Positive displacement pump

Abstract

To accurately determine an operation state, in a positive displacement pump.SOLUTION: A uniaxial eccentric screw pump 1 comprises a stator 21 for defining a pump chamber PR, a rotor 22 for transferring fluid accompanied by a cyclic motion in the pump chamber PR, and an operation state determination device 4 for determining an operation state. The operation state determination device 4 has a Doppler sensor 40 for transmitting an ultrasonic wave or a microwave toward the rotor 22, receiving a reflection wave thereof, and creating a Doppler signal, a determination image creation unit 41a for creating a determination image corresponding to the Doppler signal, and a determination unit 41b which is applied with learning processing by using teacher data so as to output the operation state when inputted with the determination image.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a positive displacement pump. [Background technology]

[0002] In a positive displacement pump, fluid is transferred by periodic motion (e.g., reciprocating motion or rotational motion) of a moving member (e.g., a piston or rotor) in a pump chamber. In such a positive displacement pump, it is desirable to detect abnormal operation from the viewpoint of preventing wear and damage. Examples of abnormal operation include empty operation, in which there is no fluid to be transferred, and air bubble mixed operation, in which air bubbles are mixed into the fluid being transferred.

[0003] Patent Document 1 discloses a positive displacement pump that transmits microwaves toward a moving member and receives the reflected waves to generate a Doppler signal, and determines the operating state based on the Doppler signal. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2016-079971 A Summary of the Invention [Problem to be solved by the invention]

[0005] In the positive displacement pump of Patent Document 1, the operating state is determined using the peak value, the number of peaks, or the integral value of the Doppler signal. However, this determination method does not fully utilize the information contained in the Doppler signal, and there is room for improvement in the determination accuracy.

[0006] An object of the present invention is to determine the operating state of a positive displacement pump with high accuracy. [Means for solving the problem]

[0007] The present invention relates to A containment member defining a pump chamber; a moving member that moves a fluid in accordance with a periodic motion in the pump chamber; A driving state determination device for determining a driving state; Equipped with The driving state determination device includes: a Doppler sensor that transmits ultrasonic waves or microwaves toward the moving member and receives the reflected waves to generate a Doppler signal; a determination image generating unit that generates a determination image corresponding to the Doppler signal; a judgment unit that has undergone learning processing using teacher data so as to output the driving state when the judgment image is input; The present invention provides a positive displacement pump having a

[0008] According to this configuration, the Doppler signal is converted into a judgment image and used as input data for the judgment unit, so the operating state can be judged from a comprehensive perspective compared to when the operating state is judged simply from fragmentary information such as the peak value, the number of peaks, or the integral value of the Doppler signal. Even if the judgment image is difficult to distinguish visually, abnormal operation can be judged easily and with high accuracy by using the trained judgment unit. Therefore, the operating state of the positive displacement pump can be judged with high accuracy.

[0009] The determination image generating unit is An autocorrelation function may be generated from the Doppler signal. The autocorrelation function may be extracted over a time range that is an arbitrary natural number multiple of the motion period of the motion member.

[0010] According to this configuration, instead of using the graph of the Doppler signal as it is as the judgment image, the judgment image based on the autocorrelation function is used, so that the characteristic parts of the signal waveform can be clearly identified. Here, the autocorrelation function is obtained by IQ detecting the Doppler signal and performing complex correlation function processing, and then finding the correlation coefficient.

[0011] The determination image generating unit is A radar chart may be generated in which the autocorrelation function is displayed in polar coordinates by synchronizing the time range with a phase angle of 0 to 360°.

[0012] According to this configuration, by making the autocorrelation function into a radar chart and using it as the judgment image, it is possible to improve the judgment accuracy compared to when the autocorrelation function itself is used as the judgment image. As for the improvement in judgment accuracy, an experiment was conducted and it was confirmed that the rate of erroneous judgments was reduced to one third before and after making it into a radar chart.

[0013] The judgment image generating unit may extract a portion of the radar chart within the following angle range An, where N is the arbitrary natural number. TIFF2025025525000002.tif1389

[0014] According to this configuration, the amount of input data and teacher data can be reduced compared to the case where the judgment image of the entire phase angle from 0 to 360° is input. This is because characteristic data appears periodically because the moving member moves periodically. The angle range An is set by conducting an experiment and confirming characteristic data that changes greatly depending on the operating state. Therefore, even if the amount of input data and teacher data is limited, the deterioration of the estimation accuracy can be suppressed to a certain extent.

[0015] The determination image generating unit is A power spectrum may be generated by performing a Fourier transform on the Doppler signal; A spectral envelope may be generated by connecting frequency points corresponding to integer multiples of the motion period of the motion member in the power spectrum, The spectral envelope may be extracted at a predetermined frequency range.

