Analysis system using rotary volumetric single-shaft eccentric pump, method, and computer program, and analysis device for use in analysis system

The system analyzes object characteristics using a rotary displacement uniaxial eccentric pump with a sensor and learning model to enhance quality and reproducibility in food manufacturing, addressing the limitations of existing methods by leveraging pump transfer information.

JP2024160593APending Publication Date: 2024-11-14YAMAGATA UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2023075762
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-01
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

Existing methods for analyzing the characteristics of objects transferred by rotary positive displacement uniaxial eccentric pumps are inadequate for ensuring quality and reproducibility in food manufacturing processes, particularly in 3D food printing, as they rely on complex calculations and do not effectively utilize information from the pump transfer process.

Method used

A system comprising a rotary displacement uniaxial eccentric pump with a stator and rotor, equipped with a sensor to detect behavior during transfer, and an analysis device to analyze characteristics using sensor information, including a learning model to determine object properties.

Benefits of technology

Enables accurate analysis of object characteristics, improving quality and reproducibility in food manufacturing by utilizing pump transfer information, allowing for real-time monitoring and adjustment of processing conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024160593000001_ABST
    Figure 2024160593000001_ABST
Patent Text Reader

Abstract

To provide a system, a method, a computer program or the like for analyzing the properties of an object by utilizing information obtained from a pump during the transfer of the object.SOLUTION: A system includes: a rotary volumetric single-shaft eccentric pump that includes a stator with a female-threaded through-hole and a male-threaded rotor that rotates eccentrically while inserted into the through-hole; a sensor that detects the behavior observed in the stator during the transfer of an object via the single-shaft eccentric pump; and an analysis device that analyzes the properties of the object by utilizing sensor information obtained through the sensor.SELECTED DRAWING: Figure 2
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a system, method, and computer program for analyzing the characteristics of an object transported by a rotary positive displacement, single-axis eccentric pump. [Background technology]

[0002] Progressive cavity pumps (PCPs) (hereinafter, when simply referred to as "pump", it means a progressive cavity pump) are equipped with a stator and a rotor, and can transport a certain amount of liquid, slurry, powder, etc. by rotating the rotor inside the stator. This type of pump can transport various objects, such as slurry liquid containing solid particles, and has been widely used in various fields. However, since the intended purpose of the pump was the transport of objects itself, which was its original purpose, research themes have mainly been focused on the transport structure of the pump itself, such as the cavity shape of the stator and the transport accuracy of objects.

[0003] The present invention is directed to analyzing the characteristics of an object transferred by a pump, not the pump transfer structure itself, and is an invention based on a new perspective, and therefore there are no prior art documents. Note that Non-Patent Documents 1 and 2 described below are documents related to a physical reservoir, which will be described later, and Non-Patent Document 3 is a document related to software used to extract features. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] "Reservoir Computing", Kohei Tanaka, Ryosuke Nakane, Akira Hirose (Morikita Publishing Co., Ltd., March 2021) [Non-Patent Document 2] "Introduction to Soft Robotics", supervised by the Soft Robotics Research Group, supervised by the Robotics Society of Japan, edited by Koichi Suzumori et al. (Ohmsha, January 2021) [Non-Patent Document 3] Christ, M., Braun, N., Neuffer, J., & Kempa-Liehr, AW (2018). Time series feature extraction on basis of scalable hypothesis tests (tsfresh-a python package). Neurocomputing, 307, 72-77. Summary of the Invention [Problem to be solved by the invention]

[0005] In food processing technology, which is one of the important research themes of the present inventors, new food production methods such as 3D food printing are being developed every day. 3D food printing makes it possible to easily produce a variety of foods with a high degree of freedom in shape, but the quality and reproducibility of foods produced by this method are greatly dependent on the characteristics of the ingredients (ex. type, viscosity, concentration) that are the processing materials. The present inventors have conducted extensive research, particularly focusing on the process of transporting ingredients using a pump, and as a result, have discovered that the characteristics of ingredients can be analyzed using information obtained from a pump during the transport of ingredients, rather than from the ingredients themselves. This analysis result can greatly contribute to maintaining the quality and improving the reproducibility of not only the ingredients themselves, but also the final product food processed using such ingredients. Therefore, the object of the present invention is to provide a system, method, computer program, etc. that analyzes the characteristics of an object using information obtained from a pump during the transport of the object. It should be noted that, although conventional general fluid numerical analysis techniques for calculating the state of a fluid can obtain rheological properties and dynamics, and can visualize the fluid state in detail and dynamically, these approaches are computationally intensive and not necessarily effective for the purpose of simply mapping the type of fluid being transported. Therefore, conventional general knowledge is insufficient to solve the problem of the present invention. [Means for solving the problem]

[0006] In order to solve the above problems, a system according to one aspect of the present invention is characterized by comprising: a rotary displacement type uniaxial eccentric pump having a stator with a female threaded insertion hole and a male threaded rotor that rotates eccentrically while inserted into the insertion hole; a sensor that detects the behavior of the stator when an object is transported by the uniaxial eccentric pump; and an analyzer that analyzes characteristics of the object using sensor information obtained through the sensor. According to the system of this aspect, the characteristics of the object can be analyzed using information obtained from the pump during the transport of the object.

[0007] A method according to one aspect of the present invention is used for a rotary displacement type single-shaft eccentric pump having a stator with a female-threaded insertion hole and a male-threaded rotor that rotates eccentrically while inserted into the insertion hole, and is characterized by comprising the steps of: detecting, by a sensor, the behavior of the stator when an object is transported by the single-shaft eccentric pump; and analyzing, by an analyzer, the characteristics of the object using sensor information obtained through the sensor.

[0008] In addition, a method according to another aspect of the present invention is a method used in a system including a rotary displacement type uniaxial eccentric pump having a stator with a female threaded insertion hole and a male threaded rotor that rotates eccentrically relative to the stator while inserted into the insertion hole, a sensor that detects behavior of the stator when an object is transported by the uniaxial eccentric pump, and an analyzer that analyzes characteristics of the object using sensor information obtained through the sensor, characterized in that the method includes the steps of: acquiring the sensor information; acquiring object information related to the object; and generating a learning model that outputs characteristics of the object in response to input of the sensor information, based on the sensor information and the object information as training data.

[0009] Furthermore, a method according to another aspect of the present invention is a method used in a system including a rotary displacement type single-shaft eccentric pump having a stator with a female threaded insertion hole and a male threaded rotor that rotates eccentrically relative to the stator while inserted into the insertion hole, a sensor that detects the behavior of the stator when an object is transported by the single-shaft eccentric pump, and an analyzer that analyzes characteristics of the object using sensor information obtained through the sensor, characterized in that the method includes the steps of: acquiring the sensor information; acquiring object information related to the object; and outputting characteristics of the object in response to input of the sensor information using a learning model generated based on the sensor information and the object information as training data.

[0010] A computer program according to one aspect of the present invention is a computer program used in a rotary displacement type single-shaft eccentric pump having a stator with a female-threaded insertion hole and a male-threaded rotor that rotates eccentrically while inserted into the insertion hole, characterized in that the computer is caused to analyze characteristics of the object by using sensor information obtained through a sensor that detects behavior occurring in the stator when the object is transported by the single-shaft eccentric pump.

