DE102022108962A1 - Valve arrangement and method - Google Patents

Abstract

The invention relates to a valve arrangement (2) for ventilating a first pressure chamber (10) of a pneumatic actuator (3) in order to carry out an actuation of an actuator member (11) of the pneumatic actuator (3) in accordance with an actuation specification, in particular a position specification, wherein the valve arrangement (2) is designed to detect a first pressure curve over time (p10) of a first pressure assigned to the first pressure chamber (10) during actuation in order to provide first pressure curve data (61) which depict the first pressure curve over time, and wherein the valve arrangement (2) has a machine learning model (31) and is designed to determine movement quality information (70) using the machine learning model (31) based on the first pressure curve data (61). indicates a quality of a movement of the actuator member (11) carried out during actuation.

Description

The invention relates to a valve arrangement for ventilating a first pressure chamber of a pneumatic actuator in order to carry out an actuation of an actuator member of the pneumatic actuator in accordance with an actuation specification, in particular a position specification.

For example, the actuation specification specifies that the actuator member should be moved to a first end position, in particular starting from a second end position.

The actuation preferably takes place in an operating mode in which the consumption of compressed air is reduced. Such an operating mode should also be referred to as a compressed air saving mode. For example, during actuation in the first pressure chamber, a first target pressure is achieved which is lower than a maximum pressure that can be achieved by ventilating the first pressure chamber (for example the available supply pressure). For example, the valve arrangement is designed to end the ventilation of the first pressure chamber before the actuation specification is met, for example before the actuator member has reached the second end position, and/or before the pressure in the first pressure chamber reaches the maximum achievable pressure - for example the supply pressure - has reached. In order to move the actuator member from the second end position to the first end position, it is preferably not necessary to ventilate the first pressure chamber until the actuator member has reached the first end position. In order to save compressed air, the ventilation of the first pressure chamber is preferably ended before the actuator member has reached the first end position, and the actuator member then (nevertheless) still reaches the first end position.

The DE 10 2020 204 484 A1 relates to a compressed air supply device which is designed to calculate a ventilation period and to ventilate a first pressure chamber of a pneumatic actuator in accordance with the calculated ventilation period in order to cause an actuation of the actuator member in accordance with the actuation specification.

One object of the invention is to improve the operation and/or monitoring of the pneumatic actuator.

The object is achieved by a valve arrangement according to claim 1. The valve arrangement is designed to detect a first time pressure curve of a first pressure assigned to the first pressure chamber during the actuation of the actuator member in order to provide first pressure curve data which represent the first time pressure curve. The valve arrangement has a machine learning model. The valve arrangement is designed to use the machine learning model to determine movement quality information based on the pressure curve data, which indicates a quality of a movement of the actuator member carried out during actuation.

The movement of the actuator member during actuation can have a variety of different characteristics. For example, the movement can have different numbers and/or different speed fluctuations. Furthermore, the movement can start at different speeds and/or be completed at different speeds. The different characteristics can have one or more causes - e.g. a reduction in the compressed air used, i.e. in particular the compressed air saving mode, an orientation of the pneumatic actuator relative to gravity, a load on the pneumatic actuator, a dimensioning of the pneumatic actuator and/or wear on the pneumatic actuator.

By determining the movement quality information, at least one characteristic of the movement of the actuator member can be detected, based on the pressure curve data - in particular based on the first time pressure curve (and optionally a second time pressure curve). Appropriately, no position of the actuator member detected with a position sensor is used to determine the movement quality information. Optionally, there is no position sensor for detecting a position of the actuator member.

The movement quality information preferably includes one or more movement quality indicators that express a quality of the movement carried out by the actuator member. For example, a higher value of a motion quality metric represents higher quality or a lower value of a motion quality metric represents lower quality.

The operation of the pneumatic actuator can expediently be monitored using the movement quality information. Furthermore, it is possible, based on the movement quality information, to adapt ventilation of the pneumatic actuator, in particular of the first pressure chamber and/or a second pressure chamber, for further actuation of the actuator member, in particular in such a way that the movement quality information for the further actuation is better movement quality than for the previous movement. Optionally, the valve arrangement is designed to record the movement quality information using the machine learning model during the current actuation determine and adapt the current activity based on the movement quality information.

The valve arrangement determines the movement quality information using the machine learning model, in particular using an artificial neural network. It has been found that a machine learning model, in particular an artificial neural network, is well suited to mapping one or more temporal pressure curves onto the movement quality information.

Advantageous further training is the subject of the subclaims.

The invention further relates to a method for operating the valve arrangement, comprising the steps: carrying out the actuation of the actuator member of the pneumatic actuator, detecting, during the actuation, the first time course of pressure in order to provide the first pressure course data, determining, using the machine Learning model, the movement quality information based on the pressure curve data.

• 1 eine schematische Darstellung eines Systems mit einer Ventilanordnung,
• 2 ein Schaubild mit zeitlichen Geschwindigkeitsverläufen eines Aktorglieds,
• 3 ein Schaubild mit zeitlichen Druckverläufen eines pneumatischen Aktors,
• 4 ein Blockdiagramm eines Machine-Learning-Modells,
• 5 ein Blockdiagramm eines neuronalen Netzes, und
• 6 ein Flussdiagramm eines Verfahrens zum Betrieb einer Ventilanordnung.

Further exemplary details as well as exemplary embodiments are explained below with reference to the figures. This shows

• 1 a schematic representation of a system with a valve arrangement,
• 2 a diagram with the temporal speed curves of an actuator member,
• 3 a diagram with pressure curves over time of a pneumatic actuator,
• 4 a block diagram of a machine learning model,
• 5 a block diagram of a neural network, and
• 6 a flowchart of a method for operating a valve arrangement.

The 1 shows an exemplary system 1 with a valve arrangement 2 and a pneumatic actuator 3. Optionally, the system 1 further comprises a higher-level controller 4, a pneumatic line arrangement 5, a pressure sensor device 6, a compressed air source 7, a compressed air sink 8 and / or one of the pneumatic actuator 3 actuation object to be actuated 9.

The system 1 represents an exemplary application environment for the valve arrangement 2. The valve arrangement 2 can preferably also be provided on its own.

