Automated root cause diagnostics on a smart pump

Abstract

The present disclosure relates to a pump system. This disclosure proposes a method using a machine learning algorithm (MLA) in a centrifugal pump for performing automated root cause diagnostics. The method comprises: obtaining one or more sensor readings from one or more sensors that are integrated into the pump; automatically diagnosing a root cause of a failure condition of the pump, and identifying one or more remedial actions for the root cause, based on the one or more sensor readings. Further, determining whether an autonomous execution of the one or more remedial actions by the pump is possible, and, if the autonomous execution is possible, automatically executing the one or more remedial actions by the pump. Alternatively, if the autonomous execution is not possible, notifying a management entity that involvement of an operator of the pump is required.

Description

[0001] AUTOMATED ROOT CAUSE DIAGNOSTICS ON A SMART PUMP

[0002] TECHNICAL FIELD

[0003] This disclosure relates generally to the field of fluid handling and specifically to centrifugal pump systems. More particularly, the disclosure pertains to smart pump systems equipped with advanced diagnostics and monitoring capabilities for improved reliability and maintenance.

[0004] BACKGROUND

[0005] Pumps are essential components in various industrial and residential applications, serving critical functions in systems ranging from simple water circulation to complex chemical processing. Despite their utility, pump systems are susceptible to failures arising from a plurality of root causes, such as mechanical wear, electrical faults, and operational errors. Traditional diagnostic methods involve manual inspections and routine maintenance, which can be both time-consuming and costly, and often fail to preemptively identify potential failures.

[0006] There is thus an increasing demand for pump systems that not only efficiently manage the fluid handling process, but also integrate sophisticated diagnostic tools to effectively troubleshoot and predict potential issues. Such smart pump systems could significantly enhance the ability to quickly identify the root causes of pump failures onsite. This capability is crucial for the timely and effective maintenance of not just the affected pump, but also for preemptively managing the health of similar pumps within the same network or system, thus ensuring operational continuity and reducing downtime.

[0007] SUMMARY

[0008] In view of the above, this disclosure aims to introduce a smart pump system equipped with advanced diagnostics and monitoring capabilities. One objective is to enable automated root cause diagnostics directly on the pump. Another objective is to allow the pump to take corrective measures automatically in real-time. Yet another objective is to train a machine learning algorithm (MLA) to be leveraged to enable automated root cause diagnostics and remediation.

[0009] These and other objectives are achieved by the solution of this disclosure as described in the independent claims. Advantageous implementations are further defined in the dependent claims.

[0010] According to a first aspect of the disclosure, a method using a MLA in a centrifugal pump for performing automated root cause diagnostics is provided. The method comprises: obtaining one or more sensor readings from one or more sensors that are integrated into the centrifugal pump; automatically diagnosing a root cause of a failure condition of the centrifugal pump, and identifying one or more remedial actions for the root cause, based on the one or more sensor readings; determining whether an autonomous execution of the one or more remedial actions by the centrifugal pump is possible; if the autonomous execution is possible, automatically executing the one or more remedial actions by the pump; or if the autonomous execution is not possible, notifying a management entity that involvement of an operator of the centrifugal pump is required.

[0011] This disclosure provides a method using the MLA in the centrifugal pump for performing automated root cause diagnostics, thereby enhancing the diagnostic and predictive maintenance capabilities of the pump. The MLA allows the pump to be capable of monitoring in-use data in real time (the sensor readings), identifying potential root causes when facing failures, and automatically executing corrective actions if permitted. For unknown issues detected by the MLA, or issues requiring human intervention, the pump is configured to notify the management entity, e.g., located at a command center (e.g., if a large leak is detected, it may need to shut down the pump and notify the customer through the command center).

[0012] The MLA may generally be referred to as an algorithm or learning procedure. The pump may comprise a processor and a memory. The MLA may be stored in the memory and may be executed by the processor. The MLA may comprise a set of rules and / or may employ statistical techniques to identify patterns and / or correlations in the input data (here, the input data comprises the one or more sensor readings), and may use these patterns and / or correlations to pinpoint potential causes of anomalies or failures, i.e., the root causes. The MLA may learn from the patterns and / or correlations, and may make decisions or predictions without being explicitly programmed for the specific task. For example, the MLA may learn by processing the input data (the sensor readings) and using it to train a model, wherein the model may comprise a neural network. The model may represent the learned patterns and / or correlations, and the model can be used by the MLA to make the decisions or predictions autonomously based on new input data (new sensor readings). The model may be trained based on the learned patterns and / or correlations in the historical sensor readings - and additionally from training data - enabling it to analyze new sensor readings, identify deviations, and suggest likely root causes based on past correlations and outcomes. Training the MLA may refer to training the model within the MLA. The MLA may accordingly encompass both the definition of the model and the process of training the model.

[0013] Optionally, the one or more remedial actions comprise adjusting one or more operational parameters controlling the centrifugal pump, and autonomously executing the identified one or more remedial actions by the centrifugal pump comprises automatically adjusting the one or more operational parameters of the centrifugal pump.

[0014] For instance, the MLA can detect a deviation from the standard operating parameters. It can thus detect the root cause and if the root cause allows it, it can further take corrective or remedial actions at the centrifugal pump, e.g., correcting the deviation from the standard operating parameters.

[0015] Optionally, the method further comprises sending an event notification to the management entity after the one or more remedial actions have been executed, wherein the event notification comprises one or more of the following:

[0016] - the diagnosed root cause of the failure condition,

[0017] - the one or more remedial actions, and a limitation for the autonomous execution of the one or more remedial actions by the centrifugal pump. In some cases, it is possible that the failure condition could only be partially fixed by the centrifugal pump, i.e., using the MLA, and a manual inspection or repair from a technician is still needed. The notification may thus indicate to the management entity the diagnosed root cause, the autonomous actions that have been taken, and remaining or unsolved issues. For instance, the notification may indicate the necessity to replace worn-out components that are not capable of being autonomously replaced by the centrifugal pump system.

