Method for determining operating information of metering pump

Abstract

Disclosed herein is an embodiment of a method for determining operational information of a metering pump, the metering pump comprising a dose chamber, a displacement member, and a drive motor for driving the displacement member, receiving a plurality of detection values indicative of an indicated amount of activation intensity of the displacement member at a respective position of the displacement member; operation information is calculated according to a machine learning model trained to output the operation information in response to receiving a plurality of input values derived from a detected value of the indicated amount.

Description

Technical Field

[0001] The invention relates to a method for determining operating information of a metering pump. Background Art

[0002] Metering pumps or dosing pumps are used to supply and dose precise amounts of liquids. These metering pumps usually have a movable displacement member, for example in the form of a membrane or piston driven by a drive motor via a drive system that transfers the rotational movement of the motor into a linear movement of the displacement member.

[0003] For many applications, it is desirable to determine operating information of a metering pump, for example to detect the presence of a fault or to calculate the effective stroke length or other operating parameters of the pump.

[0004] EP3591226 discloses a metering pump comprising a control device designed in such a way that it detects the current position of a displacement element, detects the torque of the electric drive motor at several positions of the displacement element, and monitors the torque in relation to the position of the displacement element. This prior art metering pump comprises an analysis module which either compares the torque or pressure curve detected over time or compares the detected pressure or torque curve with a previously stored sample curve.

[0005] However, it is still desirable to provide a method for determining operational information of a metering pump that is applicable to different types of metering pumps or can at least be effectively adapted for use with different types of metering pumps. It is also generally desirable to provide such a method that is reliable and computationally efficient and can be configured in a cost-effective manner. Summary of the invention

[0006] Therefore, it remains desirable to provide a method for determining operational information of a metering pump that solves one or more of the above-mentioned problems and / or has other benefits, or at least provides an alternative to existing solutions.

[0007] According to one aspect, an embodiment of a method for determining operating information of a metering pump is disclosed herein, the metering pump comprising a metering chamber, a displacement member, and a drive motor for driving the displacement member. Various embodiments of the method include: a) receiving, during operation of the metering pump, a plurality of detection values, in particular a sequence of detection values, of an indicator of an intensity of activation of the displacement member at respective positions of the displacement member; b) calculating the operational information according to a machine learning model, which is configured (particularly trained) to output the operational information in response to receiving a plurality of input values ​​(particularly a sequence of input values) obtained according to the detection value of the indicating quantity.

[0008] The inventors have recognized that by employing a trained machine learning model, it is possible to reliably determine operating information of a metering pump based on a detected value of an indicator of the intensity of activation of the displacement member at a corresponding position of the displacement member during operation of the metering pump. Various embodiments of the method can be effectively adapted for use with different types of metering pumps.

[0009] During operation of various embodiments of the metering pump, the drive motor moves the displacement member, preferably in a reciprocating manner, so that the displacement member increases and decreases the volume of the dosing chamber through its movement. When the dosing chamber is filled with an incompressible liquid, the volume change of the dosing chamber defines the volume of liquid delivered. In the presence of air or cavitation, the medium inside the dosing chamber is compressible. In this case, the change in volume is different from the volume of liquid delivered. In general, the volume delivered is equal to the volume change multiplied by the effective stroke length.

[0010] During operation of the metering pump, the pressure or a related indicator (e.g., a related force or torque) in the metering chamber can be detected and recorded, and the pressure or the related indicator in the metering chamber can be used to determine the operating condition or other operating information of the metering pump. The pressure or the related indicator can be measured continuously or intermittently (especially periodically) as a sequence (especially a time series) of detection values. The drive motor can be an electric drive motor.

[0011] The relevant force or torque may be the force acting on the displacement member (in particular by the drive motor). The force or torque acting on the displacement member is related to, in particular substantially proportional to, the pressure inside the dosing chamber. Thus, while some embodiments use the detected pressure in the dosing chamber as an indicator quantity, other embodiments use the detected force or torque as an indicator quantity.

[0012] For the purpose of detecting pressure, the metering pump may include a pressure sensor for detecting the pressure in the dosage chamber. In an alternative, the pressure may be calculated based on the driving torque or force provided by the drive motor. The calculation may be based on the understanding of the mechanical connection between the drive motor and the displacement member. The driving torque or force may be, for example, measured by a corresponding torque sensor or force sensor, respectively, or may be derived according to the electric value (electric values, potential value) of the drive motor. Typically, other examples of the indicated amount include motor current, motor voltage, or other quantities that may be derived from motor current and / or motor voltage or otherwise related thereto, and / or other quantities related to the motor load.

[0013] The inventors have discovered that a properly trained machine learning model can determine various operating information of a metering pump from multiple detected values ​​of an indicated quantity. Examples of such operating information include identification of discrete operating states and prediction of continuous valued operating parameters. Various embodiments of the methods disclosed herein allow such operating information to be determined without applying expert knowledge to identify how to derive various operating states or parameter values ​​based on the measured indicated quantities. Therefore, various embodiments of the methods disclosed herein use different types of machine learning models.

[0014] In particular, the operational information may include a classification of operational conditions of the metering pump, in particular a classification of error conditions or faults of the metering pump. The operational condition may be a current condition or even a predicted future condition, such as a prediction of an impending error condition that may occur in the near future. To this end, the machine learning model may include a classification model that is trained to output an identifier of one of a plurality of discrete categories. Examples of categories may include predetermined error conditions, such as "cavitation", "bubbles", "leakage conditions", etc. In some embodiments, the classification model may be trained to output an estimated likelihood that one or more error conditions and / or other categories of operational conditions exist or may occur in the future (particularly in the near future).

[0015] Alternatively or additionally, in some embodiments, the operational information includes the value of an operational parameter. Thus, the machine learning model may include a regression model trained to output the value of a continuous-valued operational parameter. Examples of operational parameters include: discharge pressure, effective stroke length, and discharge flow rate. For example, one minus the effective stroke length is an indication of the amount of air in the dosing chamber.

