Sensor device for monitoring a centrifugal pump

Abstract

The invention relates to a sensor device (7) for monitoring a centrifugal pump (1), comprising at least one signal input (9) including a current transformer input, an analog input and / or a digital input, a low power wide area network (LPWAN) radio interface (13) and a processor-based data processing unit (12), wherein the data processing unit (12) is configured to acquire at least one monitoring signal characterizing the state of the centrifugal pump (1) at least every second by means of the at least one signal input (8), to perform an analytical and data-reducing preprocessing of the at least one monitoring signal with regard to an alarm state of the centrifugal pump (1), a statistical evaluation of the state of the centrifugal pump (1) and / or a state monitoring and prediction of the centrifugal pump (1),and transmitting at least one pre-processed monitoring signal no more frequently than every 5 minutes via the LPWAN radio interface (13) to a cloud infrastructure (14) set up for post-processing the at least one transmitted pre-processed monitoring signal.

Description

BE2025 / 5154 2 Monitoring systems enable detailed recording of operating parameters, but require extensive cabling, a continuous power supply, and complex data processing. In addition, such monitoring systems often generate large amounts of data, the transmission and analysis of which consumes considerable resources and storage space.5 More recently, wireless monitoring solutions have been developed that offer more flexible installation and improved scalability. In particular, cellular-based technologies enable data transmission over long distances. However, these technologies present specific challenges in transmitting with minimal energy expenditure.10 One problem with conventional solutions is a high frequency of measurement transmissions, which leads to unnecessary energy consumption and high network load. 15 Description of the invention Starting from this situation, the present task is to propose a sensor device for monitoring a centrifugal pump and a corresponding method,which enables energy-efficient and wireless data transmission with flexible and even retrospective installation. The problem of the invention is solved by the features of the independent claims. Advantageous embodiments are specified in the dependent claims. 25 Accordingly, the problem is solved by a sensor device for monitoring a centrifugal pump, comprising at least one signal input including a current transformer input, an analog input and / or a digital input, a Low Power Wide Area Network (LPWAN) radio interface and a processor-based data processing device, 30 wherein the data processing device is configured to acquire, at least every second, a monitoring signal characterizing a state of the centrifugal pump by means of at least one signal input, and to perform an analyzing and / or data-reducing preprocessing of at least one monitoring signal, particularly with regard to an alarm state of the centrifugal pump.to perform a statistical evaluation of the condition of the centrifugal pump and / or condition monitoring and prediction of the centrifugal pump, and to transmit at least one pre-processed monitoring signal not more often than every 5 minutes via the LPWAN radio interface to a cloud infrastructure for post-processing of at least one transmitted, pre-processed monitoring signal. In other words, it is a sensor device for monitoring a centrifugal pump, which is in particular equipped with a variety of signal inputs, 15 including at least a current transformer input, an analog input or a digital input. The sensor device also includes an LPWAN radio interface, which enables energy-efficient and wireless transmission of data. The processor-based data processing unit of the sensor device is configured to receive at least one monitoring signal at short time intervals, at least once per second.from which at least its signal input is received and processed. The captured monitoring signals are subjected in particular to data-reducing preprocessing. The preprocessing also includes the analysis of the monitoring signals with regard to a possible alarm state of the centrifugal pump, a statistical evaluation with regard to a pump state and / or continuous condition monitoring and prediction. After preprocessing, the monitoring signals are sent at defined intervals, which are not shorter than 5 minutes, via the LPWAN radio interface to a cloud infrastructure. The cloud infrastructure is specifically designed to further process the transmitted, preprocessed monitoring signals, for example, to analyze them further and, if necessary, to initiate further measures. 