Method for detecting abnormalities in a submerged membrane separation device

Abstract

To provide an abnormality detection method of an immersion type membrane separation device where the load of a watchman is reduced and the condition of a filtration membrane can be quickly and exactly diagnosed even when the watchman is absent.SOLUTION: Abnormality detection processing is performed based on a transmembrane pressure difference value where different dispersion suppression processing is performed based on the dispersion characteristic of transmembrane pressure difference preliminarily obtained by the preliminary operation of a normal immersion type membrane separation device when abnormality detection is detected based on transmembrane pressure difference obtained from the immersion type membrane separation device alternately repeating filtration operation and relaxation operation and when performing abnormality detection processing based on the transmembrane pressure difference value obtained from the actually operated immersion type membrane separation device.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present invention relates to a method for detecting an abnormality in a submerged membrane separation device. [Background technology]

[0002] In various water treatment plants, such as those for wastewater treatment, a submerged membrane separation device is placed in a treatment tank, and the water to be treated in the treatment tank is suction-filtered using a filtration membrane incorporated in the submerged membrane separation device, thereby obtaining treated water that has been separated into solid and liquid.

[0003] In the submerged membrane separation device, an aeration device is installed below the membrane separation device, and a filtration operation is performed in which suction filtration is performed using the filtration membrane while aeration is performed by the aeration device. However, as time passes, clogging and accumulation of solid components occur in the filtration membrane, and the filtration efficiency decreases. Therefore, the device is equipped with a management device that operates and manages the operation so that the filtration operation and the relaxation operation, in which suction filtration is stopped while aeration from the aeration device is maintained, are repeated at predetermined time intervals, and the filtration membrane surface is cleaned by the upward flow of bubbles from the aeration device and the gas-liquid mixed phase of the water to be treated, just as in the filtration operation.

[0004] The management device is configured to measure the pressure difference between the front side (primary side) and the back side (secondary side) of the filtration membrane, i.e., the transmembrane pressure difference, and determine that backpressure cleaning or chemical cleaning is necessary when the transmembrane pressure exceeds a management threshold value or increases rapidly.

[0005] For example, Patent Document 1 proposes a filtering member cleaning system that can perform back pressure cleaning on the filtering member at an effective timing while suppressing the occurrence of clogging of the filtering member in a filtering treatment device.

[0006] The filtration member cleaning system has a filtration member that passes water to be treated from a primary side to a secondary side to filter the water, and a primary side area and a secondary side area partitioned by the filtration member, and is equipped with a filtration treatment device that filters the water to be treated by passing the water to be treated from the primary side area to the secondary side area, a treated water distribution means that distributes the water to be treated from the primary side area to the secondary side area, a cleaning liquid supplying means that supplies cleaning liquid to the filtration member from the secondary side to the primary side, a pressure difference measuring means that measures the pressure difference between the primary side area and the secondary side area, a determination means that is capable of setting a predetermined threshold value and determines whether to start the cleaning liquid supplying means based on the set threshold value, and a control means that controls the start of the cleaning liquid supplying means based on the determination result by the determination means, and the determination means is configured to reset the threshold value after the supply of cleaning liquid by the cleaning liquid supplying means has been completed to a value obtained by adding a predetermined offset value to a post-cleaning pressure difference value calculated based on a measured pressure difference value, which is the pressure difference measured by the pressure difference measuring means.

[0007] In recent years, from the viewpoint of improving the efficiency of managing filtration membranes, remote monitoring systems are being constructed that include a management device that is installed in a water treatment plant and manages a membrane separation device so as to switch between filtration operation and relaxation operation at predetermined time intervals, and a remote monitoring device that is communicatively connected to the management device and collects and manages the transmembrane pressure differences of each filtration membrane sampled by each management device at predetermined time intervals.

[0008] Based on the time-series transmembrane pressure difference of each membrane separation device collected by the remote monitoring device, a trend graph of the transmembrane pressure difference is displayed on a monitor, and multiple monitors visually check the trend graph to determine whether or not an abnormality has occurred in the separation membrane of each membrane separation device.

[0009] However, as the number of membrane separation devices to be monitored increased, the processing load on the monitors also increased, and the time allowed for the monitors to visually inspect each trend graph to determine whether something was abnormal or normal was limited, making it difficult to make accurate and prompt judgments.

[0010] Therefore, it is possible to increase the number of supervisors, but training many supervisors so that they can stably make the same judgments takes a lot of time, and there is a problem that the judgment results inevitably fluctuate due to individual differences. Since the operating conditions of the membrane separation device vary from facility to facility, and the trends of the trend graph of the transmembrane pressure displayed on the display device also vary accordingly, sufficient experience is required to be able to properly judge whether something is normal or abnormal in a short time.

[0011] There was also the inconvenience that abnormalities could go unnoticed at night or on holidays when the supervisor was not present. [Prior art documents] [Patent documents]

[0012] [Patent Document 1] JP 2011-31145 A Summary of the Invention [Problem to be solved by the invention]

[0013] In view of the above-mentioned problems, the applicant has proposed a method for detecting abnormalities in a submerged membrane separation device that reduces the burden on monitors and enables quick and accurate diagnosis of the condition of the filtration membrane even when a monitor is not present (Patent Application No. 2020-213492).

[0014] The method for detecting an anomaly in the submerged membrane separation device includes a data collection step of measuring and collecting the transmembrane pressure at predetermined time intervals for a submerged membrane separation device that alternately repeats a filtration operation and a relaxation operation; a state separation step of separating a group of transmembrane pressure measurement values ​​for a first period including a target period for detecting an anomaly from the time-series transmembrane pressure measurement values ​​collected in the data collection step into a first group of transmembrane pressure measurement values ​​during the filtration operation and a second group of transmembrane pressure measurement values ​​during the relaxation operation; and an anomaly determination step of calculating a dispersion state for the first group of transmembrane pressure measurement values ​​and / or the second group of transmembrane pressure measurement values ​​from a group of transmembrane pressure measurement values ​​for a second period shorter than the first period, and determining whether or not the submerged membrane separation device is abnormal based on the dispersion state, while shifting the process in time series.

[0015] However, when the time-series transmembrane pressure measurement values ​​collected in the data collection process exhibit large fluctuation characteristics, at least one of the first transmembrane pressure measurement value group (G1) and the second transmembrane pressure measurement value group (G2) separated into states in the state separation process may be erroneously determined to be abnormal even though it is actually normal.

[0016] For example, a false abnormality may be detected when the measured transmembrane pressure varies due to switching the suction volume of the suction pump used to extract permeate from the filtration membrane, or when the measured transmembrane pressure during filtration operation fluctuates due to fluctuations in the water level in the treatment tank in which the filtration membrane is immersed.

[0017] In view of the above-mentioned problems, an object of the present invention is to provide an abnormality detection method for a submerged membrane separation device that reduces the burden on the monitor and enables the condition of the filtration membrane to be quickly and accurately diagnosed even when the monitor is absent. [Means for solving the problem]

[0018] In order to achieve the above-mentioned object, a first characteristic configuration of the anomaly detection method for a submerged type membrane separation apparatus according to the present invention includes a data collection step of measuring and collecting a transmembrane pressure at predetermined time intervals for a submerged type membrane separation apparatus that alternately repeats a filtration operation and a relaxation operation, a state separation step of separating a group of transmembrane pressure measurement values ​​for a first period (P1) including a target period for detecting an anomaly from among the time-series transmembrane pressure measurement values ​​collected in the data collection step into a first transmembrane pressure measurement value group (G1) during the filtration operation and a second transmembrane pressure measurement value group (G2) during the relaxation operation, and a state separation step of calculating a dispersion state for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2) from a transmembrane pressure measurement value group (G11, G21) during a second period (P2) that is shorter than the first period (P1), and calculating a state of the submerged type membrane separation apparatus based on the dispersion state. and an abnormality determination step of repeatedly determining whether the apparatus is abnormal, with the process being shifted in time series. The method for detecting an abnormality in a submerged membrane separation apparatus includes a preliminary data collection step of performing preliminary operation of the submerged membrane separation apparatus in a normal state to measure and collect a transmembrane pressure at predetermined time intervals, and a preliminary evaluation step of calculating a first preliminary evaluation value (SV1) indicating a fluctuation characteristic of the transmembrane pressure based on the time-series transmembrane pressure measurement values ​​in the normal state collected in the preliminary data collection step, prior to the data collection step. When the first preliminary evaluation value (SV1) is greater than a predetermined first preliminary evaluation threshold (SV1th), the state separation step is performed on a group of transmembrane pressure measurement values ​​obtained by normalizing a group of transmembrane pressure measurement values ​​for the first period (P1) based on the transmembrane pressure measurement values ​​collected in the preliminary data collection step.

