Remote monitoring system

Abstract

To provide a remote monitoring system which can perform optimal remote monitoring of a state of an object to be monitored.SOLUTION: A remote monitoring system 1 comprises an imaging apparatus 9 for generating image data, and an image analysis system 10 for analyzing an abnormal state of treated water 3 on the basis of the image data. The image analysis system 10 comprises an edge server 15 and a cloud server 17. The cloud server 17 comprises a Web application 19 and is configured to cause the Web application 19 to display an analysis result of the abnormal state and the image data. The image analysis system 10 stores an inference model constructed by machine learning using training data including at least the image data, and is configured to input the image data to the inference model and infer the abnormal state of the treated water 3 in accordance with the inference model.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a remote monitoring system for monitoring the state of a solid-liquid separated material and the state of water to be treated. [Background technology]

[0002] Conventionally, dehydrators have been used to squeeze suspensions (e.g., sludge) discharged from liquid treatment facilities such as sewage treatment plants, sewage treatment plants, and industrial wastewater treatment plants to separate water from the suspension (i.e., dewater it). In order to determine whether the treatment by such dehydrators was satisfactory or not, a person had to visually check the state of the separated matter such as dehydrated cake and the state of the treated water such as the suspension.

[0003] There are also diagnostic systems for devices or facilities that use machine learning, and systems that simply monitor the status of facilities remotely (for example, SCADA: Supervisory Control And Data Acquisition). SCADA is often used for one-to-one monitoring, with one SCADA for each site, and is often on-premise (hardware and software are installed on-site).

[0004] The applicant has been developing an AI system for remote monitoring of dewatering equipment, and has reported on a remote monitoring system equipped with AI that can determine whether the operating status of sludge treatment equipment (dewatering machine) is normal or abnormal based on images taken by a remote monitoring camera. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2017-182529 A [Patent Document 2] Patent No. 2613360 [Patent Document 3] Patent No. 5351092 [Non-Patent Document 1] "Expanding business opportunities with SCADA x IoT", [online], [searched February 18, 2021], Internet,<https: / / www.jte.co.jp / package / watcher / lp / iot / > [Non-Patent Document 2] Tomohiro Itakura, Katsuko Kusumoto, Daisuke Koga, "Development of a remote monitoring AI system for dewatering equipment," Proceedings of the Sewerage Research Conference, published on July 22, 2020 Summary of the Invention [Problem to be solved by the invention]

[0006] In order to visually judge the condition of the separated material and the treated water, it was necessary to regularly visit the site to carry out inspections, which required time and manpower. Since abnormalities (for example, the dehydration state of the dehydrated cake is not desirable) are judged by appearance, it is difficult to directly measure abnormal conditions with conventional sensors, and automation was not possible. If an abnormality in the equipment state (in the case of a dehydrator, the dehydration state) is overlooked (the moisture content is not desirable), it will affect downstream equipment in the plant, which places a large mental burden on the supervisor.

[0007] The above-mentioned Patent Document 1 discloses a means for connecting a host device and a display device via a network, but does not mention a function for remotely monitoring plant equipment or a function for multiple people to check the status of one piece of plant equipment through a web application. Since this is only possible on a closed network, remote monitoring is not possible even if you want to, and work efficiency cannot be improved.

[0008] The above-mentioned Patent Document 3 discloses a system for remote monitoring using a public line (Internet), but does not mention transmitting data including an image of the water 3 to be treated.

[0009] The above-mentioned non-patent document 2 discloses a very simple system that distinguishes between normal and abnormal equipment conditions, but does not describe a specific method for deploying this system in various treatment plants.

[0010] Therefore, an object of the present invention is to provide a remote monitoring system capable of optimally remotely monitoring the state of separated materials and water to be treated in various treatment plants. [Means for solving the problem]

[0011] In one aspect, a remote monitoring system for monitoring the condition of water to be treated is provided, comprising: an imaging device that images the water to be treated and generates image data; and an image analysis system that analyzes an abnormal condition of the water to be treated based on the image data, wherein the image analysis system comprises an edge server electrically connected to the imaging device; and a cloud server connected to the edge server via a network, the cloud server having a Web application, and the cloud server configured to cause the Web application to display the analysis results of the abnormal condition and the image data, wherein the image analysis system stores an inference model constructed by machine learning using training data including at least the image data, and the image analysis system is configured to input the image data into the inference model, infer an abnormal condition of the water to be treated according to the inference model, classify the analysis results into multiple stages, and normalize and calculate the analysis results.

