JP7436305B2 - drying system - Google Patents

Description

The present invention relates to a drying system that dries materials to be dried, such as sludge and slurry, while stirring, and particularly relates to a drying system that can perform optimal automatic operation.

DESCRIPTION OF RELATED ART Conventionally, the drying apparatus which conveys and dries materials to be dried, such as sludge, while stirring is known (for example, refer patent document 1). This type of drying device uses a shaft provided with multiple rows of blades (paddles) for stirring and transporting the material to be dried. As the shaft rotates, the blades provided on the shaft also rotate around the shaft, whereby the material to be dried supplied to one end of the shaft is conveyed to the other end of the shaft while being stirred. During this stirring and transportation, the material to be dried is heated and dried.

In this type of drying device, the blades fixed to two shafts overlap each other to enhance the stirring effect. Drying equipment is required to be easy to operate, such as stabilizing the moisture content of the dried material after drying, and reliably preventing poor dispersion and overdrying of the dried material. There is a need for a new drying device that can reduce initial costs.

Japanese Patent Application Publication No. 2014-131784 Japanese Patent Application Publication No. 4-157000 Japanese Patent Application Publication No. 4-338300 JP2011-67787A Japanese Patent Application Publication No. 2016-70545

However, the moisture content of the material to be dried discharged from the drying device may vary depending on various factors such as the properties of the material to be dried fed into the drying device and the operating state of the drying device. Therefore, in order to optimize the operation of the drying equipment, the operator can check the condition of the material to be dried such as sludge that is fed into the drying equipment, and the current state of the drying material through the inspection port installed in the drying equipment or the discharge port of the drying equipment. The dried material is visually checked and the operation of the drying equipment is adjusted based on the condition of the material to be dried. However, operation adjustment based on visual observation of the material to be dried requires extensive experience and know-how of the operator. Furthermore, the operating conditions may vary depending on the operator, and as a result, the moisture content of the material to be dried discharged from the drying apparatus may not be stable.

The present invention acquires the condition of the dried material as image data, sound data, or vibration data, determines whether the operating condition is good or bad based on this data, and keeps the moisture content of the obtained dried material within the target range. The purpose of the present invention is to provide a drying system that can be stored and stably maintained.

In one aspect, there is provided a drying system for drying an object to be dried, which is constructed using a drying device that dries the object to be dried, an imaging device that generates image data of the object to be dried, and a machine learning algorithm. A control system having a learned model is provided, and the control system inputs image data of a storage device storing the learned model and the image data of the object to be dried to the learned model, and the drying object is discharged from the drying device. A drying system is provided, comprising a processing device that executes calculations for outputting, from the learned model, optimal operating parameters for the drying system that can keep the moisture content of the material to be dried within a target range. Ru.

In one aspect, the control system inputs image data of the object to be dried and state data of the drying system to the learned model, and adjusts the moisture content of the object to be dried discharged from the drying device to a target range. The processing device is configured to cause the processing device to perform an operation for outputting, from the learned model, optimal operating parameters of the drying system that can be accommodated within the range.
In one aspect, the control system inputs image data of the object to be dried and state data of a dewatering system disposed upstream of the drying system into the learned model, and inputs image data of the object to be dried and state data of a dewatering system arranged upstream of the drying system, The processing device is configured to cause the processing device to perform calculations for outputting, from the learned model, optimal operating parameters for the drying system that can keep the moisture content of the material to be dried within a target range.

In one embodiment, a drying system for drying an object to be dried includes a drying device for drying the object to be dried, a sound detector that detects sound generated from the object to be dried and generates sound data. , a control system having a learned model constructed by a machine learning algorithm, the control system including a storage device storing the learned model, inputting the sound data to the learned model, and controlling the drying device. A dryer comprising: a processing device that executes calculations for outputting, from the learned model, optimal operating parameters for the drying system that can keep the moisture content of the dried material discharged from the drying system within a target range; system is provided.

In one aspect, the control system inputs the sound data and the state data of the drying system into the learned model, and keeps the moisture content of the material to be dried discharged from the drying device within a target range. The processing device is configured to cause the processing device to perform calculations for outputting optimal operating parameters for the drying system from the learned model.
In one aspect, the control system inputs sound data of the object to be dried and state data of a dewatering system disposed upstream of the drying system into the learned model, and The processing device is configured to cause the processing device to perform calculations for outputting, from the learned model, optimal operating parameters for the drying system that can keep the moisture content of the material to be dried within a target range.

In one aspect, there is provided a drying system for drying an object to be dried, comprising: a drying device for drying the object to be dried; an elastic member disposed at a position where the object to be dried contacts; and a drying system fixed to the elastic member. and a control system having a vibration detector that generates vibration data of the elastic member and a learned model constructed by a machine learning algorithm, the control system including a storage device storing the learned model; The vibration data is input to the trained model, and the trained model outputs optimal operating parameters for the drying system that can keep the moisture content of the material to be dried discharged from the drying device within a target range. A drying system is provided that includes a processing device that performs operations for.

In one aspect, the control system inputs the vibration data and the state data of the drying system into the learned model, and keeps the moisture content of the material to be dried discharged from the drying device within a target range. The processing device is configured to cause the processing device to perform calculations for outputting optimal operating parameters for the drying system from the learned model.
In one aspect, the control system inputs vibration data of the object to be dried and state data of a dewatering system disposed upstream of the drying system into the learned model, and The processing device is configured to cause the processing device to perform calculations for outputting, from the learned model, optimal operating parameters for the drying system that can keep the moisture content of the material to be dried within a target range.

In one aspect, a drying system for drying an object to be dried includes: a drying device for drying the object to be dried; an imaging device for generating image data of the object to be dried dried by the drying device; A control system includes a trained model constructed by a machine learning algorithm, and the control system includes a storage device storing the trained model, inputs the image data to the trained model, and controls the drying system. A drying system is provided that includes a processing device that executes a calculation for outputting an operation index value indicating whether the operation state is normal or abnormal from the learned model.

In one aspect, the control system is configured to calculate a moving average of the driving index values.
In one aspect, the control system determines that an abnormality is occurring in the operation of the drying system when the moving average of the operating index values exceeds a threshold.
In one aspect, the control system resets the operation abnormality determination operation of the drying system when the moving average of the operating index values decreases to a predetermined normal level.

According to the present invention, a trained model constructed by a machine learning algorithm generates optimal operating parameters that can keep the moisture content of the material to be dried discharged from the drying device within a target range, and By applying this to the drying equipment, the overall operation of the drying system can be optimized.

In particular, according to the present invention, instead of the operating parameters conventionally determined by experienced operators based on visual inspection of the dried material, the drying system uses a trained model to determine the moisture content of the obtained dried material. Optimal operating parameters can be generated that can keep the rate within the target range. Therefore, the drying system is capable of adjusting the operating parameters of the drying device in real time as accurately as, or more accurately than, an experienced operator.

1 is a diagram illustrating one embodiment of a drying system. FIG. FIG. 2 is a top view of the drying device shown in FIG. 1; FIG. 3 is a diagram of the paddle viewed from the axial direction of the shaft. FIG. 2 is a schematic diagram showing an example of a trained model. FIG. 7 illustrates another embodiment of a drying system. FIG. 7 is a diagram showing yet another embodiment of a drying system. FIG. 1 is a diagram illustrating one embodiment of a dehydration system. It is a figure showing other embodiments of a dehydration system. It is a figure which shows yet another embodiment of a dehydration system. It is a graph which shows the time change of the operation index value output from the 2nd learned model during operation of a drying system. 11 is a graph showing temporal changes in the moving average of the driving index values shown in FIG. 10;

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a diagram illustrating one embodiment of a drying system. This drying system is suitable for drying turbid residues obtained by dewatering suspensions such as sludge and slurry. In particular, the drying system is suitably used for drying a dehydrated cake discharged from a dehydrator such as a screw press.

The drying system includes an input conveyor 1 that conveys a material to be dried such as a dehydrated turbid residue, a drying device 5 that receives the material to be dried conveyed by the input conveyor 1 and dries the material to be dried, and a drying device. an imaging device 7 that generates image data of the object to be dried before being put into the drying device 5; an imaging device 8 that generates image data of the object to be dried in the drying device 5; The drying system includes an imaging device 9 that generates image data, and a control system 10 that controls the operation of the drying system based on the image data.

Suspended residue is a material with low water content that remains after a suspension, such as sludge, is dewatered by a dewatering device (eg, a screw press or belt press). The dehydrated turbid residue is commonly referred to as cake. Specific examples of sludge include sludge generated during the treatment of sewage, human waste, and industrial wastewater. Furthermore, examples of suspensions other than sludge include industrial waste or slurry generated during the manufacture of industrial products such as foodstuffs, cosmetics, and paper.

FIG. 2 is a top view of the drying device 5 shown in FIG. 1. As shown in FIG. 2, the drying device 5 includes a housing 15 that forms a drying chamber 14 therein, two stirring rotors 17, and an electric motor 18 that rotates these stirring rotors 17. Each stirring rotor 17 includes a shaft 17A and a plurality of paddles 17B fixed to the shaft 17A. Both ends of the shaft 17A extend through the housing 15 and are rotatably supported by bearings (not shown). Paddle 17B is arranged within drying chamber 14. Paddles 17B are arranged along the axial direction of shaft 17A.

Each paddle 17B is inclined with respect to the direction perpendicular to the axial direction of the shaft 17A. Furthermore, the inclination angle of half of all the paddles 17B is different from the inclination angle of the other half of the paddles 17B. These paddles 17B having different inclination angles are arranged alternately along the axial direction of the shaft 17A.

FIG. 3 is a diagram of the paddle 17B viewed from the axial direction of the shaft 17A. As shown in FIG. 3, the paddle 17B has a fan shape. Each paddle 17B is fixed to the outer peripheral surface of the shaft 17A, and a pair of two paddles 17B surround the outer peripheral surface of the shaft 17A. When viewed from the axial direction of the shafts 17A, the outermost edge of the paddle 17B fixed to one shaft 17A overlaps the outermost edge of the paddle 17B fixed to the other shaft 17A. Such an overlapping arrangement improves the efficiency of stirring the material to be dried. Actually, the paddle 17B fixed to one shaft 17A is not in contact with the paddle 17B fixed to the other shaft 17A.

As shown in FIG. 2, two gears 20 are respectively fixed to the two shafts 17A. These gears 20 mesh with each other. Therefore, the two shafts 17A, ie, the two stirring rotors 17, rotate in opposite directions. The electric motor 18 is connected to one of the two shafts 17A via a torque transmission mechanism 22. The torque transmission mechanism 22 is composed of a combination of a sprocket and a chain, a combination of a pulley and a belt, or the like. When the electric motor 18 is activated, the two stirring rotors 17 rotate in opposite directions. The electric motor 18 has a built-in inverter (not shown). The electric motor 18 is connected to a control system 10, and the operation of the electric motor 18, ie the rotational speed of the stirring rotor 17, is controlled by the control system 10.

