between the robot and a mechanical object. By applying the teachings of this invention, many types of anomalies can be detected, even unexpected unstable behavior caused by changes in the plant environment that lead to instability. This applies whether the disturbance originates from an external source or to changes in the parameters of the controlled system (changes in the mechanical structure). Specifically, collisions can be detected at an early stage. In the case of a collision, the control error e y Consequently, the anomaly detection signal w will suddenly increase. This state change will be detected, and control actions will be limited according to parameterization.

[0052] Specifically, in the case of industrial robots, this invention can detect the loss of a tripod arm, or detect erroneous parameterization as an anomaly, or detect erroneously parameterized tools, or detect mechanical changes over time (friction, bearing damage, loosening, etc.), and can also detect changes in the drive system (e.g., belt wear, blockage, slipping clutch, etc.) to indicate when the model parameters (run-in behavior of mechanical structures, temperature effects, process changes, etc.) should be adjusted.

[0053] Depending on the operating mode (centralized, distributed), other versions of the 2DoF controller can also be envisioned, especially such as... Figure 4and Figure 5 As shown. Specifically, as Figure 4 As shown, in the case of distributed control (single-axis operation), the feedforward control u FF It is based on the setpoint y in controller 2 itself. set Certainly. However, in the case of centralized control (such as when operating a set of axes, i.e., an "axis group"), feedforward control can also be centrally computed, i.e., computed outside the two-degree-of-freedom controller, thus the tuple [y set ,u FF It is fed to the corresponding driver, such as Figure 5 As shown. Therefore, the first control output u FF The calculation can be performed outside the two-degree-of-freedom controller 2, preferably through the inverse model of the technical system 1, which is then fed as input to the two-degree-of-freedom controller 2 and processed by the feedforward path 3. Alternatively, the feedforward path 3 can use the inverse model of the technical system 1, based on the setpoint y. set Calculate the first control output u FF .

[0054] Figure 6 Possible implementations of the invention based on adjusted constraint values ​​are further illustrated. This constraint can be implemented in the feedforward branch, the feedback branch, or it can apply to the entire manipulated variable, i.e., the summation control variable u. However, other options for implementing the constraint (symmetric, asymmetric, time-dependent, process-dependent, etc.) are also conceivable.

[0055] In addition, as in Figure 6 As shown, in the process of adjusting multiple control parameters k p k i u max Before at least one control parameter, the anomaly detection signal w can be filtered by a predetermined filter F in a preferred manner, preferably a low-pass filter, or a band-pass filter, or a high-pass filter, or a notch filter.

[0056] Similarly, Figure 6 As shown, after passing through filter F, outliers are connected to the limiting element lim. To implement this invention, as... Figure 6 The limiting element lim, parameterized by at least one limiting parameter lim1, shown can be set in the two-degree-of-freedom controller 2 to limit the first control output u. FF and / or limit the second control output u FB and / or limit the total control variable u, such that at least one limiting parameter u lim It can be considered as multiple control parameters k used to parameterize the two-degree-of-freedom controller 2 p k i u maxone of the two cases, and adjusting at least one limit parameter u in dependence on the anomaly detection signal w lim .

[0057] Figure 7 A second embodiment of the application is shown, which is based on an adjusted limit value. In Figure 6 and Figure 7 In both cases shown, one conceivable way of implementing the application is to provide a first normal value liml, lim2, lim3 of the limit value in the case where no anomaly is detected, and to provide a second set of values with a smaller absolute value in the case where an anomaly is detected and needs to be reduced.

Claims

1. A method of controlling an output (y) of a technical system (1) to a predetermined setpoint (y by means of a two-degree-of-freedom controller (2), the two-degree-of-freedom controller (2) having a feedforward part (3) providing a first control output (u and a feedback part (4) providing a second control output (u, the two-degree-of-freedom controller (2) being parameterized by a plurality of control parameters (k, k, u, the feedback part (4) calculating the second control output (u from a control error (e, the control error representing a deviation between the output (y) of the technical system (1) and the predetermined setpoint (y, the total sum control variable (u) acting on the technical system (1) to control the output (y) to the setpoint (y being determined from the first control output (u and the second control output (u, characterized in that set ) is calculated from the control error (e FF ) and the first control output (u FB ) is calculated from the control error (e p ) and the second control output (u i ) is calculated from the control error (e max ) and the total sum control variable (u y ) is calculated from the first control output (u FB ) and the second control output (u set ) is calculated from the control error (e FF ) and the total sum control variable (u FB ) is calculated from the first control output (u set ) and the second control output (u Anomaly detection time point (t) during the output (y) of the control system (1) w At point t, from the anomaly detection time point (t) w The control error (e) present at point ) y The abnormality detection signal (w) is derived, and the plurality of control parameters (k) used to parameterize the two-degree-of-freedom controller (2) are adjusted according to the abnormality detection signal (w). p k i u max At least one control parameter in ), and from the anomaly detection time point (t w (Starting with) the adjustment of at least one control parameter (k) p k i u max The parameterized two-degree-of-freedom controller (2) is used to control the output (y) of the control system (1).