[0012] In one aspect, the cloud server is connected to a user terminal via the network, and the cloud server is configured to send an email to the user terminal notifying that the condition of the treated water is abnormal if the cloud server determines that the condition of the treated water is abnormal. In one aspect, the cloud server is connected to a warning light via the network, and the cloud server is configured to notify via the warning light that the condition of the treated water is abnormal when the cloud server determines that the condition of the treated water is abnormal. In one aspect, the Web application is configured to be able to set an abnormality condition for the cloud server to determine that the state of the water to be treated is abnormal. In one aspect, the remote monitoring system further includes a sensor that measures a physical quantity that indirectly indicates the condition of the treated water, and the image analysis system is configured to further input the measurement data of the physical quantity into the inference model and infer an abnormal condition of the treated water according to the inference model. In one embodiment, the sensor is any one or more of a temperature sensor, a humidity sensor, a water quality sensor, a sound sensor, a vibration sensor, and an odor sensor. In one aspect, the inference model is stored in the edge server, and the edge server is configured to input at least the image data into the inference model and infer an abnormal state of the treated water according to the inference model, and is configured to transmit the analysis results and the image data to the cloud server via the network. In one aspect, the edge server is configured to transmit the analysis results and the image data at different transmission frequencies. In one aspect, the edge server is configured to transmit the image data to the cloud server via the network, the inference model is stored in the cloud server, and the cloud server is configured to input at least the image data into the inference model and infer an abnormal state of the treated water according to the inference model.

[0013] In one aspect, a remote monitoring system for monitoring the state of a solid-liquid separated product is provided, the remote monitoring system comprising: a solid-liquid separation device that performs solid-liquid separation of water to be treated to produce the separated product; an imaging device that images the separated product and generates image data; and an image analysis system that analyzes an abnormal state of the separated product based on the image data. The image analysis system comprises an edge server electrically connected to the imaging device; and a cloud server connected to the edge server via a network. The cloud server comprises a Web application. The cloud server is configured to cause the Web application to display an analysis result of the abnormal state and the image data. The image analysis system stores an inference model constructed by machine learning using training data that includes at least the image data. The image analysis system is configured to input the image data into the inference model, infer an abnormal state of the separated product according to the inference model, classify the analysis results into a plurality of stages, and normalize and calculate the analysis results.

[0014] In one aspect, the cloud server is connected to a user terminal via the network, and the cloud server is configured to send an email to the user terminal notifying the user that the separated object's condition is abnormal if the cloud server determines that the separated object's condition is abnormal. In one aspect, the cloud server is connected to a warning light via the network, and the cloud server is configured to notify via the warning light that the separated object's condition is abnormal if the cloud server determines that the separated object's condition is abnormal. In one embodiment, the web application is configured to allow setting of abnormal conditions that the cloud server uses to determine that the state of the separated object is abnormal. In one aspect, the remote monitoring system further comprises a sensor that measures a physical quantity that indirectly indicates the state of the separated object, and the image analysis system is configured to further input the measurement data of the physical quantity into the inference model and infer an abnormal state of the separated object according to the inference model. In one embodiment, the sensor is any one or more of a temperature sensor, a humidity sensor, a water quality sensor, a sound sensor, a vibration sensor, and an odor sensor. In one aspect, the inference model is stored on the edge server, and the edge server is configured to input at least the image data into the inference model and infer an abnormal state of the isolate according to the inference model, and is configured to transmit the analysis results and the image data to the cloud server via the network. In one aspect, the edge server is configured to transmit the analysis results and the image data at different transmission frequencies. In one aspect, the edge server is configured to transmit the image data to the cloud server via the network, the inference model is stored on the cloud server, and the cloud server is configured to input at least the image data into the inference model and infer an abnormal state of the isolate according to the inference model. Effect of the Invention

[0015] According to the present invention, the image data of the separated material or the water to be treated and the analysis results of the image data using machine learning can be confirmed by a web application. As a result, the condition of the separated material or the water to be treated can be optimally monitored remotely. [Brief description of the drawings]

