Gob supply device

Abstract

We provide a gob supply device that can predict gob quality defects. [Solution] The gob supply device 1 includes a cooling device 26 that sprays coolant onto the upper funnel 22, a gob timing sensor 34A that detects the gob timing, which is the timing of gobs G falling from each of the multiple passage members 25 onto the corresponding rough mold 5, and the gob length, which is the vertical length of the gob at that time, a glass temperature sensor 36 that detects the temperature of the molten glass in the fore hearth 2, and a gob monitoring device 37 that monitors the state of the gobs falling from the multiple passage members onto the corresponding rough mold. The gob monitoring device estimates the state of the gobs based on at least the gob timing, gob length, and the temperature of the molten glass in the fore hearth, and reports the state of the gobs.

Description

Technical Field

[0006] , , ,

[0005] , ,

[0001] The present invention relates to a gob feeder that supplies gobs from a forehearth to a mold.

Background Art

[0002] Patent Document 1 discloses a gob feeder that supplies gobs from a forehearth to a bottle-making machine mold. The gob feeder has optical observation means for observing gobs that fall by their own weight from an orifice provided at the bottom of the forehearth, obtains measurement data such as the volume, weight, surface shape, length, and thickness of the gobs, and determines whether a quality defect has occurred by comparing the measurement data with quality standard data. And when a quality defect has occurred, the gob feeder identifies the operating factor that causes it and controls the device for operating the operating factor.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the gob feeder according to Patent Document 1 has a problem that it cannot detect a quality defect until a quality defect actually occurs in the gob. Therefore, even if control is performed to repair the operating factor, there is a problem that the quality defect of the gob continues until the operating factor is actually repaired. Therefore, it is preferable to predict the quality defect of the gob and repair the operating factor before the quality defect occurs.

[0005] In view of the above background, an object of the present invention is to provide a gob feeder that can predict a quality defect of a gob. ​​​​​To solve the above problems, one aspect of the present invention is a gob supply device (1) comprising: a shear (21) that cuts molten glass (M) being pushed downward by a plunger (15) from an orifice (13) at the bottom of a spout (11) provided in a fore hearth (2) to generate gobs (G); a scoop (24) provided below the shear that changes the supply direction of the gobs falling from the shear; a plurality of passage members (25) selectively connected to the scoop and supplying the gobs to a plurality of corresponding molds (5); and a plurality of passage members that supply the gobs to the corresponding molds (5) from each of the plurality of passage members The system includes a gob timing sensor (34A) for detecting the gob timing, which is the timing at which the gob falls into the mold; a gob length sensor (34B) for detecting the gob length, which is the vertical length of the gob; a glass temperature sensor (36) for detecting the temperature of the molten glass in the fore hearth; and a gob monitoring device (37) for monitoring the state of the gob that falls from a plurality of passage members into the corresponding mold. The gob monitoring device estimates the state of the gob based on at least the gob timing, the gob length, and the temperature of the molten glass in the fore hearth, and reports the state of the gob.

[0007] According to this embodiment, a gob supply device can be provided that can predict quality defects in gobs. Based on notifications regarding the condition of the gobs from the gob monitoring device, operators can take appropriate action before any abnormalities occur in the gobs.

[0008] In the above embodiment, the gob monitoring device may generate the anomaly index of the gob using a machine learning model that outputs an anomaly index of the gob in response to inputs of the gob timing, the gob length, and the temperature of the molten glass, and may report the anomaly index.

[0009] In this embodiment, the condition of the gob is quantified by an abnormality index, allowing the worker to appropriately recognize the condition of the gob.

[0010] In the above embodiment, the gob monitoring device may have a scoop temperature sensor for detecting the temperature of the scoop, and may estimate the state of the gob based on at least the gob timing, the gob length, the scoop temperature, and the temperature of the molten glass in the fore hearth. Alternatively, the gob monitoring device may generate an anomaly index for the gob using a machine learning model that outputs an anomaly index for the gob in response to inputs of at least the gob timing, the gob length, the scoop temperature, and the temperature of the molten glass, and may report the anomaly index.

[0011] In the above embodiment, the gob monitoring device includes a funnel provided below the shear to guide the direction of the gob's fall, and a funnel temperature sensor for detecting the temperature of the funnel. The gob monitoring device may estimate the state of the gob based on at least the gob timing, the gob length, the temperature of the funnel, and the temperature of the molten glass in the fore hearth, and report the state of the gob. Alternatively, the gob monitoring device may generate an anomaly index for the gob using a machine learning model that outputs an anomaly index for the gob in response to at least the input of the gob timing, the gob length, the temperature of the funnel, and the temperature of the molten glass, and report the anomaly index. Furthermore, the funnel comprises an upper funnel provided below the shear and a lower funnel provided below the upper funnel, the funnel temperature sensor detects the temperature of the upper funnel or the lower funnel, and the gob monitoring device generates an anomaly index for the gob using a machine learning model that outputs an anomaly index for the gob in response to inputs of at least the gob timing, the gob length, the temperature of the upper funnel or the lower funnel, and the temperature of the molten glass, and may report the anomaly index.