[0016] The machine learning model receives multiple input values ​​representing the detection value of the indicated quantity and / or input values ​​obtained from the detection value of the indicated quantity. To this end, the machine learning model can be configured to receive different types of representations of multiple input values ​​and / or additional input values. In some embodiments, the machine learning model is configured to receive multiple input values ​​of the indicated quantity (e.g., a sequence or array of input values), each of the multiple input values ​​is associated with a corresponding position of the displacement member, and wherein the machine learning model is configured to output the operation information in response to receiving at least the multiple input values ​​(e.g., in the form of input data pairs as described below or in other forms). The input value can be a detection value or a value obtained from the detection value, such as a value obtained by noise reduction processing, averaging of multiple detection values, etc.

[0017] In some embodiments, for example, when the input values ​​always represent the value of the indicated quantity at the same corresponding predetermined position of the displacement member, or the value of the indicated quantity at a predetermined time during the cyclic movement of the displacement member, a one-dimensional representation of the indicated quantity can be used as input data for a machine learning model, thereby allowing a memory-efficient representation, which can be particularly beneficial for embedded implementations. Therefore, in some embodiments, each of the multiple input values ​​is associated with a corresponding predetermined position of the displacement member and / or a corresponding predetermined time during the cyclic movement of the displacement member. Therefore, the input values ​​represent a sequence of values, wherein the sequence has a predetermined phase relationship with the cyclic movement of the displacement member.

[0018] In other embodiments, the machine learning model is configured to receive a plurality of pairs of input data, each pair of input data comprising a position of a displacement member and a corresponding value of an indication amount at said position, and wherein the machine learning model is configured to output said operation information in response to receiving said plurality of pairs of input data. Thus, detection values ​​at different positions may be used.

[0019] In the above and other embodiments, the machine learning model can therefore receive a representation of a so-called pressure-stroke curve that relates pressure or a similar indicator to the current position of the displacement member. The pressure-stroke diagram can be represented as a closed curve in a pressure-position coordinate system. Alternatively, the pressure-stroke diagram can be represented as an open curve representing pressure (or other indicator) as a function of time or phase along the cyclic movement of the displacement member.

[0020] In some embodiments, the machine learning model may receive a two-dimensional array of input values ​​representing a pressure-stroke graph, e.g., an array of image pixels representing an image of a pressure-stroke curve. For example, the input may be represented as a raster image of a representation of a pressure-stroke graph. Each pixel is represented as a number between 0 and 1 indicating the black level of the pixel. Other embodiments may use a different representation of the pressure-stroke graph, e.g., a more compact representation.

[0021] To this end, in order to detect the position of the displacement member, the metering pump can include a suitable mechanism for detecting the current position of the displacement member. For example, the metering pump can include a position sensor, or the drive motor can be a stepper motor, so that the position can be determined by counting the rotation angle of the drive motor.

[0022] Thus, in some embodiments, the method comprises: - receiving position data indicative of a monitored position of the displacement member during operation of the metering pump, or calculating position data based on at least received detection values ​​of the indicated quantity; - calculating a plurality of input values, for example a sequence of values ​​of the indicating quantity or a plurality of pairs of input data, based on the received detection values ​​of the indicating quantity and based on the received or calculated position data.

[0023] While some metering pumps allow the position of the displacement member to be determined during operation, other types of metering pumps do not provide this information. Therefore, it is desirable to provide a method applicable to a wider range of metering pumps. To this end, in some embodiments, the machine learning model is configured to receive a time series of detection values ​​of the indicated amount at corresponding time points, and in response to receiving the time series of the detection positions of the detection values ​​of the indicated amount, output the operating information. In particular, in some embodiments, the machine learning model is configured to output the operating information based only on the time series of the detection positions of the received detection values ​​of the indicated amount. It should be understood that when the cycle time (i.e., stroke cycle) of the movement of the displacement member is known and when the position of the displacement member is known for a reference time, the information carried by representing the input data as a time series is the same as that carried by representing the indicated amount as a function of the position. However, as described above, not all types of pumps can easily obtain this information. However, the inventors have recognized that a properly trained machine learning model can determine useful operating information based only on the indicated amount, i.e., without measuring the position of the displacement member.

[0024] Specifically, in some embodiments, the machine learning model includes a first machine learning model and a second machine learning model, the first machine learning model being configured to calculate an input representation of a plurality of input values ​​of the indicated quantity based on a time series of detection values ​​of the indicated quantity at corresponding time points received during the operation of the metering pump, each input value indicating the value of the indicated quantity at a corresponding (particularly corresponding predetermined or otherwise known) position of the displacement member; the second machine learning model can therefore be configured to output operational information in response to receiving the calculated input representation. Specifically, the first machine learning model can be trained to determine the phase and / or period of the cyclic motion of the displacement member according to the time series of detection values ​​of the indicator function. To this end, the first machine learning model can receive a value or pressure value of another indicated quantity obtained during a time window, the time window can have an unknown starting point relative to the stroke cycle of the pump and / or an unknown length relative to the stroke cycle of the pump. The first machine learning model can output a pressure value corresponding to a time window of a stroke of the pump (or, if the first machine learning model receives a value of another indicated quantity, output the value of the other indicated quantity), the time window starting at a predetermined point (e.g., bottom dead point) of the stroke cycle. The second machine learning model may then receive as its input a pressure (or other indicator) value corresponding to a single stroke. Alternatively, the first machine learning model may output a pressure value (or, if the first machine learning model receives a value of another indicator, output the value of the other indicator) for a time window of a predetermined duration corresponding to a predetermined number of strokes or otherwise relative to the duration of a stroke cycle. Thus, the second machine learning model may receive a pressure value or other indicator value corresponding to the predetermined duration.

[0025] Providing separate first and second machine learning models can produce a more compact, more memory-efficient representation of the entire model. In particular, when the machine learning model includes, for example, multiple models for detecting respective operating conditions or for estimating respective parameters, the machine learning model may include a single first machine learning model that performs phase detection of a time series and multiple second machine learning models, each second machine learning model receiving the output of the first machine learning model as input. However, it should be understood that in other embodiments, the first machine learning model and the second machine learning model can be combined into a single machine learning model, which can therefore receive a time series of values ​​of an indication quantity, wherein the time series has an unknown phase shift and / or an unknown duration relative to the cyclic motion of the displacement member. The combined machine learning model can be trained to output operating information directly based on the time series of unknown phase and / or duration. Therefore, the training set for such a combined machine learning model may include input time series of different relative phase shifts and / or durations relative to the cyclic motion of the displacement member.

