Apparatus and method for detecting a fault condition of a pump for pumping a fluid

Abstract

Disclosed herein are embodiments of an apparatus for detecting a fault condition of a pump for pumping a fluid, the apparatus comprising: a sensor for measuring a physical property of the fluid, a housing accommodating the sensor and configured for being mounted with said sensor in operational connection with the liquid, a signal processing unit accommodated within the housing and configured to receive a sensor signal from the sensor and to process at least the received sensor signal to detect an occurrence of the fault condition, a communications interface accommodated within the housing and configured to communicate a fault signal indicative of the detected occurrence of the fault condition to a system external to the housing.

Description

[0001] APPARATUS AND METHOD FOR DETECTING A FAULT CONDITION OF A PUMP FOR

[0002] PUMPING A FLUID

[0003] Technical field

[0004] The present disclosure relates to a method and apparatus for detecting a fault condition of a pump.

[0005] Background

[0006] When operating a hydraulic system that includes one or more pumps it is desirable to reliably detect certain fault conditions, in particular the occurrence of hydraulic shock and / or the occurrence of dry-run and / or other fault conditions, such as cavitation or the a degradation of certain pump components.

[0007] Dry-run, sometimes also referred to as dry running, is an undesirable condition that occurs when a pump operates without an adequate amount of fluid. In most pump technologies, this may lead to cavitation or other undesirable consequences for the pump, including the risk of critical damage to the internal pumping elements such as impellers, lobes, gears, casings, seals, drives such as magnetic drives, and bearings.

[0008] Hydraulic shock, often also referred to as water hammer or fluid hammer, refers to a pressure surge or wave caused when a fluid in motion, in particular a liquid, is forced to suddenly stop, change direction or otherwise experience a sudden change in momentum. Hydraulic shock may e.g. occur in a fluid system when a valve in a pipe closes suddenly causing a pressure wave to propagate along the pipe. Hydraulic shock is a common cause of failure in pumps. A pump continuously experiencing hydraulic shock may eventually suffer from fatigue wear. If a hydraulic shock event is potent enough, even a single event can damage the pump. Aside from the pump, hydraulic shock events can also cause serious damage to other parts of a fluid system, e.g. to sensors and / or the like. Hydraulic shock may also cause noise and vibration in a fluid system and may even result in pipe rupture or collapse. On this background, it is generally desirable to be able to detect the occurrence of fault conditions such as dry-run, hydraulic shock and / or other fault conditions such as degradations of pump components, cavitation etc.

[0009] It is generally desirable to provide an apparatus for detecting a fault condition of a pump that is easy to install and that provides a fast and reliable detection of the fault condition.

[0010] US 2021 / 215645A1 discloses a mechanism for generating an alert of a water hammer event in a steam pipe, the mechanism comprising: sampling an accelerometer coupled to a steam pipe to provide accelerometer data; determining that the accelerometer data meets or exceeds a threshold; and generating an alert that a water hammer event has occurred based at least in part on the accelerometer data.

[0011] On this background, it remains desirable to provide an improved or at least an alternative method and / or apparatus for detecting a fault conditions in a fluid system, preferably an apparatus that can detect multiple types of fault conditions. It is further generally desirable to provide a method and / or apparatus for detecting fault conditions that is accurate, robust, fast and cost efficient.

[0012] Summary

[0013] According to one aspect, disclosed herein are embodiments of an apparatus for detecting a fault condition of a pump for pumping a fluid, the apparatus comprising:

[0014] - a sensor for measuring a physical property of the fluid,

[0015] - a housing accommodating the sensor and configured for being mounted with said sensor in operational connection with the liquid,

[0016] - a signal processing unit accommodated within the housing and configured to receive a sensor signal from the sensor and to process at least the received sensor signal to detect an occurrence of the fault condition,

[0017] - a communications interface accommodated within the housing and configured to communicate a fault signal indicative of the detected occurrence of the fault condition to a system external to the housing. Accordingly, as the sensor, the signal processing unit and the communications interface are accommodated in the same housing, the apparatus is compact and easy to install. In particular, the apparatus directly outputs a fault signal that is indicative of the occurrence of the detected fault condition and that can readily be used by an external system, external to the apparatus, e.g. by a pump controller of the pump or by a control system for controlling a hydraulic system of which the pump may be a part. The external system can use the fault signal without the need for further signal processing, i.e. without the need for programming the external system to interpret the received sensor data. Moreover, the fault signal typically only requires a relatively small communications bandwidth and can be communicated efficiently and reliably. For example, the fault detection does not rely on the transmission of large amounts of sensor data, e.g. on the transmission of sensor signals at high sampling rates. Hence, the apparatus provides a robust and reliable detection of fault conditions even in situations where reliable, high- bandwidth communication is impractical, difficult or expensive to establish, or otherwise not available. It will be appreciated that some embodiments of the apparatus may, in addition to the fault signal, also output the sensor signal and / or other signals to an external system. However, the fault detection does not rely on such communication or on the sensor signal being reliably communicated at a certain rate.

[0018] In some embodiments, the external system may be configured, based on the fault signal indicating a detected fault condition, to adjust operation of a fluid system, in particular of the pump, and / or to implement other countermeasures in order to avoid, or at least reduce, the occurrence of the fault condition, or to otherwise mitigate undesirable consequences of the occurrence of the fault condition.

[0019] As the housing is configured to be mounted to a pump, the apparatus can easily be installed. In particular, many pumps include a sensor port for attaching a sensor, e.g. a pressure sensor or a temperature sensor. Embodiments of the apparatus may readily be connected to such ports of existing pumps. The detection of the fault condition is based at least on a measured property of the fluid being pumped by the pump. To this end, the sensor is in operational, in particular fluid, connection to the fluid in the pump. Therefore, a reliable detection of various fault conditions may be achieved without having to rely on measurements of external effects of the fault condition, e.g. vibrations or sound near the pump, etc. Such external effects may be more difficult to attribute to the fault condition to be detected, as they may also have other causes.

[0020] Generally, in some embodiments, the fluid may be water or another liquid.

[0021] Generally, the sensor signal may be a time-dependent, i.e. represent the physical property at respective times.