2025 / 5154 BE2025 / 5154 4 The sensor device for monitoring a centrifugal pump includes at least one signal input, which can be designed as a current transformer input, analog input and / or digital input,so that the sensor device can detect various types of monitoring signals, for example, electrical current values ​​for power monitoring, analog vibration signals for detecting mechanical irregularities, or digital signals from the centrifugal pump. The current transformer input is preferably intended for connecting a current transformer, also called a current sensor or current measuring transformer, to measure electrical currents and prepare them for further processing. The current transformer converts high electrical currents, which, for example, flow through a mains connection cable of the centrifugal pump, into smaller, proportional currents or voltages, which can then be detected reliably and accurately by at least one signal input of the sensor device. An example of a monitoring signal that at least characterizes the condition of the centrifugal pump is a monitoring of the motor current of the centrifugal pump. If the electric motor of the centrifugal pump draws more current than in regular operation, this may indicate mechanical problems such as a blocked shaft,Wear or a defective bearing indicates. To reliably measure this motor current, the current transformer is placed around one of the current-carrying conductors of the mains connection cable. The current transformer generates a weaker signal that is fed into the sensor device via the current transformer input and analyzed there. Another possible application is load monitoring in industrial pump systems, where current consumption serves as an indicator of the overall efficiency of the pump system. Through continuous recording and analysis of the current consumption, the sensor device can detect deviations earlier and help to improve the energy efficiency of the pump system and reduce maintenance costs. The advantage of the current transformer input or the current transformer itself is that non-contact measurement is possible. The LPWAN radio interface enables energy-efficient wireless data transmission via protocols such as LoRaWAN, LTE-ModernNB-IoT,which makes the 30 sensor device particularly suitable for retrofitting in remote or difficult-to-access areas. At least one monitoring signal 2025 / 5154 BE2025 / 5154 5 describes a state of the centrifugal pump, for example, a current consumption to identify load changes or frequency spectra to detect cavitation effects. At least one monitoring signal is subjected to analyzing and / or data-reducing preprocessing, which may include, for example, fault signal detection by pattern matching, long-term statistical evaluation of operating trends, or predictive maintenance forecasting by detecting deviations from normal operating values. The term analytical preprocessing means, in particular, that the sensor device examines the incoming monitoring signals in order to identify meaningful characteristics. For example, the sensor device can detect patterns in power consumption, vibration patterns, or temperature changes.which indicate an emerging fault such as an imbalance, cavitation or a seized bearing. This analysis can be based on various methods, for example frequency analyses, in particular Fast Fourier Transform, statistics, for example mean or standard deviation, or threshold tests, for example detection of a critical current increase. The term data-reducing preprocessing means, in particular, that the captured and analyzed monitoring signals are compressed to such an extent that the monitoring signals can be processed and transmitted with minimal storage and energy expenditure, without losing relevant information. For example, instead of a continuous current signal with a high sampling rate, the sensor device can only record statistical parameters, such as average current per minute, modified waveforms, such as reduced frequency resolution, or binary states, such as "normal operation" or "error".pass on. An example is the monitoring of the electric motor of the centrifugal pump, where, instead of transmitting the entire current profile to the cloud in real time, only whether the current consumption is within a normal range or whether a significant increase has been detected is reported. 30 2025 / 5154 BE2025 / 5154 6 This analyzing and / or data-reducing preprocessing minimizes network load, extends the battery life of wireless sensors for capturing at least one monitoring signal, and improves the efficiency of downstream data analysis in the cloud. 5 The preprocessed monitoring signals are sent to the cloud infrastructure via the LPWAN interface no more frequently than every 5 minutes, which is specifically designed for further analysis, long-term storage, and possible control measures. The processor-based data processing unit is in particular a central computing module of the sensor device, which is responsible for the acquisition,Data-reducing analysis is specifically responsible for reducing the amount of data subsequently transmitted and transmitting the monitoring signals as a reduced data set. The processor-based data processing device typically consists of a microcontroller, MCU, or microprocessor, which is specifically designed to process the monitoring signals, especially in real time, to filter out relevant information by means of preprocessing, and to transmit only essential information. The cloud infrastructure refers in particular to a networked, scalable IT platform that provides computing power, storage, and / or software services and is accessible via the Internet or private networks. In the sensor device described here, the cloud infrastructure serves in particular as a central data processing and analysis platform that further processes the received monitoring signals.long-term storage and / or intelligent diagnostics are enabled. Through the cloud infrastructure, powerful data analysis algorithms can be used for post-processing, for example, to evaluate trends, perform historical comparisons, and / or detect critical anomalies, such