Method for determining the positions of movable measuring objects, displacement measuring system and valve arrangement

Abstract

A method for determining the respective positions of at least one first movable object (16) and a second movable object (18) of a displacement measuring system (20) is described, wherein the second object (18) is arranged to be movable independently of the first object (16), in particular coaxially or translationally, and wherein the objects (16, 18) each generate a measurable signal. The displacement measuring system (20) comprises a sensor arrangement (26) with at least one sensor (28, 29, 30), wherein the sensor arrangement (26) provides an output signal in at least one first measuring direction. The objects (16, 18) are movable within the detection range of the at least one sensor (28, 29, 30).In one process step, a data set (36) containing position data (38) is provided for a multitude of predefined combinations of the position of the first object (16) and the second object (18) within the detection range of the sensors (28, 29, 30). The current output signals (e.g. 1x , B 1z , B 2x , B 2z , B 3x , B 3z The readings from the sensors (28, 29, 30) in the first and second measuring directions are recorded, and based on the current output signals, a machine learning model determines the current position of the first measuring object (16) and the second measuring object (18) based on the position data (38). Furthermore, a displacement measuring system (20) and a valve unit (10) are provided.

Description

[0001] The invention relates to a method for determining the respective position of at least one movably mounted measuring object and a second movably mounted measuring object of a displacement measuring system, as well as a displacement measuring system and a valve arrangement. The method and the displacement measuring system are used in particular to determine the opening state of a valve, especially a double-seat valve. Furthermore, the invention relates to a displacement measuring system and a valve arrangement.

[0002] Position measuring systems for valves and corresponding methods for determining the positions of movable measuring objects of a corresponding position measuring system are generally known.

[0003] In a displacement measurement system with multiple measurement objects, the respective position of each measurement object is determined separately in its own signal processing path, which means increased effort when there are multiple measurement objects.

[0004] It is therefore an object of the invention to provide an optimized method for determining the respective position of several measuring objects of a displacement measuring system.

[0005] This problem is solved according to the invention by a method for determining the respective position of at least one first movable measuring object and a second movable measuring object of a displacement measuring system. The method is particularly useful for determining the opening state of a valve, for example, a double-seat valve.

[0006] The second object being measured is arranged independently of the first object being movable relative to it, in particular coaxially or translationally, wherein the objects being measured each generate a measurable signal, in particular a magnetic field. The objects being measured are, for example, mounted linearly, rotatingly, and / or circumferentially on a circular path.

[0007] The displacement measuring system comprises a sensor arrangement with at least one sensor, wherein the sensor device provides an output signal in at least a first measuring direction and a second measuring direction. The objects being measured are movable within the detection range of the at least one sensor.

[0008] In a process step of the method according to the invention, a data set containing position data for a plurality of predetermined combinations of the position of the first and second objects within the detection range of the at least one sensor is provided. The current output signals of the at least one sensor in the first and second measuring directions are acquired, and based on these current output signals, a machine learning model determines the current position of the first and second objects using the position data.

[0009] An advantage of the method according to the invention is that, using the machine learning method, the current position of the measured objects can be determined particularly quickly and with high accuracy based on the current output signals and the position data. Transformation of the output signals or further calculations with the output signals are not strictly necessary. Specifically, the output signals do not need to be processed in a separate signal processing path. Furthermore, despite the use of a machine learning model and the application of artificial intelligence, relatively low computing power is sufficient to carry out the method.

[0010] Capturing the output signals of the sensor array in at least two measurement directions allows for a sufficiently accurate assignment of the measured output signals to a specific position of each of the first and second objects being measured. The assignment becomes more precise the greater the distance between the first and second objects. This is because, with increasing distance, the portion of the output signal caused by one specific object becomes more easily distinguishable from the portion caused by another. Thus, at least in areas where the objects are sufficiently far apart, every possible combination of output signals can be assigned to a specific position of the first and second objects.Ambiguity in the output signals can only occur if the objects being measured are arranged very close to each other, for example, less than 10 mm apart. Such ambiguity can be minimized through design measures.

