Water treatment system and water treatment method

Abstract

This water treatment system comprises: a flocculation tank in which a flocculant is added to water to be treated and flocculates are formed thereby; a camera for acquiring a plurality of images of the flocculates formed in the flocculation tank; an information derivation unit which derives, from the plurality of images, numerical information on a characteristic quantity relating to a flocculation state of the flocculates; a fractal-dimension prediction unit which applies the characteristic quantity included in the numerical information to a prediction expression and derives fractal-dimension values of the flocculates; a pump for regulating the amount of the flocculant to be added to the water to be treated, on the basis of the derived fractal-dimension values; and a membrane filtration part where the water to be treated is filtered with a membrane. The numerical information includes quantified information for three-dimensionally capturing the flocculates on the basis of the plurality of flocculate images.

Description

Water treatment system and water treatment method

[0001] The present invention relates to a water treatment system and a water treatment method.

[0002] Water treatment technology using membrane filtration is used in a variety of fields because it is safe, high-quality, and can separate impurities from water using a compact, low-cost process. However, the following technical issues currently exist with membrane filtration:

[0003] First, it is necessary to be able to respond to changes in water quality. For example, when filtering river water, the quality of river water can change suddenly due to heavy rainfall caused by climate change. Even in such cases, technology is needed to enable stable membrane filtration of river water.

[0004] Second, it must be able to handle highly difficult-to-treat water. While membrane filtration has the filtration accuracy to provide high-quality treated water, membrane clogging can occur when filtering water containing large amounts of organic matter and fine particles. Therefore, there is a problem that it is difficult to apply membrane filtration to the treatment of such highly difficult-to-treat water.

[0005] Thirdly, it is required to be able to handle filtration with high flux. Flux (permeable water volume per unit area and unit time, L / m ２ By increasing the speed (hours / hour), the device can be made more compact and less expensive.

[0006] One possible method for solving these problems is a flocculation membrane filtration method that combines the addition of a flocculant with membrane filtration. This method can provide the following solutions to the above problems.

[0007] For the first problem, by adjusting the amount of coagulant added to the water to be treated in response to fluctuations in water quality, stable membrane filtration may be possible regardless of water quality fluctuations.For the second problem, even when the water to be treated contains a large amount of organic matter or fine particles, membrane clogging can be suppressed by adding a coagulant to flocculate these particles before filtering.For the third problem, by appropriately controlling the characteristics of the flocculated flocs, high-flux filtration conditions may be achieved.

[0008] In conventional coagulation membrane filtration methods, flocs are formed by adding a coagulant, the flocs are then allowed to settle in a settling tank, and the supernatant water is then filtered. In this case, a relatively large tank is required to ensure sufficient water surface loading (retention time) in the settling tank. This creates a problem of installation space. Furthermore, the amount of coagulant required tends to be large, resulting in high costs. Conventional coagulation membrane filtration methods are based on the premise that the flocs are passed through a settling tank after coagulation, settled, and then filtered by membrane, so control is performed according to the settling properties of the flocs. Patent Document 1, for example, describes a coagulation and settling control device and control method capable of forming flocs with good settling properties. Patent Document 2 also describes a method of evaluating the state of flocs in order to form flocs with good settling properties.

[0009] On the other hand, there is a type of coagulation membrane filtration method that, unlike conventional coagulation and sedimentation methods, does not settle the flocs in a settling tank, but instead injects a coagulant and then directly filters the injected coagulant through a membrane. This coagulation direct membrane filtration method does not require a settling tank and requires a smaller amount of coagulant added. This has the advantage of reducing the overall initial and running costs of the water treatment facility. However, in the current situation, there has been insufficient research into coagulant dosage control devices and appropriate dosages for coagulation direct membrane filtration that enable stable membrane filtration operation. For example, Patent Document 3 proposes a technology for injecting coagulants so that the ratio of the turbidity of the raw water to the metal ion concentration is maintained at a certain level. Furthermore, Patent Document 4 proposes controlling the amount of coagulant injected based on the fractal dimension of the flocs.

[0010] In Patent Document 3, the coagulant is injected so that the ratio of the turbidity of the raw water to the metal ion concentration of the membrane filtration is equal to or greater than a certain level. Therefore, it is necessary to provide a washing unit that performs chemical washing using an acidic chemical, and periodic washing with the acidic chemical is also required. In contrast, in Patent Document 4, the amount of coagulant injected is controlled based on the fractal dimension of the coagulated flocs, so a washing unit for performing chemical washing is not required.

[0011] Patent Literature 4 discloses acquiring image data of flocs, using a prediction model trained by machine learning using the image data of the flocs and a known fractal dimension as training data, and outputting a measurement value of the fractal dimension of the flocs from the image data of the flocs using this prediction model. However, because a convolutional neural network (CNN) based on artificial intelligence, such as Alex net, is used to create the trained model for machine learning, the details of the weighting of each piece of information obtained from the image data are black-boxed and incomprehensible to humans. In other words, the criteria for determining whether the prediction model and the output fractal dimension are appropriate are black-boxed. Therefore, in fields such as medicine and drinking water, where an incorrect judgment can have a significant impact, it is problematic that it is not possible to explain to users whether the judgment criteria were appropriate.

[0012] Japanese Patent No. 6730467 Japanese Patent No. 6262092 Japanese Patent Application Laid-Open No. 2020-131055 Japanese Patent Application Laid-Open No. 2022-64136

[0013] An object of the present invention is to provide a water treatment system and a water treatment method that are capable of flocculating and directly filtering the water to be treated by membrane filtration, and that allow the prediction formula used to derive the fractal dimension of the flocculants to be subsequently verified.

[0014] According to one aspect of the present invention, a water treatment system includes an aggregate forming unit that forms aggregates in the water to be treated by adding a coagulant to the water, an image acquiring unit that acquires multiple images of the aggregates formed in the aggregate forming unit, an information derivation unit that derives numerical information about feature quantities related to the aggregation state of the aggregates from the multiple images acquired by the image acquiring unit, a fractal dimension prediction unit that applies the feature quantities included in the numerical information derived by the information derivation unit to a prediction formula stored in a memory unit to derive a fractal dimension value for the aggregates, an addition amount adjustment unit that adjusts the amount of coagulant added to the water to be treated based on the fractal dimension value derived by the fractal dimension prediction unit, and a membrane filtration unit that performs membrane filtration on the water to be treated containing the aggregates. The numerical information includes digitized information for capturing the aggregates in three dimensions based on the multiple images of the aggregates.

[0015] According to another aspect of the present invention, a water treatment method includes an aggregate formation step of forming aggregates in the water to be treated by adding a coagulant to the water to be treated, an image acquisition step of acquiring a plurality of images of the aggregates formed in the aggregate formation step, an information derivation step of deriving numerical information on feature quantities related to the aggregation state of the aggregates from the plurality of images acquired in the image acquisition step, a fractal dimension prediction step of deriving a fractal dimension value of the aggregates by applying the feature quantities included in the numerical information derived in the information derivation step to a prediction formula stored in a memory unit, an addition amount adjustment step of adjusting an addition amount of coagulant to the water to be treated based on the fractal dimension value derived in the fractal dimension prediction step, and a membrane filtration step of membrane-filtering the water to be treated containing the aggregates. The numerical information includes digitized information for capturing the aggregates three-dimensionally based on the plurality of images of the aggregates.

