KR20230105496A - Automatic operation system and method of belt press type dehydrator using AI based on image analysis and machine learning - Google Patents

Abstract

An automatic operation system of a belt press type dehydrator using AI based on machine learning is disclosed. The disclosed automatic operation system includes a belt press type dehydrator composed of a belt dewatering unit in which mixed sludge is dehydrated while passing between a lower dewatering belt and an upper dewatering belt, and a gravity dewatering unit that supplies the mixed sludge to the lower dewatering belt; An image acquisition unit for acquiring a sludge image by photographing the mixed sludge and transmitting the obtained image to the AI control unit in real time; An agitation tank for receiving and mixing the sludge and the coagulant by the sludge pump and the coagulant pump, and supplying the mixed sludge to the gravity dewatering part; a sludge flowmeter installed on a first connection passage connecting the sludge pump and the stirring tank to detect a current input amount of sludge supplied; a coagulant flowmeter installed on a second connection passage connecting the coagulant pump and the stirring tank to detect a current input amount of the coagulant supplied; It consists of a VVVF (Variable Voltage Variable Frequency) inverter device for driving and controlling the sludge pump and the flocculant pump, and a PID controller for transmitting a pump control signal to the VVVF inverter device to control the discharge flow rate of each of the sludge pump and the flocculant pump a pump control unit; Receiving the sludge image from the image acquisition unit, acquiring a sludge shape image from the received sludge image, inputting the sludge shape image to a machine learning learning model to derive a probability for each type of sludge agglomeration state, and It is characterized in that it comprises a; AI control unit for calculating the coagulant correction input amount according to the probability, and transmitting the coagulant correction input amount to the pump control unit.

Description

Automatic operation system and method of belt press type dehydrator using AI based on image analysis and machine learning}

The present invention relates to an automatic operation system and method for a belt press type dehydrator using machine learning-based AI, and more particularly, to derive a probability for each type of sludge agglutination state through machine learning of a shape image of a mixed sludge mixed with a coagulant, and , It relates to an automatic operation system and method for a belt press dehydrator using machine learning-based AI that automatically controls the amount of coagulant to be optimal according to the probability of each type of sludge flocculation state derived.

In general, wastewater and water treatment processes inevitably generate by-products called sludge, and these sludges are configured to be dehydrated to facilitate collection.

Accordingly, various dewatering devices for dewatering waste sludge have been developed and used, and among them, a belt press type dewatering device with excellent maintenance and dehydration performance is most often used.

The belt press type dehydrator feeds and transfers the sludge between the lower dewatering belt and the upper dehydrating belt composed of dehydration filter cloth, and after the lower dehydration belt and the upper dehydration belt are combined, the pressure roller pressurizes the sludge to form a cake. It is configured to dehydrate. .

The belt press type dewatering machine mixes the sludge with a polymer (polymer) coagulant at the front to minimize the volume of the caked dewatering sludge, and this mixing process is an important factor that determines the dewatering efficiency.

If the amount of coagulant input is large, sparsely and irregularly lumped dewatered sludge is generated, and if the amount of coagulant input is small, the sludge becomes thin and contains more water, resulting in a problem that the overall volume and moisture content increase.

In particular, since the properties of the initial sludge supplied change frequently, the amount of coagulant input must be appropriately adjusted each time.

However, in the past, workers visually observed the sludge whenever the mixing concentration changed and manually adjusted the coagulant pump based on experience to adjust the input amount of the polymer coagulant, which lowered the treatment efficiency and reduced the cost due to the use of unnecessary coagulant. A lot is happening.