[0011] If the sensor arrangement contains only one sensor, the sensor provides at least one output signal in a first and a second measuring direction. If multiple sensors are present, the sensors provide one output signal in at least one measuring direction.

[0012] Optionally, the process can be implemented in two stages: in the first stage, the machine learning model approximately determines the position of the objects being measured, and in the second stage, the accuracy of the position determination is increased. This can be achieved by further processing the portion of the output signals assigned to a specific object in the second stage, for example, in a signal processing path.

[0013] The position data, for example, contains a large number of data points, with each data point containing a comparison value for each individual output signal from each sensor. For example, with a total of four sensors, each with two output signals, the data set contains eight comparison values.

[0014] Additional measuring objects can be added to the measuring system if required.

[0015] For example, the position data is provided once in advance and only needs to be read out for a current position determination, which also contributes to accelerating the process.

[0016] According to one embodiment, the position data is generated by having the first and second objects traverse all possible combinations of positions along the measuring range. The sensor output signals can be recorded for each possible position of the objects. In this way, the position data can be generated particularly easily and with high accuracy.

[0017] The set of position data includes, for example, a family of characteristic curves that are tailored to an individual displacement measurement system.

[0018] For example, to generate position data, the first object is moved step by step into successive, discrete positions, and the second object is moved completely along its possible range of motion for each discrete position of the first object. In this way, all possible positions of the two objects that can occur during operation of the linear encoder system can be reached.

[0019] The resolution is, for example, 0.5 mm, meaning that the first object being measured is moved in 0.5 mm increments to create the position data.

[0020] The second object being measured is moved, for example, to within 5 mm of the first object. Such a minimum distance is generally always present between the objects being measured due to their spatial dimensions.

[0021] It is conceivable that the position data is generated outside the linear encoder and transferred to and stored in a control unit of the linear encoder. This has the advantage that a set of position data suitable for a large number of linear encoders of the same type only needs to be created once. This applies, for example, to linear encoders with sensors and measuring objects from the same batch from a manufacturer.

[0022] In an alternative embodiment, the position data can be calculated. Analytical or empirically determined formulas can be used for this purpose. For example, a sufficiently accurate, empirically determined fitting curve for a measured characteristic curve is conceivable.

[0023] A combination of both methods of providing the position data is also conceivable.

[0024] The selection of the characteristic curves and parameters to be used from the set of position data and / or the determination of the position based on the position data itself is carried out, for example, by a suitable control routine. This is usually stored in the linear encoder itself, for example, as software in a control unit of the linear encoder.

[0025] According to one embodiment, the at least one sensor is a magnetic field sensor, more precisely a Hall sensor. Hall sensors can detect changes in the magnetic field very precisely and are therefore particularly well suited for displacement measurement systems.

[0026] The current position of each of the two objects can be determined using a machine learning method, in particular a nearest-neighbor classification or regression. This allows the position of the objects to be determined even faster, as fewer data points need to be compared.

[0027] In nearest neighbor regression, the current output signals are compared with the position data, and based on the comparison, a current position of the first and second measurement objects is determined.

[0028] Alternatively, random forest classification or regression can be applied. Other suitable methods of shallow learning or deep learning (for example, using neural networks), ridge regression, or support vector machines are also conceivable. All these methods, using a set of positional data, lead to an accurate determination of the current position of both objects.

[0029] In addition, artificial intelligence methods can also be used to detect any inconsistencies or anomalies in the data set containing position data and to hide or correct them.

[0030] According to one embodiment, the amount of position data obtained is reduced by a downsampling method. Downsampling reduces the measurement resolution, but this can be compensated for by weighting in nearest-neighbor regression or other suitable shallow or deep learning methods. This variant is particularly advantageous when the method is to be performed on a microcontroller within the linear encoder itself, and not in a spatially separate control device located away from the encoder.

[0031] The process is carried out, for example, on a microcontroller, which is typically mounted on a circuit board on which the sensors are attached. This contributes to a particularly compact design.