[0019] When the immersion type membrane separation device to be subjected to abnormality determination is operating in a normal state, a preliminary data collection process is executed, and based on the transmembrane differential pressure obtained at that time, a preliminary evaluation process is executed, and a first preliminary evaluation value (SV1) indicating the fluctuation characteristics of the transmembrane differential pressure is calculated. When the first preliminary evaluation value (SV1) is greater than the first preliminary evaluation threshold value (SV1th), in order to avoid misjudgment due to the influence of the fluctuation characteristics, the group of transmembrane differential pressure measurement values in the first period (P1) is normalized based on the transmembrane differential pressure measurement values collected in the preliminary data collection process, and a state separation process is executed on the normalized group of transmembrane differential pressure measurement values. Therefore, the influence of the variation in the transmembrane differential pressure collected in the data collection process can be suppressed, and appropriate abnormality determination becomes possible.

[0020] In addition to the first characteristic configuration described above, the second characteristic configuration is that the preliminary evaluation process separates the time-series transmembrane differential pressure measurement values in the normal state collected in the preliminary data collection process into three preliminary transmembrane differential pressure measurement value groups (SG1, SG2, SG3) by performing non-hierarchical cluster analysis based on machine learning without a teacher, and among the centroid positions CG1, CG2, CG3 (CG1 < CG2 < CG3) of each preliminary transmembrane differential pressure measurement value group (SG1, SG2, SG3), the ratio obtained by (CG3 - CG2) / (CG2 - CG1) is calculated as the first preliminary evaluation value (SV1).

[0021] The time-series transmembrane differential pressure measurement values in the normal state collected in the preliminary data collection process are separated into three preliminary transmembrane differential pressure measurement value groups (SG1, SG2, SG3). The preliminary transmembrane differential pressure measurement value groups (SG1, SG2, SG3) include at least the filtration operation time and the relaxation operation time.

[0022] Among the centroid positions CG1, CG2, and CG3 (CG1 < CG2 < CG3) of the differential pressure measurement value groups (SG1, SG2, SG3) between the respective preliminary membranes, the ratio of the centroid distances obtained by (CG3 - CG2) / (CG2 - CG1) is calculated as the first preliminary evaluation value (SV1), so that the degree of variation specific to the immersion type membrane separation device for which preliminary data has been collected can be objectively evaluated. When such a preliminary evaluation value (SV1) is greater than a predetermined first preliminary evaluation threshold value (SV1th), it is determined that the variation is large, and the differential pressure measurement value group during the first period (P1) is normalized.

[0023] In addition to the second characteristic configuration described above, the third characteristic configuration is that in the preliminary evaluation step, when the first preliminary evaluation value (SV1) is greater than a predetermined second preliminary evaluation threshold value (SV2th) which is a value smaller than the first preliminary evaluation threshold value (SV1th), for the time-series differential pressure measurement values between the membranes in the normal state collected in the preliminary data collection step, further perform non-hierarchical cluster analysis based on machine learning without a teacher to separate them into two preliminary differential pressure measurement value groups (SG4, SG5) between the membranes, calculate the number of data of the differential pressure measurement values between the membranes in each preliminary differential pressure measurement value group (SG4, SG5) as the second preliminary evaluation value (SV2), and when any of the second preliminary evaluation values (SV2) is less than a predetermined threshold data number (CVth), perform the state separation step on the differential pressure measurement value group obtained by normalizing the differential pressure measurement value group during the first period (P1).

[0024] When the first preliminary evaluation value (SV1) is greater than a predetermined second preliminary evaluation threshold value (SV2th) which is a value smaller than the first preliminary evaluation threshold value (SV1th), that is, conceptually, when it is determined that the variation is moderate, the second preliminary evaluation value (SV2) is calculated. That is, the differential pressure measurement values collected in the preliminary data collection step are separated into two preliminary differential pressure measurement value groups (SG4, SG5) during the filtration operation and the relaxation operation, and the number of data of the differential pressure measurement values between the membranes in each preliminary differential pressure measurement value group (SG4, SG5) is calculated as the second preliminary evaluation value (SV2). Then, if any of the second preliminary evaluation values ​​(SV2) is less than a predetermined threshold number of data (CVth), it is determined that there is a possibility of large variation, and a state separation process is performed on the transmembrane pressure measurement value group normalized from the transmembrane pressure measurement value group for the first period (P1).

[0025] The fourth characteristic configuration is, in addition to the third characteristic configuration described above, that when both of the second preliminary evaluation values ​​(SV2) are equal to or greater than the threshold data number (CVth), the abnormality determination process calculates an average value of the group of transmembrane pressure measurement values ​​for the second period (P2) as an average transmembrane pressure measurement value for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2), calculates a dispersion state for the second period from the group of average transmembrane pressure measurement values ​​including each average transmembrane pressure measurement value calculated in the second period (P2) before the shift, and determines whether or not the submerged membrane separation device is abnormal based on the dispersion state.

[0026] If both of the second preliminary evaluation values ​​(SV2) are equal to or greater than the threshold number of data (CVth), it is determined that there is a sufficient number of data and that the variation is not so great that normalization processing is necessary, and the variation is calculated based on the average transmembrane pressure measurement value, which is the average value of the transmembrane pressure values ​​included in the second period (P2).

[0027] The fifth characteristic configuration, in addition to the third or fourth characteristic configuration described above, is that when the first preliminary evaluation value (SV1) is evaluated in the preliminary evaluation process to be smaller than a predetermined second preliminary evaluation threshold (SV2th), which is a value smaller than the first preliminary evaluation threshold (SV1th), a noise removal process is executed to remove noise data from the first transmembrane pressure difference measurement value group (G1) and / or the second transmembrane pressure difference measurement value group (G2) before the state separation process.

[0028] When the first preliminary evaluation value (SV1) is smaller than the second preliminary evaluation threshold (SV2th), it is determined that the variation in the transmembrane pressure measurement value is small, and a noise removal process is executed without performing normalization processing or averaging processing of the transmembrane pressure.

[0029] The sixth characteristic configuration is, in addition to the first characteristic configuration described above, that the preliminary evaluation step performs a non-hierarchical cluster analysis based on unsupervised machine learning on the time series of transmembrane pressure measurement values ​​in a normal state collected in the preliminary data collection step to separate the time series of transmembrane pressure measurement values ​​in a filtration operation (SG6) and a second preliminary transmembrane pressure measurement value group in a relaxation operation (SG7), and calculates the one having the larger standard deviation between the first and second preliminary transmembrane pressure measurement value groups (SG6, SG7) as the first preliminary evaluation value (SV1).

[0030] In order to determine the degree of variation in the transmembrane pressure measurement values, the time series of transmembrane pressure measurement values ​​under normal conditions collected in the preliminary data collection process may be separated into a first preliminary transmembrane pressure measurement value group (SG6) during filtration operation and a second preliminary transmembrane pressure measurement value group (SG7) during relaxation operation, the standard deviation of each may be calculated, and the one with the larger standard deviation may be used as the first preliminary evaluation value (SV1).

[0031] The seventh characteristic configuration is, in addition to the sixth characteristic configuration described above, that the preliminary evaluation step calculates the number of data points of the transmembrane pressure measurement values ​​of the two preliminary transmembrane pressure measurement value groups (SG6, SG7) as a second preliminary evaluation value (SV2) when the first preliminary evaluation value (SV1) is greater than a predetermined second preliminary evaluation threshold (SV2th) which is smaller than the first preliminary evaluation threshold (SV1th), and executes the state separation step after normalizing the transmembrane pressure measurement value groups for the first period (P1) when either of the second preliminary evaluation values ​​(SV2) is smaller than a predetermined threshold data number (CVth).