[0026] The present disclosure relates to different aspects, including the methods, corresponding devices, systems, methods and / or products described above and below, each of which can bring one or more benefits and advantages described in association with one or more other aspects, and each of which has one or more embodiments corresponding to the embodiments described in combination with one or more of the other aspects and / or the embodiments disclosed in the attached claims.

[0027] Specifically, according to one aspect, an embodiment of a computer-implemented method for creating a trained machine learning model is disclosed herein. The method includes: a) obtaining a set of training data items, each training data item comprising a plurality of input values ​​and a corresponding target output, the plurality of input values ​​indicating an indication amount, the indication amount indicating an activation intensity of a displacement member of a metering pump at a corresponding position of the displacement member during operation of the metering pump, and the corresponding target output indicating operational information observable during the operation of the metering pump; b) training a machine learning model based on the obtained training data set to output operational information in response to receiving multiple input values.

[0028] Generally, for the purposes of this disclosure, the term "trained machine learning model" refers to a machine learning model having a set of parameters (such as weights) that have been adapted based on a set of training data using a suitable training algorithm (such as an unsupervised training algorithm or a supervised training algorithm). Similarly, the term "training a machine learning model" refers to the process of adapting a machine learning model based on training data. In particular, training, and in particular adaptation of model parameters of a machine learning model, can be performed using supervised learning based on a training data set, wherein each training data item is labeled by a corresponding target output.

[0029] Various embodiments of the methods disclosed herein may be computer-implemented. Therefore, embodiments of a data processing system configured to perform the steps of the methods described herein are disclosed herein. Specifically, a data processing system may have a program code stored thereon, which is suitable for causing the data processing system to perform the steps of the methods described herein when executed by the data processing system. The data processing system may be implemented as a single computer or other data processing device, or as a distributed system including multiple computers and / or other data processing devices (e.g., client-server systems, cloud-based systems, etc.). The data processing system may include a data storage device for storing computer programs and detector data. The data processing system may include a communication interface for receiving detection values ​​and / or other types of sensor data. In some embodiments, the data processing system may be partially or completely implemented as a processing unit that is appropriately programmed or otherwise configured, such as a control device for controlling the operation of a metering pump. Therefore, a portion of the data processing system or the entire data processing system may be housed in the housing of a metering pump, such as as a part of a control device for controlling the operation of a metering pump or as a separate processing unit. Alternatively or additionally, the data processing system may include one or more data processing devices outside the metering pump. The data processing system may receive a sensed value indicative of the quantity from the metering pump or in other ways, such as from a separate pressure sensor.

[0030] According to one aspect, embodiments of a metering pump are disclosed herein. Various embodiments of the metering pump include a displacement member, a drive motor for driving the displacement member, and a data processing system as disclosed above and disclosed below. In particular, the metering pump may include a processing unit that may be integrated into or separated from a control device configured to control the operation of the metering pump; the processing unit may be configured to perform the steps of the method described herein. The processing unit of the pump may be used alone as a stand-alone device or as part of a distributed data processing system, for example, in collaboration with an external data processing system such as a portable data processing device and / or with a remote host computer and / or with a cloud-based architecture to perform embodiments of the process described herein. The processing unit may be separated from or partially or completely integrated into the control device for controlling the operation of the pump. The pump may also include an integrated sensor configured to measure an indicated amount or to measure an amount from which an indicated amount can be obtained.

[0031] According to another aspect, an embodiment of a system is disclosed herein, the system comprising a metering pump as disclosed herein and a data processing system, wherein the metering pump comprises a dosing chamber, a displacement member, and a drive motor for driving the displacement member. In some embodiments, the data processing system is separate from the metering pump and comprises an interface for receiving a plurality of detection values ​​indicating activation strength of the displacement member at corresponding positions of the displacement member during operation of the metering pump. Alternatively, the metering pump comprises the data processing system.

[0032] Another aspect disclosed herein relates to an embodiment of a computer program configured to cause a data processing system to perform the actions of the methods described above and below. The computer program may include program code means, which, when executed on a data processing system, are suitable for causing the data processing system to perform the actions of the methods disclosed above and below. The computer program may be stored on a computer-readable storage medium, in particular a non-transitory storage medium, or implemented as a data signal. The non-transitory storage medium may include any suitable circuit or device for storing data, such as RAM, ROM, EPROM, EEPROM, flash memory, magnetic or optical storage devices (such as CD ROM, DVD, hard disk, etc.). BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The preferred embodiments will be described in more detail with reference to the accompanying drawings, in which

[0034] Figure 1 An example of a dosage or metering pump is schematically shown.

[0035] Figure 2 An example of a pressure-stroke diagram showing a pressure-stroke curve of a metering pump is schematically shown.

[0036] Figure 3 Examples of pressure-stroke curves for a metering pump are shown in the presence of various faults or other operating conditions.

[0037] Figure 4 A flow chart of a training method for creating a trained machine learning model for subsequent use in a method for determining operating information of a metering pump is schematically shown.

[0038] Figure 5 A flow chart of a method for determining operating information of a metering pump is schematically shown.

[0039] Figures 6 to 8 An example of a machine learning model is schematically shown.

[0040] Figures 9 to 11 An example of a prediction result obtained by the method for determining operation information of a metering pump is shown. DETAILED DESCRIPTION

[0041] As an example of a dosing or metering pump, Figure 1 A diaphragm pump is shown schematically. It must be understood that the present invention can be implemented in a similar manner with other types of dosing pumps, for example metering pumps or dosing pumps that use a piston instead of a membrane as a displacement member. Figure 1 The pump shown has a pump chamber or dosing chamber 2, whose side wall is formed by a membrane 4. The membrane 4 is a displacement member. By the displacement of the membrane 4, the volume inside the dosing chamber 2 can be increased to fill the dosing chamber 2, and the volume inside the dosing chamber 2 can be reduced to discharge liquid from the dosing chamber 2. A suction valve 6 is arranged on the lower side of the dosing chamber 2, and a pressure valve 8 is arranged on the opposite side. Both valves are designed as non-return valves. In this example, the spherical valve element closes the valve by gravity. However, in addition, a biasing element, such as a spring, can be provided. During operation, liquid is sucked from the liquid container 3 through the suction valve 6 into the dosing chamber 2 via the suction line 5, and discharged from the dosing chamber 2 through the pressure valve 8. The liquid is discharged from the pressure valve 8 via the pressure line 9 and the pressure loading valve 7, for example, into the pipeline 11 of the facility. The pressure loading valve 7 in the pressure line 9 defines the pressure in the pressure line 9, that is, the pressure on the outlet side of the pressure valve 8 is maintained at a predetermined pressure. The pressure is set by the pressure loading valve 7. A pulsation damper 13 is connected to the supply line 9 for equalizing pressure pulsations occurring in the outlet or pressure line 9 .