The system 1, in particular the valve arrangement 2, is expediently designed for industrial use, in particular for industrial automation, for example for factory automation and/or process automation. System 1 is, for example, an industrial automation system. The system 1, in particular the valve arrangement 2, is expediently designed to actuate, for example move, grip and/or tension, the actuating object 9 by means of the actuator 3 as part of an industrial process, in particular an industrial manufacturing process.

The valve arrangement 2 serves to ventilate a first pressure chamber 10 of the pneumatic actuator 3 in order to carry out an actuation of an actuator member 11 of the pneumatic actuator 3 according to an actuation specification. The actuation specification includes, for example, a position specification. The position specification specifies, for example, a position, in particular an end position, into which the actuator member 11 is to be moved.

The valve arrangement 2 has a valve device 14. The valve device 14 exemplarily includes a first working outlet 21. Optionally, the valve device 14 further comprises a second working outlet 22. The first working outlet 21 is connected to the line arrangement 5 via a first line 23, for example a hose first pressure chamber 10 of the pneumatic actuator 3 pneumatically connected. The second working output 22 is pneumatically connected to a second pressure chamber 20 of the pneumatic actuator 3 via a second line 24, for example a hose, of the line arrangement 5.

The valve device 14 is expediently designed to ventilate, vent and/or block the working outlets 21, 22 independently of one another.

The valve device 14 includes one or more valve units 15. The valve units 15 include, by way of example, a first valve unit 16, a second valve unit 17, a third valve unit 18 and/or a fourth valve unit 19. Purely by way of example, the valve units 15 are connected as a full bridge.

By way of example, the first valve unit 16 is a 2/2-way valve, in particular a 2/2-way switching valve. Preferably, several or all of the valve units 15 are designed as 2/2-way valves, in particular 2/2-way switching valves. By way of example, the first valve unit 16 and/or all valve units 15 are designed as piezo valves.

The first valve unit 16 is connected between the compressed air source 7 and the first working output 21. The second valve unit 17 is connected between the compressed air sink 8 and the first working output 21. The third valve unit 18 is connected between the compressed air source 7 and the second working output 22. The fourth valve unit 19 is connected between the compressed air sink 8 and the second working output 22.

According to an alternative embodiment, the valve units 15 can be designed as 2/3-way valves.

According to a further alternative embodiment, the first valve unit and/or the second valve unit is each designed as a 3/3-way switching valve. The first valve unit expediently provides the first working output connected to the first pressure chamber and the second valve unit provides the second working output connected to the second pressure chamber. The first pressure chamber can be selectively ventilated, vented or blocked via the first valve unit. The second pressure chamber can either be ventilated, vented or blocked via the second valve unit. The second pressure chamber and/or the second valve unit is optionally present.

The valve arrangement 2 is expediently designed as a valve island. The valve device 14 is expediently designed as a valve module, in particular as a disk-shaped valve module. According to a preferred embodiment, the valve arrangement 2 comprises a connection plate on which a plurality of disk-shaped valve modules are arranged next to one another, the valve device 14 being formed by one or two of the valve modules. A valve device 14 with expediently four valve units 15 is expediently present in each valve module.

The valve arrangement 2 expediently includes a control unit 12. The control unit 12 expediently includes a microcontroller or is designed as a microcontroller. The control unit 12 is expediently integrated in the connection plate or in a valve module. The control unit 12 is preferably arranged in a control module, which is expediently connected to the connection plate or a valve module (in particular communicatively). As an example, the control module is arranged on the connection plate. The control unit 12, in particular the control module, is expediently connected in a communicative manner between a communication connection 32 of the valve arrangement 2 and the valve device 14.

The actuator 3 is designed as an example of a pneumatic cylinder. As an example, the actuator 3 is designed to be double-acting. The actuator 3 includes the actuator member 11, which can be moved into two different positions, in particular two different end positions, by applying compressed air. The actuator 3 includes the first pressure chamber 10 and the second pressure chamber 20, through whose ventilation the actuator member 11 can be actuated. By way of example, the actuator member 11 can be moved into a first position, in particular a first end position, by venting the first pressure chamber 10 (and expediently venting the second pressure chamber 20). Furthermore, the actuator member 11 can be moved into a second position that is different from the first position, in particular a second end position, by venting the first pressure chamber 10 (and expediently venting the second pressure chamber 20).

According to a possible embodiment, the actuator 3 can be designed to be single-acting and, for example, comprise a spring element that pushes the actuator member 11 into the second position. In the single-acting version, the actuator 3 expediently does not include a second pressure chamber.

The actuator member 11 is designed, for example, as a piston arrangement and in particular comprises a piston 25 and/or a piston rod 26 coupled, in particular fastened, to the piston 25. The piston 25 delimits the first pressure chamber 10 (and, if present, the second pressure chamber 20). The actuating object 9 is coupled to the actuator member 11, in particular the piston rod 26, and is actuated by a movement of the actuator member 11.

The higher-level controller 4 is communicatively connected to the valve arrangement 2, in particular the control unit 12. By way of example, the higher-level controller 4 is communicatively connected to the communication connection 32, for example via a fieldbus. The higher-level controller 4 is designed, for example, as a programmable logic controller, PLC, and/or as a cloud server. The higher-level controller 4 serves in particular to provide the control unit 12 with the actuation specification.

The pressure sensor device 6 is used to detect a first pressure that is assigned to the first pressure chamber 10 and/or a second pressure that is assigned to the second pressure chamber 20. The first pressure expediently depends on a pressure present in the first pressure chamber 10. For example, the first pressure is the pressure present at the first work output 21. Optionally, the first pressure is the pressure present in the first pressure chamber 10. Conveniently, the second print depends on one in the two th pressure chamber 20 present pressure. For example, the second pressure is the pressure present at the second work outlet 22. Optionally, the second pressure is the pressure present in the second pressure chamber 20. The pressure sensor device 6 is expediently part of the valve arrangement 2. In particular, the pressure sensor device 6 is integrated in the valve arrangement 2. The pressure sensor device 6 is expediently communicatively connected to the control unit 12. Optionally, the valve arrangement 2 includes a circuit board on which the pressure sensor device 6 and the control unit 12 are arranged. The pressure sensor device 6 preferably comprises a pressure sensor per valve module or a pressure sensor per work output 21, 22. For example, the first line 23 (for example a first hose) is pneumatically connected between a first pressure sensor assigned to the first work output 21 and the first pressure chamber 10. By way of example, the first line 24 (for example a second hose) is pneumatically connected between a second pressure sensor assigned to the second working output 22 and the second pressure chamber 20.