The present applicant (Yoocheon Engineering) connects a sludge pump and a coagulant pump to an agitator in Publication Patent Publication No. 10-1997-0064684 and Registered Utility Model No. 20-0219763 to supply and mix the sludge and coagulant, but a water level sensor in the sludge supply container By installing, the detection of this water level sensor is sent to the control unit, and the control unit controls the sludge and coagulant pump so that the sludge height becomes the set value and the coagulant input amount is optimal. did

However, controlling the sludge and coagulant pump so that the input amount of the coagulant is optimal through the detection of the water level in the sludge supply trough does not achieve a satisfactory level of control of the input amount even though the accuracy is low, so the automatic operation of the dehydrator is not achieved in practice. there was a problem i couldn't

Patent Publication No. 10-1997-0064684 Registered Utility Model No. 20-0219763 Publication No. 10-2012-0137796

The present invention was devised to solve the conventional problems as described above, and the shape image of the mixed sludge mixed with the flocculant was derived through machine learning to derive the probability for each type of sludge agglomeration state, and the flocculant The purpose is to provide an automatic operation system and method for a belt press type dehydrator using machine learning-based AI that automatically controls the amount of input to be optimal.

However, the object of the present invention is not limited thereto, and even if not explicitly mentioned, the purpose or effect that can be grasped from the solution or embodiment of the problem is also included therein, of course.

In order to achieve the above object, the automatic operation system of a belt press type dewatering machine using machine learning-based AI of the present invention is a belt dewatering unit in which dehydration is performed while the mixed sludge mixed with the coagulant passes between the lower dewatering belt and the upper dewatering belt. and a belt press type dehydrator comprising a gravity dewatering unit that primarily dehydrates and supplies the mixed sludge to the lower dewatering belt; an image acquisition unit that acquires a sludge image by photographing the mixed sludge being dewatered in the belt press type dehydrator and transmits the obtained image to an AI control unit in real time; An agitation tank for receiving and mixing the sludge and the coagulant by the sludge pump and the coagulant pump, and supplying the mixed sludge to the gravity dewatering unit; a sludge flowmeter installed on a first connection passage connecting the sludge pump and the stirring tank to detect a current input amount of sludge supplied; a coagulant flowmeter installed on a second connection passage connecting the coagulant pump and the stirring tank to detect a current input amount of the coagulant supplied; It consists of a VVVF (Variable Voltage Variable Frequency) inverter device for driving and controlling the sludge pump and the flocculant pump, and a PID controller for transmitting a pump control signal to the VVVF inverter device to control the discharge flow rate of each of the sludge pump and the flocculant pump a pump control unit; Receiving the sludge image from the image acquisition unit, obtaining a sludge shape image from the received sludge image, inputting the sludge shape image to a machine learning learning model to derive a probability for each type of sludge agglomeration state, and It is characterized in that it comprises a; AI control unit for calculating the coagulant correction input amount according to the probability, and transmitting the calculated coagulant correction input amount to the pump control unit.

The gravity dewatering part is provided with a supply tank for receiving mixed sludge from the stirring tank at one side, and while rotating on the side of the supply tank, the mixed sludge is rotated and moved through the outer circumferential surface, and the moisture in the mixed sludge is passed through a dehydration hole formed on the outer circumferential surface. It is composed of a drum thickener that removes and supplies the mixed sludge by dropping it on the lower dewatering belt, and the image acquisition unit obtains an image of the mixed sludge rotatingly moved through the outer circumferential surface of the drum thickener. It may be configured to transmit to the AI control unit.

The AI control unit captures the real-time sludge image received from the image acquisition unit at regular intervals to obtain a sludge shape image, and inputs the obtained sludge shape image to a machine learning model to determine the agglomeration state of deriving a probability for each type of sludge aggregation state wealth; A supply amount calculation unit that receives the current input amount of the coagulant from the flocculant flowmeter and calculates the corrected amount of the coagulant input by increasing or decreasing the current input amount of the coagulant at a predetermined rate according to the probability of each type of the sludge flocculation state.

The flocculation state determination unit derives a probability for each type of sludge aggregation state consisting of coagulant normal A type, coagulant excessive coagulant type B, coagulant insufficient coagulant C type, and sludge supply poor D type with respect to the sludge shape image, and the supply amount calculation unit calculates the sludge It may be configured to calculate the correction input amount of the coagulant according to the type having the highest probability in the probability of each type of aggregation state and transmit it to the pump control unit.