[0032] However, the microcontroller can also be located on a separate circuit board.

[0033] The object is further solved according to the invention by a displacement measuring system, in particular for a valve arrangement, with at least one movably mounted first measuring object and a movably mounted second measuring object, wherein the second measuring object is arranged to be movable independently of the first measuring object relative to it, in particular coaxially or translationally to it, and wherein the measuring objects each generate a measurable signal, in particular a magnetic field. The displacement measuring system further comprises a sensor arrangement comprising at least one sensor, wherein the measuring objects are movable within the detection range of the at least one sensor, and a control unit configured to acquire the output signals of the at least one sensor.The control unit contains a data set with position data for a multitude of predefined combinations of the positions of the first and second objects within the detection range of the at least one sensor. The control unit also contains a machine learning model configured to determine the current positions of the first and second objects based on the position data and the current output signals of the at least one sensor. Thus, the displacement measuring system is configured to perform the previously described method according to the invention. The advantages described in connection with the method also apply accordingly to the displacement measuring system.

[0034] The at least one sensor can comprise at least two sub-sensors that detect magnetic field components in a first measurement direction and a second measurement direction orthogonal to this measurement direction, and each provide output signals in the first and second measurement directions. Alternatively, at least two sensors can be present, each capable of detecting a magnetic field component in one measurement direction, with the measurement directions being orthogonal to each other.

[0035] The first measurement direction is defined, for example, as the direction that coincides with the direction of movement of the object being measured. It should be noted that the sub-sensor located within the sensor's surface whose measurement direction is perpendicular to the object's direction of movement typically delivers only a very weak signal, since the object is preferably positioned centrally with respect to the Hall sensor's surface.

[0036] Therefore, the sensor sub-sensor with the measurement direction on the sensor surface that produces a strong signal is naturally selected. In this application, this direction is arbitrarily defined as the x-direction of the displacement measuring system.

[0037] The second measurement direction is the measurement direction perpendicular to the first measurement direction and extending into the depth of the sensor, which is defined in this application as the z-direction.

[0038] Due to manufacturing processes, the properties of the sub-sensors for the individual spatial directions, such as sensitivity, offset, and drift, vary in sensors integrated into semiconductor chips, especially Hall sensors. The y- and x-sub-sensors, located on the surface of the sensor, exhibit similar properties, while the properties of the z-sub-sensor, located in the depth of the Hall sensor, differ more significantly.

[0039] Thus, the measurement signals in the x and z directions differ significantly and are therefore easily distinguishable. For this reason, choosing the x and z directions as the first and second measurement directions is advantageous.

[0040] In addition, the y-direction can of course be used as a third measurement direction.

[0041] It is also conceivable that there are several sensors, each encompassing a magnetic field component in only one measurement direction.

[0042] According to one embodiment, the multiple sensors are arranged along the movement path of the measured objects, for example, one behind the other on a straight line. The distances between the sensors can be the same or different. These distances can be taken into account in the position data.

[0043] For example, three or more sensors are provided in the displacement measuring system. This allows for a sufficiently large detection range for at least two objects being measured.

[0044] The objects being measured are each formed, for example, by a permanent magnet with exactly two poles, which is linearly polarized. Such magnets are readily available and inexpensive.

[0045] The problem is further solved according to the invention by a valve arrangement, in particular a double-seat valve, with a control unit comprising a displacement measuring system according to the invention, and with a first valve spindle and a second valve spindle, wherein the objects to be measured are each arranged at an end of the valve spindles projecting into the detection range of the at least one sensor. In such a valve arrangement, the open state of the valve arrangement can be determined by means of the displacement measuring system based on the position of the objects to be measured.

[0046] The valve arrangement is, for example, a process valve or a control valve. In the case of a process valve, the ends of the valve spindles, which are fitted with the measuring objects, can protrude into a control head of the process valve.

[0047] The second valve spindle is preferably designed as a hollow spindle, arranged concentrically to the first valve spindle and movable independently of the first valve spindle. This also contributes to a compact design of the valve.