[0016] According to this configuration, instead of using the graph of the Doppler signal as it is as the determination image, a determination image based on the spectral envelope is used, making it possible to clearly capture the characteristic portions of the signal waveform.

[0017] The determination image generating unit is A radar chart may be generated in which the spectrum envelope is displayed in polar coordinates by synchronizing the predetermined frequency range with a phase angle of 0 to 360°.

[0018] According to this configuration, by making the spectrum envelope into a radar chart and using it as the judgment image, it is possible to improve the judgment accuracy compared to when the spectrum envelope itself is used as the judgment image. As for the improvement in judgment accuracy, an experiment was conducted and it was confirmed that the rate of erroneous judgments was reduced by three-quarters after making the radar chart. Effect of the Invention

[0019] According to the present invention, the operating state of a positive displacement pump can be determined with high accuracy. [Brief description of the drawings]

[0020] [Figure 1] FIG. 1 is a vertical cross-sectional view of a positive displacement pump according to one embodiment. [Diagram 2] 2 is a first cross-sectional view of a sensor mounting hole portion of the positive displacement pump of FIG. 1; [Diagram 3] FIG. 2 is a second cross-sectional view of the sensor mounting hole portion of the positive displacement pump of FIG. 1 . [Figure 4] FIG. 2 is a block diagram of a driving state determination device. [Diagram 5] Doppler signal data during normal operation. [Figure 6] Doppler signal data during idle operation. [Figure 7] Doppler signal data during operation with air bubbles present. [Figure 8] Doppler signal data when operation is stopped. [Figure 9] Graph of the autocorrelation function during normal operation. [Figure 10] Graph of the autocorrelation function during idle running. [Figure 11] Graph of the autocorrelation function during operation with air bubbles. [Figure 12] Graph of the autocorrelation function during shutdown. [Figure 13]Judgment image of the graph of the autocorrelation function during normal operation. [Figure 14] Judgment image of the graph of the autocorrelation function during idle running. [Figure 15] Judgment image of the graph of the autocorrelation function during operation with air bubbles. [Figure 16] Judgment image of the graph of the autocorrelation function during shutdown. [Figure 17] A radar chart judgment image of the autocorrelation function during normal operation. [Figure 18] A radar chart image of the autocorrelation function during idle operation. [Figure 19] A radar chart image of the autocorrelation function during operation with air bubbles. [Figure 20] Judgment image of the radar chart of the autocorrelation function during shutdown. [Figure 21] A graph of the power spectrum during normal operation. [Figure 22] A graph of the power spectrum during idle operation. [Diagram 23] Graph of the power spectrum during operation with air bubbles. [Figure 24] Graph of power spectrum during shutdown. [Diagram 25] Judgment image of the spectrum envelope during normal operation. [Figure 26] Judgment image of the spectral envelope during idle operation. [Figure 27] Judgment image of the spectral envelope during operation with air bubbles. [Figure 28] Judgment image of the spectral envelope during shutdown. [Figure 29] A radar chart image of the spectrum envelope during normal operation. [Diagram 30] A radar chart image of the spectral envelope during idle operation. [Diagram 31] A radar chart image of the spectral envelope during operation with air bubbles. [Diagram 32] Judgment image of the radar chart of the spectral envelope during shutdown. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings. The following description is merely illustrative in nature and is not intended to limit the present invention, its application, or its uses. The drawings are schematic, and the ratios of dimensions may differ from the actual ones.

[0022] (First embodiment) FIG. 1 is a vertical cross-sectional view that typically shows a uniaxial eccentric screw pump (positive displacement pump) 1 according to a first embodiment.

[0023] The uniaxial eccentric screw pump 1 has a function of transporting a fluid supplied inside. The type of fluid is not particularly limited, but is generally a liquid. The uniaxial eccentric screw pump 1 has a driving machine 2 such as a motor (schematically shown) provided at one end in the longitudinal direction, a pump body 3 provided at the other end, and an operating state determination device 4.

[0024] The pump body 3 has a connection casing 10 that serves as an exterior, and a stator casing 20. The connection casing 10 and the stator casing 20 are concentrically arranged and connected to each other by bolts and nuts.

[0025] The connection casing 10 is made of a metallic material in a cylindrical shape, and houses a coupling rod 11 for power transmission. A connection port 13 is provided on the side of the connection casing 10. A fluid supply source 5, which is shown in schematic form, is connected to the connection port 13. Fluid is supplied from the supply source 5 through the connection port 13 into the inside of the connection casing 10.

[0026] The stator casing 20 is made of a metal material or a synthetic resin material in a cylindrical shape, and houses a stator (housing member) 21 and a rotor (moving member) 22 therein.

[0027] An end stud 30, which serves as a fluid discharge port, is attached to the end of the stator casing 20. The stator casing 20 and the end stud 30 are arranged concentrically and connected to each other with bolts and nuts. These connected parts are sealed with packing or the like (not shown).