The supply amount calculation unit calculates the coagulant correction input amount identical to the coagulant current input amount without increase or decrease compared to the coagulant current input amount, when the type with the highest probability in the probability of each type of sludge coagulation state is the coagulant phase type A, and the sludge coagulation If the type with the highest probability in the probability of each state type is the excess flocculant type B, the corrected amount of coagulant input is calculated so as to be lowered at a preset input reduction ratio with respect to the current amount of coagulant input, and the highest probability in the probability of each type of sludge flocculation state When the type is the coagulant shortage type C, it may be configured to calculate the corrected coagulant input amount so as to increase at a preset input increase rate with respect to the current coagulant input amount.

The supply amount calculation unit, when the type with the highest probability in the probability of each type of sludge coagulation state is the sludge supply failure type D, sets both the corrected input amount of the sludge as 0 along with the corrected amount of the coagulant and sends it to the pump control unit, so that the sludge pump And the coagulant pump may be configured so that both drive is stopped.

The flocculation state determination unit derives a probability for each type of sludge aggregation state consisting of coagulant normal A type, coagulant excessive coagulant type B, coagulant insufficient coagulant C type, and sludge supply poor D type with respect to the sludge shape image, and the supply amount calculation unit calculates the sludge It is configured to calculate the coagulant correction input amount according to the calculation formula (1) based on the probability of each type of aggregation state, and the calculation formula (1) is,

(Where A is the current input amount of coagulant, Q1 is the input increase rate, P1 is the probability of type C with insufficient coagulant, Q2 is the rate of decrease in input, P2 is the probability of type B with excess coagulant).

According to the above, the probabilities for each type of sludge flocculation state are derived by machine learning using machine learning through the image of the shape of the dewatered sludge, and the amount of coagulant is automatically adjusted based on the probability for each type of agglutination state. The dehydrator can reduce the moisture content of the sludge to a satisfactory level and efficiently maintain the automatic operation state of the dehydrator so that the sludge can be dehydrated, thereby significantly reducing the treatment cost of the dehydrated sludge.

In addition, the various beneficial advantages and effects of the present invention are not limited to the above description, and will be more easily understood in the process of describing specific embodiments of the present invention.

1 is a conceptual diagram showing an automatic operation system for a belt press type dehydrator using AI based on machine learning according to a first embodiment of the present invention;
2 is a control block diagram of an automatic operation system of a belt press type dehydrator using AI based on machine learning according to a first embodiment of the present invention;
Figure 3 is an image photograph showing the sludge shape image of the present invention classified into sludge agglomeration state types,
4 is a diagram showing a screen of a probability display monitor configured in a control panel of the present invention;
5 is a conceptual diagram showing an automatic operation system for a belt press type dehydrator using AI based on machine learning according to a second embodiment of the present invention;
6 is a control block diagram of an automatic operation system for a belt press type dehydrator using AI based on machine learning according to a second embodiment of the present invention.

The above objects, other objects, features and advantages of the present invention will be easily understood through the following preferred embodiments in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments described herein and may be embodied in other forms. Rather, the embodiments introduced herein are provided so that the disclosed content will be thorough and complete and the spirit of the present invention will be sufficiently conveyed to those skilled in the art.

In this specification, when an element is referred to as being on another element, it means that it may be directly formed on the other element or a third element may be interposed therebetween. Also, in the drawings, the thickness of components is exaggerated for effective description of technical content.

Embodiments described in this specification will be described with reference to cross-sectional views and/or plan views, which are ideal exemplary views of the present invention. In the drawings, the thicknesses of films and regions are exaggerated for effective explanation of technical content. Accordingly, the shape of the illustrated drawings may be modified due to manufacturing techniques and/or tolerances. Therefore, embodiments of the present invention are not limited to the specific shapes shown, but also include changes in shapes generated according to manufacturing processes. For example, an etched region shown at right angles may be round or have a predetermined curvature. Accordingly, the regions illustrated in the drawings have attributes, and the shapes of the regions illustrated in the drawings are for exemplifying a specific form of a region of an element and are not intended to limit the scope of the invention. Although terms such as first and second are used to describe various components in various embodiments of the present specification, these components should not be limited by these terms. These terms are only used to distinguish one component from another. Embodiments described and illustrated herein also include complementary embodiments thereof.