[0048] However, it is also conceivable that the valve spindles are arranged next to each other in a different valve design.

[0049] Further advantages and features of the invention will become apparent from the following description and the accompanying drawings. The drawings show: - Fig. 1 schematically a part of a valve arrangement according to the invention with a displacement measuring system according to the invention, - Fig. 2 a travel path of the first and second measuring objects when creating the position data, - Fig. 3 schematically a displacement measuring system according to the invention, and - Fig. 4 output signals from sensors of the displacement measuring system for each of the two measured objects.

[0050] Fig. Figure 1 schematically shows part of a valve arrangement 10.

[0051] For example, the valve arrangement 10 is a process valve or a control valve. In the exemplary embodiment, a process valve is illustrated, wherein the valve arrangement 10 in the exemplary embodiment is a double-seat valve.

[0052] The valve arrangement 10 can be pneumatically, hydraulically or electrically driven.

[0053] The valve arrangement 10 comprises a first valve spindle 12 and a second valve spindle 14.

[0054] The second valve spindle 14 is designed as a hollow spindle and is arranged concentrically to the first valve spindle 12.

[0055] At one end of the valve spindles 12, 14, which is not shown in the figures for the sake of simplicity, a closing element is arranged which interacts with a valve seat.

[0056] The general operating principle of double seat valves is well known, therefore a detailed description is omitted below.

[0057] A double-seat valve allows two separate lines to be selectively connected or disconnected fluidically. Such valves are used, for example, in food processing plants, as they enable cleaning of the system during operation. To clean one line, the connection to the other line simply needs to be closed.

[0058] At one end of the valve spindles 12, 14, which is facing away from a valve seat, a first measuring object 16 and a second measuring object 18 are arranged.

[0059] Consequently, both measuring objects 16, 18 are mounted for linear movement. More precisely, the second measuring object 18 is arranged independently of the first measuring object 16 and is coaxially movable to it.

[0060] In an alternative embodiment, which is not shown for the sake of simplicity, the objects being measured can also be mounted on a circular path and / or rotating or in another way.

[0061] The measuring objects 16, 18 are part of a displacement measuring system 20 for the valve arrangement 10.

[0062] The objects being measured, 16 and 18, are magnets, more precisely permanent magnets.

[0063] The objects being measured, 16 and 18, each generate a measurable signal, in particular a magnetic field.

[0064] Furthermore, the valve arrangement 10 includes a control unit 22. In the exemplary embodiment, the control unit 22 is implemented by a control head 24 of the valve arrangement 10.

[0065] In this case, the position measuring system 20 is integrated into the control head 24. However, it is also conceivable that the control unit 22 or the position measuring system 20 are arranged separately from the valve assembly 10.

[0066] The displacement measuring system 20 comprises, in addition to the measuring objects 16, 18, a sensor arrangement 26, which includes several sensors 28, 29, 30, in particular magnetic field sensors, in the exemplary embodiment three Hall sensors. The measuring objects 16, 18 are movable within the detection range of the sensors 28, 29, 30.

[0067] More precisely, the valve spindles 12, 14, with their end to which one of the measuring objects 16, 18 is attached, protrude into the detection range of the sensors 28, 29, 30, in particular into the control head 24.

[0068] Optionally, at least one of the sensors 28, 29, 30 is arranged such that the range of motion of both measured objects 16, 18 overlaps at least partially with the detection range of this sensor. In the exemplary embodiment, this applies to the second sensor 29 when viewed from below.

[0069] Sensors 28, 29, and 30 are arranged on a common circuit board 32. Circuit board 32 is, for example, a circuit board already present in the control unit 22.

[0070] A microcontroller 33 can also be arranged on the circuit board 32.

[0071] In the direction of movement of the measuring objects 16, 18, the sensors 28, 29, 30 are arranged in a series one after the other, in particular such that the detection ranges of two adjacent sensors 28, 29, 30 overlap.

