Control method and program for electrostatic separation device, and electrostatic separation device

Abstract

A technique for determining sorting conditions to be used in an electrostatic separator according to the state of raw materials is disclosed. The process performed by the electrostatic separator includes a step (S310) of acquiring raw material state data from an output signal of a sensor, a step (S320) of using the acquired raw material state data to execute a predetermined analysis process for analyzing the raw material state and detecting changes in the raw material state, a step (S330) of determining whether one or more changes in the raw material state have occurred, a step (S340) of acquiring input data if the raw material state has changed (YES in step S330), a step (S350) of determining sorting conditions using a learning model, and a step (S370) of changing the sorting conditions if the sorting conditions are different (YES in step S360).

Description

[Technical Field]

[0001] This disclosure relates to electrostatic sorting, and more specifically to a technique for determining the conditions for electrostatic sorting. [Background technology]

[0002] Plastics discarded in the market often contain multiple types of plastics. To recycle plastics, it is necessary to separate each type of plastic to a high purity. For example, when considering the recycling of home appliances (air conditioners, refrigerators, etc.), plastics such as polypropylene (hereinafter referred to as "PP"), acrylonitrile-butadiene-styrene copolymer (hereinafter referred to as "ABS"), and polystyrene (hereinafter referred to as "PS") are mixed together, so it is necessary to separate each plastic to a high purity. One method of separating plastics to a high purity is a sorting method called electrostatic separation.

[0003] In electrostatic sorting, multiple types of plastic pieces are mixed together, causing each piece to become triboelectrically charged. After triboelectric charging, the charged state (polarity of the static electricity, amount of charge, etc.) and the force exerted by the electric field differ depending on the type of plastic. These properties are used to sort each plastic piece.

[0004] In electrostatic separation, there is a problem in that the purity and recovery rate change depending on the mixing ratio of the raw materials introduced (e.g., ABS:PS = 6:4), particle size, and the amount of charge of the plastic pieces after triboelectric charging (hereinafter collectively referred to as "raw material condition"). To address this problem, a technology has been disclosed that adjusts the sorting conditions of electrostatic separation (for example, the position of the partition plate, the magnitude of the voltage, etc.) based on the raw material condition (see, for example, Japanese Patent Application Publication No. 2008-238140 (Patent Document 1) and Japanese Patent Application Publication No. Hei 7-328482 (Patent Document 2)). [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2008-238140 [Patent Document 2] Japanese Patent Application Publication No. 7-328482 [Overview of the project] [Problems that the invention aims to solve]

[0006] The state of the raw materials fed into the electrostatic separator changes due to the switching of input lots. For example, if raw material A from one supplier is fed into the electrostatic separator for 30 minutes, and then raw material B from another supplier is fed into the same separator, the raw material state will remain the same for the first 30 minutes, but after 30 minutes, the raw material state will change. This is because the raw material state of raw material A and raw material B are different. For example, raw material A may contain more ABS than PS, while raw material B may contain more PS than ABS.

[0007] To optimize the recovery rate and / or purity of plastics in accordance with changes in the raw material state, it is necessary to detect changes in the raw material state and adjust the sorting conditions each time a change occurs according to the raw material state after detection. However, according to the technology disclosed in Japanese Patent Publication No. 2008-238140 and Japanese Patent Publication No. Hei 7-328482, the electrostatic sorting apparatus or its operator cannot know when to switch the sorting conditions. Furthermore, when the raw material state changes, some characteristic quantities change while others do not (for example, the mixing ratio may change, but the particle size may not), and such points must also be taken into consideration. Therefore, a technology for detecting changes in the raw material state is needed. In addition, a technology is needed to determine the optimal sorting conditions for the electrostatic sorting apparatus based on the raw material state after the detection of such changes.

[0008] This disclosure is made in view of the above-mentioned background, and one of its purposes is to provide a technology for detecting changes in the state of raw materials and determining sorting conditions according to the state of the raw materials after detection. [Means for solving the problem]

[0009] According to an embodiment, a method executed by a computer for controlling an electrostatic separation device is provided. This method includes steps of acquiring, in time series, raw material data including information on raw materials input into the electrostatic separation device; detecting the presence or absence of a change in the raw material state from the raw material data acquired in the time series; based on the stepwise change in the raw material state being detected, using the raw material data acquired after the change as input data for a learning model that predicts the separation result by the electrostatic separation device; determining the separation conditions for the raw materials by the electrostatic separation device based on the input data and the learning model; and controlling the electrostatic separation device with the determined separation conditions.

Effect of the Invention

[0010] According to one aspect, separation conditions can be determined according to the raw material state.