[0012] In the above embodiment, the gob monitoring device may issue an alarm when the anomaly index is equal to or greater than the anomaly judgment value.

[0013] In this embodiment, the worker can take appropriate action before an abnormality occurs in the gob based on an alarm from the gob monitoring device. [Effects of the Invention]

[0014] According to the present invention, it is possible to provide a gob supply device that can predict gob quality defects. [Brief explanation of the drawing]

[0015] [Figure 1] Diagram of the Gob Supply Device [Figure 2] Block diagram of the Goblin Surveillance System [Figure 3] Graph showing the relationship between gob timing and anomaly index according to the first embodiment. [Figure 4] Graph showing the relationship between gob timing and anomaly index according to the second embodiment. [Figure 5] Graph showing the relationship between the temperature of the scoop and the anomaly index according to the second embodiment. [Figure 6] Graph showing the relationship between gob timing and anomaly index according to the third embodiment. [Figure 7] Graph showing the relationship between gob length and anomaly index according to the third embodiment. [Figure 8] Graph showing the relationship between the temperature of the molten glass in the fore hearth and the anomaly index according to the third embodiment. [Figure 9] Graph showing the relationship between the temperature of the upper funnel and the anomaly index according to the third embodiment. [Modes for carrying out the invention]

[0016] Hereinafter, the first embodiment of the present invention will be described. As shown in FIG. 1, the gob supply device 1 according to the first embodiment is provided on the downstream side of the forehearth 2 and supplies the gob G to the bottle making machine 3. The bottle making machine 3 is provided with a plurality of sections for forming gobs including sections 3A and 3B in parallel. The number of sections is not particularly limited as long as it is plural, and may be, for example, 8 sections, 10 sections, etc. Each of the sections 3A and 3B includes a rough mold 5 and a finishing mold.

[0017] The forehearth 2 is connected to the downstream side of the glass melting furnace 7 via the working chamber 8. The glass melting furnace 7 melts raw materials such as silica sand and glass cullets to generate molten glass M. The glass melting furnace 7 may be a combustion type or an electric heating type furnace. The working chamber 8 distributes the molten glass M supplied from the glass melting furnace 7 to the plurality of forehearths 2. The working chamber 8 is also called a distribution chamber. Each forehearth 2 defines a flow path extending substantially horizontally from the working chamber 8. A temperature control device9 is provided in the space above the liquid level of the molten glass M in the forehearth 2. The temperature control device9 is preferably installed on the ceiling or side surface of the forehearth 2. The temperature control device9 is preferably, for example, a burner using natural gas as fuel. The molten glass M is heated by the flame ejected from the temperature control device9.

[0018] A spout 11 is provided at the downstream end of the forehearth 2. The spout 11 has a cylindrical orifice 13 provided on its bottom wall 12, a tube 14 extending vertically above the orifice 13, and a plunger 15 slidably provided in the tube 14. The tube 14 is preferably coupled to the upper wall of the spout 11. When the plunger 15 moves toward the orifice 13, the molten glass M is pushed downward from the orifice 13.

[0019] The gob feeder 1 includes a shear 21, an upper funnel 22, a lower funnel 23, a scoop 24, a plurality of passage members 25, and a first cooling device 26. The plurality of passage members 25 are provided corresponding to each section 3A, 3B of the bottle making machine 3. Since the upper funnel 22 and the lower funnel 23 are not essential components, they may be omitted. Only the upper funnel​​​​​​​​​​​​​​​The scoop 24 is located below the shear 21 and changes the direction of supply of gobs G falling from the shear 21. More specifically, the scoop 24 is located below the lower funnel 23. The scoop 24 changes the direction of supply of gobs G. More specifically, the scoop 24 directs the direction of travel of gobs G falling through the lower funnel 23 toward one of the multiple passage members 25. The scoop 24 is formed in an arched groove shape and is supported so as to be rotatable about a vertical axis. The scoop 24 rotates under the driving force of an actuator. The upper end of the scoop 24 is located below the lower funnel 23, and the lower end of the scoop 24 rotates to connect with one of the multiple passage members 25.