[0042] The membrane 4 is moved in a reciprocating manner via a connecting rod 10. In order to drive the connecting rod 10 in a reciprocating manner, a drive motor is provided, in particular an electric drive in the form of an electric drive motor 12 (e.g., a stepper motor). The rotating drive motor 12 moves the connecting rod 10 via an eccentric drive 14, which converts the rotational movement into a linear reciprocating movement. The eccentric drive 14 is connected to the electric drive motor 12 via a gear drive 16. The connecting rod 10 is connected to the eccentric drive 14 at a connection point 18, which is eccentrically spaced from the axis of rotation x of the eccentric drive 14. If the eccentric drive 14 rotates in the direction of rotation R, this causes the connecting rod 10 to move linearly in the direction S. In addition, in this example, a spring 20 is arranged in the drive. The spring 20 is a compression spring connected to the connecting rod 10, so that when the connecting rod 10 moves backward to the direction S1, thereby moving the membrane 4 to the retracted position, the spring 20 is compressed. The spring 20 is able to accumulate energy during the suction stroke. This energy is released during the pressure stroke when the connecting rod 10 moves forwards together with the membrane 4, i.e. in the advanced position in the direction S2. In this way, the spring 20 smoothes the torque applied by the electric drive motor 12 during the entire stroke. It must be understood that a spring can also be arranged which is compressed during the pressure stroke and acts as a return spring. Furthermore, the invention can also be implemented without the spring 20.

[0043] The dosage pump has a control device 22 for controlling the electric drive motor 12. The control device 22 includes a monitoring module 24 for monitoring the operation of the dosage pump. The control device 22 may include conventional electronic components, in particular a CPU or other processing unit, a storage device and one or more software applications for controlling the dosage pump. The software application may be stored on the storage device and used for execution on the CPU. The monitoring module 24 may preferably be implemented as a software module. In this example, the monitoring module 24 is integrated into the control device 22. However, it would be possible to transmit information to an external computing or monitoring device, in particular a cloud device acting as the monitoring module 24. To this end, the control device 22 may include a communication interface 26 for wired and / or wireless communication.

[0044] The monitoring module 24 is configured to continuously or intermittently record the pressure P in the dosing chamber 2 and the position of the displacement member. Figure 1The position of the membrane 4 of the diaphragm pump can be recorded as a representation of the pressure-stroke curve in the pressure-stroke diagram. In order to detect the position of the membrane 4 along the direction S, in this example, an encoder 28 that detects the angular position of the rotor of the drive motor 12 is used. In addition, the specific position of the drive or displacement member can be detected, for example, by a single sensor, and other positions can be calculated based on the known speed of the displacement member and the time elapsed. In addition, a stepper motor can also be used instead of a special encoder. Based on the angular position, the transmission ratio of the gear drive 16 and the geometric design of the eccentric drive 14 are known, and the position in the direction S can be calculated. The pressure P in the dosing chamber 2 can be detected by a pressure sensor 30, or by detecting the torque of the drive motor 12 or the force acting in the drive and indirectly detecting based on the force F acting on the membrane 4, or detecting in other ways. In this example, the pressure sensor 30 is arranged at the dosing chamber 2 and is connected to the control device 22. In the case where force or torque is detected as an indicated quantity instead of pressure, this force or torque can be recorded continuously at the position of the displacement member instead of recording the pressure, since the pressure is related to the force, in particular is proportional to the force, or to the torque, in particular is related to the torque multiplied by a term that depends on the position of the eccentric drive.

[0045] The control device 22 also includes a processing module 50 configured to implement a trained machine learning model. The processing module 50 can be implemented, for example, as a software module executed by the CPU of the control device 22, or in other ways. The trained machine learning module can include a suitable representation of the model structure and the parameter values ​​of the model parameters (e.g., the weights of the neural network model). In this example, the processing module 50 including the trained machine learning module is integrated into the control device 22. However, the processing module 50 and / or the machine learning module can be implemented on a data processing system outside the control device of the pump. To this end, the control device 22 can exchange information with an external data processing system (e.g., a cloud computing architecture) used as a processing module and / or implementing the machine learning module. During operation, the processing module 50 receives the recorded pressure values ​​and the associated positions of the displacement members from the monitoring module 24. The processing module 50 optionally processes the received information and feeds it into the trained machine learning module, which in turn returns the corresponding operation information as described herein. The processing module may be configured to display information on a display of the control unit and / or to issue an alarm in the event of a detected fault condition and / or to forward information and / or any issued alarms to an external device or system via the communication interface 26, etc. Examples of processes performed by the processing module 50 and the machine learning module are described in more detail below. The trained machine learning module may be commissioned with the control device 22 during installation, or it may be subsequently loaded onto the control device 22, for example via the communication interface 26.

[0046] It should be understood that embodiments of the methods disclosed herein may also be implemented to calculate operational information for other types of metering pumps. The machine learning model may be integrated into such a pump or implemented by an external computing device (e.g., by a separate control unit that can be communicatively coupled to the pump and / or configured to measure the indicated amount). In addition, the machine learning model may be implemented by a remote data processing system (e.g., a cloud service) that is configured to receive (one or more) indicated amounts from the pump and / or from a separate sensor and calculate operational information as described herein.

[0047] As will be discussed further below, some embodiments of metering pumps may not be able to monitor the position of the displacement member. Such pumps may only be able to monitor the pressure in the dosing chamber or another related indicator. Therefore, in such embodiments, the machine learning model may receive pressure or other indicator as its only input.

