WO2021104637A1

WO2021104637A1 - Method for determining the working point of a pump

Info

Publication number: WO2021104637A1
Application number: PCT/EP2019/082998
Authority: WO
Inventors: Jean-Paul JOHN; Stefan NIEDERBERGER; Thomas Folsche
Original assignee: Cp Pumpen Ag
Priority date: 2019-11-28
Filing date: 2019-11-28
Publication date: 2021-06-03

Abstract

A method for determining the working point of a pump (1) during its operation comprising the steps of: - measuring the amplitudes (A) of the pump's vibrations with a vibration detector (6) that is mounted to the outside of a housing (2) of a pump (1) and transmitting the measured amplitudes (A) to an evaluation unit (7); - selecting the amplitudes (As) of predefined frequency sections (Afs); - providing at least one mathematical model (MQ) for determining the delivery volume (Q) of the pump; - providing at least one mathematical model (MH) for determining the pumping head (H) of the pump; - selecting one delivery volume model (MQ) and one pumping head model (MH) based on the measured amplitudes (A); and - determining the delivery volume (Q) and the pumping head (H) of the pump by inserting the selected amplitudes (As) in the selected mathematical models (MQ;MH).

Description

METHOD FOR DETERMINING THE WORKING POINT OF A PUMP

TECHNICAL FIELD OF THE INVENTION

The current invention relates to a method for determining the working point of a pump during operation in the entire operational window of the pump, in particular, the rotation speed, the pumping head and the delivery volume of a centrifugal pump, such as a radial flow pump with an impeller .

DESCRIPTION OF THE RELATED ART

Known embodiments of such a method can only determine the optimal working point of a pump with the help of a vibration detector. In the optimal working point, the pump produces the least vibrations. The more the pump is operated in a working point that is away from the optimal working point, i.e. in part load or overload, the more the vibrations increase. If the pumping head is to be determined, a pressure sensor close to the pump is required and if the delivery volume is to be determined, a flow meter is required. Both these sensors are in contact with the medium to be pumped. Depending on the medium to be pumped, the installation of these sensors is complex and therefore expensive. Additionally, due to the contact with the medium to be pumped, the sensors tend to wear. SUMMARY OF THE INVENTION

In the current invention, an objective is to provide a method for monitoring the operating point of a pump during its operation. This problem is solved by a method with the features of claim 1. Further embodiments of the method, as well as an evaluation system for determining the working point of a pump during its operation are defined by the features of further claims. A method for determining the working point of a pump during its operation according to the invention comprises the steps of: measuring the amplitudes of the pump's vibrations in the time domain during a predefined time period with a vibration detector that is mounted to the outside of a housing of a pump and that is connected to an evaluation unit; transmitting the measured amplitudes to the evaluation unit; selecting the amplitudes of predefined frequency sections; providing at least one mathematical model for determining the delivery volume of the pump; providing at least one mathematical model for determining the pumping head (H) of the pump; selecting one delivery volume model and one pumping head model based on the measured amplitudes; and determining the delivery volume and the pumping head of the pump by inserting selected amplitudes in the selected mathematical models.

Such a method has the advantage that only an evaluation unit together with a vibration detector, mounted to the outside of the pump's housing is required to be able to determine the working point of the pump during its operation. There is no flow meter obstructing the flow within the system and there are no pressure sensors before and after the pump reaching into the flow path of the medium to be pumped. The vibration detector can be, for example, an accelerometer or any kind of vibration detector with which the vibration's amplitudes can be recorded in the time domain. Instead of having predefined frequency sections, there can be predefined discrete frequencies at which the amplitudes of the vibrations are selected.

In one embodiment, the method comprises the steps of: determining the rotation speed of the pump by using the measured amplitudes; - creating a data vector with the rotation speed and the measured amplitudes; transforming the measured amplitudes from the time domain into the frequency domain.

In one embodiment, the selection of the one delivery volume model and the one pumping head model is done based on the root mean square of the measured amplitudes in the frequency domain, the measured amplitude at least at one predefined dominant frequency section and the determined rotation speed.

In one embodiment, the delivery volume and the pumping head are determined by additionally inserting the rotation speed in the selected mathematical models.

