Abnormality prediction device, filtration processing system, abnormality prediction method and program

Abstract

To predict abnormality in a membrane module having one or a plurality of hollow fiber membranes.SOLUTION: An abnormality prediction device includes: a sound wave data acquisition part for acquiring sound wave data based on a sound wave from a membrane module; a storage part for storing the sound wave data; a feature value calculation part for calculating a predetermined feature amount from the sound wave data; an abnormality prediction part for predicting abnormality of the membrane module, on the basis of the predetermined feature amount in a first period, and the feature amount in a second period later than the first period; and an output part for outputting an abnormality prediction result of the abnormality prediction part.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an abnormality prediction device, a filtration processing system, an abnormality prediction method, and a program. [Background technology]

[0002] Patent Document 1 describes a "crack occurrence prediction method that accurately sets the frequency of a specific abnormal sound that occurs immediately before a workpiece cracks and detects the abnormal sound." [Prior art document] [Patent documents] [Patent Document 1] JP 11-221760 [Patent Document 2] JP 2020-134229 Summary of the Invention

[0003] In a first aspect of the present invention, there is provided an abnormality prediction device for predicting an abnormality in a membrane module having one or more hollow fiber membranes, comprising: an acoustic data acquisition unit for acquiring acoustic data based on acoustic waves from the membrane module; a memory unit for storing the acoustic data; a feature calculation unit for calculating predetermined feature values ​​from the acoustic data; an abnormality prediction unit for predicting an abnormality in the membrane module based on the feature values ​​for a predetermined first period and the feature values ​​for a second period after the first period; and an output unit for outputting the abnormality prediction result of the abnormality prediction unit.

[0004] In the above-mentioned abnormality prediction device, the acoustic data acquisition unit may include a raw data acquisition unit that acquires raw data obtained by sampling acoustic waves from the membrane module at a predetermined sampling rate, and an acoustic data generation unit that extracts and generates the acoustic data from the raw data at a predetermined period.

[0005] In any of the above abnormality prediction devices, the raw data acquisition unit may acquire the raw data at a sampling rate of 4 kHz or more and 400 kHz or less.

[0006] In any of the above abnormality prediction devices, the acoustic data generation unit may generate the acoustic data at a period of not less than 5 seconds and not more than 1 hour.

[0007] In any of the above abnormality prediction devices, the acoustic data generation unit may generate the acoustic data by extracting the maximum value of the raw data during the period.

[0008] In any of the above abnormality prediction devices, the acoustic data generation unit may generate the acoustic data by extracting an average value of the raw data during the period.

[0009] In any of the above abnormality prediction devices, the raw data acquisition unit may be provided outside the membrane module.

[0010] In any of the above abnormality prediction devices, the acoustic data acquisition unit may acquire the acoustic data while a filtration treatment system using the membrane module is in operation.

[0011] In any of the above abnormality prediction devices, the acoustic data acquisition unit may acquire the acoustic data when the membrane module is being cleaned.

[0012] In any of the abnormality prediction devices described above, each of the first period and the second period may include a time for cleaning the membrane module.

[0013] In any of the above abnormality prediction devices, the feature calculation unit may calculate, as the feature, a moving average of the sound wave data over a predetermined time interval.

[0014] In any of the above-mentioned abnormality prediction devices, the abnormality prediction unit may predict an abnormality in the membrane module when the ratio of the characteristic amount in the second period to the maximum value of the characteristic amount in the first period exceeds a predetermined threshold.

[0015] In any of the above-mentioned abnormality prediction devices, the abnormality prediction unit may predict an abnormality in the membrane module when the ratio of the characteristic amount in the second period to the average value of the characteristic amount in the first period exceeds a predetermined threshold.

[0016] In any of the above-described anomaly prediction devices, the anomaly prediction unit may predict an anomaly in the membrane module by inputting the feature amount in the second time period into a machine learning model in which an explanatory variable is the feature amount in the first time period and an objective variable is the degree of anomaly of the membrane module. The explanatory variables may include operating conditions of a filtration treatment device, an operating state of the filtration treatment device, types of materials used in the membrane module, types of the membrane module, and properties of the raw liquid supplied to the membrane module.

[0017] In any of the above anomaly prediction devices, the machine learning model may be ocSVM or tradGAN. However, the type of the machine learning model is not particularly limited and may be a learning model using support vectors, MetricLearning, clustering, a neural network that learns data representations, a generative model, or a neural network architecture that uses an attention mechanism. The learning model using support vectors is not particularly limited and may be ocSVM. MetricLearning is not particularly limited and may be TripletLoss. The clustering is not particularly limited and may be k-means clustering or Mahalanobis distance clustering. The neural network that learns data representations is not particularly limited and may be AE. The generative model is not particularly limited and may be GAN, VAE, or tradGAN. The neural network architecture that uses an attention mechanism is not particularly limited and may be Transformer, AnomalyTransformer, or LLM.

[0018] In a second aspect of the present invention, there is provided a filtration treatment system including any one of the abnormality prediction devices described above and a filtration treatment device having the membrane module.

[0019] In the filtration treatment system, the membrane module may have a potting section for fixing the one or more hollow fiber membranes, and the abnormality prediction section may predict an abnormality in the potting section.

[0020] In any of the above filtration treatment systems, the potting material of the potting portion may include at least one of urethane or epoxy.

[0021] Any of the above filtration treatment systems may include a control device that controls the operation of the filtration treatment device. The abnormality prediction device may predict an abnormality in the membrane module while the filtration treatment device is being controlled by the control device.

[0022] In a third aspect of the present invention, there is provided an abnormality prediction method for predicting an abnormality in a membrane module having one or more hollow fiber membranes, comprising the steps of acquiring acoustic data based on acoustic waves from the membrane module, storing the acoustic data, calculating predetermined features from the acoustic data, predicting an abnormality in the membrane module based on the features for a predetermined first period and the features for a second period after the first period, and outputting the abnormality prediction result.

[0023] In a fourth aspect of the present invention, there is provided a program that, when executed by a computer, causes the computer to function as an abnormality prediction device that predicts abnormalities in a membrane module having one or more hollow fiber membranes, the abnormality prediction device comprising: an acoustic data acquisition unit that acquires acoustic data based on acoustic waves from the membrane module; a memory unit that stores the acoustic data; a feature calculation unit that calculates predetermined feature values ​​from the acoustic data; an abnormality prediction unit that predicts abnormalities in the membrane module based on the feature values ​​for a predetermined first period and the feature values ​​for a second period that is later than the first period; and an output unit that outputs the abnormality prediction results of the abnormality prediction unit.

[0024] The above summary of the invention does not list all of the features of the present invention, and subcombinations of these features may also be inventions. [Brief explanation of the drawings]

[0025] [Figure 1] 1 shows an example of the configuration of a filtration processing system 10. [Figure 2] An example of the configuration of the abnormality prediction device 300 is shown together with the filtration processing device 100 and the control device 200. [Figure 3] An example of the configuration of the membrane module 150 is shown together with the acoustic data acquisition unit 310. [Figure 4A] 10 shows an example of raw data acquired by the raw data acquisition unit 312. [Figure 4B] 10 shows an example of an enlarged view of raw data acquired by the raw data acquisition unit 312. [Figure 5A] 10 shows an example of sound wave data generated by the sound wave data generating unit 314. [Figure 5B] 10 shows an example of sound wave data generated by the sound wave data generating unit 314. [Figure 6] 10 shows an example of the feature amount calculated by the feature amount calculation unit 330. [Figure 7] An example of data used for abnormality prediction by the abnormality prediction unit 340 is shown below. [Figure 8]1 shows an example of a flowchart of an abnormality prediction method. [Figure 9] A modified example of the configuration of the abnormality prediction device 300 is shown together with the filtration processing device 100 and the control device 200. [Figure 10A] An example of the degree of anomaly output by ocSVM is shown below. [Figure 10B] An example of the degree of anomaly output by tradGAN is shown below. [Figure 11] 1 illustrates an example computer 1000 in which aspects of the present invention may be embodied in whole or in part. DETAILED DESCRIPTION OF THE INVENTION

[0026] The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the scope of the invention according to the claims. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0027] FIG. 1 shows an example of the configuration of a filtration processing system 10. Note that the blocks shown are functionally separated functional blocks and do not necessarily correspond to the actual device configuration. That is, a block shown as one block in this diagram does not necessarily have to be configured by one device. Also, blocks shown as separate blocks in this diagram do not necessarily have to be configured by separate devices.