[0032] If the first preliminary evaluation value (SV1) is greater than the second preliminary evaluation threshold (SV2th), the number of data points of the transmembrane pressure measurement values ​​of the two preliminary transmembrane pressure measurement value groups (SG6, SG7) may be set as the second preliminary evaluation value (SV2), and if either of the second preliminary evaluation values ​​(SV2) is smaller than a predetermined threshold data number (CVth), the transmembrane pressure measurement value groups for the first period (P1) may be normalized and then a state separation process may be executed.

[0033] The eighth characteristic configuration of the same is, in addition to the seventh characteristic configuration described above, that when both of the second preliminary evaluation values ​​(SV2) are equal to or greater than the threshold data number (CVth), the abnormality determination process calculates an average value of the group of transmembrane pressure measurement values ​​for the second period (P2) as an average transmembrane pressure measurement value for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2), calculates a dispersion state for the second period (P2) from the group of average transmembrane pressure measurement values ​​including each average transmembrane pressure measurement value calculated in the second period (P2) before the shift, and determines whether or not the submerged membrane separation device is abnormal based on the dispersion state.

[0034] Furthermore, if both of the second preliminary evaluation values ​​(SV2) are equal to or greater than the threshold number of data (CVth), it is determined that there is a sufficient number of data and that the variation is not so great that normalization processing is necessary, and an average transmembrane pressure measurement value, which is the average value of the transmembrane pressure values ​​included in the second period (P2), is calculated.

[0035] The ninth characteristic configuration, in addition to the seventh or eighth characteristic configuration described above, is that when the first preliminary evaluation value (SV1) is evaluated in the preliminary evaluation process to be smaller than a predetermined second preliminary evaluation threshold (SV2th), which is a value smaller than the first preliminary evaluation threshold (SV1th), a noise removal process is executed to remove noise data from the first transmembrane differential pressure measurement value group (G1) and / or the second transmembrane differential pressure measurement value group (G2) before the state separation process.

[0036] When the first preliminary evaluation value (SV1) is smaller than the second preliminary evaluation threshold (SV2th), it is determined that the variation in the transmembrane pressure measurement value is small, and a noise removal process is executed without performing normalization processing or averaging processing of the transmembrane pressure.

[0037] The tenth characteristic configuration of the same is that, in addition to any one of the first to ninth characteristic configurations described above, the variance state calculated in the abnormality determination process is an output value of a predetermined evaluation function whose variable is the variance or standard deviation for the transmembrane pressure measurement values ​​for the second period (P2), and the operating state is determined to be abnormal when the output value exceeds a predetermined threshold value.

[0038] An eleventh characteristic configuration of the same is, in addition to any one of the first to tenth characteristic configurations described above, configured to repeatedly execute a series of steps from the state separation step to the abnormality determination step every third period (P3), and the first period (P1) is set to a period longer than the sum of the second period (P2) and the third period (P3). Effect of the Invention

[0039] As described above, according to the present invention, it is possible to provide an abnormality detection method for a submerged membrane separation device that reduces the burden on the monitor and enables the condition of the filtration membrane to be quickly and accurately diagnosed even when the monitor is absent. [Brief description of the drawings]

[0040] [Figure 1] FIG. 1 is an explanatory diagram of a remote monitoring system to which the present invention is applied; [Diagram 2] (a) is an explanatory diagram of a group of transmembrane pressure measurement values ​​collected in the data collection process, and (b) is an explanatory diagram of a group of first transmembrane pressure measurement values ​​during filtration operation and a group of second transmembrane pressure measurement values ​​during relaxation operation, which are state-separated in the state separation process. [Diagram 3] (a) is an explanatory diagram of a group of first transmembrane pressure measurement values ​​during filtration operation in which state separation is performed in a state separation step, (b) is an explanatory diagram of a calculation process for the dispersion state, and (c) is an explanatory diagram of an evaluation value calculated based on the dispersion state. [Figure 4] A flowchart showing the procedure of a first embodiment of a method for detecting an abnormality in a submerged membrane separation device. [Diagram 5] 1 shows a first embodiment of an anomaly detection method for a submerged membrane separation device, in which (a) is an explanatory diagram of a group of transmembrane pressure measurement values ​​collected in a data collection step, (b) is an explanatory diagram of a first group of transmembrane pressure measurement values ​​during filtration operation and a second group of transmembrane pressure measurement values ​​during relaxation operation, which are state-separated in a state separation step, and (c) is an explanatory diagram of the output characteristics of an evaluation function and an anomaly determination threshold value determined in an anomaly determination step. [Figure 6] A flowchart showing a procedure for selecting an embodiment of a method for detecting an abnormality in a submerged membrane separation device. [Figure 7] FIG. 1 is an explanatory diagram of a group of transmembrane pressure measurement values ​​collected in a preliminary data collection step when the transmembrane pressure during filtration operation differs in two stages. [Figure 8] A flowchart showing the processing procedure of a second embodiment of the method for detecting an abnormality in a submerged membrane separation device. [Figure 9] 1 shows a second embodiment of the method for detecting an anomaly in a submerged membrane separation device, in which (a) is an explanatory diagram of a group of transmembrane pressure measurement values ​​collected in a data collection step, (b) is an explanatory diagram of a first group of transmembrane pressure measurement values ​​during filtration operation and a second group of transmembrane pressure measurement values ​​during relaxation operation, which are state-separated in a state separation step, and (c) is an explanatory diagram of the output characteristic of an evaluation function and an anomaly determination threshold value determined in an anomaly determination step. [Figure 10] A flowchart showing the procedure of a third embodiment of the method for detecting an abnormality in a submerged membrane separation device. [Figure 11] 1 is a flowchart showing another embodiment and illustrating a procedure for selecting an aspect of a method for detecting an abnormality in a submerged membrane separation device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0041] Hereinafter, an embodiment of the method for detecting an abnormality in a submerged type membrane separation device according to the present invention will be described. FIG. 1 shows a remote monitoring system to which the method for detecting an abnormality in a submerged membrane separation device is applied.

[0042] The remote monitoring system 100 includes a management device 3 (3A, 3B, . . . , 3N) that manages a submerged membrane separation device 2 (2A, 2B, . . . , 2N) (hereinafter also simply referred to as "membrane separation device 2") installed in a water treatment plant 1 (1A, 1B, . . . , 1N), a remote monitoring device 4 that is a cloud server communicatively connected to the management device 3 (3A, 3B, . . . , 3N) via the Internet or the like, and a diagnostic device 10. Note that the remote monitoring device 4 may have the function of the diagnostic device 10, or may be communicatively connected to the remote monitoring device 4 via the Internet or the like.

[0043] Each management device 3 (3A, 3B, ..., 3N) manages the operation of a single or multiple series of membrane separation devices 2 (2A, 2B, ..., 2N) installed in the water treatment plant 1 (1A, 1B, ..., 1N), by repeating, for example, a 9-minute filtration operation and a 1-minute relaxation operation as one unit.

[0044] During filtration operation, treated water is sucked and filtered through the filtration membrane while being aerated by an aeration device installed below the membrane separation device 2, and during relaxation operation, the suction and filtration of treated water through the filtration membrane is stopped while maintaining the state of aeration by the aeration device. Even during relaxation operation, the filtration membrane surface is cleaned by the air bubbles from the aeration device installed below the membrane separation device 2 and the upward flow of the gas-liquid mixed phase of the water to be treated.

[0045] Each management device 3 is configured to measure the transmembrane pressure at one-minute intervals via a pressure sensor and store the measured transmembrane pressure in an internal memory, and also to transmit operating information including sampling data of the measured transmembrane pressure to a remote monitoring device 4 via the Internet or the like at predetermined time intervals.

[0046] The remote monitoring device 4 stores the sampling data of the transmembrane pressure, i.e., the transmembrane pressure measurement values, transmitted from each management device 3 in a database 4A, and is configured to provide the sampling data of the transmembrane pressure stored in the database 4A, for example in the form of a trend graph, in response to a viewing request from a field staff member or the like using an information terminal such as a smartphone.

[0047] The remote monitoring device 4 is installed with an abnormality detection program that detects whether the submerged membrane separation devices 2 (2A, 2B, ..., 2N) installed in each water treatment plant 1 (1A, 1B, ..., 1N) are operating normally or whether any abnormality has occurred based on the transmembrane pressure measurement values ​​stored in the database 4A.