[0024] Each passage member 25 has a linearly extending trough 25A and a deflector 25B connected to the downstream end of the trough 25A. The trough 25A and deflector 25B are groove-shaped troughs. Viewed from above, each trough 25A extends radially around the scoop 24. Each trough 25A slopes downward radially from the scoop 24 side. The deflector 25B curves downward from the lower end of the trough 25A. Below the lower end of each deflector 25B are provided a plurality of rough molds 5 of the bottle-making machine 3. Each rough mold 5 is a mold for forming parisons. An opening is provided at the top of each rough mold 5 for receiving gobs G. Each deflector 25B guides the gobs G toward the corresponding opening of the rough mold 5. The upstream end of each trough 25A is selectively connected to the scoop 24 and receives a supply of gobs G from the scoop 24. In this way, each passage member 25 supplies gobs G to the corresponding plurality of rough molds 5. The trough 25A and the deflector 25B may be formed integrally.

[0025] With the above configuration, the gob G cut by the shear 21 falls through the upper funnel 22 and lower funnel 23, is guided to the passage member 25 selected by the scoop 24, and is supplied to the corresponding rough mold 5.

[0026] The first cooling device 26 cools the scoop 24 and the multiple passage members 25. The first cooling device 26 sprays coolant onto the scoop 24, causing the coolant to flow down the scoop 24 and then the multiple passage members 25. The flowing coolant cools the scoop 24 and the multiple passage members 25. If the gob supply device 1 has an upper funnel 22 and a lower funnel 23, the first cooling device 26 may spray coolant onto the upper funnel 22, causing the coolant to flow down the upper funnel 22, then the lower funnel 23, the scoop 24, and then the multiple passage members 25. The flowing coolant cools the upper funnel 22, the lower funnel 23, the scoop 24, and the multiple passage members 25. If the upper funnel 22 is omitted, the first cooling device 26 may spray the coolant onto the lower funnel 23, causing the coolant to flow down from the lower funnel 23 to the scoop 24 and then to the multiple passage members 25. In other embodiments, the first cooling device 26 may be a coolant piping (water jacket) provided on the back of the scoop 24 through which the coolant circulates. The coolant may be water.

[0027] The shear 21 may be cooled by a second cooling device 33. The second cooling device 33 may cool the shear 21 by spraying a coolant onto the shear 21. In other embodiments, the first cooling device 26 and the second cooling device 33 may be common, and the second cooling device 33 may spray a coolant onto the shear 21. In this case, the coolant sprayed onto the shear 21 flows down from the shear 21 in the order of the upper funnel 22, the lower funnel 23, the scoop 24, and the multiple passage members 25.

[0028] As shown in Figure 2, the gob supply device 1 includes a gob timing sensor 34A, a gob length sensor 34B, a scoop temperature sensor 35, a molten glass temperature sensor 36, and a gob monitoring device 37. If the gob supply device 1 has an upper funnel 22, the gob supply device 1 may also have an upper funnel temperature sensor 38. In addition, the gob supply device 1 may have a plurality of passage member temperature sensors 39.

[0029] The gob timing sensor 34A detects the gob timing, which is the timing of gobs G falling from each of the multiple passage members 25 into the corresponding rough mold 5. The gob timing sensor 34A may be configured, for example, by a line sensor camera (line scan camera) that photographs the space between the lower ends of the multiple passage members 25, i.e., the lower ends of the deflectors 25B, and the corresponding rough mold 5. In this embodiment, the gob timing is expressed as an angle with the period for molding one bottle being one cycle (360 degrees). In each of the multiple parallel sections, one cycle (360 degrees) is the period from the timing of a gob entering the rough mold (gob-in) until the next time a gob enters the rough mold. The gob-in timings of each section are offset from each other. The gob timing is the timing when the lower end of the gob G passes a predetermined position from the lower end of the passage member 25, i.e., the lower end of the deflectors 25B. However, it is not limited to this, and an appropriate threshold may be set for each section, and it may be determined that there is a delay in the timing of when the gob G is fed into the rough mold if the threshold is exceeded. The Gob timing sensor 34A may use, for example, a Gob monitoring device such as International Publication WO2024 / 024163.

[0030] The gob length sensor 34B detects the gob length, which is the vertical length of the gob G. The gob length sensor 34B may include a camera equipped with, for example, a CMOS sensor, and a calculation unit that obtains the gob length from the image acquired by the camera. The gob length sensor 34B may acquire an image of the gob G between the shear 21 and the upper funnel 22. If the upper funnel 22 is omitted, the gob length sensor 34B may acquire an image of the gob G between the shear 21 and the lower funnel 23. In another embodiment, the gob length sensor 34B may acquire an image of the gob G between the lower ends of a plurality of passage members 25 and the corresponding rough mold 5. In yet another embodiment, the gob timing sensor 34A may also function as the gob length sensor 34B. In this case, the length of the gob G may be obtained based on the timing of the lower end passing and the timing of the upper end passing, which are obtained by the gob timing sensor 34A.