[0028] The filtration processing system 10 may include a raw liquid tank 20, a filtrate tank 30, a concentrated liquid tank 40, a cleaning liquid tank 60, and an air tank 70. The filtration processing system 10 may include a filtration processing device 100, a control device 200, and an abnormality prediction device 300.

[0029] The filtration system 10 may be a system for producing drinking water by filtering out turbidity and / or parasites contained in the water to be treated (e.g., river water), a system for concentrating valuable substances in the liquid to be treated (e.g., soy sauce), or a system for removing impurities from the liquid to be treated (e.g., wine), but is not limited to these.

[0030] The raw solution tank 20 stores the raw solution 22. The raw solution 22 may be seawater, brine, river water, secondary sewage treatment water, RO concentrated water, groundwater, industrial wastewater, soy sauce, alcoholic beverages such as beer or wine, a liquid containing fungi, starch sugar, antibodies, proteins, vaccines, or enzymes, an inorganic colloid slurry, an organic emulsion, paint, pharmaceutical water, or ultrapure water. The raw solution 22 may be temperature-controlled. The raw solution 22 may be heated or cooled, and the temperature range of the raw solution 22 may be any value. The temperature of the raw solution 22 may be controlled to any temperature during operation based on the filtration cost calculated from the properties of the raw solution, filtration efficiency, the properties of the required filtrate 32, the equipment or energy required for temperature control, operating efficiency, etc.

[0031] The filtrate tank 30 stores the filtrate 32 recovered by the filtration device 100 filtering the raw liquid 22. The filtration process performed by the filtration device 100 will be described later.

[0032] The concentrated liquid tank 40 stores the concentrated liquid 42 that remains unfiltered by the filtration processing device 100. Note that the filtration processing system 10 does not necessarily have to include the concentrated liquid tank 40. In this case, the concentrated liquid 42 that remains unfiltered by the filtration processing device 100 may be returned to the stock solution tank 20. That is, the concentrated liquid tank 40 and the stock solution tank 20 may refer to the same thing. Alternatively, the filtration processing system 10 may include both the stock solution tank 20 and the concentrated liquid tank 40, with a portion of the concentrated liquid 42 stored in the concentrated liquid tank 40 and the remaining concentrated liquid 42 returned to the stock solution tank 20.

[0033] The cleaning liquid tank 60 stores a cleaning liquid 62 used for cleaning. For example, the cleaning liquid 62 may be water adjusted to a desired temperature or an aqueous solution containing sodium hypochlorite. However, the type of cleaning liquid contained in the cleaning liquid 62 is not limited to these.

[0034] The temperature of the raw solution 22 and / or the cleaning solution 62 may be adjusted by any equipment. Methods for heating the tank include, but are not limited to, a steam jacket, an electric heater, a heater coil, or a combination of a heat exchanger and a circulation pump. Steam or a liquid at any temperature may be directly introduced into the tank. The temperature may also be adjusted by a heat exchanger installed in the piping.

[0035] The air tank 70 stores air for air cleaning. Air cleaning will be described later. The filtration treatment system 10 does not necessarily have to include the air tank 70, and air cleaning does not necessarily have to be performed.

[0036] In this example, the filtration system 10 is shown with one each of the undiluted solution tank 20, concentrated solution tank 40, filtrate tank 30, cleaning solution tank 60, and air tank 70, but the number of each tank and the ratio of the number of each tank that the filtration system 10 has are not limited to this, and the filtration system 10 does not necessarily have to have these facilities. As an example, the filtration system 10 may have multiple cleaning solution tanks 60.

[0037] The filtration processing device 100 may filter the stock solution 22. The filtration processing device 100 may separate the filtrate 32 recovered by filtering the stock solution 22 from the concentrated solution 42 that remains unfiltered. The filtration processing device 100 may supply the filtrate 32 to the filtrate tank 30 and supply the concentrated solution 42 to the concentrated solution tank 40. However, the filtration processing device 100 may supply a portion of the concentrated solution 42 to the concentrated solution tank 40 and return the remaining concentrated solution 42 to the stock solution tank 20, or may return all of the concentrated solution 42 to the stock solution tank 20.

[0038] The filtration processing device 100 includes a membrane module 150. The filtration processing device 100 may include a pump 110, a temperature measurement unit 112, a pressure measurement unit 114, and a flow rate measurement unit 116. The filtration processing device 100 may include a pipe 120 and a valve 122.

[0039] In this example, one membrane module 150 is illustrated as the membrane module 150 included in the filtration processing device 100, but the number of membrane modules 150 included in the filtration processing device 100 is not limited to this. The filtration processing device 100 may have multiple membrane modules 150. As an example, the filtration processing device 100 may have 16 membrane modules 150 per set of pumps. Furthermore, the filtration processing device 100 does not necessarily have all of the pumps, valves, measuring units, and piping, and may have multiple components of each depending on the purpose. The ratio of the numbers of pumps, valves, measuring units, and membrane modules included in the filtration processing device 100 is also not limited to the example illustrated.

[0040] The membrane module 150 has hollow fiber membranes for filtering the raw liquid 22. Details of the membrane module 150 will be described later.

[0041] The membrane module 150 may be connected to the raw liquid tank 20 via piping 120. The membrane module 150 may be connected to the filtrate tank 30 via piping 120. The membrane module 150 may be connected to the concentrate tank 40 via piping 120. However, if the filtration treatment system 10 does not include the concentrate tank 40, the membrane module 150 may be connected to the raw liquid tank 20 via piping 120.

[0042] The pump 110 may be a driving source for supplying each liquid from each tank to the membrane module 150. The temperature measurement unit 112 may measure the liquid temperature. The pressure measurement unit 114 may measure the pressure in the pipe 120. The flow rate measurement unit 116 may measure the flow rate in the pipe 120. The valve 122 may open and close each pipe 120.

[0043] The compressor 174 may be a driving source for supplying air from the air tank 70 to the membrane module 150 via the pipe 120. The compressor 174 may supply air to the membrane module 150 during air cleaning.

[0044] The location of each measuring unit is not limited to the example shown in the figure. Each measuring unit may be located at any position where measurement is desired. For example, a temperature measuring unit 112 may be provided in each of the stock solution tank 20, the filtrate tank 30, and the concentrate tank 40. The liquid temperatures of the stock solution 22, the filtrate 32, and the concentrate 42 may be assumed to be almost the same.

[0045] The control device 200 may control the operation of the filtration processing device 100. For example, the control device 200 controls the opening and closing of the valve 122 and the pump pressure of the pump 110. The control device 200 may notify the abnormality prediction device 300 that the control device 200 is currently controlling the filtration processing device 100.

[0046] The abnormality prediction device 300 predicts an abnormality in the membrane module 150. Details of the abnormality prediction device 300 and the abnormality prediction method using the abnormality prediction device 300 will be described later. The abnormality prediction device 300 may predict an abnormality in the membrane module 150 while the filtration treatment device 100 is being controlled by the control device 200.