Brief Description of the Drawings

[0011] [Figure 1] It is a diagram showing the configuration of an electrostatic separation device 100 according to Embodiment 1. [Figure 2] It is a block diagram showing the hardware configuration of a computer 200 that functions as a control device 110 according to an embodiment. [Figure 3] It is a flowchart exemplifying a part of the process until the separation conditions are determined in the control device 110 according to an embodiment. [Figure 4] It is a diagram exemplifying the time series data acquired for three lots of raw materials. [Figure 5] It is a diagram for explaining an example of a method for detecting a change in the raw material state. [Figure 6] It is a diagram showing the input data and output data to a learning model according to an embodiment. [Figure 7] It is a diagram showing the relationship between the recovery rate and purity predicted using the learning model at each of the seven positions of the partition plate 108. [Figure 8]It is a diagram showing an outline of the configuration of an electrostatic separation device 800 according to Embodiment 2. [Figure 9] It is a flowchart showing a part of the processing executed by the CPU 1 of the computer 200 that realizes the control device 110 included in the electrostatic separation device 800.

Embodiments for Carrying Out the Invention

[0012] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description, the same parts are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed descriptions thereof will not be repeated.

[0013] Embodiment 1. FIG. 1 is a diagram showing the configuration of an electrostatic separation device 100 according to Embodiment 1. Referring to FIG. 1, the electrostatic separation device 100 includes a charging unit 101, a charging lamp 102, a vibrating feeder 103, electrodes 104 and 105, a DC power supply 106, a collection box 107, a partition plate 108, a sensor 109, and a control device 110.

[0014] The electrostatic separation device 100 electrostatically separates a plastic piece group 111 in which a plurality of types (for example, two types) of plastic pieces 111a and 111b having different charging characteristics as raw materials to be separated (hereinafter sometimes referred to as "pieces to be separated") into plastic pieces 111a and plastic pieces 111b. In the following example, it is described assuming that the plastic piece 111a is ABS and the plastic piece 111b is a PS piece.

[0015] The input section 101 supplies the pre-dried plastic pieces 111 to the charging cylinder 102. When the charging cylinder 102 rotates continuously at a predetermined rotational speed, the plastic pieces 111 are agitated. Inside the charging cylinder 102, multiple types of plastic pieces 111a and 111b mixed in the plastic pieces 111 become charged by friction with each other. Each of the charged plastic pieces 111a and 111b has a polarity (positive or negative) and charge amount according to the triboelectric charging series. In this example, the ABS plastic pieces 111a become positively charged, and the PS plastic pieces 111b become negatively charged. The charged plastic pieces 111a and 111b are supplied to the rear end of the upper surface of the vibrating feeder 103. The positively charged plastic pieces 111a and the negatively charged plastic pieces 111b attract each other by electrostatic force and pair up.

[0016] The vibrating feeder 103 performs an up-and-down pushing vibration, pushing the plastic pieces 111a and 111b forward while moving them up and down. Through this up-and-down pushing vibration, the vibrating feeder 103 disengages the pairing of the plastic pieces 111a and 111b and feeds them in the X direction in Figure 1. The plastic pieces 111a and 111b fed in the X direction fall from the tip of the vibrating feeder 103. The charging cylinder 102 and the vibrating feeder 103 constitute a charging section that charges each of the plastic pieces 111a and 111b before they fall.

[0017] Electrodes 104 and 105 are the first and second electrodes, respectively, and are formed in a flat plate shape. Electrodes 104 and 105 are arranged in the X direction in Figure 1, and are positioned opposite each other, straddling the path through which the plastic pieces 111a and 111b fall. A ground voltage GND is applied to electrode 104. A DC power supply 106 applies a preset DC voltage between electrodes 104 and 105, generating an electrostatic field between them. When the plastic pieces 111a and 111b, whose pairing has been broken by the vibrating feeder 103, fall between electrodes 104 and 105, each plastic piece falls while being attracted to one of the two electrodes 104 and 105 by an electrostatic force corresponding to its charge state (polarity, amount of charge). In other words, each plastic piece traces a parabolic trajectory corresponding to its charge state and falls to a different position. In this example, plastic piece 111a is positively charged and therefore falls towards electrode 104. On the other hand, plastic piece 111b is negatively charged and therefore falls towards electrode 105. Electrodes 104, 105 and DC power supply 106 constitute an electric field generating unit that applies an electrostatic field to each charged plastic piece, causing each piece to fall to a position corresponding to its charge state. A collection box 107 is located below electrodes 104 and 105 and collects the plastic pieces 111a and 111b that have fallen from the vibrating feeder 103 through the space between electrodes 104 and 105. The collection box 107 is formed in the shape of a rectangular parallelepiped, with an opening at its top. The opening of the collection box 107 is rectangular in shape, with its longest side oriented in the X direction in Figure 1. A partition plate 108 is positioned inside the collection box 107 parallel to the YZ plane in Figure 1 and is movable in the X direction in Figure 1.

[0018] Each plastic piece 111a, 111b that falls from the vibrating feeder 103 through the space between electrodes 104 and 105 is collected in one of two collection chambers 107a or 107b depending on its charge state. In this example, plastic piece 111a is positively charged and is collected in collection chamber 107a. On the other hand, plastic piece 111b is negatively charged and is collected in collection chamber 107b.

[0019] In Figure 1, one partition plate 108 is installed in the collection box 107, and two collection chambers 107a and 107b collect plastic pieces 111a and 111b, respectively. However, the number of partition plates is not limited to the number exemplified in Figure 1. Multiple partition plates can be installed depending on the number of types of plastic to be sorted.

[0020] Sensor 109 acquires data that affects the sorting results (recovery rate, purity, etc.) of the plastic pieces 111 fed from the input unit 101 to the vibrating feeder 103, including the mixing ratio, charge amount, particle size, and other data. When acquiring data on the mixing ratio, an infrared sensor is used as sensor 109. This is because an infrared sensor can identify the type of plastic by irradiating the plastic with infrared light and analyzing its spectrum. When acquiring data on the charge amount, a charge amount sensor using a through-type Faraday cage is used as sensor 109. When acquiring data on particle size, one method is to have a camera acquire an image of the plastic pieces 111 and calculate the particle size by image analysis. In this embodiment, three types of data—mixing ratio, charge amount, and particle size—are exemplified, but the data that affects the sorting results of the plastic are not limited to these exemplified data, and other data may be used. In addition, either one type of sensor 109 or multiple types of sensors 109 can be employed.

[0021] The control device 110 detects changes in the raw material state by performing predetermined analysis processing using data obtained from the sensor 109, and controls the sorting conditions according to the detected raw material state. In this embodiment, the sorting conditions are, for example, the magnitude of the voltage between electrodes 104 and 105, or the position of the partition plate 108, but the control device 110 may use other physical quantities as sorting conditions. Furthermore, the control device 110 may simultaneously control one or more sorting conditions based on the raw material data. For example, the control device 110 can simultaneously control the voltage and the position of the partition plate.

[0022] [Configuration of Computer 200] Now, with reference to Figure 2, the specific configuration of the control device 110 will be described. Figure 2 is a block diagram showing the hardware configuration of a computer 200 that functions as a control device 110 according to a certain embodiment.

[0023] Computer 200 includes, as its main components, a CPU 1 for executing programs, a mouse 2 and a keyboard 3 for receiving input instructions from the user of Computer 200, RAM 4 for volatilely storing data generated by the execution of programs by the CPU 1 or data input via the mouse 2 or keyboard 3, a storage device 5 for non-volatilely storing data, a communication interface (I / F) 7, and a monitor 8. Each component is connected to the others by a data bus.

[0024] Processing in computer 200 is achieved through the cooperation of each piece of hardware and software (computer programs) executed by CPU 1. Such software may be pre-stored in storage device 5. In other situations, the software may be stored on a CD-ROM (Compact Disc Read-Only Memory) or other recording media and distributed as a computer program product. Alternatively, the software may be provided as a downloadable application program by an information provider connected to the so-called Internet. Such software is read from the recording media by an optical disc drive (not shown) or other reading device, or downloaded via a communication interface 7, and then temporarily stored in storage device 5. The software is read from storage device 5 by CPU 1 and stored in RAM 4 in the form of an executable program. CPU 1 then executes the program.

[0025] The components of the computer 200 shown in Figure 2 are common. Therefore, one essential part of the technical concept disclosed herein is the software stored in the RAM 4, storage device 5, CD-ROM, or other recording media, or software downloadable via a network. This recording media may include non-volatile, computer-readable data recording media. Since the operation of each hardware component of the computer 200 is well known, a detailed explanation will not be repeated.

[0026] Furthermore, the recording medium is not limited to CD-ROMs, FDs (Flexible Disks), and hard disks, but may also include magnetic tape, optical discs (MO (Magnetic Optical Disc) / MD (Mini Disc) / DVD (Digital Versatile Disc)), IC (Integrated Circuit) cards (including memory cards), optical cards, mask ROMs, EPROMs (Electronically Programmable Read-Only Memory), EEPROMs (Electronically Erasable Programmable Read-Only Memory), flash ROMs, and other semiconductor memory media that permanently store programs.

[0027] The term "program" as used here includes not only programs that can be directly executed by the CPU, but also programs in source code format, compressed programs, encrypted programs, and so on.

[0028] In other aspects, the control device 110 may be implemented by dedicated hardware that performs the processing that the control device 110 performs, instead of a general-purpose computer. In this case, the hardware may be implemented by, for example, a single circuit, a complex circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination thereof.

[0029] [Determining the selection criteria] Next, with reference to Figure 3, the method from the control device 110 to determining the sorting conditions will be explained. Figure 3 is a flowchart illustrating some of the processes that the control device 110, according to a certain embodiment, performs until it determines the sorting conditions.

