Screw configuration estimation device, screw configuration estimation method, and computer program

Through computer programs and object detection learning models, the arrangement position and type of screw parts are automatically detected, which solves the omissions caused by manual detection of assembly errors in the dual-axis extruder, and improves the accuracy and efficiency of assembly.

CN118984768APending Publication Date: 2024-11-19THE JAPAN STEEL WORKS LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202380031730.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-03-31
Filing Date
2023-02-15
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

In a dual-shaft extruder, since the prior art relies on manual inspection of assembly errors in screw parts, there is a possibility of missing assembly errors, resulting in damage to the engaged rotating screw.

Method used

Using computer programs and object detection learning models, we use computer programs and object detection learning models to obtain the appearance data of screw parts, estimate the arrangement position and type of screw parts, and realize automatic assembly error detection.

Benefits of technology

Effectively estimate the arrangement position and type of screw parts, reduce the omissions of assembly errors, avoid screw damage, and improve assembly accuracy and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118984768A_ABST
    Figure CN118984768A_ABST
Patent Text Reader

Abstract

A screw configuration estimation device, which estimates the configuration of a screw for an extruder formed by combining a plurality of screw parts, comprises: an acquisition unit that acquires appearance data indicating the appearance of the plurality of screw parts or screws arranged; and a calculation unit that, on the basis of the acquired appearance data, executes calculation processing that estimates the arrangement position and type of each of the plurality of screw components constituting the screw.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a screw configuration estimating device, a screw configuration estimating method and a computer program. Background Art

[0002] The screw of the extruder is composed of a combination of various screw parts. The inspection of screw part assembly errors is performed visually. In the case of a twin-screw extruder, the inspection of screw part assembly errors is performed by the feel of the two screws combined and rotated.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Publication No. 2018-43419 Summary of the invention

[0006] However, since the confirmation is done manually, there is a possibility that assembly errors may be missed. In a twin-screw extruder, if a screw with an assembly error is installed in the device and the device is operated, there is a problem that the screw that is engaged and rotated is damaged.

[0007] An object of the present disclosure is to provide a screw configuration estimating device, a screw configuration estimating method, and a computer program capable of estimating the arrangement position and type of each of a plurality of screw parts constituting a screw used in an extruder.

[0008] According to one aspect of the present disclosure, a screw composition estimating device is a screw composition estimating device that estimates the composition of a screw for an extruder formed by combining a plurality of screw parts, and comprises: an acquisition unit that acquires appearance data representing the appearance of the plurality of arranged screw parts or the screw; and a calculation unit that performs calculation processing to estimate the arrangement position and type of each of the plurality of screw parts constituting the screw based on the acquired appearance data.

[0009] A screw configuration estimating method according to one aspect of the present disclosure is a screw configuration estimating method for estimating the configuration of a screw for an extruder formed by combining a plurality of screw parts, the screw configuration estimating method comprising: acquiring appearance data representing the appearance of the plurality of arranged screw parts or the screw; and estimating the arrangement position and type of each of the plurality of screw parts constituting the screw based on the acquired appearance data.

[0010] A computer program according to one aspect of the present disclosure is a computer program for causing a computer to execute a process of estimating the configuration of a screw for an extruder including a combination of a plurality of screw parts, the computer program causing the computer to execute the following processes: acquiring appearance data representing the appearance of the plurality of arranged screw parts or the screw; and estimating the arrangement position and type of each of the plurality of screw parts constituting the screw based on the acquired appearance data.

[0011] Effects of the Invention

[0012] According to the present disclosure, it is possible to estimate the arrangement position and type of each of a plurality of screw parts constituting a screw used in an extruder. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a block diagram showing a configuration example of the screw configuration estimating device according to the first embodiment.

[0014] Figure 2 It is a schematic diagram showing a configuration example of a screw.

[0015] Figure 3 : is a conceptual diagram showing an object detection learning model according to Embodiment 1.

[0016] Figure 4 This is a conceptual diagram showing a screw parts list.

[0017] Figure 5 This is a flowchart showing the screw configuration estimation method according to the first embodiment.

[0018] Figure 6 : is a schematic diagram showing the appearance data import screen.

[0019] Figure 7 is a conceptual diagram showing object detection results.

[0020] Figure 8 is a table showing object detection results.

[0021] Fig. 9 This is a conceptual diagram showing arrangement data of the screw configuration.

[0022] Fig.10 This is a flowchart showing the assembly error determination process according to the first embodiment.