The two shafts 17A are hollow shafts, and each paddle 17B is also a hollow paddle. That is, a space extending in the axial direction is formed inside each shaft 17A, and a space is also formed inside each paddle 17B. The space within the shaft 17A and the space within the paddle 17B communicate with each other. The drying system includes two steam supply lines 50 each connected to one end of two shafts 17A. Steam is supplied into shaft 17A from steam supply line 50, flows through the space within shaft 17A and the space within paddle 17B, and is discharged from the other end of shaft 17A. The shaft 17A and paddle 17B are heated by this steam, and their surfaces become hot.

Two steam supply lines 50 are connected to a boiler 51. The pressure and flow rate of steam flowing through the steam supply line 50 are controlled by a steam pressure control valve 52 installed in the steam supply line 50. The steam pressure regulating valve 52 is connected to the control system 10 , and the steam pressure regulating valve 52 is controlled by the control system 10 . More specifically, the pressure of steam supplied into stirring rotor 17 is controlled by control system 10 via steam pressure regulating valve 52 .

Returning to FIG. 1, the housing 15 includes an input port 25 into which the material to be dried is input, and an outlet 26 from which the dried material is discharged. The drying system includes a drying gas supply line 30 connected to the housing 15 and a drying gas exhaust line 31. The drying gas is supplied into the housing 15 through the drying gas supply line 30 , contacts the object to be dried within the housing 15 , and is exhausted from the housing 15 through the drying gas exhaust line 31 . The dry gas supplied from the dry gas supply line 30 is a high-temperature gas that has been previously dehumidified and has been previously heated by a heat source (not shown) such as steam or a heater.

A temperature measuring device 32 and a flow rate measuring device 33 are attached to the dry gas supply line 30. The temperature measuring device 32 is configured to measure the temperature of the dry gas flowing through the dry gas supply line 30, and the flow rate measuring device 33 is configured to measure the flow rate of the dry gas flowing through the dry gas supply line 30. has been done. Furthermore, a temperature measuring device 35 is attached to the dry gas discharge line 31 as well to measure the temperature of the dry gas flowing through the dry gas discharge line 31. Temperature measuring device 32 , flow measuring device 33 , and temperature measuring device 35 are connected to control system 10 and measure the temperature and flow rate of dry gas supplied into housing 15 and discharged from housing 15 . Measurements of the temperature of the drying gas are sent to the control system 10.

The drying device 5 includes a temperature measuring device 40 that measures the surface temperature of the housing 15 and a pressure measuring device 41 that measures the pressure inside the housing 15. The temperature measuring device 40 and the pressure measuring device 41 are connected to the control system 10 such that the measured value of the surface temperature of the housing 15 and the measured value of the pressure inside the housing 15 are sent to the control system 10. .

The drying device 5 includes a weir 56 disposed within the housing 15 and a vertical movement device 58 that moves the weir 56 up and down. The weir 56 is disposed within the housing 15, and is located downstream of the paddle 17B in the flow direction of the material to be dried. The interior of the housing 15 is divided into a drying chamber 14 and a discharge chamber 60 by a weir 56. The discharge port 26 is located below the discharge chamber 60. The material to be dried in the drying chamber 14 is moved over the upper edge of the weir 56 to the discharge chamber 60. The weir 56 has a function of controlling the amount and residence time of the material to be dried in the drying chamber 14 . That is, when the weir 56 is moved upward (increasing the height of the weir 56), the amount and time of the material to be dried that stays in the drying chamber 14 increases, and as a result, the material to be dried is further dried. The vertical movement device 58 is connected to the control system 10, and the operation of the vertical movement device 58, that is, the height of the weir 56 (the vertical position of the weir 56) is controlled by the control system 10.

The operation of the drying device 5 is as follows. Steam is supplied into the stirring rotor 17 from the steam supply line 50 while the electric motor 18 rotates the two stirring rotors 17 in opposite directions. The stirring rotor 17 is heated by the steam, and the surface of the stirring rotor 17 becomes high temperature. Additionally, drying gas flows into the drying chamber 14 through the drying gas supply line 30, flows through the drying chamber 14, and is discharged through the drying gas exhaust line 31. The material to be dried is introduced into the drying chamber 14 through the input port 25. The material to be dried is moved toward the discharge chamber 60 little by little by the paddle 17B of the rotating stirring rotor 17. While being moved by the stirring rotor 17, the material to be dried is dried by contact with the high temperature stirring rotor 17 and contact with dry gas. The dried material moves over the weir 56 into the discharge chamber 60 and is discharged from the drying device 5 through the discharge port 26. A discharge conveyor 64 is disposed below the discharge port 26, and the dried material passes through the discharge port 26, falls onto the discharge conveyor 64, and is conveyed by the discharge conveyor 64.

The drying system includes an imaging device 7 that generates image data of the object to be dried before being input into the drying device 5. The imaging device 7 is arranged near the input port 25 of the drying device 5, and is arranged so as to generate image data of the material to be dried before being dried. In one embodiment, the imaging device 7 may be arranged to generate image data of the material to be dried on the input conveyor 1 or a dewatering device (not shown) located upstream of the input conveyor 1. (2) may be placed near the outlet. The imaging device 7 is connected to a control system 10, and image data of the object to be dried generated by the imaging device 7 is sent to the control system 10.

The drying system further includes an imaging device 8 that generates image data of the object to be dried in the drying device 5. The imaging device 8 is fixed to the upper wall of the housing 15 of the drying device 5 and is disposed facing the inside of the drying device 5, more specifically, the inside of the drying chamber 14. In one embodiment, a plurality of imaging devices 8 may be arranged along the axial direction of the stirring rotor 17. The imaging device 8 is arranged so as to generate image data of the dried material in the process of being dried, that is, the dried material being stirred by the stirring rotor 17 . The imaging device 8 is connected to a control system 10, and image data of the object to be dried generated by the imaging device 8 is sent to the control system 10.

The drying system further includes an imaging device 9 that generates image data of the object dried by the drying device 5. The imaging device 9 is arranged near the outlet 26 of the drying device 5. More specifically, the imaging device 9 is disposed facing the discharge port 26 and is disposed so as to generate image data of the object to be dried discharged from the drying chamber 14 of the drying device 5. In one embodiment, the imaging device 9 may be located within or outside the evacuation chamber 60. Furthermore, in one embodiment, the imaging device 9 may be arranged to generate image data of the dried material on the discharge conveyor 64. The imaging device 9 is connected to a control system 10 , and image data of the object to be dried generated by the imaging device 9 is sent to the control system 10 .

The imaging device 7, the imaging device 8, and the imaging device 9 are digital cameras equipped with an image sensor (eg, a CCD image sensor or a CMOS image sensor) that can generate a still image or a continuous image of an object. Alternatively, at least one of the imaging device 7, the imaging device 8, and the imaging device 9 may be an infrared camera or a hyperspectral camera that can image an object by separating light into wavelengths. .

The moisture content of the material to be dried dried by the drying device 5 changes depending on the state of the material to be dried introduced into the drying device 5 and the operating state of the drying device 5. It is preferable that the moisture content of the material to be dried obtained by the drying device 5 is within an appropriate range. If the moisture content of the material to be dried is too high, the weight of the material to be dried will be large and the processing cost will increase. On the other hand, if the moisture content of the material to be dried is too low, the material to be dried will easily scatter.

Therefore, the control system 10 is configured to acquire image data of the object to be dried from the imaging device 7, the imaging device 8, and the imaging device 9, and to optimize the operating parameters of the drying system based on these image data. ing. The control system 10 inputs a storage device 10a storing a learned model constructed by a machine learning algorithm and image data to the learned model, and adjusts the moisture content of the dried material discharged from the drying device 5 to a target range. The apparatus includes a processing device 10b that executes calculations for outputting optimal operating parameters of the drying system that can be accommodated within the learned model from the learned model.

The storage device 10a includes a main storage device that can be accessed by the processing device 10b, and an auxiliary storage device that stores programs, learned models, and data. The main storage device is, for example, a random access memory (RAM), and the auxiliary storage device is a storage device such as a hard disk drive (HDD) or a solid state drive (SSD). The processing device 10b includes a CPU (central processing unit), a GPU (graphic processing unit), or the like.

Control system 10 is comprised of at least one computer. The at least one computer may be one server or multiple servers. The control system 10 may be an edge server connected to the imaging device 7, the imaging device 8, and the imaging device 9 via a communication line, or may be an edge server connected to the imaging device 7, the imaging device 8, and the imaging device 9 via a network such as the Internet. or a fog computing device (gateway, fog server, router, etc.) installed in a network connected to imager 7, imager 8, and imager 9. There may be. The control system 10 may be a plurality of servers connected via a network such as the Internet. For example, the control system 10 may be a combination of an edge server and a cloud server.

The trained model is composed of a neural network. The storage device 10a stores a program for constructing a trained model according to a machine learning algorithm. The processing device 10b constructs a learned model by executing operations according to instructions included in the program. Building a trained model according to a machine learning algorithm involves optimizing parameters such as neural network weights.

Examples of machine learning algorithms include SVR method (support vector regression method), PLS method (Partial Least Squares), deep learning method, random forest method, or decision tree method. However, deep learning methods are particularly suitable. Deep learning is a learning method based on a neural network with multiple hidden layers. In this specification, machine learning using a neural network including an input layer, two or more hidden layers, and an output layer is referred to as deep learning. By using the deep learning method, it is now possible to use a computer to determine the condition of an object to be dried, which has previously been determined based on human eyes and experience, based on image data of the object to be dried.

The learned model keeps the past image data of the dried material generated by the imaging device 7, the imaging device 8, and the imaging device 9 and the moisture content of the dried material discharged from the drying device 5 within the target range. The model is built according to a machine learning algorithm using learning data that includes multiple combinations of optimal operating parameters for the drying system. The past image data of the dried material is an explanatory variable, and the optimal operating parameter of the drying system that can keep the moisture content of the dried material discharged from the drying device 5 within the target range is the objective variable. The optimal operating parameters constituting the learning data are correct data, and are operating parameters determined by a skilled operator based on the condition of the material to be dried and the moisture content of the material to be dried discharged from the drying device 5. . Learning data is also called training data or teacher data.

In order to optimize the parameters (weights, etc.) of the trained model, training data containing numerous combinations of image data of the dried object and optimal operating parameters that can keep the moisture content of the cake within the target range. will be prepared. The control system 10 uses this learning data to construct a learned model using a machine learning algorithm. In addition to weights, the parameters of a trained model may include bias. The trained model constructed in this way is stored in the storage device 10a.