Optionally, the pressure sensor device can be arranged on the actuator 3.

Preferably, the system 1, in particular the valve arrangement 2 and/or the actuator 3, does not include a position sensor and/or a position measuring system for measuring the position of the actuator member 3, in particular for measuring the position at and/or between the two end positions.

• Beispielsweise empfängt die Ventilanordnung 2 von der Steuerung 4 eine erste Betätigungsvorgabe, die vorgibt, dass das Aktorglied 11 von der zweiten Endlage in die erste Endlage zu versetzen ist. Gemäß der ersten Betätigungsvorgabe belüftet die Ventilanordnung 2 über den ersten Arbeitsausgang 21 die erste Druckkammer 10 und/oder die Ventilanordnung 2 entlüftet über den zweiten Arbeitsausgang 22 die zweite Druckkammer 20, so dass sich das Aktorglied von der zweiten Endlage in die erste Endlage bewegt. Die Bewegung des Aktorglieds 11 von der zweiten Endlage in die erste Endlage erfolgt in eine erste Bewegungsrichtung.

The operation of the valve arrangement 2 will be discussed in more detail below:

• For example, the valve arrangement 2 receives a first actuation specification from the controller 4, which specifies that the actuator member 11 is to be moved from the second end position to the first end position. According to the first actuation specification, the valve arrangement 2 ventilates the first pressure chamber 10 via the first working output 21 and/or the valve arrangement 2 vents the second pressure chamber 20 via the second working output 22, so that the actuator member moves from the second end position to the first end position. The movement of the actuator member 11 from the second end position to the first end position takes place in a first direction of movement.

For example, the valve arrangement 2 receives a second actuation specification from the controller 4, which specifies that the actuator member 11 is to be moved from the first end position to the second end position. According to the second actuation specification, the valve arrangement 2 ventilates the second pressure chamber 20 via the second working output 22 and/or the valve arrangement ventilates the first pressure chamber 10 via the second working output 22, so that the actuator member moves from the first end position to the second end position. The movement of the actuator member 11 from the first end position to the second end position takes place in a second direction of movement.

In the following, the pressure chamber that drives the actuator member 11 in the direction of movement of the movement to be carried out will be referred to as the drive chamber and/or the pressure chamber that counteracts the movement to be carried out by the actuator member 11 will be referred to as the counter chamber.

The valve arrangement 2 is preferably designed to adopt a compressed air saving mode. The valve arrangement 2 is expediently designed (particularly in the compressed air saving mode) to ventilate the drive chamber up to a target pressure and not beyond when the actuator member 11 is actuated. The target pressure is, for example, lower than a maximum pressure that can be achieved by ventilating the first pressure chamber 10. For example, the target pressure is lower than the supply pressure provided by the compressed air source 7.

The valve arrangement 2 is preferably designed to end the ventilation of the drive chamber when the actuator member 11 is actuated before the actuation specification is fulfilled, for example before the actuator member 11 has reached an end position specified by the actuation specification, and/or before the pressure in the Drive chamber has reached the maximum achievable pressure - for example the supply pressure.

In this way, the actuation specification (for example the displacement of the actuator member 11 into a predetermined end position) can be fulfilled with a reduced consumption of compressed air, in particular compared to the case in which the ventilation is stopped until the actuation specification is fulfilled - for example until the end of the Actuator member 11 movement to be carried out - is continued.

Optionally, pulsed ventilation of the first pressure chamber 10 and/or the second pressure chamber 20 can be used in the compressed air saving mode.

The 2 shows a diagram with a first speed curve 41 (shown as a solid line) and a second time speed curve 42 (shown as a dashed line) of the actuator member 11. The time is plotted on the horizontal axis of the diagram and the speed is on the vertical axis of the diagram of the actuator member 11 is applied.

The first speed curve 41 was recorded during a first movement of the actuator member 11 from the second end position to the first end position. The second speed curve 42 was recorded during a second movement of the actuator member 11 from the second end position to the first end position. By way of example, less compressed air was used to actuate the actuator member 11 for the second movement than for the first movement.

The first speed curve 41 has a higher quality of movement than the second speed curve 42. For example, in the first speed curve 41, the speed of the actuator member 11 fluctuates less strongly than in the second speed curve 42. Furthermore, in the first speed curve 41, the movement of the actuator member 11 is completed more quickly than in the second speed curve. The period of time from the start of ventilation of the drive chamber to the end of the movement of the actuator member - i.e., for example, until the actuator member has reached an end position - should also be referred to as the actuation period. By way of example, the actuation duration 43 of the first speed curve 41 is shorter than the actuation duration 44 of the second speed curve 42.

The 3 shows a diagram with time pressure curves of the pneumatic actuator 3. The time pressure curves include a first drive chamber pressure curve 45 (shown as a solid line) and a first time counter-chamber pressure curve 46 (shown as a solid line). The first drive chamber pressure curve 45 and the first counter-chamber pressure curve 46 were recorded during the first movement of the actuator member 11 - so they correspond to the first speed curve 41 2 associated. The 3 also shows a second drive chamber pressure curve 47 (shown as a dashed line) and a second temporal counter-chamber pressure curve 48 (shown as a solid line). The second drive chamber pressure curve 47 and the second counter-chamber pressure curve 48 were recorded during the second movement of the actuator member 11 - so they correspond to the second speed curve 42 2 associated.

As explained in more detail below, the valve arrangement 2 is preferably designed to determine the quality of movement based on temporal pressure curves (such as the pressure curves 45 and 46, or 47 and 48), and in particular without knowledge of the speed curves of the actuator member 11 and/or position curves of the actuator member 11 to determine the movement carried out by the actuator member 11. The quality of movement is expediently determined by the control unit 12.

The valve arrangement 2 is designed to detect a first pressure curve over time of the first pressure assigned to the first pressure chamber 10 during the actuation of the actuator member 11 in order to provide first pressure curve data 61 which represent the first pressure curve over time. For example, the valve arrangement 2 detects the first pressure curve over time with the pressure sensor device 6. The one in the 3 shown first drive chamber pressure curve 45 and the second drive chamber pressure curve 47 are examples of the recorded first time pressure curve. The valve arrangement 2 expediently provides the first time pressure curve and/or the first pressure curve data 61 as a respective time series of pressure values.