Terms used in this specification are for describing embodiments, and are not intended to limit the present invention. In this specification, singular forms also include plural forms unless specifically stated otherwise in a phrase. The terms 'comprises' and/or 'comprising' used in the specification do not exclude the presence or addition of one or more other elements.

In describing the specific embodiments below, various specific contents are prepared to more specifically describe the invention and aid understanding. However, a reader having knowledge in this field to the extent of being able to understand the present invention can recognize that it can be used without these various specific details. In some cases, it is mentioned in advance that parts that are commonly known in describing the invention and are not greatly related to the invention are not described in order to prevent confusion for no particular reason in explaining the present invention.

Hereinafter, with reference to FIGS. 1 and 2, an automatic driving system and method for a belt press dehydrator using AI based on machine learning according to a first embodiment of the present invention will be described.

The automatic operation system of a belt press type dehydrator using AI based on machine learning is a belt press type dehydrator 10, an image acquisition unit 20, an agitation tank 30, a pump control unit 40, and an AI control unit 50. It is configured to include, and the pump control unit 40 and the AI control unit 50 may be configured in the control panel (A).

The belt press type dehydrator 10 is configured to include a belt dewatering unit 11 and a gravity dehydration unit 15.

In the belt dewatering unit 11, when mixed sludge mixed with sludge and coagulant is lifted from the gravity dehydration unit 15 to the lower dewatering belt 12 and supplied, it is transported on the lower dehydration belt 12 made of dehydration filter cloth. A belt pressure type dewatering unit that enters between the lower dewatering belt 12 and the upper dewatering belt 13, and dewatering is performed by the pressing force of the pressure roller 14 when the lower dewatering belt 12 and the upper dewatering belt 13 are combined. , This belt dewatering unit 11 is a well-known technology, and a detailed description thereof will be omitted.

The gravity dewatering unit 15 supplies the mixed sludge from the agitating tank 30, removes moisture from the mixed sludge to perform a primary dehydration process, and removes the mixed sludge that has undergone primary dehydration in the belt dewatering unit 11. configured to supply

The gravity dehydration unit 15 has a supply tank 17 receiving mixed sludge from the agitation tank 30 on one side, and rotates on the side of the supply tank 17 while rotating and moving the mixed sludge through the outer circumferential surface. It is composed of a drum thickener (16) for supplying the mixed sludge by dropping it onto the lower dewatering belt so that the moisture in the mixed sludge is dehydrated by gravity through the dewatering hole formed in the dewatering hole to remove moisture.

The drum thickener 16 is made in the form of a circular drum, and has a plurality of dewatering holes formed on the outer circumferential surface. As the drum thickener 16 rotates, the mixed sludge accommodated in the supply tank 17 is rotated by the outer circumferential surface of the drum thickener 16. Gravity dehydration is performed while being transferred, and the mixed sludge subjected to gravity dehydration is discharged to the bottom and supplied onto the lower dewatering belt 12.

The image acquisition unit 20 is composed of a camera means such as a webcam, acquires a shape image image of the mixed sludge rotated along the outer circumferential surface of the drum seek 16, and transmits the obtained mixed sludge image to the AI control unit 50. configured to transmit.

The stirring tank 30 is configured to receive the sludge and the flocculant respectively by the sludge pump 32 and the flocculant pump 34, mix the supplied sludge and flocculant, and supply the mixed mixed sludge (sludge + flocculant) to the supply pipe It is configured to be supplied to the supply tank 17 of the gravity dehydration unit 15 through.

In addition, a sludge flowmeter 36 for detecting the current injection amount of sludge is configured on the first connection passage connecting the stirring tank 30 and the sludge pump 32, and connecting the stirring tank 30 and the coagulant pump 34 A coagulant flowmeter 38 for detecting the current injection amount of the coagulant is configured on the second connection passage. The sludge flowmeter 36 is configured to transmit the detected current injection amount of sludge to the AI control unit 50, and the coagulant flowmeter 38 is configured to transmit the detected current injection amount of the flocculant to the AI control unit 50.