The control system 10 acquires image data from the imaging device 7, the imaging device 8, and the imaging device 9 at a predetermined period, and stores it in the storage device 10a. The control system 10 inputs image data to the trained model, and outputs optimal operating parameters that can keep the moisture content of the material to be dried within the target range from the trained model.

FIG. 4 is a schematic diagram showing an example of a learned model. The learned model is a neural network having an input layer 201, a plurality of hidden layers (also referred to as intermediate layers) 202, and an output layer 203. Although the trained model shown in FIG. 4 has four hidden layers 202, the configuration of the trained model is not limited to the embodiment shown in FIG. 4. A trained model may have five or more hidden layers 202.

Image data is input to the input layer 201 of the learned model. More specifically, the numerical value of each pixel making up the image data is input to the input layer 201. If the image data is color image data, numerical values representing red, green, and blue of each pixel are input to corresponding nodes (neurons) of the input layer 201. A control system 10 composed of at least one computer executes calculations according to a multilayer perceptron algorithm constituting a neural network, and the output layer 203 of the learned model is a water-containing product of the dried material discharged from the drying device 5. Outputs a numerical value representing the operating parameter that allows the rate to fall within the target range. However, the configuration of the learned model shown in FIG. 4 is one example, and the present invention is not limited to the example shown in FIG. 4.

Specific examples of operating parameters output from the learned model are as follows.
・Excess or deficiency of the flow rate of the material to be dried to the drying device 5 within the appropriate range → Increase, maintain, or decrease the conveyance speed of the input conveyor 1 ・Excess or deficiency of the height of the weir 56 with respect to the appropriate range → By the vertical movement device 58 Raising or lowering the weir 56, or maintaining the current height of the weir 56 - Excess or deficiency of the pressure of steam supplied to the stirring rotor 17 within the appropriate range → Increase or maintain the opening degree of the steam pressure control valve 52, or decrease/Excess/deficiency of the rotational speed of the stirring rotor 17 to the appropriate range → Increase, maintain, or decrease the rotational speed of the electric motor 18

The operating parameters described above are examples, and the operating parameters output from the learned model may be only any one of the operating parameters described above.

Conventionally, a skilled operator visually judged the condition of the material to be dried and appropriately set the operating parameters based on experience, whereas in the present embodiment, the control system 10 Input image data of an object into a trained model, and have the trained model output optimal operating parameters. By applying these operating parameters to the drying system, the overall operation of the drying system can be optimized.

In this embodiment, all of the image data generated by the imaging device 7, the imaging device 8, and the imaging device 9 are included in the learning data, but in one embodiment, the learning data includes the imaging device 7, the imaging device 8, and the imaging device 9. and image data of the object to be dried generated by one or two of the imaging devices 9. A trained model is constructed using this training data. The constructed trained model is input with current image data of the same type as the image data included in the training data (ie, used to construct the trained model). For example, if the image data generated by the imaging device 8 is included in the learning data and the image data generated by the imaging devices 7 and 9 are not included in the learning data, the imaging device 8 The current image data of the dried object generated by is input to the trained model.

In order to output more optimized operating parameters from the trained model, the learning data used to construct the trained model includes past image data of the dried object as well as past image data of the dried object. It may also include status data of the drying system at the time it was generated. A specific example of the drying system status data is as follows.
・Temperature of the material to be dried put into the drying device 5 (measured by a measuring device not shown)
・Flow rate of the material to be dried fed into the drying device 5 (conveying speed of the feeding conveyor 1)
- Moisture content of the material to be dried fed into the drying device 5 (measured by a measuring device not shown)
- Rotation speed of stirring rotor 17 of drying device 5 (rotation speed of electric motor 18)
・Torque current value of the electric motor 18 of the drying device 5 ・Height of the weir 56 (setting value of the vertical movement device 58)
- Pressure of steam injected into stirring rotor 17 (opening degree of steam pressure control valve 52)
・Flow rate of steam injected into stirring rotor 17 (opening degree of steam pressure control valve 52)
-Surface temperature of drying device 5 (measured by temperature measuring device 40)
- Pressure inside the drying device 5 (measured by the pressure measuring device 41)
- Temperature of the dry gas supplied to the housing 15 (measured by the temperature measuring device 32)
- Flow rate of dry gas supplied to the housing 15 (measured by the flow rate measuring device 33)
- Temperature of the dry gas discharged from the housing 15 (measured by the temperature measuring device 35)

The items of the drying system status data described above are examples, and the drying system status data may include only any of the above items.

The learned model includes past image data of the object to be dried generated by the image capturing device 7, image capturing device 8, and image capturing device 9, and state data of the drying system when the past image data of the object to be dried was generated. , is constructed according to a machine learning algorithm using learning data including multiple combinations of optimal operating parameters of the drying system that can keep the moisture content of the dried material discharged from the drying device 5 within the target range. . The past image data of the dried material and the state data of the drying system when the past image data of the dried material were generated are explanatory variables, and the moisture content of the dried material discharged from the drying device 5 is The optimal operating parameters of the drying system that can be kept within the target range are the objective variables. During operation of the drying system, the control system 10 inputs the current image data of the object to be dried generated by the imaging device 7, the imaging device 8, and the imaging device 9 and the current state data of the drying system to the trained model. , causes the processing device 10b to execute calculations for outputting optimal operating parameters of the drying system from the model.

In one embodiment, the trained model is based on image data of the past dried object generated by any one or two of the imaging device 7, the imaging device 8, and the imaging device 9, and the image data of the past dried object. A plurality of state data of the drying system when the image data of the object is generated and optimal operating parameters of the drying system that can keep the moisture content of the object to be dried discharged from the drying device 5 within the target range. It may be constructed according to a machine learning algorithm using training data containing combinations. During operation of the drying system, the control system 10 displays image data of the current object to be dried generated by any one or two of the imaging device 7, the imaging device 8, and the imaging device 9, and the drying system. The current state data of is input to the learned model, and the processing device 10b is caused to execute calculations for outputting the optimal operating parameters of the drying system from the learned model.

Next, another embodiment of the drying system will be described with reference to FIG. FIG. 5 is a diagram showing another embodiment of the drying system. The configuration and operation of this embodiment, which are not particularly described, are the same as those of the embodiment described with reference to FIGS. 1 to 4, and therefore, redundant explanation thereof will be omitted.

In this embodiment, instead of image data, data of sounds generated from the object to be dried is used. That is, the drying system includes a sound detector 71 that detects the sound of an object to be dried that is generated when it is put into the drying device 5 and generates sound data, and a sound detector 71 that detects the sound of the object to be dried that is generated when it is put into the drying device 5 and generates sound data. A sound detector 72 that detects sound and generates sound data, and a sound detector 73 that detects the sound of the drying object generated when it is discharged from the drying chamber 14 of the drying device 5 and generates sound data. We are prepared. The specific configurations of the sound detector 71, the sound detector 72, and the sound detector 73 are not particularly limited, but in this embodiment, a microphone that converts sound into an electrical signal is used. The sound data generated by the sound detector 71, the sound detector 72, and the sound detector 73 may be, for example, sound waveform data showing the relationship between amplitude and time, or sound waveform data showing the relationship between frequency and sound loudness. It is data. Sound data is data consisting of digital signals.

Sound detector 71 is placed within drying chamber 14 . More specifically, the sound detector 71 is disposed below the input port 25 of the housing 15 so that the material to be dried, which is input through the input port 25, is connected to the material to be dried already accommodated in the housing 15. Detects the sound of the dried material that is generated when it falls onto the sediment, and generates sound data. In one embodiment, a collision member (not shown) is arranged within the housing 15 and below the input port 25, and the sound detector 71 detects the drying material generated when falling onto the collision member. The sound may be detected and sound data may be generated. The sound detector 71 is connected to the control system 10, and the sound data generated by the sound detector 71 is sent to the control system 10.

At least a portion of the sound detector 72 is located inside the drying chamber 14 . More specifically, the sound detector 72 is fixed to the upper wall of the housing 15. The sound detector 72 detects the sound of the dried material generated during the process of drying in the drying device 5 (inside the drying chamber 14), that is, the sound of the dried material generated when the dried material is stirred by the stirring rotor 17. Detects and generates sound data. The sound detector 72 is connected to the control system 10 so that the sound data generated by the sound detector 72 is sent to the control system 10.

The sound detector 73 is disposed below the discharge port 26 of the housing 15, detects the sound generated when the material to be dried discharged from the drying chamber 14 falls onto the discharge conveyor 64, and collects sound data. generate. In one embodiment, the sound detector 73 may be located within the exhaust chamber 60 or within the exhaust outlet 26. Additionally, in one embodiment, a collision member (not shown) is arranged within the discharge chamber 60 or within or below the discharge outlet 26, and the sound detector 73 detects when the collision member falls onto the collision member. Sound data may be generated by detecting the sound generated by the dried material. The sound detector 73 is connected to the control system 10, and the sound data generated by the sound detector 73 is sent to the control system 10.

In general, the sound generated by the dried material differs depending on whether the moisture content of the dried material is high or low. Specifically, when the moisture content of the material to be dried is high, a wet sound is generated, whereas when the moisture content of the material to be dried is low, a dry sound is generated. Therefore, the sound of the dried material generated during loading, drying, and discharging changes depending on the moisture content of the dried material.

The control system 10 is configured to obtain sound data of the object to be dried from a sound detector 71, a sound detector 72, and a sound detector 73, and to optimize operating parameters of the drying system based on the sound data. has been done. The control system 10 inputs a storage device 10a storing a learned model constructed by a machine learning algorithm and sound data to the learned model, and adjusts the moisture content of the dried material discharged from the drying device 5 to a target range. The apparatus includes a processing device 10b that executes calculations for outputting optimal operating parameters of the drying system that can be accommodated within the learned model from the learned model.

The trained model uses past sound data of the material to be dried generated by the sound detector 71, the sound detector 72, and the sound detector 73 and the moisture content of the material to be dried discharged from the drying device 5 within the target range. The model is built according to a machine learning algorithm using learning data that includes multiple combinations of the optimal operating parameters of the drying system that can be accommodated. The past sound data of the dried material is an explanatory variable, and the optimal operating parameter of the drying system that can keep the moisture content of the dried material discharged from the drying device 5 within the target range is the objective variable. The optimal operating parameters constituting the learning data are correct data, in which a skilled operator can listen to the sound of loading the material to be dried, the sound of drying the material, the sound of discharging the material, and the sound of drying. These are operating parameters determined based on the moisture content of the material to be dried discharged from the device 5. Learning data is also called training data or teacher data.