[0028] A sensor mounting hole 23 is provided on the side surface of the stator casing 20. A Doppler sensor 40, the details of which will be described later, is attached to the sensor mounting hole 23.

[0029] The coupling rod 11 extends approximately in the longitudinal direction of the uniaxial eccentric screw pump 1, with one end connected to the coupling mechanism 12 and the other end connected to the rotor 22. Power from the driving machine 2 is transmitted to the rotor 22 via the coupling mechanism 12 and the coupling rod 11.

[0030] The stator 21 defines a pump chamber PR, and fluid is transported within the pump chamber PR. The stator 21 is disposed along the inner circumferential surface of the stator casing 20, and extends over the entire length of the stator casing 20. The stator 21 is formed into a tubular (e.g., cylindrical) shape from an elastic material such as rubber or resin that is appropriately selected depending on the fluid to be transported. Specifically, the stator 21 may be made of silicone rubber or fluororubber. The inner circumferential surface of the stator 21 is formed into a k-thread single-stage or multi-stage female screw shape.

[0031] The rotor 22 is disposed within the stator 21 (i.e., within the pump chamber PR). The rotor 22 transports the fluid in accordance with its periodic motion within the pump chamber PR. The rotor 22 is a shaft made of a metal material such as stainless steel and has a k-1 thread shape with a single or multi-stage male screw. The rotor 22 receives power from the driver 2, rotates on its own axis inside the stator 21, and revolves around the inner circumferential surface of the stator 21. In other words, the rotor 22 eccentrically rotates within the stator 21, thereby transporting the fluid in the pump chamber PR toward the end stud 30 and discharging the fluid from the end stud 30.

[0032] The flow of fluid in the uniaxial eccentric screw pump 1 configured as described above is as follows. Fluid is supplied from the supply source 5 into the connecting casing 10. The fluid in the connecting casing 10 is sucked into the pump chamber PR as the rotor 22 rotates eccentrically (forward rotation), and is transported toward the end stud 30 and discharged from the end stud 30. When the rotor 22 is rotated in the reverse direction, the fluid is sucked in through the end stud 30 and discharged from the connection port 13.

[0033] 2 and 3 are first and second cross-sectional views of the sensor mounting hole portion of the positive displacement pump in FIG. 1, and show the motion of the rotor 22 in the cross section of the stator 21. FIG. 2 shows the rotor 22 when the rotation angle is 0°, and FIG. 3 shows the rotor 22 when the rotation angle is 180°. These rotation angles are set for the convenience of explanation. The figure indicated by the symbol U in FIGS. 2 and 3 shows the ultrasonic wave of the Doppler sensor 40.

[0034] 2 and 3, the cross-sectional shape of rotor 22 is circular. The cross-sectional shape of the inner peripheral surface of stator 21 is elliptical, and the cross-sectional shape of the outer peripheral surface is circular. The cross-sectional shape of the inner peripheral surface of stator 21 has opposing linear portions 21a, 21b and semicircular arc portions 21c, 21d connecting the end points of linear portions 21a, 21b.

[0035] 2, when the rotation angle of rotor 22 is 0°, rotor 22 is in contact with, of the two arc-shaped portions 21c, 21d, arc-shaped portion 21d that is farther from Doppler sensor 40. When rotor 22 rotates eccentrically from this state, in the transverse cross section of stator 21, rotor 22 moves along linear portions 21a, 21b toward arc-shaped portion 21c that is closer to Doppler sensor 40.

[0036] 3, when the rotation angle of rotor 22 reaches 180°, rotor 22 comes into contact with arc-shaped portion 21c. When rotor 22 further rotates eccentrically from this state, the moving direction of rotor 22 is reversed, and rotor 22 moves along linear portions 21a and 21b toward arc-shaped portion 21d on the side farther from Doppler sensor 40.

[0037] 2, when the rotation angle reaches 360°, rotor 22 again comes into contact with arc-shaped portion 21d on the side farther from Doppler sensor 40. In this manner, rotor 22 reciprocates within pump chamber PR so as to approach and move away from Doppler sensor 40 in the cross section of stator 21. Note that while rotor 22 reciprocates in the up-down direction in FIGS. 2 and 3, the direction of reciprocating motion of rotor 22 changes in the cross section of stator 21 at another position in the longitudinal direction.

[0038] FIG. 4 shows a block diagram of the driving state determination device 4.

[0039] The operating state determination device 4 determines the operating state of the uniaxial eccentric screw pump 1. In this embodiment, the operating state determination device 4 determines whether the uniaxial eccentric screw pump 1 is in a normal operating state, an empty operating state, an operation with air bubbles present, or a stopped state. This determination may be made in real time or at regular time intervals.

[0040] In this embodiment, the driving state determination device 4 includes a Doppler sensor 40, a control device 41, and a display unit .