The pump control unit 40 includes a variable voltage variable frequency (VVVF) inverter device 42 that drives and controls the sludge pump and the coagulant pump, and a PID controller 44 that transmits a control signal to the VVVF inverter device 42 to control the pump. ), and receives the corrected input amount of coagulant and the corrected input amount of sludge from the I control unit 50, and the sludge pump 32 and the coagulant pump 34 so that the corrected input amount of the coagulant and the corrected input amount of the sludge are received. ) is configured to control the discharge of

The AI control unit 50 receives the sludge shape image from the image acquisition unit 20, determines the sludge shape image through a machine learning learning model, determines the correction amount of the coagulant input that can achieve the optimal dewatering of the mixed sludge, and determines the coagulant By transmitting the input correction amount to the PID controller 44, the coagulant pump 34 is controlled to supply the sludge and coagulant to the agitator 30 at an optimal mixing ratio, so that the mixed sludge having the optimal mixing ratio is produced in the belt press type dehydrator ( 10) is supplied to the automatic operation control.

Specifically, the AI control unit 50 is configured to include an aggregation state determining unit 52 and a supply amount calculating unit 54.

The agglomeration state determination unit 52 captures the sludge image received in real time from the image acquisition unit 20 at regular intervals to obtain a sludge shape image, and inputs the obtained sludge shape image to a machine learning learning model to determine the sludge aggregation state by type It is configured to derive probabilities.

Specifically, the machine learning learning model built in the flocculation state determination unit 52 adjusts the sludge shape image to a 400X300 image size, so that flocculant normal A type, flocculant excessive flocculant type B, flocculant insufficient flocculant type C, It is composed of a data set derived from a learning model in the form of a ckpt file by constructing a data set with four types of sludge aggregation state consisting of type D of sludge supply failure and learning with a standard convolutional neural network (CNN) structure.

Here, the data set for each type of sludge flocculation state is shown in Table 1 below.

classification TypeLabel image features number of images coagulant top A mixed sludge good 1,830 too much coagulant B Mixed sludge solidifies unevenly 1,774 lack of coagulant C Mixed sludge thinning 1,790 Coagulant Supply Amount D Poor sludge supply and cleaning 1,828

For example, referring to FIG. 3, among the images for each type of agglomeration state, type A is a form in which the mixed sludge transported by the drum thickener 16 is evenly spread and moved with a normal (optimal) amount of coagulant input, and type B is The mixed sludge transported by the drum thickener 16 is irregularly and sparsely clumped due to an excessive amount of coagulant input, and the mixed sludge transported by the drum thickener 16 is too thin due to an insufficient amount of coagulant input. It is a form.

In addition, although not attached to FIG. 3, in D type, the amount of mixed sludge supplied to the supply tank 15 is very excessive, so the mixed sludge cannot be transported by the drum thickener 16, and the drum thickener 16 The mixed sludge is not moved by the drum thickener 16 because it is overflowing or the amount of mixed sludge supplied to the supply tank 15 is very small, so there is little or no mixed sludge on the drum thickener 16. .

The flocculation state determination unit 52 captures the sludge image received in real time from the image acquisition unit 20 at regular intervals to obtain a sludge shape image, and inputs the obtained sludge shape image into a machine learning learning model to obtain flocculant normal type A , Deriving the probability for each type of sludge flocculation state consisting of type B with excessive coagulant, type C with insufficient flocculant, and type D with poor sludge supply.

In the present invention, a probability display monitor may be configured as a display module on a control panel (A) installed at a work site accessible by a manager and displaying the probability for each type of sludge agglomeration state. Accordingly, the flocculation state determination unit 52 may be configured to output the derived probability for each type of sludge flocculation state to the probability display monitor. At this time, the probability display monitor may be configured so that the sludge image obtained in real time from the image acquisition unit 20 is output together, and the probability for each type of sludge aggregation state is overlapped and displayed on the sludge image (see FIG. 4).