[0030] In step S310, the CPU 1 of the control device 110 acquires raw material state data from the output signal of the sensor 109. More specifically, the sensor 109 acquires one or more raw material state data (mixing ratio, charge amount, etc.) as time-series data. The frequency (sampling rate) at which raw material state data is acquired is not particularly limited, but in order to improve the detection accuracy of data changes described later, it is desirable to acquire raw material state data in short cycles, such as at one-second intervals. This frequency can be determined, for example, according to the relationship between the size of the raw material and the feed rate. If the particle size of the raw material is small and the feed rate of the raw material is fast, the frequency of acquiring the raw material state may be set to a high frequency (for example, 2 times / second rather than 1 time / second). Conversely, if the size of the raw material is large and the feed rate is slow, the frequency may be set to a low frequency (for example, 0.1 times / second rather than 1 time / second).

[0031] In step S320, CPU1 uses the acquired raw material state data to execute a predetermined analysis process for analyzing the raw material state and detects changes in the raw material state. More specifically, CPU1 analyzes the time-series data obtained in step S310 to determine whether or not there has been a change in the raw material state.

[0032] Here, we will explain the changes in time-series data with reference to Figure 4. Figure 4 is an example of time-series data acquired for three lots of raw materials. Figure 4 illustrates that three types of data—mixing ratio, charge amount, and particle size—are acquired by each sensor 109. The horizontal axis of Figure 4 is time (in minutes), and it is assumed that the raw material states from three different lots (Lot 1 to Lot 3) lasted for 30 minutes, 10 minutes, and 30 minutes, respectively. When switching from Lot 1 to Lot 2, it is assumed that the charge amount and particle size remain unchanged while the mixing ratio changes, and when switching from Lot 2 to Lot 3, it is assumed that the mixing ratio and particle size remain unchanged while the charge amount changes. Although each lot is clearly indicated in Figure 4, in reality it is not possible to know when the lot switches, so the control device 110 needs to determine whether or not a switch has occurred from the changes in the data.

[0033] Furthermore, even if the lot does not change, variations may occur in the raw material state data (measured values) within the same lot. This variation is due to errors that occur when a portion of multiple plastic pieces is selected. For example, if 100 plastic pieces are measured for 10,000 pieces of plastic that are fed in per minute, sampling error occurs in estimating a population of 10,000 pieces from a sample of 100 pieces. In addition, the measurement error of the sensor 109 itself can also be a factor in the variation. Therefore, the control device 110 needs to take such variations into consideration when detecting whether or not there is a change in the raw material state. Below, one method for detecting a change in the raw material state will be described, but this is merely one example, and any method that can detect a change in the raw material state is acceptable.

[0034] To detect changes in the raw material state, the control device 110 needs to detect step-like changes in each data point. Step-like changes include, for example, the change in mixing ratio when switching from the first lot to the second lot, as illustrated in Figure 4, and the change in charge amount when switching from the second lot to the third lot. As mentioned above, data within the same lot contains sampling errors and measurement errors relative to the true value. It is assumed that sampling errors and measurement errors follow the following normal distribution.

[0035]

number

[0036] Under this assumption, the control device 110 applies Hotelling's T to the time-series data. 2 The law is applied. Therefore, refer to Figure 5 for specific hoteling T 2 The method of applying the law will be explained. Figure 5 is a diagram illustrating an example of a method for detecting changes in the state of raw materials. Case (I) in Figure 5 assumes that no change in the state of raw materials occurred, while Case (II) assumes that a change in the state of raw materials occurred due to a lot change.

[0037] In the case of case (I), as initially enclosed in block 510, the control device 110 receives n1 consecutive data X at each time point t-n1, t-n1+1, ..., t. t-n1 ,X t-n1+1 ,···,X t A block is selected (hereinafter referred to as a group), and the sample mean m and sample standard deviation s are calculated using the data within that group. The value n1 can be set arbitrarily and can be set by, for example, the data sampling rate (in Figure 5, n1=10).

[0038] Next, the control device 110 processes the data X at time point t+1 following time point t at the end of the group. t+1 Regarding X t+1 Using the value of , the sample mean m, and the sample standard deviation s, the anomaly score a(X) is calculated as follows: t+1 )

[0039]

Number

[0040] Note that since the sample standard deviation s is constant for each sensor, it may be determined in advance. In the case of the Hotelling's T 2 method, in the case of univariate data, it is known that the degree of abnormality a approximately follows a chi-square distribution (χ 2 distribution) with 1 degree of freedom. In the χ 2 distribution, the larger the value, the smaller the detection probability. Therefore, the control device 110 can determine that the larger the degree of abnormality a, the greater the deviation of X t from the previous group, and can detect the occurrence of a change in the raw material state. t+1

[0041] Returning to FIG. 5, in the case of case (I), since the deviation between the data of the group surrounded by the block 510 and X t+1 is small, the degree of abnormality a (X t+1 ) becomes small. In the case of case (II), since the deviation between the data of the group surrounded by the block 520 and X t+1 is large, the degree of abnormality a (X t+1 ) becomes large. The control device 110 can detect the presence or absence of a change in the raw material state from this change in the degree of abnormality. The abnormality determination threshold value of the degree of abnormality a can be arbitrarily set according to the χ 2 distribution. Further, the control device 110 does not need to determine the presence or absence of a change only based on the value of the degree of abnormality a (X t+1 ), and can also determine the presence or absence of a change in consideration of other degrees of abnormality a (X t+2 ) or the degree of abnormality a (X t+3 ).

[0042] Referring to FIG. 3 again, in step S330, the CPU 1 determines whether there has been a change in one or more raw material states. If the CPU 1 determines that there has been a change in the raw material state (YES in step S330), it switches the control to step S340. Otherwise (NO in step S330), the CPU 1 returns the control to step S310.

[0043] In step S340, CPU1 acquires raw material state data as input data to be input to a pre-generated learning model. More specifically, CPU1 acquires data (input data) to be input to the learning model in order to determine sorting conditions based on the data after the change in the raw material state. The input data may be raw material state data for a single point immediately after the change, or the average value of several points of data after the change may be used as input data. In other situations, if multiple types of sensors are used, that is, if multiple types of data are acquired, CPU1 will acquire the input data for each.

[0044] Now, with reference to Figure 6, we will describe a learning model that uses the input data. Figure 6 is a diagram showing input data to a learning model according to a certain embodiment and output data from the learning model.

[0045] As shown in Figure 6, the learning model is a model for predicting sorting results from raw material state data and sorting condition data. The learning model is created based on learning data obtained in advance from the electrostatic sorting device (hereinafter, the three types of data are collectively referred to as "learning data"). The control device 110 may collect additional learning data during operation and update the learning model. A machine learning model can be considered as an example of a learning model; for example, a machine learning model that can predict sorting results with high accuracy using raw material state data and sorting condition data as input data is used. Here, a machine learning model is a model for predicting an appropriate output from input data, and more specifically, models such as multiple regression analysis, random forest, neural networks, and deep learning are candidates. In general, each model requires parameters to predict the result, and these parameters are, for example, partial regression coefficients in multiple regression analysis. The parameters need to be determined in advance based on the learning data. As the sorting results will involve multiple data points such as recovery rate and purity, if you are predicting the recovery rate, you should create a learning model for predicting the recovery rate in advance, and if you are predicting purity, you should create a learning model for predicting purity in advance.

[0046] Referring again to Figure 3, in step S350, CPU 1 inputs the data into the learning model to determine the optimal sorting conditions. More specifically, control device 110 uses the input data acquired in step S340 and the learning model to determine sorting conditions for optimizing the sorting results. The definition of optimization of sorting results varies depending on the material to be sorted and the customer's requirements. Below, we define optimization of sorting results as satisfying the plastic purity standard value set in advance by the customer and maximizing the recovery rate, and explain how to determine the sorting conditions.

[0047] As an example of determining sorting conditions, the method by which the control device 110 determines the position of the partition plate from a learned model will be described. The settable range of the partition plate position is predetermined in advance. The end of the vibrating feeder 103 of the electrostatic sorting device 100 is taken as the origin, and the rightward direction of the X-axis in Figure 1 is considered positive, and the leftward direction is considered negative. For example, seven patterns of position x can be set in 10mm increments, such as x = +40mm, +50mm, +60mm, +70mm, +80mm, +90mm, and +100mm.

[0048] Therefore, with reference to Figure 7, the predicted recovery rate and predicted purity at each position of the partition plate 108 will be explained. Figure 7 is a diagram showing the relationship between the recovery rate and purity predicted using the learning model at each of the seven patterns of positions of the partition plate 108.

[0049] In a given scenario, the control device 110 uses input data (raw material state data), the position of the partition plate 108, and a learning model to predict the recovery rate and purity for each of the seven possible positions of the partition plate 108 (No. 1 to 7). In this case, assuming the purity standard is set to 95%, the three patterns where the predicted purity is equal to or greater than the standard are No. 1 to No. 3. Of the three conditions, condition No. 3 yields the highest predicted recovery rate, so the control device 110 selects +60mm as the position of the partition plate 108.

[0050] Referring again to Figure 3, in step S360, CPU1 determines whether the sorting conditions currently in operation for electrostatic sorting are different from the sorting conditions determined in step S350. If CPU1 determines that these sorting conditions are different (YES in step S360), it switches control to step S370. Otherwise, i.e., if the conditions are the same (NO in step S360), CPU1 returns control to step S310.

[0051] In step S370, the CPU 1 changes the sorting conditions used for electrostatic sorting to the sorting conditions determined in step S350. The control device 110 feeds back the sorting conditions to the electrostatic sorting device 100 so that if the sorting conditions are different, the electrostatic sorting device 100 operates based on the newly determined sorting conditions. For example, when changing the position of the partition plate 108, the operator of the electrostatic sorting device 100 may change its position manually, or the control device 110 may send a positioning signal to an actuator (not shown) that changes the position of the partition plate 108, and the actuator may move the partition plate 108 according to the positioning signal.

[0052] As described above, the electrostatic separator 100 in this embodiment can detect the timing of changes in the raw material state through data collected by the sensor 109 and analysis processing performed by the control device 110. Furthermore, the electrostatic separator 100 can determine the sorting conditions according to the changed raw material state by using a learning model. As a result, the electrostatic separator 100 can optimize the recovery rate and purity even if the raw material state changes on a lot-by-lot basis.

[0053] In this embodiment, an example was described in which plastic piece 111a is ABS and plastic piece 111b is PS, but the scope of application of the disclosed technical concept is not limited to this example. For example, the group of plastic pieces 111 may consist of two or more pieces to be sorted, each having different electrostatic properties. In addition to ABS (plastic piece 111a) and PS (plastic piece 111b), other types of plastic pieces 111 may include, for example, PP, PE (polyethylene), PET (polyethylene terephthalate), and PVC (polyvinyl chloride).

[0054] Furthermore, while the above embodiment illustrates a case where the electrostatic separator 100 separates multiple types of plastic pieces 111, the objects to be separated are not limited to this. In other aspects, the electrostatic separator 100 can separate multiple types of pieces to be separated that have different charging characteristics. The pieces to be separated may be made of materials other than plastic, as long as they are electrically charged. Also, while the above embodiment shows an example where the electrostatic separator 100 separates two types of pieces to be separated, the electrostatic separator 100 according to other aspects can separate three or more types of pieces to be separated. In this case, the partition plates 108 can be arranged so that the collection box 107 has as many compartments as there are types of pieces to be separated to be collected.

[0055] Embodiment 2. The electrostatic separator 100 according to Embodiment 1 detects changes in the raw material state by measuring the raw material state, such as the mixing ratio and charge amount, using sensors and analyzing the obtained data. However, changes in the raw material state can also be detected by other methods.

[0056] Therefore, with reference to Figure 8, the electrostatic separation device 800 according to Embodiment 2 will be described. Figure 8 is a diagram showing an overview of the configuration of the electrostatic separation device 800 according to Embodiment 2. Components identical to those of the electrostatic separation device 100 according to Embodiment 1 are given the same numbers. The functions of these components are also the same. Therefore, the description of the same components will not be repeated.

[0057] The electrostatic sorting device 800 according to Embodiment 2 further comprises an input information management device 810 in addition to the configuration of the electrostatic sorting device 100 according to Embodiment 1.

[0058] The input information management device 810 maintains data (hereinafter also referred to as "input data") indicating whether or not a new lot of raw material has been fed from the input unit 101 into the charged cylinder 102. For example, while the same lot is being fed from the input unit 101 into the charged cylinder 102, the input information management device 810 outputs "0" as input data. On the other hand, when the lot is switched and a new lot of raw material is fed from the input unit 101 into the charged cylinder 102, the input information management device 810 outputs "1" as input data. Note that the format of the input data is not limited to 0 and 1 as exemplified above, and other formats may be used.

[0059] In this embodiment, a lot change is notified to the input information management device 810 by, for example, operation of a control screen (not shown) by an operator or other human input. In other cases, a lot change may be notified to the input information management device 810 automatically. For example, a sensor (not shown) provided in the input unit 101 may detect the input of a new raw material, and the detection signal may be transmitted to the input information management device 810 to notify it of a lot change. In this case, the sensor may be implemented as, for example, a barcode reader that reads a barcode attached to a container holding the raw material to identify the lot of the raw material, but other detection means may be used as the sensor.

[0060] Referring to Figure 9, the control structure of the control device 110 according to this embodiment will be described. Figure 9 is a flowchart showing a part of the processing performed by the CPU 1 of the computer 200 that implements the control device 110 of the electrostatic sorting device 800. Note that the same step numbers are assigned to the same processing as in Embodiment 1. Therefore, the explanation of the same processing may not be repeated.

[0061] In step S910, the CPU 1 acquires input data from the input information management device 810. In this embodiment, the input data is input to the control device 110 as time-series data. For example, while input of the same lot continues, the input information management device 810 outputs 0 as input data to the control device 110. At the moment the lot changes, the input information management device 810 outputs 1 as input data to the control device 110.

[0062] In step S920, CPU1 determines whether or not there has been a change in the raw material state (a change in lot). For example, if the input data received from the input information management device 810 is "0", CPU1 determines that the lot of the raw material being input remains the same. On the other hand, if the input data is "1", CPU1 determines that the lot has been switched. If CPU1 determines that there has been a change in the raw material state (YES in step S920), it switches control to step S930. Otherwise (NO in step S920), CPU1 returns control to step S910.

[0063] In step S930, CPU1 determines whether there is a time difference between the initial input and the electrostatic sorting. The electrostatic sorting device 800 switches the subsequent processing content depending on whether there is a time difference between the initial input and the electrostatic sorting. In most cases, plastic raw materials go through multiple processes from the time they are input until they are electrostatically sorted. These multiple processes include, for example, a crushing process to further crush the plastic pieces that are the raw materials, a specific gravity sorting process to sort the plastics by type based on their specific gravity, and a foreign matter removal process to remove foreign matter such as rubber. If these processes occur between the raw material input process and the electrostatic sorting process, a time difference will occur before the input raw materials are sent to the electrostatic sorting process. If such a time difference is expected, the average time required from the raw material input process to the electrostatic sorting input is determined in advance. If CPU1 determines that there is a time difference (YES in step S930), it switches the control to step S940. On the other hand, if CPU1 determines that there is no time difference, that is, if it determines that the first process the input raw material goes through is the electrostatic separation process (NO in step S930), CPU1 switches the control to step S340.