[0018] Optionally, the step of notifying the management entity that the involvement of the operator of the centrifugal pump is required comprises: providing the failure condition and / or the one or more identified remedial actions to the management entity, and / or providing the one or more sensor readings to the management entity.

[0019] For instance, the management entity may receive, if the MLA is unable to determine a potential root cause, a notification to such effect, along with parameter data illustrating the deviation for further analysis at the management entity.

[0020] Optionally, the failure condition, the one or more identified remedial actions, and / or the one or more sensor readings are provided to the management entity through a communication network.

[0021] It may be understood that the centrifugal pump and / or the sensors are communicatively coupled to enable communication for transmitting real-time sensed data to the management entity.

[0022] Optionally, the communication network comprises one of the following types: Internet, wide-area communication network, local area communication network, and a private communication network.

[0023] It may be understood that the centrifugal pump can be connected to the communication network in different ways - either through cables or radio waves, depending on what’s available or preferable in a given situation. Optionally, the method further comprises updating the MLA using a feedback loop mechanism, after the one or more remedial actions have been executed, based on a result of the executed remedial actions.

[0024] Possibly, the outcomes of unknown issues are fed back to update a model of the MLA to improve the accuracy and effectiveness of the MLA.

[0025] Optionally, the method is performed by a control entity integrated into the centrifugal pump.

[0026] Notably, the control entity is preloaded with the MLA comprising a trained model.

[0027] Optionally, the centrifugal pump maybe a booster pump of a set of booster pumps, and the method may further comprise operating the centrifugal pump to maintain a pressure setpoint and / or a flow setpoint, and to share a fluid load among the set of booster pumps.

[0028] This cooperative operation enables more precise regulation of system pressure and / or flow, enhances energy efficiency by dynamically adjusting pump activity based on demand, and extends the operational lifespan of each pump through balanced load distribution.

[0029] According to a second aspect of the disclosure, a method for training an MLA for performing automated root cause diagnostics in a centrifugal pump is provided. The method comprises: monitoring one or more sensor readings from one or more sensors that are integrated into the centrifugal pump; analyzing the one or more sensor readings to determine an abnormal reading that indicates a failure condition of the centrifugal pump; determining a root cause of the failure condition, and identifying a remedial action for the root cause based on the abnormal reading; and training the MLA based on the determined root cause, the identified remedial action, and the abnormal data.

[0030] This disclosure further proposes a training procedure for training the MLA that is to be loaded into the centrifugal pump, or a control entity of the centrifugal pump. The MLA is trained to process the sensed data and extract meaningful features that can be used to (i) determine the root cause of failures, and (ii) identify remedial actions. In other words, the MLA is trained to establish a baseline for normal operational patterns, to detect deviations from this norm to determine a potential root cause, and to identify remedial actions associated with the determined potential root cause.

[0031] Optionally, the method further comprises pre-training the MLA using a training dataset, wherein the training dataset comprises real-case data associated with a known failure condition and a respective remedial action.

[0032] Optionally, the training dataset may be obtained from a storage medium connected to the management entity. The storage medium may include a database that contains known issues / root cause, their corresponding data signatures, and the remedial action taken to solve the failure.

[0033] Optionally, the training dataset comprises one or more of the following data:

[0034] - event type data indicating whether an event is a false positive or a true root cause,

[0035] - operating parameter data indicating the one or more sensor readings of the one or more sensors of the centrifugal pump,

[0036] - true root cause data identifying an actual identified root cause,

[0037] - action data indicating at least one remedial action, decision interval data indicating a time at which the known failure condition was detected or occurred.

[0038] It may be understood that each type of data stored in the training dataset plays a role in building a comprehensive understanding of system operations, issues, and maintenance practices. Together, these data help in training the MLA - in particular a model of the MLA - more effectively by providing real-world examples and outcomes, thereby improving system reliability and performance through detailed analytics and adaptive learning, and enhancing predictive maintenance capabilities by understanding patterns and causes of failures. For instance, using historical information to train the MLA helps in refining the accuracy of the diagnostic algorithms by learning from the past, such as past mistakes (false positives) and successes (accurate identification of root causes), operational data are critical for monitoring the centrifugal pump's performance, detecting anomalies, and diagnosing issues. They form the basis of operational insights and are essential for ongoing system evaluation.

[0039] Optionally, the method further comprises performing a feedback loop mechanism to continuously update the MLA based on outcomes from implemented remedial actions.

[0040] Tracking the implemented remedial actions helps in evaluating the effectiveness of different response strategies and maintaining a history of interventions for each system component, which can be useful for training the MLA to improve the accuracy and effectiveness of the MLA.

[0041] Optionally, the step of performing the feedback loop mechanism comprises: receiving operator feedback related to an accuracy of determining the root cause and identifying the remedial action; and updating the MLA by training the MLA based on the operator feedback.

[0042] The feedback from the operator involves providing insights, corrections, or assessments based on the output of the MLA. For instance, if the MLA makes a prediction or decision that the operator knows is incorrect, they would provide feedback indicating the error. It ensures that the MLA remains relevant and efficient in dynamic environments or changing conditions.

[0043] Optionally, the method is performed by a management entity for managing one or more centrifugal pumps, and wherein the method comprises receiving, by the management entity, the one or more sensed readings from the one or more sensors through a communication network.

[0044] Possibly, the MLA is trained by the management entity. The management entity obtains the operational data for training the MLA through the communication network. Optionally, the communication network comprises one of the following types: Internet, wide-area communication network, local area communication network, and a private communication network.

[0045] Optionally, the method is performed by a control entity integrated into the centrifugal pump.

[0046] In another example, the MLA may be trained by the control entity of the centrifugal pump. The control entity may obtain the training dataset from the management entity through the communication network.

[0047] According to a third aspect of this disclosure, a system is provided. The system comprises: a centrifugal pump configured to pump a fluid; one or more sensors integrated into the centrifugal pump and configured to monitor one or more operational parameters of the centrifugal pump; and a control entity integrated into the centrifugal pump, wherein the control entity is configured to perform the method according to the first aspect, and / or to perform the method according to the second aspect.