The supply amount calculation unit 54 calculates the corrected coagulant input amount according to the probability for each type of sludge aggregation state derived from the flocculation state determination unit 52, and transmits the calculated coagulant corrected input amount to the pump control unit 40, so that the pump control unit 40 ), the coagulant pump 34 injects the coagulant into the agitation tank 30 with the corrected coagulant input amount, so that the mixing ratio of the sludge and the coagulant in the agitation tank 30 can be maintained in an appropriate state.

The supply amount calculation unit 54 may be configured to calculate the corrected amount of coagulant input according to the type with the highest probability in the probability of each type of sludge flocculation state and transmit it to the pump control unit.

For example, the supply amount calculation unit 54 calculates that when the probability for each type of sludge coagulation state is 10% of flocculant normal type A, 80% of flocculant excess type B, 5% of flocculant insufficient type C, and 5% of sludge supply poor type D, the probability is The corrected coagulant input amount is calculated according to the highest type B, and the corrected coagulant input amount is calculated so that the corrected coagulant input amount is lowered by a predetermined input reduction ratio compared to the current coagulant input amount received from the coagulant flowmeter 38.

In addition, the supply amount calculation unit 54 has the highest probability of sludge coagulation condition type A type 5%, flocculant excess B type 5%, flocculant insufficient C type 85%, sludge supply poor D type 10%, The corrected coagulant input amount is calculated according to the type C. The corrected coagulant input amount can be calculated so that the coagulant corrected input amount is increased by a preset input increase rate compared to the current coagulant input amount received from the coagulant flowmeter 38.

In addition, when type A is the highest in the probability of each type of sludge coagulation state, the current coagulant input amount is calculated without adding or subtracting the current coagulant input amount, and the corrected coagulant input amount is calculated and transmitted to the pump control unit 40, thereby changing the coagulant current input amount It is possible to maintain the current input of flocculant without and without.

On the other hand, the supply amount calculation unit 54 determines that the device is broken when type D is the highest in the probability of each type of sludge coagulation state, sets both the coagulant correction input amount and the sludge correction input amount to 0, and transmits it to the pump control unit 50, By using the pump control unit 50 to stop both the sludge pump 32 and the coagulant pump 34, equipment inspection by a manager can be performed.

As described above, the supply amount calculation unit 54 of the present invention calculates the corrected input amount of the coagulant according to the type with the highest probability in the probability of each type of sludge coagulation state and transmits it to the pump control unit 40, so that the optimal gelatinization ratio for sludge and coagulant It was configured to control the discharge amount of the coagulant pump 34 to have, but is not limited thereto, and as another example, the supply amount calculation unit 54 of the present invention calculates the following formula (1) based on the probability of each type of sludge aggregation state It can be configured to calculate the coagulant correction dosage by.

... 산출식(1)

... formula (1)

Here, A is the current input amount of coagulant (flow unit), Q1 is the input increase rate, P1 is the probability of type C with insufficient coagulant, Q2 is the rate of decrease in input, and P2 is the probability of type B with excess coagulant.

In this case, even if the type with the highest probability in the probability for each type of sludge aggregation state derived by the machine learning learning model in the flocculation state determination unit 52 is flocculant normal type A, the supply amount calculation unit 54 is the same as the current input amount of flocculant Instead of calculating the corrected input amount, it can be configured to calculate the corrected input amount (K) to the flocculation by the calculation formula (1) according to the probability of excess coagulant type B and the probability of insufficient coagulant type C.

For example, the probability for each type of sludge flocculation state derived by the machine learning learning model in the flocculation state determination unit 52 is 70% of flocculant normal type A, 15% of flocculant excessive type B, 10% of flocculant insufficient type C, and poor sludge supply When D type is derived as 5%, the current coagulant input amount is 100, the preset input reduction ratio is 0.2, and the preset input increase ratio is 0.2, the corrected coagulant input amount (K) calculated by the formula (1) is 100 + (( 100 x 0.2 x 0.1) - (100 x 0.2 x 0.15) gives 99.