[0023] Fig.11 This is a schematic diagram showing another example of the method of importing appearance data.

[0024] Fig.12 This is a flowchart showing the assembly error determination process according to the second embodiment.

[0025] Fig.13 This is a block diagram showing a configuration example of a screw configuration estimating device according to the third embodiment.

[0026] Fig.14 This is a conceptual diagram showing a category discrimination detection learning model according to the third embodiment.

[0027] Fig.15 This is a flowchart showing a screw configuration estimation method according to the third embodiment. DETAILED DESCRIPTION

[0028] The following is a description of specific examples of a screw configuration estimation device, a screw configuration estimation method, and a computer program according to an embodiment of the present invention with reference to the accompanying drawings. It should be noted that the present invention is not limited to these examples, but is illustrated by the claims, and is intended to include all changes within the meaning and scope equivalent to the claims. In addition, at least a portion of the following embodiments and modifications may be combined arbitrarily.

[0029] Hereinafter, the present invention will be described in detail based on the accompanying drawings showing embodiments of the present invention.

[0030] (Implementation Method 1)

[0031] Figure 1 1 is a block diagram showing a configuration example of a screw configuration estimating device 1 according to Embodiment 1. The screw configuration estimating device 1 according to Embodiment 1 is a computer that performs calculation processing to estimate the configuration of a screw 3 composed of a plurality of screw parts 3a used in an extruder A.

[0032] The extruder A includes: a cylinder having a heater; a screw 3 which is provided in the cylinder so as to be drivable in a rotation direction and which melts, plasticizes and kneads a raw material resin; and a rotary motor which rotates the screw 3 .

[0033] Figure 2 : is a schematic diagram showing a configuration example of the screw 3. The screw 3 is configured by combining and integrating a plurality of types of screw parts 3a to form one screw 3. For example, the screw 3 is configured by arranging and combining a helical screw-shaped forward screw element for conveying the raw material in the forward direction, a reverse screw element for conveying the raw material in the reverse direction, a kneading element for kneading the raw material, and the like in an order and position according to the characteristics of the raw material.

[0034] The screw configuration estimating device 1 includes a computing unit 11 , a storage unit 12 , a communication unit 13 , a display unit 14 , and an operating unit 15 as a hardware configuration.

[0035] The operation unit 11 is a processor having operation circuits such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), an ASIC (Application Specific Integrated Circuit), and an FPGA (Field-Programmable Gate Array), an internal storage device such as a ROM (Read Only Memory), a RAM (Random Access Memory), and an I / O terminal. The operation unit 11 may include one or more operation circuits such as a GPU (Graphics Processing Unit), a TPU (Tensor Processing Unit), or an AI chip (semiconductor for AI) dedicated to image processing related to object detection and image recognition. The operation unit 11 performs the estimation process of the screw structure by reading and executing the computer program (program product) 16 stored in the storage unit 12. Each functional unit of the screw structure estimation device 1 may be implemented in software, or partially or entirely in hardware.

[0036] The storage unit 12 is, for example, a memory such as a hard disk, an EEPROM (Electrically Erasable Programmable ROM), or a flash memory. The storage unit 12 stores various programs executed by the operation unit 11 and various data necessary for the processing of the operation unit 11. In the present embodiment, the storage unit 12 stores at least a computer program 16 executed by the operation unit 11. In addition, the storage unit 12 also stores an object detection learning model 17 and a screw component table 18. The details of the object detection learning model 17 and the screw component table 18 will be described later.

[0037] The computer program 16 may be written into the storage unit 12 at the manufacturing stage of the screw configuration estimation device 1, or may be distributed from another information processing device via a network. The screw configuration estimation device 1 acquires the computer program 16 through communication and writes it into the storage unit 12. The computer program 16 may be recorded in a readable manner on a recording medium 10 such as a semiconductor memory such as a flash memory, an optical disk, a magneto-optical disk, or a magnetic disk. The screw configuration estimation device 1 reads the computer program 16 and stores it in the storage unit 12.

[0038] The communication unit 13 is a communication circuit that transmits and receives data with the measuring device 2 by wire or wirelessly. The communication unit 13 may be a circuit that reads data from a memory of the measuring device 2 or a storage medium retrieved from the measuring device 2. The communication unit 13 functions as an acquisition unit that acquires appearance data representing the appearance of the plurality of arranged screw parts 3a or the screw 3.