[0022] In some embodiments, the sensor is or comprises a pressure sensor and the physical property is or comprises a fluid pressure of the fluid. The inventors have realized that various fault conditions can reliably be detected based on pressure measurements of the fluid pressure of the fluid being pumped by the pump, in particular based on detectable features or signatures of the time-dependent pressure signal. Accordingly, various embodiments of the apparatus may be configured, e.g. by suitable programming, to detect different types of fault conditions based at least in part on sensor signals of the same sensor. Some embodiments may even be configured to concurrently monitor the pump for multiple fault conditions based completely or in part on the same sensor signal.

[0023] In some embodiments, the apparatus may comprise alternative or additional sensors and / or the sensor may be configured to measure additional or alternative physical quantities. Accordingly, the sensor signal may be indicative of one or more physical quantities. For example, the apparatus may comprise a pressure sensor and a temperature sensor, either as separate sensors or as a combined sensor that is configured to measure fluid pressure and temperature of the fluid. In some embodiments, the apparatus is configured to detect the fault condition based on the pressure signal alone while, in other embodiments, the apparatus is configured to detect the fault condition based on the pressure signal in combination with one or more other sensor signals, e.g. in combination with a temperature signal.

[0024] In some embodiments, the time-dependent sensor signal is indicative of the measured physical property at a rate high enough to detect variations of the physical property at frequencies of at least 1 kHz, such as at least 4 kHz, such as at least 6 kHz, such as at least between 1kHz and 3kHz, such as at least between 1 kHz and 5 kHz. The inventors have realized that a particularly reliable detection of various fault conditions, in particular of hydraulic shock and dry-run, may be provided when the physical property is monitored over time at a sufficiently high rate. In particular, when the sensor is a pressure sensor, it is preferable that the sensor is capable of detecting pressure changes at temporal resolutions high enough to detect pressure changes in the acoustic frequency range, in particular at acoustic frequencies up to 3 kHz, such as up to 5kHz, such as up 7 kHz or even higher acoustic frequencies. Specifically, the inventors have realized that a particularly reliable distinction between various fault conditions may be obtained when the physical property is monitored over time at a sufficiently high rate, i.e. with at temporal resolutions high enough to detect pressure changes in the acoustic frequency range, in particular at acoustic frequencies up to 3 kHz, such as up to 5kHz, such as up 7 kHz or even higher acoustic frequencies, such as up to 12 kHz or even up to 20 kHz. With a sufficiently high temporal resolution, the system can reliably distinguish between different error conditions (such as hydraulic shock, cavitation, dry run, etc.), in particular when the detected signal features lie in a boundary region between different error conditions.

[0025] In some embodiments, the fault condition is one of the following fault conditions: hydraulic shock, dry-run, thermal shock, a degradation of more or more components of the pump, and / or cavitation. For example, the condition of various components of the pump, e.g. the condition of one or more bearings and / or the condition of a pump impeller and / or the condition of a drive such as a magnetic drive, may degrade over time. At least some embodiments of the apparatus disclosed herein are configured to detect such degraded conditions based on measured time-varying pressure signals. A degradation of a component of the pump may be detected by monitoring the condition of said component and by triggering a fault signal when the monitored condition is degraded beyond a predetermined acceptable level, e.g. based on a decision tree, a neural network, a Kalman filter, and / or the like. It will be appreciated that some embodiments of the apparatus disclosed herein may be configured to detect two or more of the above fault conditions and / or other fault conditions, in particular fault conditions affecting the time-varying pressure inside the pump.

[0026] In some embodiments, the signal processing unit is configured to:

[0027] - derive one or more signal features, in particular one or more predetermined signal features, from the received sensor signal,

[0028] - apply a classification process to at least the derived one or more signal features, wherein the classification process is configured to output a fault condition indicator. In some embodiments, the derived one or more signal features may be the only inputs to the classification process while, in other embodiments, classification process may receive additional inputs, e.g. one or more signal features of sensor signals from one or more additional sensors and / or other measured or recorded parameters. For example, in some embodiments, the classification process only receives one or more pressure signal features of a measured pressure signal while, in other embodiments, the process receives the one or more pressure signal features of the pressure signal and, additionally, a temperature signal or one or more temperature signal features derived from the temperature signal.

[0029] The inventors have realized that a reliable detection of various fault conditions may be performed based on a small number of respective predetermined signal features, thereby allowing the apparatus to reliably perform the fault detection even when the apparatus only includes limited computational resources and / or a limited power supply. This is in particular the case when the computed features are features of the measured signal in time domain, i.e. features whose computation do not require a transformation of the signal into frequency domain, e.g. by Fast Fourier Transformation or otherwise. For various fault conditions, in particular dry run or hydraulic shock, the inventors have found the computation of one or two signal features, in particular one or two signal features of the time-domain sensor signal, to be sufficient to provide a reliable detection. Nevertheless, alternative embodiments may use more than two signal features, e.g. between 3 and 5 signal features or even more.

[0030] Examples of signal features include: an average, e.g. a moving average such as over a moving time window of predetermined length, or a peak value of the measured physical quantity, a variance of the measured physical quantity, e.g. over a moving time window of predetermined length. a feature, e.g. a peak value, of a time derivative of the measured physical quantity.

[0031] Examples of a classification process include a classification decision tree, e.g. a classification decision tree trained by a machine-learning process, thereby providing a computationally efficient and memory-efficient classification. Other suitable classification processes include other types of machine-learning processes, e.g. neural networks or the like. In some embodiments, the process may employ a Kalman filter or another suitable method for detecting or estimating the relevant signal features, or otherwise.

[0032] In embodiments, where the apparatus is configured to detect an occurrence of two or more fault conditions, the detection of the occurrence of each of the fault conditions may be based on respective signal features, different from the signal features used for the detection of the other fault condition(s), or the detection of two or more fault conditions may partly or completely be based on the same set of one or more signal features.

[0033] The inventors have realized that a particularly reliable detection of hydraulic shock may be obtained based on detected or computed peak values of a measured pressure signal, in particular in combination with detected or computed peak values of a time-derivative of the measured pressure signal. A particularly reliable detection of hydraulic shock may be obtained when the pressure signal is measured at a sufficiently high sampling rate, thereby allowing an accurate peak detection or computation of the peak values.