[0048] Optionally, the system further comprises a management entity, configured to receive a notification from the control entity, wherein the notification indicates a requirement for operator involvement, or is an event notification from the control entity after the one or more remedial actions have been executed. The event notification comprises one or more of the following information:

[0049] - the diagnosed root cause of the failure condition,

[0050] - the one or more remedial actions, and a limitation for the autonomous execution of the one or more remedial actions by the centrifugal pump.

[0051] Optionally, the management entity is further configured to receive the diagnosed root cause of the failure condition and the one or more remedial actions from the control entity, and / or receive the one or more sensor readings from the one or more sensors. Optionally, the system is a fluid system comprising a plurality of centrifugal pumps, the system being configured to maintain a pressure setpoint and / or a flow setpoint and to distribute fluid load among the plurality of centrifugal pumps.

[0052] Notably, the multi-pump arrangement introduces redundancy, allowing continuous system operation even if one pump is offline, for example, due to maintenance or failure. These advantages make the application particularly suitable for high-demand or mission-critical fluid systems, such as those used in high-rise buildings, municipal water supplies, or industrial process environments, where reliability, scalability, and energy optimization are essential.

[0053] According to a fourth aspect of this disclosure, a method of controlling a fluid system comprising a plurality of centrifugal pumps is provided. The method comprises controlling each of the centrifugal pumps using the method according to the first aspect or any implementation form of the first aspect.

[0054] Optionally, the plurality of centrifugal pumps form a booster system configured to maintain a pressure setpoint and / or a flow setpoint, and the method further comprises: operating the plurality of centrifugal pumps cooperatively to maintain the pressure setpoint and / or the flow setpoint of the booster system by distributing fluid load among the plurality of centrifugal pumps.

[0055] According to a fifth aspect of this disclosure, a computer program product, stored on a non-transitory computer-readable medium, is provided. The computer program product comprises instructions that, when executed by a processor, perform the method according to the first aspect, or the method according to the second aspect.

[0056] All steps that are performed by the various entities described in the present application as well as the functionalities described to be performed by the various entities are intended to mean that the respective entity is adapted to or configured to perform the respective steps and functionalities. Even if, in the following description of specific embodiments, a specific functionality or step to be performed by external entities is not reflected in the description of a specific detailed element of that entity that performs that specific step or functionality, it should be clear for a skilled person that these methods and functionalities can be implemented in respective software or hardware elements or any kind of combination thereof.

[0057] BRIEF DESCRIPTION OF DRAWINGS

[0058] The above-described aspects and implementation forms will be explained in the following description of specific embodiments in relation to the enclosed drawings, in which

[0059] FIG. 1 shows a method according to an embodiment of this disclosure;

[0060] FIG. 2 shows a schematic system according to an embodiment of this disclosure;

[0061] FIG. 3 shows an in-use flow of the MLA according to an embodiment of this disclosure;

[0062] FIG. 4 shows a method according to an embodiment of this disclosure; and

[0063] FIG. 5 shows a training flow of the MLA according to an embodiment of this disclosure.

[0064] DETAILED DESCRIPTION OF EMBODIMENTS

[0065] Illustrative embodiments of methods for training and using an MLA for performing automated root cause diagnostics in a centrifugal pump are described with reference to the figures. Although this description provides a detailed example of possible implementations, it should be noted that the details are intended to be exemplary and in no way limit the scope of the application.

[0066] In this disclosure, an embodiment / example may refer to other embodiments / examples. For example, any description including but not limited to terminology, element, process, explanation, and / or technical advantage mentioned in one embodiment / example is applicable to the other embodiments / examples. The same elements are labeled with the same reference signs and may function similarly or likewise. FIG. 1 shows a method 100 using an MLA in a centrifugal pump for performing automated root cause diagnostic, according to an embodiment of this disclosure. The method 100 comprises a step 101 of obtaining one or more sensor readings from one or more sensors that are integrated into the centrifugal pump. The method 100 further comprises a step 102 of automatically diagnosing a root cause of a failure condition of the centrifugal pump, and identifying one or more remedial actions for the root cause, based on the one or more sensor readings. Further, the method 100 comprises a step 103 of determining whether an autonomous execution of the one or more remedial actions by the centrifugal pump is possible. If the autonomous execution is possible, the method 100 further comprises a step 104 of automatically executing the one or more remedial actions by the centrifugal pump. Or, if the autonomous execution is not possible, the method 100 further comprises a step 105 of notifying a management entity that involvement of an operator of the centrifugal pump is required.

[0067] Advancements in sensor technology, data analytics, and connectivity now allow for the development of smart pumps that can monitor various operational parameters in real time. These systems can analyze data to detect anomalies that precede failures, thereby enabling immediate corrective actions and informed maintenance decisions. Despite these advancements, there remains a significant need for improved smart pump systems that are more reliable, user-friendly, and capable of self-diagnosing complex issues effectively and efficiently. This disclosure seeks to fulfill this need by providing a method using an MLA in a pump for performing automated root cause diagnostics, thereby enhancing the diagnostic and predictive maintenance capabilities of the pump.

[0068] It may be understood that once the MLA has been trained, it is loaded into the pump, and the root cause detection and remedial action can be now handled by the pump directly. Embodiments of this disclosure discuss a training phase of the MLA and an in-use phase of the MLA.

[0069] Notably, the pump discussed in the present application is a centrifugal pump, which operates by converting rotational kinetic energy from a motor into hydrodynamic energy, efficiently moving fluids through the system using a rotating impeller and a volute casing. For the ease of description, it may only be referred to as "the pump," but it is understood to be the centrifugal pump. In particular, the centrifugal pump described in this disclosure maybe used in a variety of fluid systems. A fluid system refers to a collection of fluid-handling components that interact with each other and their surroundings to transport, regulate, or utilize fluids, including liquids, gases, or combinations thereof. These systems encompass not only the physical components but also the dynamic behavior and properties of the fluids as they move and exert forces.