[0048] Figure 2 An example of a pressure-stroke diagram is schematically shown, which depicts a pressure-stroke curve that can be detected by the monitoring module 24 in general or otherwise. The abscissa shows the stroke length S in percentage, i.e., the linear movement of the membrane 4 between the position where it represents the minimum volume of the dosing chamber 2 and the position where the maximum volume of the dosing chamber 2 is defined. The ordinate shows the pressure P detected by the pressure sensor 30. 0% of the stroke corresponds to the lower dead center 32, and 100% of the stroke length corresponds to the upper dead center 34. The curve illustrates four phases (phase, stage) of the membrane movement. The lower part of the curve represents the suction phase 36, the part where the pressure on the left side increases rapidly represents the compression phase 38, the upper part represents the discharge phase 40, and the part where the pressure on the right side decreases rapidly represents the expansion phase 42, wherein the internal pump volume expands. The expansion phase 42 corresponds to the movement of the membrane 4 in the direction S1 together with the suction phase 36, and the compression phase 38 and the discharge phase 40 form a pressure stroke in the direction S2.

[0049] When the monitoring module 24 of the control device 22 continuously or intermittently records or monitors the pressure and the associated displacement values, changes in the pressure-stroke curve over time or over a number of strokes can be detected by the monitoring device. Different problems or faults that may occur in the metering pump have different effects on the course of the curve in the pressure-stroke diagram.

[0050] Figure 3Examples of pressure-stroke curves of a dosing pump are shown in the presence of different faults or other operating conditions, such as cavitation, the presence of air bubbles, etc. It will be appreciated that other indicated quantities, such as force or torque, may be represented in a similar manner as a function of stroke length and therefore associated with different forms of indicated (quantity)-stroke diagrams, which may be used to detect operating conditions and estimate operating parameters in a similar manner.

[0051] Various embodiments of the methods disclosed herein provide an effective way to reliably detect such problems or faults based on recorded pressure or other indicators and displacement values.

[0052] Figure 4 A flow chart of a training method for creating a trained machine learning model for subsequent use in a method for determining operating information of a metering pump is schematically shown.

[0053] Initially, the process obtains a set of training data items. To this end, in step S1, the process obtains a set of input sequences. Each input sequence represents the value of an indicator, which indicates the activation intensity of the displacement member of the metering pump at the corresponding position of the displacement member during the operation of the metering pump. For the purpose of the following description, the embodiments of the methods and devices disclosed herein are described with primary reference to pressure as the indicator. However, it should be understood that various methods and devices can also use other indicators instead of or in addition to pressure.

[0054] The input sequence can represent the measured pressure value P(t 0 )、P(t 1 ),…,P(t n ), where time t 0 , ..., tn are the corresponding times during at least one stroke cycle of the metering pump. The number of recorded values ​​n, n>1 may depend on the rate at which the metering pump records values. Alternatively, the input sequence may represent a sequence of paired pressure and position data: (P 1 ,S 1 )、(P 2 ,S 2 ),…,(P n ,S n ), where each pair (Pi, Si) represents a pressure and a corresponding position, in particular the position of a displacement component measuring the pressure. It should be understood that, for example, when the measured pressure value always represents the pressure value at the corresponding predetermined position, the position S i may not need to be explicitly included in the input sequence. Instead, the input sequence can be represented as a pressure vector p[i], i=1, ..., n, where index i lists the predetermined position S iIt will also be appreciated that other embodiments may use other representations of the input sequence, such as using measured torque or force rather than pressure. In one embodiment, the input sequence may be represented as an array of value / data pairs. In other embodiments, the input sequence may be represented in a different manner, such as as an image of a pressure-stroke curve, where the pressure-stroke curve represents a sequence of pairs of pressure and associated position data.

[0055] The input sequence can be obtained by operating a plurality of metering pumps under various operating conditions and by recording pressure and / or position data from corresponding sensors of the pumps. For example, the recorded pressure and / or position data can be obtained by a monitoring module of the metering pumps, for example, as described above in conjunction with Figure 1 The recorded data may be received from the pump via a suitable communication interface of the pump, or they may be stored locally and later retrieved from a data storage device of the pump or a separate monitoring unit.

[0056] In step S2, each of the obtained input sequences is labeled with one or more target output values ​​indicating operation information of a pump that is operated when the input sequence has been obtained.

[0057] The target output value may be obtained in various ways, such as by manually determining an operating condition of the pump, such as an operating fault, and by assigning a category identifier representing the determined operating condition to the target value. For example, some operating conditions, such as cavitation, may be induced, for example, by blocking the inlet of the pump. In other embodiments, the target value may be obtained by performing a reference measurement of, for example, the output pressure or effective stroke length of the metering pump. For example, the target value of the effective stroke length may be determined from flow measurements and from pump speed and pump volume.

[0058] In any case, the process can store the obtained target output values ​​associated with the corresponding input sequences to which they are related, thereby creating a set of labeled training data items. Each training data item includes an input sequence and a corresponding target output, the input sequence indicating an indicator indicating the activation intensity of the displacement member of the metering pump at the corresponding position of the displacement member during the operation of the metering pump, and the corresponding target output indicating the operating information that can be observed during the operation of the metering pump. It should be understood that the training data can include multiple target values ​​for each input sequence, for example, identifiers for classifying operating conditions and / or corresponding values ​​of one or more operating parameters. Therefore, the training data can be used to train a single machine learning model or different machine learning models to output corresponding types of operating information.

[0059] In step S3, the obtained training data is used to train a machine learning model to output operation information in response to receiving multiple input values ​​(particularly in response to receiving an input sequence of input values). The machine learning model may be a feedforward neural network, such as a convolutional neural network.

[0060] For example, a convolutional neural network may receive an input indicating a pressure-stroke curve. The pressure-stroke curve may be represented as a sequence of 1D data points, a sequence of 2D data points, or a two-dimensional input representation of a pressure-stroke graph depicting the pressure-stroke curve, or otherwise. In this case, the machine learning model may be a feed-forward neural network, such as a convolutional neural network.

[0061] When the input values ​​are represented as a time series, the machine learning model can receive a one-dimensional sequence of inputs. In this case, the machine learning model can be a feedforward neural network, a recurrent neural network, or another suitable type of machine learning model.

[0062] Typically, when the input to a machine learning model is a sequence or a one-dimensional array, the network can be a dense, fully connected network, a one-dimensional convolutional network, and / or a recurring network, such as a long short-term memory (LSTM) network, or a combination thereof. For example, the first layer(s) of the network can be convolutional, and the following layers are fully connected. In one example, convolutional layers and LSTM layers can be combined, for example such that the first layer(s) are 1D convolutional, the subsequent layer(s) are LSTM(layers), and the further layer(s) are dense(layers).