[0039] The measuring device 2 is a device capable of obtaining appearance data representing the appearance of the arranged plurality of screw parts 3a or the screw 3. The communication device is, for example, a camera or a video camera that photographs the screw 3. The communication device may be a smartphone, a tablet terminal, a laptop PC (personal computer), or the like having a light receiving lens and a photographing element. The measuring device 2 having a photographing function sends image data showing the appearance of the screw 3 as appearance data to the screw configuration estimation device 1. The image data is data including pixel values ​​of a plurality of pixels arranged in the horizontal direction and the vertical direction.

[0040] There is no particular limitation on the method of photographing the long screw 3. The entire screw 3 may be photographed in one image, or images taken of one end and the other end of the screw 3 may be synthesized to form a single photographed image including the entire screw 3. In addition, the method may be configured to scan from one end of the screw 3 to the other end to photograph and synthesize a single photographed image including the entire screw 3. The photographed image of the screw 3 may also be a dynamic image.

[0041] The measuring device 2 may be a distance measuring sensor that measures the distance to a plurality of points on the surface of the screw 3 to obtain point cloud data. The distance measuring sensor includes an infrared sensor such as LiDAR. The infrared sensor includes a light emitting element that emits infrared rays and a light receiving element that receives infrared rays irradiated to and reflected from the screw 3. The light emitting element is, for example, an infrared laser such as a vertical cavity surface emitting laser (VCSEL: Vertical Cavity Surface Emitting LASER), and irradiates the screw 3 in a dot pattern arranged vertically and horizontally. The light receiving element is, for example, a CMOS image sensor. The infrared sensor calculates the distance to the screw 3 based on the round-trip time from irradiation toward the screw 3 to being reflected back. The measuring device 2 calculates the distance to each dot pattern, and sends the point cloud data as three-dimensional information of the screw 3 as appearance data to the screw composition estimation device 1. The point cloud data is, for example, data of a plurality of points on the surface of the screw 3 represented by three-dimensional coordinates.

[0042] The screw configuration estimation device 1 acquires image data or point cloud data as appearance data sent from the measurement device 2 via the communication unit 13 and stores it in the storage unit 12. The calculation unit 11 can convert the point cloud data into voxel data. In addition, the calculation unit 11 can convert the point cloud data or voxel data into two-dimensional image data.

[0043] In order to simplify the description, an example of estimating the screw configuration using two-dimensional image data will be described below. Hereinafter, a two-dimensional image of the screw 3 obtained by photographing or ranging is appropriately referred to as a screw image, and two-dimensional image data is referred to as screw image data.

[0044] The display unit 14 is a display device such as a liquid crystal display panel or an organic EL display panel, etc. The display unit 14 displays the estimation result of the screw structure according to the first embodiment and the like under the control of the calculation unit 11 .

[0045] The operation unit 15 is an input device such as a touch panel, a mouse, an operation button, or a keyboard that receives operations by a user who uses the screw configuration estimating device 1 .

[0046] It should be noted that the screw configuration estimation device 1 may be a multi-computer including a plurality of computers, or a virtual machine constructed virtually by software. In addition, part or all of the screw configuration estimation device 1 may be configured as a cloud server.

[0047] Figure 3 1 is a conceptual diagram showing an object detection learning model 17 according to Embodiment 1. The object detection learning model 17 includes, for example, a convolutional neural network (CNN) that has been learned by deep learning. The object detection learning model 17 has an input layer 17a to which screw image data is input, an intermediate layer 17b to extract feature quantities of the screw image data, and an output layer 17c to output an inference result related to the detected object. The object detection learning model 17 is, for example, a YOLO model.

[0048] Each layer of the object detection learning model 17 has multiple nodes. The nodes of each layer are connected by edges. Each layer has an activation function (response function) and the edge has a weight. The value output from the node of each layer is calculated based on the value of the node of the previous layer, the weight of the edge and the activation function of the layer. The weight of the edge can be changed by learning.

[0049] The input layer 17 a of the object detection learning model 17 has a plurality of nodes, receives input of screw image data, ie, pixel values ​​of individual pixels constituting an image of the surface of the screw 3 , and transfers the input pixel values ​​to the intermediate layer 17 b .

[0050] The middle layer 17b has multiple groups of convolutional layers (CONV layers), pooling layers and fully connected layers. The convolutional layer filters the values ​​output by the nodes of the previous layer to extract the feature map. The pooling layer reduces the feature map output by the convolutional layer to obtain a new feature map.

[0051] The output layer 17c has nodes that output the final inference results about the object detected from the screw image. The inference results include the center coordinate position and the vertical and horizontal dimensions of the bounding box surrounding the object, the object detection score showing the accuracy that the image surrounded by the bounding box is the image of the object, and the category score showing the accuracy that the object belongs to a specific category.