[0034] Moreover, the inventors have realized that a particularly reliable detection of dry-run may be obtained based on detected average values of a measured pressure signal in combination with a detected variance of the measured pressure signal, in particular of a bandpass-filtered measured pressure signal. A particularly reliable detection of dry-run may be obtained when the pressure signal is measured at a sufficiently high sampling rate, so as to determine an accurate variance measure.

[0035] In some embodiments, the signal processing unit is further configured to derive information about one or more additional operational conditions of the pump, in particular operational conditions not related to a current occurrence of a fault condition. For example, the derived information about one or more additional operational conditions of the pump may include one or more of the following: a motor shaft speed, an indication of a current bearing condition, an indication of a current impeller condition, an indication of a current drive condition, such as a magnetic drive condition. One or more of the above and / or other types of additional operational information may be derived from measured pressure signals as disclosed herein and / or from other types of sensor signals. For example, the apparatus may be configured to output information about the current condition of one or more pump components, e.g. of bearings, the impeller, a drive and / or the like, thereby allowing to plan maintenance or other preventive measures to reduce the risk of fault conditions occurring, e.g. to reduce the risk that the monitored pump components are degraded so much as to amount to a fault condition. The current condition of a pump component may be indicative of a degree of wear or of another operational condition of the component. For example, the information may be computed from one or more derived signal features of the sensor signal, e.g. by feeding the derived one or more signal features into a regression decision tree or other suitable model, e.g. a neural network model or the like.

[0036] The present disclosure relates to different aspects including the apparatus described above and in the following, corresponding apparatus, systems, methods, and / or products, each yielding one or more of the benefits and advantages described in connection with one or more of the other aspects, and each having one or more embodiments corresponding to the embodiments described in connection with one or more of the other aspects and / or disclosed in the appended claims.

[0037] In particular, according to one aspect, disclosed herein are embodiments of a method of detecting a fault condition of a pump that is pumping a fluid, the method comprising:

[0038] - receiving a pressure signal indicative of a fluid pressure of the fluid,

[0039] - deriving one or more signal features from the received pressure signal,

[0040] - applying a classification process to the derived one or more signal features, wherein the classification process is configured to output a fault condition indicator.

[0041] It is noted that features of various embodiments of the method described herein may be implemented at least in part in software or firmware and carried out on a data processing unit or other data processing system caused by the execution of program code means such as computer-executable instructions. Alternatively, the features of the computer-implemented method may be implemented by an otherwise suitably configured data processing unit.

[0042] Accordingly, another aspect disclosed herein relates to embodiments of a data processing unit configured to perform the acts of the method described herein. To this end, the data processing unit may have stored thereon program code configured to cause, when executed by the data processing unit, the data processing unit to perform the acts of the method described herein. The data processing unit may include a memory for storing a suitable computer program.

[0043] Here and in the following, the term data processing unit comprises any circuit and / or device suitably adapted to perform the above functions. The term data processing unit comprises general- or special-purpose programmable microprocessors, Digital Signal Processors (DSP), Application Specific Integrated Circuits (ASIC), Programmable Logic Arrays (PLA), Field Programmable Gate Arrays (FPGA), Graphical Processing Units (GPU), special purpose electronic circuits, etc., or a combination thereof. The data processing unit may be a data processing unit integrated into an embodiment of the apparatus disclosed herein.

[0044] Yet another aspect disclosed herein relates to embodiments of a computer program configured to cause a data processing unit to perform the acts of the computer- implemented method described above and in the following. A computer program may comprise program code means adapted to cause a data processing unit to perform the acts of the computer-implemented method disclosed above and in the following when the program code means are executed on the data processing unit. The computer program may be stored on a computer-readable storage medium, in particular a nontransient storage medium, or embodied as a data signal. The non-transient storage medium may comprise any suitable circuitry or device for storing data, such as a RAM, a ROM, an EPROM, EEPROM, flash memory, magnetic or optical storage device, such as a CD ROM, a DVD, a hard disk, and / or the like.

[0045] Brief description of the drawings

[0046] The above and other aspects will be apparent and elucidated from the embodiments described in the following with reference to the drawing in which:

[0047] FIG. 1 shows a schematic block diagram of an embodiment of an apparatus disclosed herein when operationally connected to a pump.

[0048] FIG. 2 shows a flow diagram of a process for detecting dry-run of a pump.

[0049] FIG. 3 shows an example of a received pressure signal.

[0050] FIG. 4 shows an example of a decision tree.

[0051] FIGs. 5 and 6 show examples of signal features computed during the process of FIG. 2. FIG. 7 shows a flow diagram of a process for detecting hydraulic shock in a pump.

[0052] FIG. 8 shows a perspective exploded view of a pressure sensing assembly including an embodiment of an apparatus disclosed herein.

[0053] FIG. 9 shows an exploded cross section view of the pressure sensing assembly of FIG. 8 according to the line VII-VII defined in FIG. 8. FIG. 10 shows a perspective view of a vertical multistage pump with a moulded body of a pressure sensing assembly inserted into the pump at the top of the fluid section of the pump.

[0054] Detailed description

[0055] FIG. 1 shows a schematic block diagram of an embodiment of an apparatus disclosed herein when operationally connected to a pump.

[0056] The apparatus, generally designated by reference numeral 6, comprises a housing 600, a sensor 610, a processing unit 620 and a communication interface 630.

[0057] The housing is configured to be mounted to a pump 100 with the sensor in operational connection to the fluid being pumped by the pump. To this end, the housing 600 may be configured to be directly mounted to the pump, e.g. to a housing of the pump. Alternatively, the housing 600 may be configured to be mounted to the pump via a fitting or other intermediate component. To this end, the housing 600 may comprise a connector for connecting the housing 600 to the pump directly or to a fitting or other intermediate component, which in turn is configured to be connected to the pump.