Hereinafter, with reference to FIGS. 5 and 6, an automatic driving system and method for a belt press type dehydrator using AI based on machine learning according to a second embodiment of the present invention will be described.

The second embodiment is different from the first embodiment only in that the thickness sensing unit 60 is added, but the rest of the configuration is the same, so the same configuration is given the same reference numerals on the drawings, and the description is omitted. Only configurations with .

The thickness sensor 60 is installed on one side of the lower portion of the gravity dehydration unit 15 and measures the thickness of the mixed sludge on the lower dewatering belt 12 supplied from the gravity dehydration unit 15 to the lower dewatering belt 12 and transported. It is configured to transmit the measured thickness to the AI control unit 60.

In the present invention, the thickness sensor 60 is composed of a laser irradiation type thickness sensor, irradiates a laser toward the lower desorption valve 12, receives the reflected laser, and measures the thickness of the mixed sludge transported on the lower dewatering belt 12. It is possible to measure and transmit the measured thickness to the AI control unit 60.

At this time, the thickness sensor 60 measures the thickness of the mixed sludge transferred on the lower dewatering belt 12 multiple times, calculates the average measured thickness for the plurality of measured thicknesses, and transmits the result to the AI control unit 60. can be configured to

The AI control unit 60 compares the average measured thickness of the mixed sludge received from the thickness sensor 60 with a preset normal thickness range of the sludge, and when the average measured thickness of the mixed sludge is out of the normal sludge thickness range, at the same time, the aggregated state If the probability of flocculant normal type A is 70% or more in the probability of each type of sludge aggregation state derived by the determination unit 52 by the machine learning learning model, the sludge pump 32 and the flocculant pump 34 are both stopped. Pump control unit 40 It is configured to transmit a pump stop signal to.

In the above, the present invention has been shown and described in relation to preferred embodiments for illustrating the principles of the present invention, but the present invention is not limited to the configuration and operation as shown and described. Rather, those skilled in the art will appreciate that many changes and modifications to the present invention can be made without departing from the spirit and scope of the appended claims. Accordingly, all such appropriate changes and modifications and equivalents should be regarded as belonging to the scope of the present invention.

10...Belt press dehydrator
11...belt dewatering part
12... Lower dewatering belt
13... upper dewatering belt
14... pressure roller
15... Gravity dewatering unit
16... Drum Thickener
17...supply tank
20 ... image acquisition department
30 ... stirring tank
32 ... sludge pump
34 ... coagulant pump
36 ... sludge flow meter
38 ... coagulant flow meter
40 ... pump control unit
42...VVVF inverter device
44...PID controller
50 ... AI control unit
52 ... collection state determination unit
54 ... Supply amount calculation unit
60 ... thickness sensing unit

Claims (7)