In one embodiment, the pump is a centrifugal pump with an impeller arranged in the housing and wherein the rotation speed of the pump corresponds to the rotation speed of the impeller. The centrifugal pump can be driven by an electrical motor, such as for example an asynchronous motor, which can be directly connected to the power grid or by means of frequency converter. The motor can be directly connected to the pump, i.e. to the impeller of the pump or by means of a magnetic coupling.

In one embodiment, the measured amplitudes from the time domain are transformed into the frequency domain by means of a fast Fourier transformation.

In one embodiment, the predefined frequency sections or the predefined discrete frequencies comprise the passing frequency of the inlet edges of the impeller and the passing frequency of the outlet edges of the impeller. The passing frequency of the inlet edges is created by the periodical passing of the inlet edges close to the structure that holds the impeller in place at the inlet of the pump. This structure can comprise one or more baffles or bars. If there are several baffles or bars, they can be arranged evenly distributed around the circumference of the pump's inlet. They extend from the centre towards the periphery. With an increasing number of structures or impeller blades, the passing frequency of the inlet edges increase as well. Accordingly, the passing frequency of the outlet edges is created by the periodical passing of the outlet edges close to the at least one cut-off at the outlet of the pump. As there is at least one cut-off and at least one bar or baffle, the passing frequency of the inlet edges is either equal to, a multiple of or a fraction of the passing frequency of the outlet edges.

In one embodiment, the at least one mathematical model for determining the delivery volume of the pump comprises at least one part load model and at least one overload model and the at least one mathematical model for determining the pumping head of the pump comprises at least one part load model and at least one overload model. With the split of the models in the best point of operation of the pump, the accuracy of the determination of the delivery volume and the pumping head can be increased. It would also be possible to split the entire range of the pump's performance curve in three or more sections to further increase the accuracy of the determination of the working point of a pump during its operation.

In one embodiment, the one delivery volume model and the one pumping head model are selected additionally based on the root mean square of the measured amplitudes in the frequency domain.

In one embodiment, the delivery volume model and the pumping head model comprise terms that comprise at least one of: the rotation speed, the root mean square of the measured amplitudes in the frequency domain, the variance of the measured amplitudes in the frequency domain, the measured amplitudes at the passing frequency of the inlet edges of the impeller, and the measured amplitudes at the passing frequency of the outlet edges of the impeller.

In one embodiment, the creation of the at least one mathematical model for determining the pumping head of the pump comprising the steps of: providing a pump with a housing, a vibration detector and an evaluation unit in a test plant with a flow meter, a first pressure sensor arranged at the pump's inlet and a second pressure sensor arranged at the pump's outlet, wherein the vibration detector is mounted to the outside of the housing and wherein the flow meter, the pressure sensors and the vibration detector are connected to the evaluation unit; measuring the amplitudes of the vibrations in the time domain with the vibration detector at several points on the pump's performance curve during a predefined time period; determining the delivery volume with the flow meter and determining the pumping head with the two pressure sensors at the same points on the pump's performance curve and during the same time period; transforming the measured amplitudes from the time domain into the frequency domain by means of a fast Fourier transformation; dividing the complete frequency domain of each data set in discrete sections; selecting frequency sections with dominant amplitudes; determining the rotation speed of the pump at each measurement point by using the measured amplitudes; creating a first data set with the rotation speed (n), the measured amplitudes in the time domain, the variance of the measured amplitudes, the root mean square of the measured amplitude, the dominant amplitudes in the frequency domain and the delivery volume at each measurement point; - creating a second data set with the rotation speed, the measured amplitudes in the time domain, the variance of the measured amplitudes, the root mean square of the measured amplitudes, the dominant amplitudes in the frequency domain and the pumping head at each measurement point; defining the delivery volume and the pumping head of the best point of operation of the pump; creating a part load model for the delivery volume by using the part of the first data set where the value of the delivery volume equals or is below the value of the delivery volume of the best point of operation of the pump and a regression method, wherein the rotation speed, the variance of the measured amplitudes, the root mean square of the measured amplitudes and the dominant amplitudes are input variables and the corresponding delivery volume is the output variable; creating an overload model for the delivery volume by using the part of the first data set where the value of the delivery volume equals or is above the value of the delivery volume of the best point of operation of the pump and a regression method, wherein the rotation speed, the variance of the measured amplitudes, the root mean square of the measured amplitudes and the dominant amplitudes are input variables and the corresponding delivery volume is the output variable; creating a part load model for the pumping head by using the part of the second data set where the value of the pumping head equals or is above the value of the pumping head of the best point of operation of the pump and a regression method, wherein the rotation speed, the variance of the measured amplitudes, the root mean square of the measured amplitudes and the prominent dominant amplitudes are input variables and the corresponding pumping head is the output variable; creating an overload model for the pumping head by using the part of the second data set where the value of the pumping head equals or is below the value of the pumping head of the best point of operation of the pump and a regression method, wherein the rotation speed, the variance of the measured amplitudes, the root mean square of the measured amplitudes and the prominent dominant amplitudes are input variables and the corresponding pumping head is the output variable; and eliminating irrelevant variables in the models through backwards selection. In one embodiment, the models are being created by using a Ridge regression method or by using an ordinary least square regression method.