[0016] [Figure 1] FIG. 1 is a schematic diagram illustrating an embodiment of a remote monitoring system. [Diagram 2] FIG. 13 is a schematic diagram showing another embodiment of the remote monitoring system. [Diagram 3] FIG. 13 is a schematic diagram showing another embodiment of the remote monitoring system. [Figure 4] FIG. 4(a) shows image data of the separated object 5 when the image analysis system 10 infers that it is normal, and FIG. 4(b) shows image data of the separated object 5 when the image analysis system 10 infers that it is abnormal. [Diagram 5] FIG. 13 shows image data of the separated material as turbid residue fed into a sludge dryer, and the analysis results of each image data. [Figure 6] FIG. 13 is a diagram showing image data of water in an aquarium tank as the water to be treated, and analysis results of each image data. [Figure 7] FIG. 2 illustrates one embodiment of a display of a web application. [Figure 8] FIG. 13 is a schematic diagram showing another embodiment of the remote monitoring system. [Figure 9] FIG. 1 shows an example of a method for monitoring the separated product as a dehydration filtrate. [Figure 10] FIG. 13 is a schematic diagram showing another embodiment of the remote monitoring system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. 1 is a schematic diagram showing one embodiment of a remote monitoring system. The remote monitoring system 1 of this embodiment is a remote monitoring system for monitoring the state of a separated material obtained by solid-liquid separation. The remote monitoring system 1 includes a solid-liquid separation device 7 for performing solid-liquid separation of the water to be treated 3 to generate a separated material 5, an imaging device 9 for imaging the separated material 5 and generating image data, and an image analysis system 10 for analyzing an abnormal state of the separated material 5 based on the image data.

[0018] Examples of the water to be treated 3 include suspensions (e.g., sludge) discharged from liquid treatment facilities such as sewage treatment plants, sewage treatment plants, industrial wastewater treatment plants, and water treatment facilities for breeding organisms. In this embodiment, the solid-liquid separation device 7 is a dehydrator that separates water from the water to be treated 3 (i.e., dehydrates it) to generate a separated matter 5. Specifically, the solid-liquid separation device 7 is a screw press.

[0019] Examples of the separated material 5 include a turbid residue (e.g., cake) remaining after removing liquid from the water to be treated 3 (e.g., sludge) and a separated liquid (dehydrated filtrate) separated from the water to be treated 3. In one embodiment, the separated material 5 as a turbid residue falls onto the conveyor 12 and is transported by the conveyor 12.

[0020] As the solid-liquid separator 7, other types of dehydrators such as a filter press, a centrifuge, a multi-disk type dehydrator (e.g., a slit saver), a sand filter, a pressurized flotation treatment device, a mechanical thickener, a gravity type sludge settling device, and a protein skimmer device may be used in addition to the screw press. Furthermore, in one embodiment, the solid-liquid separator 7 may be a drying device that generates a separated matter 5 (e.g., a turbid residue) by heating the water to be treated 3 and evaporating the water.

[0021] The imaging device 9 is a digital camera equipped with an image sensor (e.g., a CCD image sensor or a CMOS image sensor) capable of generating still or continuous images of an object. Alternatively, the imaging device 9 may be a hyperspectral camera capable of imaging an object by decomposing light into wavelengths. In this embodiment, the imaging device 9 is arranged facing the separated object 5 on the conveyor 12 and configured to be able to image the separated object 5 on the conveyor 12. In one embodiment, the imaging device 9 may be arranged facing the solid-liquid separation device 7, may be configured to be able to image the separated object 5 before it falls onto the conveyor 12, or may be configured to be able to image the separated object 5 after it falls from the conveyor 12.

[0022] As shown in FIG. 2, in one embodiment, the imaging device 9 may be arranged facing the water to be treated 3 and configured to be able to image the water to be treated 3, i.e., the water to be treated 3 before entering the solid-liquid separation device 7. In this embodiment, the remote monitoring system 1 functions as a remote monitoring system for monitoring the state of the water to be treated 3. An example of the water to be treated 3 imaged by the imaging device 9 is a suspension stored in a storage tank. In one embodiment, the water to be treated 3 imaged by the imaging device 9 may be water in a tank of an aquarium. In this case, the imaging device 9 is arranged facing the tank of the aquarium. When the water to be treated 3 is water in a tank of an aquarium, the remote monitoring system 1 does not include the solid-liquid separation device 7 and the conveyor 12. In another embodiment, the imaging device 9 may be arranged in the tank or the storage tank. In another embodiment, the water to be treated 3 imaged by the imaging device 9 may be treated water that has been treated by biological treatment of wastewater (for example, activated sludge treatment or fluidized carrier method). In this case, the imaging device 9 is arranged facing the tank after the biological treatment.

[0023] The image analysis system 10 includes an edge server 15 and a cloud server 17 connected to the edge server 15 via a network 16 such as the Internet. The edge server 15 includes a storage device 15a that stores a program for creating a model by executing machine learning described below and using the model, a calculation device 15b that executes calculations according to instructions included in the program, and a transmission / reception unit 15c that transmits image data and data such as analysis results of an abnormal state of the separated material 5 or the water to be treated 3 described below to the cloud server 17. Hereinafter, the separated material 5 and the water to be treated 3 may be collectively referred to as the monitored object. The edge server 15 is placed at the site where the monitored object is placed.