In order to optimize the parameters (weights, etc.) of the trained model, training data containing numerous combinations of sound data of the dried object and optimal operating parameters that can keep the moisture content of the cake within the target range. will be prepared. The control system 10 uses this learning data to construct a learned model using a machine learning algorithm. In addition to weights, the parameters of a trained model may include bias. The trained model constructed in this way is stored in the storage device 10a.

The control system 10 acquires sound data from the sound detector 71, the sound detector 72, and the sound detector 73 at a predetermined period, and stores it in the storage device 10a. The control system 10 inputs the sound data to the trained model, and outputs from the trained model optimal operating parameters that can keep the moisture content of the material to be dried within the target range.

In this embodiment, all of the sound data generated by the sound detector 71, the sound detector 72, and the sound detector 73 are included in the learning data, but in one embodiment, the learning data is It may also include sound data of the object to be dried generated by one or two of the sound detector 72 and the sound detector 73. A trained model is constructed using this training data. The constructed trained model is input with current sound data of the same type as the sound data included in the training data (that is, used to construct the trained model). For example, if the sound data generated by the sound detector 72 is included in the learning data, but the sound data generated by the sound detectors 71 and 73 are not included in the learning data, during the operation of the drying system, The current sound data of the object to be dried generated by the sound detector 72 is input to the trained model.

In order to output more optimized operating parameters from the model, the training data used to construct the trained model includes sound data of past drying objects as well as sound data of past drying objects. It may also include status data of the drying system at the time of the drying system. A specific example of the state data of the drying system is the same as in the embodiment described above. The trained model is based on the sound data of the past drying object generated by the sound detector 71, the sound detector 72, and the sound detector 73, and the sound data of the drying system when the past sound data of the drying object was generated. Using learning data including multiple combinations of state data and optimal operating parameters of the drying system that can keep the moisture content of the material to be dried discharged from the drying device 5 within the target range, the process is performed according to a machine learning algorithm. Constructed. The sound data of the past drying object and the state data of the drying system when the past sound data of the drying object were generated are explanatory variables, and the moisture content of the drying object discharged from the drying device 5 is The optimal operating parameters of the drying system that can be kept within the target range are the objective variables. During operation of the drying system, the control system 10 uses the current sound data of the object to be dried generated by the sound detector 71, the sound detector 72, and the sound detector 73 and the current state data of the drying system as a trained model. , and causes the processing device 10b to execute calculations for outputting the optimal operating parameters of the drying system from the learned model.

In one embodiment, the trained model uses past sound data of the to-be-dried object generated by any one or two of the sound detector 71, the sound detector 72, and the sound detector 73, and the past The state data of the drying system when the sound data of the dried material is generated, and the optimal operating parameters of the drying system that can keep the moisture content of the dried material discharged from the drying device 5 within the target range. may be constructed according to a machine learning algorithm using training data including multiple combinations of. During operation of the drying system, the control system 10 receives sound data of the current object to be dried, which is generated by any one or two of the sound detector 71, the sound detector 72, and the sound detector 73. , inputs current state data of the drying system into the learned model, and causes the processing device 10b to execute calculations for outputting optimal operating parameters of the drying system from the learned model.

The embodiment described above with reference to FIG. 5 may be combined with the embodiment described with reference to FIGS. 1 to 4. That is, the drying system shown in FIG. 5 may further include at least one of the imaging device 7, the imaging device 8, and the imaging device 9 of the embodiment shown in FIG. In this case, the trained model is constructed according to a machine learning algorithm using learning data including past sound data and past image data. During operation of the drying system, both audio and image data are input to the trained model.

Next, another embodiment of the drying system will be described with reference to FIG. 6. FIG. 6 is a diagram showing another embodiment of the drying system. The configuration and operation of this embodiment, which are not particularly described, are the same as those of the embodiment described with reference to FIGS. 1 to 4, and therefore, redundant explanation thereof will be omitted.

In this embodiment, vibration data is used instead of image data. That is, the drying system includes an elastic member 81 with which an object to be dried that is put into the drying device 5 comes into contact, a vibration detector 82 that is fixed to the elastic member 81 and generates vibration data of the elastic member 81, and a vibration detector 82 that is fixed to the elastic member 81 and generates vibration data of the elastic member 81, and The apparatus includes an elastic member 85 with which the material to be dried discharged from the drying chamber 14 contacts, and a vibration detector 86 fixed to the elastic member 85 and generating vibration data of the elastic member 85. Although the specific configurations of the vibration detector 82 and the vibration detector 86 are not particularly limited, in this embodiment, a vibration sensor or an acceleration sensor that converts vibration into an electrical signal is used.

The vibration detector 82 is fixed to an elastic member 81 disposed below the input port 25 of the housing 15. Elastic member 81 is arranged within drying chamber 14 . In this embodiment, the elastic member 81 is plate-shaped. One end of the elastic member 81 is fixed to the side wall of the housing 15, and the vibration detector 82 is fixed to the other end of the elastic member 81. The material to be dried that is input through the input port 25 falls onto the elastic member 81 , and the vibration detector 82 vibrates together with the elastic member 81 . The vibration detector 82 detects vibrations generated when the object to be dried falls onto the elastic member 81, and generates vibration data. Vibration detector 82 is connected to control system 10 such that vibration data generated by vibration detector 82 is sent to control system 10 .

The vibration detector 86 is fixed to an elastic member 85 disposed below the outlet 26 of the housing 15. In this embodiment, the elastic member 85 is plate-shaped. One end of the elastic member 85 is fixed to a fixed member (not shown), and a vibration detector 86 is fixed to the other end of the elastic member 85. In one embodiment, elastic member 85 and vibration detector 86 may be located within discharge chamber 60 or within discharge port 26 .

The material to be dried discharged from the drying chamber 14 falls onto the elastic member 85, and the vibration detector 86 vibrates together with the elastic member 85. The vibration detector 86 detects vibrations generated when the object to be dried falls onto the elastic member 85, and generates vibration data. Vibration detector 86 is connected to control system 10 such that vibration data generated by vibration detector 86 is sent to control system 10 .

The vibration data generated by the vibration detector 82 and the vibration detector 86 is, for example, data indicating the relationship between amplitude and time, and is data consisting of a digital signal. Generally, the manner in which the dried material vibrates when it collides with the elastic member 81 and the elastic member 85 is different depending on whether the moisture content of the dried material is high or low. Specifically, when the moisture content of the material to be dried is high, larger lumps come into contact with the elastic members 81 and 85, and the period of vibration tends to become shorter. On the other hand, when the moisture content of the material to be dried is low, smaller lumps come into contact with the elastic members 81 and 85, and the period of vibration tends to become longer. Therefore, the vibration data of the elastic member 81 and the elastic member 85 changes depending on the moisture content of the material to be dried.

Control system 10 is configured to obtain vibration data from vibration detector 82 and vibration detector 86 and to optimize operating parameters of the drying system based on the vibration data. The control system 10 inputs a storage device 10a storing a learned model constructed by a machine learning algorithm and vibration data to the learned model, and adjusts the moisture content of the dried material discharged from the drying device 5 to a target range. The apparatus includes a processing device 10b that executes calculations for outputting optimal operating parameters of the drying system that can be accommodated within the learned model from the learned model.

The learned model uses past vibration data of the elastic members 81 and 85 generated by the vibration detector 82 and the vibration detector 86 and the moisture content of the dried material discharged from the drying device 5 to be within a target range. It is a model built according to a machine learning algorithm using learning data that includes multiple combinations of the optimal operating parameters of the drying system. The past vibration data is an explanatory variable, and the optimal operating parameter of the drying system that can keep the moisture content of the material to be dried discharged from the drying device 5 within the target range is the objective variable. The optimal operating parameters constituting the learning data are the correct data, which is created by a skilled operator using vibration data when the material to be dried is input, vibration data when the material to be dried is discharged, and data on the vibration when the material is discharged from the drying device 5. These are operating parameters determined based on the moisture content of dry matter. Learning data is also called training data or teacher data.

In order to optimize the parameters (weights, etc.) of the trained model, numerous combinations of past vibration data of the elastic members 81 and 85 and optimal operating parameters that can keep the moisture content of the cake within the target range are performed. Learning data including . The control system 10 uses this learning data to construct a learned model using a machine learning algorithm. In addition to weights, the parameters of a trained model may include bias. The trained model constructed in this way is stored in the storage device 10a.

The control system 10 acquires vibration data from the vibration detector 82 and the vibration detector 86 at a predetermined period, and stores it in the storage device 10a. The control system 10 inputs the vibration data of the elastic members 81 and 85 to the learned model, and outputs from the learned model optimal operating parameters that can keep the moisture content of the dried material within the target range.

In this embodiment, the learning data includes vibration data generated by both vibration detector 82 and vibration detector 86; however, in one embodiment, the learning data includes vibration data generated by both vibration detector 82 and vibration detector 86. It may also include vibration data generated by. A trained model is constructed using this training data. The constructed trained model is input with current vibration data of the same type as the vibration data included in the training data (i.e., used to construct the trained model). For example, if vibration data generated by vibration detector 86 is included in the learning data and vibration data generated by vibration detector 82 is not included in the learning data, then during operation of the drying system, vibration data generated by vibration detector 86 is included in the training data. The current vibration data of the elastic member 85 generated by is input to the learned model.

In order to output more optimized operating parameters from the model, the training data used to construct the trained model includes past vibration data as well as state data of the drying system when the past vibration data was generated. May include. A specific example of the state data of the drying system is the same as in the embodiment described above.

The learned model includes past vibration data of the elastic members 81 and 85 generated by the vibration detector 82 and the vibration detector 86, state data of the drying system when the past vibration data was generated, and the drying device 5. It is constructed according to a machine learning algorithm using learning data that includes multiple combinations of optimal operating parameters of the drying system that can keep the moisture content of the dried material discharged from the drying system within the target range. The past vibration data and the state data of the drying system when the past vibration data was generated are explanatory variables, and it is possible to keep the moisture content of the dried material discharged from the drying device 5 within the target range. The optimal operating parameters of the drying system are the objective variables. During operation of the drying system, the control system 10 inputs current vibration data of the elastic members 81, 85 generated by the vibration detector 82 and vibration detector 86 and current state data of the drying system into the trained model. , causes the processing device 10b to execute calculations for outputting optimal operating parameters of the drying system from the model.

In one embodiment, the trained model is based on historical vibration data generated by one of vibration detector 82 and vibration detector 86 and state data of the drying system at the time the past vibration data was generated. , is constructed according to a machine learning algorithm using learning data including multiple combinations of optimal operating parameters of the drying system that can keep the moisture content of the dried material discharged from the drying device 5 within the target range. Good too. During operation of the drying system, control system 10 inputs current vibration data generated by the one of vibration detectors 82 and vibration detectors 86 and current state data of the drying system into the trained model. , causes the processing device 10b to execute calculations for outputting optimal operating parameters of the drying system from the learned model.