[0041] The Doppler sensor 40 transmits ultrasonic waves or microwaves toward the rotor 22 and receives the reflected waves, generating a Doppler signal according to the Doppler frequency. Here, the Doppler frequency is the difference between the frequency of the transmitted ultrasonic waves and the frequency of the received ultrasonic waves. More specifically, the Doppler sensor 40 outputs a sine wave Doppler signal according to the Doppler frequency, and the frequency of the Doppler signal changes according to the Doppler frequency. In addition, the intensity (amplitude) of the Doppler signal changes in principle according to the energy of the reflected wave. However, when the Doppler frequency is 0, the intensity (amplitude) of the Doppler signal is 0 regardless of the energy of the reflected wave.

[0042] In this embodiment, an example will be described in which the Doppler sensor 40 is of an ultrasonic type. Even if the Doppler sensor 40 is of a microwave type, the processing by the control device 41 described below is substantially the same, except that the waveform (attenuation rate and attenuation tendency) of the Doppler signal changes. For example, the attenuation rate of ultrasonic waves is smaller in water than in air, whereas the attenuation rate of microwaves is larger in water than in air. Therefore, substantially the same results can be obtained by processing that takes this into account.

[0043] During normal operation, the Doppler sensor 40 transmits and receives ultrasonic waves with a fluid interposed between the Doppler sensor 40 and the rotor 22. That is, during normal operation, the ultrasonic waves pass through the fluid in the pump chamber PR on their way from the Doppler sensor 40 to the rotor 22.

[0044] Ultrasonic waves have the property of being reflected by metal members, and some of the waves are attenuated as they pass through rubber, resin, etc., before reaching the rotor 22. Therefore, when the stator casing 20 is made of metal, it is sufficient to drill holes only in the stator casing 20 in order to transmit ultrasonic waves from the Doppler sensor 40, have them reflected by the rotor 22, and receive the reflected waves with the Doppler sensor 40, and there is no need to drill holes in the stator 21, which is made of rubber, resin, etc.

[0045] However, depending on the material and thickness of stator 21, such as rubber or resin, ultrasonic waves may be attenuated in stator 21. In this case, for example, a hole of an appropriate depth may be drilled in stator 21 and Doppler sensor 40 may be disposed in the hole. Furthermore, as will be described later, Doppler sensor 40 may be disposed inside stator 21 when control device 41 receives a Doppler signal output from Doppler sensor 40 by wireless communication.

[0046] Figures 5 to 8 show experimental data of Doppler signals during normal operation, empty operation, operation with air bubbles, and operation stop, respectively. Figures 5 to 8 show Doppler signals subjected to IQ detection (quadrature detection), that is, I (In-phase) signals and Q (Quadrature) signals are shown superimposed. In Figures 5 to 8, the I signal is shown by a gray solid line, and the Q signal is shown by a black dashed line.

[0047] In Figures 5 to 8, the horizontal axis indicates time (seconds) and the vertical axis indicates signal intensity. In Figures 5 to 8, the scales of the vertical and horizontal axes are aligned, but the position of the rotor 22 at the start point (time 0 seconds) is not aligned. This is the same in the following figures. During normal operation, idle operation, and operation with air bubbles (Figures 5 to 7), the rotation speed of the rotor 22 is set to 60 revolutions per minute, and when operation is stopped (Figure 8), the rotation speed of the rotor 22 is set to 0, that is, the rotor 22 is stopped. Furthermore, water was used as the fluid to be transferred in the experiments.

[0048] In normal operation, the pump chamber PR is completely filled with water, and no impurities such as air bubbles are present. Referring to Fig. 5 (normal operation), generally flat portions with low signal strength and steeply rising peak portions appear periodically and alternately. The flat portions indicate that the Doppler sensor 40 has not detected the rotor 22, i.e., that the ultrasonic waves are being absorbed by the water in the pump chamber PR. More specifically, they indicate that the rotor 22 is generally in a position other than that shown in Fig. 3.

[0049] The peak portion indicates that the Doppler sensor 40 has detected the rotor 22, i.e., that the ultrasonic waves are being reflected by the rotor 22 without being absorbed by the fluid in the pump chamber PR. More specifically, it indicates that the rotor 22 is approximately in the position shown in Fig. 3. Note that since the fluid interposed between the Doppler sensor 40 and the rotor 22 is in the form of a thin film, not only when the rotor 22 is completely in the position shown in Fig. 3, but also when the rotor 22 is slightly shifted from the position shown in Fig. 3, the ultrasonic waves pass through the thin film of fluid, reach the rotor 22 and are reflected, causing the Doppler sensor 40 to detect the rotor 22.

[0050] In this way, during normal operation, the degree of fluid filling in the pump chamber PR and the movement of the rotor 22 can be determined based on the waveform of the Doppler signal.