[0064] In step S940, CPU1 waits for the time difference before processing. CPU1 waits for the average time calculated in advance from the raw material input process to the electrostatic separation process. After waiting for this average time, the input raw material is sent to the electrostatic separation process. After that, CPU1 switches control to step S340.

[0065] In step S340, CPU1 acquires raw material state data to input into the learning model. If there is a time lag, CPU1 waits for the average time and then collects data on the raw material state using the signal output from sensor 109. If there is no time lag, CPU1 collects data on the raw material state at the time the change was detected in step S920.

[0066] In step S350, CPU1 inputs the collected data into the learning model to determine the optimal selection criteria.

[0067] In step S360, CPU1 determines whether the sorting conditions currently in operation for electrostatic sorting are different from the sorting conditions determined in step S350. If these sorting conditions are different (YES in step S360), CPU1 changes the sorting conditions used for electrostatic sorting to the sorting conditions determined in step S350 (step S370). On the other hand, if these sorting conditions are the same (NO in step S360), CPU1 continues to acquire input data without changing the sorting conditions (step S310).

[0068] As described above, the electrostatic sorting apparatus 800 according to Embodiment 2 includes an input information management device 810, which can reduce the processing load on the control device 110.

[0069] In other words, the electrostatic separation device 100 according to the above-described embodiment 1 requires predetermined data analysis processing each time data on the raw materials being input is acquired, which increases the load on the CPU 1 of the computer 200 that implements the electrostatic separation device 100.

[0070] In contrast, the control device 110 according to this second embodiment detects lot changes (changes in raw material state) using input data from the input information management device 810. As illustrated above, the input data is displayed using the smallest bits, such as 0 or 1, so the load on the CPU 1 of the computer 200 that constitutes the control device 110 is smaller than the load on the control device 110 according to the first embodiment.

[0071] Some of the technical features disclosed above can be summarized as follows: [Configuration 1] According to one embodiment, a method is provided that is performed by a computer controlling an electrostatic separator. This method includes the steps of: acquiring raw material data in a time series, including information about the raw materials to be fed into the electrostatic separator; detecting whether there has been a change in the state of the raw materials from the raw material data acquired in the time series; using the raw material data acquired after the change as input data for a learning model that predicts the sorting results by the electrostatic separator, based on the detection of a step-like change in the state of the raw materials; determining the sorting conditions for the raw materials by the electrostatic separator based on the input data and the learning model; and controlling the electrostatic separator with the determined sorting conditions.

[0072] [Configuration 2] A method performed by a computer controlling an electrostatic separator, according to another embodiment, includes the steps of: acquiring input lot data relating to lots of raw materials to be fed into the electrostatic separator; detecting a change in the raw material lot from the input lot data; using the raw material data acquired after the change as input data for a learning model that predicts the sorting result by the electrostatic separator, based on the detection of the change in the raw material lot; determining the sorting conditions for the raw materials by the electrostatic separator based on the input data and the learning model; and controlling the electrostatic separator with the determined sorting conditions.

[0073] [Configuration 3] A method following a certain phase further includes, in addition to Configuration 2, the step of reading a pre-measured time required from the time the raw material is introduced until it is sent to the electrostatic separation process. The step of using the raw material data as input data for the learning model includes using raw material data acquired after the pre-measured time has elapsed from the time the lot is switched as input data for the learning model.

[0074] [Configuration 4] In a certain case, in addition to any of the above configurations, the controlling step includes controlling the electrostatic sorting device with the determined sorting conditions based on the fact that the determined sorting conditions are different from the current sorting conditions used up to the determination of the sorting conditions.

[0075] [Configuration 5] A method following a certain phase further includes, in addition to any of the above configurations, the steps of: collecting raw material data including information on the raw materials to be fed into the electrostatic separator; sorting condition data representing sorting conditions that can be adjusted in the electrostatic separator; and sorting result data representing the sorting results of the raw materials sorted based on the adjusted sorting conditions; and determining the parameters of the learning model using the collected raw material data, sorting condition data, and sorting result data.

[0076] [Configuration 6] In a certain case, in addition to any of the above configurations, the raw material data includes one of the following: the mixing ratio of materials contained in the raw material fed into the electrostatic separator; data relating to the electrostatic properties of the raw material; and data relating to the size of the crushed pieces constituting the raw material.

[0077] [Configuration 7] In a certain aspect, in addition to any of the above configurations, the electrostatic separator is equipped with partition plates for distinguishing and collecting each of the multiple materials contained in the raw material. The sorting condition data includes data relating to the position of the partition plates, data relating to the voltage of the electrostatic separator, data relating to the recovery rate of the material to be sorted from the raw material, and data relating to the purity of the material to be sorted.