as an impending bearing failure due to increasing irregularities in power consumption. Post-processing can include, or is preferred over pre-processing, the statistical long-term evaluation of operational trends or predictive maintenance forecasting. If such an anomaly is detected, the cloud can send an automated maintenance signal to maintenance personnel and / or create a pro-30 active maintenance plan. 2025 / 5154 BE2025 / 5154 7 Furthermore, the cloud infrastructure can simultaneously store and compare transmitted, pre-processed monitoring signals from hundreds of sensor devices to identify system-wide patterns or error patterns across multiple locations. The cloud infrastructure can consist of public,private or hybrid cloud systems exist. Advantages of cloud infrastructure include high computing power, scalability, automatic software updates, access from anywhere, and / or the ability to develop more precise fault diagnoses and maintenance strategies for centrifugal pumps through artificial intelligence (AI) and machine learning. At the same time, the cloud reduces the local hardware requirements of the sensor device, as complex data processing and long-term storage are outsourced to the cloud. According to a preferred further education, the sensor device includes two or three current transformer inputs, at least one analog output, at least one digital output, and / or an RS485 interface. The analog input is preferred for an input- The RS485 interface, also known as EIA-485 or RS-485, is an industry standard for a physical interface for asynchronous serial data transmission. The RS485 interface conforming to the TIA-485-A standard is preferred.The sensor device is designed in particular according to ANSI / TIA / EIA-485-A-98(R2003). Furthermore, the sensor device preferably features a USB-C interface, especially for programming, whereby the USB-C interface is preferably designed according to IEC62680-1-3 (July 2016). Such further development enables versatile integration of the sensor device into different application areas and / or monitoring and control systems. According to another preferred design, the analyzing and / or data-reducing preprocessing includes data compression and / or selection of at least one monitoring signal. Data compression can be performed, for example, by lossless methods such as delta coding, where only changes between successive monitoring signals are stored, or by run-length coding, where long segments of identical monitoring signals are compressed. Alternatively, lossy compression can be used natively, such as by Fast Fourier Transform (FFT).for the conversion of time-based monitoring signals into a frequency spectrum, where only the relevant frequency component is stored and transmitted. The selection of monitoring signals can be implemented, for example, by a threshold method, in which only monitoring signals above or below a defined limit are processed further, or by event-based selection, in which the sensor device only transmits monitoring signals when unusual behavior of the centrifugal pump is detected, such as sudden vibrations or power fluctuations. Adaptive selection can also be used, in which a recording rate varies depending on the operating state of the centrifugal pump, for example, a higher recording frequency in the case of noticeable changes and a reduced frequency during normal operation. These methods contribute to increasing the energy efficiency of the sensor device, saving storage space, and optimizing wireless communication.in particular to reduce the amount of data and improve the efficiency of transmission over the LPWAN radio interface.15 According to a further preferred design, the analyzing and / or data-reducing preprocessing includes amplitude and / or time reduction of the least significant monitoring signal. Amplitude reduction can be achieved by quantization, in which continuous monitoring signals are mapped to a limited number of levels, so that finer monitoring signal variations are neglected, or by smoothing techniques such as moving-average filters, which eliminate small fluctuations and retain only relevant monitoring signals. Alternatively, an amplitude clipping method can be used, in which only monitoring signals are over- Data is stored at or below certain threshold values ​​in order to detect only significant changes in a pump's state. A reduction in time can be achieved through downsampling, where monitoring signals are recorded only at longer time intervals.For example, from 1 ms to 100 ms, which reduces the amount of data that needs to be generated and processed. Another possibility is event-driven sampling, in which monitoring signals are only recorded or processed when a specific change in the signal waveform is detected, such as a sudden pressure or vibration deviation of the centrifugal pump. These designs significantly reduce the data load, extend the battery life of a battery-powered sensor device, and optimize the use of available bandwidth for wireless transmission via the LPWAN radio interface, thereby increasing processing and transmission efficiency. Another preferred design is a data transmission rate of the least tense monitoring signal≥300, 500, 1000 or 2000 greater than at least a transmitted,pre-processed monitoring signal. This design allows for significant data reduction. A large amount of raw data can be captured, analyzed, and compressed by the sensor