[0016] 1 is a diagram illustrating a schematic configuration of a water treatment system according to an embodiment. 2 is a diagram illustrating a schematic configuration of a membrane filtration unit provided in the water treatment system. 3 is a characteristic diagram illustrating a relationship between a fractal dimension and an irreversible fouling index in (a) and a relationship between a fractal dimension and a reversible fouling index in (b). 4 is a diagram illustrating a water treatment method performed by the water treatment system.

[0017] Hereinafter, embodiments will be described with reference to the accompanying drawings. Note that the following embodiments are examples of specific embodiments of the present invention and are not intended to limit the technical scope of the present invention.

[0018] <Water Treatment System> First, the configuration of a water treatment system 1 according to this embodiment will be described with reference to Figures 1 and 2. The water treatment system 1 is used to obtain treated water by adding a flocculant to the water to be treated to flocculate impurities in the water, and then filtering the water containing the impurity flocculates through a membrane. Examples of the water to be treated include, but are not limited to, river water, lake water, groundwater, and industrial wastewater.

[0019] As shown in FIG. 1, the water treatment system 1 mainly comprises a coagulation tank 20 for forming coagulants in the water to be treated, a coagulant addition unit 10 for introducing a coagulant into the coagulation tank 20, a membrane filtration unit 50 for membrane filtering the water to be treated containing the coagulants, and a controller 30.

[0020] The coagulation tank 20 is a tank for coagulating impurities such as organic matter and fine particles contained in the water to be treated. The water to be treated before coagulation treatment is introduced into the coagulation tank 20. The coagulation tank 20 is provided with an agitation mechanism 21 for agitating the water to be treated stored in the coagulation tank 20.

[0021] The flocculant addition unit 10 is a unit for adding a flocculant to the water to be treated in the coagulation tank 20. The flocculant addition unit 10 includes a flocculant source 11, an addition line 12 connecting the flocculant source 11 and the coagulation tank 20, and a pump 13 installed on the addition line 12. When the flocculant is added to the water to be treated by the flocculant addition unit 10, impurities contained in the water to be treated coagulate in the coagulation tank 20 to form coagulates (flocs). In other words, a large number of floc particles gather to form coagulates. The flocculant addition unit 10 and the coagulation tank 20 function as a coagulate formation unit that forms coagulates in the water to be treated by adding a coagulant to the water to be treated.

[0022] The pump 13 is configured to be adjustable in rotation speed, and the rotation speed is adjusted by a command from the controller 30 (addition amount control unit 34), thereby adjusting the addition amount of flocculant sent to the flocculation tank 20 through the flocculant addition unit 10.

[0023] The type of flocculant is not particularly limited, but examples that can be used include polyaluminum chloride, aluminum sulfate, ferric chloride, ferric sulfate, and polysilica iron. In the flocculation tank 20, the water to be treated after the addition of the flocculant is stirred by the stirring mechanism 21, thereby forming flocs (flocculated flocs) in the water to be treated. The flocculation tank 20 may further be provided with a pH adjuster adding section that adds a predetermined pH adjuster.

[0024] A connection line L1 is connected to the coagulation tank 20. The other end of the connection line L1 is connected to the membrane filtration section 50. The water to be treated, which contains the coagulation obtained in the coagulation tank 20, is sent to the membrane filtration section 50 through the connection line L1. In other words, the coagulation tank 20 is installed upstream of the membrane filtration section 50, and no settling tank for settling the coagulated flocs is installed between the coagulation tank 20 and the membrane filtration section 50.

[0025] A collection line L2 is connected to the connection line L1 to collect a portion of the water to be treated that flows through the connection line L1. The water to be treated that flows through the collection line L2 is returned to the coagulation tank 20.

[0026] A monitoring unit 24 is provided in the collection line L2. The monitoring unit 24 is made of a transparent material, and allows the state of the treated water flowing through the collection line L2 to be observed from outside. The monitoring unit 24 forms a flow path with, for example, a rectangular cross section, and has a front surface 24a and a rear surface 24b that face each other in one direction. The front surface 24a and the rear surface 24b are each made of a transparent flat plate and are spaced a predetermined distance apart in the one direction.

[0027] A camera 27 is disposed outside the front surface 24a, and markers 25a and 25b are provided on the front surface 24a and the rear surface 24b, respectively. These markers 25a and 25b are disposed so as to fall within the photographing range of the camera 27 when the camera 27 photographs the water to be treated in the monitoring unit 24 from the front surface 24a side. The camera 27 constitutes an image acquisition unit that acquires multiple images of the aggregates.

[0028] The camera 27 photographs the monitoring unit 24 in the one direction, and the photographed image is stored in the controller 30. At this time, when the camera 27 focuses on a floc particle or aggregate in the water to be treated, information about the degree of focus on the markers 25a and 25b on the front and back sides is also stored in the controller 30 for each photographed image.

[0029] The membrane filtration unit 50 performs membrane filtration of the water to be treated that contains aggregates, and is configured, for example, by an external pressure type hollow fiber membrane module. Note that the membrane filtration unit 50 is not limited to a hollow fiber membrane module.

[0030] Fig. 2 schematically shows the internal configuration of the membrane filtration unit 50. As shown in Fig. 2, the membrane filtration unit 50 includes a hollow fiber membrane bundle 53 having a plurality of bundle-shaped hollow fiber membranes 52, and a housing 51 that accommodates the hollow fiber membrane bundle 53.

[0031] The housing 51 is a hollow cylindrical container that is long in the vertical direction and includes a cylindrical side portion 51B, an upper surface portion 51A that closes the upper end opening of the side portion 51B, and a lower surface portion 51C that closes the lower end opening of the side portion 51B. A fixing member 54 is provided inside the housing 51. By fixing the fixing member 54 to the side portion 51B, the space inside the housing 51 is liquid-tightly divided into two spaces. As described below, the water to be treated that has flowed through the connection line L1 is introduced into the first space S1 located below the fixing member 54. As described below, the treated water that has passed through the hollow fiber membrane 52 flows into the second space S2 located above the fixing member 54.

[0032] The lower surface 51C of the housing 51 is provided with an inlet port 46 connected to the downstream end of the connection line L1. Therefore, the water to be treated is introduced into the first space S1 through the inlet port 46. The side surface 51B is provided, just above the lower surface 51C, with a primary air inlet 45 for introducing air and a drain outlet 47 for draining water. Therefore, by introducing air into the first space S1 through the primary air inlet 45, bubbling cleaning of the hollow fiber membranes 52 can be performed. Furthermore, water can be drained from the first space S1 through the drain outlet 47.

[0033] The upper surface portion 51A is provided with an outlet port 55 for discharging the treated water from the second space S2. The upstream end of the delivery line L3 is connected to the outlet port 55. The treated water that has been membrane filtered in the membrane filtration portion 50 is delivered to a predetermined destination through the delivery line L3.