In one embodiment, the pump is a centrifugal pump with an impeller arranged in the housing and wherein the rotation speed of the pump corresponds to the rotation speed of the impeller .

In one embodiment, the selected frequency sections comprise the passing frequency of the inlet edges of of the impeller and the passing frequency of the outlet edges of the impeller.

The features of the above-mentioned embodiments of the method for determining the working point of a pump during its operation can be used in any combination, unless they contradict each other. An evaluation system for determining the working point of a pump during its operation according to the invention comprises a vibration detector adapted to be mounted to the outside of a housing of the pump and an evaluation unit enabling the performance of the method of one of the above embodiments.

In one embodiment, the transmittal of the measured amplitudes to the evaluation unit is realized by cable or wireless . BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the current invention are described in more detail in the following with reference to the figures.

These are for illustrative purposes only and are not to be construed as limiting. It shows

Fig. 1 a perspective view of a radial flow pump with an impeller and an accelerometer mounted to the pump housing; and

Fig. 2 a schematic cross section through the pump of figure 1.

DETAILED DESCRIPTION OF THE INVENTION

Figure 1 shows a perspective view of a radial flow pump 1 with an impeller 5 and an accelerometer 6 mounted to the pump housing 2 and figure 2 shows a schematic cross section through the pump 1 of figure 1. The radial flow pump 1 comprises a central inlet 3, a radial outlet 4 and an impeller 5 that is arranged in the housing 2. At the inlet 3, radially extending bars 30 are holding the inlet side of the impeller 5 in place. In the depicted embodiment, there are three bars 30 that are evenly distributed around the circumference of the inlet opening, respectively the inlet channel. The accelerometer 6 is arranged in the vicinity of the cut-off 20 of the pump 1 and is connected to the evaluation unit 7. The impeller 5 comprises several blades 50 that are arranged around the axis of rotation, extending from an inner region towards an outer region. Each blade 50 comprises an inlet edge 51 on the inlet side and an outlet edge 52 on the outer side. During the operation of the pump, vibrations are generated resulting from pressure fluctuations due to the inlet edges 51 passing the bars 30 periodically and due to the outlet edges 52 passing the cut-off 20 periodically. With three bars 30 at the inlet and one cut-off 20 at the outlet, the passing frequency of the inlet edges 51 is three times the passing frequency of the outlet edges 52.

The vibrations measured with the accelerometer 6 are transmitted to the evaluation unit 7 where these measurements are used for further processes.

The rotation speed is identified by a combination of cross correlations. Firstly, the measured amplitudes in the time domain A(t) are recorded. Secondly, a rough speed range is determined from the recorded data using the combination of cross-correlations. Thirdly, a further combination of cross-correlations is performed to identify the speed range more precisely. Fourthly, the correlation values from the third step are interpolated using cubic spline interpolation. Finally, the maximum is selected. The index of this correlation corresponds to the speed.

To decide whether the pump is operating in part load or overload, the root mean square of the measured amplitudes in the frequency domain RMS(A(f)), the amplitude at the passing frequency of the outlet edges A(fba) and the current rotation speed n of the pump are used. In part load operation, the pumping head H is higher than the one in the best point of operation of the pump and the delivery volume Q is lower than the one in the best point of operation of the pump. Consequently, in overload operation, the pumping head is lower than in the best point of operation and the delivery volume is higher than in the best point of operation. In part load operation, the RMS(A(f)) increase significantly with lower deliver volume values. In overload operation, the A(fba) increase significantly with higher delivery volume values. The rotation speed decreases with increasing delivery volume values.