device as monitoring signals before only a greatly reduced amount of data is transmitted to the cloud infrastructure as pre-processed monitoring signals. An example is the monitoring of the power consumption of the centrifugal pump, where the monitoring signal is captured as a raw signal with a sampling rate of 10 kHz, while after pre-processing only a compressed status value or a statistical parameter, again average or peak value per minute, is transmitted as a pre-processed monitoring signal. Similarly, as another example, a vibration sensor can continuously measure at 1 kHz, whereby only when critical thresholds are reached or anomalies are detected is a summarized information signal sent as a pre-processed monitoring signal. Another example is a Fourier analysis of a pump noise.From millions of raw data points, only conspicuous frequency changes are extracted as monitoring signals and transmitted as a pre-processed monitoring signal. Such advanced training improves the efficiency of wireless transmission over LPWAN and simultaneously ensures that only relevant information for monitoring the centrifugal pump is retained. Specifically, the advanced training allows for significant data reduction before transmission and conservation of bandwidth and energy consumption. According to a further preferred design, at least one transmitted, pre-processed monitoring signal is of data type Boolean. After this further development, at least one monitoring signal, after pre-processing, can only assume two 20 possible states, for example 0 or 1, true or false, or on or off. This design enables maximum data reduction, since instead of complex raw data, only a binary value is sent as monitoring signals.which significantly reduces transmission time and energy consumption. A concrete example is the detection of an alarm condition of the centrifugal pump, where the sensor device can continuously measure vibrations, current consumption, or temperature and process these monitoring signals. Instead of sending the monitoring signals with high resolution, only a Boolean value is transmitted, indicating whether a critical threshold has been exceeded, for example, 1 equals alarm or 0 equals no alarm. Another example is condition monitoring based on pattern evaluation, where an internal model of the sensor device analyzes whether the centrifugal pump is operating in a normal or regular, or faulty, operating mode and sends only a Boolean signal as a result, for example, "pump is running perfectly" equals 0 or "maintenance required" equals 1. Further training is particularly suitable for energy-efficient LPWAN networks, as the data load is reduced to a minimum.while still reliably transmitting decision-relevant information. In other words, such a binary representation of the monitoring signals significantly reduces the complexity of the transmitted data. According to another preferred design, the LPWAN radio interface is set up to transmit at least one pre-processed monitoring signal via a Long Range Wide Area Network (LoRaWAN) network protocol, a Long Term Evolution for Machines (LTE-M) network protocol, and / or a Narrow Band IoT network protocol. LoRaWAN is a radio protocol specifically developed for the Internet of Things (IoT) that operates in license-free frequency bands and is optimized for long ranges and low energy consumption, allowing sensors to send monitoring signals over several kilometers to gateways, cloud infrastructure, or the like.for example, for monitoring centrifugal pumps in remote industrial plants without wired infrastructure. 30 LTE-M is based on the LTE mobile communications standard and offers higher bandwidth and lower latency than LoRaWAN, making it suitable for applications requiring nearer real-time data transmission, such as monitoring pump vibrations in real time within a municipal waterworks. NarrowBandIoT (NB-IoT), on the other hand, is a narrowband mobile communications technology that can penetrate particularly deep into buildings or underground areas, enabling reliable monitoring of centrifugal pumps in basements or shafts. NB-IoT also allows for high scalability with many networked data processing devices within a network. By supporting the various LPWAN protocols, the sensor device can be flexibly adapted to different use cases.by selecting a suitable transmission technology depending on the infrastructure and requirements. 10 According to a further preferred configuration, transmission occurs continuously and / or periodically and / or no more often than every 10, 15, 20, 30 or 60 minutes. Such further development enables flexible adaptation to different monitoring requirements. Continuous transmission means, in particular, that the sensor device sends the pre-processed monitoring signal in real time or with minimal delay, which is particularly advantageous in safety-critical applications, for example, when a sudden increase in the current draw of the centrifugal pump indicates an impending blockage and an immediate warning must be issued. 