[0031] The scoop temperature sensor 35 detects the temperature of the scoop 24. The upper funnel temperature sensor 38 detects the temperature of the upper funnel 22. The upper funnel temperature sensor 38 may also detect the temperature of the lower funnel 23. Multiple passage member temperature sensors 39 measure the temperature of each passage member 25. The passage member temperature sensors 39 may measure the temperature of at least one of the trough 25A and the deflector 25B. In this embodiment, the passage member temperature sensor 39 measures the temperature of the trough 25A. The upper funnel temperature sensor 38, the scoop temperature sensor 35, and the multiple passage member temperature sensors 39 may be, for example, radiation thermometers or thermocouples.

[0032] The molten glass temperature sensor 36 detects the temperature of the molten glass M in the fore hearth 2. The molten glass temperature sensor 36 may be, for example, a protective tube type thermocouple. The molten glass temperature sensor 36 is positioned at a predetermined location within the fore hearth 2. The molten glass temperature sensor 36 may measure the temperature, for example, on the center line of the fore hearth 2, at a position 10 to 1500 cm upstream from the orifice 13, and at a depth of 5 to 40 cm from the liquid surface of the molten glass M. The position of the molten glass temperature sensor 36 may be appropriately changed according to the size of the fore hearth 2. In addition, the fore hearth 2 may include multiple regions in a plan view, and the molten glass temperature sensor 36 may measure the temperature of the molten glass M in each region. For example, the fore hearth 2 may include nine regions in a plan view: upstream central, upstream left, upstream right, midstream central, midstream left, midstream right, downstream central, downstream left, and downstream right. Multiple molten glass temperature sensors 36 may measure the temperature of the molten glass M at the center of each region in a plan view. The molten glass temperature sensors 36 may measure the temperature at a predetermined depth from the liquid surface of the molten glass M, or they may measure the temperature of the liquid surface of the molten glass M.

[0033] The gob monitoring device 37 monitors the state of gobs G falling from multiple passage members 25 onto corresponding rough molds 5. The gob monitoring device 37 is a computer composed of an MPU (microprocessor), ROM, RAM, etc. The gob monitoring device 37 performs state monitoring processing to detect the state of gobs G by executing calculation processing according to the program using the MPU. The gob monitoring device 37 may be configured as a single piece of hardware, or as a unit consisting of multiple pieces of hardware. Furthermore, at least a part of each functional part of the gob monitoring device 37 may be implemented by hardware such as LSIs, ASICs, FPGAs, etc., or by a combination of software and hardware. The program may be stored in a non-volatile storage device such as an HDD or flash memory, or it may be stored in a removable storage medium such as a DVD or CD-ROM and installed in the storage device of the gob monitoring device 37 when the storage medium is read by a reader. Alternatively, the program may be downloaded and installed in the storage device of the gob monitoring device 37 via a communication line such as the internet. The gob monitoring device 37 may be built, for example, by a cloud server. The Gob monitoring device 37, which is a computer, performs the status monitoring process described below based on its program.

[0034] The gob timing sensor 34A, gob length sensor 35B, scoop temperature sensor 35, upper funnel temperature sensor 38, molten glass temperature sensor 36, and multiple passage member temperature sensors 39 are connected to the gob monitoring device 37 via a wired or internet-based communication line. The gob monitoring device 37 acquires the gob timing of each section based on the signal from the gob timing sensor 34A. The gob monitoring device 37 acquires the gob length based on the signal from the gob length sensor 35B. The gob monitoring device 37 acquires the temperature of the scoop 24 based on the signal from the scoop temperature sensor 35. The gob monitoring device 37 acquires the temperature of the upper funnel 22 based on the signal from the upper funnel temperature sensor 38. The gob monitoring device 37 acquires the temperature of the molten glass M at a predetermined location in the fore hearth 2 based on the signal from the molten glass temperature sensor 36. The gob monitoring device 37 acquires the temperatures of the multiple passage members 25 based on the signals from the multiple passage member temperature sensors 39.

[0035] The gob monitoring device 37 is connected to an output device 41 via a wired or internet-based communication line. The output device 41 may be, for example, a display, speaker, warning light, etc. The output device 41 may also be a mobile device such as a smartphone or mobile phone. The mobile device may be carried by the worker.