[0063] The training of the machine learning model can be based on a suitable training algorithm known in the art, such as a back-propagation algorithm.

[0064] The training process produces a trained machine learning model that can then be used to predict operational information about the metering pump in response to receiving a plurality of input values ​​as described herein. To this end, a representation of the trained machine learning model can be stored in a processing or control device of the metering pump or on another data processing system external to the pump, such as as a computer program module. For example, a representation of the trained machine learning model can be loaded as a processing module 50 or in (e.g., as described above in conjunction with Figure 1 The method can be implemented in other ways (described above) to a metering pump, or in other ways.

[0065] Figure 5 A flow chart of a method for determining operating information of a metering pump is schematically shown.

[0066] In an initial step S4, the process records pressure and / or position data from a corresponding sensor of a metering pump for which operational information is to be determined. The process may record data during normal operation of the pump and represent the recorded data as an input sequence of recorded values ​​or other suitable form. It should be understood that in other embodiments, the process may obtain a detection value of another type of indicator quantity, which indicates the activation intensity of the displacement member at the corresponding position of the displacement member during the operation of the metering pump. In addition, in some embodiments, the process may only obtain the value of an indicator function; in other embodiments, the process may record the value of the indicator function and the corresponding position data. The process feeds the obtained data to a suitable processing unit, which receives the indicator function value and optionally the corresponding position data as input.

[0067] In the subsequent step S5, the process derives a plurality of input values ​​(particularly a sequence of input values) based on the detected values ​​of the indicator quantity, and feeds the input values ​​to a trained (e.g., based on the combined Figure 4 The process described above may be used to derive the input values ​​into a machine learning model (trained according to the process described above). For example, deriving the input values ​​may involve using the detection values ​​directly, or it may involve appropriate scaling of the input values ​​and / or other pre-processing steps, such as suitable filtering, etc. The specific pre-processing steps may depend on the representation of the input values ​​to be fed into the machine learning model, for example, whether the trained machine learning model is configured to receive a 2D representation of a pressure-stroke curve, a time series of (optionally normalized) pressure values, etc.

[0068] The trained machine learning model outputs one or more corresponding output values ​​in response to receiving a sequence of input values. Depending on the type of operating information represented by the output of the trained machine learning model, the output value may have different forms. For example, when the operating information represents the type of operating condition (e.g., a fault type), the machine learning model may be a classification model, and the output value may represent a category identifier that represents the operating condition of the metering pump. When the operating information represents a continuous quantity, the trained machine learning model may be a regression model, and the output value may be a continuous value representing the continuous quantity. In some embodiments, the machine learning model may output multiple output values. To this end, the machine learning model may include multiple part-models, each part-model being trained to output a specific type of operating information. Different part-models may receive the same or different representations of input values.

[0069] In step S6, the process may then return the output value directly from the trained machine learning model, or calculate the derived output value based on the output of the machine learning model. In some embodiments, the process may perform one or more actions in response to the output value, for example, issuing an alarm in response to a predicted error condition, or even performing one or more control actions for a pump of another component of the system in response to the output of the machine learning model.

[0070] Figures 6 to 8 An example of a machine learning model used in an embodiment of the method disclosed herein is schematically shown.

[0071] Specifically, Figure 6 A machine learning model 120 is shown which receives as input 110 a representation of a pressure-stroke curve, i.e., a series of data points (P i ,S i ), i = 1, ..., n, each data point represents a pair of values ​​(P i ,S i ), which represents the pressure P i and the corresponding displacement position S i . A machine learning model can treat a series of data points as two-tuples (P i ,S i ), as a 2D image of a pressure-stroke diagram, or in another suitable form. The machine learning model outputs the operational information 130 that it was trained to calculate. Examples of operational information include classification of operational conditions of the metering pump, in particular classification of error conditions of the metering pump. Other examples include values ​​of operational parameters, in particular continuous valued parameters, such as discharge pressure, effective stroke length, and discharge flow rate.

[0072] Figure 7 A machine learning model 150 is shown, which receives a pressure value P i A 1D representation of a pressure sensor is used as input 140 and is trained to output output values ​​130 indicative of operational information. The 1D representation may represent a pressure value detected at a corresponding position during a stroke cycle (or a pressure value derived from the detected value) or a time series of a series of detected or derived pressure values. In some embodiments, the 1D representation may represent a single stroke cycle starting at a predetermined position during a stroke cycle, that is, the 1D representation may be a representation of a pressure-stroke curve. In other embodiments, the 1D representation may represent another duration, such as more or less than a single stroke cycle; alternatively or additionally, the 1D representation may represent a time window starting at an unknown position during a stroke cycle.

[0073] To this end, in some embodiments, the machine learning model 150 may be or include a recurrent network. Alternatively, the machine learning model may be or include a feedforward network configured to receive a fixed-length one-dimensional input sequence representing pressure values ​​sampled over a period of predetermined length.

[0074] In any case, the machine learning model 150 outputs an output value 130 in response to receiving the time series as input. Figure 6 As described, the output value may indicate a classification of an operating state of the metering pump or indicate an operating parameter of the metering pump.

[0075] It is generally desirable to provide a machine learning model that can be easily adapted to different types of pumps, particularly without requiring large training data sets to be obtained for each individual type of pump.

[0076] For example, some metering pumps cannot output position information about the displacement of the displacement member. They can only output a pressure value or another type of indication. Therefore, it is desirable to provide a method that can also calculate the operating information of such metering pumps.

[0077] In particular, the process may receive a time series of detected pressure values ​​without information about how the time series relates to the movement of the displacement member. In particular, the process may not know the relative phase of the time series with respect to the stroke cycle, and / or the process may not know the duration of the stroke cycle with respect to the time series, i.e., how many cycles the time series covers.