[0048] In other words, the remote monitoring device 4 also functions as the diagnostic device 10, and the method for detecting an anomaly in a submerged membrane separation apparatus of the present invention is executed by each management device 3 and the remote monitoring device 4. The diagnostic results are stored in a database 4A and can be viewed in response to a viewing request from a field worker or the like using an information terminal such as a smartphone. The diagnostic device 10 that executes the method for detecting an anomaly in a submerged membrane separation apparatus may be configured separately from the remote monitoring device 4, and may be configured as a general-purpose personal computer that can communicate with the remote monitoring device 4 via the Internet or the like.

[0049] Hereinafter, the method for detecting an abnormality in a submerged type membrane separation device according to the first embodiment will be described. The method for detecting an abnormality in a submerged type membrane separation device includes a data collection step, a state separation step, a noise removal step, and an abnormality determination step.

[0050] The data collection step is a step of measuring and collecting the transmembrane pressure of each submerged membrane separation device 2 at predetermined time intervals. The transmembrane pressures sampled at one-minute intervals via a pressure sensor are collected by each management device 3 and stored in an internal memory. The stored transmembrane pressures are transmitted to a remote monitoring device 4 via the Internet together with other data at a predetermined time, for example, about once every half day, and stored in a database 4A provided in the remote monitoring device 4.

[0051] The other data includes an ID code for uniquely identifying the water treatment plant 1, an ID code for uniquely identifying the membrane separation device, and further includes the sampling time of each transmembrane pressure, etc. If multiple systems of membrane separation devices 2 are installed in the water treatment plant 1, the ID code for the membrane separation devices 2 of each system is included.

[0052] FIG. 2(a) shows an example of multiple transmembrane pressure measurement values ​​sampled at one-minute intervals, i.e., a group of transmembrane pressure measurement values. The figure shows a state in which the transmembrane pressure measurement values ​​sampled and plotted in time series are connected by a straight line, with a mixture of transmembrane pressure measurement values ​​during both filtration and relaxation operations, and a noise signal that has been mixed into the signal line of the pressure sensor being superimposed. Note that the transmembrane pressure measurement values ​​are negative values, and the absolute value of the transmembrane pressure measurement value increases toward the bottom in the figure. Therefore, in the following explanation, the lower the transmembrane pressure, the higher the transmembrane pressure.

[0053] The state separation process is a process of separating a group of transmembrane pressure measurement values ​​during a first period (P1) including a target period for detecting abnormalities from the time-series transmembrane pressure measurement values ​​collected in the data collection process, into a first group of transmembrane pressure measurement values ​​during filtration operation (G1) and a second group of transmembrane pressure measurement values ​​during relaxation operation (G2).

[0054] The state separation step is configured to separate the first transmembrane pressure measurement value group and the second transmembrane pressure measurement value group into two clusters by performing a non-hierarchical cluster analysis based on unsupervised machine learning. As the non-hierarchical cluster analysis, k-means can be suitably used.

[0055] In the k-means method, first, the virtual center positions of the two clusters are initially set, and then the close ones are separated as clusters belonging to each virtual center position based on the Euclidean distance between the transmembrane pressure measurement value and the virtual center of gravity. A new center of gravity is calculated from the transmembrane pressure measurement values ​​included in the same cluster. Next, the Euclidean distance between the new center of gravity and the transmembrane pressure measurement value is calculated, and the process of separating the transmembrane pressure measurement values ​​into the cluster with the center of gravity that is closer is repeated until convergence, thereby separating the transmembrane pressure measurement values ​​into a first transmembrane pressure measurement value group (G1) and a second transmembrane pressure measurement value group (G2).

[0056] FIG. 2(b) shows the results of separating the transmembrane pressure measurement values ​​in FIG. 2(a) into a first transmembrane pressure measurement value group (G1) during filtration operation and a second transmembrane pressure measurement value group (G2) during relaxation operation using the k-means method. The darker areas indicate the second transmembrane pressure measurement value group (G2), and the lighter areas indicate the first transmembrane pressure measurement value group (G1). Isolated points that are separated from the mass are scattered in each of the first transmembrane pressure measurement value group (G1) and the second transmembrane pressure measurement value group (G2). The isolated points are noise. The noise includes noise signals mixed into the signal lines of the pressure sensors and transmembrane pressures detected during transitions from filtration operation to relaxation operation and from relaxation operation to filtration operation.

[0057] The noise removal process is a process of removing noise data from the first transmembrane pressure measurement value group and / or the second transmembrane pressure measurement value group during a second period (P2) that is shorter than the first period (P1), and more specifically, a process of removing data that deviates from a specified upper and lower range for each of the first transmembrane pressure measurement value group and the second transmembrane pressure measurement value group.

[0058] For example, the top 95% or more of data with high values ​​and the bottom 5% of data with low values ​​are removed as noise from the number of data in the first transmembrane pressure measurement group and / or the second transmembrane pressure measurement group. It is also possible to remove data that deviates above or below an appropriate range set, for example, from 5% to 95% of the possible range of transmembrane pressure measurement values, as noise. Note that this range is not particularly limited to the range from 5% to 95%, and may be set as appropriate.

[0059] The abnormality determination process is a process of calculating a dispersion state from a first group of transmembrane pressure differential measurement values ​​(G11) and / or a second group of transmembrane pressure differential measurement values ​​(G21) for a second period (P2) that is shorter than the first period (P1) from which noise has been removed in the noise removal process, and determining whether or not the submerged membrane separation device is abnormal based on the dispersion state.

[0060] By separating the cluster into a group of first transmembrane pressure measurement values ​​(G1) during filtration operation and a group of second transmembrane pressure measurement values ​​(G2) during relaxation operation, it becomes possible to perform abnormality diagnosis focused on filtration operation and abnormality diagnosis focused on relaxation operation.

[0061] FIG. 3(a) illustrates a first transmembrane pressure measurement value group (G1) during filtration operation separated in the state separation step, and FIG. 3(b) illustrates a second period (P2) in which the noise removal step is performed and the dispersion state is calculated, as a dashed or solid rectangle. The noise removal step is performed on the transmembrane pressure measurement value group (G11) included in the time width of the dashed or solid rectangle among the first transmembrane pressure measurement value group (G1), and then the dispersion state is calculated. The dashed rectangle indicates the second period (P2) that was performed in the past, and the solid rectangle indicates the second period (P2) that is currently being performed.

[0062] In this embodiment, the first period (P1) is set to 24 hours (one day) and the second period (P2) is set to one hour, but the values ​​are not limited to these values, and it is also possible to set the first period (P1) to 12 hours (half a day) and the second period (P2) to 0.5 hours. Furthermore, the noise removal process is performed and the dispersion state is calculated for each period obtained by shifting the second period (P2) in a chronological order at predetermined time intervals (for example, one minute) shorter than the second period (P2).

[0063] For example, when calculating the dispersion state for each period obtained by shifting the second period (P2) chronologically in units of one minute, 60 dispersion states are calculated in one hour. The predetermined time shorter than the second period (P2) is not limited to one minute, but may be set appropriately to five minutes or the like. The transmembrane pressure measurement value group used in the noise removal step for each period of the second period (P2) is based on the original data before noise removal.

[0064] As shown in Figure 3(c), it is preferable to adopt the output value, i.e., evaluation value, of a predetermined evaluation function whose variable is the variance or standard deviation for the transmembrane pressure measurement values ​​for the second period (P2) as the dispersion state used for abnormality judgment, and if the output value exceeds a predetermined threshold value Ref (abnormality judgment threshold), it is judged that the operating state is abnormal.

[0065] The evaluation function is not particularly limited, and a linear function with the variance or standard deviation as a variable can be used as appropriate. In this embodiment, a linear function is used in which the variance or standard deviation as a variable is multiplied by a predetermined coefficient. When the coefficient is 1, the variance or standard deviation becomes the evaluation value.

[0066] 3(b), in region R where the variation in the transmembrane pressure measurement values ​​gradually increases, the evaluation value exceeds the anomaly determination threshold, i.e., it is determined that an anomaly has occurred. This focuses on the fact that if stability is lost for some reason in the second period (P2), the variance or standard deviation of the transmembrane pressure measurement values ​​in the second period (P2) increases.