[0034] The delivery line L3 is provided with a secondary air inlet 48. When backwashing the hollow fiber membrane bundle 53, air is introduced into the second space S2 through the secondary air inlet 48 from an air compressor (not shown).

[0035] The hollow fiber membrane bundle 53 is a one-end-free type that is suspended from a fixing member 54 and placed in the first space S1. That is, each hollow fiber membrane 52 has an open upper end 52B and a sealed lower end 52A, and while the upper end 52B is fixed to the fixing member 54, the lower end 52A is not fixed and is free. In the one-end-free type hollow fiber membrane bundle 53, each hollow fiber membrane 52 can swing independently, making it less likely for turbid components to accumulate. Note that, for example, epoxy resin or urethane resin can be used as the resin for sealing the hollow fiber membranes 52. Note that the hollow fiber membrane bundle is not limited to the one-end-free type, and may also be a both-end-fixed type in which both ends are fixed.

[0036] The upper end 52B opens into the second space S2. Therefore, the space inside the hollow fiber membranes 52 is connected to the second space S2. Therefore, the water to be treated that contains aggregates and flows into the first space S1 flows into the internal space of the hollow fiber membranes 52 through the membrane holes of the hollow fiber membranes 52. Since the aggregates are filtered at this time, the treated water after the aggregates have been removed flows into the second space S2. In other words, the hollow fiber membrane bundle 53 is of an external pressure filtration type.

[0037] The controller 30 is composed of a microcomputer including a CPU that executes arithmetic processing, a ROM that stores processing programs and data, and a RAM that temporarily stores data, etc. By executing the processing program, the controller 30 functions as an information derivation unit 31, a storage unit 32, a fractal dimension prediction unit 33, and an addition amount control unit 34.

[0038] The information derivation unit 31 derives numerical information about feature quantities related to the aggregation state of the aggregates from multiple images captured by the camera 27. Specifically, feature quantities related to the aggregation state include, for example, the shape, size, density, number, and color of the aggregates or floc particles. This numerical information about feature quantities is information that numerically represents various feature quantities that are input into a prediction formula (described below) when deriving the value of the fractal dimension of the aggregate. Examples of feature quantities include the particle diameter of the floc particles, the size of the aggregate, the position of the areal center of gravity of the floc particles that make up the aggregate, the number of floc particles, the color intensity of the floc particles, and the edge length of the floc particles. However, it is not necessary for all of these feature quantities to be included in the numerical information; it is sufficient if at least some of the feature quantities are included.

[0039] Furthermore, the feature quantities included in the numerical information are associated with the focus of the markers 25a and 25b so that the aggregate can be captured three-dimensionally. That is, in each image captured at different focal positions in the depth direction, information about the focus of the marker 25a on the front surface 24a and information about the focus of the marker 25b on the rear surface 24b are associated with information indicating the feature quantities included in the numerical information. Each image is captured at a different focal position in the depth direction. Therefore, by comparing the numerical information about the feature quantities obtained in each image, it is possible to capture the feature quantities related to the aggregation state of the aggregate three-dimensionally. Note that each image captured at different focal positions in the depth direction does not represent images of aggregates in different aggregation states, but represents images of a single aggregate in a certain aggregation state. In other words, each image represents multiple images of an aggregate having a certain fractal dimension. Because the fractal dimension of an aggregate varies depending on the aggregation state, multiple images are obtained of an aggregate in the same aggregation state so that the fractal dimension of the aggregate can be identified. However, images of aggregates in different aggregation states may also be prepared so that feature amounts can be compared between aggregates in different aggregation states (aggregates having different fractal dimensions).

[0040] It can be said that information about the degree of focus on each marker 25a, 25b is substituted by information about the focal position of the camera 27 in the depth direction. Therefore, the amount of change in particle diameter of floc particles between images when multiple images are combined together, taken when the focus is adjusted to different depth positions, can be used as information for three-dimensionally grasping the aggregate in terms of its size. Furthermore, the amount of change in size of the aggregate between multiple images when the focus is adjusted to different depth positions can be used as information for three-dimensionally grasping the aggregate in terms of its size. Furthermore, the amount of change in position of the areal center of gravity of the floc particles between multiple images when the focus is adjusted to different depth positions can be used as information for three-dimensionally grasping the aggregate in terms of its size and shape. Furthermore, the amount of change in a numerical value representing the color intensity of the floc particles between multiple images when the focus is adjusted to different depth positions can be used as information for three-dimensionally grasping the aggregate in terms of its shape and density. Furthermore, the edge length (perimeter) of the floc particles between multiple images when focused at different positions in the depth direction can be used as information to capture the aggregates in three dimensions in terms of their shape, size, and density.

[0041] Furthermore, feature quantities relating to the aggregation state may be compared between aggregates in different aggregation states. For example, the number of floc particles may be compared between aggregates in different aggregation states. In this case, too, the information can be used to three-dimensionally grasp each of the aggregates in different aggregation states. Similarly, the shape, size, density, or color of the aggregates in different aggregation states may be compared.

[0042] A prediction formula for deriving the value of the fractal dimension of an aggregate is stored in the storage unit 32. This prediction formula is used to calculate the fractal dimension of the aggregate from the feature amount derived by the information derivation unit 31.

[0043] The prediction formula was obtained through preliminary experiments. In the preliminary experiments, water to be treated containing flocculants of various levels of fractal dimension was prepared by changing the flocculation conditions, such as the amount of flocculant added and pH. The water to be treated containing the flocculants was then photographed using an image processing device and a camera (OMRON FH-2050), and the feature values ​​obtained from the image data were quantified. At this time, image processing built into the image processing device may be applied to emphasize the feature values. Possible image processing methods that can be used at this time include specifying the processing range, expanding floc particles, contracting floc particles, and color adjustment filter processing.

[0044] "Fractal dimension" is a concept that represents the overall density of flocs. Water Research, 114, (2017), 88-103, examines the relationship between the effective density of flocs and floc size when kaolin and polyaluminum chloride are used. As the size of the flocs increases, the effective density decreases, and as the size of the flocs decreases, the effective density increases. Fractal dimension is a concept that represents the slope of the relationship between the size of the flocs and the effective density. When the fractal dimension is large, flocs with high effective density are likely to be formed regardless of the particle size. On the other hand, when the fractal dimension is small, flocs with low effective density are likely to be formed regardless of the particle size.

[0045] The fractal dimension of the water to be treated (reference fractal dimension) may be measured using a particle size distribution measuring device. Specifically, the particle size distribution measuring device irradiates the water to be treated with laser light having a wavelength of 650 nm or 405 nm, and measures the scattering intensity I and the scattering vector q of small-angle forward scattering at that time. The scattering vector q of small-angle forward scattering is expressed by the following equation (1): q = (2π / λ) sin(θ / 2) ... (1) where λ is the wavelength of the incident light and θ is the scattering angle. The scattering intensity I and the scattering vector q of small-angle forward scattering have a relationship expressed by the following equation (2): I(q) ∝ q - D ... (2) Therefore, the scattering intensity I and the scattering vector q can be plotted on a double logarithmic graph, and the slope (D) obtained when linearly approximating the scattering intensity I and the scattering vector q can be calculated as the reference fractal dimension.