The valve arrangement 2 has a machine learning model 31. The valve arrangement 2 is designed to use the machine learning model 31 to determine movement quality information 70 based on the first pressure curve data, which indicates a quality of movement during actuation carried out movement of the actuator member 11 indicates. The machine learning model 31 is preferably stored on the control unit 12. The movement quality information 70 expediently relates to the entire movement of the actuator member 11 carried out during the actuation, i.e., for example, to the entire movement of the actuator member 11 from the second end position to the first end position.

Preferably, the valve arrangement 2 is further designed to detect a second pressure curve over time of the second pressure assigned to the second pressure chamber 20 during the actuation of the actuator member 11 in order to provide second pressure curve data 62 which represent the second pressure curve over time. For example, the valve arrangement 2 detects the second pressure curve over time with the pressure sensor device 6. The one in the 3 shown second drive chamber pressure curve 46 and the second drive chamber pressure curve 48 are examples of the detected second time pressure curve. The valve arrangement 2 expediently provides the second time pressure curve and/or the second pressure curve data 62 as a respective time series of pressure values.

The valve arrangement 2, in particular the control unit 12, is preferably designed to determine the movement quality information 70 using the machine learning model 31 based on the first pressure curve data 61 and the second pressure curve data 62.

The movement quality information 70 will be discussed in more detail below. The movement quality information 70 expediently includes one or more movement quality Key figures. For example, the movement quality information includes an actuation time 71, a fluctuation information 72, a rebound information 73 and/or a starting time 74.

First about the actuation time 71: The actuation time 71 is expediently a movement quality indicator and is provided in particular as a numerical value. The actuation time 71 expediently indicates how long it takes to actuate the actuator member 11. The actuation time 71 therefore preferably corresponds to the period of time from the start of the actuation - for example from the start of the ventilation of the first pressure chamber 10 - to the end of the movement of the actuator member 11 - for example until the time at which the actuator member 11 is the first End position reached. The ones in the 2 The actuation durations 43, 44 shown are examples of the actuation time 71. Furthermore, the actuation time 71 can indicate how long the movement of the actuator member 11 lasts and in this case can also be referred to as the running time. Optionally, the control unit 12 is designed to compare the actuation time 71 with a previously determined actuation time in order to determine whether the actuation time has improved - that is, reduced. Optionally, the control unit 12 is designed to calculate a cycle time of the pneumatic actuator 3 or the system 1 based on the actuation time 71. Optionally, the control unit 12 is designed to compare the cycle time with a previously determined cycle time in order to determine whether the cycle time has improved - i.e. reduced.

Now to the fluctuation information 72: The fluctuation information 72 is expediently a movement quality key figure and is provided in particular as a numerical value. The fluctuation information 72 indicates how much the speed of the actuator member 11 fluctuates during actuation. For example, a high value of the fluctuation information 72 corresponds to a strong fluctuation in speed, and a low value corresponds to a weak fluctuation in speed. The fluctuation information 72 is in particular a measure of how evenly the actuator member 11 moves during actuation, in particular when moving from the second end position to the first end position.

Now to the rebound information 73: The rebound information 73 indicates whether the actuator member 11 rebounds in an end position (in particular in the end position to be reached) when actuated. For an actuation that is intended to cause a movement of the actuator member 11 from the second end position to the first end position, the rebound information 73 indicates, for example, whether the actuator member 11 rebounds in the first end position. The rebound information 73 can, for example, be provided as binary information, i.e. in particular indicate whether there is a rebound or whether there is no rebound. The rebound information 73 is expediently a movement quality key figure and is optionally provided as a numerical value.

Now to the departure time 74: The departure time 74 is expediently a movement quality key figure and is provided in particular as a numerical value. The start time 74 indicates when the actuator member 11 begins to move during actuation. For example, the starting time 74 corresponds to a period of time from the start of ventilation of the drive chamber (for example the first pressure chamber 10) to the point in time at which the actuator member 11 begins to move. In the 2 an exemplary departure time 74 is shown.

Optionally, the movement quality information 70 has a (in particular binary) movement quality indicator 75. The movement quality indicator 75 can expediently assume two states, in particular only two states. The movement quality indicator 75 expediently indicates whether the movement quality is sufficient or whether the movement quality is not sufficient. Sufficient movement quality can also be referred to as good movement quality. Insufficient quality of movement can also be referred to as poor quality of movement. For example, the quality of movement in the 2 shown first speed curve 41 is sufficient and the movement quality of the second speed curve 42 is not sufficient.

In the following, preprocessing of the first temporal pressure profile and/or the second temporal pressure profile will be discussed in more detail. The term “pressure progression over time” is intended to mean the first and/or second pressure progression over time. The term “pressure history data” is intended below to mean the first pressure history data 61 and/or the second pressure history data 62. In particular, the preprocessing explained below is applied to the first pressure time curve, the second pressure time curve or to both the first pressure time curve and the second pressure time curve.

The pressure curve over time is expediently recorded by the control unit 12. The control unit 12 preferably calculates the pressure history data based on the recorded pressure history over time. In particular, the control unit 12 carries out preprocessing of the pressure curve over time in order to supply the pressure curve data receive. Alternatively, the pressure history data is equal to the recorded pressure history over time.

The valve arrangement 2, in particular the control unit 12, is preferably designed to record the pressure curve over time as a time series. For example, the pressure curve over time includes a large number (for example at least 100, at least 500 or at least 1000) pressure values that follow one another in time. The valve arrangement 2, in particular the control unit 12, is preferably designed to subject the time series to preprocessing in order to determine the pressure curve data. The valve arrangement 2 is preferably designed to record the first time pressure curve as a first time series and to subject the time series to preprocessing in order to determine the first pressure curve data.

Optionally, the preprocessing includes a normalization of the pressure curve over time in relation to the supply pressure or in relation to a maximum pressure value of the pressure curve over time, in particular in such a way that all pressure values of the pressure curve over time are between 0 and 1. For example, as part of the preprocessing, each pressure value of the pressure curve over time is divided by the supply pressure or by the maximum pressure value.