[0067] As shown by the dashed arrow in FIG. 3(b), it is preferable to configure the series of steps from the state separation step to the abnormality determination step to be repeatedly executed every time the third period (P3) elapses. In this case, the first period P1 (24 hours) to be targeted is the first period P1 (24 hours) after the third period P3 (6 hours) has elapsed since the first period P1 that was the target last time. The first period (P1) is set to a period longer than the sum of the second period (P2) and the third period (P3). Note that, in the methods for detecting an abnormality in a submerged membrane separation device according to the second and third aspects described later, it is preferable to configure the series of steps from the state separation step to the abnormality determination step to be repeatedly executed every time the third period (P3) elapses.

[0068] During the first period (P1) including the target period for detecting an abnormality, the state separation process to the abnormality determination process are repeated every time the third period (P3) shorter than the first period (P1) elapses. Therefore, the accuracy of the state separation is ensured by the state separation process using the first transmembrane pressure measurement value group during the first period (P1) set longer than the third period (P3), and the calculation load of the diagnosis device 10 can be distributed even if the number of submerged membrane separation devices to be diagnosed increases by setting the third period (P3) to a predetermined period, such as to match the timing of monitoring by a monitor. In this example, the third period (P3) is set to 6 hours. Therefore, a series of processes from the state separation process to the abnormality determination process is repeated every 6 hours.

[0069] Furthermore, the same process is performed on the second transmembrane pressure measurement value group (G2), and an abnormality determination is performed for each of the filtration operation and the relaxation operation. Note that since the type of abnormality detected from the first transmembrane pressure measurement value group (G1) and the type of abnormality detected from the second transmembrane pressure measurement value group (G2) may differ, it is not essential to execute the series of steps for both the first and second transmembrane pressure measurement value groups (G1, G2), and it is sufficient to execute the series of steps for either one depending on the abnormal state to be detected.

[0070] FIG. 4 shows the procedure of the method for detecting an abnormality in a submerged type membrane separation device according to the first embodiment described above. That is, the method for detecting an abnormality in a submerged membrane separation device that alternately repeats filtration operation and relaxation operation includes a data collection step (SA1) of measuring and collecting the transmembrane pressure of the submerged membrane separation device at predetermined time intervals, and a state separation step (SA2) of separating a group of transmembrane pressure measurement values ​​during a first period (P1) including a target period for detecting an abnormality from the time-series transmembrane pressure measurement values ​​collected in the data collection step into a first group of transmembrane pressure measurement values ​​during filtration operation and a second group of transmembrane pressure measurement values ​​during relaxation operation.

[0071] Then, a second period (P2) shorter than the first period (P1) is set (SA3), a noise removal process is performed to remove noise data from the first transmembrane pressure measurement value group and the second transmembrane pressure measurement value group during the second period (P2) (SA4), and a process is performed to calculate the dispersion state for the first transmembrane pressure measurement value group and the second transmembrane pressure measurement value group during the second period (P2) from which noise has been removed (SA5).

[0072] In order to execute each process from step SA3 to step SA5 throughout the entire target period, the second period P2 is shifted chronologically at predetermined time intervals and repeated (SA6), and an abnormality determination process is executed (SA7, SA8, SA9, SA10) to determine whether the submerged membrane separation device is normal or abnormal based on each obtained dispersion state and on a predetermined abnormality determination threshold.

[0073] In the flowchart shown in FIG. 4, an example is shown in which the second period (P2) is shifted at a predetermined time interval so as to cover the entire period of the first period (P1), and if the evaluation value exceeds the abnormality determination threshold in any one of the second periods (P2), it is determined that an abnormality has occurred. However, it may also be configured to determine that an abnormality has occurred if the number of times that the evaluation value exceeds the abnormality determination threshold in the second period (P2) reaches a predetermined number.

[0074] As described using the dashed arrow in FIG. 3(b), the abnormality determination process may be repeated in a third period (P3) that is shorter than the first period (P1).

[0075] Hereinafter, a method for detecting an anomaly in a submerged type membrane separation device according to a different aspect from the method for detecting an anomaly in a submerged type membrane separation device according to the first aspect, and a judgment for selecting the different aspect of the anomaly detection method will be described.

[0076] As shown in Figure 5(a), in some submerged membrane separation devices, during filtration operation in which treated water is sucked through the filtration membrane, the suction volume of the suction pump may be switched in two stages depending on the time period when the amount of water being treated is high, etc. In this example, the pressure is -20 kPa during relaxation operation, approximately -65 kPa during filtration operation when the pump suction volume is low, and approximately -85 kPa during filtration operation when the pump suction volume is high.

[0077] As shown in Figure 5(b), when the state separation process described in the first embodiment is performed on transmembrane pressure measurement values ​​exhibiting such characteristics, data variation due to switching of the suction volume by the suction pump appears in the first transmembrane pressure measurement value group during filtration operation. Therefore, as shown in Figure 5(c), even if the behavior is essentially normal, in regions R1 and R2 where the variation is large, the evaluation value may exceed the abnormality determination threshold in the abnormality determination process, leading to a risk of an erroneous determination that an abnormality exists.

[0078] Even in a submerged membrane separation device in which the amount of suction by the pump is kept constant, if the water level in the treatment tank in which the filtration membrane is submerged fluctuates, the first transmembrane pressure measurement group during filtration operation fluctuates accordingly, which may lead to a similar erroneous judgment. Furthermore, when switching between filtration operation and relaxation operation, a similar erroneous judgment may be made even if an intermediate value of the transmembrane pressure during filtration operation and relaxation operation is measured. Even in such a case, it is necessary to appropriately judge an abnormality without causing an erroneous judgment.

[0079] FIG. 6 shows a procedure for selecting a specific determination method that enables appropriate abnormality determination depending on the operating environment in which the submerged membrane separation apparatus is installed. An appropriate abnormality determination method is selected from a plurality of abnormality determination methods based on the transmembrane pressure measured while the submerged membrane separation device to be determined to be abnormal is operating normally, as described in detail below.

[0080] The method for detecting an abnormality in an immersion type membrane separation device includes a preliminary data collection step (SB1) and preliminary evaluation steps (SB2 to SB6), and is configured to select an appropriate abnormality detection method from a plurality of abnormality detection methods including the abnormality detection method of the first aspect described above based on the results of the preliminary evaluation steps.

[0081] A preliminary data collection step (SB1) in which a preliminary operation is performed on an immersion type membrane separation device in a normal state, and the intermembrane differential pressure is measured and collected at predetermined time intervals (in this embodiment, set every minute in accordance with the process of step SA1 in FIG. 4), and a preliminary evaluation step (SB2 to SB6) in which a first preliminary evaluation value (SV1) indicating the fluctuation characteristics of the intermembrane differential pressure is calculated based on the time-series intermembrane differential pressure measurement values in the normal state collected in the preliminary data collection step are executed prior to a data collection step (for example, step SA1 in FIG. 4) for actually detecting an abnormality.

[0082] In the preliminary evaluation step, by performing non-hierarchical cluster analysis based on unsupervised machine learning, for example, non-hierarchical cluster analysis by the k-means method described above, on the time-series intermembrane differential pressure measurement values in the normal state collected in the preliminary data collection step (SB1), the values are separated into three preliminary intermembrane differential pressure measurement value groups (SG1, SG2, SG3) (SB2), and among the centroid positions CG1, CG2, CG3 (CG1 < CG2 < CG3) of each preliminary intermembrane differential pressure measurement value group (SG1, SG2, SG3), the centroid distance ratio obtained by (CG3 - CG2) / (CG2 - CG1) is calculated as the first preliminary evaluation value (SV1) (SB3).

[0083] For example, as shown in FIG. 7, when the suction amount by the suction pump is switched in two stages during the filtration operation of sucking and filtering the treated water from the filter membrane, the centroid position (CG1) of the preliminary intermembrane differential pressure measurement value group (SG1) during the relaxation operation, the centroid position (CG2) of the preliminary intermembrane differential pressure measurement value group (SG2) during the filtration operation corresponding to the low suction pressure, and the centroid position (CG3) of the preliminary intermembrane differential pressure measurement value group (SG3) during the filtration operation corresponding to the high suction pressure are obtained, and the ratio obtained by (CG3 - CG2) / (CG2 - CG1) is calculated as the first preliminary evaluation value (SV1).