[0011] Furthermore, a computer program according to another aspect of the present invention is a computer program used in a system including a rotary displacement type uniaxial eccentric pump having a stator with a female threaded insertion hole and a male threaded rotor that rotates eccentrically while inserted into the insertion hole, a sensor that detects behavior of the stator when an object is transported by the uniaxial eccentric pump, and an analyzer that analyzes characteristics of the object using sensor information obtained through the sensor, characterized in that the computer acquires the sensor information, acquires object information related to the object, and generates a learning model that outputs characteristics of the object in response to input of the sensor information based on the sensor information and the object information as training data. Effect of the Invention

[0012] According to the present invention, there are provided a system, a method, a computer program, and the like, which utilize information obtained from a pump during the transport of an object to analyze the characteristics of the object. [Brief description of the drawings]

[0013] [Figure 1] FIG. 2 is a block diagram showing a configuration example of the present system. [Diagram 2] FIG. 2 is a perspective view showing an example of the appearance of the present system together with peripheral devices. [Diagram 3] FIG. 2 is a schematic perspective view showing an example of the basic configuration of a pump included in the present system, together with a sensor installed in the pump. [Figure 4]FIG. 2 is a front view showing an example of the appearance of a pump including optional configurations. [Diagram 5] FIG. 2 is an exploded perspective view of the pump, including optional features. [Figure 6] FIG. 5 is a bottom view of the pump shown in FIG. [Figure 7] FIG. 2 is a bottom view of the pump with the bottom spacer removed. [Figure 8] 4A to 4C are diagrams illustrating the function of a vibration waveform processing unit. [Figure 9] FIG. 13 is a diagram illustrating an example of a Fourier transformed amplitude spectrum. [Figure 10] FIG. 13 is a diagram illustrating an example of a Fourier transformed amplitude spectrum. [Figure 11] FIG. 1 shows a conceptual diagram of physical reservoir calculation. [Figure 12] FIG. [Figure 13] FIG. 1 illustrates the modeling procedure used to manufacture the rotor. [Figure 14] FIG. 1 illustrates the modeling procedure used to manufacture the rotor. [Figure 15] FIG. 1 illustrates the modeling procedure used to manufacture the rotor. [Figure 16] FIG. 1 illustrates the modeling procedure used to manufacture the rotor. [Figure 17] FIG. 1 illustrates the modeling procedure used to manufacture the rotor. [Figure 18] FIG. [Figure 19] FIG. [Figure 20] FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Figure 21] FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Figure 22] FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Diagram 23]FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Figure 24] FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Diagram 25] FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Figure 26] FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Figure 27] FIG. 1 illustrates the modeling procedure used to manufacture the stator. [Figure 28] 4A to 4C are diagrams illustrating a procedure for forming the gel stator of Example 1. [Figure 29] 1A to 1C are diagrams illustrating a procedure for forming the silicone rubber stators of Examples 2 to 4. [Diagram 30] FIG. 2 is a diagram showing the structure of a sensor used in the examples. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, exemplary embodiments for carrying out the present invention will be described in detail with reference to the drawings. However, the materials, sizes, shapes, and relative positions of components described in the following embodiments are arbitrary unless essential to the present invention, and can be changed according to the configuration of the device to which the present invention is applied or various conditions. In addition, terms indicating specific directions or positions (e.g., terms including "upper", "lower", "side", and "end") are used as necessary, but the use of these terms is for the purpose of facilitating understanding of the invention with reference to the drawings, and the technical scope of the present invention is not limited by the meanings of these terms. Therefore, unless otherwise specified, the scope of the present invention is not limited to the embodiments specifically described below.

[0015] 1. System Configuration Fig. 1 shows a block diagram of an example of the configuration of this system, and Fig. 2 shows an example of the external appearance of this system together with peripheral devices in a perspective view. Fig. 3 shows a schematic perspective view of an example of the basic configuration of a pump included in the system shown in Fig. 2 together with a sensor installed in the pump. Fig. 4 shows a front view of an example of the external appearance of a pump including optional configurations, and Fig. 5 shows an exploded perspective view of the same.

[0016] As clearly shown in Figures 1 and 2, this system 1 includes a rotary displacement type uniaxial eccentric pump 2 capable of transporting an object such as a slurry liquid, and a control device 3 (Figure 1). In addition to controlling the pump 2, the control device 3 also functions as an analysis device that analyzes the characteristics of an object using sensor information obtained through a sensor 4 installed in the pump 2.

[0017] 2. Pump As clearly shown in Fig. 3, the pump 2 includes a stator 21 and a rotor 22 as basic components, similar to a commercially available rotary displacement type uniaxial eccentric pump. In addition to these basic components (essential components), the pump 2 also includes optional components (the stator 21 is also shown for convenience) shown in Fig. 2, Fig. 4, and Fig. 5. Each of the members included in these basic components and optional components has a substantially bilaterally symmetrical shape.

[0018] The stator 21 has a generally rectangular shape, with the top 21a and bottom 21b being generally square. A female-threaded insertion hole 210 is provided in the stator 21 along the longitudinal direction connecting the top 21a and bottom 21b, penetrating substantially through the center of the stator 21. The rotor 22 rotates eccentrically while inserted into the insertion hole 210.

[0019] As the rotor 22 rotates, a series of sealed spaces (cavities) 212 are formed between the inner surface 211 of the insertion hole 210 of the stator 21 and the outer surface 221 of the rotor 22, generating suction force and successively creating new sealed spaces. As these sealed spaces 212 move, the object is transported from the inlet side to the outlet side, and in this system 1, from the upper side to the lower side of the stator 21.

[0020] The rotational force of the rotor 22 can be obtained, for example, through a motor device 5 located above the pump 10. The motor device 5 may be mounted on a workbench 11 together with the pump 10. A shaft 51 extending from the motor device 5 is connected to an end 22a of the rotor 22. A torque measuring device 12 may be provided in parallel with the motor device 5. The torque measuring device 12 measures the torque generated in the shaft 51 when the rotor 22 is rotated. The measured torque can also be used, for example, to measure the viscosity of an object.

[0021] The stator 21 is entirely surrounded by a casing 23 and spacers 24a to 24c. The spacers 24a and 24b are the same member, i.e., they have the same size and shape and are made of the same material. When assembling the device, the spacers 24a and parts (241a, 241b) of the spacers 24b are installed at the top and bottom of the casing 23, respectively, in a state where they are taken inside the casing 23. Meanwhile, the spacer 24c is completely housed inside the casing 23 together with part (241b) of the spacer 24b on the bottom side of the casing 23. A storage member 27 capable of storing objects is further provided on the upper part of the spacer 24a.

[0022] The casing 23 mainly includes four plate-like portions 230 surrounding the side surfaces of the stator 21, an annular flange portion 23a provided on the side of the top portion 21a of the stator 21, and an annular flange portion 23b provided on the side of the bottom portion 21b of the stator 21. The annular flange portion 23a and the annular flange portion 23b have the same size and shape. The annular flange portions 23a and 23b are each provided with a pair of ear portions 232a and 232b on opposing edges.