[0046] Once a sufficient amount of image data of the water to be treated, associated with a reference fractal dimension value, has been prepared, a prediction formula is created by associating feature quantities obtained from the image data with the calculated fractal dimension. The feature quantities used here can be obtained from image data of the original image or image data of an image that has undergone at least one type of processing. Examples of feature quantities include the size (area) of the aggregates (aggregated flocs), the position of the area center of gravity of the floc particles, the number of floc particles of a given size, the color and its shade or deviation of shade of the floc particles, the edge length or color and its shade of the floc particles, etc. Furthermore, parameters obtained by performing arithmetic operations using at least one of these feature quantities may also be used to create the prediction formula. Of the feature quantities, it is preferable to use a combination of the size (area), number of aggregates, and color shade of the aggregates obtained from the image data and the image data that has undergone some processing. More preferably, the number of aggregates having a predetermined size, the number of aggregates having a predetermined size in an image (processed image) after expansion of the aggregates, the number of aggregates having a predetermined size in an image (processed image) after contraction of the aggregates, the color intensity of the aggregates, the color intensity of the aggregates in an image (processed image) after expansion of the aggregates, and the color deviation of the aggregates are preferably used. However, these feature quantities need to be associated with information indicating the degree of focus on the markers 25a and 25b. Furthermore, to summarize the feature quantities into a controllable number of digits, each feature quantity may be multiplied by N or 1 / N (N is a natural number) and used to create the prediction formula. The prediction formula may be expressed, for example, as the number of floc particles × coefficient (or order) + size of floc particle × coefficient (or order) + color intensity of floc particle × coefficient (or order) + focus 1 × coefficient (or order) + focus 2 × coefficient (or order) + offset value. That is, the prediction formula only needs to clearly state the relationship between the fractal dimension and the quantified feature amount using at least one of a coefficient and an order that represent the weighting of the data.

[0047] Furthermore, since the creation of a prediction formula requires the processing of a huge amount of information, machine learning software or the like may be used. Specifically, free software such as Rapid Miner Studio may be used. In creating a prediction formula, it is often more efficient to use machine learning to process big data.

[0048] The fractal dimension prediction unit 33 applies the feature amount included in the numerical information derived by the information derivation unit 31 to a prediction formula to derive the value of the fractal dimension for the aggregate.

[0049] The addition amount control unit 34 controls the pump 13 based on the value of the fractal dimension predicted by the fractal dimension prediction unit 33. The addition amount control unit 34 and the pump 13 constitute an addition amount adjustment unit 37 that adjusts the amount of flocculant added to the coagulation tank 20.

[0050] The storage unit 32 may store correlation data between the fractal dimension of the aggregates and the reversible fouling and irreversible fouling when membrane filtration is performed in the membrane filtration unit 50. For example, the storage unit 32 may store correlation data between the fractal dimension and the irreversible fouling index as shown in (a) of Figure 3, or may store correlation data between the fractal dimension and the reversible fouling index as shown in (b) of Figure 3.

[0051] The irreversible fouling index is the change per hour in the transmembrane pressure detected at the earliest timing after repeated physical cleaning during membrane filtration operation. In other words, it is an index that represents the change in the transmembrane pressure immediately after each repeated cleaning of the hollow fiber membrane. This index is based on the transmembrane pressure measured immediately after the SS components accumulated on the membrane surface are removed by physical cleaning, and therefore indicates the transmembrane pressure when the SS components accumulated on the membrane surface are at their lowest. In other words, it does not indicate the transmembrane pressure due to the fouling components adhering to the membrane surface, but rather indicates the degree of clogging of the membrane pores themselves. The irreversible fouling index is preferably 0 Pa / h or more and 88 Pa / h or less, more preferably 0.8 Pa / h or more and 88 Pa / h or less, and most preferably 0.8 Pa / h or more and 44 Pa / h or less. Under conditions where clogging of 88 Pa / h or more occurs, stable operation over a long period (several months or years) is impossible. On the other hand, the less clogging of the membrane, the more stable the operation, and it is most preferable that there is no clogging (0 Pa / h).

[0052] The transmembrane pressure (TMP) of the hollow fiber membrane module is obtained from the difference between the pressure detected by the primary pressure sensor P1, which is provided in the connection line L1 and detects the primary pressure of the membrane filtration section 50, and the pressure detected by the secondary pressure sensor P2, which is provided in the delivery line L3 and detects the secondary pressure of the membrane filtration section 50.

[0053] On the other hand, the reversible fouling index is the change in weight (g / h) (mass change) of SS components accumulating in the hollow fiber membrane bundle 53 during membrane filtration operation. Specifically, the change in weight of SS components adhering to and accumulating in the hollow fiber membrane bundle 53 during continuous filtration operation under actual operating conditions is measured, and the change per unit time is quantified. The reversible fouling index is preferably less than 33 g / h and 0 g / h or more, more preferably less than 33 g / h and 0.8 g / h or more, and most preferably less than 25 g / h and 0.8 g / h or more. Under conditions where accumulation of 33 g / h or more occurs, stable operation over a long period (several months or years) is impossible. On the other hand, the lower the accumulation, the more stable operation is possible, and no accumulation (0 kg) is most preferable. The reversible fouling index does not necessarily have to be obtained by measuring the change in the mass of the module over time; for example, another correlated index, such as the change in accumulation amount predicted from log data, may be used.

[0054] The addition amount control unit 34 may control the pump 13 to obtain a value of the fractal dimension of the aggregates such that the irreversible fouling index falls within a range of 0 Pa / h or more and less than 88 Pa / h, and the reversible fouling index falls within a range of 0 g / h or more and less than 33 g / h. That is, the fractal dimension may be controlled so that the value is preferably 1.9 or more and 2.2 or less, and most preferably 1.93 or more and 2.13 or less. In other words, it has been found that both the reversible fouling index and the irreversible fouling index are highly correlated with the fractal dimension of the aggregates. Therefore, by using the fractal dimension of the aggregates as an index, operation within preferred ranges of the reversible fouling index and the irreversible fouling index is possible.

[0055] <Water Treatment Method> Next, a water treatment method using the water treatment system 1 according to this embodiment will be described with reference to FIG.

[0056] First, a flocculant is added to the water to be treated containing impurities to form flocs (floc formation step, step ST10). Specifically, the pump 13 of the flocculant addition unit 10 is operated at a predetermined rotation speed to add the flocculant to the water to be treated in the flocculation tank 20. The water to be treated is then stirred by the stirring mechanism 21, causing the impurities to flocculate and forming flocs (flocs) in the water to be treated.

[0057] The water to be treated, in which flocculants have formed in the flocculation tank 20, is sent through a connecting line L1 to the membrane filtration unit 50. This water to be treated is subjected to membrane filtration in the membrane filtration unit 50 (membrane filtration step, step ST20). The treated water filtered in the membrane filtration unit 50 is sent to a demand destination through a delivery line L3. In this method, after a flocculant is added to the water to be treated, the water to be treated is directly subjected to membrane filtration without performing a flocculation precipitation operation.