20. Periodic transmission, on the other hand, occurs particularly at fixed time intervals, such as every 30 minutes or every hour, which is useful for applications where an immediate response is not required.such as a long-term analysis of wear patterns or efficiency trends of the centrifugal pump. Limiting a transmission rate to certain minimum intervals, such as 10, 15, or 60 minutes, helps to minimize the energy consumption of the sensor device and network load, especially for battery-powered sensor devices used in hard-to-reach areas. For example, the sensor device in a remote water treatment plant can send operating parameters of the centrifugal pumps every 20 minutes to save energy, while an industrial high-performance centrifugal pump in a factory is monitored continuously or every 10 minutes to be able to react to deviations early. 2025 / 5154 BE2025 / 5154 12 According to another preferred design, the recording is carried out continuously and / or periodically and / or at least every 500, 250, 100, 50, 25, 5 or 1 ms. Continuous recording means in particular that the monitoring signal is recorded without interruption in real time.This is particularly important in highly dynamic processes, for example, when the centrifugal pump exhibits strong load fluctuations and sudden changes in power consumption or vibrations need to be detected. Periodic acquisition, on the other hand, takes place particularly at regular intervals, approximately every 100 or 500 milliseconds, which can be sufficient if only long-term trends or medium-term changes, such as a gradual increase in power consumption due to wear, are to be monitored. Choosing a higher sampling rate, for example, 1 ms, is particularly relevant for applications where rapid events such as pressure surges or cavitation effects need to be detected in order to adjust the centrifugal pump control accordingly. A lower sampling rate of 500 ms or more, however, is useful if only slowly varying operating conditions are to be monitored.For example, to determine general energy consumption over extended periods. Through flexible configuration of the acquisition rate, the sensor device can be optimized between high data resolution and energy-saving operation, depending on requirements, to ensure application-optimized data acquisition. 20 According to a further preferred design, the sensor device has a fixing device for mounting the sensor device on a DIN rail, in particular in a control cabinet. The fixing device can consist of a standardized snap-in or clamp fastening, which allows the sensor device to be mounted on the DIN rail without additional tools and removed again if necessary. The DIN rail, also called a DIN rail, is in particular a standardized metal profile rail that is used in control cabinets and distribution boxes to securely and space-savingly mount electrical and electronic components such as relays, control units or sensor devices. The DIN rail is frequently used in automation technology,used in plant engineering or electrical distribution systems. A control cabinet is generally a small housing that accommodates and protects electrical and electronic components, such as controls, fuses, or sensors, for example in production plants, waterworks, or industrial pump systems. By mounting the sensor device on the DIN rail in the control cabinet, it can be integrated directly into existing control systems, even retroactively, thus ensuring a robust and professional installation, for example for monitoring centrifugal pumps in a water treatment plant or in an industrial cooling circuit. In this respect, further development facilitates a compact and standards-compliant installation of the sensor device in industrial environments. The problem of the invention is further solved by a centrifugal pump comprising a sensor device as described above,where at least one signal input is connected to 10 of the centrifugal pump. A centrifugal pump is generally defined as a fluid pump that uses rotary motion and dynamic forces to pump predominantly liquids. In addition to tangential acceleration of the liquid, centrifugal force occurring in a radial flow is used for pumping, so that such pumps are also referred to as centrifugal pumps. The centrifugal pump is preferably used for a hydraulic system of a building or other applications, especially as a sewage or wastewater pump. In the regular operation of the centrifugal pump, a motor housing for the centrifugal pump 20 can be arranged above a pump housing, in which the impeller, driven by the motor via a shaft, for pumping the liquid and the centrifugal chamber are provided. The motor housing can be permanently connected to the pump housing and / or be designed as a single unit. Likewise, the centrifugal pump and the motor can each have their own shaft.where the shafts can be connected to each other via a coupling. Preferably, the shaft extends from the motor housing into the pump housing on one drive side and / or is fixedly connected to the shaft on the drive side of the impeller. Accordingly, the suction opening for the fluid to be pumped is preferably located at the bottom of the pump housing. Preferably, the motor housing and / or the pump housing is made of metal, in particular cast iron or stainless steel, ceramic, and / or plastic. The fluid preferably comprises water or another liquid medium such as wastewater. The fluid may contain solids such as dirt, impurities, or Debris of any kind, in particular feces, sediment, dirt, sand, or even smaller pieces of wood, brush, textiles, or rags, or the like, which can cause a blockage in the pump housing. 