2. The method of claim 1, wherein, The feedforward path (3) uses an inverse model of the technical system (1) to calculate the first control output (u FF ) from the setpoint (y set ).

3. The method of claim 1, wherein, The first control output (u FF ) is calculated outside the two-degree-of-freedom controller (2), preferably by means of an inverse model of the technical system (1), is fed as input to the two-degree-of-freedom controller (2) and is pre-processed by the feedforward path (3).

4. The method according to any of the preceding claims, characterized in that, by summing the first control output (u FF ) and the second control output (u FB ), the sum control variable (u) is calculated from the first control output (u FF ) and the second control output (u FB ).

5. The method according to any of the preceding claims, characterized in that, The anomaly detection signal (w) is derived from the control error (e y ) as an absolute value (|e|) of the control error (e y ), preferably a scaled absolute value, or the anomaly detection signal (w) is derived from the control error (e y ) as a squared value (e y ) of the control error (e 2 ), preferably a scaled squared value, or the anomaly detection signal (w) is derived from the control error (e y ) as an L-norm (L(e)) of the control error (e y ), preferably a scaled L-norm.

6. The method according to any of the preceding claims, characterized in that, before adjusting the at least one of the plurality of control parameters (k p , k i , u max ), filtering the anomaly detection signal (w) by means of a predetermined filter (F), which is preferably a low-pass filter, or a band-pass filter, or a high-pass filter, or a notch filter.

7. The method according to any of the preceding claims, characterized in that, A linear function or a non-linear function is provided to describe a correlation between the anomaly detection signal (w) and the at least one control parameter (k p , k i , u max ) from the plurality of control parameters (k p , k i , u max ).

8. The method according to any of the preceding claims, characterized in that, The two-degree-of-freedom controller (2) provides a limiting element (lim) parameterized by at least one limiting parameter (lim1) to limit the first control output (u). FF ) and / or limit the second control output (u FB ) and / or limit the summation control variable (u), the at least one limiting parameter (u) lim ) are the plurality of control parameters (k) used to parameterize the two-degree-of-freedom controller (2). p k i u max One of the following, and adjust the at least one limiting parameter (u) according to the anomaly detection signal (w). lim ).

9. The method according to any of the preceding claims, characterized in that, The two-degree-of-freedom controller (2) is implemented as a discrete-time two-degree-of-freedom controller (2), the control output (u FB , u FF ) being continuously calculated at equidistantly spaced discrete control time points (t k ), the anomaly detection signal (w) being continuously calculated at equidistantly spaced discrete detection time points (t w ).

10. The method of claim 9, wherein, The discrete control time point (t k ) and the detection time point (t w ) coincide.

11. The method according to any of the preceding claims, characterized in that, in case the absolute value of the abnormality detection signal (w) exceeds a predetermined abnormality threshold value (W), performing the adjustment of the at least one control parameter of the plurality of control parameters (k p , k i , u max ).

12. The method of claim 11, wherein, reducing the absolute value of the at least one of the plurality of control parameters (k p , k i , u max ) to adjust the at least one of the plurality of control parameters (k p , k i , u max ) in case the abnormality detection signal (w) exceeds the predetermined abnormality threshold value (W).

13. The method according to any of the preceding claims, characterized in that, A planar motor (PM) is controlled as the technical system (1), or a long-stator linear motor (LLM) is controlled as the technical system (1), or an industrial robot is controlled as the technical system (1), the detected anomaly preferably indicates a collision of the robot with a static mechanical object, or a machine tool is controlled as the technical system (1).

14. The method according to any of the preceding claims, characterized in that, The feedback part (4) only comprises a proportional controller (k p ) without an integral part.

15. A control system with a control unit on which a two-degree-of-freedom controller (2) is implemented to control an output (y) of a technical system (1) to a predetermined setpoint (y set ), the two-degree-of-freedom controller (2) having a feedforward part (3) providing a first control output (u FF ) and a feedback part (4) providing a second control output (u FB ), the two-degree-of-freedom controller (2) being parameterized by a plurality of control parameters (k p , k i , u max ), the feedback part (4) being designed to calculate the second control output (u FB ) as a function of a control error (e y ) representing a deviation between the output (y) of the technical system (1) and the predetermined setpoint (y set ), the two-degree-of-freedom controller (2) being further designed to determine, as a function of the first control output (u FF ) and the second control output (u FB ), a sum control variable (u) acting on the technical system (1) to control the output (y) to the setpoint (y set ), characterized in that, The control unit is further designed to derive, at an anomaly detection point (t w ) during the control of the output (y) of the technical system (1), an anomaly detection signal (w) from the control error (e w ) present at the anomaly detection point (t y ), to adjust at least one control parameter of the plurality of control parameters (k p , k i , u max ) used to parameterize the two-degree-of-freedom controller (2) in accordance with the anomaly detection signal (w), and to control the output (y) of the technical system (1) from the anomaly detection point (t w ) by means of the two-degree-of-freedom controller (2) parameterized by the at least one adjusted control parameter (k p , k i , u max ). ​​​​​​​​​​​​

Images