[0023] The bottom 21b of the stator 21 is elastically supported by the spacers 24b and 24c inside the casing 23. The spacer 24b has an annular flange portion 240b having an outer diameter and shape corresponding to the annular flange portion 23b of the casing 23. The annular flange portion 240b has a pair of ears 242b on the opposing edges, and an annular rectangular member 241b protruding upward is further provided at the center position of the annular flange portion 240b. Furthermore, the spacer 24c is provided in a state of being stacked on the annular rectangular member 241b. The spacer 24c has an outer diameter and shape corresponding to the annular rectangular member 241b, and is inserted into the inside of the casing 25 together with the annular rectangular member 241b. On the other hand, the top of the stator 21 is elastically supported by the spacer 24a. The spacer 24a has an annular flange portion 240a having an outer diameter and shape corresponding to the annular flange portion 23a of the casing 23. The annular flange portion 240a has a pair of ears 242a on opposing edges, and an annular rectangular member 241a protruding toward the bottom is provided at the center of the annular flange portion 240a. The annular rectangular member 241a is inserted into the casing 25 at a position facing the spacers 24b and 24c in the vertical direction, with the stator 21 sandwiched between them. As a result, the stator 21 is stably and elastically accommodated inside the casing 23 by utilizing the spacers 25a to 25c.

[0024] The storage member 27 includes a main body portion 28 and an annular flange portion 29. The main body portion 28 is composed of a substantially cylindrical portion 28a and a funnel-shaped portion 28b that tapers downstream, i.e., from the upper side to the lower side of the stator 21. The annular flange portion 29 has an outer diameter and a shape corresponding to the annular flange portion 23a of the casing 23 and the annular flange portion 240a of the spacer 24a, and has a pair of ears 292a on opposing edges, similar to the annular flange portion 23a and the annular flange portion 240a.

[0025] The casing 23, the spacers 24a, 24b, and the storage member 27 are aligned by being screwed at the top through the through holes 233a, 243a, 293a of the ears 232a, 242a, 292a of the respective annular flanges, and at the bottom through the through holes 233b, 243b of the ears 232b, 242b. When aligned, the storage member 27 communicates with the through hole 244a provided in the annular rectangular member 241a of the spacer 24a and the insertion hole 210 of the stator 21 through the through hole 280a provided in the funnel portion 28b, and further communicates with the through hole 244c provided in the spacer 24c and the through hole 244b provided in the spacer 24b.

[0026] 6 shows a bottom view of the pump shown in FIG. 4, and FIG. 7 shows a bottom view of the pump with the spacer 24b removed. As is clear from these figures, the rectangular through-hole 244c (see FIG. 7) in the spacer 24c and the circular through-hole 244b (see FIG. 6) in the spacer 24b are both significantly larger than the insertion hole 210 of the stator 21. Thus, the object is smoothly transferred from the inlet side to the discharge side through these through-holes. The rotor 22 extends to a position approximately on the inlet side of the through-hole 280a of the storage member 27 while being inserted through the insertion hole 210 of the stator 21 and the through-hole 244a of the spacer 24a. In actual use, a shaft 51 extending from the motor device 5 is connected to the upper end 223 of the rotor 22.

[0027] 3. Sensor, stator and rotor materials Unlike a general pump, the stator 21 of the pump 2 is provided with a sensor 4. The sensor 4 can sense the behavior, such as vibration, of the stator 21 when the pump 2 transfers an object. The vibration includes information about the properties of the object that are generated according to the difference in type, concentration, viscosity, etc. of the object being transferred. The sensor 4 is installed at a position where such information can be easily sensed. For example, it is preferable to install the sensor 4 on the outer surface of the stator 21, that is, between the outer surface of the stator 21 and the casing 23. As shown in FIG. 3, the sensor 4 may be installed on a side surface perpendicular to the major axis of the elliptical insertion hole 210 formed in the top 21a and bottom 21b of the stator, or on a side surface perpendicular to the minor axis. It is sufficient to install the sensor 4 on one side of the stator 21, but a pair of sensors may also be installed on the opposite side. Although the manufacturing cost increases, the behavior can be sensed more accurately. Of course, the sensor 4 may be installed on all sides.

[0028] The sensor 4 uses, for example, a piezoelectric element that detects the piezoelectric effect in which a voltage is generated by compression or expansion. Among piezoelectric elements, a thin type such as a piezoelectric film sensor is particularly preferable. By using a thin type, even if the piezoelectric element is installed between the outer surface of the stator 21 and the casing 23, it becomes easy to install it, and it is possible to prevent the device from becoming large. However, various sensors other than piezoelectric elements can be used as long as they can detect the behavior occurring in the stator 21 when the pump 2 transports an object.

[0029] In order to facilitate the detection of the behavior by the sensor 4, the material of the stator 21 is preferably one that easily generates the behavior required for analysis and easily propagates the behavior. For this reason, in selecting the material of the stator 21 in this system, attention is paid to the Young's modulus of the stator 21, and further to the relationship with the viscosity of the object to be transferred. In this system, a slight overlap (meaning the overlap "δ" in the embodiment described later) is provided between the stator 21 and the rotor 22, and this overlap is used to accurately measure the pressure that the object transferred by the rotor 22 presses against the stator 21. Note that, although some conventional pumps have such an overlap, it is not always the case. This pressure is easily affected by the Young's modulus of the stator 21 and the viscosity of the object to be transferred, so the inventors carried out development while paying attention to these values. Conventional stators are made of rubber and resin, more specifically, materials are important and selected from acrylonitrile butadiene rubber, hydrogenated acrylonitrile butadiene rubber, chlorinated polyethylene, ethylene propylene rubber, butyl rubber, fluororubber, perfluororubber, urethane rubber, silicone rubber, polyamide, polytetrafluoroethylene, etc., but for the above reasons, this system 1 focuses on not only the material but also the relationship between the Young's modulus of the stator 21 and the viscosity of the object to be transported. On the other hand, the manufacturing material of the rotor 22 does not have a significant effect on the detection of behavior in this system, so various commonly used materials such as stainless steel, titanium, ceramics, Hastelloy (registered trademark), Carpenter (registered trademark), and Silicolloy (registered trademark) can be used.

[0030] 4. Control device (1) Overview The control device 3 may be configured by any of hardware, DSP (Digital Signal Processor), and software. When configured by software, for example, it is configured with a cloud server, a GPU of a computer, a CPU, RAM, ROM, etc., and is realized by the operation of a computer program stored in a recording medium such as a RAM, a ROM, a hard disk, or a semiconductor memory. More specifically, the computer program used in the present system may be provided in a state stored in a non-transitory computer-readable storage medium including software for controlling one or more electronic data processing devices. Computer-readable storage media include, but are not limited to, floppy disks, magnetic hard disk drives, solid-state hard disks, flash memories, USB thumb drives, random access memories (RAM), read-only memories (ROM), optical disks, magneto-optical disks, and processor register files. For example, the term computer-readable storage medium also includes various types of recording media that can be accessed by a computer via a network or a communication link.

[0031] The control device 3 includes a control unit 300 that controls the entire device, a sensor information generation unit 305 that generates sensor information, a sensor information acquisition unit 320 that acquires sensor information, an object information acquisition unit 330 that acquires object information related to objects, a learning model generation unit 340 that generates a learning model 341, and an adjustment unit 360 that adjusts the operation of the pump 2, as well as a memory unit 350, a communication unit 351, an input / output unit 352, a display unit 353, and a reading unit 354.