[0051] In the empty operation, the pump chamber PR is completely filled with air, i.e., there is no water. Referring to Fig. 6 (empty operation), the size of the peak portion is smaller than that of Fig. 5. This is because the transmission and reception of ultrasonic waves by the Doppler sensor 40 is not performed at a point that has no area, but within an area that has a certain area, and because the attenuation rate of ultrasonic waves is greater in air than in water.

[0052] In the air bubble mixing operation, the pump chamber PR is filled with water and air bubbles are mixed in. Referring to Fig. 7 (air bubble mixing operation), the waveform is randomly disturbed in parts. This is because the transmission and reception of ultrasonic waves are disturbed by air bubbles randomly present in the water.

[0053] In the stopped state, the pump chamber PR is completely filled with water, but the rotor 22 is not moving. Referring to Fig. 8 (stopped state), only a generally flat portion with low signal intensity is present. In other words, the ultrasonic waves continue to be absorbed by the fluid in the pump chamber PR.

[0054] The control device 41 performs arithmetic processing and controls the entire uniaxial eccentric screw pump 1. The control device 41 includes, for example, a CPU (Central Processing Unit) or MPU (Micro Processing Unit) that cooperates with software to realize a predetermined function. The control device 41 may be configured with hardware circuits such as dedicated electronic circuits or reconfigurable electronic circuits designed to realize a predetermined function, or may be configured with various semiconductor integrated circuits. Examples of various semiconductor integrated circuits include a CPU, an MPU, a microcomputer, a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and an ASIC (Application Specific Integrated Circuit). The control device 41 may also include a storage device such as a RAM (Random Access Memory) and a ROM (Read Only Memory). Specifically, the control device 41 may be configured with, for example, an information processing device such as a desktop personal computer, a notebook computer, a workstation, or a tablet terminal, or a printed circuit board having an equivalent function.

[0055] The control device 41 realizes a predetermined function by reading out the stored data and programs and performing various arithmetic processing. The programs executed by the control device 41 may be provided from the outside using a communication unit that communicates according to a predetermined communication standard, or may be stored in a portable recording medium.

[0056] The control device 41 is connected to the Doppler sensor 40 via a cable in order to receive the Doppler signal generated by the Doppler sensor 40. Note that, for example, the Doppler sensor 40 may be connected to a wireless transmitter and the control device 41 may be connected to a wireless receiver, so that the control device 41 receives the Doppler signal generated by the Doppler sensor 40 via wireless communication.

[0057] The control device 41 includes, as functional components, a determination image generating unit 41a and a determination unit 41b. These are realized by the cooperation of the above hardware and software. These may also be interpreted as corresponding circuits.

[0058] The determination image generating unit 41a generates a determination image corresponding to the Doppler signal. In this embodiment, the determination image generating unit 41a generates an autocorrelation function from the Doppler signal as shown in Figs. 5 to 8, and cuts out the autocorrelation function in a time range that is an arbitrary natural number N times the motion period of the rotor 22. Here, the autocorrelation function is obtained by IQ detecting the Doppler signal and performing complex correlation function processing, and finding the correlation coefficient.

[0059] Figures 9 to 12 respectively show the autocorrelation functions during normal operation, empty operation, operation with air bubbles, and when operation is stopped. Specifically, Figures 9 to 12 respectively show the autocorrelation functions generated from the Doppler signals of Figures 5 to 8. In Figures 9 to 12, the horizontal axis represents time, and the vertical axis represents the correlation coefficient.

[0060] Figures 13 to 16 show judgment images generated by extracting the autocorrelation functions of Figures 9 to 12 over a time range twice the motion period (two rotations) of rotor 22. That is, here, the arbitrary natural number N is set to 2. Specifically, Figures 13 to 16 show judgment images generated by extracting region XIII of Figure 9, region XIV of Figure 10, region XV of Figure 11, and region XVI of Figure 12 and displaying the vertical axis in a logarithmic scale.

[0061] The judgment images in Figures 13 and 14 are roughly symmetrical, and reproducibility can be confirmed with data for two rotations of the rotor 22. The judgment image in Figure 15 is not symmetrical, reflecting the randomness caused by air bubbles being mixed in. The judgment image in Figure 16 only shows a slight wavy curve at the bottom, reflecting the absence of any area where the signal strength increases due to the operation being stopped.

[0062] In this embodiment, the determining unit 41b outputs the driving state using the determination images as shown in FIGS. 13 to 16 as input data, and determines the driving state.

[0063] The judgment unit 41b is subjected to a learning process using teacher data so as to output an operating state when a judgment image is input. Specifically, the judgment unit 41b has a trained model that has machine-learned the relationship between the input teacher data (judgment image) and the operating state (normal operation, empty operation, operation with air bubbles, or operation stop). The trained model may be expressed as a discriminator, a classifier, a predictor, or the like. The learning may be performed by a neural network, a support vector machine, a decision tree, or a combination of these. The neural network may also utilize multi-layered deep learning. In this embodiment, experimental results using deep learning will be described below.