[0036] The gob monitoring device 37 may be connected to a control device 42 via a wired or internet-based communication line. The control device 42 controls the actuators of the temperature control device 9, the first cooling device 26, the second cooling device 33, the plunger 15, the shear 21, and the scoop 24. The control device 42 is a computer composed of an MPU (microprocessor), ROM, RAM, etc. The control device 42 controls each device by executing calculations according to a program using the MPU. The gob monitoring device 37 and the control device 42 may be composed of a single computer. The control device 42 may be connected to an input device 44 for receiving operator input. The control device 42 may also be connected to an output device 41.

[0037] The gob monitoring device 37 estimates the state of the gob G based on at least the gob timing, gob length, the temperature of the scoop 24, and the temperature of the molten glass M in the fore hearth 2. In the first embodiment, the gob monitoring device 37 generates an anomaly index for the gob G using a first machine learning model that outputs an anomaly index for the gob G in response to inputs of at least the gob timing, gob length, and the temperature of the molten glass M in the fore hearth 2. The anomaly index is a value between 0 and 1, and a larger value indicates that an anomaly has occurred in the gob G. Here, an anomaly refers to a defect caused by the gob G occurring in the bottle formed by the bottle making machine 3. Defects include, for example, streaks formed on the surface of the bottle or molding defects in the bottle.

[0038] The first machine learning model may consist of known neural network models, deep learning models, etc. The training data for creating the first machine learning model includes at least gob timing, gob length, the temperature of the molten glass M in the fore hearth 2, and an anomaly index. This data may be acquired at predetermined time intervals during the operation of the gob supply device 1 and the bottle making machine 3. The anomaly index may be acquired by inspecting the bottles formed by the bottle making machine 3 at predetermined time intervals. In this case, the anomaly index may be set to 1 if there is a defect, and to 0 if there is no defect. Depending on the degree of the defect, different values ​​greater than 0 and less than or equal to 1 may be set for the anomaly index.

[0039] The gob monitoring device 37 inputs the gob timing, gob length, and temperature of the molten glass M in the fore hearth 2 to the first machine learning model at predetermined time intervals for each section 3A and 3B, and obtains an anomaly index as an output. In this way, the gob monitoring device 37 obtains an anomaly index corresponding to each section 3A and 3B at predetermined time intervals. The gob monitoring device 37 outputs the obtained anomaly index to the output device 41. The output device 41 notifies the worker by at least one of image, video, or audio. If the output device 41 is a mobile terminal, the notification may be made by chat or email.

[0040] The gob monitoring device 37 determines whether each of the abnormality indices corresponding to sections 3A and 3B is greater than or equal to an abnormality judgment value. The abnormality judgment value is preferably set to, for example, 0.2 to 0.7. When the abnormality indices are greater than or equal to the abnormality judgment value, the state of the gob G in the corresponding sections 3A and 3B deteriorates, and defects may occur. When the abnormality indices are less than the abnormality judgment value, the state of the gob G in the corresponding sections 3A and 3B is normal, and defects are unlikely to occur. In other embodiments, abnormality determination may be performed by comparing the rate of change or the derivative of the abnormality indices with the judgment value.

[0041] If the anomaly index corresponding to a specific section 3A among multiple sections exceeds the anomaly judgment value, the gob monitoring device 37 controls the output device 41 to issue an alarm. The output device 41 alerts the worker with at least one of an image, video, or sound. At this time, the output device 41 should announce the number of the specific section 3A in which the anomaly index exceeded the anomaly judgment value. The worker can recognize from the alarm output from the output device 41 that the condition of the gob G corresponding to the specific section 3A has deteriorated. This allows the worker to take action such as replacing the trough 25A, increasing the cooling effect of the first cooling device 26, or changing the temperature of the fore hearth 2, thereby preventing defects.

[0042] To verify the capabilities of the gob monitoring device 37, the abnormality index during the actual operation of the gob supply device 1 was checked. The gob supply device 1 was operated continuously for 70 hours, with measurements taken every minute. The start of operation was set to 0:00, and the end of operation to 70:00. During this operating period, in the first section 3A, defects caused by gobs G in the molded bottles increased from approximately 50 hours. After 52 hours and 30 minutes, the first cooling device 26 increased its cooling action in response to operator input. As a result, the temperature of the scoop 24 decreased from 52 hours and 30 minutes. The first cooling device 26 sprays coolant toward the upper funnel 22. The sprayed coolant falls from the upper funnel 22, cooling the lower funnel 23, the scoop 24, and multiple troughs 25A. The amount of coolant sprayed by the first cooling device 26 is controlled by the control device 42. The control device 42 determines the amount of coolant injected by the first cooling device 26 based on the difference between the temperature of the scoop 24 and the target temperature. The control device 42 increases the amount of coolant injected when the temperature of the scoop 24 is higher than the target temperature, and decreases the amount of coolant injected when the temperature of the scoop 24 is lower than the target temperature. After 52 hours and 30 minutes, the operator operated the input device 44 to lower the target temperature of the scoop 24 by 2°C. After 52 hours and 30 minutes, defects decreased.