A belt press type dehydrator consisting of a belt dewatering unit in which the mixed sludge mixed with the coagulant is dewatered while passing between the lower dewatering belt and the upper dewatering belt, and a gravity dewatering unit that primarily dehydrates and supplies the mixed sludge to the lower dewatering belt. ;
an image acquisition unit that acquires a sludge image by photographing the mixed sludge being dewatered in the belt press type dehydrator and transmits the obtained image to an AI control unit in real time;
An agitation tank for receiving and mixing the sludge and the coagulant by the sludge pump and the coagulant pump, and supplying the mixed sludge to the gravity dewatering unit;
a sludge flowmeter installed on a first connection passage connecting the sludge pump and the stirring tank to detect a current input amount of sludge supplied;
a coagulant flowmeter installed on a second connection passage connecting the coagulant pump and the stirring tank to detect a current input amount of the coagulant supplied;
It consists of a VVVF (Variable Voltage Variable Frequency) inverter device for driving and controlling the sludge pump and the flocculant pump, and a PID controller for transmitting a pump control signal to the VVVF inverter device to control the discharge flow rate of each of the sludge pump and the flocculant pump a pump control unit;
Receiving the sludge image from the image acquisition unit, obtaining a sludge shape image from the received sludge image, inputting the sludge shape image to a machine learning learning model to derive a probability for each type of sludge agglomeration state, and An automatic operation system of a belt press dehydrator using machine learning-based AI, characterized in that it comprises: an AI control unit that calculates the corrected coagulant input amount according to probability and transmits the calculated coagulant corrected input amount to the pump control unit.
According to claim 1,
The gravity dewatering part is provided with a supply tank for receiving mixed sludge from the stirring tank at one side, and while rotating on the side of the supply tank, the mixed sludge is rotated and moved through the outer circumferential surface, and the moisture in the mixed sludge is passed through a dehydration hole formed on the outer circumferential surface. It is composed of a drum thickener that removes and supplies the mixed sludge by dropping it on the lower dewatering belt,
The image acquisition unit acquires an image of the mixed sludge that is rotated through the outer circumferential surface of the drum thickener and transmits the image to the AI control unit. system.
According to claim 1,
The AI control unit,
A flocculation state determination unit that captures the real-time sludge image received from the image acquisition unit at regular intervals to obtain a sludge shape image, and inputs the obtained sludge shape image to a machine learning model to derive a probability for each type of sludge agglomeration state;
A supply amount calculation unit that receives the current input amount of the coagulant from the coagulant flowmeter and calculates the corrected amount of the coagulant input by increasing or decreasing the current input amount of the coagulant at a predetermined rate according to the probability of each type of the sludge flocculation state; machine learning characterized in that it is configured to include Automatic operation system of belt press type dehydrator using AI based system.
According to claim 3,
The flocculation state determination unit derives a probability for each type of sludge flocculation state consisting of normal coagulant type A, excessive flocculant type B, insufficient coagulant type C, and poor sludge supply type D for the sludge shape image,
The supply amount calculation unit is configured to calculate the coagulant correction input amount according to the type with the highest probability in the probability of each type of sludge coagulation state and transmit it to the pump control unit Belt press type dehydrator using machine learning-based AI, characterized in that automatic driving system.
According to claim 4,
The supply amount calculation unit,
When the type with the highest probability in the probability of each type of sludge flocculation state is the coagulant normal type A, the coagulant correction input amount is calculated as the same as the coagulant current input amount without an increase or decrease compared to the coagulant current input amount,
When the type with the highest probability in the probability of each type of sludge flocculation state is the excess coagulant B type, the corrected coagulant input amount is calculated so that the coagulant input amount is lowered by a preset input reduction ratio with respect to the coagulant current input amount,
When the type with the highest probability in the probability of each type of sludge flocculation state is the coagulant deficiency type C, the corrected coagulant input amount is calculated so as to increase at a preset input increase rate with respect to the coagulant current input amount Machine learning based, characterized in that configured to calculate Automatic operation system of belt press type dehydrator using AI.
According to claim 4,
The supply amount calculation unit, when the type with the highest probability in the probability of each type of sludge coagulation state is the sludge supply failure type D, sets both the corrected input amount of the sludge as 0 along with the corrected amount of the coagulant and sends it to the pump control unit, so that the sludge pump And the automatic operation system of the belt press type dehydrator using machine learning-based AI, characterized in that the coagulant pump is configured to stop driving.
According to claim 3,
The flocculation state determination unit derives a probability for each type of sludge flocculation state consisting of normal coagulant type A, excessive flocculant type B, insufficient coagulant type C, and poor sludge supply type D for the sludge shape image,
The supply amount calculation unit is configured to calculate the coagulant correction input amount according to the calculation formula (1) based on the probability of each type of the sludge aggregation state,
The calculation formula (1) is,

(Where A is the current input amount of coagulant, Q1 is the input increase rate, P1 is the probability of type C with insufficient coagulant, Q2 is the rate of decrease in input, P2 is the probability of type B with excess coagulant)
An automatic operation system of a belt press type dehydrator using machine learning-based AI, characterized in that it consists of.

Images