Optionally, the preprocessing also includes a reversal of the pressure curve over time. The reversal corresponds, for example, to a reflection on a horizontal axis. Preferably, the reversal is carried out only for the first pressure curve over time and not for the second pressure curve over time.

P* stellt zweckmäßigerweise eine transformierte Druck-ZeitReihe dar. Die Ermittlung der Bewegungsqualität-Information erfolgt zweckmäßigerweise auf Basis von P*.For example, preprocessing includes gradient boosting. With gradient amplification, the gradients of the pressure curve over time are amplified. For example, the control unit 12 carries out the gradient amplification according to the following equation (1), where the pressure values of the pressure curve over time are used as "p", and a and b are expediently real parameters: $$\text{P*}=\left(\text{p}+\text{a}\right)/\left(\text{p}+\text{b}\right)$$

P* expediently represents a transformed pressure-time series. The movement quality information is expediently determined on the basis of P*.

Gradient enhancement can also be referred to as contrast enhancement. The gradient amplification serves in particular to make unevenness, for example "bumps", and/or peaks in the pressure curve over time more pronounced. Optionally, the gradient amplification is only carried out for selected areas of the pressure curve over time - in particular not for the entire pressure curve over time.

Optionally, the preprocessing includes a decimation of the pressure curve over time. In particular, the number of pressure values of the pressure curve over time is reduced, for example at least by a factor of 10 or at least by a factor of 20. For example, the number of pressure values is reduced to a maximum of 200 or a maximum of 100. Expediently, the decimation does not take place uniformly - for example, the control unit 12 defines one or more areas of the pressure curve over time in which a higher temporal resolution - i.e. more pressure values per unit of time - remains after the decimation than in other areas of the pressure curve over time.

As a result of the preprocessing, the first pressure history data 61 and/or the second pressure history data 62 are expediently available to the control unit. The first pressure history data 61 and/or the second pressure history data 62 are available in particular as a respective time series, for example with a maximum of 200 or a maximum of 100 values each. The first pressure history data 61 and/or the second pressure history data 62 serve as input variables for the machine learning model.

The machine learning model 31 will be discussed in more detail below.

The machine learning model 31 preferably maps the first pressure history data 61 and/or the second pressure history data 62 onto the movement quality information 70. Preferably, the first pressure curve data 61 and/or the second pressure curve data 62 are the only input variables of the machine learning model 31. For example, the machine learning model 31 comprises at least one artificial neural network 50, in particular a convolutional neural network. Network and/or a Long Short-Term Memory network. The at least one artificial neural network 50 maps the first pressure curve data 61 and/or the second pressure curve data 62 onto the movement quality information 70.

The machine learning model 31 preferably includes a plurality of neural networks 50, which map the first pressure history data 61 and/or the second pressure history data 62 onto the movement quality information. In particular, a separate neural network 50 is available for each movement quality key figure and/or for the movement quality indicator. The 4 shows an example Technical design of the machine learning model 31 with five neural networks 50.

By way of example, a neural network 51 is present in order to map the first pressure curve data 61 and/or the second pressure curve data 62 to the actuation time 71.

This neural network 51 can also be referred to as the first neural network 51.

By way of example, a neural network 52 is present in order to map the first pressure curve data 61 and/or the second pressure curve data 62 onto the fluctuation information 72. This neural network 52 can also be referred to as a second neural network 52.

By way of example, a neural network 53 is present in order to map the first pressure history data 61 and/or the second pressure history data 62 onto the rebound information 73. This neural network can also be referred to as a third neural network 53.

By way of example, a neural network 54 is present in order to map the first pressure curve data 61 and/or the second pressure curve data 62 to the start time 74. This neural network 54 can also be referred to as a fourth neural network 54.

By way of example, a neural network 55 is present in order to map the first pressure history data 61 and/or the second pressure history data 62 onto the movement quality indicator 75. This neural network 55 can also be referred to as a fifth neural network 55.

One, several or all of the aforementioned neural networks 50, 51, 52, 53, 54, 55 are expediently designed to provide the movement quality information 70 (or the respective movement quality key figure or the movement quality indicator) only on the basis of the first pressure curve. Data 61 and/or the second pressure history data 62 is to be determined - preferably only on the basis of the first pressure history data 61 and the second pressure history data 62, in particular together. For example, one, several or all of the aforementioned neural networks 50, 51, 52, 53, 54, 55 are designed to generate the movement quality information 70 (or the respective movement quality key figure or the movement quality indicator) only on the basis of the first pressure curve. To determine data 61 or the second pressure history data 62, or preferably based on the first pressure history data 61 and the second pressure history data 62 together.

The 5 shows an exemplary structure of the artificial neural network 50. The artificial neural network 50 explained below is expediently the first artificial neural network 51. Furthermore, the artificial neural network 50 explained below can be the artificial neural network 52, 53, 54 and /or 55 trade.

By way of example, the artificial neural network 50 includes an input layer 81, which serves in particular to receive the input data - i.e., for example, the first pressure curve data 61 and/or the second pressure curve data 62 - and, for example, to determine the size of the input data.

By way of example, the artificial neural network 50 includes several convolution layers 82A, 82B, 82C, which expediently define several filters, in particular the filters explained below, and are designed to recognize features, in particular the features explained below. The convolution layers 82A, 82B, 82C are connected to the input layer 81.

By way of example, the artificial neural network 50 includes a dense layer 83. The dense layer 83 is expediently designed to provide the motion quality information based on the outputs from the convolution layers 82A, 82B, 82C.

The artificial neural network 50 preferably comprises at least one filter for detecting a uniform movement of the actuator member 11, a filter for detecting that the actuator member 11 has reached an end position and/or a filter for detecting an end of a movement of the actuator member 11. Optionally, this can artificial neural network 50 have one or more filters in order to recognize one or more of the following features in the course of the movement of the actuator member 11: pre-ventilation, displacement, acceleration, braking, post-ventilation.

Preferably, the artificial neural network 50 includes at least two of the above-mentioned filters. The valve arrangement 2 is expediently designed to determine the movement quality information 70 by a weighted addition of features that were recognized using the at least two filters.

Optionally, the artificial neural network 50 is designed as a convolutional neural network. The artificial neural network expediently comprises two paths on which filters of different sizes are present, which preferably consist of are designed to evaluate certain changes in curvature and/or to ignore disturbances.