[0047] The abnormality prediction device 300 of this example predicts an abnormality in the membrane module 150 while the filtration treatment device 100 is being controlled by the control device 200. This eliminates the need to perform a separate test to check for an abnormality in the membrane module 150. Furthermore, when a separate test is performed to check for an abnormality in the membrane module 150, the abnormality may only be detected after an abnormality has occurred in the membrane module 150. On the other hand, the abnormality prediction device 300 of this example predicts an abnormality in the membrane module 150 and can prevent an abnormality in the membrane module 150 from occurring.

[0048] FIG. 2 shows an example of the configuration of an anomaly prediction device 300 together with a filtration processing device 100 and a control device 200. The anomaly prediction device 300 includes an acoustic data acquisition unit 310, a memory unit 320, a feature calculation unit 330, an anomaly prediction unit 340, and an output unit 350. Note that the blocks shown are functionally separated functional blocks and do not necessarily correspond to the actual device configuration. In other words, blocks shown as one block in this diagram do not necessarily have to be configured by one device. Furthermore, blocks shown as separate blocks in this diagram do not necessarily have to be configured by separate devices.

[0049] The acoustic data acquisition unit 310 acquires acoustic data based on acoustic waves from the membrane module 150. The acoustic data acquisition unit 310 may supply the acquired acoustic data to the storage unit 320.

[0050] The acoustic data acquisition unit 310 may acquire acoustic data during operation of the filtration treatment system 10 using the membrane module 150. This eliminates the need to perform a separate test to check for abnormalities in the membrane module 150.

[0051] The ultrasonic data acquisition unit 310 may include a raw data acquisition unit 312 and an ultrasonic data generation unit 314 .

[0052] The raw data acquisition unit 312 may acquire raw data by sampling sound waves from the membrane module 150 at a predetermined sampling rate. The raw data acquisition unit 312 may acquire raw data at a sampling rate of 4 kHz or more and 400 kHz or less. For example, the raw data acquisition unit 312 acquires raw data at a sampling rate of 48 kHz, 96 kHz, or 192 kHz. However, the sampling rate at which the raw data acquisition unit 312 acquires raw data is not limited to these. The sampling rate at which the raw data acquisition unit 312 acquires raw data may be determined based on the accuracy required for abnormality prediction, the type of membrane module 150, etc.

[0053] The sonic data generating unit 314 may extract and generate sonic data from the raw data at a predetermined cycle. The sonic data generating unit 314 may generate sonic data at a cycle of 5 seconds or more and 1 hour or less. For example, the sonic data generating unit 314 generates sonic data at a cycle of 10 seconds, 20 seconds, or 60 seconds. The cycle at which the sonic data generating unit 314 generates sonic data may be determined based on the accuracy required for abnormality prediction, the storage capacity of the storage unit 320, etc.

[0054] The sonic data generation unit 314 may generate sonic data by extracting the maximum value of the raw data over a predetermined period. As another example, the sonic data generation unit 314 may generate sonic data by extracting the average value of the raw data over a predetermined period. That is, the sonic data acquisition unit 310 may acquire sonic data that has been subjected to so-called edge processing on the raw data acquired by the raw data acquisition unit 312. However, the sonic data acquisition unit 310 may also acquire raw data that has not been subjected to edge processing as sonic data.

[0055] In this example, the ultrasonic data acquisition unit 310 has been described as having the raw data acquisition unit 312 and the ultrasonic data generation unit 314, and thus performing edge processing on the raw data, but the entity that performs edge processing on the raw data is not limited to this. For example, the ultrasonic data acquisition unit 310 may acquire ultrasonic data that has been edge processed outside the ultrasonic data acquisition unit 310.

[0056] The ultrasonic data acquiring section 310 may supply the ultrasonic data that has been subjected to edge processing to the storage section 320. However, the ultrasonic data acquiring section 310 may also supply the raw data acquired by the raw data acquiring section 312 to the storage section 320 as ultrasonic data.

[0057] The storage unit 320 stores the sound wave data. That is, the storage unit 320 may store sound wave data that has undergone edge processing.

[0058] The storage unit 320 of this example stores the sonic data acquired by the sonic data acquisition unit 310, i.e., the edge-processed sonic data. This allows the amount of data to be reduced compared to storing raw data. For example, if the sampling rate at which the raw data acquisition unit 312 acquires raw data is 192 kHz and the sonic data generation unit 314 generates sonic data every 10 seconds, the amount of data to be stored can be reduced to 1 / 1,920,000 times the amount of data stored when storing raw data. However, the storage unit 320 may also store raw data as sonic data.

[0059] The feature amount calculation unit 330 calculates a predetermined feature amount from the acoustic data. The feature amount calculation unit 330 may calculate a moving average of the acoustic data over a predetermined time interval as the feature amount. However, the feature amount calculated by the feature amount calculation unit 330 is not limited to this. The feature amount calculation unit 330 may also calculate a maximum value of the acoustic data over a predetermined time interval as the feature amount. The feature amount calculation unit 330 may also calculate a moving average and / or maximum value of the acoustic data over a time interval of 30 minutes or more and one week or less as the feature amount. The time interval is not particularly limited, but may be 30 minutes, I (I is a natural number from 1 to 23), one day, or two days. Here, I is a natural number from 1 to 23, but may also be other values. The time interval at which the feature amount calculation unit 330 calculates the feature amount may be set depending on the type of membrane module 150, the operating conditions of the filtration processing device 100, etc.

[0060] The feature amount calculation unit 330 may calculate multiple feature amounts. As an example, the feature amount calculation unit 330 may calculate multiple types of feature amounts. For example, the feature amount calculation unit 330 calculates a moving average and a maximum value of sound wave data over a predetermined time interval. As another example, the feature amount calculation unit 330 may calculate feature amounts over multiple predetermined time intervals. For example, the feature amount calculation unit 330 calculates a feature amount over a first time interval (e.g., one day) and a feature amount over a second time interval (e.g., five days). The feature amount calculation unit 330 may calculate feature amounts over three or more time intervals.

[0061] The feature calculation unit 330 in this example calculates predetermined feature amounts from the sound wave data. That is, the feature calculation unit 330 may further calculate a moving average and / or a maximum value for the sound wave data from which the average value or maximum value has been extracted for the raw data. However, the feature calculation unit 330 may also calculate predetermined feature amounts from the raw data as sound wave data.

[0062] The abnormality prediction unit 340 predicts an abnormality in the membrane module 150 based on the characteristic values ​​for a predetermined first period and the characteristic values ​​for a second period that follows the first period. The predetermined first period may be a period during which it is estimated that no abnormality has occurred in the membrane module 150. For example, the predetermined first period may be a predetermined period after the filtration processing system 10 starts operating, or a predetermined period after the membrane module 150 is replaced with a new one. The predetermined second period may be a predetermined period during which an abnormality in the membrane module 150 is predicted. That is, the abnormality prediction unit 340 may predict an abnormality in the membrane module 150 by comparing the characteristic values ​​for the first period during which it is estimated that no abnormality has occurred in the membrane module 150 with the characteristic values ​​for the second period during which an abnormality in the membrane module 150 is predicted.