[0084] Note that the fluctuation of the suction pressure during the filtration operation may be caused not only by the switching of the suction pressure by the suction pump but also by the fluctuation of the water level in the treatment water tank in which the filtration membrane is immersed. Further, even if the suction pressure during the filtration operation is constant, the differential pressure between membranes measured during the pre-data collection step (SB1) may include the differential pressure between membranes at an intermediate level when switching between the relaxation operation and the filtration operation.

[0085] That is, the time-series differential pressure between membranes measured in the normal state collected in the pre-data collection step (SB1) is separated into at least three pre-differential pressure between membranes measurement value groups (SG1, SG2, SG3) including at least the filtration operation and the relaxation operation. Among the centroid positions CG1, CG2, CG3 (CG1 < CG2 < CG3) of each pre-differential pressure between membranes measurement value group (SG1, SG2, SG3), the ratio obtained by (CG3 - CG2) / (CG2 - CG1) is calculated as the first pre-evaluation value (SV1), whereby the degree of variation peculiar to the immersion type membrane separation device can be objectively evaluated.

[0086] When the first pre-evaluation value (SV1) is greater than the first pre-evaluation threshold value (SV1th) (SB4, right branch), in order to avoid misjudgment due to the influence of large fluctuation characteristics, the second mode described below is adopted. The second mode is a mode in which the differential pressure between membranes measurement value group in the first period (P1) in the first mode described above is normalized based on the differential pressure between membranes measurement value collected in the pre-data collection step, and the state separation step is executed for the normalized differential pressure between membranes measurement value group. As a result of suppressing the influence of the variation of the differential pressure between membranes collected in the data collection step, appropriate abnormality determination becomes possible.

[0087] If the first preliminary evaluation value (SV1) is smaller than the first preliminary evaluation threshold (SV1th) and larger than a predetermined second preliminary evaluation threshold (SV2th) which is smaller than the first preliminary evaluation threshold (SV1th), in other words, if the variation is conceptually judged to be moderate (SB4, downward branch), a non-hierarchical cluster analysis (k-means) based on unsupervised machine learning is further performed on the time series of transmembrane pressure measurement values ​​in the normal state collected in the preliminary data collection process, to separate them into two preliminary transmembrane pressure measurement value groups (SG4, SG5) corresponding to filtration operation and relaxation operation, and the number of data points of the transmembrane pressure measurement values ​​in each preliminary transmembrane pressure measurement value group (SG4, SG5) is calculated as the second preliminary evaluation value (SV2) (SB5).

[0088] Specifically, a 10-minute window is set for the transmembrane pressure measurement values ​​of each preliminary transmembrane pressure measurement value group (SG4, SG5), and the process of counting the number of data included in the window is executed for each window frame shifted by one minute in the time axis direction. If the number of data included in each window frame is less than a predetermined number (the number of data is set to less than 5 in this embodiment. Since the transmembrane pressure value is sampled at one-minute intervals, the number of data included in the window frame is a maximum of 10) due to classification into the other preliminary transmembrane pressure measurement value group or the transmembrane pressure measurement value itself being missing, the number of data is set to 0, and if the number of data is equal to or more than a predetermined number (5 or more in this embodiment), the number of data is set to 1. This process is repeated, and the sum of the numbers of data found for each window frame is set as the second preliminary evaluation value (SV2). The width of the window frame is not particularly limited and may be set appropriately.

[0089] If any one of the second preliminary evaluation values ​​(SV2) is less than a predetermined threshold data number (CVth) (SB6, right branch), the second mode is adopted, and if both of the second preliminary evaluation values ​​(SV2) are equal to or greater than the threshold data number (CVth), it is determined that there is a sufficient number of data and the variation is not so large that normalization processing is necessary (SB6, lower branch), and the third mode described below is adopted. The third mode is a mode in which the transmembrane pressure measurement value group for the second period (P2) in the above-mentioned first mode is averaged.

[0090] Hereinafter, the method for detecting an abnormality in a submerged type membrane separation device according to the second embodiment will be described with reference to FIG. The anomaly detection method for a submerged type membrane separation device according to the second aspect includes a data collection step (SC1), a noise removal step (SC2), a normalization processing step (SC3), a state separation step (SC4), and anomaly determination steps (SC5 to SC11).

[0091] In the data collection step (SC1), the transmembrane pressure of the submerged membrane separation device is measured and collected at predetermined time intervals. In the noise removal step (SC2), noise data is removed from the time-series transmembrane pressure measurement values ​​collected in the data collection step. In the noise removal step, data that deviates from a predetermined upper and lower range for the transmembrane pressure measurement value is removed. As in the first embodiment, for example, the top 95% or more data with high values ​​and the bottom 5% data with low values ​​are removed as noise, based on the number of data in the transmembrane pressure measurement value group. In addition, for example, data that deviates above or below the appropriate range set to 5% to 95% of the possible range of the transmembrane pressure measurement value can also be removed as noise. Note that this range is not particularly limited to the range of 5% to 95%, and may be set appropriately.

[0092] In the normalization process (SC3), the transmembrane pressure measurement values ​​collected in the data collection process are normalized based on the maximum and minimum measurement values ​​sampled in advance for the submerged membrane separation device that is the target of abnormality judgment. For example, the values ​​are set to 0 kPa during relaxation operation and 1 kPa during filtration operation.

[0093] In the state separation step (SC4), a group of transmembrane pressure measurements normalized in the normalization step (SC3) for a first period (P1) including a target period for detecting an abnormality is separated into a first transmembrane pressure measurement group (G1) during filtration operation and a second transmembrane pressure measurement group (G2) during relaxation operation. As in the first embodiment, the state separation step is configured to separate the first transmembrane pressure measurement group (G1) and the second transmembrane pressure measurement group (G2) into two clusters by performing non-hierarchical cluster analysis based on unsupervised machine learning. The k-means method can be suitably used as the non-hierarchical cluster analysis.

[0094] In the abnormality determination step, as in the first embodiment, for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2) separated in the state separation step (SC4), a dispersion state is calculated from the transmembrane pressure measurement value group (G11 and / or G21) for a second period (P2) shorter than the first period (P1), and it is determined whether or not the submerged membrane separation device is abnormal based on the dispersion state (SC5, SC6).

[0095] In order to execute each process of steps SC5 and SC6 throughout the entire target period, the second period (P2) is shifted chronologically at predetermined time intervals and repeated (SC7), and an abnormality determination process is executed (SC8, SC9, SC10, SC11) to determine whether the submerged membrane separation device is normal or abnormal based on each obtained dispersion state and on a predetermined abnormality determination threshold.

[0096] FIG. 9(a) shows a characteristic diagram in which the transmembrane pressure measurement values ​​of FIG. 5(a) processed in the first mode have been normalized. The values ​​are normalized to 0.0 kPa during relaxation operation and to −1.0 kPa during filtration operation in which the pump suction volume is large. FIG. 9(b) shows the result of state separation of the data shown in FIG. 9(a). By performing normalization, the variance or standard deviation calculated in response to fluctuations in the pump suction volume does not fluctuate significantly, and as shown in FIG. 9(c), by appropriately setting the abnormality determination threshold, erroneous determination is prevented.

[0097] Hereinafter, the abnormality detection method for the submerged type membrane separation device according to the above-mentioned third aspect will be described in detail. The anomaly detection method for a submerged type membrane separation device according to the third aspect includes a data collection step (SD1), a noise removal step (SD2), a state separation step (SD3), an averaging process step (SD4 to SD6), and an anomaly determination step (SD7 to SD13).

[0098] In the data collection step (SD1), the transmembrane pressure of the submerged membrane separation device is measured and collected at predetermined time intervals. In the noise removal step (SD2), noise data is removed from the time-series transmembrane pressure measurement values ​​collected in the data collection step. In the noise removal step, data that deviates from a predetermined upper and lower range for the transmembrane pressure measurement value is removed. As in the first embodiment, for example, the top 95% or more data with high values ​​and the bottom 5% data with low values ​​are removed as noise, based on the number of data in the transmembrane pressure measurement value group. In addition, for example, data that deviates above or below an appropriate range set to 5% to 95% of the possible range of the transmembrane pressure measurement value can also be removed as noise. Note that this range is not particularly limited to the range of 5% to 95%, and may be set appropriately.