[0058] A portion of the water to be treated flowing through the connection line L1 flows into the collection line L2. In the monitoring unit 24 of the collection line L2, the camera 27 acquires image data of the water to be treated flowing through the monitoring unit 24 (image acquisition step, step ST31). This image data is sent to the controller 30. At this time, the camera 27 captures multiple images of the same aggregate by changing the focal length.

[0059] Information derivation unit 31 of controller 30 derives numerical information about feature quantities related to the shape, size, and density of the aggregates from a plurality of images among the acquired image data (information derivation step, step ST32). The derived numerical information is stored in controller 30. The work of deriving the numerical information in step ST32 may be performed manually, or controller 30 may be configured to derive the numerical information from the images.

[0060] The numerical information is information that indicates various feature quantities numerically, and the feature quantities include, for example, the particle diameter of the floc particles, the size of the aggregate, the position of the area center of gravity of the floc particles that make up the aggregate, the number of floc particles, the color density of the floc particles, the edge length of the floc particles, etc. These feature quantities are derived in association with the degree of focus on both markers 25 a and 25 b.

[0061] The fractal dimension prediction unit 33 applies the feature quantities contained in the numerical information derived by the information derivation unit 31 to the prediction formula stored in the memory unit 32 to derive the value of the fractal dimension of the aggregate (fractal dimension prediction step, step ST33).

[0062] Based on the fractal dimension value of the flocculant derived in step ST33, the addition amount control unit 34 adjusts the amount of flocculant added to the water to be treated by the flocculant adding unit 10 (addition amount adjustment step, step ST34). In step ST34, the addition amount control unit 34 controls the pump 13 of the flocculant adding unit 10 so that the fractal dimension value of the flocculant becomes a value that falls within a range in which the reversible fouling index value when membrane filtration is performed in the membrane filtration unit 50 is equal to or greater than 0 g / h and less than 33 g / h, and the irreversible fouling index value when membrane filtration is performed in the membrane filtration unit 50 is equal to or greater than 0 Pa / h and less than 88 Pa / h.

[0063] Then, after a predetermined filtration time has elapsed (YES in step ST40), the filtration operation is temporarily stopped, and the membrane filtration section 50 is backwashed (step ST41). In step ST41, the hollow fiber membrane module is backwashed by pressurizing the air from the secondary side. Specifically, air is introduced into the housing 51 through the secondary air inlet 48 (FIG. 2), and the treated water in the second space S2 is pushed into the first space S1 by air pressurization. In this way, using air for backwashing not only increases the water recovery rate and reduces running costs, but also enhances the cleaning effect (the effect of removing SS from the membrane surface). Thereafter, bubbling cleaning or the like is performed in the first space S1, and then the filtration operation is resumed.

[0064] As described above, in this embodiment, the prediction formula to which the numerical information derived by the information derivation unit 31 is applied is not a black box. Therefore, when later verifying the derived fractal dimension value and the amount of flocculant to be added, it is possible to verify which part of the prediction formula had a problem. Moreover, the numerical information on the feature quantities related to the aggregation state of the aggregates generated by the information derivation unit 31 includes quantified information for capturing the aggregate three-dimensionally based on multiple images of the aggregate. This numerical information on the feature quantities is used by the fractal dimension prediction unit 33 when deriving the fractal dimension value of the aggregate. Therefore, even if only two-dimensional information is obtained from each of the multiple images acquired by the camera 27, it is possible to capture the aggregate three-dimensionally by correlating the information obtained from these multiple images. Therefore, the fractal dimension prediction unit 33 can derive the fractal dimension value based on the aggregate captured three-dimensionally, and the fractal dimension value corresponding to the aggregation state of the aggregate becomes a value closer to the three-dimensional shape of the aggregate.

[0065] Furthermore, in this embodiment, information indicating the size of the aggregate, etc., is associated with information indicating the degree of focus of two markers 25a, 25b arranged at different distances from camera 27. That is, by referring to the information indicating the degree of focus of two markers 25a, 25b, information indicating the position in focus in the depth direction is obtained, and this information is used to obtain information indicating the size of the aggregate at this position. Therefore, information for three-dimensionally capturing the size, etc. of the aggregate can be obtained from each image captured at different focal positions in the depth direction. For example, information for three-dimensionally capturing the size, etc. of the aggregate can be obtained from each image captured at different focal lengths in the depth direction.

[0066] In this embodiment, the weighting of the feature quantities is expressed by coefficients or orders in the prediction formula, so that when the derived fractal dimension value and the amount of flocculant to be added are verified later, it is possible to verify which coefficient or order in the prediction formula had a problem.

[0067] In this embodiment, the addition amount control unit 34 controls the pump 13. At this time, the fractal dimension value of the aggregates is controlled so that the reversible fouling index value during membrane filtration is greater than or equal to 0 g / h and less than 33 g / h, and the irreversible fouling index value falls within a range of greater than or equal to 0 Pa / h and less than 88 Pa / h. Therefore, reversible fouling and irreversible fouling can be suppressed even when the treated water containing aggregates is directly subjected to membrane filtration. In other words, the amount of accumulation of suspended solids (SS) components on the membrane surface of the membrane module in the membrane filtration unit 50 can be suppressed, and clogging of the membrane of the membrane module can be suppressed.

[0068] Example 1 Here, an example will be described in which water to be treated was actually treated and a prediction formula was created.

[0069] In Example 1, surface water from the Takahashi River was used as the water to be treated. First, water with various fractal dimensions was prepared as the water to be treated by changing the coagulation conditions, such as the amount of coagulant added and pH. Specifically, the amount of coagulant added to the water to be treated in the coagulation tank 20 was changed by changing the rotation speed of the pump 13 in the coagulant addition unit 10. Alternatively, the pH was changed by changing the amount of pH adjuster added to the water to be treated in the coagulation tank 20 by changing the rotation speed of the pump in the pH adjustment unit. The water to be treated was then stirred by the stirring mechanism 21 to form coagulates (coagulated flocs) under each coagulation condition in the water.

[0070] Next, the fractal dimension, which serves as a reference for each aggregate, was measured for the water to be treated containing aggregates under each aggregation condition using a particle size distribution analyzer. Specifically, a SALD-7500 nano (manufactured by Shimadzu Corporation) was used as the particle size distribution analyzer. This analyzer can measure the angle and scattered light intensity of the light scattered when a laser beam is irradiated onto aggregated flocs contained in the water to be treated. From this data, the fractal dimension of the aggregates under each aggregation condition can be calculated. Since this calculation result can be said to be the true fractal dimension, it is used as the fractal dimension (reference).

[0071] Furthermore, for the same water to be treated as that used to measure the reference fractal dimension, images of aggregates contained in the water to be treated were taken with a camera, and the feature quantities of the aggregates obtained from the captured image data were quantified. Specifically, the water to be treated was photographed using an OMRON FH-2050 camera. At this time, the same aggregate was photographed at multiple different focal positions in the depth direction, and multiple images of the aggregate were obtained. The feature quantities of the aggregates obtained from the captured image data were then quantified.