5 The problem of the invention is furthermore solved by a method for monitoring a centrifugal pump by means of a sensor device,with the steps: Capture at least one monitoring signal characterizing the state of the centrifugal pump at at least one signal input of the sensor device at least every second, where the signal input includes a current transformer input, an analog input and / or a digital input; Analyzing and / or data-reducing preprocessing of at least one monitoring signal, in particular with regard to an alarm state of the centrifugal pump, a statistical evaluation of the state of the centrifugal pump, and / or a state monitoring and prediction of the centrifugal pump; and 15 periodically transmitting at least one preprocessed monitoring signal not more often than every 5 minutes via a Low Power Wide Area Network (LPWAN) radio interface of the sensor device to a cloud infrastructure set up for post-processing of at least one transmitted, preprocessed monitoring signal. 20 According to a preferred further education, the method includes the step: Post-processing of at least one transmitted,pre-processed monitoring signals via the cloud infrastructure. 25 Such further training enables more in-depth analyses, long-term data evaluations and / or intelligent forecasts for centrifugal pump monitoring. Post-processing can, for example, include merging and comparing transmitted, pre-processed monitoring signals over extended periods to detect wear trends or anomalies that might indicate an impending failure of the centrifugal pump. In a water treatment plant, for example, the cloud could determine from the transmitted, pre-processed monitoring signals that the current draw of the centrifugal pump is slowly increasing, indicating increasing bearing friction, and generate an early maintenance recommendation. In an industrial cooling system, the cloud can, by comparing with historical data, detect that unusual vibrations indicate cavitation or a blocked pipe.which allows targeted measures to be initiated. Furthermore, the cloud can perform automatic pattern recognition using machine learning to predict future malfunctions and optimize preventive maintenance strategies. Another example is integration into a higher-level monitoring system that links various sensor devices in a production plant so that correlations between centrifugal pump failures and other pump conditions can be detected. This cloud-based post-processing significantly improves the efficiency and reliability of centrifugal pump monitoring.while simultaneously reducing the local computational and storage requirements of the sensor device. According to another preferred embodiment, the analyzing and / or data-reducing preprocessing includes data compression and / or selection, and at least one monitoring signal. According to a preferred further development, the analyzing and / or data-reducing preprocessing includes amplitude and / or time reduction, and at least one monitoring signal. Further embodiments and advantages of the method arise by analogy with the sensor device described above. It is preferred that the sequence of process steps, unless technically required in an explicit sequence,The aforementioned sequence of process steps is particularly preferred. Brief description of the drawings 30 The invention is explained in more detail below with reference to the attached drawings and preferred embodiments. 2025 / 5154 BE2025 / 5154 16 In the drawings, Fig. 1 shows a schematic view of a centrifugal pump for carrying out a proposed method according to a preferred embodiment of the invention, Fig. 2 shows a schematic arrangement comprising the centrifugal pump according to Fig. 1 and a sensor device for monitoring the centrifugal pump according to the preferred embodiment of the invention, Fig. 3 shows a flowchart for carrying out the proposed method according to the preferred embodiment of the invention, Fig. 4 shows a diagram for acquiring and evaluating monitoring signals for current measurement of the centrifugal pump according to the preferred embodiment of the invention.Fig. 5 a diagram of a trend analysis for monitoring the power consumption of the centrifugal pump according to the preferred embodiment of the invention, 20 Fig. 6 a diagram for the proper verification of threshold values ​​at signal inputs of the sensor device according to the preferred embodiment of the invention, 25 Fig. 7 another diagram for the proper verification of threshold values ​​at signal inputs of the sensor device according to the preferred embodiment of the invention,and Fig. 8 a diagram of the course of additional information to a cloud infrastructure downstream of the sensor device-30 according to the preferred embodiment of the invention. 2025 / 5154 BE2025 / 5154 17 Detailed description of the embodiments Fig. 1 shows a schematic view of a centrifugal pump 1 for carrying out a method for monitoring centrifugal pump 1 as described below according to a preferred embodiment of the invention. The centrifugal pump 1, designed as a wastewater pump, conventionally features a pump housing 2 with a suction opening 3 located at the bottom of the pump housing 2 as an inlet. A rotor 4 is provided in the pump housing 2, which extends vertically in the drawing. The rotor 4 is part of an electric motor 5, shown only in outline, which is arranged opposite the suction opening 3. An impeller 6 is provided facing the suction opening 3, which is driven by the rotor 4 of the electric motor 5. Furthermore, the centrifugal pump 1 features a control system.not shown in Fig. 115. Fig. 2 shows a schematic arrangement comprising the centrifugal