[0099] In the state separation step (SD3), a group of transmembrane pressure measurement values ​​in a first period (P1) including a target period for detecting an abnormality among the transmembrane pressure measurement values ​​from which noise has been removed in the noise removal step (SD2) is separated into a first transmembrane pressure measurement value group (G1) during filtration operation and a second transmembrane pressure measurement value group (G2) during relaxation operation. As in the first embodiment, the state separation step is configured to separate into two clusters, the first transmembrane pressure measurement value group (G1) and the second transmembrane pressure measurement value group (G2), by performing a non-hierarchical cluster analysis based on unsupervised machine learning. As the non-hierarchical cluster analysis, k-means can be suitably used.

[0100] In the averaging process steps (SD4 to SD6), a process of calculating a moving average value from a group of transmembrane pressure measurement values ​​(G11 and / or G21) in a second period (P2) that is shorter than the first period (P1) for the first group of transmembrane pressure measurement values ​​(G1) and / or the second group of transmembrane pressure measurement values ​​(G2) separated in the state separation process (SD3) is repeated by shifting the second period (P2) chronologically at predetermined time intervals in order to execute the process for the entire period of interest.

[0101] The process of calculating the dispersion state from the transmembrane pressure measurement values ​​(G11 and / or G21) in the second period described above is repeated for the entire target period, shifting the second period (P2) chronologically at predetermined intervals. The transmembrane pressure value for calculating the dispersion state at this time is the moving average value calculated in steps SD4 to SD6.

[0102] An evaluation process is performed on the variance values ​​calculated in steps SD7 to SD9 (SD10). If the variance is less than a preset abnormality threshold, it is determined to be normal (SD11, SD12), and if it is equal to or greater than the threshold, it is determined to be abnormal (SD11, SD13).

[0103] In other words, the abnormality determination process calculates the average value of the group of average transmembrane pressure measurement values ​​for the second period (P2) as the average transmembrane pressure measurement value for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2), calculates the dispersion state for the second period (P2) from the group of average transmembrane pressure measurement values ​​including each average transmembrane pressure measurement value calculated in the second period (P2) before the shift, and determines whether or not the submerged membrane separation device is abnormal based on the dispersion state.

[0104] FIG. 11 shows another embodiment relating to the calculation and judgment of the first preliminary evaluation value (SV1) indicating the fluctuation characteristic of the transmembrane pressure explained in FIG. The method includes a preliminary data collection step (SE1) and a preliminary evaluation step (SE2 to SE5), and based on the results of the preliminary evaluation step, an appropriate anomaly detection method is selected from the anomaly detection methods of any one of the first, second, or third aspects described above.

[0105] The preliminary evaluation process separates the time series of transmembrane pressure measurement values ​​under normal conditions collected in the preliminary data collection process (SE1) into a first preliminary transmembrane pressure measurement value group (SG6) during filtration operation and a second preliminary transmembrane pressure measurement value group (SG7) during relaxation operation by performing unsupervised machine learning-based non-hierarchical cluster analysis (k-means) on the time series of transmembrane pressure measurement values ​​under normal conditions collected in the preliminary data collection process (SE1), and calculates the standard deviation of each of the first and second preliminary transmembrane pressure measurement value groups (SG6, SG7) to determine the degree of variation in the transmembrane pressure measurement values, and calculates the larger standard deviation as the first preliminary evaluation value (SV1).

[0106] When the first preliminary evaluation value (SV1) is greater than a predetermined first preliminary evaluation threshold value (SV1th), the second aspect of the anomaly detection method shown in FIG. 8 is executed.

[0107] When the first preliminary evaluation value (SV1) is smaller than the first preliminary evaluation threshold (SV1th) and larger than a predetermined second preliminary evaluation threshold (SV2th) which is smaller than the first preliminary evaluation threshold (SV1th), the number of data points of the transmembrane pressure measurement values ​​of the two preliminary transmembrane pressure measurement value groups (SG6, SG7) is calculated as a second preliminary evaluation value (SV2), and when either of the second preliminary evaluation values ​​(SV2) is smaller than a predetermined threshold data number (CVth), the transmembrane pressure measurement value groups for the first period (P1) are normalized and then a state separation step is executed. The method of calculating the number of data points of the transmembrane pressure measurement values ​​of the two preliminary transmembrane pressure measurement value groups (SG6, SG7) is the same as the method described in step SB6 of FIG. 6.

[0108] If the first preliminary evaluation value (SV1) is greater than the second preliminary evaluation threshold (SV2th), the number of data points of the transmembrane pressure measurement values ​​of the two preliminary transmembrane pressure measurement value groups (SG6, SG7) may be set as the second preliminary evaluation value (SV2), and if either of the second preliminary evaluation values ​​(SV2) is smaller than a predetermined threshold data number (CVth), the transmembrane pressure measurement value groups for the first period (P1) may be normalized and then a state separation process may be executed.

[0109] When both of the second preliminary evaluation values ​​(SV2) are equal to or greater than the threshold data number (CVth), the abnormality determination step calculates the average value of the group of transmembrane pressure measurement values ​​for the second period (P2) as the average transmembrane pressure measurement value for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2) (processing of steps SD4 to SD6 in FIG. 10), calculates the dispersion state for the second period (P2) from the group of average transmembrane pressure measurement values ​​including each average transmembrane pressure measurement value calculated in the second period (P2) before the shift (processing of steps SD7 to SD9 in FIG. 10), and determines whether or not the submerged membrane separation device is abnormal based on the dispersion state.

[0110] Furthermore, if both of the second preliminary evaluation values ​​(SV2) are equal to or greater than the threshold number of data (CVth), it is determined that there is a sufficient number of data and that the variation is not so great that normalization processing is necessary, and an average transmembrane pressure measurement value, which is the average value of the transmembrane pressure values ​​included in the second period (P2), is calculated.

[0111] If the first preliminary evaluation value (SV1) is evaluated to be smaller than a predetermined second preliminary evaluation threshold (SV2th), which is a value smaller than the first preliminary evaluation threshold (SV1th), the first aspect of the anomaly detection method is executed, including a noise removal process for removing noise data, on the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2) before the state separation process.

[0112] The dispersion state calculated in the abnormality determination process of the first to third aspects is the output value of a predetermined evaluation function whose variable is the variance or standard deviation for the transmembrane pressure measurement values ​​during the second period (P2), and if the output value exceeds a predetermined threshold value, the operating state is determined to be abnormal.

[0113] The above description is one example of the method for detecting an abnormality in a submerged membrane separation device according to the present invention, and it goes without saying that the specific aspects of each process can be appropriately modified and designed within the scope of the effects of the present invention. [Explanation of symbols]

[0114] 100: Remote monitoring system 1(1A, 1B, 1C): Water treatment plant 2: Submerged membrane separation device 3: Management device 4: Remote monitoring equipment (diagnostic equipment) 4A: Database

Claims

1. A data collection step of measuring and collecting a transmembrane pressure difference at predetermined time intervals for a submerged membrane separation apparatus that alternately repeats a filtration operation and a relaxation operation; A state separation step of separating a group of transmembrane pressure measurements during a filtration operation (G1) and a group of transmembrane pressure measurements during a relaxation operation (G2) by performing a non-hierarchical cluster analysis based on unsupervised machine learning including a k-means algorithm on a group of transmembrane pressure measurements during a first period (P1) including a target period for detecting an abnormality among the time-series transmembrane pressure measurements collected in the data collection step; an anomaly determination step of calculating a dispersion state indicated by an output value of a predetermined evaluation function including a linear function having as a variable a variance or standard deviation for a transmembrane pressure measurement group (G11, G21) in a second period (P2) shorter than the first period (P1) for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2), and determining whether or not the submerged membrane separation device is abnormal in the second period (P2) based on whether or not the dispersion state exceeds an anomaly determination threshold, the process being repeated by shifting the time series at predetermined times shorter than the second period (P2); A method for detecting an abnormality in a submerged membrane separation apparatus, comprising: A preliminary data collection step of performing a preliminary operation of the submerged membrane separation apparatus in a normal state to measure and collect the transmembrane pressure difference at predetermined time intervals; a preliminary evaluation step of performing a non-hierarchical cluster analysis based on unsupervised machine learning including k-means on the time-series transmembrane pressure measurement values ​​in a normal state collected in the preliminary data collection step, thereby separating the time-series transmembrane pressure measurement values ​​into three preliminary transmembrane pressure measurement value groups (SG1, SG2, SG3), and calculating a ratio obtained by (CG3-CG2) / (CG2-CG1) among the center positions CG1, CG2, CG3 (CG1<CG2<CG3) of each of the preliminary transmembrane pressure measurement value groups (SG1, SG2, SG3) as the first preliminary evaluation value (SV1) indicating the degree of inherent variation of the submerged membrane separation device; is carried out prior to the data collection step; When the first preliminary evaluation value (SV1) is greater than a predetermined first preliminary evaluation threshold (SV1th) that is an indicator that the degree of the inherent variation is large, the state separation process is performed on a group of transmembrane pressure measurement values ​​normalized based on a maximum measurement value and a minimum measurement value of the transmembrane pressure measurement values ​​under normal conditions collected in the preliminary data collection process for the first period (P1).