[0070] Fluid systems cover a broad range of applications, including but not limited to hydronic systems and booster systems. Hydronic systems use water as the heat transfer medium for heating or cooling purposes and typically include components such as boilers, chillers, pumps, and piping networks to circulate the heated or cooled water. Within such systems, booster systems play an integral role in maintaining adequate pressure levels to ensure sufficient flow across the system.

[0071] Booster systems are designed to increase water pressure in scenarios where the existing pressure is insufficient to meet operational demand. They are commonly used in high- rise buildings, municipal water supply networks, irrigation systems, and data centers. Depending on the design and capacity requirements, a booster system may comprise a single centrifugal pump or multiple pumps operating in coordination. Multi-pump configurations are often employed to enhance redundancy, reliability, and load distribution, with individual pumps dynamically activated or deactivated to accommodate changes in demand. For example, in large-scale installations, multiple pumps may be used to ensure continuous operation in case of failure or to efficiently manage peak load conditions. Conversely, in smaller installations, a single pump may be sufficient to meet the system's pressure and flow requirements.

[0072] The present application is applicable to both standalone centrifugal pumps and multipump arrangements operating within such fluid systems, including but not limited to booster and hydronic configurations.

[0073] The method 100 shown in FIG. i describes the use of MLA. Optionally, the method 100 maybe performed by a control entity 12 integrated into the centrifugal pump, as shown in FIG. 2. The control entity 12 may comprise a processor or processing circuitry (not shown) configured to perform, conduct, or initiate the various operations of the control entity 12 described herein. The processing circuitry may comprise hardware and / or the processing circuitry maybe controlled by software. The hardware may comprise analog circuitry digital circuitry, or both analog and digital circuitry. The digital circuitry may comprise components such as application-specific integrated circuits (ASICs), field- programmable arrays (FPGAs), digital signal processors (DSPs), or multi-purpose processors. The control entity 12 may further comprise memory circuitry, which stores one or more instruction(s) that can be executed by the processor or by the processing circuitry, in particular under the control of the software. For instance, the memory circuitry may comprise a non-transitory storage medium storing executable software code which, when executed by the processor or the processing circuitry, causes the various operations of the control entity 12 to be performed. In one example, the processing circuitry comprises one or more processors and a non-transitory memory connected to the one or more processors. The non-transitory memory may carry executable program code which, when executed by the one or more processors, causes the control entity 12 to perform, conduct, or initiate the operations or methods described herein.

[0074] FIG. 2 illustrates a schematic system according to an embodiment of this disclosure. FIG. 2 especially shows a system 1 comprising a centrifugal pump 10 configured to pump a fluid; one or more sensors 11 integrated into the centrifugal pump 10 and configured to monitor one or more operational parameters of the centrifugal pump 10; and a control entity 12 integrated into the centrifugal pump 10. Notably, the control entity 12 may be configured to perform the method steps shown in FIG. 1. The centrifugal pump 10, which has been integrated with the control entity 12 and the sensors 11, is considered a smart centrifugal pump.

[0075] In one implementation, the pump system 1 maybe a booster system comprising one or more centrifugal pumps 10 connected in parallel and configured to collectively maintain a pressure setpoint and / or a flow setpoint. Each centrifugal pump in the booster system may be equipped with its own control entity 12, allowing them to operate independently or collaboratively, including sharing fluid load and managing redundancy.

[0076] The one or more sensors n may include vibration sensors, pressure sensors, temperature sensors, and flow rate sensors that provide real-time data on pump operations.

[0077] The system 1 further comprises a management entity 13, which may be a command center server. Typically, the command center server acts as the central point of communication, and is used for monitoring, controlling, and for data analysis for managing the operations of one or more pumps 10 within a network. The management entity 13 may be connected to a Human-Machine Interface (HMI) to facilitate interactions between the system and a human operator.

[0078] It may be understood that the centrifugal pump 10 and / or the sensors 11 are communicatively coupled to enable communication for transmitting real-time sensed data to the management entity 13 in the command center. In other words, the smart pump and the management entity 13 communicate through a communication network 14. Optionally, the communication network 14 comprises one of the following types: Internet, wide-area communication network, local area communication network, and a private communication network. It maybe understood that the communication link to the communication network 14 can be wired or wireless.

[0079] The management entity 13 maybe configured to receive a notification from the control entity 12. Optionally, the notification indicates a requirement for operator involvement, or is an event notification from the control entity 12 after the one or more remedial actions have been executed.

[0080] The system 1 may further comprise a storage medium 20 connected to the management entity 13. The storage medium 20 is configured to store data received by the management entity 13 from the one or more sensors 11. Details regarding the storage medium 20 will be discussed in the later part of the application. FIG. 3 further illustrates a flow chart of how the control entity 12 uses the MLA during the in-use phase, according to an embodiment of this disclosure. The embodiment of FIG. 3 is based on the embodiment of FIG. 1. That is, the control entity 12 is at least configured to perform the method 100 shown in FIG. 1.

[0081] The MLA, or the control entity 12 using the MLA, is configured to automatically diagnose a root cause of a failure condition of the centrifugal pump 10 based on data received from the one or more sensors 11. The control entity 12 determines whether the centrifugal pump 10 is able to take corrective measures automatically, and proceeds accordingly based on the outcome of the determination. In particular, if the autonomous correction is possible, the centrifugal pump 10 automatically executes the one or more remedial actions identified by the control entity 12. If the autonomous correction is not possible, the control entity 12 is configured to notify the management entity 13 that involvement of an operator of the centrifugal pump 10 is required.

[0082] In an optional embodiment, the MLA can detect a deviation from the standard operating parameters. It can thus detect the root cause and if the root cause allows it, it can further take corrective or remedial actions at the centrifugal pump 10.

[0083] Optionally, the one or more remedial actions may comprise adjusting one or more operational parameters controlling the centrifugal pump 10. The control entity 12 may be configured to instruct the autonomously execution of the identified one or more remedial actions by the centrifugal pump 10, and for instance, instruct the automatically adjustment of the one or more operational parameters of the centrifugal pump 10.