[0078] [Configuration 8] In a certain case, in addition to any of the above configurations, the learning model includes a machine learning model for predicting the sorting result from the raw material data and the sorting condition data.

[0079] [Configuration 9] In a certain case, in addition to any of the above configurations, the step of detecting the presence or absence of the change includes the step of analyzing the raw material data acquired in a time series and the step of determining the presence or absence of the change based on the results of the analysis.

[0080] [Configuration 10] In a certain case, in addition to any of the above configurations, the analysis of the raw material data is performed according to Hotelling's T 2 It will be carried out using the law.

[0081] [Configuration 11] According to another embodiment, a program is provided that causes a computer to perform the method described in any of the above.

[0082] [Configuration 12] In yet another embodiment, an electrostatic sorting apparatus is provided, comprising a memory storing a program that causes a computer to perform the method described in any of the above, and a processor that executes the program.

[0083] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the claims rather than the foregoing description, and all modifications within the meaning and scope of equivalents of the claims are intended. [Industrial applicability]

[0084] The disclosed technical concept is applicable to electrostatic sorting equipment. [Explanation of Symbols]

[0085] 1 CPU, 2 Mouse, 3 Keyboard, 4 RAM, 5 Storage device, 8 Monitor, 100, 800 Electrostatic sorting device, 101 Input section, 102 Charged cylinder, 103 Vibration feeder, 104, 105 Electrodes, 106 DC power supply, 107 Collection box, 107a, 107b Collection chamber, 108 Partition plate, 109 Sensor, 110 Control device, 111 Plastic piece group, 111a, 111b Plastic piece, 200 Computer, 510, 520 Block, 810 Input information management device.

Claims

1. A control method for controlling an electrostatic separation device using a computer, comprising: acquiring raw material data including information on raw materials input to the electrostatic separation device in a time series manner; detecting whether or not there is a change in the state of the raw material from the raw material data acquired in time series; a step of using the raw material data acquired after the step-like change in the raw material state as input data for a learning model that predicts the sorting result by the electrostatic sorting device based on the detection of the step-like change in the raw material state; determining a condition for sorting raw materials by the electrostatic sorting device based on the input data and the learning model; and controlling the electrostatic separation device under the determined separation conditions.

2. A control method for controlling an electrostatic separation device using a computer, comprising: acquiring input lot data relating to a lot of material being input to the electrostatic separation device; detecting a change in the lot of the raw material from the input lot data; A step of using the raw material data acquired after the change in the raw material lot as input data for a learning model that predicts the sorting result by the electrostatic sorting device based on the detection of the change in the raw material lot; determining a condition for sorting raw materials by the electrostatic sorting device based on the input data and the learning model; and controlling the electrostatic separation device under the determined separation conditions.

3. The method further includes a step of reading out a time measured in advance as a time required from when the raw material is input until when the raw material is sent to the electrostatic separation step; 3. The control method according to claim 2, wherein the step of using raw material data as input data for the learning model includes a step of using raw material data acquired after the previously measured time has elapsed since the lot was changed as input data for the learning model.

4. The control method according to any one of claims 1 to 3, wherein the controlling step includes a step of controlling the electrostatic sorting device under the determined sorting conditions based on the fact that the current sorting conditions used up until the determination of the sorting conditions are different from the determined sorting conditions.

5. A step of collecting raw material data including information on raw materials input to the electrostatic separation device, sorting condition data representing sorting conditions that can be adjusted by the electrostatic separation device, and sorting result data representing a sorting result of the raw materials sorted based on the adjusted sorting conditions; The control method according to any one of claims 1 to 3, further comprising a step of determining parameters of the learning model using the collected raw material data, the sorting condition data, and the sorting result data.

6. The raw material data is A mixing ratio of materials contained in the raw material input to the electrostatic separation device; Data regarding the electrostatic properties of the feedstock; and The control method according to claim 5, further comprising data relating to the size of the crushed pieces constituting the raw material.

7. the electrostatic separation device is provided with partition plates for separately collecting each of the plurality of materials contained in the raw material; The selection condition data is Data regarding the position of the partition plate; Data relating to the voltage of the electrostatic separation device; Data on the recovery rate of materials subject to sorting from raw materials; and data relating to the purity of the material to be sorted.

8. The control method according to claim 5 , wherein the learning model includes a machine learning model for predicting the sorting result from the raw material data and the sorting condition data.

9. The step of detecting the presence or absence of a change includes: Analyzing the raw material data acquired in time series; The control method according to claim 1 , further comprising: determining whether or not there is a change based on a result of the analysis.

10. 10. The control method according to claim 9, wherein the raw material data is analyzed using Hotelling's T2 method.

11. A program for causing a computer to execute the control method according to any one of claims 1 to 3.

12. A memory storing a program for causing a computer to execute the control method according to any one of claims 1 to 3; and a processor for executing the program.