[0063] For example, the abnormality prediction unit 340 predicts an abnormality in the membrane module 150 when the ratio of the feature amount in the second time period to the maximum value of the feature amount in the first time period exceeds a predetermined threshold. The predetermined threshold may be 1.2 or greater and 5.0 or less. One example of the predetermined threshold is 1.3. If the predetermined threshold is smaller than 1.2, an abnormality in the membrane module 150 may be predicted even in a situation where no abnormality in the membrane module 150 occurs. Conversely, if the predetermined threshold is larger than 5.0, an abnormality in the membrane module 150 may occur before the abnormality in the membrane module 150 is predicted, making it impossible to prevent the abnormality in the membrane module 150 from occurring. The threshold may be set depending on the type or structure of the membrane module 150, the operating conditions of the filtration processing device 100, the type of abnormality to be predicted, etc.

[0064] As another example, the anomaly prediction unit 340 predicts an anomaly in the membrane module 150 when the ratio of the feature amount in the second time period to the average value of the feature amount in the first time period exceeds a predetermined threshold. The predetermined threshold may be 1.2 or greater and 5.0 or less. One example of the predetermined threshold is 1.3. If the predetermined threshold is smaller than 1.2, an anomaly in the membrane module 150 may be predicted even in a situation where no anomaly in the membrane module 150 occurs. Conversely, if the predetermined threshold is greater than 5.0, an anomaly in the membrane module 150 may occur before the anomaly in the membrane module 150 is predicted, making it impossible to prevent the anomaly in the membrane module 150 from occurring. The threshold may be set depending on the type of membrane module 150, the operating conditions of the filtration processing device 100, the type of anomaly to be predicted, or the like.

[0065] The abnormality prediction unit 340 may compare the feature quantity for the second time period with both the maximum value and average value of the feature quantity for the first time period. For example, the abnormality prediction unit 340 may predict an abnormality in the membrane module 150 when the ratio of the feature quantity for the second time period to the maximum value of the feature quantity for the first time period exceeds a predetermined threshold and the ratio of the feature quantity for the second time period to the average value of the feature quantity for the first time period exceeds a predetermined threshold. Alternatively, the abnormality prediction unit 340 may predict an abnormality in the membrane module 150 when the ratio of the feature quantity for the second time period to the maximum value of the feature quantity for the first time period exceeds a predetermined threshold, or when the ratio of the feature quantity for the second time period to the average value of the feature quantity for the first time period exceeds a predetermined threshold.

[0066] The same applies to the case where the feature calculation unit 330 calculates feature amounts at multiple predetermined time intervals. That is, the abnormality prediction unit 340 may compare the feature amounts for the second period with the feature amounts for the first period for both the feature amounts based on the first time interval and the feature amounts based on the second time interval, and may predict an abnormality in the membrane module 150 based on the comparison result.

[0067] The anomaly prediction unit 340 of this example compares the feature quantity of the second period with the maximum value and / or average value of the feature quantity of the first period. The feature quantity is calculated by calculating the maximum value and / or moving average of the acoustic wave data. The acoustic wave data is calculated by extracting the maximum value or average value from the raw data. Therefore, the data used by the anomaly prediction unit 340 of this example for anomaly prediction may be the raw data that has been coarse-grained up to three times.

[0068] The abnormality prediction unit 340 of this example predicts an abnormality in the membrane module 150 based on the characteristic amount for a predetermined first period and the characteristic amount for a second period that follows the first period. This eliminates the need to perform a separate test to check for an abnormality in the membrane module 150. Furthermore, the abnormality prediction unit 340 of this example can predict an abnormality in the membrane module 150 and prevent the occurrence of an abnormality in the membrane module 150, compared to when a separate test is performed to check for an abnormality in the membrane module 150.

[0069] Furthermore, the abnormality prediction unit 340 of this example predicts an abnormality in the membrane module 150 using data that is coarse-grained raw data. As a result, the memory unit 320 of this example only needs to store edge-processed sound wave data rather than storing raw data, which allows for a reduction in the amount of data stored compared to when raw data is stored to predict an abnormality in the membrane module 150 using Fourier transform, etc. Furthermore, the abnormality prediction unit 340 of this example uses less data for analysis compared to when an abnormality in the membrane module 150 is predicted using Fourier transform, etc., and therefore allows for improved analysis efficiency.

[0070] The output unit 350 outputs the abnormality prediction result of the abnormality prediction unit 340. The output unit 350 may output the abnormality prediction result to the control device 200. The output unit 350 may notify the user of the abnormality prediction result via a display device such as a monitor, may notify the user via an audio device such as a speaker, or may notify the user via a sending device such as email.

[0071] The control device 200 may control the filtration processing device 100 based on the abnormality prediction result output from the output unit 350. For example, if the abnormality prediction result output from the output unit 350 predicts that an abnormality will occur in the membrane module 150, the control device 200 may stop control of the filtration processing device 100 to replace or repair the membrane module 150, or may change the control conditions of the filtration processing device 100 to delay the occurrence of the abnormality in the membrane module 150. If the abnormality prediction result output from the output unit 350 does not predict that an abnormality will occur in the membrane module 150, the control device 200 may continue control of the filtration processing device 100 under the current control conditions. The control device 200 may automatically stop, change, or continue control of the filtration processing device 100 in response to the output of the abnormality prediction result from the output unit 350, or may be performed by a user who receives a notification from the output unit 350.

[0072] In this example, the abnormality prediction unit 340 predicts an abnormality in the membrane module 150, and the output unit 350 outputs the abnormality prediction result of the abnormality prediction unit 340. As a result, when an abnormality is predicted to occur in the membrane module 150, the occurrence of an abnormality in the membrane module 150 can be prevented by replacing or repairing the membrane module 150, or by changing the control conditions of the filtration treatment device 100.

[0073] 3 shows an example of the configuration of the membrane module 150 together with the acoustic data acquisition unit 310. The membrane module 150 has one or more hollow fiber membranes 1510. The membrane module 150 may have a case 1500 and a potting unit 1520.

[0074] The case 1500 may be provided with an opening 1502, an opening 1504, an opening 1506, and an opening 1508. The membrane module 150 may be connected to the piping 120 through the opening 1502, may be connected to the piping 120 through the opening 1504, may be connected to the piping 120 through the opening 1506, or may be connected to the piping 120 through the opening 1508. That is, the feed solution 22 may flow into the membrane module 150 through the opening 1502, the filtrate 32 may flow into and out of the membrane module 150 through the openings 1506 and 1508, and the concentrate 42 may flow out of the membrane module 150 through the opening 1504.

[0075] The hollow fiber membrane 1510 may function as a filter for filtering the raw solution 22. The hollow fiber membrane 1510 may be a porous membrane with a predetermined pore size. The hollow fiber membrane 1510 may be produced by a thermally induced phase separation (TIPS) method or a non-solvent induced phase separation (NIPS) method. However, the method for producing the hollow fiber membrane 1510 is not limited to these.

[0076] The potting portion 1520 may fix one or more hollow fiber membranes 1510. The potting material of the potting portion 1520 may include a thermosetting resin. The potting material of the potting portion 1520 may include at least one of urethane or epoxy.

[0077] The raw liquid 22 that flows into the membrane module 150 through the opening 1502 flows into the area where the hollow fiber membrane 1510 is provided, and may be filtered by passing through the hollow fiber membrane 1510 from the inside to the outside of the hollow fiber membrane 1510.

[0078] The abnormality prediction unit 340 may predict an abnormality in the potting unit 1520. The abnormality prediction unit 340 may predict a crack in the potting unit 1520. The potting unit 1520 may crack due to the cleaning process of the membrane module 150. For example, the potting unit 1520 may crack due to a temperature change caused by hot water cleaning, which is included in the cleaning process. The abnormality prediction unit 340 may predict an abnormality in the membrane module 150 caused by the cleaning process of the membrane module 150.