[0064] In the experiment, a comparative example was first confirmed in order to confirm the improvement in judgment accuracy by adopting machine learning. As a comparative example, a method of judging the driving state by setting a threshold value for the peak part of the Doppler signal without using machine learning was experimented. The rate of erroneous judgment in the comparative example was about 38%, that is, the accuracy rate was about 62%. Furthermore, the method of setting a threshold value as in the comparative example could not judge a driving state having randomness such as operation with air bubbles. Therefore, whether or not a result better than this judgment accuracy can be obtained is one indicator of whether or not it can be said to have high judgment accuracy.

[0065] In this embodiment, a total of 669 Doppler signal data were experimentally acquired during normal operation, empty operation, operation with air bubbles, and operation stop, and deep learning was performed to confirm the judgment accuracy. 75% of the 669 data were used for learning, and 25% were used for confirming the judgment accuracy. The deep learning model was constructed by the error backpropagation method, based on the known ResNet50 (convolutional neural network with a depth of 50 layers), the initial value was the known ImageNet (dataset after learning), and the parameters updated by learning were fine-tuned to narrow down to the latter 10 layers out of the total 50 layers. In addition, optimization was performed by the stochastic gradient descent method, cross entropy was used as the error function, and the activation function was the softmax function only for the final judgment part and the ReLU function for all other parts. The number of epochs was set to 150.

[0066] The accuracy of the determination was checked five times. The percentages of erroneous determinations in the five tests were 14%, 21%, 31%, 13%, and 12%, averaging about 18%. In other words, the accuracy rate was about 82%. Therefore, it can be said that good accuracy was achieved in this embodiment.

[0067] The display unit 42 is a part that displays the judgment result by the judgment unit 41b, and is configured by, for example, a liquid crystal display, an organic EL display, or a plasma display. The judgment result displayed on the display unit 42 may be not only a single operating state, but also a plurality of possible operating states and their probabilities. For example, it may be displayed as 90% normal operation, 5% empty operation, 3% air bubble mixed operation, and 1% operation stop. In addition, the display unit 42 may display the judgment image generated by the judgment image generation unit 41a together with the judgment result. The display unit 42 may be connected to the control device 41 by wire or wirelessly, similar to the above-mentioned Doppler sensor 40.

[0068] The uniaxial eccentric screw pump 1 of this embodiment provides the following advantages.

[0069] Since the Doppler signal is converted into a judgment image and used as input data for the judgment unit 41b, the operating state can be judged from a comprehensive perspective compared to when the operating state is judged simply from fragmentary information such as the peak value, the number of peaks, or the integral value of the Doppler signal. Even if the judgment image is difficult to distinguish visually, abnormal operation can be judged easily and with high accuracy by using the learned judgment unit 41b. Therefore, the operating state of the single-shaft eccentric screw pump 1 can be judged with high accuracy.

[0070] Furthermore, by using a determination image based on an autocorrelation function instead of using a graph of a Doppler signal as it is as a determination image, the characteristic portions of the signal waveform can be clearly captured.

[0071] Second embodiment The uniaxial eccentric screw pump (positive displacement pump) 1 according to the second embodiment differs from the first embodiment in the determination image generating unit 41a and the determination unit 41b. The rest of the second embodiment is substantially the same as the first embodiment. Therefore, the description of the parts shown in the first embodiment may be omitted.

[0072] In this embodiment, the judgment image generating unit 41a generates judgment images shown in Figs. 17 to 20. Figs. 17 to 20 are generated as radar charts in which the autocorrelation functions of Figs. 9 to 12 described in the first embodiment are cut out in a time range that is an arbitrary natural number N times (here, N=2) the motion cycle of the rotor 22, and the time range is synchronized with a phase angle of 0 to 360° to display the autocorrelation functions in polar coordinates. That is, Figs. 17 to 20 show the autocorrelation functions Cxx(τ) obtained by cutting out the region XIII of Fig. 9, the region XIV of Fig. 10, the region XV of Fig. 11, and the region XVI of Fig. 12 and displaying the vertical axis in logarithm, as shown in the following formula with the absolute value γ(τ) and the argument θ(τ) centered at the origin. Here, fs indicates the sampling frequency, and R indicates the number of revolutions per minute of the uniaxial eccentric screw pump 1, which is set to rotate once in the aforementioned arbitrary natural number N (here, N=2). It should be noted that only in FIG. 17, the notation of the absolute value γ(τ) and the argument θ(τ) is shown for reference, but such notation is not included in the determination image.

[0073]

number

[0074] The judgment images in Figures 17 and 18 are roughly symmetrical from top to bottom, and reproducibility can be confirmed with data for two rotations of the rotor 22. The judgment image in Figure 19 is not symmetrical from top to bottom, reflecting the randomness caused by air bubbles being mixed in. The judgment image in Figure 20 shows only a small crushed pattern in the center, reflecting the absence of areas with high signal strength due to the shutdown of the system.

[0075] In this embodiment, the determination unit 41b outputs the driving state using the determination images as shown in Figures 17 to 20 as input data, and determines the driving state. The determination unit 41b performs learning processing using the determination images as shown in Figures 17 to 20 as teacher data, and is substantially the same as the first embodiment except that the teacher data has been changed.

[0076] The confirmation of the judgment accuracy was carried out five times in the same manner as in the first embodiment. The percentages of erroneous judgments in the five times were 12%, 5%, 4%, 4%, and 5%, averaging about 6%. In other words, the accuracy rate was 94%. Therefore, in this embodiment, even better judgment accuracy was obtained.

[0077] In this embodiment, the autocorrelation function is made into a radar chart and used as the judgment image, so that the judgment accuracy can be improved compared to the case where the autocorrelation function itself is used as the judgment image. As for the improvement of the judgment accuracy, an experiment was performed as described above, and it was confirmed that the rate of erroneous judgment was reduced to one third before and after the radar chart was made.

[0078] As a modification of this embodiment, when the above-mentioned arbitrary natural number is set to N, the determination image generating unit 41a may extract the following portion of the angle range An of the radar charts of FIGS.

[0079]

number

[0080] When N=2 is used as an arbitrary natural number as in the above embodiment, the angle range An is 45° to 135° and 225° to 315°. For example, when comparing the angle ranges in Fig. 17 and Fig. 18, it can be seen that the shape in Fig. 17 protrudes upward and downward compared to Fig. 18, and a difference is clearly seen in the angle range. In contrast, there is no significant difference in the shape protruding left and right in Figs. 17 and 18. Therefore, it is expected that the part of the angle range An in the judgment image will function advantageously for judgment.

[0081] According to this modification, the amount of input data and teacher data can be reduced compared to the case where a judgment image of the entire phase angle from 0 to 360° is input. This is because characteristic data appears periodically due to the periodic motion of the rotor 22. The angle range An is set by conducting an experiment and confirming characteristic data that changes greatly depending on the operating state. Therefore, even if the amount of input data and teacher data is limited, the deterioration of the estimation accuracy can be suppressed to a certain extent.

[0082] Third embodiment In the single-shaft eccentric screw pump (positive-displacement pump) 1 according to the third embodiment, a determination image generating unit 41a and a determination unit 41b are different from those in the first and second embodiments. Therefore, the description of the parts shown in the first and second embodiments may be omitted.

[0083] In this embodiment, the judgment image generating unit 41a generates a power spectrum by Fourier transforming the Doppler signal, generates a spectral envelope that connects points of frequencies in the power spectrum that correspond to integer multiples of the motion period of the rotor 22, and generates a judgment image by extracting the spectral envelope within a predetermined frequency range.

[0084] Figures 21 to 24 respectively show power spectra during normal operation, empty operation, operation with air bubbles, and when operation is stopped. Specifically, Figures 21 to 24 respectively show power spectra generated from the Doppler signals of Figures 5 to 8 described in the first embodiment. In Figures 21 to 24, the horizontal axis indicates frequency (Hz) and the vertical axis indicates spectrum intensity. In these figures, peaks are taken at frequency points corresponding to integer multiples of the motion period. Peaks from -20 to 20 as the integers are plotted as black circles.

[0085] 25 to 28 show, as judgment images, spectral envelopes that connect the frequency points corresponding to integer multiples of the motion period of rotor 22 in the power spectra of Figs. 21 to 24, cut out in a predetermined frequency range (-20 Hz to 20 Hz). That is, Figs. 25 to 28 show, as judgment images, curves (spectral envelopes) that connect the black circles in Figs. 21 to 24. Note that, although -20 Hz to 20 Hz is used here as the predetermined frequency range, this is not particularly limited and any frequency range may be used.

[0086] 25 to 28, differences for each driving state are seen, similar to the determination images of the first embodiment (FIGS. 13 to 16) and the determination images of the second embodiment (FIGS. 17 to 20). Therefore, it is understood that the determination unit 41b can determine the driving state.

[0087] In this embodiment, the determination unit 41b outputs the driving state using the determination images as shown in Figures 25 to 28 as input data, and determines the driving state. The determination unit 41b performs learning processing using the determination images as shown in Figures 25 to 28 as teacher data, and is substantially the same as the first embodiment except that the teacher data has been changed.

[0088] The confirmation of the judgment accuracy was performed five times in the same manner as in the first embodiment. The percentages of erroneous judgments in the five times were 25%, 20%, 8%, 22%, and 20%, and the average was about 19%. In other words, the accuracy rate was 81%. Therefore, good judgment accuracy was obtained in this embodiment as well.