[0058] The housing 600 accommodates the sensor 610, the processing unit 620 and the communications interface 630. In some embodiments, the housing 600 comprises a probe member that is configured to extend into the pump when the housing is mounted to the pump. To this end, the pump may comprise a port defining an opening such that the apparatus may be connected to the port with a probe member extending through the opening into the interior of the pump housing. The probe member may accommodate a sensing element of the sensor, e.g. a pressure probe and / or a temperature probe. In some embodiments a portion of the sensing element may extend out of the housing so as to allow the sensing element to be brought into direct contact with the fluid inside the pump. In other embodiments, the sensing element is completely encapsulated within the housing. In any event, the housing preferably encapsulates the processing circuit and the interface circuit such that they are sealed from the fluid being pumped. In FIG. 1, the apparatus is shown in operational connection to a pump 100. The pump 100 includes a pump housing 130, one or more impellers 110, or other fluid moving member, and a pump drive 120. The pump 100 may be a centrifugal pump, such as a vertical, multistage centrifugal pump, or a different type of pump. The pump 100 has an inlet 111 for suction of water or a different fluid, such as of a different liquid. The pump 100 also has an outlet 112 for providing the output flow of the pump. The pump drive 120 may comprise an electric motor 121 and may comprise a motor drive circuit 122. The motor drive circuit 122 may include a frequency converter for supplying the motor with electrical energy and / or other circuitry for controlling operation of the motor 121. The motor drive circuit 122 may be connectable to a suitable power supply (not shown) in order to supply the drive circuit, e.g. a frequency converter, with electric energy.

[0059] During operation, the motor 121 drives the impeller 110 causing the pump to pump fluid from the inlet 111 to the outlet 112. To this end, the motor 121 may drive a shaft 113 of the pump, which, in turn, may drive the impeller 110 of the pump 100. While FIG.l illustrates one example of a pump, it will be appreciated that the apparatus 600 may be used to detect fault conditions of other types of pumps. For example, some embodiments of the pump may include more than one impeller.

[0060] The sensor 610 may be a pressure sensor, e.g. an absolute or differential pressure sensor, for measuring a fluid pressure of the fluid inside the pump housing.

[0061] Alternatively, the sensor may be another type of sensor for measuring another physical property of the fluid inside the pump housing. In some embodiments, the apparatus may include more than one sensor, e.g. a pressure sensor for measuring the fluid pressure and a temperature sensor for measuring the temperature of the fluid inside the pump housing.

[0062] The processing unit 620 may be an ASIC, a suitably programmed microprocessor or another circuit, in particular integrated circuit. The processing unit may be configured to receive and process a sensor signal from the sensor 610 so as to detect an occurrence of a fault condition, e.g. dry-run, hydraulic shock, a degraded (e.g. broken, worn, or the like) component of the pump, cavitation, and / or the like. Responsive to the detection of a fault condition, the processing unit forwards a fault signal to communications interface 620 via a suitable internal data connection. Examples of processes for detecting fault conditions, which may be executed by processing unit 620, will be described below with reference to FIGs. 2 through 6. The apparatus may output the fault signal as a digital and / or analog signal, e.g. as a suitable voltage signal where the voltage encodes the detection result. In some embodiments, the apparatus may further output the measured sensor signal and / or a processed, e.g. filtered, version of the measured sensor signal. If the apparatus includes more than one sensor, the apparatus may output the sensor signals from all sensors. For example, in an embodiment where the apparatus includes a pressure sensor, the apparatus may output a pressure signal along with a fault signal computed based on the measured pressure signal. In embodiments, where the apparatus includes an additional temperature sensor and / or another type of sensor, the apparatus may further output the temperature signal or other sensor signal. The apparatus may output the fault signal and, optionally, the sensor signal(s) continuously, intermittently at a suitable rate, e.g. once every 1-5 s, or otherwise.

[0063] The communications interface 620 may comprise a wired or wireless communications interface configured to provide communication of analog or digital signals to an external system 300. In FIG. 1, the external system is illustrated as an entity separate from the pump 100, but it will be appreciated that the external system may also be the motor drive circuit 122 of the pump 100 or another pump controller. Alternatively, the external system 300 may be a control system of a hydraulic system, which hydraulic system may include the pump 100. It will generally be appreciated that the communications interface 630 may be configured to forward fault signals to the pump 100 and / or to an external system separate from the pump 100.

[0064] In the following two embodiments of a process for detecting a fault condition of a pump based on measured pressure signals will be described, namely a process for detecting dry-run and a process for detecting hydraulic shock, respectively. Each or both of the processes may be performed by the processing unit 620 of FIG. 1 or otherwise.

[0065] FIG. 2 shows a flow diagram of a process for detecting dry-run of a pump. In initial step S21, the process receives a measured pressure signal. Preferably, the pressure signal has a sampling rate that is high enough, or the pressure signal is otherwise indicative of the measured physical property at a rate high enough, to detect variations of the physical property at frequencies of at least 1 kHz, such as at least 4 kHz, such as at least 6 kHz, such as at least between 1kHz and 3kHz, such as at least between 1 kHz and 5 kHz. To this end, the pressure sensor performing the pressure measurement may have sufficiently fast response characteristics. In any event, the process may receive the pressure signal as a digital signal at a sufficiently high sampling rate.

[0066] FIG. 3 shows an example of a received pressure signal.

[0067] Again referring to FIG. 2, in step S22, the process performs a bandpass filtering of the received pressure signal to obtain a band-pass filtered pressure signal. In addition to the band-pass filtered pressure signal, the process further uses the received pressure signal (or another suitably filtered version thereof) as a full-bandwidth pressure signal.

[0068] In optional step S23, the full-bandwidth pressure signal and the bandpass-filtered pressure signal are calibrated to represent respective pressure values in bar or another suitable unit for representing pressure. The calibration may be based on sensor-specific calibration parameters or otherwise. In the remainder of the description of the method of FIG. 2, when reference is made to the bandpass-filtered pressure signal or the fullbandwidth pressure signal, these terms are intended to refer to the calibrated versions of these signals, respectively, when the process indeed includes the optional calibration step S23. When the process does not include a calibration step, these terms refer to the uncalibrated versions of the respective signals.