[0084] Alternatively, if the control entity 12 is unable to take corrective action by itself (for instance, a replacement of parts is required, or an unknown issue), it notifies the command center 13 for human / operator intervention. For example, the notification can identify the root cause and identify the remedial action that needs to be taken by the command center 13. Alternatively, the notification can indicate that the control entity 12 is not able to identify the root cause and further investigation by the operator is required. Optionally, the control entity 12 maybe configured to send an event notification to the management entity 13 after the one or more remedial actions have been executed. Possibly, the event notification comprises one or more of the following:

[0085] - the diagnosed root cause of the failure condition,

[0086] - the one or more remedial actions, and a limitation for the autonomous execution of the one or more remedial actions by the centrifugal pump 10.

[0087] For example, the failure condition could only be partially fixed by the smart pump, and a manual inspection or repair from a technician is still needed. The notification may indicate to the management entity 13 the diagnosed root cause, the autonomous actions that have been taken, and remaining or unsolved issues. For instance, the notification may indicate the necessity to replace worn-out components that are not capable of being autonomously replaced by the pump system. In another example, the control entity 12 may decide to temporarily shut down the pump to reduce the risk of damage, but it needs to first alert the maintenance personnel (the management entity 13) to check and confirm.

[0088] Optionally, the step of notifying the management entity 13 that the involvement of the operator of the pump 10 is required comprises: providing the failure condition and / or the one or more identified remedial actions to the management entity 13, and / or providing the one or more sensor readings to the management entity 13.

[0089] In a possible embodiment, the management entity 13 may receive, if the MLA is unable to determine a potential root cause, a notification to such effect, along with parameter data illustrating the deviation for further analysis at the management entity 13.

[0090] Optionally, the failure condition, the one or more identified remedial actions, and / or the one or more sensor readings are provided to the management entity 13 through a communication network 14. Further information or data may be requested by the management entity 13.

[0091] It may be understood that the management entity 13, namely the command center server, continuously receives data from connected pumps 10 equipped with various sensors measuring parameters like pressure, temperature, flow rate, and vibration. This data provides real-time insights into the operation of each pump 10. Advanced analytics are applied to the collected data to assess performance trends, predict potential failures, and optimize operations. Operators use the interfaces provided by the command center server to monitor the real-time data and status of the pump systems.

[0092] This disclosure proposes a smart pump capable of conducting self-root cause diagnostics and actively reporting the results to a command center, rather than just sending raw operational data. This shift from passive data reporting to active diagnostic communication enhances system efficiency, reduces downtime, and improves maintenance practices.

[0093] The MLA allows the pump 10 to be capable of monitoring in-use data in real time, identifying potential root causes when facing failures, and automatically executing corrective actions if permitted. For unknown issues detected by the MLA or issues requiring human intervention, the centrifugal pump 10 is configured to notify the command center (e.g., if a large leak is detected, it may need to shut down the pump and notify the customer through the command center). In some cases, the outcomes of unknown issues are fed back into the system to update the MLA to improve the accuracy and effectiveness of the MLA.

[0094] The proposed system enables automatic alarm handling and may result in reduced downtime. By pushing the MLA to the centrifugal pump 10, less data is being transmitted to the command center (reduced cost, bandwidth, etc.). Automated detection of root cause at the centrifugal pump 10 helps to move towards automatic handling of events on the command center, and reduces the need for someone to analyze the events.

[0095] Automated and self-diagnostic results can provide specific insights into potential issues before they escalate into failures. For example, if the control entity 12 of the centrifugal pump 10 detects abnormal vibrations or temperatures that indicate wear, it can alert the command center before the condition worsens. This allows maintenance to be scheduled proactively, reducing the risk of unexpected breakdowns. With detailed diagnostic information, maintenance can be more targeted and efficient. Rather than performing broad, routine maintenance, technicians can address specific parts or systems identified by the diagnostics, saving time and resources.

[0096] Self-diagnostics that identify and communicate specific problems allow for quicker responses from the command center. Operators don’t have to wait for data analysis to reveal a problem; they receive an immediate alert with details about the nature of the issue, speeding up the response process.

[0097] Possibly, the control entity 12 maybe configured to perform self-diagnostics regularly, to ensure that each component of the centrifugal pump 10 is functioning optimally. This continuous monitoring helps maintain operational efficiency, as the system can adjust operations based on the health and performance data provided by the centrifugal pump 10.

[0098] Early detection of issues such as leaks, blockages, or inefficiencies can lead to cost savings and a lower environmental impact. By addressing issues early, the overall system faces less downtime. This reliability is crucial in industries where continuous operation is essential, such as water treatment facilities or manufacturing plants. Regular diagnostics and timely maintenance can extend the life of the pump. Preventing severe failures and addressing wear and tear early helps avoid the stresses that can shorten equipment life.

[0099] Over time, the data collected from regular self-diagnostics can be used to analyze longterm trends and behaviors of the pump system. This accumulated data is valuable for refining predictive maintenance algorithms and improving the overall design and operation of future pump systems. Continuous feedback on the pump’s condition helps in maintaining high standards of operation and can be used in quality assurance and improvement strategies.

[0100] FIG. 4 shows a method 200 for training an MLAfor performing automated root cause diagnostics in a centrifugal pump, according to an embodiment of the disclosure. The method 200 comprises a step 201 of monitoring one or more sensor readings from one or more sensors that are integrated into the centrifugal pump; a step 202 of analyzing the one or more sensor readings to determine an abnormal reading that indicates a failure condition of the centrifugal pump; a step 203 of determining a root cause of the failure condition, and identifying a remedial action for the root cause based on the abnormal reading; and a step 204 of training the MLA based on the determined root cause, the identified remedial action, and the abnormal data.

[0101] This disclosure further proposes a training procedure for training the MLA that is to be loaded into the centrifugal pump 10 of the system 1 shown in FIG. 2.

[0102] In a possible embodiment, the management entity 13 as shown in FIG. 2 comprises a processor (not shown) configured to train the MLA. In another embodiment, the MLA maybe trained by the control entity 12 as shown in FIG. 2.