For example, the artificial neural network 50 is designed to distinguish between different phases over time by activating one or more filters, for example a movement phase in which a movement takes place, a braking phase in which the actuator element 11 is braked, and / or an end phase in which the movement is ended.

The artificial neural network 50, in particular the dense layer 53, is preferably designed to determine, in particular to estimate, the actuation time 71 from a temporal position of an activation of filters (in particular several or all of the aforementioned filters).

The valve arrangement 2 is preferably designed, using a mathematical method from a time distribution of several or all of the aforementioned features by a weighted addition, the movement quality information 70, in particular the actuation time 71, fluctuation information 72, rebound information 73, start time 74 and / or to determine the movement quality indicator 75.

Below we will go into more detail about how ventilation of the actuator 3 can be adjusted based on the movement quality information.

Preferably, the valve arrangement 2, in particular the control unit 12, is designed to adapt a ventilation of the first pressure chamber 10 for further actuation of the actuator member 11 based on the movement quality information and/or to adapt a ventilation of a second pressure chamber 20 of the pneumatic one based on the movement quality information Adjust actuator 3 for further actuation of actuator member 11. The further actuation expediently takes place on the basis of a further (i.e. in particular a new) actuation specification and/or after completion of the present actuation.

The valve arrangement 2, in particular the control unit 12, is expediently designed to adapt the ventilation of the first pressure chamber 10 and/or the second pressure chamber 20 for further actuation in response to the movement quality information indicating that the movement quality is not sufficient. in particular to increase. The valve arrangement 2, in particular the control unit 12, is expediently designed in response to the movement quality information indicating that the movement quality is sufficient not to adjust the ventilation of the first pressure chamber 10 and/or the second pressure chamber 20 for further actuation, especially not to increase.

For example, the movement quality information includes a first actuation time of a first movement of the actuator member 11 in a first direction and a second actuation time of a second movement of the actuator member 11 in a second direction. The second direction is expediently opposite to the first direction. In this case, the movement quality information was expediently determined on the basis of pressure curves of movements of the actuator member 11 in two different directions. The movements include in particular a first movement from the second end position to the first end position and a second movement from the first end position to the second end position. For example, for each movement, a respective movement quality information is determined as explained above and the two movement quality information is then merged to obtain the movement quality information that includes the two actuation times. The first actuation time and the second actuation time are expediently examples of the actuation time 71 explained above. The valve arrangement 2 expediently determines the first actuation time and the second actuation time as explained above in connection with the actuation time 71. For example, the valve arrangement 2 carries out a first actuation of the actuator 3 in order to effect the first movement of the actuator member 11 (for example from the second end position to the first end position) and determines the first actuation time using the machine learning model 31. For example, the valve arrangement 2 carries out a second actuation of the actuator 3 in order to effect the second movement of the actuator member 11 (for example from the first end position to the second end position) and determines the second actuation time using the machine learning model 31.

The valve arrangement 2, in particular the control unit 12, is preferably designed to adapt the ventilation of the first pressure chamber 10 and/or the ventilation of the second pressure chamber 20 based on a relationship between the first actuation time and the second actuation time. The relationship is, for example, a ratio of the first actuation time to the second actuation time - that is, the division of the first actuation time by the second actuation time.

For example, the valve arrangement 2 calculates the ratio “ratio” as dividing the first actuation time “tHin” by the second actuation time “tBck” according to equation (2) shown below: $$\text{ratio}=\text{tHin/tBck}$$

The valve arrangement 2, in particular the control unit 12, is preferably designed to adapt the ventilation of the first pressure chamber 10 in that the valve arrangement 2 adapts, in particular increases, a target pressure value. The target pressure value is, for example, a pressure value of the first pressure when the actuation specification is fulfilled, in particular when the first end position is reached.

The valve arrangement 2, in particular the control unit 12, is preferably designed to adapt the ventilation of the second pressure chamber 20 in that the valve arrangement 2 adapts, in particular increases, a counterpressure. The back pressure value is a pressure value of a second pressure assigned to the second pressure chamber 20. For example, the back pressure value is the pressure value of the second pressure at the start of the actuation - that is, in particular at the start of the ventilation of the first pressure chamber.

The valve arrangement 2, in particular the control unit 12, is preferably designed to provide a target pressure value of a previous actuation of the actuator member 11 and/or a counter-pressure value of a previous actuation of the actuator member 11 when adjusting the ventilation of the first pressure chamber 10 and/or the ventilation of the second pressure chamber 20 to be taken into account. The term “previous” operation refers in particular to the operation carried out, on the basis of which the movement quality information is determined. The operation following this operation (in which the ventilation is adjusted) should be referred to as further operation.

For example, the valve arrangement 2, in particular the control unit 12, is designed to provide a difference between the supply pressure and the target pressure value of the previous actuation and/or a difference between the supply pressure when adjusting the ventilation of the first pressure chamber 10 and/or the ventilation of the second pressure chamber 20 and the back pressure value of the previous actuation must be taken into account.

For example, the valve arrangement 2 is designed to adapt, in particular increase, the ventilation in that the valve arrangement 2 adds an increment “p_target_hin_delta” to the target pressure value of the previous actuation and uses the result as the target pressure value for the further actuation (in particular the next actuation). The valve arrangement 2 expediently calculates the increment “p_target_hin_delta” according to the following equation (3), where “p1” is the supply pressure, “pEnd” is the target pressure value of the previous actuation, “c” is a parameter, in particular a constant, for example a positive number (e.g. 3), and “ratio” is the ratio explained above: $$\text{p\_target\_hin\_delta}=\left(\text{<figure-callout id="1" label="p" filenames="DE102022108962A1\_0005.png,DE102022108962A1\_0006.png" state="{{state}}">p</figure-callout>}1-\text{pEnd}\right)/\text{c*ratio}$$

For example, the valve arrangement 2 is designed to adapt, in particular increase, the ventilation in that the valve arrangement 2 adds an increment “p_target_bck_delta” to the back pressure value of the previous actuation and uses the result as a back pressure value for the further actuation (in particular the next actuation). The valve arrangement 2 expediently calculates the increment “p_target_bck_delta” according to the following equation (4), where “p1” is the supply pressure, “pini” is the back pressure value of the previous actuation, “c” is a parameter, in particular a constant, for example a positive number (e.g. 3), and “ratio” is the ratio explained above: $$\text{p\_target\_bck\_delta}=\left(\text{<figure-callout id="1" label="p" filenames="DE102022108962A1\_0005.png,DE102022108962A1\_0006.png" state="{{state}}">p</figure-callout>}1-\text{pini}\right)/c/\text{ratio}$$

The valve arrangement 2 is preferably designed to carry out a plurality of actuations of the actuator member with different target pressure values. The valve arrangement 2 is expediently designed to determine respective movement quality information with each actuation. The valve arrangement 2 is expediently designed to determine a target pressure value for further actuation based on the movement quality information and the various target pressure values. For example, the valve arrangement 2 is designed to use the target pressure value for further actuation that led to sufficient movement quality in one or more previous actuations.