[0069] In subsequent step S24, the process performs feature detection processes on the fullbandwidth pressure signal and on the bandpass-filtered pressure signal, respectively, so as to obtain first and second signal features. The signal features may be represented as respective signal feature signals. In particular, the process may perform an averaging of the full-bandwidth pressure signal to compute a moving average of the full-bandwidth pressure signal. Moreover, the process may compute a standard deviation, e.g. over a moving window, of the bandpass-filtered pressure signal. The moving average of the full-bandpass pressure signal represents a first signal feature of the received pressure signal, while the computed standard deviation of the bandpass-filtered signal represents a second signal feature of the received pressure signal. It will be appreciated that the process may use alternative or additional signal features, e.g. another measure of the variability of the bandpass-filtered pressure signal instead of a standard deviation.

[0070] In step S25, the process feeds the first and second signal features into a classification decision tree that is configured to classify each pair of signal features into one of at least two states, e.g. "dry" and "wet," respectively, i.e. the output of the classification tree is a time series of classifications. In some embodiments, the decision tree may be configured to classify each pair of signal features into more than two states, e.g. as "dry", "inconclusive" and "wet", respectively.

[0071] FIG. 4 schematically illustrates a decision tree configured to classify each pair of signal features into one of two states, namely "dry" and "wet" respectively.

[0072] In the example of FIG. 4, the decision tree has an input node based on the first signal feature and one or more intermediate nodes, which are either the first or the second signal feature. In each node, the respective signal feature is compared to a corresponding threshold. The leaf nodes of the decision tree represent the "dry" and "wet" classifications, optionally associated with respective probabilities.

[0073] It will be appreciated that, in other embodiments, the classification decision tree may have a different tree structure. In particular, the decision tree may include leaf nodes representing an "inconclusive" classification, e.g. when the signal features are close to the respective thresholds or when the determined probabilities are smaller than a threshold probability. The classification decision tree may be a trained classification tree that has previously been trained by a suitable decision tree learning process and based on a plurality of training examples. In particular, the training examples may be created from measured pressure signals that have been measured on one or more reference pumps, using one or more reference sensors. The pressure measurements may have been performed under known conditions, i.e. when the pump was either run with sufficient fluid or in a dry-run condition. The resulting measured pressure signals, labelled with the known conditions under which the pressure signals have been obtained, may then be used to train a decision tree to correctly classify the pressure signals as either "dry" or "wet".

[0074] It will be appreciated that other embodiments may employ other types of classification processes, e.g. based on neural networks or other machine-learning models. Some embodiments may employ a Kalman filter or other suitable process.

[0075] Again referring to FIG. 2, in step S26, the process performs a voting scheme to determine a final classification based on the decision tree classifications computed at different time steps. In particular, the process may determine a majority vote of the decision-tree classifications for a predetermined number of most recent time steps. The majority vote may thus indicate a classification of the signal as "dry", "wet" or "inconclusive". The overall "inconclusive" class may e.g. be used when a clear majority for either "dry" or "wet" cannot be determined and / or when the decision trees already include an "inconclusive" class. The voting scheme may be performed at a suitable rate, e.g. once every one to five seconds, or at another suitable rate.

[0076] In step S27, the process outputs the result of the voting scheme as a final detection result. The process may output the final result as a continuous signal or at a suitable rate, e.g. depending on the frequency of the execution of the voting scheme. The detection result may then be output by an embodiment of the apparatus disclosed herein as a fault signal, e.g. as described in connection with FIG. 1, or otherwise.

[0077] FIGs. 5 and 6 show examples of the signal features computed in step S24. In particular, the upper graphs in Figs. 5 and 6 show examples of the moving average of the full- bandwidth pressure signal while the lower graphs in Figs. 5 and 6 show examples of the computed standard deviation of the bandpass-filtered signal. Furthermore, in FIGs. 5 and 6, the resulting votes of an example of the voting scheme of step S26 are indicated for respective time periods. In particular, periods during which the voting scheme resulted in an overall classification of the signal as "wet" are designated by the letter "W", periods during which the voting scheme resulted in an overall classification of the signal as "dry" are designated by the letter "D", and periods during which the voting scheme resulted in an overall classification of the signal as "inconclusive" are designated by the letter "I".

[0078] FIG. 7 shows a flow diagram of a process for detecting hydraulic shock in a pump.

[0079] In initial step S71, the process receives a measured pressure signal. As described in connection with the previous example, preferably, the pressure signal has a sampling rate that is high enough, or the pressure signal is otherwise indicative of the measured physical property at a rate high enough, to detect variations of the physical property at frequencies of at least 1 kHz, such as at least 4 kHz, such as at least 6 kHz, such as at least between 1kHz and 3kHz, such as at least between 1 kHz and 5 kHz. To this end, the pressure sensor performing the pressure measurement may have sufficiently fast response characteristics. In any event, the process may receive the pressure signal as a digital signal at a sufficiently high sampling rate.

[0080] In step S72, the process computes a derivative of the received pressure signal with respect to time, i.e. the computed derivative is indicative of the speed of pressure change.

[0081] In optional step S73, the process performs a moving average filtering of the received pressure signal and of the computed derivative signal, respectively. The moving averaging may be performed over a suitable time window, e.g. 10 ms or another suitable duration. In subsequent step S74, the process performs feature detection processes on the (optionally averaged) pressure signal and on the (optionally averaged) derivative signal. In particular, the process may perform peak detection of the pressure signal and determine the maximum peak value of the detected peaks in the pressure signal. To this end, the process may fit a suitable curve, e.g. a Gaussian or other suitable function, on the signal samples in a proximity of the detected peak and compute the maximum peak value from the fitted curve. As the pressure peaks caused by water hammer may have a short duration, an accurate computation of the maximum peak values benefits from a high sampling rate of the measured pressure signal. The computed maximum peak value may thus represent the first signal feature.

[0082] Moreover, the process may perform peak detection of the derivative signal and determine the maximum peak values of the detected peaks. It will be appreciated that the derivative signal will include positive and negative peaks indicating a fast increase and decrease of the pressure signal, respectively. The computed maximum peak values of the derivative signal may thus represent a maximum absolute value of the derivative. The computed maximum peak values of the derivative signal represent the second signal feature. As the pressure peaks caused by water hammer may have a short duration, an accurate computation of the maximum peak values of the derivative signal benefits from a high sampling rate of the measured pressure signal.