[0043] Figure 3 is a graph showing the relationship between the gob timing corresponding to Section 1 3A and the anomaly index corresponding to Section 1 3A. As shown in Figure 3, the anomaly index rises from approximately 48 hours and is maintained at its maximum value from around 50 hours. Subsequently, the anomaly index decreased sharply due to the increased cooling effect of the first cooling device 26 at 52 hours and 30 minutes. From these results, it can be said that the anomaly index has a high correlation with the actual occurrence of defects. In other words, the state of gob G and the occurrence of defects can be predicted based on the anomaly index.

[0044] Gob timing, gob length, and the temperature of the molten glass M in fore hearth 2 did not individually show any significant changes related to the occurrence of defects. In other words, monitoring gob timing, gob length, and the temperature of the molten glass M in fore hearth 2 individually does not allow for the prediction of defect occurrence.

[0045] Based on the above, it was confirmed that using all of the following factors—gob timing, gob length, and the temperature of the molten glass M in fore hearth 2—rather than using each of them individually, is effective in estimating anomalies in gob G.

[0046] In the gob supply device 1 according to the second embodiment, the gob monitoring device 37 estimates the state of the gob G based on at least the gob timing, gob length, the temperature of the scoop 24, and the temperature of the molten glass M in the fore hearth 2. The gob monitoring device 37 may also generate an anomaly index for the gob G using a second machine learning model that outputs an anomaly index for the gob G in response to inputs of at least the gob timing, gob length, the temperature of the scoop 24, and the temperature of the molten glass M in the fore hearth 2.

[0047] The training data for creating the second machine learning model includes at least gob timing, gob length, temperature of scoop 24, temperature of molten glass M in fore hearth 2, and an anomaly index. These data are preferably acquired at predetermined time intervals during the operation of the gob supply device 1 and the bottle making machine 3. The anomaly index according to the second embodiment may be acquired by an inspection similar to that of the first embodiment.

[0048] To confirm the effectiveness of the second machine learning model, an anomaly index was examined for the actual operating results of the gob supply device 1 using a method similar to that of the first embodiment. The gob supply device 1 was operated continuously for 5.5 hours, with measurements taken every minute. The start of operation was defined as 0:00, and the end of operation as 5.5:00. During this operating period, in the first section 3A, defects caused by gobs G in the molded bottles increased from approximately 4 hours and 30 minutes. Subsequently, at 4 hours and 45 minutes, the first cooling device 26 increased its cooling action in response to operator input. As a result, the temperature of the scoop 24 decreased from 4 hours and 45 minutes. The first cooling device 26 sprays coolant toward the upper funnel 22. The sprayed coolant falls from the upper funnel 22, cooling the lower funnel 23, the scoop 24, and multiple troughs 25A. The amount of coolant sprayed by the first cooling device 26 is controlled by the control device 42. The control device 42 determines the amount of coolant injected by the first cooling device 26 based on the difference between the temperature of the scoop 24 and the target temperature. The control device 42 increases the amount of coolant injected when the temperature of the scoop 24 is higher than the target temperature, and decreases the amount of coolant injected when the temperature of the scoop 24 is lower than the target temperature. After 4 hours and 45 minutes, the operator operated the input device 44 to lower the target temperature of the scoop 24 by 2°C. After 4 hours and 45 minutes, defects decreased.

[0049] Figure 4 is a graph showing the relationship between the gob timing corresponding to Section 3A and the anomaly index corresponding to Section 3A. Figure 5 is a graph showing the relationship between the temperature of Scoop 24 and the anomaly index corresponding to Section 3A.

[0050] As shown in Figures 4 and 5, the anomaly index began to rise after approximately 2 hours, peaking around 4 hours and 30 minutes. Subsequently, the anomaly index decreased sharply after the cooling capacity of the first cooling device 26 increased at 4 hours and 45 minutes. These results suggest that the anomaly index has a high correlation with the actual occurrence of defects. In other words, the condition of gob G and the occurrence of defects can be predicted based on the anomaly index.

[0051] Gob timing, gob length, temperature of molten glass M in fore hearth 2, and temperature of scoop 24 did not individually show any significant changes related to defect occurrence. In other words, monitoring gob timing, gob length, temperature of molten glass M in fore hearth 2, and temperature of scoop 24 individually does not allow for prediction of defect occurrence.