Optionally, the valve arrangement 2 is designed to carry out a first actuation of the actuator 3 in a normal operating mode and to carry out a second actuation of the actuator 3 in the compressed air saving mode (in which there is expediently a pressure reduction compared to the normal operating mode), and for both actuations To determine motion quality information to determine whether the compressed air saving mode has degraded motion quality. If this is the case, the valve arrangement 2 deactivates the compressed air saving mode or increases the target pressure value and/or the counterpressure value for the next actuation in the compressed air saving mode.

Optionally, the valve arrangement 2 is designed to adapt the target pressure and/or the counterpressure based on the movement quality information, in particular based on the ratio of the first actuation time to the second actuation time sen that the first actuation time and the second actuation time approach each other.

Optionally, the valve arrangement 2 is designed to increase the target pressure value of the drive chamber for the next actuation of the actuator 3 on the basis of the fluctuation information 72 - for example in response to the fact that the fluctuation information indicates a fluctuation above a predetermined threshold value.

Optionally, the valve arrangement 2 is designed to increase the counter pressure value of the counter chamber for the next actuation of the actuator 3 in response to the rebound information indicating that a rebound is present.

The valve arrangement 2 is preferably designed to iteratively adapt the ventilation of the actuator 3, in particular the target pressure and/or the counterpressure. For example, the valve arrangement 2 is designed to calculate the movement quality information of this operation after each actuation of the actuator 3 and to adapt the ventilation of the actuator for the next actuation based on the movement quality information.

The valve arrangement 2 is preferably designed to ventilate the first pressure chamber 10 and/or the second pressure chamber 20 in a pulsed manner for the further actuation of the actuator member 11.

Optionally, the valve arrangement 2, in particular the control unit 12, is designed to carry out a diagnosis based on the movement quality information. For example, predictive maintenance can be carried out based on the diagnosis. For example, the valve arrangement 2 is designed to calculate a temporal change, in particular a tendency, of the movement quality information 70, in particular the starting time 74, and based on this to generate diagnostic information that indicates a state of the system 1, in particular of the actuator 3 and/or the valve arrangement 2. For example, the valve arrangement 2 can conclude, based on a starting time 74 that increases over time, that there is wear on the actuator 3 and generate diagnostic information that indicates this wear.

Optionally, the valve arrangement 2 is designed to detect, on the basis of the movement quality information 70, in particular based on the start time 74, which load or mass is actuated by the actuator 3 - in particular the mass of the actuating object 9. For example, the valve arrangement 2 is designed , based on the movement quality information 70, in particular the start time 74, to detect an empty journey of the actuator 3.

With reference to the 6 A method for operating the valve arrangement 2 will be described below.

The method includes a step S1, in which the actuation of the actuator member 11 of the pneumatic actuator 3 is carried out. During actuation, the first pressure curve over time is recorded in order to provide the first pressure curve data.

The method further includes a step S2, in which the movement quality information is determined based on the first pressure curve data using the machine learning model 31.

The method includes in particular a step S3, in which it is checked whether the movement quality information indicates sufficient movement quality or insufficient movement quality.

In particular, in response to the fact that the movement quality information indicates an insufficient movement quality, the method continues with step S4, in which, for further actuation of the actuator member 11, the first pressure chamber 10 is ventilated and/or the second pressure chamber 20 is ventilated is adjusted, in particular by increasing the target pressure value and/or the final pressure value.

The method expediently continues with carrying out step S1 again in order to carry out a further actuation of the actuator member 11, the further actuation being carried out with the adapted ventilation of the first pressure chamber 10 and/or the adapted ventilation of the second pressure chamber 20, in particular with the increased target pressure value and/or the increased final pressure value.

In particular, in response to the fact that the movement quality information indicates a sufficient movement quality, step S4 is expediently not carried out and the method continues by carrying out step S1 again in order to carry out a further actuation of the actuator member 11, the further actuation expediently being carried out with the same ventilation of the first pressure chamber 10 and/or the second pressure chamber 20 as the previous actuation is carried out, in particular with the same target pressure value and/or the same final pressure value as the previous ventilation.

According to a preferred embodiment, the movement quality information 70 is determined without using a flow sensor. The valve arrangement 2 expediently determines the movement quality information without knowledge of an opening cross section of a valve and/or without knowledge of a throttle position of the pneumatic actuator and/or without knowledge of the type of pneumatic actuator. In particular, the valve arrangement 2 determines the movement quality information based exclusively on the first pressure curve over time and the second pressure curve over time.

Preferably, no sensor unit, in particular no position sensor and/or no pressure sensor, is present on the actuator 3. In particular, there is no direct detection of pressures in the pressure chambers 10, 20. The valve arrangement 2 expediently detects the first pressure and/or the second pressure based on pressures measured at the working outlets 21, 22.

Preferably, the valve arrangement 2 does not use any information about a piston surface of the actuator 3 to determine the movement quality information. In particular, the valve arrangement 2 does not carry out any explicit force calculation to determine the movement quality information.

Appropriately, the valve arrangement 2, in particular the machine learning model 31, does not form any pressure differences and/or any derivatives to determine the movement quality information.

The machine learning model 31, in particular one, several or all artificial neural networks 50 of the machine learning model 31, is preferably trained with training data.

The training data is expediently obtained through simulations and/or through measurements of real pneumatic actuations of a pneumatic actuator. The training data expediently includes a large number of first pressure curve data and/or second pressure curve data, each associated with movement quality information. The machine learning model is trained using the training data to map the first pressure curve data and/or the second pressure curve data onto the movement quality information.