[0083] It will be appreciated that the process may use alternative or additional signal features. For example, in some embodiments, the process may only detect positive peaks or only negative peaks of the derivative signal. In other embodiments, the process may detect both positive and negative peaks and compute the absolute maximum peak values of both positive and negative peaks. In yet alternative embodiments, the process may detect pairs of peaks, where each pair includes a positive and a negative peak in short succession from each other. When the process detects pairs of peaks, the process may use both maximum peak values as second and third signal features, respectively.

[0084] In step S75, the first and second signal features are fed into a classification decision tree that is configured to classify each pair of signal features into one of four states, namely "no hydraulic shock" and the severity levels of a detected hydraulic shock, respectively, e.g. severity "1", "2" and "3". The classification into severity levels may be based on suitable thresholds of the first and second signal features. It will be appreciated that other embodiments may perform a classification into fewer classes or into more classes.

[0085] As in the previous example, the classification decision tree may be a trained classification tree that has been trained based on a plurality of training examples using decision tree learning. In particular, the training examples may be created from measured pressure signals measured on one or more reference pumps, using one or more reference sensors. The pressure measurements may have been performed under known conditions, i.e. when the pump was exposed to hydraulic shock at known times and under controlled conditions, in particular hydraulic shocks of known strength. The resulting measured pressure signals, labelled with the known conditions under which the pressure signals have been obtained, may then be used to train a decision tree to correctly classify the pressure signals as either "no hydraulic shock", "severity 1", "severity 2" or "severity 3".

[0086] In optional step S76, the process may compute an estimated energy associated with a detected hydraulic shock event. The computation may be based on the maximum peak values of the detected pressure peak and the detected peak in the derivative signal. The computation may be based on equations and / or algorithms known as such in the art.

[0087] In step S77, the process outputs the result of the classification of step S75 and, optionally, the estimated energy, as a final detection result. The detection result may then be output by an embodiment of the apparatus disclosed herein as a fault signal, as described in connection with FIG. 1 or otherwise.

[0088] The process may perform the hydraulic shock event detection continuously. In some embodiments, upon detection of a hydraulic shock event, the process may pause for a predetermined timeout interval following a hydraulic shock event detection, e.g. for about 0.5 s - 2 s, and then resume the detection. FIGs. 8-9 show a pressure sensing assembly including an embodiment of an apparatus disclosed herein. In the present example, the apparatus is embodied as a pressure sensor module 6.

[0089] The pressure sensing assembly 1 comprises a one-piece fitting 2, a first opening 4 for a pressure sensor module 6, a second opening 8 for a plug 10, a third opening 12 functioning as an air bleed opening and an cylindrical protrusion 14 with an external thread suitable for inserting the one-piece fitting into an opening with an internal thread of a hydraulic component in a hydraulic circuit. A fourth opening 15 is arranged in the one-piece fitting, passing through the cylindrical protrusion. The first, second, third and fourth openings are all in fluid connection with each other. For the sake of simplifying the drawings, the threads are not shown in the drawings.

[0090] An internal thread 16 is provided in the second opening and a complementary external thread 18 is formed on the plug 10. A first end of the plug is provided with a tapered portion 20 and the opposite end of the plug is provided with a hexagonal head 22. The plug is inserted into the second opening with the tapered portion first. The user can then displace the plug in the second opening via the hexagonal head 22. The second opening and the first opening are connected by a channel 24. Likewise, the third opening 12 is also in fluid communication with the channel. When the plug 10 is inserted into the second opening and screwed into its bottom position, the plug blocks the third opening. When the plug is screwed at least partially out of the second opening, the third opening comes into fluid communication with the channel. Fluid in the channel, typically coming from the hydraulic circuit via the fourth opening can then escape the hydraulic circuit via the third opening. This is used to bleed air out of the fluid circuit. Bleeding air out of a hydraulic circuit is well known in the art.

[0091] It should be noted that in the current embodiment, the third opening is provided as a hole through the one-piece fitting and arranged perpendicular to the centre axis of the second opening. However, other options for providing the same air bleed functionality can be provided instead. For example, in one embodiment, the third opening could be arranged as a channel parallel to the centre axis of the second opening. The channel could extend only partially along the second opening. When the plug is in the bottom position, the channel is blocked, but when the plug is withdrawn, the channel is unblocked and fluid can escape through the third opening.

[0092] The pressure sensing assembly further comprises fastening means 26, which are arranged to hold the pressure sensor module 6 in place in the first opening. The fastening means comprise a fastening bracket (or fastening element) 26 which is in the shape of a bent metal plate element. A first portion 28 of the fastening bracket is arranged to engage with a groove 30 in the one-piece fitting of the pressure sensing assembly. A second portion 32 of the fastening bracket is arranged to wrap over the sensor and a third portion 34 of the fastening bracket is provided with a hole 36 whereby the bracket can be held in place with a screw 38 which engages with a fifth opening 40 in the one-piece fitting.

[0093] The pressure sensor module 6 is provided with a cylindrical portion 42 which is inserted into the first opening. A pliable cylindrical sleeve 44 arranged on the cylindrical portion forms a sealed connection between the sensor and the one-piece fitting when the pressure sensor module is pressed into the first opening. The pressure sensor module also comprises a computing element arranged inside a housing 46 arranged on top of the cylindrical sensing portion. The housing also comprises a connector 48 suitable for connecting an electrical wire with a cooperating plug.

[0094] Accordingly, in the embodiment of FIGs. 8-9 and in various other embodiments of the apparatus for detecting a fault condition of a pump disclosed herein, the sensor for measuring a physical property of the fluid, is a sensor that is configured to be detachably coupled to the pump or to another component of a fluid system, thereby providing a flexible apparatus for detecting fault conditions. In particular, the sensor may be detachably coupled to the pump or another component of a fluid system, such that the sensor is in contact with, or otherwise in operational proximity to, the fluid being pumped by the pump. FIG. 10 shows a vertical multistage pump 200 of the kind which is known in the art. The pump itself will therefore not be described in detail in this specification. The pump comprises a pump portion 202 with an inlet 204, an outlet 206 and an electric motor 208 which drives the pump via a coupling (hidden inside the cover element 210). A one- piece fitting 212 of a pressure sensing assembly, e.g. the pressure sensing assembly of FIGs. 8-9 is inserted into an opening at the top 214 of the pump portion 202. As described above, the one-piece fitting has a first opening 216 suitable for receiving a pressure sensor module and a second opening 218 suitable for receiving a plug.