[0032] The control unit 300 can be configured with a CPU, a GPU, a ROM, a RAM, and the like. The memory unit 350 is composed of a hard disk or flash memory, etc., and can store information such as processing results within the control device 3, such as a learning model generated by this system, and information obtained from the external Internet. The reading unit 354 can read the computer program that defines the processing of the control device 3 from the storage unit 350 in which the computer program is recorded, and from a recording medium (not shown).

[0033] (2) Generation of sensor information The sensor information generation unit 305 includes a vibration waveform processing unit 310 , a peak extraction unit 315 , and a feature extraction unit 316 .

[0034] The vibration waveform processing unit 310 performs pre-processing on the vibration waveform obtained from the sensor 4 through the input / output unit 352. The vibration waveform processing unit 310 includes a digital conversion unit 311, a Fourier transform unit 312, a noise processing unit 313, and an inverse Fourier transform unit 314.

[0035] The function of the vibration waveform processing unit 310 will be described with reference to Fig. 8 to Fig. 10. Fig. 8(a) shows an example of a vibration waveform obtained through the sensor 4, which is a piezoelectric element, in other words, a vibration waveform 61 before pre-processing. In addition to the waveform shown in Fig. 8(a), the waveforms shown in Fig. 8(b) to (d) are displayed to the user by the display unit 353. The user can perform processing while checking the waveforms through the display unit 353.

[0036] First, the vibration waveform 61 is converted into a digital signal by the digital conversion unit 311 .

[0037] Next, the digitally converted vibration waveform is Fourier transformed by the Fourier transform unit 312, and is decomposed into frequency components including a plurality of amplitude spectra. Fig. 8(b), Fig. 9, and Fig. 10 show examples of the Fourier transformed amplitude spectra 62, 62-1 to 62-4.

[0038] Next, noise is removed from the Fourier transformed amplitude spectra 62, 62-1 to 4 by the noise processing unit 313. For the amplitude spectra 62, 62-1, and 62-2 shown in Fig. 8(b) and Fig. 9, noise removal is performed by a general method, that is, by setting an appropriate threshold, for example, 0.0012, and removing the amplitude spectrum 62 having an amplitude equal to or less than the set threshold.

[0039] In contrast, for example, in the case of the amplitude spectra 62-3 and 62-4 shown in FIG. 10, since the difference between the amplitude spectra is small, if the above-mentioned general method is used, most of the amplitude spectra will be removed, making analysis impossible. Therefore, instead of the general method, a method is adopted for removing the amplitude spectra corresponding to "n%" of the multiple peaks while giving priority to the amplitude spectra showing smaller peaks. Here, "n" is a value that should be appropriately set by the user, and for example, the value of "n" that can correctly classify all objects even for unlearned evaluation data using a learning model generated according to the method described later is used. Note that when searching for the value of "n", the search is performed in 10% increments, for example, 1%, 10%, 20%, and so on.

[0040] Finally, as shown in FIG. 8(c), the amplitude spectrum from which noise has been removed by the noise processing unit 313 is inverse Fourier transformed by the inverse Fourier transform unit 314 to restore the vibration waveform.

[0041] In the peak extraction unit 315, in order to perform machine learning, a plurality of peaks (640 in FIG. 8(c)) are extracted from the vibration waveform processed by the vibration waveform processing unit 310.

[0042] The feature extraction unit 316 extracts features to be used as sensor information from the extracted peaks 640 (FIG. 8(d)). During extraction, the feature extraction unit 316 may further extract multiple feature amounts from each of the above-mentioned peaks 640. These extracted feature amounts are used as "sensor information" to be used for creating a learning model, etc.

[0043] (3) Acquiring sensor information The sensor information acquisition unit 320 acquires the sensor information generated by the sensor information generation unit 305 .

[0044] (4) Acquiring object information The object information acquisition unit 330 acquires object information related to an object acquired from a user via the input / output unit 352, such as the type, viscosity, concentration, etc. of the object, as a correct label.

[0045] (5) Generating learning models The learning model generation unit 340 generates a learning model 341 based on the sensor information and object information as learning data acquired from the sensor information acquisition unit 320 and the object information acquisition unit 330. This learning model 341 is learned (trained) by supervised machine learning so as to output characteristics of an object corresponding to the object information, such as the type of object (e.g., an estimated accuracy that enables the type to be determined), viscosity, concentration, etc., in response to the sensor information input from the sensor information acquisition unit 320. The generated learning model 341 is stored in the storage unit 350. Furthermore, the learning model 341 may be transmitted to the management server 15 via the external Internet 13 using the communication unit 351.

[0046] The learning model 341 can be generated using known machine learning algorithms such as neural networks that include physical reservoir calculations. As disclosed in Non-Patent Documents 1 and 2 and many other documents, physical reservoir calculation is a known technique, but an outline of it will be given below for completeness. Figure 11 shows a conceptual diagram comparing physical reservoir computation with a conventional recurrent neural network. Conventionally, recurrent neural networks (RNNs) have been considered suitable for machine learning methods for time-series data. Physical reservoir computation is derived from this recurrent neural network, and is a machine learning method that does not learn anything other than the final output layer in the learning of a recurrent neural network. In other words, it is a method of replacing the nonlinear transformation function of reservoir computation with a physical system that performs calculations suitable for high-speed machine learning of time-series data. Furthermore, it was revealed that the structure of the neural network itself has sufficient generalization, so that sufficient calculations can be performed by balancing only the final output layer. Physical reservoir computation is faster and less power-consuming than software implementation, and in particular, expectations are high for industrial applications regarding its ability to process time-series data obtained from a physical system in real time, and various studies have been reported. In particular, soft matter physical reservoir calculations replace actual physical systems with soft matter materials, while multi-soft matter physical reservoir calculations replace actual physical systems with multiple soft matter materials. By combining physical memory calculations with texture evaluation of soft matter, this device constitutes a new soft machine that can detect even subtle differences in food with high accuracy.

[0047] (6) Operation of the control device After generating the learning model 341, the control device 3 can output the characteristics of the object by using the sensor information acquired through the sensor information acquisition unit 321, in other words, can analyze the characteristics of the object. The learning model 341 used by the control device 3 may be generated by its own learning model generation unit 340, or may be acquired through the management server 15 via the external Internet 13. Furthermore, the learning model 341 once acquired may be changed, updated, etc. by another learning model managed by the management server 15.

[0048] The control device 3 can also adjust the operation of the pump 2 according to the characteristics of the object. For example, if an abnormality is found in the concentration of the object as a result of the analysis, the control device 3 can perform processing such as emergency stopping of the pump 2 and issuing a warning to the user.

[0049] 5. Working Example The present invention will be described in more detail below with reference to examples, but the present invention is not limited to the following examples. The system of the examples is for analyzing the type of an object. [Example 1] (1) Pump The pump was manufactured in the following manner. 1) Rotor manufacturing A side view of the rotor is shown in Figure 12, and an illustration of the modeling procedure used to manufacture the rotor is shown in Figures 13 to 17. In these figures and the following description, "d (= D+2δ)": Rotor diameter (mm), 6.20mm ("D": Stator diameter (mm), 6.00mm / "δ": Rotor and stator overlap (mm), 0.100mm) "λ R ": Rotor pitch (mm), 10.00mm · "L R ": Length of rotor in longitudinal direction (mm), 40.00mm · "e": Eccentricity (mm), 1.00mm means.