For determining the operating point of the pump, the root mean square of the measured amplitudes in the frequency domain RMS(A(f)), the variance of the measured amplitudes in the frequency domain VAR(A(f)), the amplitude at the passing frequency of the inlet edges A(fbi), the amplitude at the passing frequency of the outlet edges A(fba) and the current rotation speed n of the pump are used.

REFERENCE SIGNS LIST

1 pump 5 impeller

2 housing 50 blade

20 cut-off 51 inlet edge

3 inlet 52 outlet edge

30 bar 6 accelerometer

4 outlet 7 evaluation unit

Claims

1. A method for determining the working point of a pump during its operation comprising the steps of: measuring the amplitudes (A[t]) of the pump's vibrations in the time domain during a predefined time period (At) with a vibration detector (6) that is mounted to the outside of a housing (2) of a pump (1) and that is connected to an evaluation unit (7); transmitting the measured amplitudes (A [t]) to the evaluation unit (7); selecting the amplitudes (As[f]) of predefined frequency sections (Afs); providing at least one mathematical model (MQ) for determining the delivery volume (Q) of the pump; providing at least one mathematical model (MH) for determining the pumping head (H) of the pump; selecting one delivery volume model (MQ) and one pumping head model (MH) based on the measured amplitudes (A[t]); and determining the delivery volume (Q) and the pumping head (H) of the pump by inserting selected amplitudes (As[f]) in the selected mathematical models (MQ;MH).

2. The method of claim 1, comprising the steps of: determining the rotation speed (n) of the pump by using the measured amplitudes (A[t]); creating a data vector with the rotation speed (n) and the measured amplitudes (A[t]); and transforming the measured amplitudes from the time domain (A [t]) into the frequency domain (A [f]).

3. The method of claim 1 or 2, wherein the selection of the one delivery volume model (MQ) and the one pumping head model (MH) is done based on the root mean square of the measured amplitudes in the frequency domain RMS(A(f)), the measured amplitude at least at one predefined dominant frequency section (Afs) and the determined rotation speed (n).

4. The method of claim 2 or 3, wherein the delivery volume (Q) and the pumping head (H) are determined by additionally inserting the rotation speed (n) in the selected mathematical models (MQ;MH).

5. The method of one of claims 1 to 4, wherein the pump is a centrifugal pump with an impeller (5) arranged in the housing (2) and wherein the rotation speed of the pump (1) corresponds to the rotation speed (n) of the impeller (5).

6. The method of one of claims 1 to 5, wherein the measured amplitudes from the time domain (A [t]) are transformed into the frequency domain (A [f]) by means of a fast Fourier transformation.

7. The method of one of claims 1 to 6, wherein the predefined frequency sections (Afs) comprise the passing frequency (fbi) of the inlet edges (51) of the impeller (5) and the passing frequency (fbo) of the outlet edges (52) of the impeller.

8. The method of claim 7, wherein the passing frequency (fbi) of the inlet edges (51) is a multiple of the passing frequency (fbo) of the outlet edges (52) or is equal to the passing frequency (fbo) of the outlet edges (52) or is a fraction of the passing frequency (fbo) of the outlet edges (52).

9. The method of one of claims 1 to 8, wherein the at least one mathematical model for determining the delivery volume (Q) of the pump comprises at least one part load model (MQP) and at least one overload model (MQO) and wherein the at least one mathematical model for determining the pumping head (H) of the pump comprises at least one part load model (MHP) and at least one overload model (MHO).

10. The method of one of claims 1 to 9, wherein the one delivery volume model (MQ) and the one pumping head model (MH) are selected additionally based on the root mean square of the measured amplitudes in the frequency domain (RMS A [f]).

11. The method of one of claims 1 to 10, wherein the delivery volume model (MQ) and the pumping head model (MH) comprise terms that comprise at least one of: the rotation speed (n), - the root mean square of the measured amplitudes in the frequency domain (RMS(A[f])), the variance of the measured amplitudes in the frequency domain (VAR(A[f])), the measured amplitudes at the passing frequency of the inlet edges of the impeller (fbi), and the measured amplitudes at the passing frequency of the outlet edges of the impeller (fbo).