[0095] As will be known to the person skilled in the art of vertical multistage pumps, the fluid will enter the pump via the inlet and then be pressed upwardly through the housing via one or more pump stages until it reaches the top 214 of the pump. The fluid will then be pushed downwardly through channels (hidden inside the pump) to reach the outlet. There is very little flow resistance through these channels, hence, the pressure at the top of the pump and the pressure at the outlet will be essentially the same. Hence, measuring the pressure at the top of the pump is a good indication of the pressure output of the pump. In prior art systems, pressure sensors are typically placed in the fluid circuit downstream of the pump. However, there are numerous advantages to placing the sensor near the top of the pump. In addition to measuring the pressure of the fluid, the pressure sensor assembly itself can be used to bleed air from the system due to the novel construction of the moulded body. Furthermore, by placing the sensor near the top of the pump, it has been found that different characteristics of the pump, in particular various fault conditions as disclosed herein, can be measured by the pressure sensor module, in addition to the pressure of the output fluid.

[0096] It is to be noted that the figures and the above description have shown the example embodiments in a simple and schematic manner. Many of the specific mechanical details have not been shown since the person skilled in the art should be familiar with these details and they would just unnecessarily complicate this description. For example, the specific materials used and the specific manufacturing procedures have not been described in detail since it is maintained that the person skilled in the art would be able to find suitable materials and suitable processes to manufacture the pressure sensing assembly described herein.

[0097] Embodiments of the method described herein can be implemented by means of hardware comprising several distinct elements, and / or at least in part by means of a suitably programmed microprocessor. In the apparatus claims enumerating several means, several of these means can be embodied by one and the same element, component or item of hardware. The mere fact that certain measures are recited in mutually different dependent claims or described in different embodiments does not indicate that a combination of these measures cannot be used to advantage.

[0098] It should be emphasized that the term "comprises / comprising" when used in this specification is taken to specify the presence of stated features, elements, steps or components but does not preclude the presence or addition of one or more other features, elements, steps, components or groups thereof.

[0099] Some embodiments of the aspects disclosed herein may be summarized as follows:

[0100] Embodiment 1: An apparatus for detecting a fault condition of a pump for pumping a fluid, the apparatus comprising:

[0101] - a sensor for measuring a physical property of the fluid,

[0102] - a housing accommodating the sensor and configured for being mounted with said sensor in operational connection with the liquid,

[0103] - a signal processing unit accommodated within the housing and configured to receive a sensor signal from the sensor and to process at least the received sensor signal to detect an occurrence of the fault condition,

[0104] - a communications interface accommodated within the housing and configured to communicate a fault signal indicative of the detected occurrence of the fault condition to a system external to the housing.

[0105] Embodiment 2: The apparatus according to embodiment 1, wherein the sensor is or comprises a pressure sensor and the at least one physical property comprises a fluid pressure of the fluid. Embodiment 3: The apparatus according to embodiment 1 or 2, wherein the sensor is or comprises a temperature sensor and the physical property comprises a temperature of the fluid.

[0106] Embodiment 4: The apparatus according to any one of the preceding embodiments, wherein the sensor signal is indicative of temporal variations of the measured physical property at frequencies of at least 1 kHz, such as at least 5 kHz, such as at least 7 kHz, such as at least 12 kHz, such as at least 20 kHz.

[0107] Embodiment 5: The apparatus according to any one of the preceding embodiments, wherein the fault condition is one of the following fault conditions: hydraulic shock, dryrun, thermal shock, a degradation of more or more components of the pump, cavitation.

[0108] Embodiment 6: The apparatus according to any one of the preceding embodiments, wherein the signal processing unit is configured to:

[0109] - derive one or more signal features from the received sensor signal,

[0110] - apply a classification process to at least the derived one or more signal features, wherein the classification process is configured to output a fault condition indicator.

[0111] Embodiment 7: The apparatus according to any one of the preceding embodiments, wherein the signal processing unit is configured to derive one or more of the following signal features from at least the received sensor signal: a time derivative of the measured at least one physical quantity; a variance of the received sensor signal.

[0112] Embodiment 8: The apparatus according to any one of the preceding embodiments, wherein the sensor signal is a pressure signal indicative of a measured fluid pressure of the fluid, wherein the fault condition includes an occurrence of hydraulic shock, and wherein the signal processing unit is configured to detect an occurrence of hydraulic shock based at least on a feature of the received pressure signal and based on a feature of a time derivate of the received pressure signal. Embodiment 9: The apparatus according to any one of the preceding embodiments, wherein the sensor signal is a pressure signal indicative of a measured fluid pressure of the fluid, wherein the fault condition includes a dry-run condition of the pump, and wherein the signal processing unit is configured to detect an occurrence of the dry-run condition based on a time-average of the received pressure signal and based on a variance of the received pressure signal.

[0113] Embodiment 10: The apparatus according to any one of the preceding embodiments, wherein the signal processing unit is further configured to derive information about one or more additional operational conditions of the pump.

[0114] Embodiment 11: The apparatus according to embodiment 10, wherein the derived information about one or more additional operational conditions of the pump includes one or more of the following: a motor shaft speed, an indication of a current bearing condition, an indication of a current impeller condition an indication of a current drive condition.

[0115] Embodiment 12: The apparatus according to any one of the preceding embodiments, wherein the sensor for measuring a physical property of the fluid is configured to be detachably coupled to the pump or to another component of a fluid system, the fluid system including the pump.

[0116] Embodiment 13: A method of detecting a fault condition of a pump pumping a fluid, the method comprising:

[0117] - receiving a pressure signal indicative of a fluid pressure of the fluid,

[0118] - deriving one or more signal features from the received pressure signal,

[0119] - applying a classification process to the derived one or more signal features, wherein the classification process is configured to output a fault condition indicator.