[0052] More specifically, the normalized screw image input to the input layer 17a is divided into multiple grids, and the position, size, object score and category score of the bounding box are calculated for each grid. The position of the bounding box in a grid is expressed as a relative position relative to the upper left vertex or center position of the grid, for example.

[0053] Assuming that the screw image is divided into S×S grids, the number of bounding boxes detected in each grid is B, and the number of object categories is C, the output obtained from the output layer 17c becomes, for example, a vector of dimension S×S×(B×5+C).

[0054] It should be noted that the output obtained from the output layer 17c includes multiple bounding boxes that overlap with each other, but through post-processing to remove the overlapping bounding boxes, such as NMS (Non-Maximum Suppression) processing, the position and size of the most likely bounding box that encloses each object of one or more objects contained in the screw image, the object detection score and category score are obtained.

[0055] The method for generating the object detection learning model 17 is described. First, training data including a plurality of screw image data and an annotation file of each screw image data is prepared. The annotation file is teacher data showing the accurate value assigned to the corresponding screw image. Specifically, the annotation file is data showing the center coordinate position, vertical and horizontal dimensions, and category of the bounding box surrounding the image of the screw part 3a included in the corresponding screw image. The category represents a group classified according to the type of the screw part 3a. The type of the screw part 3a is determined by the L / D, lead, lead angle, number of threads, etc. of the screw part 3a.

[0056] When the thread image of the training data is input into the CNN, the object detection learning model 17 can be generated by optimizing the weighting coefficients of the neural network using error back propagation method, error gradient descent method, etc., so that the error (the value of a predetermined loss function or error function) between the data output from the CNN and the data shown by the teacher data becomes smaller.

[0057] It should be noted that although YOLO is described as an example of the object detection learning model 17, R-CNN, Fast R-CNN, Faster R-CNN or other CNNs may be used to form the object detection learning model 17. Alternatively, an object detection learning model 17 using algorithms such as decision trees, random forests, and SVM (Support Vector Machine) may be used. In addition, the object detection learning model 17 may be composed of a combination of the above-mentioned multiple algorithms.

[0058] Figure 4 : is a conceptual diagram showing the screw part table 18. The screw part table 18 stores the category of the object, the label (type name) of the screw part 3a, L / D, the lead or angle (lead angle), and the number of threads in association with each other. L / D is the ratio between the length (L) and the diameter (D) in the longitudinal direction of the screw part 3a.

[0059] Figure 5 1 is a flowchart showing a method for estimating a screw structure according to Embodiment 1. The computing unit 11 of the screw structure estimation device 1 acquires appearance data from the measuring device 2 via the communication unit 13 (step 111). The appearance data is data indicating the appearance of a plurality of arranged screw parts 3a or the appearance of the screw 3. The computing unit 11 uses two-dimensional thread image data based on the appearance data.

[0060] Figure 6 1 is a schematic diagram showing an appearance data import screen 131. The computing unit 11 that has acquired the appearance data displays the appearance data import screen 31 showing the screw image. The appearance data import screen 131 includes an imported image display unit 131a and a screw configuration estimation start button 131b. The computing unit 11 displays the screw image on the imported image display unit 131a based on the acquired screw image data. When the user operates the screw configuration estimation start button 131b, the following screw configuration estimation process is started.

[0061] The calculation unit 11 inputs the screw image data to the object detection learning model 17 and outputs a feature map (step S112). The feature map includes information for specifying the position and type of the object included in the screw image.

[0062] The operation unit 11 determines the arrangement position and category of the plurality of screw parts 3a based on the feature map (step S113). Specifically, the coordinates of the center position of the bounding box around each of the plurality of screw parts 3a, the vertical and horizontal dimensions, the object detection score, the category score, etc. are obtained from the feature map. The operation unit 11 determines the position and size of the bounding box of each of the plurality of screw parts 3a, and the category based on the object detection score and the category score. For example, the category with the object detection score above the threshold and the highest category score is determined.

[0063] Figure 7 is a conceptual diagram showing the object detection results, Figure 8 is a table showing object detection results. Figure 7 In the figure, the rectangular box is a boundary box, and the numbers "1", "2", ... are numbers (No.) for identifying each of the multiple screw parts 3a determined. The center position of the boundary box is represented by the X coordinate and the Y coordinate. "FF" indicates the type of screw part 3a belonging to a specific category (positive threaded part).