QUOTES INCLUDED IN THE DESCRIPTION

This list of documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 102020204484 A1 [0004]

Claims (20)

Valve arrangement (2) for ventilating a first pressure chamber (10) of a pneumatic actuator (3) in order to carry out an actuation of an actuator member (11) of the pneumatic actuator (3) according to an actuation specification, in particular a position specification, wherein the valve arrangement (2) is formed is, during actuation, to detect a first pressure curve over time of a first pressure assigned to the first pressure chamber (10) in order to provide first pressure curve data (61) which depict the first pressure curve over time, and wherein the valve arrangement (2) via a machine Learning model (31) has and is designed to use the machine learning model (31) based on the first pressure curve data (61) to determine movement quality information (70), which has a quality of a movement carried out during operation Movement of the actuator member (11). Valve arrangement (2). Claim 1 , whereby the machine learning model (31) maps the first pressure curve data (61) onto the movement quality information (70). Valve arrangement (2) according to one of the preceding claims, wherein the movement quality information (70) comprises an actuation time (71) which indicates how long the actuation of the actuator member (11) lasts, and/or fluctuation information (72), which indicates how much the speed of the actuator member (11) fluctuates during actuation, and/or rebound information (73), which indicates whether the actuator member (11) rebounds in an end position when actuated and/or a starting time ( 74), which indicates when the actuator member (11) begins to move during actuation. Valve arrangement (2) according to one of the preceding claims, wherein the valve arrangement (2) is designed to detect a second pressure curve over time of the second pressure assigned to a second pressure chamber (20) of the pneumatic actuator (3) during the actuation of the actuator member (11), to provide second pressure history data (62) which depict the second pressure history over time, and the movement quality using the machine learning model (31) based on the first pressure history data (61) and the second pressure history data (62). -Information (70) to be determined. Valve arrangement (2) according to one of the preceding claims, wherein the valve arrangement (2) is designed to end the ventilation of the first pressure chamber (10) when the actuator member (11) is actuated before the actuation specification is fulfilled. Valve arrangement (2) according to one of the preceding claims, wherein the valve arrangement (2) is designed to adapt ventilation of the first pressure chamber (10) for further actuation of the actuator member (11) on the basis of the movement quality information (70) and/or on the basis the movement quality information (70) to adapt a ventilation of a/the second pressure chamber (20) of the pneumatic actuator (3) for further actuation of the actuator member (11). Valve arrangement (2). Claim 6 , wherein the movement quality information comprises a first actuation time of a first movement of the actuator member (11) in a first direction of movement and a second actuation time of a second movement of the actuator member (11) in a second direction of movement and the valve arrangement (2) is designed to ventilate the first pressure chamber (10) and / or the ventilation of the second pressure chamber (20) based on a relationship of the first actuation time to the second actuation time. Valve arrangement (2). Claim 7 , wherein the valve arrangement (2) is designed to adapt the ventilation of the first pressure chamber (10) in that the valve arrangement (2) adjusts a target pressure value, the target pressure value being a pressure value of the first pressure when the actuation specification is fulfilled, and / or the ventilation the second pressure chamber (20) in that the valve arrangement (2) adjusts a back pressure, the back pressure value being a pressure value of a second pressure assigned to the second pressure chamber (20). Valve arrangement (2) according to one of the Claims 6 until 8th , wherein the valve arrangement (2) is designed, when adjusting the ventilation of the first pressure chamber (10) and / or the ventilation of the second pressure chamber (20), a target pressure value of a previous actuation of the actuator member (11) and / or a counter-pressure value of a previous actuation of the actuator member (11), whereby the target pressure value is a pressure value of the first pressure when the actuation specification is fulfilled and/or where the counterpressure value is a pressure value of a second pressure assigned to the second pressure chamber. Valve arrangement (2) according to one of the Claims 6 until 9 Claims, wherein the valve arrangement (2) is designed to ventilate the first pressure chamber (10) and/or the second pressure chamber (20) in a pulsed manner for the further actuation of the actuator member (11). Valve arrangement (2) according to one of the Claims 6 until 10 , whereby the valve arrangement (2) is made is to carry out a plurality of actuations of the actuator member (11) with different target pressure values, the target pressure value being a pressure value of the first pressure when the actuation specification is fulfilled, and to determine a respective movement quality information (70) for each actuation, and on the basis of which Movement quality information (70) and the various target pressure values to determine a target pressure value for further actuation of the actuator member (11). Valve arrangement (2) according to one of the preceding claims, wherein the machine learning model (31) comprises at least one artificial neural network (50), in particular a convolutional neural network and / or a long short-term memory network. Valve arrangement (2). Claim 12 , wherein the artificial neural network (50) has at least one filter for detecting a uniform movement of the actuator member (11), a filter for detecting that the actuator member (11) has reached an end position, and / or a filter for detecting an end of a movement the actuator member (11). Valve arrangement (2). Claim 13 , wherein the artificial neural network (50) comprises at least two of the filters mentioned and the valve arrangement (2) is designed to determine the movement quality information (70) by a weighted addition of features that were recognized using the at least two filters. Valve arrangement (2) according to one of the preceding claims, wherein the valve arrangement (2) is designed to record the first time pressure curve as a first time series and to subject the first time series to preprocessing in order to determine the first pressure curve data (61). Valve arrangement (2). Claim 15 , where the preprocessing includes gradient boosting. Valve arrangement (2) according to one of the preceding claims, comprising a control unit (12) on which the machine learning model (31) is stored. Valve arrangement (2). Claim 17 , further comprising a communication port (32) and a valve device (14) for ventilating the first pressure chamber (10), the control unit (12) being communicatively connected between the communication port (32) and the valve device (14). Method for operating a valve arrangement (2) according to one of the preceding claims, comprising the steps: - Carrying out (S1) the actuation of the actuator member (11) of the pneumatic actuator (3), - during actuation, detecting the first pressure curve over time in order to provide the first pressure curve data (61), - Using the machine learning model (31), determining (S2) the movement quality information based on the first pressure curve data (61). Procedure according to Claim 19 , further comprising the step - carrying out a further actuation of the actuator member (11), wherein during the further actuation a ventilation of a first pressure chamber of the pneumatic actuator and / or a ventilation of a second pressure chamber of the pneumatic actuator is adjusted based on the movement quality information.

Images