[0079] The acoustic data acquisition unit 310 may acquire acoustic data during cleaning of the membrane module 150. The first and second periods for the abnormality prediction unit 340 to predict an abnormality may each include the time of cleaning of the membrane module 150. This allows the abnormality prediction unit 340 to predict an abnormality in the membrane module 150 based on acoustic data from a period when an abnormality in the membrane module 150 is likely to occur, thereby enabling the abnormality in the membrane module 150 to be accurately predicted.

[0080] The ultrasonic data acquiring unit 310 may be provided outside the membrane module 150. This makes it possible to predict abnormalities in the membrane module 150 using ultrasonic data acquired by the externally provided ultrasonic data acquiring unit 310, without changing the internal design of the membrane module 150.

[0081] The acoustic data acquisition unit 310 may be provided at a predetermined position outside the case 1500 of the membrane module 150. For example, the acoustic data acquisition unit 310 is provided at a position outside the case 1500 corresponding to the potting part 1520. This makes it possible to acquire acoustic data generated at a position where an abnormality in the membrane module 150 is likely to occur, and to accurately predict an abnormality in the membrane module 150.

[0082] In this example, the ultrasonic data acquisition unit 310 is provided outside the membrane module 150, but the manner in which the ultrasonic data acquisition unit 310 is provided is not limited to this. That is, in this example, the raw data acquisition unit 312 and the ultrasonic data generation unit 314 included in the ultrasonic data acquisition unit 310 are provided outside the membrane module 150, but only the raw data acquisition unit 312 may be provided outside the membrane module 150, and the ultrasonic data generation unit 314 may be provided in another position. That is, the raw data acquisition unit 312 may be provided outside the membrane module 150, and the ultrasonic data generation unit 314 does not have to be provided adjacent to the outside of the membrane module 150.

[0083] 4A shows an example of raw data acquired by the raw data acquisition unit 312. The horizontal axis represents time, and the vertical axis represents the amplitude of the raw data. In this example, the raw data acquisition unit 312 acquires the raw data at a sampling rate of 192 kHz.

[0084] 4B shows an example of an enlarged view of raw data acquired by the raw data acquisition unit 312. The acoustic data generation unit 314 may extract and generate acoustic data from the raw data at a predetermined cycle. The cycle shown in the figure is an example of the cycle at which the acoustic data generation unit 314 generates acoustic data.

[0085] For example, when the sonic data generation unit 314 extracts the maximum value of raw data over a predetermined period to generate sonic data, the sonic data generation unit 314 extracts amplitude A1 as the maximum value of the raw data over the first period. Similarly, the sonic data generation unit 314 extracts amplitude A2 as the maximum value of the raw data over the second period, and amplitude A3 as the maximum value of the raw data over the third period. Note that A2 indicates the absolute value of the amplitude of the raw data, and A2≧0. In this manner, the sonic data generation unit 314 may generate sonic data by extracting the maximum value of the raw data. The same applies when the sonic data generation unit 314 extracts the average value of the raw data to generate sonic data.

[0086] 5A shows an example of sound wave data generated by the sound wave data generation unit 314. The horizontal axis represents time, and the vertical axis represents the maximum amplitude value extracted from the raw data. In this example, the sound wave data generation unit 314 generates sound wave data by extracting the maximum value of the raw data over a predetermined period. In other words, the graph shown in FIG. 5A is sound wave data generated by extracting the maximum value from the raw data of FIG. 4A.

[0087] 5A, the data used for abnormality prediction may include data from a period when the operation of the filtration processing device 100 is temporarily suspended. The raw data acquisition unit 312 may acquire raw data from the period when the operation of the filtration processing device 100 is temporarily suspended, and the sound wave data generation unit 314 may generate sound wave data by extracting a maximum value of the raw data from the period when the operation of the filtration processing device 100 is temporarily suspended. However, after the raw data acquisition unit 312 acquires the raw data from the period when the operation of the filtration processing device 100 is temporarily suspended, the raw data from the period when the operation of the filtration processing device 100 is temporarily suspended may be deleted. Alternatively, after the sound wave data generation unit 314 extracts a maximum value of the raw data from the period when the operation of the filtration processing device 100 is temporarily suspended to generate sound wave data, the sound wave data from the period when the operation of the filtration processing device 100 is temporarily suspended may be deleted.

[0088] FIG. 5B shows an example of sonic data generated by the sonic data generation unit 314. The horizontal axis represents time, and the vertical axis represents the average amplitude extracted from the raw data. In this example, the sonic data generation unit 314 generates sonic data by extracting the average value of raw data over a predetermined period. That is, the graph shown in FIG. 5B is sonic data generated by extracting the average value from the raw data of FIG. 4A. The sonic data generation unit 314 may generate sonic data by extracting the average value of raw data during a period when the operation of the filtration processing device 100 is temporarily stopped. However, after the sonic data generation unit 314 extracts the average value of raw data during a period when the operation of the filtration processing device 100 is temporarily stopped and generates sonic data, the sonic data during the period when the operation of the filtration processing device 100 is temporarily stopped may be deleted.

[0089] FIG. 6 shows an example of a feature calculated by the feature calculation unit 330. The horizontal axis represents time, and the vertical axis represents the moving average of the maximum amplitude values ​​shown in FIG. 5A. In this example, the feature calculation unit 330 calculates a moving average of sound wave data over a predetermined time interval as a feature. In this example, the feature calculation unit 330 calculates a 13-hour moving average of the sound wave data shown in FIG. 5A. However, as described with reference to FIG. 2, various variations can be considered in the feature and time interval calculated by the feature calculation unit 330. The feature calculation unit 330 may calculate a moving average of sound wave data over a time interval that includes a period during which the operation of the filtration processing device 100 is temporarily stopped. However, after the feature calculation unit 330 calculates the moving average of sound wave data over a time interval that includes a period during which the operation of the filtration processing device 100 is temporarily stopped, the feature for the period during which the operation of the filtration processing device 100 is temporarily stopped may be deleted.

[0090] The abnormality prediction unit 340 predicts an abnormality in the membrane module 150 based on the characteristic amounts of a predetermined first period and the characteristic amounts of a second period that follows the first period. In this example, the first period is two days after the start of operation of the filtration treatment system 10, and the second period is a period that follows the first period. However, the start time and length of the first period are not limited to these.

[0091] FIG. 7 shows an example of data used for abnormality prediction by the abnormality prediction unit 340. In this example, the abnormality prediction unit 340 predicts an abnormality in the membrane module 150 when the ratio of the feature amount for the second period to the average value of the feature amount for the first period exceeds a predetermined threshold. The threshold is 1.3, for example. However, as explained in relation to FIG. 2, various variations can be considered in the abnormality prediction method by the abnormality prediction unit 340. The abnormality prediction unit 340 may use, as the feature amount for the second period, feature amounts for a period that includes a period during which the operation of the filtration processing device 100 is temporarily stopped. However, the abnormality prediction unit 340 may use, as the feature amount for the second period, feature amounts from which feature amounts for the period during which the operation of the filtration processing device 100 is temporarily stopped are deleted.

[0092] In the example of FIG. 7, at time T0, the ratio of the feature amount for the second period to the average value of the feature amount for the first period exceeds a predetermined threshold of 1.3. Therefore, the abnormality prediction unit 340 may predict that an abnormality will occur in the membrane module 150 at time T0. In reality, an abnormality in the membrane module 150 occurred between time T1 and time T2. Note that in this example, data up to the occurrence of an abnormality in the membrane module 150 is shown to confirm the effectiveness of the abnormality prediction device 300. However, in practical use, the occurrence of an abnormality in the membrane module 150 can be prevented by replacing or repairing the membrane module 150 or changing the control conditions of the filtration treatment device 100 when the occurrence of an abnormality is predicted at time T0.