[0024] The edge server 15 is composed of at least one computer. The storage device 15a includes a main storage device such as a RAM, and an auxiliary storage device such as a hard disk drive (HDD) and a solid state drive (SSD). Examples of the computing device 15b include a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). However, the specific configuration of the edge server 15 is not limited to these examples.

[0025] The imaging device 9 is electrically connected to the edge server 15. Image data of the monitored object generated by the imaging device 9 is sent to the edge server 15 and stored in the storage device 15a.

[0026] The cloud server 17 includes a storage device 17a in which programs and the like are stored, a calculation device 17b that executes calculations in accordance with instructions contained in the programs, a transmission / reception unit 17c for receiving data such as image data and analysis results of abnormal conditions of the monitored object described below, and a Web application 19 that displays the analysis results of abnormal conditions of the monitored object and image data generated by the imaging device 9.

[0027] Cloud server 17 is composed of at least one computer. Storage device 17a includes a main storage device such as a RAM, and an auxiliary storage device such as a hard disk drive (HDD) and a solid state drive (SSD). Examples of computing device 18 include a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). However, the specific configuration of cloud server 17 is not limited to these examples.

[0028] The cloud server 17 is connected to a user terminal 22 via the network 16. Examples of the user terminal 22 include a personal computer in a monitoring room installed at a location away from the cloud server 17, a smartphone, a tablet terminal, and the like. A user (monitoring person) can access the Web application 19 from the user terminal 22 via a Web browser, and can check the analysis result of the abnormal state of the monitored object and image data from the Web application 19, and can operate the Web application 19. The cloud server 17 may be connected to a plurality of user terminals 22 via the network 16. In this case, the user can access the Web application 19 from a plurality of user terminals 22. In one embodiment, the image analysis system 10 may include a plurality of cloud servers 17.

[0029] The image analysis system 10 acquires image data generated by the imaging device 9, and analyzes the abnormal state of the monitored object based on the image data. By analyzing the abnormal state of the monitored object, it is possible to determine whether the treatment by the solid-liquid separation device 7 or the water quality of the water to be treated 3 is good or bad. Specifically, the image analysis system 10 analyzes the abnormal state of the monitored object by inference using artificial intelligence (AI). The image data generated by the imaging device 9 is sent to the edge server 15 and stored in the storage device 15a.

[0030] The image analysis system 10 includes an inference model. The inference model is stored in the storage device 15a of the edge server 15. The edge server 15 inputs image data obtained from the imaging device 9 to the inference model, and infers an abnormal state of the monitored object according to the algorithm of the inference model. The inference model is a trained model created by machine learning using training data including image data (explanatory variables) generated by the imaging device 9 during a past solid-liquid separation process and inference result data as a correct answer label (objective variable) determined by an operator. The edge server 15 creates the inference model according to a machine learning algorithm using training data including the past image data and the corresponding inference result data. The machine learning algorithm is not particularly limited, but a neural network deep learning, random forest, gradient boosting method, etc. can be suitably used.

[0031] The computing device 15b of the edge server 15 executes machine learning using the training data according to the instructions included in the program stored in the storage device 15a, and constructs an inference model. The obtained inference model is stored in the storage device 15a. In one embodiment, the construction of the inference model does not have to be performed in the edge server 15. An inference model may be constructed using the training data in a computer other than the edge server 15, and the inference model may be brought to the edge server 15. The edge server 15 inputs the image data generated by the imaging device 9 during the solid-liquid separation process into the learned inference model, and infers the abnormal state of the monitored object according to the inference model. In this embodiment, the edge server 15 infers whether the state of the monitored object is abnormal or normal according to the inference model. Furthermore, the edge server 15 acquires image data from the imaging device 9, and periodically or irregularly executes machine learning using the training data including the image data and the corresponding inference result data, and updates the inference model.

[0032] The edge server 15 transmits the analysis result (i.e., the inference result of the abnormal state of the monitored object) and image data of the monitored object to the cloud server 17. The analysis result and image data are stored in the storage device 17a. The edge server 15 calculates the analysis result as a categorical variable. For example, when the edge server 15 infers that the state of the monitored object is normal according to the inference model, it calculates "1", and when the edge server 15 infers that the state of the monitored object is abnormal, it calculates "0".

[0033] In one embodiment, the edge server 15 may be configured to transmit the analysis result of the abnormal state of the monitored object and the image data at different transmission frequencies. For example, the edge server 15 may transmit the analysis result once every 10 seconds and the image data once every minute. By appropriately setting each transmission frequency, it is possible to obtain the required effect while reducing the amount of data communication.