12. The method of one of claims 1 to 11, wherein the creation of the at least one mathematical model (MH) for determining the pumping head (H) of the pump comprising the steps of: providing a pump (1) with a housing (2), a vibration detector (6) and an evaluation unit (7) in a test plant with a flow meter, a first pressure sensor arranged at the pump's inlet and a second pressure sensor arranged at the pump's outlet, wherein the vibration detector (6) is mounted to the outside of the housing (2) and wherein the flow meter, the pressure sensors and the vibration detector (6) are connected to the evaluation unit (7); measuring the amplitudes of the vibrations (A [t]) in the time domain with the vibration detector (6) at several points on the pump's performance curve during a predefined time period (At); determining the delivery volume (Q) with the flow meter and determining the pumping head (H) with the two pressure sensors at the same points on the pump's performance curve and during the same time period (At); determining the rotation speed (n) of the pump (1) at each measurement point by using the measured amplitudes (A [t]); transforming the measured amplitudes from the time domain (A [t]) into the frequency domain (A [f]) by means of a fast Fourier transformation; dividing the complete frequency domain of each data set in discrete sections (Af); selecting frequency sections (Afs) with dominant amplitudes (As[f]); creating a first data set with the rotation speed (n), the measured amplitudes (A [t]) in the time domain, the variance of the measured amplitudes (VAR_A[t]), the root mean square of the measured amplitudes (RMS_A[t]), the dominant amplitudes (As[f]) in the frequency domain and the delivery volume (Q) at each measurement point; creating a second data set with the rotation speed (n), the measured amplitudes (A [t]) in the time domain, the variance of the measured amplitudes (VAR_A [t]), the root mean square of the measured amplitudes (RMS_A[t]), the dominant amplitudes (As[f]) in the frequency domain and the pumping head (H) at each measurement point; defining the delivery volume (QB) and the pumping head (HB) of the best point of operation of the pump; creating a part load model for the delivery volume (MQP) by using the part of the first data set where the value of the delivery volume (Q) equals or is below the value of the delivery volume (QB) of the best point of operation of the pump and a regression method, wherein the rotation speed (n), the variance of the measured amplitudes (VAR_A[t]), the root mean square of the measured amplitudes (RMS_A[t]) and the dominant amplitudes (As[f]) are input variables and the corresponding delivery volume (Q) is the output variable; creating an overload model for the delivery volume (MQO) by using the part of the first data set where the value of the delivery volume (Q) equals or is above the value of the delivery volume (QB) of the best point of operation of the pump and a regression method, wherein the rotation speed (n), the variance of the measured amplitudes (VAR_A[t]), the root mean square of the measured amplitudes (RMS_A[t]) and the dominant amplitudes (As[f]) are input variables and the corresponding delivery volume (Q) is the output variable; creating a part load model for the pumping head (MHP) by using the part of the second data set where the value of the pumping head (H) equals or is above the value of the pumping head (HB) of the best point of operation of the pump and a regression method, wherein the rotation speed (n), the variance of the measured amplitudes (VAR_A[t]), the root mean square of the measured amplitudes (RMS_A[t]) and the prominent dominant amplitudes (As[f]) are input variables and the corresponding pumping head (H) is the output variable; creating an overload model for the pumping head (MHO) by using the part of the second data set where the value of the pumping head (H) equals or is below the value of the pumping head (HB) of the best point of operation of the pump and a regression method, wherein the rotation speed (n), the variance of the measured amplitudes (VAR_A[t]), the root mean square of the measured amplitudes (RMS_A[t]) and the prominent dominant amplitudes (As[f]) are input variables and the corresponding pumping head (H) is the output variable; and eliminating irrelevant variables in the models through backwards selection.

13. The method of claim 12, wherein the models are being created by using a Ridge regression method or by using an ordinary least square regression method.

14. The method of claim 12 or 13, wherein the pump is a centrifugal pump with an impeller (5) arranged in the housing (2) and wherein the rotation speed of the pump (1) corresponds to the rotation speed (n) of the impeller (5).

15. The method of one of claims 12 to 14, wherein the selected frequency sections (Afs) comprise the passing frequency (fbi) of the inlet edges (51) of the impeller (5) and the passing frequency (fbo) of the outlet edges (52) of the impeller (5).

16. An evaluation system for determining the working point of a pump (1) during its operation, the system comprising a vibration detector (6) adapted to be mounted to the outside of a housing (2) of the pump (1) and an evaluation unit (7) enabling the performance of the method of one of claims 1 to 11.

17. The evaluation system of claim 16, wherein the transmittal of the measured amplitudes (A [t]) to the evaluation unit (7) is realized by cable or wireless.