2. The preliminary evaluation step is a step of, when the first preliminary evaluation value (SV1) is smaller than the first preliminary evaluation threshold (SV1th) and larger than a predetermined second preliminary evaluation threshold (SV2th) which is an index indicating that the degree of the inherent variation is small, further performing a non-hierarchical cluster analysis based on unsupervised machine learning including k-means on the time-series transmembrane pressure measurement values ​​in a normal state collected in the preliminary data collection step, thereby separating the transmembrane pressure measurement values ​​into two preliminary transmembrane pressure measurement value groups (SG4, SG5), setting a predetermined time window for the transmembrane pressure measurement values ​​of each of the preliminary transmembrane pressure measurement value groups (SG4, SG5), and counting the number of data included in the window. the process is repeated for each window frame shifted in the time axis direction by the measurement interval in the preliminary data collection process, the number of data is set to 0 if the number of data included in each window frame is less than a predetermined number, and the number of data is set to 1 if the number is equal to or greater than the predetermined number, the sum of the numbers of data calculated for each window frame is calculated as a second preliminary evaluation value (SV2), and if any of the second preliminary evaluation values ​​(SV2) is less than a predetermined threshold number of data (CVth) set as a number sufficient for evaluation, the state separation process is performed on a group of transmembrane pressure measurement values ​​normalized from a group of transmembrane pressure measurement values ​​in the first period (P1).

3. The method for detecting an abnormality in a submerged type membrane separation apparatus as described in claim 2, wherein, when both of the second preliminary evaluation values ​​(SV2) for each of the preliminary transmembrane pressure measurement value groups (SG4, SG5) are equal to or greater than the threshold data number (CVth), the abnormality determination step calculates, for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2), an average value of the transmembrane pressure measurement value group for the second period (P2) as an average transmembrane pressure measurement value, calculates a dispersion state indicated by an output value of a predetermined evaluation function including a linear function with a variance or standard deviation for the second period as a variable from the average transmembrane pressure measurement value group including each average transmembrane pressure measurement value calculated for the second period (P2) before the predetermined time is shifted in a chronological manner, and determines whether or not the submerged type membrane separation apparatus is abnormal based on the dispersion state.

4. 4. The method for detecting an anomaly in a submerged membrane separation device according to claim 2 or 3, further comprising the steps of: when, in the preliminary evaluation step, the first preliminary evaluation value (SV1) is evaluated to be smaller than the predetermined second preliminary evaluation threshold (SV2th), which is a value smaller than the first preliminary evaluation threshold (SV1th), performing a noise removal step of removing noise data from the first transmembrane differential pressure measurement value group (G1) and / or the second transmembrane differential pressure measurement value group (G2) before the state separation step.

5. A data collection step of measuring and collecting a transmembrane pressure difference at predetermined time intervals for a submerged membrane separation apparatus that alternately repeats a filtration operation and a relaxation operation; A state separation step of separating a group of transmembrane pressure measurements during a filtration operation (G1) and a group of transmembrane pressure measurements during a relaxation operation (G2) by performing a non-hierarchical cluster analysis based on unsupervised machine learning including a k-means algorithm on a group of transmembrane pressure measurements during a first period (P1) including a target period for detecting an abnormality among the time-series transmembrane pressure measurements collected in the data collection step; an anomaly determination step of calculating a dispersion state indicated by an output value of a predetermined evaluation function including a linear function having as a variable a variance or standard deviation for a transmembrane pressure measurement group (G11, G21) in a second period (P2) shorter than the first period (P1) for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2), and determining whether or not the submerged membrane separation device is abnormal in the second period (P2) based on whether or not the dispersion state exceeds an anomaly determination threshold, the process being repeated by shifting the time series at predetermined times shorter than the second period (P2); A method for detecting an abnormality in a submerged membrane separation apparatus, comprising: A preliminary data collection step of performing a preliminary operation of the submerged membrane separation apparatus in a normal state to measure and collect the transmembrane pressure difference at predetermined time intervals; a preliminary evaluation step of performing a non-hierarchical cluster analysis based on unsupervised machine learning on the time-series transmembrane pressure measurement values ​​in a normal state collected in the preliminary data collection step to separate the transmembrane pressure measurement values ​​into a first preliminary transmembrane pressure measurement value group (SG6) during filtration operation and a second preliminary transmembrane pressure measurement value group (SG7) during relaxation operation, calculating the first preliminary evaluation value (SV1) as the first preliminary evaluation value (SV1) of the first and second preliminary transmembrane pressure measurement value groups (SG6, SG7) having a larger standard deviation, and calculating the number of data of the transmembrane pressure measurement values ​​of the two preliminary transmembrane pressure measurement value groups (SG6, SG7) as a second preliminary evaluation value (SV2) when the first preliminary evaluation value (SV1) is greater than a predetermined second preliminary evaluation threshold (SV2th) that is smaller than the first preliminary evaluation threshold (SV1th); is carried out prior to the data collection step; a state separation step being performed on a group of transmembrane pressure measurement values ​​obtained by normalizing a group of transmembrane pressure measurement values ​​for the first period (P1) based on transmembrane pressure measurement values ​​collected in the preliminary data collection step when the first preliminary evaluation value (SV1) is greater than a predetermined first preliminary evaluation threshold value (SV1th); and a state separation step being performed after normalizing the group of transmembrane pressure measurement values ​​for the first period (P1) when any of the second preliminary evaluation values ​​(SV2) is less than a predetermined threshold data number (CVth).

6. The method for detecting an abnormality in a submerged type membrane separation device as described in claim 5, wherein, when both of the second preliminary evaluation value (SV2) for the first and second preliminary transmembrane pressure measurement value groups (SG6, SG7) are equal to or greater than the threshold data number (CVth), the abnormality determination step calculates an average value of the transmembrane pressure measurement value group for the second period (P2) for the first transmembrane pressure measurement value group (G1) and / or the second transmembrane pressure measurement value group (G2) as an average transmembrane pressure measurement value, calculates a dispersion state for the second period (P2) from the average transmembrane pressure measurement value group including each average transmembrane pressure measurement value calculated in the second period (P2) before the shift, and determines whether or not the submerged type membrane separation device is abnormal based on the dispersion state.

7. 7. The method for detecting an anomaly in a submerged type membrane separation device according to claim 5 or 6, further comprising the steps of: when, in the preliminary evaluation step, it is evaluated that the first preliminary evaluation value (SV1) is smaller than a predetermined second preliminary evaluation threshold (SV2th) that is a value smaller than the first preliminary evaluation threshold (SV1th), a noise removal step is executed to remove noise data from the first transmembrane differential pressure measurement value group (G1) and / or the second transmembrane differential pressure measurement value group (G2) before the state separation step.

8. 8. The method for detecting an anomaly in a submerged membrane separation apparatus according to claim 1, wherein the dispersion state calculated in the anomaly determination step is an output value of a predetermined evaluation function including a linear function having a variance or a standard deviation for the transmembrane pressure measurement values ​​in the second period (P2) as a variable, and when the output value exceeds a predetermined threshold value, the operating state is determined to be abnormal.

9. 9. The method for detecting an anomaly in a submerged membrane separation device according to claim 1, wherein a series of steps from the state separation step to the anomaly determination step are repeatedly executed every third period (P3), and the first period (P1) is set to a period longer than the sum of the second period (P2) and the third period (P3).

Images