[0052] Based on the above, it was confirmed that using all of the following factors—gob timing, gob length, temperature of the molten glass M in fore hearth 2, and temperature of scoop 24—rather than using each of them individually is effective in estimating anomalies in gob G.

[0053] In the gob supply device 1 according to the third embodiment, the gob monitoring device 37 may estimate the state of the gob G based on at least the gob timing, gob length, the temperatures of the funnels 22 and 23, and the temperature of the molten glass M of the fore hearth 2. Specifically, the gob monitoring device 37 may generate an anomaly index for the gob G using a third machine learning model that outputs an anomaly index for the gob G in response to at least the input of the gob timing, gob length, the temperatures of the funnels 22 and 23, and the temperature of the molten glass M of the fore hearth 2. Here, the temperatures of the funnels 22 and 23 are the temperatures of the upper funnel 22 or the lower funnel 23.

[0054] The training data for creating the third machine learning model includes at least gob timing, gob length, temperature of the upper funnel 22 or lower funnel 23, temperature of the molten glass M in the fore hearth 2, and an anomaly index. These data may be acquired at predetermined time intervals during the operation of the gob supply device 1 and the bottle making machine 3. The anomaly index according to the third embodiment may be acquired by an inspection similar to that of the first embodiment.

[0055] The gob monitoring device 37 inputs the gob timing, gob length, temperature of the upper funnel 22, and temperature of the molten glass M of the fore hearth 2 to the third machine learning model at predetermined time intervals for each section 3A and 3B, and obtains an anomaly index as an output. In this way, the gob monitoring device 37 obtains an anomaly index corresponding to each section 3A and 3B at predetermined time intervals. The gob monitoring device 37 then determines whether each of the anomaly indices corresponding to each section 3A and 3B is greater than or equal to the anomaly judgment value.

[0056] To confirm the effectiveness of the third machine learning model, we examined the anomaly index against the actual operating results of Gob supply device 1, similar to the first machine learning model.

[0057] Figure 6 is a graph showing the relationship between the gob timing corresponding to the first section 3A and the anomaly index corresponding to the first section 3A. Figure 7 is a graph showing the relationship between the gob length corresponding to the first section 3A and the anomaly index corresponding to the first section 3A. Figure 8 is a graph showing the relationship between the molten glass temperature of the fore hearth 2 and the anomaly index corresponding to the first section 3A. Figure 9 is a graph showing the relationship between the temperature of the upper funnel 22 and the anomaly index corresponding to the first section 3A.

[0058] As shown in Figures 6 to 9, the anomaly index began to rise after approximately 2 hours, peaking around 4 hours and 30 minutes. Subsequently, the anomaly index decreased sharply after the replacement of trough 25A, corresponding to section 1A, at 5 hours and 15 minutes. These results suggest that the anomaly index has a high correlation with the actual occurrence of defects. In other words, the condition of gob G and the occurrence of defects can be predicted based on the anomaly index.

[0059] Gob timing, gob length, temperature of molten glass M in fore hearth 2, and temperature of upper funnel 22 did not individually show significant changes related to the occurrence of defects. In other words, monitoring gob timing, gob length, temperature of molten glass M in fore hearth 2, and temperature of upper funnel 22 individually does not predict the occurrence of defects. The correlation coefficient between gob timing and anomaly index was approximately 0.016, the correlation coefficient between gob length and anomaly index was approximately -0.064, the correlation coefficient between temperature of molten glass M in fore hearth 2 and anomaly index was approximately -0.014, and the correlation coefficient between temperature of upper funnel 22 and anomaly index was approximately -0.061. These correlation coefficients also show that monitoring gob timing, gob length, temperature of molten glass M in fore hearth 2, and temperature of upper funnel 22 individually does not predict the occurrence of defects. Furthermore, the third machine learning model will function similarly even if it is created based on the temperature of the lower funnel 23 instead of the temperature of the upper funnel 22.

[0060] This concludes the description of specific embodiments, but the present invention is not limited to the above embodiments and can be broadly modified and implemented. For example, the input to the machine learning model may include, in addition to gob timing, gob length, and the temperature of the molten glass M in the fore hearth 2, the mass of the gob G, the temperature of each of the multiple troughs 25A, and the gob temperature. The gob temperature may be obtained by a radiation thermometer or estimated based on the temperature of the molten glass M in the fore hearth 2. The mass of the gob G may be obtained by the product of the volume of the gob G and the density of the gob G. The volume of the gob G may be obtained, for example, by photographing the gob G with a camera and processing the image. The density of the gob G may be obtained based on the temperature of the gob G.