[0050] First, set the center O and create a circle C with diameter d. R (Fig. 13) Next, draw a circle C from the center O. R and the length L R Draw a straight line H (Figure 14). Then draw a circle C R Move e from the center O (Figure 15). Here, the straight line H is the path along the circle C. R The torsion angle is 360*L. R / λ R The rotor is then extruded while being twisted at 150 degrees (Figure 16) to obtain the rotor shown in Figure 17.

[0051] According to the above modeling, a rotor was shaped using a stereolithography printer Shuffle 4K (manufactured by Phrozen) using a resin material, namely, CLEAR RESIN (model number: 22343, manufactured by Wanhao).

[0052] 2) Stator manufacturing Figure 18 shows a side view of the stator, Figure 19 shows a bottom view, and Figures 20 to 27 show diagrams illustrating the modeling procedure used to manufacture the stator. In these diagrams and the following description, "l S ": Length of one side (bottom and top) of the stator in the short direction (mm), 20.00mm "D": Stator diameter (mm), 6.00mm "δ": Rotor and stator overlap (mm), 0.100mm "λ S (= 2λ R )": Stator pitch (mm), 20.00mm · "L S ": Length of the stator in the longitudinal direction (mm), 30.00mm · "e": Eccentricity (mm), 1.00mm means.

[0053] First, set the center O and create a circle C with diameter D. S1、 and Circle C S2 (Figure 20) Next, draw a circle C from the center O. S1 and the length L S Draw a straight line H (Figure 21). Then draw a circle C S1 Move 2e from center O to circle C S2 From center O to circle C S1 Then, move 2e in the opposite direction (Fig. 22). S1 and circle C S2 Draw a tangent line connecting these (Figure 23). This creates an ellipse C with a major axis of 4e+D and a minor axis of D that can correspond to the cross-sectional shape of the stator. e This is completed (Fig. 23). To make a stator from this, first, the length of one side at the center O is l SDraw a square (Fig. 25), and extrude it using a straight line H (see Fig. 21) as a path to create an ellipse (Fig. 21). Next, draw an ellipse C using the straight line H as a path. e Cut it out from the body created earlier by sweeping (Figure 27). In this case, the twist angle is 360*L. S / λ S Twist it at a right angle and cut it off.

[0054] According to the above modeling, a gel stator was shaped using EG gel made by the following method. First, the monomer N,N-dimethylacrylamide (KJ Chemicals) was prepared, and the crosslinking agent N,N'-methylenebis(acrylamide) (Fujifilm Wako Pure Chemical Industries) was added to the N,N-dimethylacrylamide while stirring it with a magnetic stirrer at room temperature and about 400 rpm. After stirring until the N,N'-methylenebis(acrylamide) was dissolved, the photoinitiator polymerization agent 2-oxoglutaric acid (Fujifilm Wako Pure Chemical Industries) was added and similarly stirred until dissolved. Then, ethylene glycol (Fujifilm Wako Pure Chemical Industries) was added as a solvent. After adding the ethylene glycol, the mixture was stirred for about 30 minutes to obtain an EG gel solution. The resulting solution contained 49.25 wt% N,N-dimethylacrylamide (KJ Chemicals), 0.38 wt% N,N'-methylenebis(acrylamide), 0.36 wt% 2-oxoglutaric acid, and 50.00 wt% ethylene glycol (Fujifilm Wako Pure Chemicals).

[0055] With reference to FIG. 28, the procedure for forming a gel stator using the above EG gel material will be described. First, a master (mold) was created (FIG. 28(a)). This master is composed of master A with a spiral shape inside and box-shaped master B. Using CLEAR RESIN as the material, master A was formed by a stereolithography printer Shuffle 4. As in the above, the formed master A was immersed in IPA (isopropyl alcohol) for 5 days. Meanwhile, master B was prepared by laser cutting an acrylic plate with a thickness of 3 mm. Silicone rubber (Ecoflex (registered trademark) 00-30) (manufactured by Smooth-On) was poured into these masters A+B (FIG. 28(b)), and the masters A+B were removed to prepare a silicone rubber mold (FIG. 28(c)). After that, an EG gel solution was poured into the completed silicone rubber mold and irradiated with UV light to form a gel stator (FIG. 28(d) and (e)).

[0056] The Young's modulus of the molded stator was measured by a compression test. The test piece to be measured was manufactured using the same method as the stator, and was cylindrical with a diameter of 29 mm and a height of 12.5 mm. The test equipment used was a benchtop material testing machine SAT-1150 (manufactured by A&D Co., Ltd.), and a load cell with a maximum allowable capacity of 500 N was used. The test was carried out at a compression speed of 5 mm / min, and three test pieces were prepared. The Young's modulus of each was calculated, and the average of the three was obtained. As a result, a value of 91 kPa was obtained.

[0057] 3) Any configuration An optional configuration of the pump was as shown in FIGS. The acrylic plate constituting the casing 23 was 3 mm thick and transparent so that the state of the object during transportation could be observed. Four plate-shaped parts were cut out of this acrylic plate using a laser processing machine, and the edges of the parts were glued together so as to surround the stator to form the casing 23. The casing 23 was only installed so as to surround the stator, and was not glued to the stator or the sensor. In this way, a structure is adopted in which the casing 23 is not fixed to the stator or the sensor, particularly to the sensor, so that it is possible to substantially prevent the sensor from impeding the detection of the behavior of the stator.

[0058] The spacers 24a to 24c were produced by creating a mold using an FDM 3D printer and then filling the mold with silicone rubber (Ecoflex (registered trademark) 00-30). The thickness (see FIG. 5) of the annular flange portion 240b of the spacer 24b provided on the bottom side of the casing 23 was set to 2 mm. The thickness (see FIG. 5) of the annular rectangular member 241b protruding from the annular flange portion 240b was set to 5 mm, and the outer frame was set to 20 mm in both the vertical and horizontal directions. The size of the outer frame corresponds to the length (mm) of one side (bottom and top) in the short direction of the stator 21. Furthermore, the diameter of the circular through hole 244b (see FIG. 6) penetrating the center of the annular flange portion 240b and the annular rectangular member 241b was set to 14 mm. The thickness (see FIG. 5) of the spacer 24c was set to 5 mm, the same as the annular rectangular member 241b, and the outer frame was also set to 20 mm in both the vertical and horizontal directions. Furthermore, the rectangular through hole 244c (see Figure 7) that passes through the center of the spacer 24c has a vertical length of 14 mm along the major axis of the elliptical insertion hole 210 formed in the bottom 21b of the stator 21, and a horizontal length of 10 mm along the minor axis.