[0093] 8 shows an example of a flowchart of an anomaly prediction method. In step S100, acoustic data based on acoustic waves from the membrane module 150 is acquired. For example, an acoustic data acquisition unit 310 acquires acoustic data based on acoustic waves from the membrane module 150. The acoustic data acquisition unit 310 may acquire raw data by sampling acoustic waves from the membrane module 150 at a predetermined sampling rate using a raw data acquisition unit 312, and may generate acoustic data by extracting it from the raw data at a predetermined cycle using an acoustic data generation unit 314.

[0094] In step S110, the sound wave data is stored. For example, the sound wave data is stored in the storage unit 320. The storage unit 320 may store sound wave data that has undergone edge processing.

[0095] In step S120, a predetermined feature amount is calculated from the sound wave data. For example, the feature amount calculation unit 330 calculates the predetermined feature amount from the sound wave data. The feature amount calculation unit 330 may calculate a moving average and / or a maximum value of the sound wave data over a predetermined time interval as the feature amount.

[0096] Steps S100 and S110 may be repeated before step S120. That is, time-series data of the acoustic data may be accumulated by repeating step S100 of acquiring acoustic data based on acoustic waves from the membrane module 150 and step S110 of storing the acoustic data. Step S120 may be executed when acoustic data for a predetermined time interval has been accumulated, and a feature value based on the predetermined time interval may be calculated.

[0097] Note that, after sonic data for a predetermined time interval has been accumulated, step S120 may be executed for each cycle of steps S100 and S110. As an example, it is assumed that the sonic data for a predetermined time interval is N pieces of data. In this case, step S120 may be executed for the first time after N pieces of data have been accumulated. For example, step S120 may not be executed while the first to (N-1)th pieces of data are being accumulated, and step S120 is executed when the Nth piece of data has been accumulated, and feature amounts based on the first to Nth pieces of data are calculated. Thereafter, step S120 may not be executed until N pieces of data from (N+1)th to 2Nth have been accumulated. In other words, step S120 may be executed when the (N+1)th piece of data has been accumulated, and feature amounts based on the second to (N+1)th pieces of data are calculated. In this way, the number of times steps S100 and S110 are repeated and the timing of execution of step S120 for each repetition may be appropriately set based on the sampling rate of the raw data acquisition unit 312, the period at which the acoustic data generation unit 314 extracts and generates acoustic data and / or the time interval for calculating the feature amount, etc.

[0098] As described above, the raw data, the acoustic data, the feature values, and / or the data used for abnormality prediction may include data from a period when the operation of the filtration processing device 100 is temporarily suspended. The data from a period when the operation of the filtration processing device 100 is temporarily suspended may be deleted in any of steps S100, S110, and S120, or may not be deleted in any of the steps. When the data from a period when the operation of the filtration processing device 100 is temporarily suspended is deleted in step S120, the data may be deleted before the feature values ​​are calculated, or the data may be deleted after the feature values ​​are calculated.

[0099] In step S130, an abnormality in the membrane module 150 is predicted based on the characteristic amount for a predetermined first period and the characteristic amount for a second period that follows the first period. For example, the abnormality prediction unit 340 predicts an abnormality in the membrane module 150 based on the characteristic amount for the predetermined first period and the characteristic amount for the second period that follows the first period.

[0100] Steps S100 to S130 may be repeatedly executed. For example, they may be repeatedly executed when it is not predicted in step S130 that an abnormality will occur in the membrane module 150, and the process may proceed to step S140 when it is predicted in step S130 that an abnormality will occur in the membrane module 150.

[0101] In step S140, the abnormality prediction result is output. For example, the output unit 350 outputs the abnormality prediction result of the abnormality prediction unit 340.

[0102] In the above, it has been explained that steps S100 to S130 may be repeatedly executed when step S130 does not predict that an abnormality will occur in the membrane module 150, but steps S100 to S140 may also be repeatedly executed regardless of the result of the abnormality prediction. That is, step S140 may be executed every time step S130 is executed, not just when it is predicted that an abnormality will occur in the membrane module 150, and the abnormality prediction result indicating whether or not an abnormality is predicted to occur in the membrane module 150 may be output.

[0103] 9 shows a modified example of the configuration of an anomaly prediction device 300 together with a filtration processing device 100 and a control device 200. The anomaly prediction device 300 of this example differs from the embodiment of FIG. 2 in that the anomaly prediction unit 340 uses a machine learning model 342. In this example, differences from the embodiment of FIG. 2 will be particularly described, and the rest may be the same as the embodiment of FIG. 2.

[0104] The anomaly prediction unit 340 may predict an anomaly in the membrane module 150 by inputting feature values ​​from the second period into a machine learning model 342, whose explanatory variables are feature values ​​from the first period and whose objective variable is the anomaly level of the membrane module 150. That is, the machine learning model 342 may be a machine learning model trained based on feature values ​​from the first period. For example, the machine learning model 342 is an ocSVM or a tradGAN. For example, the machine learning model is not particularly limited and may be a learning model using support vectors, MetricLearning, clustering, a neural network that learns data representations, a generative model, or a neural network architecture that uses an attention mechanism. The learning model using support vectors is not particularly limited and may be an ocSVM. The MetricLearning is not particularly limited and may be TripletLoss. The clustering is not particularly limited and may be k-means clustering or Mahalanobis distance clustering. The neural network that learns data representations is not particularly limited and may be AE. The generative model is not particularly limited and may be a GAN, a VAE, or a tradGAN. The neural network architecture using the attention mechanism is not particularly limited and may be a Transformer, an Anomaly Transformer, or an LLM. However, the type of machine learning model 342 is not limited to these. The explanatory variables may include the operating conditions of the filtration treatment device 100, the operating state of the filtration treatment device 100, the type of components used in the membrane module 150, the type of membrane module 150, and the liquid properties of the raw liquid 22 supplied to the membrane module 150.

[0105] The anomaly prediction unit 340 may input the feature amounts for the second period to the machine learning model 342. The machine learning model 342 may output an abnormality level of the membrane module 150 according to the input feature amounts for the second period and supply it to the anomaly prediction unit 340. The anomaly prediction unit 340 may predict an abnormality of the membrane module 150 based on the abnormality level output by the machine learning model 342. For example, the anomaly prediction unit 340 may predict that an abnormality will occur in the membrane module 150 if the slope of the moving average of the abnormality level is greater than a predetermined threshold. As another example, the anomaly prediction unit 340 may predict that an abnormality will occur in the membrane module 150 if the abnormality level is greater than a predetermined threshold.

[0106] The machine learning model 342 may be generated outside the anomaly prediction device 300 and stored in an external server or the like. However, the generation location and storage location of the machine learning model 342 are not limited to this. The machine learning model 342 may be generated by the feature calculation unit 330 and stored in the feature calculation unit 330, or may be generated by the feature calculation unit 330 and stored in the anomaly prediction unit 340, or may be generated by the feature calculation unit 330 and stored in an external server or the like. Any other combination of generation location and storage location of the machine learning model 342 may be adopted.

[0107] The machine learning model 342 may use multiple feature quantities as explanatory variables. As described in relation to FIG. 2, the feature quantity calculation unit 330 may calculate multiple feature quantities. For example, the feature quantity calculation unit 330 may calculate multiple types of feature quantities, and may calculate feature quantities for multiple predetermined time intervals. In this case, the machine learning model 342 may be trained using the multiple feature quantities for a first period as explanatory variables. The abnormality prediction unit 340 may predict an abnormality in the membrane module 150 by inputting the multiple feature quantities for a second period into the machine learning model 342.

[0108] FIG. 10A shows an example of the abnormality level output by ocSVM. In this example, the abnormality prediction unit 340 predicts that an abnormality will occur in the membrane module 150 when the slope of the moving average of the abnormality level is greater than a predetermined threshold. The abnormality prediction unit 340 may predict that an abnormality will occur in the membrane module 150 when the slope of the moving average of the abnormality level becomes greater than a predetermined threshold at time T0. Note that there is a time before time T0 when the value of the abnormality level is approximately the same as the value of the abnormality level at time T0, but at that time the slope of the moving average of the abnormality level does not exceed the threshold, so it is not necessary to predict that an abnormality will occur in the membrane module 150.

[0109] In this example, an abnormality occurred in the membrane module 150 between time T1 and time T2. In order to confirm the effectiveness of the abnormality prediction device 300, data up until the occurrence of an abnormality in the membrane module 150 is shown, but in practical use, the occurrence of an abnormality in the membrane module 150 can be prevented by replacing or repairing the membrane module 150 or changing the control conditions of the filtration treatment device 100 when the occurrence of an abnormality is predicted at time T0.

[0110] In this example, the data used for abnormality prediction includes data from a period when the operation of the filtration processing device 100 is temporarily stopped, but the data from a period when the operation of the filtration processing device 100 is temporarily stopped may be deleted at any stage before the abnormality is predicted. For example, the data from a period when the operation of the filtration processing device 100 is temporarily stopped may be deleted in any of steps S100, S110, or S120 described above.

[0111] FIG. 10B shows an example of the anomaly level output by tradGAN. In this example, the anomaly prediction unit 340 predicts that an anomaly will occur in the membrane module 150 when the slope of the moving average of the anomaly level is greater than a predetermined threshold. The anomaly prediction unit 340 may predict that an anomaly will occur in the membrane module 150 when the slope of the moving average of the anomaly level becomes greater than a predetermined threshold at time T0. Note that there is a time before time T0 when the anomaly level value is approximately the same as the anomaly level value at time T0, but at that time, the slope of the moving average of the anomaly level does not exceed the threshold, so it is not necessary to predict that an anomaly will occur in the membrane module 150.

[0112] In this example, an abnormality occurred in the membrane module 150 between time T1 and time T2. In order to confirm the effectiveness of the abnormality prediction device 300, data up until the occurrence of an abnormality in the membrane module 150 is shown, but in practical use, the occurrence of an abnormality in the membrane module 150 can be prevented by replacing or repairing the membrane module 150 or changing the control conditions of the filtration treatment device 100 when the occurrence of an abnormality is predicted at time T0.

[0113] In this example, the data used for abnormality prediction includes data from a period when the operation of the filtration processing device 100 is temporarily stopped, but the data from a period when the operation of the filtration processing device 100 is temporarily stopped may be deleted at any stage before the abnormality is predicted. For example, the data from a period when the operation of the filtration processing device 100 is temporarily stopped may be deleted in any of steps S100, S110, or S120 described above.

[0114] Various embodiments of the present invention may be described with reference to flowcharts and block diagrams, where the blocks may represent (1) stages of a process in which operations are performed or (2) sections of an apparatus responsible for performing the operations. Particular stages and sections may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable medium, and / or a processor provided with computer-readable instructions stored on a computer-readable medium. Dedicated circuitry may include digital and / or analog hardware circuitry, and may include integrated circuits (ICs) and / or discrete circuits. Programmable circuitry may include reconfigurable hardware circuitry, including logical AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, memory elements such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and the like.

[0115] A computer-readable medium may include any tangible device capable of storing instructions that are executed by an appropriate device, such that the computer-readable medium having instructions stored thereon comprises an article of manufacture containing instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable media may include electronic, magnetic, optical, electromagnetic, and semiconductor storage media. More specific examples of computer-readable media may include floppy disks, diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), compact disc read-only memory (CD-ROM), digital versatile disc (DVD), Blu-ray (RTM) disc, memory stick, integrated circuit card, and the like.

[0116] The computer readable instructions may include either assembler instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0117] The computer-readable instructions may be provided to a processor or programmable circuit of a programmable data processing device, such as a computer, locally or over a wide area network (WAN) such as a local area network (LAN) or the Internet, and the computer-readable instructions may be executed to create means for performing the operations specified in the flowcharts or block diagrams. Here, the computer may be a personal computer (PC), a tablet computer, a smartphone, a workstation, a server computer, a general-purpose computer, a special-purpose computer, or the like, or may be a computer system in which multiple computers are connected. Such a computer system in which multiple computers are connected is also called a distributed computing system, and is a broad definition of computer. In a distributed computing system, the multiple computers collectively execute a program by each executing a portion of the program and passing data between the computers as needed during program execution.

[0118] Examples of processors include a computer processor, a central processing unit (CPU), a processing unit, a microprocessor, a digital signal processor, a controller, a microcontroller, etc. A computer may have one or more processors. In a multiprocessor system with multiple processors, each processor executes a portion of a program and passes data between processors as needed during program execution, allowing the multiple processors to collectively execute a program. For example, in multitasking, each of the multiple processors may execute a portion of each task in small chunks by switching tasks at time slice intervals. In this case, which portion of a program each processor executes changes dynamically. Which portion of a program each of the multiple processors executes may also be statically determined by multiprocessor-aware programming.

[0119] 11 illustrates an example of a computer 1000 in which aspects of the present invention may be embodied, in whole or in part. A program installed on the computer 1000 may cause the computer 1000 to function as or perform operations associated with an apparatus or one or more sections of the apparatus according to an embodiment of the present invention, and / or to perform a process or steps of the process according to an embodiment of the present invention. Such a program may be executed by the CPU 1012 to cause the computer 1000 to perform specific operations associated with some or all of the blocks of the flowcharts and block diagrams described herein.

[0120] A computer 1000 according to this embodiment includes a CPU 1012, a RAM 1014, a graphics controller 1016, and a display device 1018, which are interconnected by a host controller 1010. The computer 1000 also includes input / output units such as a communication interface 1022, a hard disk drive 1024, a DVD-ROM drive 1026, and an IC card drive, which are connected to the host controller 1010 via an input / output controller 1020. The computer also includes legacy input / output units such as a ROM 1030 and a keyboard 1042, which are connected to the input / output controller 1020 via an input / output chip 1040.

[0121] The CPU 1012 operates according to programs stored in the ROM 1030 and RAM 1014, thereby controlling each unit. The graphics controller 1016 acquires image data generated by the CPU 1012 into a frame buffer or the like provided in the RAM 1014 or into the graphics controller itself, and causes the image data to be displayed on the display device 1018.

[0122] The communication interface 1022 communicates with other electronic devices via a network. The hard disk drive 1024 stores programs and data used by the CPU 1012 in the computer 1000. The DVD-ROM drive 1026 reads programs or data from a DVD-ROM 1027 and provides the programs or data to the hard disk drive 1024 via the RAM 1014. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0123] The ROM 1030 stores therein a boot program and the like that is executed by the computer 1000 upon activation, and / or programs that depend on the hardware of the computer 1000. The input / output chip 1040 may also connect various input / output units to the input / output controller 1020 via a parallel port, a serial port, a keyboard port, a mouse port, and the like.

[0124] The programs are provided by a computer-readable medium such as a DVD-ROM 1027 or an IC card. The programs are read from the computer-readable medium, installed in the hard disk drive 1024, RAM 1014, or ROM 1030, which are also examples of computer-readable media, and executed by the CPU 1012. Information processing described in these programs is read by the computer 1000, and causes cooperation between the programs and the various types of hardware resources described above. An apparatus or method may be configured by implementing information manipulation or processing in accordance with the use of the computer 1000.