[0064] Next, the calculation unit 11 specifies the type name, L / D, lead or angle (lead angle), and number of threads of each of the multiple screw parts 3a to generate arrangement data (step S114). Specifically, the calculation unit 11 uses the type of each of the multiple screw parts 3a specified as a keyword to refer to the screw part table 18 to clarify the L / D, lead or angle, and number of threads of each of the multiple screw parts 3a. Then, the calculation unit 11 stores the arrangement data of the screw configuration in the storage unit 12 (step S115).

[0065] Fig. 9 : is a conceptual diagram showing the arrangement data of the screw structure. The arrangement data is data that associates the arrangement order of the plurality of screw parts 3a constituting the screw 3, the D / L (length to diameter ratio) of the plurality of screw parts 3a, the lead or lead angle, and the number of threads. Fig. 9 In FIG. 1 , “No.” conceptually shows the arrangement order of the plurality of screw parts 3 a .

[0066] The calculation unit 11 executes a predetermined process based on the generated and stored arrangement data of the screw configuration (step S116 ).

[0067] For example, the calculation unit 11 can manage the screw parts 3a by registering the arrangement data of the screw configuration in a database, etc. For example, the information of the screw parts 3a of each screw 3 mounted on one or more extruders A can be managed.

[0068] Furthermore, the calculation unit 11 can perform behavior analysis of the extruder A and resin flow analysis based on the generated and stored arrangement data of the screws 3 .

[0069] Furthermore, the calculation unit 11 can determine an assembly error of the plurality of screw parts 3 a based on the arrangement data.

[0070] As an example of the predetermined process, the determination of assembly error of the screw component 3a in one screw 3 will be described. It is assumed that the storage unit 12 stores arrangement data of the standard screw 3. That is, the storage unit 12 stores arrangement data indicating the accurate arrangement order and types of the plurality of screw components 3a.

[0071] Fig.10 1 is a flowchart showing the assembly error determination process according to Embodiment 1. The operation unit 11 reads the arrangement data of the standard screw 3 stored in the storage unit 12 (step S131). The operation unit 11 compares the arrangement data of the screw 3 as the determination object obtained in the processing of steps S111 to S114 with the arrangement data of the standard screw 3 (step S132), and determines whether there is a difference (step S133). That is, the operation unit 11 compares the L / D, lead, angle, and number of threads of the screw parts 3a in the same arrangement order.

[0072] When it is determined that there is no difference (step S133: No), the operation unit 11 ends the assembly error determination process. When it is determined that there is a difference (step S133: Yes), the operation unit 11 notifies that there is an assembly error (step S134) and ends the determination process. For example, the operation unit 11 displays the fact that there is an assembly error of the screw part 3a on the display unit 14. The operation unit 11 preferably displays an image of the screw 3 and an image of the screw part 3a in which the difference is detected on the display unit 14. The operation unit 11 can be configured to display the screw part 3a at the assembly error position by changing the color of the screw part image, etc. In addition, the operation unit 11 can display the arrangement order of the screw parts 3a with assembly errors and the correct type of the screw parts 3a on the display unit 14.

[0073] In addition, the screw configuration estimation device 1 may include a speaker or a light emitting device, which notifies the user of the assembly error by sound or light when an assembly error of the screw part 3a is detected. The screw configuration estimation device 1 may be configured to send notification data for notifying the assembly error of the screw part 3a to the user's communication terminal or the management device of the twin-screw extruder A.

[0074] It should be noted that although an example of comparing arrangement data has been described, the arrangement positions and categories of multiple screw parts 3a constituting a standard screw 3 can also be compared with the arrangement positions and categories of multiple screw parts 3a constituting the screw 3 as the judgment object obtained in the processing of steps S111 to S113 to determine whether there is a difference.

[0075] According to the screw configuration estimating device 1 and the like according to the first embodiment configured as described above, the arrangement position and type of each of the plurality of screw components 3 a constituting the screw 3 used in the extruder A can be estimated.

[0076] In addition, the screw configuration estimation device 1 can generate configuration data that associates the configuration order, L / D, lead, angle, number of threads, etc. of the plurality of screw parts 3a constituting the screw 3. If the configuration data of the screw 3 is obtained, it can be helpful for the management of the screw parts 3a or the resin flow analysis inside the extruder A.

[0077] In addition, the thread structure estimation device 1 can determine whether there is an assembly error of the screw part 3a and notify the user of the assembly error. Specifically, by comparing the arrangement data of the accurate screw structure with the arrangement data of the screw 3 as the determination object, the assembly error of the screw part 3a can be determined. The assembly error of the screw part 3a can be determined without relying on manual work.

[0078] Furthermore, by using a machine learning model, the position and type of the screw component 3 a constituting the screw 3 can be estimated with higher accuracy.

[0079] It should be noted that, in the present first embodiment, an example of generating arrangement data of the screw 3 based on appearance data obtained by photographing or measuring the distance of the screw 3 has been described, but appearance data obtained by photographing or measuring the distance of a plurality of screw parts 3a arranged before assembly may also be used.

[0080] Fig.11 FIG. 1 is a schematic diagram showing another example of a method for importing appearance data. Fig.11 As shown, the screw image data obtained by photographing a plurality of screw parts 3a arranged at intervals is acquired. The calculation unit 11 inputs the screw image data into the object detection learning model 17 to detect each of the plurality of screw parts 3a and obtain the arrangement data of the screw 3.

[0081] In the case of a screw image obtained by photographing the assembled screw 3, there is a risk that the boundaries between the screw parts 3a cannot be accurately determined. By photographing the screw parts 3a arranged at intervals before assembly, more accurate arrangement data can be obtained.

[0082] In addition, a mark that enables image recognition may be provided on each of the plurality of screw parts 3a. The computing unit 11 can accurately distinguish each of the plurality of screw parts 3a based on the image of the mark included in the screw image. Different marks may be provided on the screw parts 3a according to the type of the screw parts 3a. The arrangement position and type of the screw parts 3a constituting the screw 3 can be detected with higher accuracy.

[0083] In addition, in the first embodiment, an example of processing two-dimensional screw image data is described, but the arrangement data can also be generated using three-dimensional data of the screw 3. The three-dimensional data of the screw 3 can be obtained based on images taken from multiple different positions and angles using a camera or the like. In addition, the three-dimensional data of the screw 3 can be obtained from point cloud data obtained by ranging. By using a learning model or matching of three-dimensional data, the calculation unit 11 can similarly detect assembly errors of the screw parts 3a.

[0084] (Implementation Method 2)

[0085] The screw configuration estimation device 1 according to the second embodiment is different from the first embodiment in that it determines the assembly error of the two screws 3 mounted on the twin-screw extruder. Since the other structures of the screw configuration estimation device 1 are the same as those of the screw configuration estimation device 1 according to the first embodiment, the same reference numerals are attached to the same parts, and detailed description is omitted.

[0086] The calculation unit 11 of the screw configuration estimation device 1 according to the second embodiment acquires appearance data of two screws 3 mounted on the twin-screw extruder, generates arrangement data of the first screw 3 and arrangement data of the second screw 3 , and stores them in the storage unit 12 .

[0087] Fig.12 2 is a flowchart showing the assembly error determination process according to Embodiment 2. The calculation unit 11 reads the arrangement data of the first screw 3 and the arrangement data of the second screw 3 stored in the storage unit 12 (step S231). Then, the calculation unit 11 compares the arrangement data of each screw 3 (step S232).

[0088] Then, the operation unit 11 determines whether there is a difference between the arrangement data of the first screw 3 and the second screw 3 (step S233). If it is determined that there is no difference (step S233: No), the operation unit 11 ends the assembly error determination process. If it is determined that there is a difference (step S233: Yes), the operation unit 11 notifies that there is an assembly error (step S234) and ends the determination process.

[0089] According to the screw configuration estimating device 1 of the second embodiment, it is possible to determine whether or not there is an assembly error by comparing the arrangement data of the two screws 3 mounted on the twin-screw extruder, and to notify if there is an assembly error.

[0090] It should be noted that, similarly to the first embodiment, the arrangement positions and types of the plurality of screw parts 3a constituting the first screw 3 can be compared with the arrangement positions and types of the plurality of screw parts 3a constituting the second screw 3. The calculation unit 11 can determine the difference between the configuration of the first screw 3 and the configuration of the second screw 3 by comparing the arrangement positions and types. Through this determination, an assembly error of the screw 3 can be detected.

[0091] (Implementation method 3)

[0092] The screw configuration estimation device 1 according to the third embodiment differs from the first embodiment in the arrangement order and the method for determining the type of the screw parts 3a. Since the remaining structure of the screw configuration estimation device 1 is the same as that of the screw configuration estimation device 1 according to the first embodiment, the same reference numerals are given to the same parts and detailed description is omitted.

[0093] Fig.13 1 is a block diagram showing a configuration example of the screw configuration estimation device 1 according to Embodiment 3. The storage unit 12 of the screw configuration estimation device 1 according to Embodiment 3 stores the type identification learning model 19 instead of the object detection learning model 17 .

[0094] Fig.14 1 is a conceptual diagram showing a type discrimination detection learning model according to Embodiment 3. The type discrimination detection learning model is an image discrimination model. The type discrimination learning model 19 includes, for example, a convolutional neural network (CNN) that has been learned by deep learning. The structure of the neural network itself is the same as that of the object detection learning model 17. The type discrimination learning model 19 has an input layer 19a, an intermediate layer 19b, and an output layer 19c. Part image data, which is data of an image portion of a screw part 3a, is input to the input layer 19a. The intermediate layer 19b extracts feature quantities of the part image data. The output layer 19c outputs accuracy data indicating the accuracy of the part image belonging to each of a plurality of categories.

[0095] A method for generating a category identification learning model 19 is described. First, training data including a plurality of part image data and teacher data showing the category to which each part image belongs is prepared. The category represents a group classified according to the type of the screw part 3a. Then, when the thread image of the training data is input into the CNN, the operation unit 11 is able to generate a category identification learning model 18 by optimizing the weighting coefficient of the neural network using an error back propagation method, an error gradient descent method, etc., so that the error (the value of a predetermined loss function or error function) between the data output from the CNN and the data shown in the teacher data becomes smaller.

[0096] Fig.15: is a flowchart showing a method for estimating a screw structure according to Embodiment 3. The operation unit 11 of the screw structure estimation device 1 acquires appearance data from the measuring device 2 via the communication unit 13 (step S311), and extracts part appearance data corresponding to the part image (step S312). As in Embodiment 1, the operation unit 11 is an operation unit that uses two-dimensional screw image data based on the appearance data. In Embodiment 3, it is desirable that the screw image data use image data obtained by photographing a plurality of screw parts 3a arranged at intervals. The part appearance data is part image data corresponding to the screw part 3a portion in the part image. The operation unit 11 extracts the part image by processing such as pattern matching.

[0097] At this stage, it is not necessary to determine the type of the screw part 3a, and it is sufficient to extract an image similar to the image of the screw part 3a. It should be noted that the operation unit 11 identifies the position of the part image by extracting the part image. That is, the operation unit 11 detects the arrangement position of multiple screw parts 3a.

[0098] Next, the operation unit 11 inputs the extracted part image to the type identification learning model 19 and outputs accuracy data belonging to each of the multiple categories (step S313). The operation unit 11 clarifies the categories of the multiple screw parts 3a based on the accuracy data output from the type identification learning model 19 (step S314).

[0099] Thereafter, similarly to the first embodiment, the arrangement data of the screws 3 is generated (step S315 ) and stored (step S316 ), and a predetermined process using the arrangement data is executed (step S316 ).

[0100] According to the screw configuration estimating device 1 of the third embodiment, similarly to the first embodiment, it is possible to estimate the arrangement position and type of each of the plurality of screw components 3 a constituting the screw 3 used in the extruder A.

[0101] It should be noted that although an example of extracting part appearance data through image processing such as pattern matching has been described, detection of part appearance images can also be performed through a machine learning model.

[0102] Furthermore, in the third embodiment, the three-dimensional data of the screw 3 may be used to perform arithmetic processing to generate the arrangement data of the screw 3 .

[0103] Furthermore, in the first to third embodiments, an example was described in which the arrangement positions and types of the plurality of screw parts 3 a constituting the screw 3 were estimated mainly using the learning model, but the estimation may be performed by rule-based image processing.

[0104] For example, the screw configuration estimation device 1 may be configured to estimate the arrangement position and type of the screw parts 3a by template matching processing. The storage unit 12 of the screw configuration estimation device 1 stores template images of multiple types of screw parts 3a in association with the types of the screw parts 3a. The calculation unit 11 determines the positions and types of each of the multiple screw parts 3a included in the screw image by template matching processing using the template image stored in the storage unit 12. The processing after determining the arrangement position and type of the multiple screw parts 3a is the same as that of the above-mentioned embodiment 1-3.

[0105] The screw configuration estimation device 1 can be configured to estimate the arrangement position and type of the screw parts 3a based on the feature quantity. The storage unit 12 of the screw configuration estimation device 1 stores the feature quantities of the plurality of types of screw parts 3a and the types of the screw parts 3a in a corresponding manner. As the feature quantity, various features such as edges, changes in shading, or positional relationships between feature points can be considered. The calculation unit 11 determines the position and type of each of the plurality of screw parts 3a included in the screw image by comparing with the feature quantity stored in the storage unit 12. The processing after clarifying the arrangement position and type of the plurality of screw parts 3a is the same as that of the above-mentioned embodiments 1-3.