[0072] Furthermore, the displacement measuring system 20 includes a control unit 34. In the exemplary embodiment, the control unit 34 is also integrated into the control unit 22, which is designed as a control head 24. However, it is also conceivable to house the control unit 34 outside the valve assembly 10, whereby the control unit 34 can be connected to the control unit 22, in particular the control head 24, via a signal line.

[0073] The control unit 34 is set up to capture the output signals of the sensors 28, 29, 30 and to determine the position of each of the two measuring objects 16, 18 based on the output signals of the sensors 28, 29, 30.

[0074] For this purpose, a data record 36 containing position data 38 for a multitude of predefined combinations of the positions of the first measurement object 16 and the second measurement object 18 within the detection range of the sensors 28, 29, 30 is stored in the control unit 34. This means that for every possible combination of positions of the measurement objects 16, 18, there exists an associated data point in the position data 38, which can be assigned to a position of the measurement objects 16, 18.

[0075] To determine the position of each of the two measurement objects 16, 18, a machine learning model is stored in the control unit 34, which is set up to determine a current position of the first measurement object 16 and the second measurement object 18 based on the current output signals of the sensors 28, 29, 30 and the position data 38.

[0076] A method for determining the position of the measured objects 16, 18 is described below using the following: Fig. 2 to 4 explained.

[0077] First, a data set with position data 38 is provided for a large number of predefined combinations of a position of the first measurement object 16 and the second measurement object 18 within the detection range of the sensors 28, 29, 30.

[0078] One way to provide the position data 38 is to calculate it.

[0079] Another way to provide the position data 38 is to determine it through measurements, as described in Fig. 2 is illustrated.

[0080] To generate the position data, the first and second measurement objects 16, 18 traverse all possible combinations of positions along the measuring range. The output signals from sensors 28, 29, 30 are recorded and stored. Once the measurement objects 16, 18 have traversed all possible combinations of positions, a uniquely identifiable data point of the position data 38 is available for each possible position of the measurement objects 16, 18.

[0081] In the illustrated embodiment, the first object 16 is moved stepwise into successive, discrete positions to generate the position data 38, and the second object 18 is moved completely along its possible range of motion for each discrete position of the first object 16. In this way, all possible combinations of positions can be systematically approached.

[0082] The range of motion of the second measuring object 18 increases stepwise as the first measuring object 16 is moved, as described in Fig. 2 is evident. Fig. 2. For illustrative purposes, the individual motion curves are provided with the reference symbols corresponding to the assigned measuring object.

[0083] The position data 38, for example, are created outside the distance measuring system 20 and transferred to the control unit 34 of the distance measuring system 20 and stored there.

[0084] Fig. Figure 3 schematically illustrates the displacement measuring system 20. Fig. 1.

[0085] Out of Fig. 3 shows that each sensor 28, 29, 30 comprises at least two sub-sensors 40, 42, which measure magnetic field components in a first measurement direction B x and a second one perpendicular to this measurement direction B z capture and output signals in the first and second measurement directions B x , B zdelivery.

[0086] Specifically, sensor 30 delivers the output signals B 1x , B 1z , sensor 29 provides the output signals B 2x , B 2z and sensor 28 provides the output signals B 3x , B 3z .

[0087] It is conceivable that each sensor 28, 29, 30 additionally transmits a signal B 1y , B 2y , B 3y in a third measuring direction B y delivers.

[0088] Fig. 4 shows the output signals B 1x , B 1z , B 2x , B 2z , B 3x , B 3z The sensors 28, 29, 30 for the first measurement object 16 (left) and the second measurement object 18 (right) when these are moved separately along the entire measurement path. More precisely, the output signals are split for better illustration to show the respective contributions of the measurement objects 16, 18 to a total output signal.

[0089] In the exemplary embodiment, the measured objects 16, 18, considered individually, generate an identical output signal.

[0090] Since the measured objects 16, 18 in the displacement measuring system 20 are always detected in combination with each other, the actual total output signal of each sensor 28, 29, 30 in a respective measuring direction results from a superposition of the two individual components of the output signals shown. The in Fig. The four separately shown output signals are therefore only for illustrative purposes; in reality, the components of the output signal attributable to the first measurement object 16 and the second measurement object 18 cannot be individually recorded, but only in the form of the total output signal.

[0091] The exact value of the superimposed signals depends on the exact positions of the measured objects 16, 18.

[0092] In Fig.Figure 4 illustrates an example scenario where the first object being measured is 16 in position p1 and the second object being measured is 18 in position p2.

[0093] The respective output signals of sensors 28, 29, 30 are illustrated by dots.

[0094] If the first object being measured 16 is in position p1 and the second object being measured 18 is in position p2, the total output signal of a sensor 28, 29, 30 is obtained by superimposing the two individual output signals.

[0095] To determine the position of the individual measuring objects 16, 18, the current output signals B are used. 1x , B 1z , B 2x , B 2z , B 3x , B 3z The sensors 28, 29, 30 are recorded in the first measuring direction and the second measuring direction and optionally in a third measuring direction, more precisely the respective superimposed total output signals.

[0096] Based on the current output signals B 1x , B 1z , B 2x , B 2z , B 3x , B 3z A machine learning model is used to determine the current position of the first measurement object 16 and the second measurement object 18 based on the position data.

[0097] The following section uses nearest-neighbor classification as an example. However, nearest-neighbor regression, random forest classification or regression, or suitable shallow learning or deep learning methods are also conceivable.

[0098] In nearest neighbor classification, the current output signals B are first considered. 1x , B 1z , B 2x , B 2z , B 3x , B 3z Sensors 28, 29, 30 detected.

[0099] These output signals are then compared with each data point of the position data. Each data point contains one corresponding to each current output signal B. 1x , B 1z , B 2x , B 2z , B 3x , B 3z corresponding value.

[0100] For example, the output signals are processed in the form of a vector. In this case, the individual data points of the position data are also vectors.

[0101] For each comparison, a distance d is taken. i the current output signals are determined by the respective data point of the position data.

[0102] The data point of the position data that corresponds to all output signals B 1x , B 1z , B 2x , B 2z , B 3x , B 3zThe data point with the smallest distance best represents the current position of the measured objects 16, 18. This data point can be determined, for example, by determining the current position of each individual output signal B. 1x , B 1z , B 2x , B 2z , B 3x , B 3z A difference is calculated between this value and the corresponding value stored in the data point, and these differences are summed. The data point with the smallest difference best represents the position of the measured objects 16 and 18.

[0103] Nearest-neighbor classification makes it possible to reduce the resolution of the position data. Furthermore, determining the position of the first and second measured objects can be done particularly quickly, as fewer data points need to be compared.

[0104] It is also conceivable to perform an averaging over the nearest data points of the position data, for example, over the three nearest data points. This is also referred to as a k nearest neighbor (kNN) method, where k represents the number of data points compared. Such a procedure is recommended, for example, if the measurement data used to determine the position data is noisy or weak.

[0105] Another possible optimization is the reduction of position data through a downsampling method. This results in a reduced measurement resolution, which can, however, be compensated for by weighting the averaging. That is, the accuracy of the predicted position is improved by averaging over the most probable predicted positions, weighted by their respective probabilities.

[0106] This is particularly advantageous if the procedure is to be carried out on a microcontroller, especially on a microcontroller with limited memory.

[0107] In this embodiment, the method is carried out on the microcontroller 33, which is already integrated into the displacement measuring system 20 and on which the sensors 28, 29, 30 are also arranged. The microcontroller 33 does not have the capabilities of an external control device.

Claims

[1] Method for determining the respective position of at least one first movable measuring object (16) and one second movable measuring object (16) of a displacement measuring system (20), in particular a method for determining the opening state of a valve, in particular a double-seat valve, wherein the second object of measurement (18) is arranged to be movable independently of the first object of measurement (16) relative to it, in particular coaxially or translationally to it, and wherein the objects of measurement (16, 18) each generate a measurable signal, in particular a magnetic field, wherein the displacement measuring system (20) comprises a sensor arrangement (26) with at least one sensor (28, 29, 30), wherein the sensor arrangement (26) provides an output signal in at least one first measuring direction, and wherein the objects being measured (16, 18) are movable within the detection range of the at least one sensor (28, 29, 30), comprising the following steps: - a data set (36) with position data (38) for a multitude of predefined combinations of a position of the first measurement object (16) and the second measurement object (18) within the detection range of the at least one sensor (28, 29, 30) is provided, - the current output signals (B 1x , B 1z , B 2x , B 2z , B 3x , B 3z ) of at least one sensor (28, 29, 30) in the first measuring direction and the second measuring direction are recorded, and - based on the current output signals (B 1x , B 1z , B 2x , B 2z , B 3x , B 3z ) a current position of the first measurement object (16) and the second measurement object (18) is determined by a machine learning model based on the position data (38). [2] Method according to claim 1, characterized by, that the position data (38) are generated by the first and second measurement objects (16, 18) passing through all possible combinations of positions along the measurement range. [3] Method according to claim 2, characterized by , that the first measurement object (16) is moved step by step into successive, discrete positions to generate the position data (38) and the second measurement object (18) is moved completely along the possible range of motion for each discrete position of the first measurement object (16). [4] Method according to claim 2 or 3, characterized by , that the position data (38) are created outside the distance measuring system (20) and transferred to a control unit (34) of the distance measuring system (20) and stored there. [5] Method according to any one of the preceding claims, characterized by , that the at least one sensor (28, 29, 30) is a magnetic field sensor, in particular a Hall sensor. [6] Method according to any one of the preceding claims, characterized by , that a current position of the two measurement objects (16, 18) is determined according to a machine learning method, in particular a nearest neighbor classification or regression. [7] Method according to any one of the preceding claims, characterized by , that the procedure is carried out on a microcontroller (33) which is located in particular on a printed circuit board (32) on which the sensors (28, 29, 30) are mounted. [8] Displacement measuring system (20), in particular for a valve arrangement (10), with at least one movably mounted first measuring object (16) and a movably mounted second measuring object (18), wherein the second measuring object (18) is arranged to be movable independently of the first measuring object (16) relative to it, in particular coaxially or translationally to it, and wherein the measuring objects (16, 18) each generate a measurable signal, in particular a magnetic field, with a sensor arrangement (26) comprising at least one sensor (28, 29, 30), wherein the objects to be measured (16, 18) are movable within the detection range of the at least one sensor (28, 29, 30), and with a control unit (34) which is configured to capture the output signals of the at least one sensor (28, 29, 30), wherein a data set (36) with position data (38) for a plurality of predefined combinations of a position of the first measured object (16) and the second measured object (18) within the detection range of the at least one sensor (28, 29, 30) is stored in the control unit (34), and wherein a machine learning model is stored in the control unit (34) which is configured based on the current output signals (B 1x , B 1z , B 2x , B 2z , B 3x , B 3z) of the at least one sensor (28, 29, 30) based on the position data (38) to determine a current position of the first measurement object (16) and the second measurement object (18). [9] Displacement measuring system (20) according to claim 8, characterized by , that the at least one sensor (28, 29, 30) comprises at least two sub-sensors (40, 42) that detect magnetic field components in a first measurement direction and a second measurement direction orthogonal to this measurement direction and each provide output signals in the first and second measurement directions. [10] Valve arrangement (10), in particular double seat valve, with a control unit (22) comprising a displacement measuring system (20) according to claim 8 or 9, and with a first valve spindle (12) and a second valve spindle (14), wherein the measuring objects (16, 18) are each arranged at an end of the valve spindles (12, 14) projecting into the detection range of the at least one sensor (28, 29, 30). [11] Valve arrangement (10) according to claim 10, characterized by , that the second valve spindle (14) is designed as a hollow spindle, which is arranged concentrically to the first valve spindle (12) and is movable independently of the first valve spindle (12).

Images