[0125] For example, when communication is performed between the computer 1000 and an external device, the CPU 1012 may execute a communication program loaded into the RAM 1014 and instruct the communication interface 1022 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 1012, the communication interface 1022 reads transmission data stored in a transmission buffer processing area provided in the RAM 1014, the hard disk drive 1024, the DVD-ROM 1027, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes reception data received from the network to a reception buffer processing area or the like provided on the recording medium.

[0126] The CPU 1012 may also cause all or a necessary portion of a file or database stored on an external recording medium such as a hard disk drive 1024, a DVD-ROM drive 1026 (DVD-ROM 1027), an IC card, etc. to be read into the RAM 1014, and perform various types of processing on the data on the RAM 1014. The CPU 1012 then writes back the processed data to the external recording medium.

[0127] Various types of information, such as various types of programs, data, tables, and databases, may be stored on the recording medium and may undergo information processing. The CPU 1012 may perform various types of processing on data read from the RAM 1014, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequences of the programs, and write the results back to the RAM 1014. The CPU 1012 may also search for information in a file, database, etc. on the recording medium. For example, if multiple entries, each having an attribute value of a first attribute associated with an attribute value of a second attribute, are stored on the recording medium, the CPU 1012 may search for an entry that matches a condition specified by the attribute value of the first attribute from among the multiple entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0128] The above-described programs or software modules may be stored in a computer-readable medium on or near the computer 1000. A recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can also be used as a computer-readable medium, thereby providing the programs to the computer 1000 via the network.

[0129] Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be apparent to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the claims that such modifications and improvements can also be included within the technical scope of the present invention.

[0130] It should be noted that the execution order of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before," "prior to," etc., and that the processes can be performed in any order unless the output of a previous process is used in a subsequent process. Even if the operational flow in the claims, specifications, and drawings is described using "first," "next," etc. for convenience, this does not mean that the processes must be performed in this order. [Explanation of symbols]

[0131] 10 Filtration treatment system, 20 Raw liquid tank, 22 Raw liquid, 30 Filtrate tank, 32 Filtrate, 40 Concentrated liquid tank, 42 ​​Concentrated liquid, 60 Cleaning liquid tank, 62 Cleaning liquid, 70 Air tank, 100 Filtration treatment device, 110 Pump, 112 Temperature measurement unit, 114 Pressure measurement unit, 116 Flow rate measurement unit, 120 Piping, 122 Valve, 150 Membrane module, 174 Compressor, 200 Control device, 300 Anomaly prediction device, 310 Sound wave data acquisition unit, 312 Raw data acquisition unit, 314 Sound wave data generation unit, 320 Memory unit, 330 Feature calculation unit, 340 Anomaly prediction unit, 342 Machine learning model, 350 Output unit, 1000 Computer, 1010 Host controller, 1012 CPU, 1014 RAM, 1016 Graphics controller, 1018 display device, 1020 input / output controller, 1022 communication interface, 1024 hard disk drive, 1026 DVD-ROM drive, 1027 DVD-ROM, 1030 ROM, 1040 input / output chip, 1042 keyboard, 1500 case, 1502 opening, 1504 opening, 1506 opening, 1508 opening, 1510 hollow fiber membrane, 1520 potting part

Claims

1. An abnormality prediction device for predicting an abnormality in a membrane module having one or more hollow fiber membranes, an acoustic data acquisition unit that acquires acoustic data based on acoustic waves from the membrane module; a storage unit that stores the sound wave data; a feature calculation unit that calculates a predetermined feature from the sound wave data; an abnormality prediction unit that predicts an abnormality in the membrane module based on the characteristic amount in a predetermined first period and the characteristic amount in a second period that is later than the first period; an output unit that outputs the abnormality prediction result of the abnormality prediction unit; Equipped with Anomaly prediction device.

2. The ultrasonic data acquisition unit a raw data acquisition unit that acquires raw data by sampling the sound waves from the membrane module at a predetermined sampling rate; a sonic data generating unit that extracts and generates the sonic data from the raw data at a predetermined cycle; have The abnormality prediction device according to claim 1 .

3. The raw data acquisition unit acquires the raw data at a sampling rate of 4 kHz or more and 400 kHz or less. The abnormality prediction device according to claim 2.

4. The sonic data generating unit generates the sonic data at a period of 5 seconds or more and 1 hour or less. The abnormality prediction device according to claim 2.

5. The sonic data generating unit extracts the maximum value of the raw data during the period to generate the sonic data. The abnormality prediction device according to claim 2.

6. The sonic data generating unit extracts an average value of the raw data during the period to generate the sonic data. The abnormality prediction device according to claim 2.

7. The raw data acquisition unit is provided outside the membrane module. The abnormality prediction device according to claim 2.

8. The sonic data acquisition unit acquires the sonic data during operation of the filtration treatment system using the membrane module. The abnormality prediction device according to claim 1 .

9. The sonic data acquisition unit acquires the sonic data when the membrane module is cleaned. The abnormality prediction device according to claim 8.

10. Each of the first period and the second period includes a time for cleaning the membrane module. The abnormality prediction device according to claim 1 .

11. The feature amount calculation unit calculates a moving average of the sound wave data over a predetermined time interval as the feature amount. The abnormality prediction device according to claim 1 .

12. The abnormality prediction unit predicts an abnormality in the membrane module when a ratio of the characteristic amount in the second period to a maximum value of the characteristic amount in the first period exceeds a predetermined threshold. The abnormality prediction device according to claim 1 .

13. The abnormality prediction unit predicts an abnormality in the membrane module when a ratio of the characteristic amount in the second period to an average value of the characteristic amount in the first period exceeds a predetermined threshold. The abnormality prediction device according to claim 1 .

14. The abnormality prediction unit predicts an abnormality in the membrane module by inputting the feature amount in the second time period into a machine learning model in which an explanatory variable is the feature amount in the first time period and an objective variable is the degree of abnormality of the membrane module. The abnormality prediction device according to claim 1 .

15. The machine learning model is ocSVM or tradGAN The abnormality prediction device according to claim 14.

16. The abnormality prediction device according to any one of claims 1 to 15, a filtration treatment device having the membrane module; Equipped with Filtration treatment system.

17. the membrane module has a potting portion for fixing the one or more hollow fiber membranes, The abnormality prediction unit predicts an abnormality in the potting unit.

17. The filtration system of claim 16.

18. The potting material of the potting portion includes at least one of urethane and epoxy.

18. The filtration treatment system of claim 17.

19. A control device for controlling the operation of the filtration processing device is provided, The abnormality prediction device predicts an abnormality in the membrane module while the filtration treatment device is being controlled by the control device.

17. The filtration system of claim 16.

20. 1. A method for predicting an abnormality in a membrane module having one or more hollow fiber membranes, comprising: acquiring acoustic data based on acoustic waves from the membrane module; storing the acoustic wave data; calculating a predetermined feature amount from the sound wave data; predicting an abnormality in the membrane module based on the characteristic amount during a predetermined first period and the characteristic amount during a second period after the first period; A step of outputting an abnormality prediction result; Equipped with Anomaly prediction methods.

21. When executed by a computer, the computer An abnormality prediction device for predicting an abnormality in a membrane module having one or more hollow fiber membranes, an acoustic data acquisition unit that acquires acoustic data based on acoustic waves from the membrane module; a storage unit that stores the sound wave data; a feature calculation unit that calculates a predetermined feature from the sound wave data; an abnormality prediction unit that predicts an abnormality in the membrane module based on the characteristic amount in a predetermined first period and the characteristic amount in a second period that is later than the first period; an output unit that outputs the abnormality prediction result of the abnormality prediction unit; It functions as an abnormality prediction device equipped with program.

Images