[0072] The parameters representing the feature quantities of the aggregates were recorded as follows: size of the aggregate, the number of aggregates having a predetermined size, the number of floc particles, the position of the areal center of gravity of the floc particles, the number of aggregates having a predetermined size in an image (processed image) of the aggregate (aggregated floc) after expansion processing, the number of aggregates having a predetermined size in an image (processed image) of the aggregate after contraction processing, the color intensity of the aggregate, the color intensity of the aggregate in an image (processed image) of the aggregate after expansion processing, the color deviation of the aggregate, the edge length or color or shade or deviation of the floc particle, and the focus relative to markers 25a and 25b. When creating a prediction formula using at least a portion of these recorded parameters, each parameter value was multiplied by 1 / 1000.

[0073] Next, a linear regression equation (prediction equation) was created based on the data obtained above using a machine learning trained model. Specifically, using free software Rapid Miner Studio, the fractal dimension (reference) of the treated water was linked to each of the parameters obtained from the images of the treated water for all of the prepared treated water and input. Based on the input data, the software created a linear regression equation (prediction equation) that estimates the fractal dimension from the feature amount of the aggregates.

[0074] As a result, the following correlation equation 1 was obtained.

[0075]

[0076] Here, "number" refers to the number of floc particles whose cross-sectional area in a cross section including the focal position at the time of image capture is the number of pixels indicated (for example, "number of 6" is 6 pixels). Note that, because the parameter value is multiplied by 1 / 1000 when creating the prediction formula, if the number of floc particles in the image is 1000, the number of floc particles in the prediction formula will be 1. "Expansion" refers to an image that has been expanded, and "contraction" refers to an image that has been contracted. "Focus 1, 2" refers to the focus relative to markers 25a and 25b, "average density" refers to the average density of the color of the aggregates, and "average deviation" refers to the average deviation of the color of the aggregates. For example, "expansion number of 6" refers to the number of floc particles whose cross-sectional area in a cross section including the focal position at the time of image capture is 6 pixels in the expanded image.

[0077] When the error between the fractal dimension obtained by this correlation equation 1 and the fractal dimension (standard) measured using a particle size distribution measuring device was examined, the RSME (root mean square error) was 0.028±0.002. In other words, prediction was possible with an error of 0.1 or less.

[0078] Example 2 A more accurate fractal dimension prediction formula was created using parameters representing more feature quantities from the data used in Example 1. As a result, the following correlation formula 2 was obtained.

[0079]

[0080] Here, "number" refers to the number of floc particles whose cross-sectional area in a cross section including the focal position at the time of image capture is the stated number of pixels (for example, "number of 6" is 6 pixels), "expansion" refers to the image after expansion processing, "contraction" refers to the image after contraction processing, "area" refers to the cross-sectional area of ​​the floc particles in a cross section including the focal position at the time of image capture, "density" refers to the color intensity of the aggregate, "focus 1, 2" refers to the focus relative to markers 25a, 25b, "average density" refers to the average color intensity of the aggregate, and "average deviation" refers to the average deviation of the color of the aggregate. For example, "number of expanded 6" refers to the number of floc particles whose cross-sectional area in a cross section including the focal position at the time of image capture is 6 pixels in the expanded image.

[0081] When the error between the fractal dimension obtained by this correlation equation 2 and the fractal dimension (standard) measured using a particle size distribution measuring device was examined, the RSME (root mean square error) was 0.015±0.001. In other words, prediction was possible with an error of 0.1 or less.

[0082] Example 3 In Example 3, industrial water received from the Kashima Plant of Kuraray Co., Ltd. was used as the water to be treated. First, flocculation conditions such as the amount of flocculant added and pH were changed to prepare water with various fractal dimensions as the water to be treated. Specifically, raw water was collected and measured in a laboratory, and the pH was changed by changing the amount of flocculant added to the raw water or by changing the amount of pH adjuster added. The water to be treated was then stirred using the stirring mechanism 21 to form flocs (flocs) in the water under each flocculation condition.

[0083] Next, the fractal dimension (reference) of each aggregate was measured for the water to be treated containing aggregates under each aggregation condition using a particle size distribution analyzer. Specifically, a SALD-7500 nano (manufactured by Shimadzu Corporation) was used as the particle size distribution analyzer. This device can measure the angle and scattered light intensity of the light scattered when a laser beam is irradiated onto aggregates contained in the water to be treated. From this data, the fractal dimension of the aggregates under each aggregation condition can be calculated. Since this calculation result can be said to be the true fractal dimension, it is used as the reference fractal dimension.

[0084] Furthermore, for the same water to be treated as that used to measure the reference fractal dimension, images of aggregates contained in the water to be treated were taken with a camera, and the feature quantities of the aggregates obtained from the captured image data were quantified. Specifically, the water to be treated was photographed using an OMRON FH-2050 camera. At this time, the same aggregate was photographed at multiple different focal positions in the depth direction, and multiple images of the aggregate were obtained. The feature quantities of the aggregates obtained from the captured image data were then quantified.

[0085] The parameters representing the feature amounts of the aggregates were the number of aggregates having a predetermined size (e.g., 4 pixels, 6 pixels, 8 pixels, 10 pixels, etc.), the number of aggregates having a predetermined size in the image after expansion processing, the number of aggregates having a predetermined size in the image after contraction processing, the color intensity of the aggregates, and the average color intensity of the aggregates. When creating a prediction formula using each of these recorded parameters, each parameter value was multiplied by 1 / 1000.

[0086] Next, a linear regression equation (prediction equation) was created based on the data obtained above using a machine learning trained model. Specifically, using the free software Rapid Miner Studio, the fractal dimension (reference) of the treated water was linked to each of the above parameters obtained from images of the treated water for all of the prepared treated water and input. Based on the input data, the software created a linear regression equation (prediction equation) that estimates the fractal dimension from the feature amount of the aggregates, and the following correlation equation 3 was obtained.

[0087]

[0088] Furthermore, both the data for the water to be treated obtained in Example 3 and the data for the water to be treated obtained in Example 1 were input into the software, and a prediction equation was created in the same manner. As a result, the following correlation equation 4 was obtained.

[0089]

[0090] Here, "number" refers to the number of floc particles whose cross-sectional area in a cross section including the focal position at the time of image capture is the number of pixels indicated (for example, "number of 6" is 6 pixels). "Expansion" refers to an image that has been expanded, and "contraction" refers to an image that has been contracted. "Area" refers to the cross-sectional area of ​​the floc particles in a cross section including the focal position at the time of image capture. "Density" refers to the color intensity of the aggregate, "focus 1, 2" refers to the focus relative to markers 25a and 25b, "average density" refers to the average color intensity of the aggregate, and "average deviation" refers to the average deviation of the color of the aggregate. For example, "expansion number of 6" indicates that a floc particle occupies 6 pixels in the expanded image.

[0091] When the error between the fractal dimension obtained by this correlation equation 4 and the true fractal dimension measured using a particle size distribution measuring device was examined, the RSME (root mean square error) was 0.080±0.014. In other words, prediction was possible with an error of 0.1 or less.