[0106] Description of Reference Numerals

[0107] 1: Screw composition estimation device

[0108] 2: Measuring device

[0109] 3: Screw

[0110] 3a: Screw parts

[0111] 10: Recording medium

[0112] 11: Operation unit

[0113] 12: Storage

[0114] 13: Ministry of Communications

[0115] 14: Display unit

[0116] 15: Operation Department

[0117] 16: Computer Programs

[0118] 17: Object Detection Learning Model

[0119] 18: Screw parts list

[0120] 19: Category Discrimination Learning Model

[0121] 131: Appearance data import screen

[0122] 131a: Import image display unit

[0123] 131b: Screw constitutes the estimated start button

[0124] A: Extruder.

Claims

1. A screw rod configuration estimation device, It is characterized in that The structure of a screw for an extruder formed by combining a plurality of screw parts is estimated, and the screw structure estimation device comprises: an acquisition unit that acquires appearance data indicating an appearance of the plurality of arranged screw parts or the screw; and A calculation unit executes a calculation process for estimating an arrangement position and a type of each of the plurality of screw components constituting the screw based on the acquired appearance data.

2. The screw configuration estimating device according to claim 1, It is characterized in that The operation unit: Based on the estimation results regarding the arrangement positions and types of the plurality of screw parts, arrangement data is generated that associates the arrangement order of the plurality of screw parts constituting the screw, the ratio of the length and diameter of the screw parts, the lead or lead angle, and the number of threads.

3. The screw configuration estimating device according to claim 1 or 2, It is characterized in that The calculation unit determines whether the configuration of the screw to be determined is correct by comparing data on standard arrangement positions and types of the plurality of screw components with data on the arrangement positions and types of each of the plurality of screw components.

4. The screw configuration estimating device according to any one of claims 1 to 3, It is characterized in that The calculation unit determines whether there is an assembly error between the first screw and the second screw by comparing data related to the arrangement position and type of each of a plurality of screw parts constituting a first screw mounted on the twin-screw extruder with data related to the arrangement position and type of each of a plurality of screw parts constituting a second screw mounted on the twin-screw extruder.

5. The screw configuration estimating device according to any one of claims 1 to 4, It is characterized in that The operation unit inputs the appearance data acquired by the acquisition unit into the object detection learning model, and outputs data representing the arrangement position and type of each of the multiple screw parts, wherein the object detection learning model is learned to output data representing the arrangement position and type of each of the multiple screw parts when the appearance data representing the arranged multiple screw parts or the appearance of the screw is input.

6. The screw configuration estimating device according to any one of claims 1 to 4, It is characterized in that The calculation unit estimates data indicating the arrangement position and type of each of the plurality of screw components by performing rule-based image processing on the appearance data.

7. The screw configuration estimating device according to any one of claims 1 to 6, It is characterized in that the computing unit extracts component appearance data indicating the appearance of each screw component constituting the screw from the appearance data acquired by the acquiring unit, The computing unit inputs the extracted part appearance data into a type identification learning model to estimate the type of each of the plurality of screw parts, wherein the type identification learning model is learned to output data indicating the type of the screw part when the part appearance data indicating the appearance of the screw part is input.

8. The screw configuration estimating device according to any one of claims 1 to 7, It is characterized in that The appearance data includes image data obtained by photographing the plurality of screw parts or the screw, or point cloud data obtained by measuring distances to a plurality of points on the surface of the plurality of screw parts or the screw.

9. A method for estimating the composition of a screw, It is characterized in that The structure of a screw for an extruder formed by combining a plurality of screw parts is estimated, and the method for estimating the screw structure comprises: acquiring appearance data representing an appearance of the plurality of arranged screw parts or the screw; and The arrangement position and type of each of the plurality of screw components constituting the screw are estimated based on the acquired appearance data.

10. A computer program, It is characterized in that The computer program is used to cause a computer to execute a process of estimating a structure of a screw for an extruder formed by combining a plurality of screw parts, the computer program causing the computer to execute the following process: acquiring appearance data representing an appearance of the plurality of arranged screw parts or the screw; and The arrangement position and type of each of the plurality of screw components constituting the screw are estimated based on the acquired appearance data.