pump 1 according to Fig. 1 and a sensor device 7 for monitoring the centrifugal pump 1 according to the preferred embodiment of the invention. The sensor device 7 has a separate housing from the centrifugal pump 1, which can be attached to a DIN rail of a control cabinet (not shown) by means of a fixing device 8. The sensor device 7 can be battery-operated, but in this case, it is supplied with electrical energy by the control cabinet. The sensor device 7 has multiple signal inputs 9, which include several current transformer inputs, an analog input, and a digital input. In addition, an analog output, at least one digital output, an RS485 interface, and a USB-C interface are provided. One of the current transformer inputs is connected to a current transformer 10.The sensor device 7 detects a monitoring signal flowing in the mains connection cable 30 11 of the centrifugal pump 1, which characterizes a state of the centrifugal pump 1. The sensor device 7 also includes a processor-based data processing unit 12 designed as a microcontroller 2025 / 5154 BE2025 / 5154 18, only indirectly visible in Fig. 2, as well as a Low Power Wide Area Network (LPWAN) radio interface 13. The data processing unit 12 is configured to carry out the process steps shown in Fig. 3 for executing the proposed method according to the preferred embodiment of the invention. 5 In a first step 100, the detection of at least one monitoring signal characterizing a state of the centrifugal pump 1 is carried out by means of at least one signal- The input 9 of the sensor device 7 is processed at least every second by the data processing unit 12. At least one monitoring signal can be taken as described by the current absorbed by the electric motor 5 10.a rotational speed of centrifugal pump 1 etc. The acquisition is carried out continuously and periodically and preferably at least every 500, 250, 100, 50, 25, 5 or 1 ms. In a second step 200, an analytical and data-reducing preprocessing of at least one monitoring signal with regard to an alarm state of centrifugal pump 1, a statistical evaluation of the state of centrifugal pump 1 and / or a state monitoring and prediction of centrifugal pump 1 is carried out by means of the data processing device 12. This analytical and data-reducing preprocessing includes data compression and selection of at least one monitoring signal. An alarm condition of the centrifugal pump 1 occurs, for example, when the sensor device 7 detects critical operational deviations by means of the analyzing preprocessing, which indicate a malfunction, overload, or imminent failure of the centrifugal pump 1. 25 For example, a sudden increase in current consumption would indicate a blocked rotor shaft 5 or a damaged bearing,Unusual vibrations may indicate cavitation or imbalance. If the temperature of the electric motor 5 rises above a defined limit, overheating of the electric motor 5 can be detected, which can trigger an alarm and generate a maintenance message 30. The alarm status can be sent via a downstream cloud infrastructure 14 to a higher-level monitoring system to enable an immediate response. The statistical evaluation of the centrifugal pump's condition refers, for example, to a long-term analysis of the recorded monitoring signals to identify averages, trends, and deviations. For example, the sensor device can record and analyze the average current consumption, rotational speed, or temperature over several weeks to detect atypical changes in operation at an early stage. A typical application is determining an average oscillation amplitude over several operating cycles.to detect gradual changes due to mechanical wear. Similarly, a trend analysis of energy consumption can help identify inefficiencies, for example, due to deposits in pipelines that place a greater load on the centrifugal pump 1. The statistical evaluation can serve as a basis for optimized maintenance strategies and help to avoid unplanned downtime.15 Condition monitoring is carried out, for example, by a continuous analysis of the monitoring signals of the centrifugal pump 1 to ensure that the monitoring signals are operating within normal operating ranges. Various measured values ​​such as current consumption, speed, temperature, or vibrations can be monitored in real time and compared with this statistical data. Condition prediction, predictive maintenance,This largely allows, for example, machine learning methods or predefined models to predict future faults or maintenance needs based on the recorded monitoring signals. For example, an increasing current draw under constant load can predict that the centrifugal pump 1 will need maintenance in a few weeks, as bearing damage is becoming apparent. Similarly, regular analysis of frequency spectra of vibrations can provide early warning of developing imbalance before it leads to critical damage. The combination of condition monitoring and prediction enables efficient maintenance planning, reduces planned downtime, and extends the service life of the centrifugal pump 1.30 2025 / 5154 BE2025 / 5154 20 In a third step, 300, a periodic transmission of at least one pre-processed monitoring signal took place not more often than every 5 minutes via the Low Power Wide Area Network (LPWAN) radio interface 13 of the sensor device 7 to the cloud infrastructure 14.which is configured to reprocess at least one transmitted, pre-processed monitoring signal. The transmission