The embodiment described above with reference to FIG. 6 may be combined with the embodiment described with reference to FIGS. 1 to 4. That is, the drying system shown in FIG. 6 may further include at least one of the imaging device 7, the imaging device 8, and the imaging device 9 of the embodiment shown in FIG. The trained model is constructed according to a machine learning algorithm using learning data including past vibration data and past image data. During operation of the drying system, both vibration data and image data are input to the trained model.

Furthermore, the embodiment described with reference to FIG. 5 and the embodiment described with reference to FIG. 6 may be combined with the embodiment described with reference to FIGS. 1 to 4. That is, the drying system shown in FIG. 6 includes at least one of the imaging device 7, the imaging device 8, and the imaging device 9 of the embodiment shown in FIG. 1, the sound detector 71, the sound detector 72, and and a sound detector 73. In this case, the model is built according to a machine learning algorithm using training data including past vibration data, past sound data, and past image data. While the drying system is operating, vibration, sound, and image data are input to the trained model.

Furthermore, the embodiment described with reference to FIG. 5 may be combined with the embodiment described with reference to FIG. 6. That is, the drying system shown in FIG. 6 may further include at least one of the sound detector 71, the sound detector 72, and the sound detector 73 shown in FIG. In this case, the model is constructed according to a machine learning algorithm using training data including past vibration data and past sound data. During operation of the drying system, both vibration and sound data are input to the trained model.

The drying system according to each of the embodiments described above is suitably used for drying a dehydrated cake discharged from a dewatering system equipped with a dewatering device such as a screw press. Therefore, in order to output more optimized operating parameters from the trained model, in addition to at least one of image data, sound data, and vibration data, data from the dehydration system located upstream of the drying system is State data may be input to a trained model. The training data used to construct the trained model includes state data of the dehydration system in addition to at least one of past image data, past sound data, and past vibration data. The dehydration system will be explained in detail below.

The dewatering system is arranged upstream of the input conveyor 1 in FIG. 1 and is configured to dewater a suspension such as sludge. FIG. 7 shows a dehydration system that includes a suspension storage tank 101 that stores a suspension, and a flocculation device that injects a flocculant into the suspension and forms flocs by stirring the flocculant and suspension. 103, a concentrator 104 for removing liquid from the aggregate to form a concentrate, and a dehydrator 105 for further removing liquid from the concentrate comprising the concentrated aggregate to form a turbid residue. . The concentrating device 104 is a device that removes liquid from the aggregates formed by the aggregating device 103 to form a concentrate having a concentration of 3 to 15 wt%, and the dehydrating device 105 removes the liquid from the aggregates formed by the aggregating device 103. Alternatively, it is a device that removes liquid from the concentrate formed by the concentrator 104 to form a turbid residue having a concentration of 3 to 50 wt%.

A turbid residue is a low water content material that remains after removing liquid from a suspension. The turbid residue that remains after removing liquid from the suspension by dehydrator 105 is commonly referred to as cake. In the following description, the suspension will be referred to as sludge and the turbid residue will be referred to as cake. Specific examples of sludge include sludge generated during the treatment of sewage, human waste, and industrial wastewater. Furthermore, examples of suspensions other than sludge include industrial waste or slurry generated during the manufacture of industrial products such as foodstuffs, cosmetics, and paper.

In this embodiment, the concentrating device 104 is a rotating disk type deliquid device (for example, Slit Saver manufactured by Kenden Co., Ltd.). This type of thickening device 104 is preferable because it is easy to understand changes in the state of the thickened sludge, but the type of thickening device 104 is not particularly limited. For example, the concentrator 104 may be drum-type, screw-type, or belt-type. The dewatering device 105 of this embodiment is comprised of a screw press that dewaters sludge by pressurizing it. A screw press is an example of a pressure dehydrator that dewaters sludge. A press-fit type, a shaft sliding type, or a two-stage type can be used as the screw press. The dewatering device 105 shown in FIG. 7 is a sliding type screw press. In addition to the screw press, other types of pressurized dehydrators such as a filter press and a centrifugal separator can also be used.

As shown in FIG. 7, the flocculation device 103 includes a flocculation-mixing tank 107 that accommodates sludge and mixes the sludge with a flocculant, an agitator 108 for stirring the sludge in the flocculation-mixing tank 107, and a flocculation-mixing tank 107. A sludge introduction pipe (suspension introduction pipe) 114 connected to the sludge introduction pipe 114 and a pump 112 provided on the sludge introduction pipe 114 are provided. The stirrer 108 includes a stirring blade 109 disposed in the coagulation mixing tank 107 and a stirring motor 110 connected to the stirring blade 109. The suspension storage tank 101 is connected to a coagulation mixing tank 107 by a sludge introduction pipe 114. The sludge is transferred from the suspension tank 101 to the flocculation mixing tank 107 through the sludge introduction pipe 114 by the pump 112 . The flow rate of sludge to the coagulation mixing tank 107 can be adjusted by operating the pump 112.

A temperature sensor 122 is attached to the sludge introduction pipe 114. This temperature sensor 122 is configured to measure the temperature of sludge introduced into the coagulation mixing tank 107. Furthermore, a sludge concentration meter (suspension concentration meter) 124 is attached to the sludge introduction pipe 114, and the concentration of suspended solids in the sludge introduced into the coagulation mixing tank 107 is measured by the sludge concentration meter 124. .

The aggregation device 103 further includes a water supply line 126 connected to the aggregation mixing tank 107 and a flow rate control valve 127 attached to the water supply line 126. The water supply line 126 supplies water into the coagulation mixing tank 107 to dilute the sludge in the coagulation mixing tank 107 . The flow rate of water supplied to the coagulation mixing tank 107 through the water supply line 126 is regulated by a flow rate control valve 127 . In one embodiment, water supply line 126 may be connected to sludge inlet tube 114 and supply water directly to sludge inlet tube 114 .

The flocculating device 103 further includes a flocculant supply device 128 connected to the flocculating mixing tank 107. This coagulant supply device 128 is configured to inject the coagulant into the sludge in the coagulation mixing tank 107 at a predetermined flow rate. The flow rate of the flocculant injected into the sludge in the flocculation mixing tank 107 is adjusted by the flocculant supply device 128. The sludge is stirred together with the flocculant by an agitator 108. By stirring the flocculant and sludge in the flocculating mixing tank 107, flocs are formed by aggregation of suspended substances in the sludge. The flocs formed in the flocculation mixing tank 107 are generally called flocs.

Although the aggregation mixing tank 107 shown in FIG. 7 is a single-stage tank, the aggregation-mixing tank 107 may be a multi-stage tank. The intensity of stirring the sludge can be adjusted by the rotation speed of the stirrer 108. A side glass may be installed in the aggregation mixing tank 107 so that the state of the agglomerates inside can be observed.

The flocculation mixing tank 107 is connected to the thickening device 104 by a sludge transfer pipe 118. Concentrating device 104 is arranged between aggregating device 103 and dehydrating device 105. Sludge consisting of flocs formed by flocculation device 103 is transferred to thickening device 104 through sludge transfer pipe 118. The aggregate is concentrated and dehydrated by the concentrator 104. An example of the concentration device 104 is a rotating disk type deliquor device.

The outlet of the thickener 104 is arranged above the input port 135 of the dehydrator 105, and the sludge made of the concentrate formed by the thickener 104 is input into the input port 135 of the dewaterer 105. In this embodiment, the dewatering device 105 is a screw press. The dewatering device 105 as a screw press includes a filtration tube 136, a screw shaft 137 arranged concentrically within the filtration tube 136, a screw blade 138 fixed to the outer surface of the screw shaft 137, and the screw shaft 137 and the screw blade. The apparatus includes a screw motor 140 that rotates a screw motor 138 to send sludge toward a discharge chamber 141, and a shaft sliding actuator 142 connected to the screw motor 140.

The filter tube 136 is made of a perforated plate such as punched metal. One end of the filter tube 136 is sealed by a closing wall 144, and the other end of the filter tube 136 is connected to the discharge chamber 141. The input port 135 is formed in the filter tube 136 and is adjacent to the closing wall 144 .

The screw shaft 137 extends through the filter cylinder 136. The screw shaft 137 has a truncated conical shape whose diameter gradually increases toward the downstream side. The screw shaft 137 extends through the closing wall 144 , and the end of the screw shaft 137 is connected to the screw motor 140 . The screw motor 140 has a built-in inverter (not shown).

The screw blade 138 is a single blade that extends spirally along the longitudinal direction of the screw shaft 137. A minute gap is formed between the inner surface of the filter tube 136 and the screw blade 138, and the screw blade 138 can rotate without contacting the filter tube 136. The sludge introduced into the filter tube 136 from the input port 135 is transferred through the filter tube 136 toward the discharge chamber 141 by the rotating screw blades 138 .

A space in which sludge is transferred within the filter tube 136 is formed by the inner surface of the filter tube 136, the screw blades 138, and the screw shaft 137. The volume of this space gradually decreases along the direction of movement of the sludge. Therefore, as the sludge is moved through this space by the screw blades 138, it is squeezed and dewatered. The filtrate that has passed through the filter tube 136 is collected by a filtrate receiver 145 disposed below the filter tube 136 and then discharged through a drain 146.

An annular back pressure plate 150 is disposed opposite the downstream end of the filter tube 136. This back pressure plate 150 has a truncated conical shape with a tapered surface for receiving the dehydrated sludge transferred within the filter tube 136. A through hole through which the screw shaft 137 passes is formed in the center of the back pressure plate 150, and the back pressure plate 150 is arranged concentrically with the screw shaft 137. The back pressure plate 150 is not fixed to the screw shaft 137, and the back pressure plate 150 does not rotate.

The back pressure plate 150 is connected to a back pressure plate drive device 151. This back pressure plate driving device 151 is configured to move the back pressure plate 150 in the axial direction of the screw shaft 137. The gap between the back pressure plate 150 and the downstream end of the filtration tube 136 is adjusted by the back pressure plate drive device 151. The back pressure plate drive device 151 is composed of, for example, a hydraulic cylinder or an electric cylinder.

The shaft sliding actuator 142 is configured to move the screw motor 140 in the axial direction of the screw shaft 137. When the shaft sliding actuator 142 moves the screw motor 140 in the axial direction, the screw shaft 137 and the screw blades 138 connected to the screw motor 140 are moved in the axial direction of the screw shaft 137 within the filtration cylinder 136.