[0078] therefore, Figure 8 A machine learning model 850 is shown, which receives as input a representation 840 of a time series, similar to combining Figure 7 However, the time series 840 may cover more or less than a single stroke cycle, and the start of the time series relative to a stroke cycle may also be unknown. Figure 8 The machine learning model 850 of FIG. 850 includes two partial models 851 and 853, respectively. The first partial model 851 is trained to receive the time series 840 and output a representation of a pressure-stroke curve 852 that can be used as an input to the second partial model 853. The representation of the pressure-stroke curve 852 can be a pressure value P covering a single stroke cycle and starting from a predetermined position during the stroke cycle. i The 1D representation can then be used as input to the second partial model 853, for example as Figure 7 The input 140 of the model. Optionally, the representation of the pressure-stroke curve 852 can be a series of data points (P i ,S i ), i = 1, ..., n, each data point represents a pair of values ​​(Pi ,S i ), which represents the pressure P i and the corresponding displacement position S i , such as combined Figure 6 In any case, the second part model 853 can be similar to Figure 6 Model 120 or Figure 7 The model 150 is trained to output an output value 130 indicating operational information in response to receiving a representation of the pressure stroke graph 852, for example, as combined with Figure 6 or Figure 7 It will be appreciated that the representation of the pressure-stroke curve need not be identical to that shown in FIG. Figure 8 The displacement member cycle corresponds to a 1D representation of the pressure value. Figure 7 2D representation shown.

[0079] Therefore, the first partial model 851 only needs to be trained to convert the time series 840 of pressure values ​​into a corresponding representation of the pressure-stroke curve 852. Since the pressure evolves periodically with the stroke cycle time as a period, the conversion task corresponds to detecting the phase shift of the pressure time series relative to the stroke cycle and / or the period of the stroke cycle relative to the time series 840. The inventors have recognized that it is possible to train a machine learning model to perform this task. Therefore, the second partial model 853 can be effectively reused or at least adapted to different types of metering pumps, for example by a suitable transfer learning process based on relatively few additional training examples. In particular, the second partial model can even be used or adapted for use with metering pumps that provide position information about the displacement member. Similarly, the output of the first partial model 851 can be used as an input to multiple second partial models for detecting corresponding operating conditions or estimating corresponding operating parameters.

[0080] It should be understood that the first partial model can also be divided into multiple sub-models, for example, a first sub-model that receives the time series of pressure values ​​and outputs a time series with a length corresponding to a single stroke cycle, preferably a fixed length. The second partial model can receive the output of the first partial model and output a time series of phase shifts that covers a single stroke cycle starting from a predetermined position of the displacement member (e.g., "bottom dead center"). Depending on the type of pump, specifically depending on the position / cycle information available from a given pump, the first sub-model, the second sub-model, or both sub-models can be omitted.

[0081] It should be understood that in some embodiments, the machine learning model 850 can be implemented as a single model.

[0082] Figures 9 to 11An example of prediction results obtained by the method for determining operational information of a metering pump as disclosed herein is shown. Fig. 9 Results of a classification model trained to predict five different classes of error conditions for a metering pump from a representation of a pressure-stroke curve are shown. The operating conditions include a normally operating pump and a pump operating under different types of faults (such as bubbles, cavitation, etc.). Fig. 9 shows how the predicted class relates to the known actual operating conditions of the pump. Note that the individual points are intentionally slightly offset from the classification value to allow them to be visually distinguished from each other. Fig. 9 It can be seen that the classification results clearly distinguish various operating states with only a small number of misclassifications, i.e., the trained machine learning model is found to perform accurate and reliable classification.

[0083] Fig.10 and 11 The corresponding correlations for the two trained regression models are shown. Fig.10 The model is trained to predict the outlet pressure of a metering pump from the pressure-stroke curve, and Fig.11 The model is trained to predict the effective stroke length of the metering pump. From the correlation, it can be seen that the predicted values ​​by the method disclosed in this article provide an accurate estimate of the operating parameters of the pump.

[0084] Embodiments of the methods described herein may be computer-implemented. In particular, embodiments of the method may be implemented by means of hardware comprising several different elements and / or at least in part by means of a suitably programmed data processing system. In a device claim enumerating several means, several of these means may be implemented by the same element, component or item of hardware. The mere fact that certain measures are mentioned in mutually different dependent claims or described in different embodiments does not indicate that a combination of these measures cannot be used to advantage.

[0085] In summary, the various aspects disclosed herein can be summarized as follows:

[0086] Embodiment 1: A method for determining operating information of a metering pump, the metering pump comprising a metering chamber, a displacement member and a drive motor for driving the displacement member, the method comprising: a) receiving, during operation of the metering pump, a plurality of detected values ​​indicative of an amount indicative of an activation intensity of the displacement member at respective positions of the displacement member; b) calculating the operational information based on a machine learning model, the machine learning model being trained to output the operational information in response to receiving a plurality of input values ​​obtained based on the detected values ​​of the indicated quantity.

[0087] Embodiment 2: The method according to embodiment 1, wherein the indicated quantity comprises the pressure inside the dosing chamber and / or the torque of the driving motor.

[0088] Embodiment 3: The method according to any of the preceding embodiments, wherein the operating information comprises a classification of an operating condition of the metering pump, in particular a classification of an error condition of the metering pump.

[0089] Embodiment 4: A method according to Embodiment 3, wherein the machine learning model comprises a classification model, and the classification model is trained to output an identifier of one category among multiple discrete categories.

[0090] Embodiment 5: A method according to any one of the preceding embodiments, wherein the operation information includes the value of an operation parameter.

[0091] Embodiment 6: A method according to Embodiment 5, wherein the machine learning model includes a regression model, and the regression model is trained to output the value of a continuous-valued operating parameter.

[0092] Embodiment 7: The method according to embodiment 6, wherein the operating parameter indicates one or more of the following operating parameters: discharge pressure, effective stroke length and discharge flow rate.

[0093] Embodiment 8: A method according to any one of the preceding embodiments, wherein the machine learning model is configured to receive multiple input values ​​indicating a quantity, each of the multiple input values ​​is associated with a corresponding position of a displacement member, and wherein the machine learning model is configured to output the operation information in response to receiving the multiple pairs of input data.

[0094] Embodiment 9: The method according to embodiment 8 comprises: - receiving position data indicative of a monitored position of the displacement member during operation of the metering pump, or calculating the position data from at least the received detection value of the indicated quantity; - calculating a plurality of input values ​​based on the received detected values ​​of the indicated quantity and the received or calculated position data.

[0095] Embodiment 10: A method according to any one of embodiments 1 to 7, wherein the machine learning model is configured to receive a time series of detection values ​​of an indicating quantity at corresponding time points, and output the operation information in response to receiving the time series of detection positions of the detection values ​​of the indicating quantity.