occurs continuously and periodically, and no more often than every 10, 15, 20, 30, or 60 minutes. The LPWAN radio interface is configured to transmit at least one pre-processed monitoring signal using its Long Range Wide Area Network (LoRaWAN) network protocol, its Long Term Evolution for Machines (LTE-M) network protocol, and / or its Narrow Band IoT network protocol. Due to the pre-processing, the data transmission rate of at least one monitoring signal is ≥300, 500, 1000, or 2000 higher than that of the transmitted, pre-processed monitoring signal. The transmission interval of no more frequently than every 5 minutes means, for example, that data aggregation is implemented in the sensor device itself. If a change of state, for example centrifugal pump 1 on / off, is detected several times within a 5-minute transmission interval,The state changes are counted on one side and on the other, up to a certain number with their own timestamp, for example a maximum of 15 state changes in 5 minutes, aggregated in one message and transmitted to the sensor in direction 7 and the cloud infrastructure 14, so that the information is not available live, but no information is lost that is shorter than the transmission interval. Finally, post-processing of at least one transmitted, pre-processed monitoring signal can be carried out using the cloud infrastructure, for example at a control station. Post-processing can, for example, include merging and comparing transmitted, pre-processed monitoring signals over longer periods to detect wear trends or anomalies that indicate an impending failure of the centrifugal pump. It is also conceivable that pre-processing can be configured from the cloud infrastructure.to adjust, for example, threshold values ​​for changes or events for each centrifugal pump 1 individually. In addition, further processing algorithms, for example filter algorithms, could be downloaded from the cloud infrastructure 14 to the sensor device 7 or the data processing unit 12 and applied specifically to individual monitoring signals. Fig. 4 shows a diagram for acquiring and evaluating monitoring signals for current measurement of the centrifugal pump 1 according to the preferred embodiment of the invention. The monitoring signals are present in the form of currents at the signal inputs 10 and 12 to detect operating states of the centrifugal pump 1. The data processing unit 12 provides information by means of the analytical preprocessing when the measured current exceeds a configured threshold value, label P1 in Fig. 4, or falls below it, label P2 in Fig. 4.each with a timestamp. The average current between P1 and P2, as well as during ongoing operation until the next update, is then displayed. The recorded monitoring signals can be persistently stored in the sensor device 7 to survive power outages. In another variant, standard deviations for an ON phase of the centrifugal pump 1, as well as the magnitude and duration of a switch-on pulse, are also recorded and evaluated. Figure 5 shows a diagram of a trend analysis for monitoring the power consumption of the centrifugal pump 1 according to the preferred embodiment of the invention. The analysis makes it possible to identify potential problems with the centrifugal pump 1 at an early stage. The sensor device 7 or the data processing device 12 provides a counter for the number of switch-on events of the centrifugal pump 1 in the last 24 hours, and finally, a possible ongoing cycle at the time of evaluation. The counter- ler is updated and persistently saved with each update cycle,to survive a power outage. Alternatively, an alarm based on a simple threshold comparison for an average value, for example 1.5 times the nominal current, can be generated on a dashboard provided by the cloud infrastructure. In a later step, an additional component of the cloud infrastructure can perform a trend analysis of the reported average values ​​to predict problems before an absolute threshold is exceeded. Fig. 6 shows a diagram for the proper verification of threshold values ​​at signal inputs of sensor device 7 for adjustable threshold values ​​for an ON / OFF indicator 5 according to the preferred embodiment of the invention. If the measured monitoring signal is above an ON threshold, an ON event with a timestamp is sent to the cloud infrastructure 14. If the measured monitoring signal then falls below an OFF threshold, an OFF event with a timestamp is sent to the cloud infrastructure 14. Additionally, a time is transmitted,The time the signal input 9 is above the threshold 10, the so-called ON time, is sent to the cloud infrastructure 14 in two versions. As long as the analog input is above the ON threshold, the time since the threshold was exceeded is sent every 15 minutes. When the analog input falls below the OFF threshold, the time between the ON and OFF thresholds is sent. Figure 7 shows a further diagram for checking thresholds at the signal inputs 9 of the sensor device 7 for adjustable thresholds for ON / OFF detection according to the preferred embodiment of the invention, wherein alarm thresholds can be configured so that when the alarm threshold 20 is exceeded, an alarm message with a timestamp is sent to the cloud infrastructure 14. detwird. If the alarm threshold is undershot, corresponding information with a timestamp is sent to the cloud infrastructure 14 indicating that the alarm is no longer active. 25,