Next, the operation of the dewatering device 105 will be explained. Sludge consisting of the concentrate formed by the concentrator 104 is charged into the filter cylinder 136 from the input port 135. The sludge is transferred within the filter tube 136 toward the discharge chamber 141 by the rotating screw blades 138 . As the sludge is moved through the filter tube 136, it is squeezed and dewatered. The filtrate that has passed through the filter tube 136 is collected by a filtrate receiver 145 and discharged through a drain 146. The sludge is dewatered within the filter tube 136 to form a cake. The cleaning device 155 supplies cleaning liquid to the outer peripheral surface of the filter tube 136 at preset time intervals.

The cake that has moved inside the filter tube 136 is pressed against the back pressure plate 150. The cake is compressed by being prevented from moving by the back pressure plate 150. This compressed cake forms a plug 152 that seals the downstream end of the filter barrel 136. Plug 152 reduces the moisture content of the cake within filter barrel 136 by applying back pressure to the subsequent cake. While forming a plug 152 in the filter tube 136, the cake is pushed by the following cake, passes through the gap between the back pressure plate 150 and the downstream end of the filter tube 136, and gradually enters the discharge chamber 141. It is discharged. The cake is discharged from the discharge chamber 141 through a discharge port 153 provided at the bottom of the discharge chamber 141 . In this way, liquid is removed from the sludge and a cake with low moisture content is formed. The cake is conveyed to the input port 25 of the drying device 5 by the input conveyor 1 shown in FIG. 1, FIG. 5, or FIG. 6.

The gap between the back pressure plate 150 and the downstream end of the filtration tube 136 changes depending on the axial position of the back pressure plate 150, and as a result, the compressive force applied to the sludge in the filtration tube 136 changes. More specifically, when the gap between the back pressure plate 150 and the downstream end of the filtration tube 136 becomes smaller, a larger force is required to push out the plug 152, which reduces the compression applied to the sludge in the filtration tube 136. Power increases. Therefore, the compressive force applied to the sludge can be adjusted not only by the rotational speed of the screw blades 138 but also by the position of the back pressure plate 150. A pressure sensor 161 is disposed inside the filter tube 136, and the pressure of the sludge in the filter tube 136 is measured by the pressure sensor 161. The position of the pressure sensor 161 is not particularly limited as long as it is inside the filtration tube 136, and may be in the front, middle, or rear stage of the filtration tube 136. It's okay.

Plug 152 is formed from a low moisture content cake. When the moisture content of the cake forming the plug 152 decreases, the plug 152 becomes hard and may block the downstream end of the filter tube 136. In such a case, the shaft sliding actuator 142 moves the screw shaft 137 and the screw blades 138 in the axial direction, thereby forcibly discharging the hard cake with a low water content. As a result, the dehydrator 105 can operate stably and continuously.

The training data used to construct the trained model includes state data of the dehydration system in addition to at least one of past image data, past sound data, and past vibration data. A specific example of the status data of the dehydration system is as follows.
-Sludge temperature (measured by temperature sensor 122)
・Flow rate of sludge to the coagulation mixing tank 107 (estimated from the rotation speed of the pump 112)
・Sludge properties (pH, alkalinity, TS, VTS, SS, VSS, amount of suspended solids)
・Physical properties of flocculant (cation degree, molecular weight)
- Flow rate of flocculant (operating speed of flocculant supply device 128)
- Stirring speed of the stirrer 108 (rotational speed of the stirring motor 110)
・Flow rate of water to the flocculation mixing tank 107 (opening degree of flow rate control valve 127)
- Operating speed of the concentrator 104 (rotational speed of the drive motor of the concentrator 104)
- Operating torque of the concentrator 104 (current value to the drive motor of the concentrator 104)
- Flow rate of the liquid separated by the concentrator 104 (measured by a flow meter not shown)
-Sludge concentration concentrated by the thickening device 104 (measured by a concentration sensor not shown)
・Rotational speed of screw shaft 137 ・Current value (torque value) to screw motor 140
- Pressure of sludge in filter tube 136 (measured by pressure sensor 161)
・Opening degree of back pressure plate 150 (set value of back pressure plate drive device 151)
- Pressure of the back pressure plate 150 (measured by a pressure sensor not shown)
・Measurement of cake moisture content (measured by a measuring device not shown or manually)

The items of the status data of the dehydration system described above are examples, and the status data of the dehydration system may include only any of the above items.

The trained model includes at least one of past image data, past sound data, and past vibration data, state data of the dehydration system when the past data was generated, and water discharged from the drying device 5. It is constructed according to a machine learning algorithm using learning data that includes multiple combinations of optimal operating parameters of the drying system that can keep the moisture content of the dried material within the target range. Past image data, past sound data, past vibration data, and dehydration system status data are explanatory variables, and it is possible to keep the moisture content of the dried material discharged from the drying device 5 within the target range. The optimal operating parameters of the drying system are the objective variables. During operation of the drying system, the control system 10 shown in FIG. 1, FIG. 5, or FIG. The processing device 10b is caused to perform calculations for outputting the optimal operating parameters of the drying system from the learned model.

In one embodiment, the trained model includes at least one of historical image data, historical sound data, and historical vibration data, and state data of a drying system and a dehydration system at the time the historical data was generated. Machine learning algorithm It may be constructed according to During operation of the drying system, the control system 10 shown in FIG. 1, FIG. 5, or FIG. The current state data is input to the learned model, and the processing device 10b is caused to execute calculations for outputting the optimal operating parameters of the drying system from the learned model.

FIG. 8 is a diagram showing another embodiment of the dehydration system. The dewatering system shown in FIG. 8 differs from the embodiment shown in FIG. 7 in that a two-stage screw press is employed as the dewatering device 105, and a shaft-sliding screw press is employed. In this embodiment, the shaft sliding actuator 142 shown in FIG. 7 is not provided. The configuration and operation of the embodiment not particularly described are the same as those of the embodiment shown in FIG. 7, so the redundant explanation thereof will be omitted.

The dewatering device 105, which is composed of a two-stage screw press, includes a filtration tube 136, and a first screw 171 and a second screw 172, which are arranged concentrically with the filtration tube 136 within the filtration tube 136 and which transport sludge. , a first screw motor 175 that rotates a first screw 171, and a second screw motor 176 that rotates a second screw 172 independently of the first screw 171.

The first screw motor 175 and the second screw motor 176 may be directly connected to the first screw 171 and the second screw 172, respectively, or may be connected to the first screw via a torque transmission mechanism composed of a chain, a sprocket, etc. 171 and the second screw 172, respectively.

The filter tube 136 is formed from a screen (perforated plate) such as a punched metal. An input port 135 is formed at the upstream end of the filter tube 136 . Sludge introduced into the filter tube 136 from the input port 135 is transferred in a predetermined transfer direction within the filter tube 136 by the rotating first screw 171 and second screw 172. A downstream end of the filter tube 136 is connected to the discharge chamber 141 .

The second screw 172 is connected to the first screw 171 so as to be rotatable independently of the first screw 171. The first screw 171 and the second screw 172 extend through the filter cylinder 136 and the discharge chamber 141, respectively. The discharge chamber 141 is connected to the filter tube 136. A plug 177 made of cake is discharged from the filter tube 136 into the discharge chamber 141 . The length of the second screw 172 in the axial direction is shorter than the length of the first screw 171 in the axial direction. The first screw 171 has a truncated cone-shaped first screw shaft 171A whose diameter gradually increases along the sludge transfer direction, and a first screw blade 171B fixed to the outer surface of the first screw shaft 171A. are doing. The second screw 172 has a cylindrical second screw shaft 172A and a second screw blade 172B fixed to the outer surface of the second screw shaft 172A.

The upstream end of the filter cylinder 136 is sealed by a closing wall 144 . The upstream end of the first screw shaft 171A extends through the closing wall 144 and is connected to a first screw motor 175 for rotating the first screw 171. The second screw shaft 172A of the second screw 172 is arranged concentrically with the first screw shaft 171A. The outer diameter of the second screw shaft 172A is the same as the maximum diameter of the first screw shaft 171A. The downstream end of the second screw shaft 172A extends through the wall 141A that constitutes the discharge chamber 141, and is connected to a second screw motor 176 for rotating the second screw 172. The lower part of the discharge chamber 141 is connected to the discharge port 153.

The first screw motor 175 has a built-in inverter (not shown), and the motor control unit 106 is configured to be able to control the operation of the first screw motor 175 via the inverter. That is, the motor control unit 106 can control the rotation speed and rotation direction of the first screw motor 175 via the inverter. The motor control unit 106 can issue a command to the first screw motor 175 to rotate the first screw 171 independently of the second screw 172.

The second screw motor 176 is also connected to the motor control section 106. The second screw motor 176 has a built-in inverter (not shown), and the motor control unit 106 is configured to be able to control the operation of the second screw motor 176 via the inverter. That is, the motor control unit 106 can control the rotation speed and rotation direction of the second screw motor 176 via the inverter. The motor control unit 106 can issue a command to the second screw motor 176 to rotate the second screw 172 independently of the first screw 171.

The first screw blade 171B extends helically along the axial direction of the first screw shaft 171A, and the second screw blade 172B spirally extends along the axial direction of the second screw shaft 172A. The total length of the portion of the first screw 171 to which the first screw blade 171B is fixed and the portion of the second screw 172 to which the second screw blade 172B is fixed is the axial length of the filter tube 136. Same or longer.

A minute gap is formed between the inner surface of the filter tube 136 and the first screw blade 171B, so that the first screw blade 171B can rotate without contacting the filter tube 136. Similarly, a minute gap is formed between the inner surface of the filter tube 136 and the second screw blade 172B, so that the second screw blade 172B can rotate without contacting the filter tube 136. ing. The sludge introduced into the filter tube 136 from the input port 135 formed at the upstream end of the filter tube 136 can be transferred toward the discharge chamber 141 by the rotating first screw blade 171B and second screw blade 172B. can.

The pitch of the second screw blade 172B is smaller than the pitch of the first screw blade 171B. Furthermore, the number of turns of the second screw blade 172B is less than three turns. In this embodiment, the winding direction (that is, the helical direction) of the second screw blade 172B is opposite to the winding direction of the first screw blade 171B. Therefore, when sending out the sludge introduced from the input port 135 to the discharge chamber 141, the second screw 172 is rotated in the opposite direction to the first screw 171, as shown in FIG.

The winding direction of the second screw blade 172B may be the same as the winding direction of the first screw blade 171B. In this case, when sending out the sludge introduced from the input port 135 to the discharge chamber 141, the second screw 172 is rotated in the same direction as the first screw 171.

The filter tube 136 is divided into a dewatering area 1A where the first screw 171 is placed and a plug forming area 1B where the second screw 172 is placed. A space into which sludge is transferred in the dewatering area 1A is formed by the inner surface of the filter cylinder 136, the first screw blade 171B, and the first screw shaft 171A. The cross-sectional area of this transfer space gradually decreases along the sludge transfer direction, as shown in FIG. Therefore, as the sludge introduced from the input port 135 is transferred through this transfer space by the first screw blades 171B, the sludge is compressed and dewatered. The filtrate that has passed through the filter tube 136 is collected by a filtrate receiver 145 arranged below the filter tube 136. A drain 146 is connected to the filtrate receiver 145 , and the filtrate collected by the filtrate receiver 145 is discharged from the dehydrator 105 via the drain 146 .

The space into which sludge is transferred in the plug formation region 1B is formed by the inner surface of the filter cylinder 136, the second screw blade 172B, and the second screw shaft 172A. As shown in FIG. 8, the cross-sectional area of this transfer space is constant. In the plug forming region 1B, a plug 177 is formed by the sludge (ie, cake) dehydrated in the dewatering region 1A. Plug 177 reduces the moisture content of the cake in filter tube 136 by applying back pressure to the subsequent cake. While forming a plug 177 within the plug forming area 1B, the cake is pushed by subsequent cakes and is discharged little by little into the discharge chamber 141. The cake is discharged from the discharge chamber 141 through a discharge port 153 provided at the bottom of the discharge chamber 141 . In this way, liquid is removed from the sludge and a cake with low moisture content is formed. The cake is conveyed to the input port 25 of the drying device 5 by the input conveyor 1 shown in FIG. 1, FIG. 5, or FIG. 6.

A specific example of the status data of the dehydration system of this embodiment is as follows.
-Sludge temperature (measured by temperature sensor 122)
・Flow rate of sludge to coagulation mixing tank 107 (rotational speed of pump 112)
・Sludge properties (pH, alkalinity, TS, VTS, SS, VSS, amount of suspended solids)
・Physical properties of flocculant (cation degree, molecular weight)
- Flow rate of flocculant (operating speed of flocculant supply device 128)
- Stirring speed of the stirrer 108 (rotational speed of the stirring motor 110)
・Flow rate of water to the flocculation mixing tank 107 (opening degree of flow rate control valve 127)
- Operating speed of the concentrator 104 (rotational speed of the drive motor)
- Operating torque of concentrator 104 (current value to drive motor)
- Flow rate of the liquid separated by the concentrator 104 (measured by a flow meter not shown)
-Sludge concentration concentrated by the thickening device 104 (measured by a concentration sensor not shown)
- Rotational speed of the first screw 171 - Current value (torque value) to the first screw motor 175
- Rotational speed of the second screw 172 - Current value (torque value) to the second screw motor 176
- Pressure of sludge in filter tube 136 (measured by pressure sensor 161)
・Measurement of cake moisture content (measured by a measuring device not shown or manually)

The items of the status data of the dehydration system described above are examples, and the status data of the dehydration system may include only any of the above items.

The trained model includes at least one of past image data, past sound data, and past vibration data, state data of the dehydration system when the past data was generated, and water discharged from the drying device 5. It is constructed according to a machine learning algorithm using learning data that includes multiple combinations of optimal operating parameters of the drying system that can keep the moisture content of the dried material within the target range. Past image data, past sound data, past vibration data, and dehydration system status data are explanatory variables, and it is possible to keep the moisture content of the dried material discharged from the drying device 5 within the target range. The optimal operating parameters of the drying system are the objective variables. During operation of the drying system, the control system 10 shown in FIG. 1, FIG. 5, or FIG. The processing device 10b is caused to perform calculations for outputting the optimal operating parameters of the drying system from the learned model.

In one embodiment, the trained model includes at least one of historical image data, historical sound data, and historical vibration data, and state data of a drying system and a dehydration system at the time the historical data was generated. Machine learning algorithm It may be constructed according to During operation of the drying system, the control system 10 shown in FIG. 1, FIG. 5, or FIG. The current state data is input to the learned model, and the processing device 10b is caused to execute calculations for outputting the optimal operating parameters of the drying system from the learned model.

In the embodiment shown in FIGS. 7 and 8, the dehydration system includes a concentrator 104, but in one embodiment, the concentrator 104 may be omitted. In this case, the aggregates (agglomerated flocs) formed in the flocculation device 103 are transferred to the input port 135 of the dewatering device 105 through the sludge transfer pipe 118.

FIG. 9 is a diagram showing still another embodiment of the dehydration system. In the dehydration system shown in FIG. 9, a belt press type dehydration device is employed as the dehydration device 105. The above-mentioned concentrator 104 is not provided. The configuration and operation of the embodiment not particularly described are the same as those of the embodiment shown in FIG. 7, so the redundant explanation thereof will be omitted.

The flocculation device 103 includes a granulation tank 180 that stores sludge, an agitator 108 for stirring the sludge in the granulation tank 180, a sludge introduction pipe 114 connected to the granulation tank 180, and a sludge introduction pipe 114 connected to the sludge introduction pipe 114. A pump 112 is provided. The stirrer 108 includes a stirring blade 109 disposed within the granulation tank 180 and a stirring motor 110 connected to the stirring blade 109. The suspension storage tank 101 is connected to the coagulation and granulation tank 180 by a sludge introduction pipe 114. The sludge is transferred from the suspension tank 101 to the granulation tank 180 by the pump 112 through the sludge introduction pipe 114. The flow rate of sludge to the granulation tank 180 can be adjusted by operating the pump 112.

A temperature sensor 122 is attached to the sludge introduction pipe 114. This temperature sensor 122 is configured to measure the temperature of sludge introduced into the granulation tank 180. Furthermore, a sludge concentration meter (suspension concentration meter) 124 is attached to the sludge introduction pipe 114, and the concentration of suspended solids in the sludge introduced into the granulation tank 180 is measured by the sludge concentration meter 124. .

The flocculation device 103 further includes a flocculant supply device 128 connected to the granulation tank 180. This flocculant supply device 128 is configured to inject the flocculant into the sludge in the granulation tank 180 at a predetermined flow rate. The flow rate of the flocculant injected into the sludge in the granulation tank 180 is adjusted by the flocculant supply device 128. The sludge is stirred together with the flocculant by the stirring blades 109 of the stirrer 108. By stirring the flocculant and sludge in the granulation tank 180, aggregates are formed by aggregation of suspended substances in the sludge. The agglomerates formed within granulation tank 180 are generally referred to as pellets. The intensity of stirring the sludge can be adjusted by the rotation speed of the stirrer 108.

The granulation tank 180 is connected to the belt press type dewatering device 105 by a sludge transfer pipe 118. The dewatering device 105 includes an endless pretreatment filter cloth belt 182, a plurality of support rollers 183a to 183c that support the pretreatment filter cloth belt 182, an endless first filter cloth belt 186, and a first filter cloth belt. 186, an endless second filter cloth belt 190, second guide rollers 191a to 191f that support the second filter cloth belt 190, and first guide rollers 188a to 188e that support the filter cloth belt 186; It includes a plurality of pressing rollers 195A to 195E that support two filter cloth belts 190, and a drive device 198 that moves the first filter cloth belt 186 and the second filter cloth belt 190.

The pretreatment filter cloth belt 182 is arranged upstream of the squeezing rollers 195A to 195E in the sludge transfer direction. A part of the pretreatment filter cloth belt 182 is supported by support rollers 183a and 183b so as to form an inclined portion 182A inclined upward along the sludge transfer direction. Support roller 183a is connected to a drive shaft of electric motor 200. When the electric motor 200 rotates the support roller 183a, the pretreatment filter cloth belt 182 moves in the direction in which the sludge on the slope portion 182A rises.

The drive device 198 includes an electric motor 198A, a drive wheel 198B fixed to the drive shaft of the electric motor 198A, and a belt 198C as a driving force transmission member supported by the drive wheel 198B. A chain may be used instead of belt 198C. The belt 198C is connected to a first guide roller 188c and further connected to a second guide roller 191f. When the electric motor 198A rotates the drive wheel 198B, the rotation of the drive wheel 198B is transmitted to the first guide roller 188c and the second guide roller 191f by the belt 198C. The first guide roller 188c and the second guide roller 191f rotate at the same speed, so that the first filter cloth belt 186 and the second filter cloth belt 190 advance in the same direction and at the same speed.

The first filter cloth belt 186 is supported by first guide rollers 188a to 188e and a plurality of squeeze rollers 195A to 195E. The first guide roller 188d is connected to the first filter cloth tensioning device 201. The first filter cloth tensioning device 201 is configured to adjust the tension of the first filter cloth belt 186 by moving the position of the first guide roller 188d.

The second filter cloth belt 190 is supported by second guide rollers 191a to 191f and a plurality of pressing rollers 195A to 195E. The second guide roller 191a is connected to a second filter cloth tensioning device 202. This second filter cloth tensioning device 202 is configured to adjust the tension of the second filter cloth belt 190 by moving the position of the second guide roller 191a.

Sludge consisting of aggregates (pellets) formed by the flocculation device 103 is transferred onto the pretreatment filter cloth belt 182 through the sludge transfer pipe 118 . While the sludge ascends the inclined portion 182A of the pretreatment filter cloth belt 182, the liquid in the sludge falls due to gravity. This process is a gravity dewatering process.

By moving the pretreatment filter cloth belt 182, the sludge on the pretreatment filter cloth belt 182 is transferred onto the first filter cloth belt 186. The first filter cloth belt 186 and the second filter cloth belt 190 extending between the plurality of squeeze rollers 195A to 195E are opposed to each other. As the first filter cloth belt 186 is moved, sludge becomes trapped between the first filter cloth belt 186 and the second filter cloth belt 190. While the first filter cloth belt 186 and the second filter cloth belt 190 are wound around pressing rollers 195A to 195E, the sludge is sandwiched between the first filter cloth belt 186 and the second filter cloth belt 190 and is squeezed. . As a result, the sludge becomes a cake (turbid residue) with a low moisture content. The cake with a low moisture content is discharged from the discharge port 153. The cake is conveyed to the input port 25 of the drying device 5 by the input conveyor 1 shown in FIG. 1, FIG. 5, or FIG. 6.

A filtrate receiver 145 is arranged below the squeeze rollers 195A to 195E. The filtrate removed from the sludge is collected by a filtrate receiver 145. A drain 146 is connected to the filtrate receiver 145 , and the filtrate collected by the filtrate receiver 145 is discharged from the dehydrator 105 via the drain 146 .