[0096] Embodiment 11: A method according to Embodiment 10, wherein the machine learning model includes a first machine learning model and a second machine learning model, the first machine learning model being configured to calculate a plurality of input values ​​of the indicated quantity based on a time series of detection values ​​of the indicated quantity at corresponding time points received during operation of the metering pump, each input value indicating the value of the indicated quantity at a corresponding position of the displacement member; and the second machine learning model being configured to output operation information in response to receiving the calculated plurality of input values.

[0097] Embodiment 12: A computer-implemented method for creating a trained machine learning model for use in the method according to embodiment 1, the training method comprising: a) obtaining a set of training data items, each training data item comprising a plurality of input values ​​and a corresponding target output, the plurality of input values ​​indicating an indication amount, the indication amount indicating an activation intensity of a displacement member of a metering pump at a corresponding position of the displacement member during operation of the metering pump, and the corresponding target output indicating operational information observable during the operation of the metering pump; b) training a machine learning model based on the obtained training data set to output operational information in response to receiving multiple input values.

[0098] Embodiment 13: A data processing system configured to execute the steps of the method defined in any one of Embodiments 1 to 12.

[0099] Embodiment 14: A metering pump comprising a metering chamber, a displacement member, a drive motor for driving the displacement member, and a data processing system as defined in Embodiment 13.

[0100] Embodiment 15: A system comprising a metering pump as defined in embodiment 13 and a data processing system; wherein the metering pump comprises a dosing chamber, a displacement member, and a drive motor for driving the displacement member.

[0101] Embodiment 16: A system according to embodiment 15, wherein the data processing system is separate from the metering pump and the data processing system includes an interface for receiving multiple detection values ​​indicating the amount of activation intensity of the displacement member at corresponding positions of the displacement member during operation of the metering pump.

[0102] Embodiment 17: The system of embodiment 15, wherein the metering pump further comprises a data processing system.

[0103] Embodiment 18: A computer program, comprising computer program code, which is configured to cause a data processing system to perform the steps of the method according to any one of embodiments 1 to 12 when the computer program code is executed by the data processing system.

[0104] It should be emphasized that the terms "comprises" and "comprising" when used in this specification are used to specify the presence of stated features, elements, steps or parts, but do not exclude the presence or addition of one or more other features, elements, steps, parts or groups thereof.

Claims

1. A method for determining operating information of a metering pump, the metering pump comprising a metering chamber, a displacement member and a drive motor for driving the displacement member, wherein the method comprises: a) receiving, during operation of the metering pump, a plurality of detected values ​​indicative of an intensity of activation of the displacement member at respective positions of the displacement member; b) calculating the operation information according to a machine learning model, wherein the machine learning model is trained to output the operation information in response to receiving a plurality of input values ​​obtained according to the detection value of the indication quantity.

2. The method according to claim 1, wherein: The indicated quantity comprises the pressure inside the dosing chamber and / or the torque of the drive motor.

3. A method according to any one of the preceding claims, wherein: The operating information comprises a classification of an operating condition of the metering pump, in particular a classification of an error condition of the metering pump.

4. The method according to claim 3, wherein: The machine learning model includes a classification model that is trained to output an identifier for one of a plurality of discrete categories.

5. A method according to any one of the preceding claims, wherein: The operation information includes values ​​of operation parameters.

6. The method according to claim 5, wherein: The machine learning model includes a regression model that is trained to output values ​​of continuous-valued operational parameters.

7. The method according to claim 5 or 6, wherein: The operating parameters indicate one or more of the following operating parameters: discharge pressure, effective stroke length, and discharge flow rate.

8. A method according to any one of the preceding claims, wherein: The machine learning model is configured to receive a plurality of input values ​​of the indication quantity, each of the plurality of input values ​​being associated with a corresponding position of the displacement member, and wherein the machine learning model is configured to output the operational information in response to receiving at least the plurality of input values.

9. The method according to claim 8, comprising: - receiving position data indicative of a monitored position of the displacement member during operation of the metering pump, or at least calculating position data from received detection values ​​of the indicated quantity; - calculating the plurality of input values ​​based on the received detection values ​​of the indication quantity and the received or calculated position data.

10. The method according to claim 8 or 9, wherein: The machine learning model is configured to receive a plurality of pairs of input data, each pair of input data comprising a position of the displacement member and a corresponding value of the indicated quantity at the position, and wherein the machine learning model is configured to output the operation information in response to receiving the plurality of pairs of input data.

11. The method according to any one of claims 1 to 9, wherein: The machine learning model is configured to receive a time series of detection values ​​of the indicative quantity at corresponding time points, and output the operation information in response to receiving the time series of detection values ​​of the indicative quantity.

12. The method according to claim 11, wherein: The machine learning model includes a first machine learning model and a second machine learning model, the first machine learning model being configured to calculate a plurality of input values ​​of the indicated quantity based on a time series of detection values ​​of the indicated quantity at corresponding time points received during operation of the metering pump, each input value indicating a value of the indicated quantity at a corresponding position of the displacement member; the second machine learning model being configured to output the operation information in response to receiving the calculated plurality of input values.

13. A computer-implemented method for creating a trained machine learning model for use in the method of any one of claims 1-12, the training method comprising: a) obtaining a set of training data items, each training data item comprising a plurality of input values ​​and a corresponding target output, the plurality of input values ​​indicating an indication amount indicating an activation intensity of a displacement member of the metering pump at a corresponding position of the displacement member during operation of the metering pump, and the corresponding target output indicating operational information observable during the operation of the metering pump; b) training a machine learning model based on the obtained training data set to output operational information in response to receiving multiple input values.

14. A data processing system configured to perform the steps of the method defined in any one of claims 1 to 13.

15. A metering pump comprising a dosing chamber, a displacement member, a drive motor for driving the displacement member, and a data processing system as defined in claim 14.

16. A system comprising a metering pump and a data processing system as defined in claim 15; wherein: The metering pump comprises a metering chamber, a displacement member, and a driving motor for driving the displacement member.

17. The system of claim 16, wherein: The data processing system is separate from the metering pump and includes an interface for receiving a plurality of detection values ​​indicative of an activation intensity of the displacement member at respective positions of the displacement member during operation of the metering pump.

18. The system of claim 16, wherein: The metering pump also includes the data processing system.

19. A computer program comprising computer program code configured to, when executed by a data processing system, cause the data processing system to perform the steps of the method according to any one of claims 1 to 13.

Images