[0089] In this embodiment, instead of using a graph of the Doppler signal as it is as a determination image, a determination image based on a spectral envelope is used, making it possible to clearly capture the characteristic portions of the signal waveform.

[0090] (Fourth embodiment) In the uniaxial eccentric screw pump (positive displacement pump) 1 according to the fourth embodiment, the determination image generating unit 41a and the determination unit 41b are different from those in the first to third embodiments. Therefore, the description of the parts shown in the first to third embodiments may be omitted.

[0091] In this embodiment, the determination image generating unit 41a generates the determination images shown in Figs. 29 to 32. Figs. 29 to 32 are generated as radar charts in which the spectrum envelope connecting the black circles in Figs. 21 to 24 is cut out in a predetermined frequency range (here, -20 Hz to 20 Hz) of the rotor 22 and the predetermined frequency range (here, -20 Hz to 20 Hz) is synchronized with a phase angle of 0 to 360° to display the spectrum envelope in polar coordinates. That is, Figs. 29 to 32 are radar charts in which the spectrum envelope connecting the black circles in Figs. 21 to 24 is displayed in polar coordinates in the same manner as Figs. 17 to 20. The method of generating the radar chart is substantially the same as that in the second embodiment.

[0092] In this embodiment, the determination unit 41b outputs the driving state using the determination images as shown in Figures 29 to 32 as input data, and determines the driving state. The determination unit 41b performs learning processing using the determination images as shown in Figures 29 to 32 as teacher data, and is substantially the same as the first embodiment except that the teacher data has been changed.

[0093] The confirmation of the judgment accuracy was performed five times in the same manner as in the first embodiment. The percentages of erroneous judgments in the five times were 13%, 14%, 20%, 13%, and 12%, and the average was about 14%. In other words, the accuracy rate was 86%. Therefore, good judgment accuracy was obtained in this embodiment as well.

[0094] In this embodiment, by forming the spectrum envelope into a radar chart and using it as the judgment image, it is possible to improve the judgment accuracy compared to the case where the spectrum envelope itself is used as the judgment image. As for the improvement in judgment accuracy, an experiment was performed as described above, and it was confirmed that the rate of erroneous judgments was reduced to three-quarters before and after the radar chart was formed.

[0095] Although specific embodiments of the present invention have been described above, the present invention is not limited to the above embodiments, and various modifications can be made within the scope of the present invention.

[0096] In each of the above embodiments, the uniaxial eccentric screw pump 1 has been described as an example of a positive displacement pump, but the positive displacement pump may be another type of pump, such as a rotary pump or a piston pump. [Explanation of symbols]

[0097] 1. Single-shaft eccentric screw pump (positive displacement pump) 2 Drive unit 3 Pump body 4. Driving condition determination device 5 Source 10 Connection casing 11 Coupling rod 12 Coupling mechanism 13 Connection port 20 Stator casing 21 Stator 21a,21b Straight section 21c,21d arcuate part 22 Rotor 23 Sensor mounting hole 30 End stud 40 Doppler Sensor 41 Control device 41a Judgment image generation unit 41b Judgment part 42 Display section

Claims

1. A containment member defining a pump chamber; a moving member that moves a fluid in accordance with a periodic motion in the pump chamber; A driving state determination device for determining a driving state; Equipped with The driving state determination device includes: a Doppler sensor that transmits ultrasonic waves or microwaves toward the moving member and receives the reflected waves to generate a Doppler signal; a determination image generating unit that generates a determination image corresponding to the Doppler signal; a judgment unit that has undergone learning processing using teacher data so as to output the driving state when the judgment image is input; A positive displacement pump having

2. The determination image generating unit is generating an autocorrelation function from the Doppler signal; 2. The positive displacement pump according to claim 1, wherein the autocorrelation function is extracted over a time range that is an arbitrary natural number multiple of the motion period of the motion member.

3. The determination image generating unit is 3. The positive displacement pump according to claim 2, wherein a radar chart is generated by displaying the autocorrelation function in polar coordinates by synchronizing the time range with a phase angle of 0 to 360 degrees.

4. The positive displacement pump according to claim 3 , wherein the determination image generating unit extracts a portion of the radar chart within a following angle range An, where N is the arbitrary natural number.

5. The determination image generating unit is generating a power spectrum by performing a Fourier transform on the Doppler signal; A spectral envelope is generated by connecting frequency points corresponding to integer multiples of the motion period of the motion member in the power spectrum; 2. The positive displacement pump of claim 1, wherein the spectral envelope is clipped at a predetermined frequency range.

6. The determination image generating unit is 6. The positive displacement pump according to claim 5, wherein a radar chart is generated in which the spectrum envelope is displayed in polar coordinates by synchronizing the predetermined frequency range with a phase angle of 0 to 360 degrees.

Images