A specific example of the status data of the dehydration system of this embodiment is as follows.
-Sludge temperature (measured by temperature sensor 122)
・Flow rate of sludge to granulation tank 180 (rotational speed of pump 112)
・Sludge properties (pH, alkalinity, TS, VTS, SS, VSS, amount of suspended solids)
・Physical properties of flocculant (cation degree, molecular weight)
- Flow rate of flocculant (operating speed of flocculant supply device 128)
- Stirring speed of the stirrer 108 (rotational speed of the stirring motor 110)
- Movement speed of the first filter cloth belt 186 and the second filter cloth belt 190 (rotational speed of the electric motor 198A)
・Positions of guide rollers 188d and 191a

The items of the status data of the dehydration system described above are examples, and the status data of the dehydration system may include only any of the above items.

The trained model includes at least one of past image data, past sound data, and past vibration data, state data of the dehydration system when the past data was generated, and water discharged from the drying device 5. It is constructed according to a machine learning algorithm using learning data that includes multiple combinations of optimal operating parameters of the drying system that can keep the moisture content of the dried material within the target range. Past image data, past sound data, past vibration data, and dehydration system status data are explanatory variables, and it is possible to keep the moisture content of the dried material discharged from the drying device 5 within the target range. The optimal operating parameters of the drying system are the objective variables. During operation of the drying system, the control system 10 shown in FIG. 1, FIG. 5, or FIG. The processing device 10b is caused to perform calculations for outputting the optimal operating parameters of the drying system from the learned model.

In one embodiment, the trained model includes at least one of historical image data, historical sound data, and historical vibration data, and state data of a drying system and a dehydration system at the time the historical data was generated. Machine learning algorithm It may be constructed according to During operation of the drying system, the control system 10 shown in FIG. 1, FIG. 5, or FIG. The current state data is input to the learned model, and the processing device 10b is caused to execute calculations for outputting the optimal operating parameters of the drying system from the learned model.

In one embodiment, the control system 10 is configured to determine whether the operation of the drying system is normal or abnormal, in addition to or in lieu of the trained model described above that outputs optimal operating parameters for the drying system. The trained model may be provided. If an abnormality occurs in the operating state of the drying system, the desired moisture content of the material to be dried may not be achieved, which may have a significant impact on subsequent processing. Although such operational abnormalities do not occur frequently, when an operational abnormality occurs in the drying system, it is important to correctly detect it.

Therefore, the present embodiment provides a drying system that can correctly detect operational abnormalities in the drying system using a trained model. In the following explanation, the above-mentioned trained model that outputs the optimal operating parameters of the drying system will be referred to as the first trained model, and the trained model that will determine whether the operation of the drying system is normal or abnormal will be referred to as the first trained model. 2 is called a trained model.

A second trained model is stored in the storage device 10a of the control system 10. In one example, the second trained model is comprised of a neural network. A program for constructing a second learned model according to a machine learning algorithm is stored in the storage device 10a. The processing device 10b constructs the second trained model by executing operations according to instructions included in the program. The details of the second trained model are the same as those of the first trained model, so a redundant explanation will be omitted.

The second learned model uses learning data including multiple combinations of past image data generated by the imaging device 9 and operation index values indicating that the operating state of the drying system is normal or abnormal. It is a model built according to machine learning algorithms. As shown in FIG. 1, the imaging device 9 is arranged to generate image data of the material to be dried by the drying device 5. In one example, the imaging device 9 is arranged above a discharge conveyor 64 for carrying out the dried material dried by the drying device 5, and is arranged to generate image data of the dried material on the discharge conveyor 64. It's okay.

The past image data generated by the imaging device 9 is an explanatory variable, and the operating index value indicating whether the operating state of the drying system is normal or abnormal is an objective variable. Whether the operating state of the drying system is normal or abnormal is determined by the operator. Examples of operation index values include a first value (e.g., 0) indicating that the operating state of the drying system is normal, and a second value (e.g., 1) indicating that the operating state of the drying system is abnormal. It will be done. Causes of the abnormal operating state of the drying system include unexpected fluctuations in the moisture content of the material to be dried, clogging of the material to be dried in the drying device 5, and the like.

In order to optimize the parameters (weights, etc.) of the second learned model, a large number of image data generated by the imaging device 9 and operation index values indicating whether the operating state of the drying system is normal or abnormal are combined. Learning data (also referred to as training data or teacher data) including combinations is prepared. The control system 10 uses this learning data to construct a second trained model using a machine learning algorithm. The parameters of the second trained model may include bias in addition to weights. The second learned model constructed in this way is stored in the storage device 10a.

During operation of the drying system, the control system 10 acquires current image data from the imaging device 9 at predetermined intervals and stores it in the storage device 10a. The control system 10 inputs image data to a second trained model, and uses a processing device 10b to perform calculations for outputting an operation index value indicating whether the operating state of the drying system is normal or abnormal from the second trained model. have it executed.

FIG. 10 is a graph showing temporal changes in the operating index value output from the second learned model during operation of the drying system. In FIG. 10, the vertical axis represents the driving index value, and the horizontal axis represents the driving time. The interval between outputs of the determination results by the second trained model is set as appropriate. Although FIG. 10 shows an example in units of seconds, the intervals may be set in units of minutes, hours, or days. In the first half of the graph in FIG. 10, the numerical value 0 continues, indicating that the operation is normal, but gradually the numerical value 1, indicating abnormal operation, is output. As shown in FIG. 10, while the same numerical value is continuously output from the second learned model, different numerical values may be output from time to time. This is because the second learned model could not correctly determine the operating state of the drying system.

FIG. 11 is a graph showing temporal changes in the moving average of the driving index values shown in FIG. In FIG. 10, the vertical axis represents the moving average of driving index values, and the horizontal axis represents driving time. The control system 10 calculates a moving average of the driving index values output from the second trained model. The time interval of the moving average is set as appropriate. Control system 10 compares the moving average of the operating index values to a threshold value and determines that an abnormality has occurred in the operation of the drying system when the moving average of the operating index values exceeds the threshold value. The control system 10 then issues an alarm indicating an abnormal operation of the drying system. Since the operator can know from the alarm that an abnormality has occurred in the operation of the drying system, the operator can eliminate the abnormality in operation by taking appropriate measures for the processing system.

As can be seen from FIG. 11, the moving average of the driving index values continues to fluctuate even after once exceeding the threshold value, so it may exceed or fall below the threshold value. Therefore, once the control system 10 determines that an abnormality has occurred in the operation of the drying system, it does not take any action to determine the abnormality in operation and issues an alarm unless the moving average of the operation index values falls to a predetermined normal level. Do not release. When the operational abnormality of the drying system is resolved, the moving average of the operational index values gradually decreases. When the moving average of the operating index values falls to a predetermined normal level, the control system 10 resets the operation abnormality determination operation of the drying system and restarts the operation abnormality determination operation. Furthermore, the control system 10 cancels the alarm.

According to this embodiment, the control system 10 can correctly determine that an abnormality has occurred in the drying system based on the operation index value output from the second learned model.

The control system 10 uses either or both of a first trained model that outputs optimal operating parameters of the drying system and a second trained model that determines whether the operation of the drying system is normal or abnormal. You may prepare.

The embodiments described above have been described to enable those skilled in the art to carry out the invention. Various modifications of the above embodiments can be naturally made by those skilled in the art, and the technical idea of the present invention can be applied to other embodiments. Therefore, the invention is not limited to the described embodiments, but is to be construed in the broadest scope according to the spirit defined by the claims.

1 Input conveyor 5 Drying device 7 Imaging device 8 Imaging device 9 Imaging device 10 Control system 10a Storage device 10b Processing device 14 Drying chamber 15 Housing 17 Stirring rotor 17A Shaft 17B Paddle 18 Electric motor 20 Gear 22 Torque transmission mechanism 25 Inlet port 26 Discharge port 30 Dry gas supply line 31 Dry gas discharge line 32 Temperature measuring device 33 Flow rate measuring device 35 Temperature measuring device 40 Temperature measuring device 41 Pressure measuring device 50 Steam supply line 51 Boiler 52 Steam pressure control valve 56 Weir 58 Vertical movement device 60 Discharge chamber 64 Discharge conveyor 71 Sound detector 72 Sound detector 73 Sound detector 81 Elastic member 82 Vibration detector 85 Elastic member 86 Vibration detector

Claims (4)

A drying system for drying an object to be dried, the drying system comprising:
a drying device for drying the object to be dried;
an imaging device that generates image data of the object to be dried;
Equipped with a control system that has a trained model built using machine learning algorithms,
The control system inputs a storage device storing the learned model and image data of the object to be dried into the learned model, and sets the moisture content of the object to be dried discharged from the drying device within a target range. a processing device that performs calculations for outputting optimal operating parameters of the drying system that can be accommodated within the learned model from the learned model;
The control system inputs image data of the object to be dried and state data of a dehydration system disposed upstream of the drying system into the learned model, and controls the control system of the object to be dried that is discharged from the drying device. A drying system configured to cause the processing device to perform a calculation for outputting, from the learned model, optimal operating parameters for the drying system that can keep the moisture content within a target range. A drying system for drying an object to be dried, the drying system comprising:
a drying device that transports and dries the material to be dried while stirring;
an imaging device that generates image data of the object to be dried;
Equipped with a control system that has a trained model built using machine learning algorithms,
The control system inputs a storage device storing the learned model and image data of the object to be dried into the learned model, and sets the moisture content of the object to be dried discharged from the drying device within a target range. a processing device that performs calculations for outputting optimal operating parameters of the drying system that can be accommodated within the learned model from the learned model;
The drying system, wherein the operating parameters include whether the pressure of steam supplied to a stirring rotor for stirring the material to be dried in the drying apparatus is over or under an appropriate range. A drying system for drying an object to be dried, the drying system comprising:
a drying device that transports and dries the material to be dried while stirring;
an imaging device that generates image data of the object to be dried;
Equipped with a control system that has a trained model built using machine learning algorithms,
The control system inputs a storage device storing the learned model and image data of the object to be dried into the learned model, and sets the moisture content of the object to be dried discharged from the drying device to a target range. a processing device that performs calculations for outputting optimal operating parameters of the drying system that can be accommodated within the learned model from the learned model;
A drying system, wherein the operating parameters include an excess or deficiency of a height of a weir for controlling an amount and residence time of the material to be dried in the drying apparatus with respect to an appropriate range. A drying system for drying an object to be dried, the drying system comprising:
a drying device that transports and dries the material to be dried while stirring;
an imaging device that generates image data of the object to be dried;
Equipped with a control system that has a trained model built using machine learning algorithms,
The control system inputs a storage device storing the learned model and image data of the object to be dried into the learned model, and sets the moisture content of the object to be dried discharged from the drying device within a target range. a processing device that performs calculations for outputting optimal operating parameters of the drying system that can be accommodated within the learned model from the learned model;
The drying system, wherein the operating parameters include an excess or deficiency of a rotational speed of a stirring rotor for stirring the material to be dried in the drying apparatus with respect to an appropriate range.

Images