[0092] It should be noted that the embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The present invention is not limited to the above-described embodiments, and various modifications and improvements are possible without departing from the spirit of the present invention. For example, in the above-described embodiment, one camera 27 and two markers 25a, 25b are used to obtain information on the depth direction of the aggregate in the image. Alternatively, two cameras may be used to photograph the aggregate from different directions to obtain information on the depth direction of the aggregate.

[0093] Here, the above embodiment will be outlined.

[0094] (1) The water treatment system according to the embodiment includes an aggregate forming unit that forms aggregates in the water to be treated by adding a coagulant to the water, an image acquiring unit that acquires multiple images of the aggregates formed in the aggregate forming unit, an information derivation unit that derives numerical information about feature quantities related to the aggregation state of the aggregates from the multiple images acquired by the image acquiring unit, a fractal dimension prediction unit that applies the feature quantities included in the numerical information derived by the information derivation unit to a prediction formula stored in a memory unit to derive a fractal dimension value for the aggregates, an addition amount adjustment unit that adjusts the amount of coagulant added to the water to be treated based on the fractal dimension value derived by the fractal dimension prediction unit, and a membrane filtration unit that performs membrane filtration on the water to be treated containing the aggregates. The numerical information includes digitized information for capturing the aggregates in three dimensions based on the multiple images of the aggregates.

[0095] In the water treatment system, the prediction formula to which the numerical information derived by the information derivation unit is applied is not a black box. Therefore, when later verifying the derived fractal dimension value and the amount of flocculant added, it is possible to verify which part of the prediction formula was problematic. Moreover, the numerical information used by the fractal dimension prediction unit to derive the fractal dimension value for the aggregate, i.e., the numerical information on the feature quantities related to the aggregation state of the aggregate derived by the information derivation unit, includes quantified information for capturing the aggregate three-dimensionally based on multiple images of the aggregate. Therefore, even if only two-dimensional information is obtained from each of the multiple images acquired by the image acquisition unit, it is possible to capture the aggregate three-dimensionally by correlating the information obtained from these multiple images. Therefore, the fractal dimension prediction unit can derive the fractal dimension value based on the aggregate captured three-dimensionally. Therefore, the fractal dimension value corresponding to the aggregation state of the aggregate is closer to the three-dimensional shape of the aggregate.

[0096] (2) The image acquisition unit may include a camera that captures images of the water to be treated containing the aggregates, and the water treatment system may further include at least two markers that are located within the camera's image capture range and at different distances from the camera. In this case, the numerical information may include information indicating at least one feature value obtained from an original image of the water to be treated or a processed image obtained by processing the original image: the size of the aggregates, the position of the areal center of gravity of the floc particles that make up the aggregates, the number of the floc particles, the color or shade or deviation thereof of the floc particles, and the edge length, color, shade or deviation thereof of the floc particles. Furthermore, the information derivation unit may derive the numerical information from the multiple images of the aggregates captured by the camera at different focal lengths, and the at least one feature value in the numerical information may be associated with the degree of focus of the at least two markers.

[0097] In this aspect, when the numerical information includes information about the size of the aggregate as a feature, the feature is derived in association with information indicating the degree of focus of at least two markers positioned at different distances from the camera. That is, by referring to the information indicating the degree of focus of at least two markers, information indicating the position in focus in the depth direction is obtained. Therefore, information for three-dimensionally grasping the size, etc., of the aggregate can be obtained from each image captured at different focal positions in the depth direction. For example, when information indicating the size of an aggregate captured multiple times with different focal positions in the depth direction is referenced, this information can be used as information for three-dimensionally grasping the aggregate in terms of its size. Furthermore, when information about the position of the areal center of gravity of the floc particles constituting the aggregate is included, this information can be used as information for three-dimensionally grasping the floc particles in terms of their size and shape. Furthermore, the amount of change in color, shade, or deviation of the floc particles between images captured at different focal positions in the depth direction can be used as information for three-dimensionally grasping the aggregate in terms of its shape and density. Furthermore, the change in edge length (perimeter), color, shade, or deviation of a flock particle between images captured at multiple focal positions in the depth direction can be used as information for three-dimensionally grasping the shape, size, and density of the agglomerate. The number of flock particles may also be compared between agglomerates in different agglomeration states. Even in this case, the change in edge length (perimeter), color, or deviation of a flock particle may also be used as information for three-dimensionally grasping the agglomerate in different agglomeration states. The size of the agglomerate, the position of the center of gravity of the area of ​​the flock particle, the color of the flock particle, or the edge length of the flock particle may also be compared between agglomerates in different agglomeration states.

[0098] Furthermore, at least one of the following information may be referenced for a certain focal position: information indicating the size of the aggregate; information indicating the position of the areal center of gravity of the floc particles constituting the aggregate; information indicating the color or shade or deviation thereof of the floc particles; and information indicating the edge length or color or shade or deviation thereof of the floc particles. In this case, at least one other piece of information may be referenced for a different focal position. In this case, too, the information may be used as information for capturing the aggregate three-dimensionally.

[0099] (3) In the prediction formula, the relationship between the fractal dimension and the feature amount may be specified using at least one of a coefficient and an order that indicates weighting of data.

[0100] In this embodiment, the weighting of the features is expressed by coefficients and orders in the prediction formula, so when verifying the derived fractal dimension value and the amount of flocculant to be added later, it is possible to verify which coefficient or order in the prediction formula had a problem.

[0101] (4) The addition amount adjustment unit may adjust the addition amount of the flocculant so that the fractal dimension value of the aggregates is such that the reversible fouling index value when membrane filtration is performed in the membrane filtration unit is 0 g / h or more and less than 33 g / h, and the irreversible fouling index value when membrane filtration is performed in the membrane filtration unit is 0 Pa / h or more and less than 88 Pa / h.

[0102] In this embodiment, reversible fouling and irreversible fouling can be suppressed even when the treated water containing aggregates is directly filtered through a membrane, i.e., the amount of suspended solids (SS) components that accumulate on the membrane surface of the membrane module in the membrane filtration section can be suppressed, and clogging of the membrane of the membrane module can be suppressed.

[0103] (5) The reversible fouling index may be an index indicating the amount of change in mass of SS components accumulating on the membrane surface of the membrane module per unit time during membrane filtration by the membrane filtration section.

[0104] (6) The irreversible fouling index may be an index indicating the amount of change per unit time in the transmembrane pressure detected at the earliest timing after repeated physical cleaning of the membrane filtration section.

[0105] (7) The water treatment system may include a connection line that connects the flocculation formation unit and the membrane filtration unit to each other and does not include a settling tank for settling the flocs.

[0106] (8) The water treatment method according to the embodiment includes an aggregate formation step of forming aggregates in the water to be treated by adding a coagulant to the water to be treated, an image acquisition step of acquiring a plurality of images of the aggregates formed in the aggregate formation step, an information derivation step of deriving numerical information on feature quantities related to the aggregation state of the aggregates from the plurality of images acquired in the image acquisition step, a fractal dimension prediction step of deriving a fractal dimension value of the aggregates by applying the feature quantities included in the numerical information derived in the information derivation step to a prediction formula stored in a memory unit, an addition amount adjustment step of adjusting an addition amount of coagulant to the water to be treated based on the fractal dimension value derived in the fractal dimension prediction step, and a membrane filtration step of performing membrane filtration on the water to be treated containing the aggregates. The numerical information includes digitized information for capturing the aggregates three-dimensionally based on the plurality of images capturing the aggregates.