[0120] Embodiment 14: The method according to embodiment 13, wherein the fault condition includes a dry-run condition of the pump, and wherein the one or more signal features include a time-average of the received pressure signal and a variance of the received pressure signal.

[0121] Embodiment 15: The method according to embodiment 14, wherein deriving comprises:

[0122] - performing a band-pass-filtering of the received pressure signal to obtain a band-pass- filtered pressure signal,

[0123] - deriving a first signal feature representative of a moving average of the received pressure signal,

[0124] - deriving a second signal feature representative of a standard-deviation of the band- pass-filtered pressure signal.

[0125] Embodiment 16: The method according to any one of embodiments 13 through 15, wherein the fault condition includes an occurrence of hydraulic shock, and wherein the one or more signal features include a peak value of the received pressure signal over a predetermined moving time window and a peak value of a time derivative of the received pressure signal over a predetermined moving time window.

[0126] Embodiment 17: The method according to embodiment 16, wherein deriving comprises:

[0127] - deriving a first signal feature representative of a peak value of the received pressure signal over a moving time window to derive a peak value,

[0128] - deriving a second signal feature representative of a peak value of a time derivative of the received pressure signal over a predetermined moving time window.

[0129] Embodiment 18: The method according to embodiment 14 through 17, wherein the classification process receives two signal features and is configured to classify the pressure signal based on the received two signal features into at least a first class representative of the presence of a fault condition and a second class representative of an absence of said fault condition.

[0130] Embodiment 19: The method according to embodiment 18, wherein the classification process is further configured to classify the pressure signal according to a set of predetermined severity levels.

Claims

Claims1. An apparatus for detecting a fault condition of a pump for pumping a fluid, the apparatus comprising:- a sensor for measuring a physical property of the fluid,- a housing accommodating the sensor and configured for being mounted with said sensor in operational connection with the liquid,- a signal processing unit accommodated within the housing and configured to receive a sensor signal from the sensor and to process at least the received sensor signal to detect an occurrence of the fault condition,- a communications interface accommodated within the housing and configured to communicate a fault signal indicative of the detected occurrence of the fault condition to a system external to the housing.

2. The apparatus according to claim 1, wherein the sensor is or comprises a pressure sensor and the at least one physical property comprises a fluid pressure of the fluid.

3. The apparatus according to any one of the preceding claims, wherein the sensor is or comprises a temperature sensor and the physical property comprises a temperature of the fluid.

4. The apparatus according to any one of the preceding claims, wherein the sensor signal is indicative of temporal variations of the measured physical property at frequencies of at least 1 kHz, such as at least 5 kHz, such as at least 7 kHz, such as at least 12 kHz, such as at least 20 kHz.

5. The apparatus according to any one of the preceding claims, wherein the fault condition is one of the following fault conditions: hydraulic shock, dry-run, thermal shock, a degradation of more or more components of the pump, cavitation.

6. The apparatus according to any one of the preceding claims, wherein the signal processing unit is configured to:- derive one or more signal features from the received sensor signal,- apply a classification process to at least the derived one or more signal features, wherein the classification process is configured to output a fault condition indicator.

7. The apparatus according to any one of the preceding claims, wherein the signal processing unit is configured to derive one or more of the following signal features from at least the received sensor signal: a time derivative of the measured at least one physical quantity; a variance of the received sensor signal.

8. The apparatus according to any one of the preceding claims, wherein the sensor signal is a pressure signal indicative of a measured fluid pressure of the fluid, wherein the fault condition includes an occurrence of hydraulic shock, and wherein the signal processing unit is configured to detect an occurrence of hydraulic shock based at least on a feature of the received pressure signal and based on a feature of a time derivate of the received pressure signal.

9. The apparatus according to any one of the preceding claims, wherein the sensor signal is a pressure signal indicative of a measured fluid pressure of the fluid, wherein the fault condition includes a dry-run condition of the pump, and wherein the signal processing unit is configured to detect an occurrence of the dry-run condition based on a time-average of the received pressure signal and based on a variance of the received pressure signal.

10. The apparatus according to any one of the preceding claims, wherein the signal processing unit is further configured to derive information about one or more additional operational conditions of the pump.

11. The apparatus according to claim 10, wherein the derived information about one or more additional operational conditions of the pump includes one or more of the following: a motor shaft speed, an indication of a current bearing condition, an indication of a current impeller condition, an indication of a current drive condition.

12. The apparatus according to any one of the preceding claims, wherein the sensor for measuring a physical property of the fluid is configured to be detachably coupled to the pump or to another component of a fluid system, the fluid system including the pump.

13. A method of detecting a fault condition of a pump pumping a fluid, the method comprising:- receiving a pressure signal indicative of a fluid pressure of the fluid,- deriving one or more signal features from the received pressure signal,- applying a classification process to the derived one or more signal features, wherein the classification process is configured to output a fault condition indicator.

14. The method according to claim, wherein the fault condition includes a dry-run condition of the pump, and wherein the one or more signal features include a timeaverage of the received pressure signal and a variance of the received pressure signal.

15. The method according to claim 14, wherein deriving comprises:- performing a band-pass-filtering of the received pressure signal to obtain a band-pass- filtered pressure signal,- deriving a first signal feature representative of a moving average of the received pressure signal,- deriving a second signal feature representative of a standard-deviation of the band- pass-filtered pressure signal.

16. The method according to any one of claims 13 through 15, wherein the fault condition includes an occurrence of hydraulic shock, and wherein the one or more signal features include a peak value of the received pressure signal over a predetermined moving time window and a peak value of a time derivative of the received pressure signal over a predetermined moving time window.

17. The method according to claim 16, wherein deriving comprises:- deriving a first signal feature representative of a peak value of the received pressure signal over a moving time window to derive a peak value,- deriving a second signal feature representative of a peak value of a time derivative of the received pressure signal over a predetermined moving time window.

18. The method according to claim 14 through 17, wherein the classification process receives two signal features and is configured to classify the pressure signal based on the received two signal features into at least a first class representative of the presence of a fault condition and a second class representative of an absence of said fault condition.

19. The method according to claim 18, wherein the classification process is further configured to classify the pressure signal according to a set of predetermined severity levels.

Images