[0034] FIG. 3 is a schematic diagram showing another embodiment of the remote monitoring system 1. The configuration and operation of this embodiment that are not particularly described are the same as those of the embodiment described with reference to FIG. 1 and FIG. 2, so the overlapping description will be omitted. In this embodiment, a model is created by executing machine learning, and a program for using the model is stored in the storage device 17a. In this embodiment, the edge server 15 transmits image data of the monitored object to the cloud server 17. In this embodiment, the inference model is stored in the storage device 17a of the cloud server 17. The cloud server 17 inputs the image data transmitted from the edge server 15 to the inference model, and infers an abnormal state of the monitored object according to the algorithm of the inference model.

[0035] The inference model is a trained model created by machine learning using training data including image data (explanatory variables) generated by the imaging device 9 during a past solid-liquid separation process and inference result data as a correct label (objective variable) determined by an operator. The cloud server 17 creates the inference model according to a machine learning algorithm using the training data including the past image data and the corresponding inference result data.

[0036] The computing device 17b of the cloud server 17 executes machine learning using the training data according to the instructions included in the program stored in the storage device 17a, and constructs an inference model. The obtained inference model is stored in the storage device 17a. In one embodiment, the construction of the inference model does not have to be performed in the cloud server 17. An inference model may be constructed using the training data in a computer other than the cloud server 17, and the inference model may be brought to the cloud server 17. The cloud server 17 inputs the image data generated by the imaging device 9 during the solid-liquid separation process into the learned inference model, and infers the abnormal state of the monitored object according to the inference model. In this embodiment, the cloud server 17 infers whether the state of the monitored object is abnormal or normal according to the inference model. Furthermore, the cloud server 17 acquires image data from the edge server 15, and executes machine learning periodically or irregularly using the training data including the image data and the corresponding inference result data, and updates the inference model. The analysis result (i.e., the inference result of the abnormal state of the monitored object) and the image data of the monitored object are stored in the storage device 17a. The cloud server 17 calculates the above analysis result as a categorical variable.

[0037] FIG. 4 is a diagram showing an example of image data of the separated material 5 and an analysis result. FIG. 4(a) shows image data of the separated material 5 when the image analysis system 10 (edge ​​server 15 or cloud server 17) infers that the separated material 5 is normal, and FIG. 4(b) shows image data of the separated material 5 when the image analysis system 10 infers that the separated material 5 is abnormal. The separated material 5 shown in FIG. 4 is a turbid residue. The separated material 5 in FIG. 4(a) has a low moisture content and is dry, so it is evenly scattered on the conveyor. The separated material 5 in FIG. 4(b) has a high moisture content and is sparsely scattered with sludge attached.

[0038] In one embodiment, the image analysis system 10 (edge ​​server 15 or cloud server 17) may classify and calculate the above analysis results into a plurality of levels (for example, 3 levels to 5 levels). FIG. 5 is a diagram showing image data of the separated material 5 as turbid residue and the analysis results of each image data. The image data shown in FIG. 5 is an image of the separated material 5 that has been fed into a sludge dryer (not shown) by the conveyor 12. In the example shown in FIG. 5, the image analysis system 10 classifies the analysis results of the abnormal state of the separated material 5 into 5 levels, from level 1 to level 5, based on the moisture content (degree of dryness) inferred from the image data. In FIG. 5, the moisture content decreases from level 1 to level 5.

[0039] The image analysis system 10 calculates the analysis results when classified into a plurality of stages as categorical variables. For example, the image analysis system 10 calculates "1" when the moisture content level is 1, and calculates "3" when the moisture content level is 3. In one embodiment, the image analysis system 10 may calculate the analysis results of the abnormal state by normalizing (0 to 1, or 0 to 100). In one embodiment, for example, when a separated material 5 with a moisture content level of 3 is desired to be obtained, if the operating state changes to level 2 or level 4, the cloud server 17 may issue an abnormality notification as described below, or the user may change the operating parameters.

[0040] By classifying the analysis results into multiple stages, it is possible to flexibly respond to the different requirements of each site. Furthermore, by normalizing, it is possible to standardize the display screen. As a result, versatility is increased, the number of sites where the system can be applied increases, and the cost and time required to build the system can be reduced.

[0041] Fig. 6 is a diagram showing image data of water in an aquarium tank as the treated water 3, and the analysis results of each image data. In the example shown in Fig. 6, the image analysis system 10 classifies the analysis results of the image data into three levels: "clean", "slightly murky", and "murky".

[0042] 7 is a diagram showing an embodiment of the display of the Web application 19. The arithmetic device 17b of the cloud server 17 has a graph display function, an abnormality notification function, an abnormality condition setting function, etc. The cloud server 17 causes the Web application 19 to display a graph 25 showing the analysis result (inference result) of the abnormal state of the monitored object, image data 27 generated by the imaging device 9, etc.