[0103] The MLA is trained to process the sensed data and extract meaningful features that can be used to (i) determine the root cause of failures, and (ii) identify remedial actions. In other words, the MLA is trained to establish a baseline for normal operational patterns, to detect deviations from this norm to determine a potential root cause, and to identify remedial actions associated with the determined potential root cause.

[0104] FIG. 5 shows a training flow of the MLA according to an embodiment of this disclosure. The embodiment of FIG. 5 is based on the embodiment of FIG. 4.

[0105] Optionally, the method 200 shown in FIG. 4 further comprises pre-training the MLA using a training dataset, wherein the training dataset comprises real-case data associated with a known failure condition and a respective remedial action.

[0106] Optionally, the training dataset comprises one or more of the following data:

[0107] - event type data indicating whether an event is a false positive or a true root cause,

[0108] - operating parameter data indicating the one or more sensor readings of the one or more sensors of the centrifugal pump 10,

[0109] - true root cause data identifying an actual identified root cause, - action data indicating at least one remedial action, decision interval data indicating a time at which the known failure condition was detected or occurred.

[0110] It may be understood that “Event type” refers to categorizing incidents or alerts generated by the system. A "false positive" means the system identified what it thought was a problem, but this was an error— no actual problem existed. A "true root cause" refers to a genuine issue correctly identified by the system. Using this information to train the MLA helps in refining the accuracy of the diagnostic algorithms by learning from past mistakes (false positives) and successes (accurate identification of root causes).

[0111] Operating parameters include all the operational data collected from sensors on the centrifugal pump 10, such as temperature, pressure, flow rate, and any other relevant parameters. These data points are critical for monitoring the pump's performance, detecting anomalies, and diagnosing issues. They form the basis of operational insights and are essential for ongoing system evaluation.

[0112] True root cause refers to the actual identified root cause. This item specifies the definitive cause of an issue after it has been investigated and confirmed. Recording the true root causes is crucial for understanding why problems occur and for training staff or automated systems to prevent future occurrences. This data is also valuable for refining predictive maintenance strategies.

[0113] Action data includes steps performed by the management entity 13 or by the customer / servicing agent, including any parts exchanged. This records all actions taken in response to issues, including troubleshooting steps, repairs, and parts replacements, whether these actions were taken remotely by the command center or on-site by a technician. Tracking these actions helps in evaluating the effectiveness of different response strategies and maintaining a history of interventions for each system component, which can be useful for training the MLA.

[0114] A decision interval may refer to the timestamps interval where the problem can be seen in the data, or where the problem happened, i.e., the specific time intervals captured in the data where the problem manifests or occurred. Storing the decision intervals allows for precise analysis of the conditions leading up to, during, and after an event. It aids in time-based analysis, helping to pinpoint when particular issues are more likely to occur and under what conditions.

[0115] It may be understood that each type of data stored in the training dataset plays a role in building a comprehensive understanding of system operations, issues, and maintenance practices. Together, these data help in training the MLA more effectively by providing real-world examples and outcomes, thereby improving system reliability and performance through detailed analytics and adaptive learning, and enhancing predictive maintenance capabilities by understanding patterns and causes of failures.

[0116] Optionally, the training dataset is obtained from the storage medium 20 shown in FIG. 2. Notably, the storage medium 20 shown in FIG. 2 may include a database that contains known issues / root cause, their corresponding data signatures, and the remedial action taken to solve the failure. How the storage medium 20 is populated is not limited in this application, and it may be populated using historical data or supplemented through the in-use loop-back mechanism. In the loop-back mechanism, data generated during the operation of the system is fed back into the database in realtime or near-real-time. This allows the system to continuously update and refine its dataset based on current operational feedback. The storage medium 20 essentially contains the training dataset for training the MLA.

[0117] Optionally, the method further comprises performing a feedback loop mechanism to continuously update the MLA based on outcomes from implemented remedial actions.

[0118] Optionally, the step of performing the feedback loop mechanism comprises: receiving operator feedback related to an accuracy of determining the root cause and identifying the remedial action; and updating the MLA by training the MLA based on the operator feedback.

[0119] The following includes a non-exhaustive list of root causes that could be included in the database to help identify and address issues in a centrifugal pump system: • Valve before pump is closed - no flow alarm (e.g., the valve located before the pump is closed, preventing any fluid from reaching the pump, which triggers a no-flow alarm)

[0120] • Valve after pump is closed - no flow alarm (e.g., the valve located after the pump is closed, blocking the flow of fluid out of the pump, resulting in a no-flow alarm)

[0121] • Broken / Worn out component (e.g., any part of the pump system, such as impellers, bearings, or seals, is damaged or has deteriorated over time, impacting the pump's functionality)

[0122] • Poor control selection (e.g., the controls configured for operating the pump are not optimal for the pump's application or the system's demands, leading to inefficiency or malfunction)

[0123] • Sensor malfunction (e.g., one or more sensors that monitor the pump's operating conditions (like pressure or flow sensors) are not working correctly, giving false readings or no data)

[0124] • System degraded (e.g., the overall pump system has deteriorated due to age, lack of maintenance, or harsh operating conditions).

[0125] • Air in the system (e.g., air trapped within the pump or pipes can cause decreased efficiency, sputtering flow, and potential damage to the pump)

[0126] • Fast closing valve - water hammer (e.g., a valve that closes too quickly can cause a water hammer, a pressure surge when the fluid in motion is forced to stop or change direction suddenly).

[0127] • No water (e.g., the pump is running without water (or the intended fluid), which can lead to a dry-run situation, potentially damaging the pump)

[0128] • Dry-run detection (e.g., the system has detected that the pump is operating without fluid flow, which is critical to prevent damage due to overheating and wear).