[0059] (2) Sensors and acquisition of sensor information The sensor used was the DT1-028K (manufactured by TE Connectivity Measurement Specialties), a thin-film piezoelectric film sensor with excellent flexibility and responsiveness. Details can also be obtained via the following URL (as of the filing date). For reference, the materials at the URL are attached at the end of the specification. https: / / www.digikey.jp / ja / datasheets / te-connectivity-measurement-specialties / te-connectivity-measurement-specialties-piezo_film_guide Figure 30 shows a schematic exploded perspective view of the sensor structure, obtained through the above URL. The sensor is in the shape of a vertically long rectangular sheet as a whole. The dimensions indicated by the symbols in the figure are as follows: TIFF2024160593000002.tif21156

[0060] As shown in FIG. 3, the sensor is attached to one side of the stator, more specifically, to one side perpendicular to the major axis of the elliptical insertion hole 210 formed in the top 21a and bottom 21b of the stator, over substantially the entire side. The attachment position is aligned to the center in the left-right direction, and in the up-down direction, the lower side of the wiring part 41 (see FIG. 3), i.e., the part about 11 mm from the upper edge of the sensor 4, is aligned with the upper surface edge of the stator. In this case, the sensor is arranged symmetrically along the transfer direction of the object transferred by the rotor and with the insertion hole 210 included in the center in the width direction perpendicular to the transfer direction, and as a result, it is possible to more efficiently and accurately detect, for example, vibration, which is a behavior occurring in the stator when the object is transferred by the pump 2. A removable double-sided tape was used to attach the sensor to the stator. Using the above sensor, we set it to send data 120 times per second.

[0061] (3) Processing by the control device (analysis device) 1) Analysis subjects We calculated the estimation accuracy for seven types of objects, namely, (a) butter, (b) honey, (c) mayonnaise, (d) meringue, (e) olive oil, (f) silicone oil, and (g) sugar water, in order to analyze their properties, in this case the types of objects. Meringue and sugar water were prepared by the researchers themselves, and the rest were commercially available products. Meringue was made by adding 10g of sugar to one beaten egg white. More specifically, the egg white was beaten until it foamed to a certain extent, 1 / 3 of the sugar was added, and the mixture was further beaten until it became finely textured, after which the remaining 2 / 3 was added and further beaten. Sugar water was made by mixing 200g of sugar with 100g of water.

[0062] Furthermore, the viscosity of each of these objects was measured using a tuning fork vibration viscometer SV-10A (manufactured by A&D Corporation). As a result, at room temperature (approximately 25 degrees), the following values ​​were obtained: (a) butter: 3200 mPa·s, (b) honey: 6640 mPa·s, (c) mayonnaise: 3740 mPa·s, (d) meringue: 2920 mPa·s, (e) olive oil: 67.4 mPa·s, (f) silicone oil: 618 mPa·s, and (g) sugar water: 304 mPa·s.

[0063] 2) Sensor information generation After filling the stator with each object, the rotor was rotated about 30 times at a maximum rotation speed of 20 rpm, and three sets of data (vibration waveforms) were obtained for each object through the sensor. Each set contains approximately 8,000 pieces of data. The objects in the stator are rotated approximately once by the rotor's rotation about 30 times. A vertical rotation type test machine TL901 (manufactured by Trinity Labs) was used to rotate the rotor 22. Two of the three sets of data obtained for each object were used as learning data for generating a learning model, and the remaining set was used as evaluation data for evaluating the performance of the generated learning model. In addition, the evaluation data was also used to obtain the value of "n" described above, as necessary.

[0064] The learning data and evaluation data (corresponding to the vibration waveform 61 shown in (a) of FIG. 8) acquired through the sensor were digitally converted into duodecimal numbers using an A / D converter ADS1015 manufactured by TEXAS INSTRUMENTS, which is an example of the digital conversion unit 311. However, since the range of the converted data is from -2048 to +2048, it was normalized to the range of -1 to +1 to adapt it to machine learning.

[0065] Next, the digitally converted vibration waveform was Fourier transformed by the Fourier transform section 312 to decompose it into frequency components, in other words, into a plurality of amplitude spectra (corresponding to the amplitude spectrum 62 shown in FIG. 8(b)).

[0066] Next, noise was removed by the noise processing unit 313. The noise was removed by a method in which a threshold value of 0.001 was set and amplitude spectra having amplitudes equal to or less than the threshold value were removed.

[0067] Finally, the amplitude spectrum from which noise has been removed by the noise processing unit 313 is inverse Fourier transformed by the inverse Fourier transform unit 314 to restore the vibration waveform.

[0068] The peak extraction section 315 extracts approximately 470 peaks (corresponding to 640 in FIG. 8C) from the vibration waveform 64 processed by the vibration waveform processing section 310.

[0069] The feature extraction unit 316 (corresponding to (d) in FIG. 8) further extracts about 40 feature amounts from each of these extracted peaks, and increases the feature amounts to obtain sensor information. tsfresh (see Non-Patent Document 3) was used for the feature extraction unit 316. The sensor information obtained by applying tsfresh, i.e., the feature amounts obtained from each peak, was further standardized in terms of the ratio of the feature amounts using StandardScaler() of tsfresh.

[0070] 3) Acquiring object information The object information was obtained from the user through the input / output unit 352. The object information included seven types of objects, namely, (a) butter, (b) honey, (c) mayonnaise, (d) meringue, (e) olive oil, (f) silicone oil, and (g) sugar water. These were used as correct labels in the learning model.

[0071] 4) Generation of learning model The learning model generation unit 340 acquired the sensor information and object information for the learning data from the sensor information acquisition unit 320 and the object information acquisition unit 330, and generated the learning model 341 based on the acquired information. 80% of the learning data was used for training, and the remaining 20% ​​was used for testing.

[0072] The learning model 341 was trained to output the characteristics of an object in response to sensor information input from the sensor information acquisition unit 320. More specifically, the learning model was set to display which of the seven types of objects the object transported by the pump is in a "percentage (%)" corresponding to the estimation accuracy in relation to other objects (the sum of the percentages displayed for each object is 100%), and after learning, the learning model was trained to display the largest "percentage (%)" for objects that match the object information acquired from the object information acquisition unit 330 so that all items can be correctly classified based on the percentage.

[0073] The average accuracy rate for the training data at the time of generating the learning model, more specifically, the average accuracy rate for the seven types of objects, was 88.0%. Here, even the lowest accuracy rate for butter was 75%, which was a high value.

[0074] 5) Analysis results When the evaluation data was displayed with the trained learning model 341 and classified based on the above-mentioned "proportion (%)," a relatively high estimation accuracy was obtained for all the objects, and all the objects were correctly classified. The proportion of (d) meringue, which had the lowest estimation accuracy, was 44.0%, and it was clearly classified from the second highest proportion of (c) mayonnaise, which was 29.0%.

[0075] [Example 2] Except for the points specifically noted below, this embodiment can be considered to be the same as the first embodiment.

[0076] 1) The stator Instead of the EG gel stator used in Example 1, a silicone rubber (Ecoflex (registered trademark) 00-30) stator was used. The procedure for forming this silicone rubber stator will be described with reference to FIG. 29. First, a spiral mold A and a box-shaped mold B are prepared (FIG. 29(a)). Mold A was formed using a stereolithography printer Shuffle 4K, using CLEAR RESIN as the material. The formed master A was then immersed in IPA (isopropyl alcohol) for five days. If this process was neglected, the silicone rubber would not harden when it was poured in later. Meanwhile, mold B was made by laser cutting a 3 mm thick acrylic plate. Silicone rubber (Ecoflex (registered trademark) 00-30) was poured into this mold (FIG. 29(b)), and the silicone rubber stator was formed (FIG. 29(c)). The Young's modulus of the molded stator was measured by a compression test, resulting in a value of 176 kPa.