[0043] As shown in Fig. 7, the analysis results are displayed on the Web application 19 by the graph display function as a graph 25 in which the horizontal axis represents the monitoring time of the monitored object and the vertical axis represents numbers (categorical variables) indicating the analysis results. In the example shown in Fig. 7, the analysis results for the past 24 hours are displayed in chronological order. In this embodiment, when the image analysis system 10 (edge ​​server 15 or cloud server 17) infers that the state of the monitored object is normal, "1" is displayed, and when the image analysis system 10 infers that the state of the monitored object is abnormal, "0" is displayed. The user can check not only the latest analysis result but also the progress of past analysis results on the graph 25.

[0044] Next, the abnormality notification function will be described. Cloud server 17 judges whether the state of the monitored object is normal or abnormal based on the analysis result according to the inference model. When cloud server 17 judges the state of the monitored object to be abnormal, it sends an email to user terminal 22 notifying that the state of the separated object is abnormal. An abnormality notification setting button 30 is displayed on web application 19, and the user can switch between starting and stopping the abnormality notification by operating abnormality notification setting button 30.

[0045] In one embodiment, as shown in Fig. 8, the cloud server 17 may be linked to a warning light 29 via the network 16. The warning light 29 may be connected to the cloud server 17 via the user terminal 22, or may be connected to the cloud server 17 without via the user terminal 22. In one embodiment, when the cloud server 17 determines that the state of the monitored object is abnormal, it notifies the user that the state of the monitored object is abnormal via the warning light 29. Specifically, when the cloud server 17 determines that the state of the monitored object is abnormal, it turns on the warning light 29.

[0046] In this embodiment, when an abnormality occurs, notification is given by email or warning light 29, so the user can become aware of the situation even when not looking at the screen of Web application 19. Furthermore, the user does not need to judge whether the condition of the monitored object is good or bad. In addition, by using Web application 19, one facility can be checked from multiple remote locations, and since one person does not need to monitor, the load of monitoring can be shared. As a result, the user's burden can be reduced and shared. The user does not need to monitor all the time, and the user's work time can be reduced.

[0047] In one embodiment, the edge server 15 or the cloud server 17 may be configured to calculate a confidence level according to the inference model. The confidence level is an index value indicating the validity of the inference result, and is expressed as a numerical value from 0 to 100. The confidence level may also be expressed as a numerical value from 0 to 1.

[0048] Next, the abnormal condition setting function will be described. Web application 19 is configured to be able to set abnormal conditions for cloud server 17 to determine that the state of the monitored object is abnormal. Specifically, the user sets the abnormal conditions by setting numbers in an abnormal condition setting window 32 displayed on Web application 19. Abnormal condition setting window 32 includes a moving average setting window 32a, an alarm level setting window 32b, and a safety level setting window 32c.

[0049] Examples of abnormal conditions include the following: When any of the following conditions is met, cloud server 17 determines that the state of the monitored object is abnormal. (i) When the image analysis system 10 infers that the condition of the monitored object is abnormal according to the inference model. (ii) The image analysis system 10 infers that the state of the monitored object is abnormal, and the degree of certainty is equal to or greater than a predetermined threshold value. (iii) When the categorical variable of the analysis result is classified into multiple stages, the categorical variable is above or below a predetermined threshold. (iv) When the moving average value over time of a categorical variable in the analysis results exceeds or falls below a predetermined threshold value. (v) When the waveform that represents the analysis results in a time series (for example, a graph that represents categorical variables in a time series; hereinafter, simply referred to as the analysis waveform) is in an abnormal state. Specifically, cloud server 17 stores the analysis waveforms from the past few days in storage device 17a, and calculates a reference waveform (ideal analysis waveform) from the analysis waveforms from the past few days. Cloud server 17 compares the current analysis waveform with the reference waveform, and if the degree of match falls below a predetermined set value, determines that the current state of the monitored object is abnormal.

[0050] In one embodiment, the abnormal condition is set by inputting a number in the alarm level setting window 32b. For example, if one wishes to set the above-mentioned (i) as the abnormal condition, one inputs "1" in a separately prepared setting file, and inputs the threshold value to be used for the judgment in the alarm level setting window 32b. If one wishes to set the above-mentioned (iii) as the abnormal condition, one inputs "3" in the setting file, and inputs the threshold value to be used for the judgment in the alarm level setting window 32b.

[0051] In one embodiment, the thresholds of (ii) to (iv) above are set by inputting a corresponding number in the safety level setting window 32c. In one embodiment, if the user wants to consider a stage where the warning level threshold is reached once and then recovered by 0.2 or more as safe, the user inputs "0.2" in the safety level setting window 32c.

[0052] In the above (iv), the time range of the moving average can be set in the moving average setting window 32a.

[0053] In this manner, in the present embodiment, since the abnormality conditions can be easily set, the present embodiment can be applied to many sites. Since the level at which an abnormality notification is desired varies depending on the site, the ability to set the level individually improves usability.

[0054] FIG. 9 is a diagram showing an example of a method for monitoring the separated material 5 as the dehydrated filtrate. In this embodiment, the dehydrated filtrate is imaged by an imaging device 9 and classified into multiple stages. The image analysis system 10 outputs an inference result of contamination to clarification based on the image data of the separated material 5. Since the state of the separated material 5 is affected by the amount of dehydrated polymer injected, it is possible to confirm whether the amount of polymer injected is appropriate, excessive, or insufficient based on the classification result of the separated material 5. By determining a desired state, the amount of polymer can be adjusted in accordance with fluctuations in the dehydrated filtrate. In one embodiment, the cloud server 17 may notify an abnormality if the amount of polymer is excessive or insufficient.

[0055] For example, if it is desired to maintain a "slightly clear" state, and the inference result for the separated material 5 becomes "polluted" due to a change in the state of the current sludge fed into the solid-liquid separation device 7, an alarm for a shortage of polymer is output (notifying an abnormality) and the amount of polymer injected is increased. If the filtrate is polluted, the color of the dehydrated filtrate will be dark black or brown and cloudy. If the filtrate is clear, it will be light brown or have a transparent appearance. This difference in appearance is learned by the inference model through machine learning and used for inference.

[0056] Fig. 10 is a schematic diagram showing another embodiment of the remote monitoring system 1. The configuration and operation of this embodiment that are not particularly described are the same as those of the embodiment described with reference to Figs. 1 to 9, so duplicated descriptions will be omitted. The remote monitoring system 1 of this embodiment further includes a sensor 34 that measures a physical quantity that indirectly indicates the state of the monitored object. The sensor 34 is electrically connected to the edge server 15. The physical quantity measured by the sensor 34 is sent to the edge server 15 and stored in the storage device 15a.

[0057] In this embodiment, the edge server 15 inputs the image data obtained from the imaging device 9 and the measurement data of the physical quantities into the inference model, and infers an abnormal state of the monitored object according to the algorithm of the inference model. The inference model of this embodiment is a trained model created by machine learning using training data including image data generated by the imaging device 9 during a past solid-liquid separation process, the measurement data of the physical quantities (explanatory variables), and inference result data as a correct label (objective variable) determined by an operator. The image analysis system 10 creates an inference model according to a machine learning algorithm using training data including past image data, past measurement data of the physical quantities, and corresponding inference result data. Furthermore, the edge server 15 acquires image data from the imaging device 9 and acquires measurement values ​​of the physical quantities from the sensor 34, and periodically or irregularly executes machine learning using training data including the image data, the measurement data of the physical quantities, and the corresponding inference result data to update the inference model. The embodiment described with reference to FIG. 3 can also be applied to this embodiment. In one embodiment, the edge server 15 transmits image data of the monitored object and measurement data of the above-mentioned physical quantities to the cloud server 17, and the cloud server 17 inputs the image data and measurement data of the above-mentioned physical quantities transmitted from the edge server 15 into an inference model, and may infer an abnormal state of the monitored object according to the algorithm of the inference model.

[0058] Examples of physical quantities indirectly indicating the state of the object to be monitored include the temperature of the object to be monitored and / or the solid-liquid separator 7, the humidity of the object to be monitored, the water quality of the water to be treated 3, the sound emitted from the solid-liquid separator 7, the vibration of the solid-liquid separator 7, and the smell of the object to be monitored. Examples of the sensor 34 include a temperature sensor, a humidity sensor, a water quality sensor, a sound sensor, a vibration sensor, and an odor sensor. In one embodiment, the remote monitoring system 1 may include a plurality of sensors 34. The remote monitoring system 1 may include any one of a temperature sensor, a humidity sensor, a water quality sensor, a sound sensor, a vibration sensor, and an odor sensor, or may include a plurality of sensors of a temperature sensor, a humidity sensor, a water quality sensor, a sound sensor, a vibration sensor, and an odor sensor. By further using the above physical quantities as training data, it is possible to more accurately infer an abnormal state.

[0059] According to the above-described embodiment, the image data of the separated material 5 or the water to be treated 3 and the analysis results of the image data using machine learning can be confirmed by the Web application 19. As a result, optimal remote monitoring of the state of the separated material 5 or the water to be treated 3 can be performed. Furthermore, according to the above-described embodiment, the image data and the analysis results can be confirmed from multiple locations using the Web application 19. Even with a small-scale facility, pinpoint AI analysis of images and remote monitoring of monitored objects are possible, making it suitable for water treatment.

[0060] The above-described embodiments have been described for the purpose of enabling a person having ordinary skill in the art to practice the present invention. Various modifications of the above-described embodiments are naturally possible for a person skilled in the art, and the technical idea of ​​the present invention can be applied to other embodiments. Therefore, the present invention is not limited to the described embodiments, but is to be interpreted in the broadest scope according to the technical idea defined by the claims. [Explanation of symbols]

[0061] 1. Remote monitoring system 3. Untreated water 5 Separate 7 Solid-liquid separator 9. Imaging device 10. Image Analysis System 12 Conveyor 15 Edge Server 16 Network 17 Cloud Server 19 Web Applications 22 User terminal 25 Graphs 27 Image data 29 Warning light 30 Abnormality notification setting button 32 Abnormal condition setting window 34 Sensors

Claims

1. An imaging device that captures an image of the water to be treated and generates image data; An image analysis system that analyzes an abnormal state of the water to be treated based on the image data, an edge server including an inference model constructed by machine learning using training data including at least the image data; A cloud server having a web application and causing the web application to display an analysis result of an abnormal state and image data; Equipped with A remote monitoring system characterized in that the image data is input into the inference model and an abnormal state of the treated water is inferred using the inference model.

2. A remote monitoring system for monitoring a state of water to be treated, comprising: An imaging device that captures an image of the water to be treated and generates image data; An image analysis system that analyzes an abnormal state of the water to be treated based on the image data, The image analysis system includes: an edge server electrically connected to the imaging device; A cloud server connected to the edge server via a network, the cloud server includes a Web application, and the cloud server is configured to cause the Web application to display an analysis result of the abnormal state and the image data; The image analysis system stores an inference model constructed by machine learning using training data including at least the image data, The image analysis system includes: The image data is input to the inference model, and an abnormal state of the water to be treated is inferred according to the inference model; The analysis results are classified into multiple stages, The remote monitoring system is configured to calculate and normalize the analysis result.

3. the cloud server is connected to a user terminal via the network; The remote monitoring system of claim 1 or 2, wherein the cloud server is configured to send an email to a user terminal notifying that the state of the treated water is abnormal when the cloud server determines that the state of the treated water is abnormal.

4. The cloud server is connected to the warning light via the network, The remote monitoring system of claim 1 or 2, wherein the cloud server is configured to notify via the warning light that the state of the treated water is abnormal when the cloud server determines that the state of the treated water is abnormal.

5. The remote monitoring system according to claim 3 or 4, wherein the web application is configured to be able to set an abnormality condition for the cloud server to determine that a state of the water to be treated is abnormal.

6. The system further includes a sensor for measuring a physical quantity indirectly indicating a state of the water to be treated, A remote monitoring system as described in any one of claims 1 to 5, wherein the image analysis system is configured to further input measurement data of the physical quantity into the inference model and infer an abnormal state of the treated water according to the inference model.

7. The remote monitoring system according to claim 6 , wherein the sensor is one or more of a temperature sensor, a humidity sensor, a water quality sensor, a sound sensor, a vibration sensor, and an odor sensor.

8. The inference model is stored on the edge server; A remote monitoring system as described in any one of claims 1 to 7, wherein the edge server is configured to input at least the image data into the inference model and infer an abnormal state of the treated water according to the inference model, and is configured to transmit the analysis results and the image data to the cloud server via the network.

9. The remote monitoring system according to claim 8 , wherein the edge server is configured to transmit the analysis results and the image data at different transmission frequencies.

10. the edge server is configured to transmit the image data to the cloud server via the network; The inference model is stored on the cloud server; A remote monitoring system as described in any one of claims 1 to 7, wherein the cloud server is configured to input at least the image data into the inference model and infer an abnormal state of the treated water according to the inference model.

11. A solid-liquid separation device that separates the water to be treated into a solid-liquid separation product; an imaging device that captures an image of the separated object and generates image data; an image analysis system that analyzes an abnormal state of the separated object based on the image data; an edge server including an inference model constructed by machine learning using training data including at least the image data; A cloud server having a web application and causing the web application to display an analysis result of an abnormal state and image data; Equipped with A remote monitoring system characterized in that the image data is input into the inference model, and an abnormal state of the separated object is inferred using the inference model.

Images