[0129] • Water hammer (e.g., similar to the issue with a fast-closing valve, this occurs when a sudden change in fluid velocity creates a shock wave through the system, potentially causing extensive damage)

[0130] • Low system pressure - cavitation (e.g., the pressure in the system drops too low, causing cavitation, where vapor bubbles form and collapse within the pump, leading to significant damage) • High liquid temperature - cavitation (e.g., high fluid temperatures lower the pressure at which vapor forms, leading to cavitation under conditions that normally wouldn’t cause it)

[0131] • Faulty wiring (e.g., electrical wiring issues can lead to improper pump operation or failures due to incorrect or intermittent power supply)

[0132] • Pump spinning the wrong way around (e.g., the pump rotates in the opposite direction from what is intended, usually due to incorrect wiring or configuration, reducing its effectiveness or causing damage)

[0133] • Incorrect installation (e.g., the pump or associated components are not installed according to specifications, leading to operational problems or failures)

[0134] • Incorrect configuration (e.g., the settings or setup of the pump system are not aligned with operational needs or manufacturer’s recommendations, causing poor performance or damage)

[0135] Notably, a variety of sensors in pump systems are used to monitor different parameters, ensuring efficient operation and early detection of potential issues. Each type of sensor serves a specific purpose, helping in the overall management and maintenance of the system. The present disclosure may apply to the variety of sensors in pump systems. A list of common types of sensors found in pump systems may be as follows:

[0136] • Flow sensor, which measures the rate at which fluid is moving through the pipes. It helps in assessing whether the pump is operating within its designed flow rate and detecting issues like blockages or leaks.

[0137] • Pressure sensor, which monitors the pressure within the pump system and pipelines. It is critical for ensuring that the pump operates within safe pressure limits and for detecting pressure drops or spikes that could indicate problems like pipe bursts, clogged filters, or faulty valves.

[0138] • Temperature sensor, which measures the temperature of the fluid being pumped and / or the pump components. It is important to monitor the thermal performance of the pump, preventing overheating, and ensuring that fluid temperatures remain within specified limits for safe operation.

[0139] • Vibration sensor, which detects vibrations in the pump mechanism. It is used to identify imbalances, bearing failures, or misalignments that could lead to mechanical wear or failure. • Level sensor, which measures the level of fluid in a tank or reservoir. It is essential for preventing dry running of the pump, managing tank capacities, and controlling the filling or emptying processes.

[0140] • Speed sensor, which monitors the rotational speed of the pump shaft. It is vital for ensuring the pump operates at the correct speed, optimizing efficiency, and preventing speed-related malfunctions.

[0141] • Current / power sensor, which monitors the power consumption of the pump. For instance a pump having a high power consumption at a given operating point compared to before may indicate that the motor is working harder to run at the same operating point. This may, e.g., be due to clogging.

[0142] Notably, there are also other types of sensors that, while not directly related to water pumps, play vital roles in various specialized applications. These may include position sensors, pH sensors, conductivity sensors, gas sensors, chemical composition sensors, etc. This disclosure may also apply to those sensors.

[0143] Additionally, the database may comprise data for each root cause. The following list includes a non-exhaustive list of data that could be included in the database:

[0144] • Motor power (e.g., i and 2 - measured in different places)

[0145] • Motor speed

[0146] • Motor current

[0147] • Motor temperature

[0148] • Electronics temperature

[0149] • Liquid temperature

[0150] • Outlet pressure

[0151] • Inlet pressure

[0152] • Pressure difference

[0153] • Vibration - through the accelerometers (measured in different spots)

[0154] • Hydrophone data

[0155] • Other data from the motor such as, start and stop counter, control mode, warnings, alarms, ...

[0156] • Estimated flow

[0157] • Estimated head

[0158] • Estimated efficiency • Estimated hydraulic power

[0159] • Ultrasonic data

[0160] • Ambient temperature

[0161] • Ambient humidity

[0162] Table i shows an example correspondence table of root causes, corresponding data, and related sensors.

[0163] According to another embodiment of this disclosure, a system is provided. The system maybe the system 1 shown in FIG. 2. The system 1 comprises: a pump 10 configured to pump a fluid; one or more sensors 11 integrated into the pump 10 and configured to monitor one or more operational parameters of the pump 10; and a control entity 12 integrated into the pump 10, wherein the control entity 12 is configured to perform the method according to the previous embodiments related to the use of the MLA, for instance, the embodiments shown in FIG. 1 and FIG. 3. Optionally, the control entity 12 may further be configured to perform the method according to the previous embodiments related to the training of the MLA, for instance, the embodiments shown in FIG. 4 and FIG. 5.

[0164] Optionally, the system 1 further comprises a management entity 13, configured to receive a notification from the control entity 12, wherein the notification indicates a requirement for operator involvement, or is an event notification from the control entity after the one or more remedial actions have been executed. The event notification comprises one or more of the following information:

[0165] - the diagnosed root cause of the failure condition,

[0166] - the one or more remedial actions, and a limitation for the autonomous execution of the one or more remedial actions by the pump.

[0167] Optionally, the management entity 13 is further configured to receive the diagnosed root cause of the failure condition and the one or more remedial actions from the control entity 12, and / or receive the one or more sensor readings from the one or more sensors 11.

[0168] According to another embodiment of this disclosure, a computer program product, stored on a non-transitory computer-readable medium, is provided. The computer program product comprises instructions that, when executed by a processor, perform the method according to the previous embodiments related to the use of the MLA, for instance, the embodiments shown in FIG. 1 and FIG. 3.

[0169] According to another embodiment of this disclosure, a computer program product, stored on a non-transitory computer-readable medium, is provided. The computer program product comprises instructions that, when executed by a processor, perform the method according to the previous embodiments related to the training of the MLA, for instance, the embodiments shown in FIG. 4 and FIG. 5.

[0170] Notably, the centrifugal pump 10 discussed in this disclosure may be a smart pump, e.g., a pump integrated with processing capabilities, sensors, software, and connectivity features that is able to perform automated monitoring and control, and 1 particularly benefits from the above-described advantages provided by the control entity 12.

[0171] The present disclosure has been described in conjunction with various embodiments as examples as well as implementations. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed matter, from the studies of the drawings, this disclosure and the independent claims. In the claims as well as in the description the word “comprising” does not exclude other elements or steps and the indefinite article “a” or “an” does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation.

Claims

Claims1. A method (100) using a machine learning algorithm, MLA, in a centrifugal pump (10) for performing automated root cause diagnostics, the method comprising: obtaining one or more sensor readings from one or more sensors (n) that are integrated into the centrifugal pump (10); automatically diagnosing a root cause of a failure condition of the centrifugal pump (10), and identifying one or more remedial actions for the root cause, based on the one or more sensor readings; determining whether an autonomous execution of the one or more remedial actions by the centrifugal pump (10) is possible; if the autonomous execution is possible, automatically executing the one or more remedial actions by the centrifugal pump (10); or if the autonomous execution is not possible, notifying a management entity (13) that involvement of an operator of the centrifugal pump (10) is required.

2. The method (100) according to claim 1, wherein the one or more remedial actions comprise adjusting one or more operational parameters controlling the centrifugal pump (10), and autonomously executing the identified one or more remedial actions by the centrifugal pump (10) comprises: automatically adjusting the one or more operational parameters of the centrifugal pump (10).

3. The method (100) according to claim 1 or 2, comprising: sending an event notification to the management entity (13) after the one or more remedial actions have been executed, wherein the event notification comprises one or more of the following:- the diagnosed root cause of the failure condition,- the one or more remedial actions, and a limitation for the autonomous execution of the one or more remedial actions by the centrifugal pump (10).

4. The method (100) according to one of the claims 1 to 3, wherein notifying the management entity (13) that the involvement of the operator of the centrifugal pump (10) is required comprises: providing the failure condition and / or the one or more identified remedial actions to the management entity (13), and / or providing the one or more sensor readings to the management entity (13).

5. The method (100) according to claim 4, wherein the failure condition, the one or more identified remedial actions, and / or the one or more sensor readings are provided to the management entity (13) through a communication network (14).

6. The method (100) according to claim 5, wherein the communication network (14) comprises one of the following types: Internet, wide-area communication network, local area communication network, and a private communication network.

7. The method (100) according to one of the claims 1 to 6, comprising: updating the MLA using a feedback loop mechanism, after the one or more remedial actions have been executed, based on a result of the executed remedial actions.

8. The method (100) according to one of the claims 1 to 7, wherein the method is performed by a control entity (12) integrated into the centrifugal pump (10).

9. The method (100) according to one of the claims 1 to 8, wherein the centrifugal pump (10) is a booster pump of a set of booster pumps, the method further comprising: operating the centrifugal pump (10) to maintain a pressure setpoint and / or a flow setpoint, and to share a fluid load among the set of booster pumps.

10. A method (200) for training a machine learning algorithm, MLA, for performing automated root cause diagnostics in a centrifugal pump (10), the method comprising: monitoring one or more sensor readings from one or more sensors (11) that are integrated into the centrifugal pump (10); analyzing the one or more sensor readings to determine an abnormal reading that indicates a failure condition of the centrifugal pump (10);determining a root cause of the failure condition, and identifying a remedial action for the root cause based on the abnormal reading; and training the MLA based on the determined root cause, the identified remedial action, and the abnormal data. n. The method (200) according to claim 10, comprising: pre-training the MLA using a training dataset, wherein the training dataset comprises real-case data associated with a known failure condition and a respective remedial action.

12. The method (200) according to claim 11, wherein the training dataset comprises one or more of the following data:- event type data indicating whether an event is a false positive or a true root cause,- operating parameter data indicating the one or more sensor readings of the one or more sensors (11) of the centrifugal pump (10),- true root cause data identifying an actual identified root cause,- action data indicating at least one remedial action, decision interval data indicating a time at which the known failure condition was detected or occurred.

13. The method (200) according to one of the claims 10 to 12, comprising: performing a feedback loop mechanism to continuously update the MLA based on outcomes from implemented remedial actions.

14. The method (200) according to claim 13, wherein performing the feedback loop mechanism comprises: receiving operator feedback related to an accuracy of determining the root cause and identifying the remedial action; and updating the MLA by training the MLA based on the operator's feedback.

15. A system (1) comprising: a centrifugal pump (10) configured to pump a fluid;one or more sensors (n) integrated into the centrifugal pump (10) and configured to monitor one or more operational parameters of the centrifugal pump (10); a control entity (12) integrated into the centrifugal pump (10), wherein the control entity (12) is configured to perform the method (100) according to one of the claims 1 to 6, and / or to perform the method (200) according to one of the claims 7 to 13.

16. The system (1) according to claim 15, further comprising: a management entity (13), configured to receive a notification from the control entity (12), wherein the notification indicates a requirement for operator involvement, or is an event notification from the control entity (12) after the one or more remedial actions have been executed, wherein the event notification comprises one or more of the following information:- the diagnosed root cause of the failure condition,- the one or more remedial actions, and a limitation for the autonomous execution of the one or more remedial actions by the centrifugal pump (10).

17. The system (1) according to claim 16, wherein the management entity (13) is further configured to: receive the diagnosed root cause of the failure condition and the one or more remedial actions from the control entity (12), and / or receive the one or more sensor readings from the one or more sensors (11).

18. The system (1) according to any one of the claims 15 to 17, wherein the system (1) is a fluid system comprising a plurality of centrifugal pumps (10), the system (1) being configured to maintain a pressure setpoint and / or a flow setpoint, and to distribute fluid load among the plurality of centrifugal pumps (10).

19. A method of controlling a fluid system comprising a plurality of centrifugal pumps, the method comprising controlling each of the centrifugal pumps using the method according to any of claims 1 to 9.

20. The method according to claim 19, wherein the plurality of centrifugal pumps form a booster system configured to maintain a pressure setpoint and / or a flow setpoint, and the method further comprises: operating the plurality of centrifugal pumps cooperatively to maintain the pressure setpoint and / or the flow setpoint of the booster system by distributing fluid load among the plurality of centrifugal pumps.

21. A computer program product, stored on a non-transitory computer-readable medium, comprising instructions that, when executed by a processor, perform the method (100) according to one of the claims 1 to 9, or the method (200) according to one of the claims 10 to 14.

Images