[0077] 2) Sensor information generation Since the difference between the amplitude spectra obtained from the vibration waveform processing unit 310 was small, the noise processing unit 313 did not use a general noise removal method, but used a method of removing a number of amplitude spectra equivalent to "n%" of the multiple peaks while giving priority to amplitude spectra showing smaller peaks. More specifically, using a learning model generated according to a method described later, the value of "n" that correctly classified all objects for the unlearned evaluation data, which is "10" in this embodiment, was used to remove a number of small amplitude spectra equivalent to "10%" of the multiple peaks.

[0078] 3) Generation of learning model The average accuracy rate for the training data at the time of generating the learning model was 77.0%. However, the accuracy rates for (f) silicone oil and (g) sugar water were slightly lower at 58% and 36%, respectively.

[0079] 4) Analysis results When the system equipped with the trained learning model 341 was used to display the above-mentioned "percentage (%)" for the evaluation data and classify the data based on the percentage, the estimation accuracy was low in some areas, and as a result, (a) butter was mistakenly classified as (d) meringue, and (b) honey was mistakenly classified as (e) olive oil. However, other objects could be correctly classified. Thus, although some of the classification results contained errors, the above-mentioned "percentage (%)" for (a) butter was 39.0%, while the above-mentioned "percentage (%)" for (d) meringue was 40.0%, so the estimation accuracy was not so low and the degree of error was not so large. On the other hand, the above-mentioned "percentage (%)" for (b) honey was 0.0%, while the above-mentioned "percentage (%)" for (e) olive oil was 77.0%, so the estimation accuracy was significantly low. However, taken as a whole, the method is practical, and the above errors can be easily corrected by devising a learning method and appropriately adjusting the value of "n" set in the noise processing section.

[0080] The results are summarized in Table 1 below. In Table 1, "Percentage (%)" indicates the percentage (corresponding to the estimation accuracy) displayed for each of the seven types of objects. For (a) and (b), the percentages for misclassifications (d) and (e) are also added. From the results of Examples 1 and 2, it is inferred that in estimation (classification) of unlearned data when the viscosity of the object is about 67 to 6641 mPa s, it is preferable to set the Young's modulus of the stator to about 90 to 177 kPa.

[0081] [Table 1]

[0082] [Example 3] Except for the points specifically noted below, this embodiment can be considered to be the same as the first embodiment. (1) Stator A stator made of silicone rubber (HTV-2000) (Engraving Japan) was used instead of the stator made of EG gel used in Example 1. The molding procedure was the same as that for silicone rubber (Ecoflex (registered trademark) 00-30) described in Example 2. The compression test described above gave a Young's modulus of 816 kPa for HTV-2000.

[0083] (2) Processing by the control device (analysis device) 1) Analysis subjects For five types of objects, namely, (h) broccoli, (i) carrot, (j) komatsuna, (k) pumpkin, and (l) tomato, and their respective ink materials ((h) broccoli: broccoli fine powder manufactured by Mikasa Sangyo Co., Ltd., (i) carrot: carrot fine powder manufactured by Mikasa Sangyo Co., Ltd., (j) komatsuna: komatsuna fine powder manufactured by Mikasa Sangyo Co., Ltd., (k) pumpkin: pumpkin fine powder manufactured by Mikasa Sangyo Co., Ltd., (l) tomato: ripe tomato powder manufactured by Tamachan Shop Co., Ltd.), the estimation accuracy was obtained for each object in order to analyze the type of these objects, as in Example 1. Each ink material was made by adding water to food powder in a powder:water ratio of 1:3 and mixing to prepare 100g of ink material. Viscosity measurements were also carried out on these objects. For the measurements, a viscoelasticity measuring device MCR-302E (Anton Paar) was used, and the shear rate was set to 0.1-0.5 [1 / s] during the measurements. Furthermore, the viscosity was measured at a shear rate of 20 rpm in rpm terms. This is because the rotor was rotated at 20 rpm when sensing was carried out. As a result, for (h) broccoli, it was 7.13×10 5 mPa·s, (i) for carrots, 2.02×10 5 mPa s, (j) for Japanese mustard spinach, 3.41×10 5 mPa s, (k) for pumpkin, 2.15×10 5 mPa·s, (l) for tomatoes, 2.68×10 5 The value obtained was mPa·s.

[0084] 2) Acquiring object information The object information was acquired from the user through the input / output unit 352. The object information included five types of objects, namely, (h) broccoli, (i) carrot, (j) Japanese mustard spinach, (k) pumpkin, and (l) tomato. These were used as correct labels in the learning model.

[0085] 3) Generation of learning model At the time of generating the learning model, the average accuracy rate for the training data, more specifically, for the five types of objects, was 82.0%. However, the accuracy rate for the carrot, which was the lowest, was 58%.

[0086] 4) Analysis results When the above "proportion (%)" of the evaluation data was displayed and classification was performed based on the proportion using a system equipped with the trained learning model 341, good estimation accuracy was obtained for all the objects, and as a result, all the objects were correctly classified. However, although the classification was correct, the above "proportion (%)" for (h) broccoli was 45.1%, while the above "proportion (%)" for (j) komatsuna was 43.1%, and the estimation accuracy was not very high. However, it would be relatively easy to improve the accuracy by devising a learning method, etc.

[0087] [Example 4] Except for the points specifically noted below, this embodiment can be considered to be the same as Example 3. 1) The stator Instead of the silicone rubber (HTV-2000) stator used in Example 3, a silicone rubber (Ecoflex (registered trademark) 00-10) (manufactured by Smooth-On) stator was used. The molding procedure was the same as that for the silicone rubber (Ecoflex (registered trademark) 00-30) described in Example 2. The compression test described above gave a Young's modulus of 84 kPa for Ecoflex (registered trademark) 00-10.

[0088] 2) Generation of learning model At the time of generating the learning model, the average accuracy rate for the training data was 97%. Even for the lowest accuracy rate, carrot, a high value of 83% was achieved.

[0089] 3) Analysis results When the system equipped with the trained learning model 341 was used to display the above "proportion (%)" of the evaluation data and classify it based on the proportion, a relatively high estimation accuracy was obtained for all of the data, and as a result, all of the objects were correctly classified. For the above "proportion (%)" of (l) tomato, which had the lowest estimation accuracy, a value of 54.8% was obtained, and it was possible to clearly classify it from (h) broccoli, which had the second highest "proportion (%)" of 43.7%.

[0090] The results are summarized in Table 2 below. In Table 2, the "percentage (%)" indicates the percentage (corresponding to the estimation accuracy) displayed for each of the five types of objects, as in Table 1. From the results of Examples 3 and 4, it can be seen that the viscosity of the object is 2.0×10 5 ~7.0×10 5 In estimation (classification) of unlearned data when the Young's modulus is about mPa·s, it is estimated that it is preferable to set the Young's modulus of the stator to about 83 to 817 kPa.