[0107] (9) In the image acquisition step, the plurality of images of the aggregates may be acquired using a camera that captures images of the water to be treated containing the aggregates and at least two markers that are positioned within the camera's image capture range and at different distances from the camera. The numerical information may include information indicating at least one feature value among the size of the aggregates, the position of the areal center of gravity of the floc particles constituting the aggregates, the number of the floc particles, the color or shade or deviation thereof of the floc particles, and the edge length, color, shade or deviation thereof of the floc particles, obtained from an original image of the water to be treated or a processed image obtained by processing the original image. In this case, in the information derivation step, the numerical information may be derived from the plurality of images captured by the camera at different focal lengths, and the at least one feature value may be associated with the degree of focus of the at least two markers.

[0108] (10) In the water treatment method, the prediction formula may specify the relationship between the fractal dimension and the feature quantity using at least one of a coefficient and an order representing weighting of data.

[0109] (11) In the addition amount adjustment step, the addition amount of the flocculant may be adjusted so that the fractal dimension value of the aggregates is such that the reversible fouling index value when membrane filtration is performed in the membrane filtration step is 0 g / h or more and less than 33 g / h, and the irreversible fouling index value when membrane filtration is performed in the membrane filtration step is 0 Pa / h or more and less than 88 Pa / h.

[0110] As explained above, direct membrane filtration of the treated water by flocculation is possible, and the fractal dimension of the flocs can be derived using a prediction formula that can be verified later.

[0111] This application is based on Japanese Patent Application No. 2024-015019 filed with the Japan Patent Office on February 2, 2024, the contents of which are incorporated herein by reference.

Claims

1. A water treatment system comprising: an aggregate formation unit that forms aggregates in the water to be treated by adding a coagulant to the water; an image acquisition unit that acquires multiple images of the aggregates formed in the aggregate formation unit; an information derivation unit that derives numerical information about features related to the aggregation state of the aggregates from the multiple images acquired by the image acquisition unit; a fractal dimension prediction unit that applies the features included in the numerical information derived by the information derivation unit to a prediction formula stored in a memory unit to derive a fractal dimension value for the aggregates; an addition amount adjustment unit that adjusts the amount of coagulant to be added to the water to be treated based on the fractal dimension value derived by the fractal dimension prediction unit; and a membrane filtration unit that performs membrane filtration on the water to be treated containing the aggregates, wherein the numerical information includes digitized information for capturing the aggregates in three dimensions based on the multiple images in which the aggregates are captured.

2. The water treatment system of claim 1, wherein the image acquisition unit includes a camera that photographs the water to be treated containing the agglomerates, the water treatment system further includes at least two markers that are provided within the range of photography by the camera and positioned at different distances from the camera, the numerical information includes information indicating at least one feature value among the size of the agglomerates, the position of the center of gravity of the area of the floc particles that make up the agglomerates, the number of the floc particles, the color or shade or deviation thereof of the floc particles, and the edge length or color or shade or deviation thereof of the floc particles, obtained from an original image of the water to be treated or a processed image obtained by processing the original image, and the information derivation unit derives the numerical information from the multiple images obtained when the agglomerates are photographed by the camera at different focal lengths, and in the numerical information, the at least one feature value is associated with the degree of focus of the at least two markers.

3. A water treatment system as described in claim 1 or 2, wherein the prediction formula specifies the relationship between the fractal dimension and the feature quantity using at least one of a coefficient and an order representing the weighting of the data.

4. A water treatment system as described in any one of claims 1 to 3, wherein the addition amount adjustment unit adjusts the addition amount of the coagulant so that the fractal dimension value of the agglomerates is such that the reversible fouling index value when membrane filtration is performed in the membrane filtration unit is 0 g / h or more and less than 33 g / h, and the irreversible fouling index value when membrane filtration is performed in the membrane filtration unit is in the range of 0 Pa / h or more and less than 88 Pa / h.

5. The water treatment system described in claim 4, wherein the reversible fouling index is an index that indicates the mass change of SS components that accumulate per unit time on the membrane surface of the membrane module during membrane filtration by the membrane filtration section.

6. A water treatment system as described in claim 4 or 5, wherein the irreversible fouling index is an index that indicates the amount of change per unit time in the transmembrane pressure difference detected at the earliest timing after repeated physical cleaning of the membrane filtration section.

7. A water treatment system as described in any one of claims 1 to 6, comprising a connecting line that connects the flocculation formation section and the membrane filtration section to each other and does not have a sedimentation tank installed for settling the flocs.

8. A water treatment method comprising: an aggregate formation step of forming aggregates in the water to be treated by adding a coagulant to the water; an image acquisition step of acquiring multiple images of the aggregates formed in the aggregate formation step; an information derivation step of deriving numerical information about features related to the aggregation state of the aggregates from the multiple images acquired in the image acquisition step; a fractal dimension prediction step of applying the features included in the numerical information derived in the information derivation step to a prediction formula stored in a memory unit to derive a fractal dimension value of the aggregates; an addition amount adjustment step of adjusting the amount of coagulant to be added to the water to be treated based on the fractal dimension value derived in the fractal dimension prediction step; and a membrane filtration step of membrane-filtering the water to be treated containing the aggregates, wherein the numerical information includes digitized information for capturing the aggregates in three dimensions based on the multiple images in which the aggregates are captured.

9. The water treatment method of claim 8, wherein in the image acquisition step, the multiple images of the aggregates are acquired using a camera that photographs the water to be treated containing the aggregates and at least two markers that are set within the camera's photography range and are positioned at different distances from the camera; the numerical information includes information indicating at least one feature value among the size of the aggregates, the position of the area center of gravity of the floc particles that make up the aggregates, the number of the floc particles, the color or shade or deviation thereof of the floc particles, and the edge length or color or shade or deviation thereof of the floc particles, obtained from the original image of the water to be treated or a processed image obtained by processing the original image; and in the information derivation step, the numerical information is derived from the multiple images photographed by the camera at different focal lengths; and in the numerical information, the at least one feature value is associated with the degree of focus of the at least two markers.

10. A water treatment method according to claim 8 or 9, wherein the prediction formula specifies the relationship between the fractal dimension and the feature quantity using at least one of a coefficient representing the weighting of data and an order.

11. A water treatment method described in any one of claims 8 to 10, wherein in the addition amount adjustment step, the amount of coagulant added is adjusted so that the fractal dimension value of the agglomerates is such that the reversible fouling index value when membrane filtration is performed in the membrane filtration step is 0 g / h or more and less than 33 g / h, and the irreversible fouling index value when membrane filtration is performed in the membrane filtration step is in the range of 0 Pa / h or more and less than 88 Pa / h.

Images