[0061] The machine learning model may output an anomaly index based on inputs of at least gob timing, gob length, temperature of molten glass M, and temperatures of multiple troughs 25A. Alternatively, the machine learning model may output an anomaly index based on inputs of at least gob timing, gob length, temperature of upper funnel 22, temperature of molten glass M, and temperature of scoop 24.

[0062] Furthermore, the temperature of the molten glass M in the forehearth 2, used as input to the machine learning model, may include the temperatures of the molten glass M at multiple measurement points within the forehearth 2. For example, the temperatures of the molten glass M at multiple measurement points along the center line of the forehearth 2 may be used. The measurement points should be arranged at equal intervals along the center line. By measuring the temperature of the molten glass M at multiple measurement points in the forehearth 2, the state of the molten glass M can be recognized more accurately.

[0063] The input to the machine learning model may include gob timing, gob length, the temperature of the molten glass M in the fore hearth 2, and the temperature of the scoop 24, as well as the ceiling temperature of the work chamber 8 located upstream of the fore hearth 2. The ceiling temperature of the work chamber 8 is preferably measured at multiple points. The ceiling temperature of the work chamber 8 is closely related to the temperature of the molten glass M passing through the work chamber 8. Therefore, by using the ceiling temperature of the work chamber 8 as input to the machine learning model, the state of the molten glass M can be recognized more accurately.

[0064] The input to the machine learning model may include, in addition to gob timing, gob length, temperature of molten glass M in the fore hearth 2, and temperature of scoop 24, at least one of the set point variable (SV) and process variable (PV) of the temperature control device 9 in the fore hearth 2. If the temperature control device 9 is a burner, the set point and process variable may be the fuel flow rate. Multiple temperature control devices 9 may be provided in the fore hearth 2. The set point and process variable of the temperature control device 9 are closely related to the temperature of molten glass M in the fore hearth 2. Therefore, by using the set point and process variable of the temperature control device 9 as input to the machine learning model, the state of the molten glass M can be recognized more accurately. [Explanation of symbols]

[0065] 1: Goblin Supply Device 2: Forehaas 3: Bottle making machine 11: Spout 13: Orifice 15: Plunger 21: Shear 22: Upper Funnel 23: Lower Funnel 24: Scoop 25: Trough 26: 1st cooling device 34A: Gob timing sensor 34B: Gob length sensor 35: Scoop temperature sensor 36: Molten glass temperature sensor 37: Goblin Surveillance Device 38: Upper funnel temperature sensor 39: Passageway component temperature sensor G: Gob M: Molten glass

Claims

1. It is a goblin supply device, A shear cuts the molten glass, which is pushed downward by a plunger through an orifice located at the bottom of the spout in the fore hearth, to create a gob, A scoop provided below the shear, which changes the supply direction of the gob falling from the shear, Multiple passage members selectively connected to the scoop and supplying the gob to a corresponding number of molds, A gob timing sensor detects the gob timing, which is the timing of the gob falling from each of the multiple passage members into the corresponding mold, A gob length sensor detects the gob length, which is the vertical length of the gob, A glass temperature sensor for detecting the temperature of the molten glass in the fore hearth, The system includes a gob monitoring device that monitors the state of the gobs that fall from multiple passage members into the corresponding molds, The gob monitoring device generates an anomaly index for the gob using a machine learning model that outputs an anomaly index for the gob in response to inputs of the gob timing, the gob length, and the temperature of the molten glass in the fore hearth, and the gob supply device reports the anomaly index.

2. It has a scoop temperature sensor that detects the temperature of the scoop, The gob supply device according to claim 1, wherein the machine learning model outputs the anomaly index in response to at least the input of the gob timing, the gob length, the scoop temperature, and the molten glass temperature.

3. A funnel provided below the shear and guiding the direction of the gob's descent, It includes a funnel temperature sensor that detects the temperature of the funnel, The gob supply device according to claim 1, wherein the machine learning model outputs the anomaly index in response to at least the input of the gob timing, the gob length, the funnel temperature, and the molten glass temperature.

4. The funnel comprises an upper funnel provided below the shear and a lower funnel provided below the upper funnel. The funnel temperature sensor detects the temperature of the upper funnel or the lower funnel, The gob supply device according to claim 3, wherein the machine learning model outputs the anomaly index in response to at least the input of the gob timing, the gob length, the temperature of the upper funnel or the lower funnel, and the temperature of the molten glass.

5. The gob supply device according to any one of claims 1 to 4, wherein the gob monitoring device gives an alarm when the abnormality index is equal to or greater than the abnormality judgment value.

Images