[0091] [Table 2]

[0092] Further aspects, features and advantages of the present invention will become readily apparent from the following detailed description, which merely sets forth a number of specific embodiments and examples, including the best mode contemplated for carrying out the invention. The present invention is also capable of other and different embodiments, and its numerous details can be modified in various obvious respects, without departing from the spirit and scope of the present invention. Accordingly, the drawings and description are by way of illustration only, and not of limitation.

[0093] For example, in the embodiment described above, the type of object is exemplified as the object to be analyzed by the present system, but the present system can also be used to analyze the viscosity, concentration, and other properties of an object. For example, the present system can be used as a system for quality control of an object by detecting a change in the viscosity or concentration of an object during the transfer of a certain type of object. Therefore, the object to be analyzed by the present system is not limited to the type of object.

[0094] In the above-described embodiment, EG gel and silicone rubber (HTV-2000, Ecoflex (registered trademark) 00-10, Ecoflex (registered trademark) 00-30) are used as the material of the stator 21, but other materials can be used as long as they are capable of generating the vibrations required for analysis and propagating these vibrations. Therefore, the material of the stator 21 of the present invention is not limited to these.

[0095] Furthermore, in the embodiment described above, a resin material is used as the material of the rotor 22, but as already explained, the manufacturing material of the rotor 22 does not significantly affect the detection of behavior in this system, so various other commonly used materials can be used. [Explanation of symbols]

[0096] 1 System 2. Pump 3 Control device (analysis device) 4 Sensors 5. Motor device 13 Network 15 Management Server 21 Stator 22 Rotor 23 Casing 25a, b, c Spacer 27 Storage material 300 Control section 305 Sensor information generation unit 310 Vibration waveform processing unit 311 Digital Conversion Department 312 Fourier Transform Section 313 Noise Processing Unit 314 Inverse Fourier Transform 315 Peak detector 316 Feature Extraction Unit 320 Information Acquisition Department 321 Sensor information acquisition unit 322 Object information acquisition unit 340 Learning model generation unit 341 Learning Model 350 Memory unit 351 Communication unit 360 Adjustment unit JPEG2024160593000005.jpg219170

Claims

1. A stator having a female-threaded insertion hole; a male-threaded rotor that rotates eccentrically while being inserted into the insertion hole; and A sensor that detects a behavior of the stator when an object is transported by the single-shaft eccentric pump; an analysis device that analyzes characteristics of the object using sensor information obtained through the sensor; A system comprising:

2. When the viscosity of the object is 67 to 6641 mPa·s, the Young's modulus of the stator is set to 90 to 177 kPa, and the viscosity of the object is 2.0×10 5 ~7.0 x 10 5 The system of claim 1 , wherein the Young's modulus of the stator is set to 83 to 817 kPa for mPa·s.

3. The system of claim 1 , wherein the sensor senses vibrations occurring in the stator.

4. The analysis device includes: A digital conversion unit that converts the vibration waveform obtained through the sensor into a digital signal; A Fourier transform unit that performs a Fourier transform on the vibration waveform digitally converted by the digital converter; a noise processing unit that removes a number of amplitude spectra corresponding to n% of the multiple peaks of the amplitude spectrum obtained by performing the Fourier transform in the Fourier transform unit, while giving priority to amplitude spectra showing smaller peaks; an inverse Fourier transform unit that restores the vibration waveform by performing an inverse Fourier transform on the amplitude spectrum from which noise has been removed by the noise processing unit; The system of claim 3 .

5. The system of claim 3 , wherein the sensor is a piezoelectric element.

6. The system of claim 5 , wherein the piezoelectric element is disposed on an outer surface of the stator.

7. The system of claim 1 , wherein the analysis device further comprises an adjustment unit that adjusts operation of the single-axis eccentric pump in response to an analyzed characteristic of the object.

8. The analysis device includes: A sensor information acquisition unit that acquires the sensor information; an object information acquisition unit that acquires object information relating to the object; The system according to claim 1 , further comprising a learning model generation unit that generates a learning model that outputs characteristics of the object in response to input of the sensor information, based on the sensor information and the object information as training data.

9. The system according to claim 8 , further comprising a communication unit that transmits the learning model generated by the learning model generation unit to a management server.

10. The analysis device includes: A sensor information acquisition unit that acquires the sensor information; an object information acquisition unit that acquires object information relating to the object; The system according to claim 1 , further comprising a learning model that is generated based on the sensor information and the object information as training data, and that outputs characteristics of the object in response to input of the sensor information.

11. The system of claim 10 , wherein the learning model is a learning model generated in another system.

12. The analysis device for use in the system according to claim 1 .

13. A stator having a female-threaded insertion hole; a male-threaded rotor that rotates eccentrically while being inserted into the insertion hole, detecting a behavior of the stator by a sensor when the single-shaft eccentric pump transports an object; analyzing a characteristic of the object by an analysis device using sensor information obtained through the sensor; 23. A method comprising:

14. A stator having a female-threaded insertion hole; a male-screw type rotor that is inserted into the insertion hole and rotates eccentrically with respect to the stator; A sensor that detects a behavior of the stator when an object is transported by the single-shaft eccentric pump; an analysis device that analyzes characteristics of the object using sensor information obtained through the sensor; 1. A method for use in a system comprising: acquiring the sensor information; obtaining object information relating to the object; generating a learning model that outputs a characteristic of the object in response to input of the sensor information, based on the sensor information and the object information as training data; A method for providing the above.

15. The method of claim 14 , comprising transmitting the generated learning model to a management server.

16. A stator having a female-threaded insertion hole; a male-screw type rotor that is inserted into the insertion hole and rotates eccentrically with respect to the stator; A sensor that detects a behavior of the stator when an object is transported by the single-shaft eccentric pump; an analysis device that analyzes characteristics of the object using sensor information obtained through the sensor; 1. A method for use in a system comprising: acquiring the sensor information; obtaining object information relating to the object; outputting a characteristic of the object in response to input of the sensor information, using a learning model generated based on the sensor information and the object information as training data; A method for providing the above.

17. The method of claim 16 , wherein the learned model is a learned model generated in another system.

18. converting a vibration waveform obtained through the sensor into a digital signal; performing a Fourier transform on the vibration waveform digitally converted by the digital converter; removing a number of amplitude spectra corresponding to n% of the multiple peaks of the amplitude spectrum obtained by the Fourier transform, while giving priority to amplitude spectra showing smaller peaks; Restoring the vibration waveform by performing an inverse Fourier transform on the noise-removed amplitude spectrum; 18. The method of any of claims 13 to 17, comprising:

19. A stator having a female-threaded insertion hole; a male-threaded rotor that rotates eccentrically while being inserted into the insertion hole, On the computer, analyzing characteristics of the object using sensor information obtained through a sensor that detects a behavior of the stator when the object is transported by the single-shaft eccentric pump; A computer program comprising:

20. A stator having a female-threaded insertion hole; a male-threaded rotor that rotates eccentrically while being inserted into the insertion hole; and A sensor that detects a behavior of the stator when an object is transported by the single-shaft eccentric pump; an analysis device that analyzes characteristics of the object using sensor information obtained through the sensor; A computer program for use in a system comprising: On the computer, Acquire the sensor information; obtaining object information relating to the object; generating a learning model that outputs a characteristic of the object in response to input of the sensor information, based on the sensor information and the object information as training data; A computer program comprising: