Quality determination method of seaming cap and capping device

Abstract

To detect defective capping with high accuracy.SOLUTION: The present invention relates to a quality determination method of a seaming cap that performs quality determination of a non-defective cap or a defective cap, and to a capping device. A plurality of non-defective torque waveforms, which plot torque values detected at each prescribed time when a non-defective cap is seamed, and a plurality of defective torque waveforms, which plot torque values detected at each prescribed time when a defective cap is seamed, are prepared in advance. A histogram indicating an occurrence frequency for each torque value is created from the non-defective torque waveforms and the defective torque waveforms. An algorithm for quality determination is created by selecting the occurrence frequencies at a plurality of prescribed torque values from the histogram as feature quantities. When a new cap is seamed on a container, the occurrence frequencies at the plurality of torque values are calculated from the torque waveform, and the calculated occurrence frequencies are analyzed by the algorithm to determine the quality.SELECTED DRAWING: Figure 5

Description

Technical Field

[0001] The present invention relates to a method for determining the quality of a tightening cap and a capping device, and more particularly to a method for determining the quality of a tightening cap for detecting capping defects such as so-called bell caps and bell release caps, and a capping device.

Background Art

[0002] Conventionally, when attaching a cap to a container such as a PET bottle, the screw portion formed on the inner periphery of the cap is screwed and tightened against the mouth portion of the container having a screw portion formed on the outer periphery. As a capping device for tightening a cap on such a container, there is known one including a container holding means for holding the container, a capping head for gripping the cap, and a servo motor for rotating the capping head (Patent Documents 1 and 2). When tightening a cap on a container using such a capping device, a capping defect called a so-called bell cap may occur in which the screw portion of the cap rides up on the screw portion of the container and the cap is attached obliquely. To detect such a bell cap, Patent Document 1 photographs the container and the cap with a camera to detect the inclination of the cap, and Patent Document 2 measures the torque waveform acting on the servo motor during capping to detect the difference between normal and bell cap states.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] Here, in the process of generating the above brake cap, the screwed part that has been screwed up once may return to a normal screwed state and then be properly capped. Hereinafter, when such screwing up of the screwed part occurs but then returns to normal, it will be referred to as a brake release cap. When such a brake release cap occurs, since there is a risk that debris due to damage to the screwed part may occur when the screwed part is screwed up, it is desirable to handle such a brake release cap as a defective product. However, since the above brake release cap has the same appearance as a normal capping result, it is not possible to perform a pass / fail determination using a camera as in Patent Document 1 above, and even when measuring the torque waveform as in Patent Document 2, there is a problem that it is more difficult to determine the brake release cap than the brake cap. In view of such problems, the present invention provides a method for determining the quality of a tightening cap and a capping device that can more accurately detect defects in the tightening cap.

Means for Solving the Problems

[0005] That is, the method for determining the quality of a tightening cap according to the invention of claim 1 is a method for determining the quality of a tightening cap for determining whether a cap with a screwed part formed thereon is a good-quality cap tightened in a normal state or a defective cap tightened in an abnormal state when the cap with the screwed part is tightened on a container having a screwed part formed at its mouth part. In this method, In advance, a plurality of good-quality torque waveforms obtained by plotting torque values detected at predetermined time intervals when a good-quality cap is tightened and a plurality of defective torque waveforms obtained by plotting torque values detected at predetermined time intervals when a defective cap is tightened are prepared respectively, and a histogram showing the occurrence frequency for each torque value is created from the above good-quality torque waveform and defective torque waveform. The occurrence frequencies at a plurality of predetermined torque values are selected as feature quantities from the above histogram, and an algorithm for determining quality is created based on the plurality of feature quantities. When newly tightening a cap onto a container, the quality is determined by calculating the occurrence frequency of a plurality of torque values from a torque waveform obtained by plotting torque values detected at predetermined time intervals, and analyzing the calculated occurrence frequency according to the above algorithm. Moreover, the capping device according to the invention of claim 3 includes a container holding means for holding a container having a threaded portion formed at its mouth, a capping head for gripping a cap having a threaded portion, a servo motor for rotating the capping head, a torque detecting means for detecting the torque acting on the capping head at predetermined time intervals, a torque waveform creating means for creating a torque waveform by plotting the torque values detected by the torque detecting means, and a control means for controlling the servo motor. In the capping device for tightening the cap gripped by the capping head onto the container with a predetermined torque, the control means creates a histogram showing the occurrence frequency for each torque value from a plurality of good product torque waveforms obtained by plotting torque values detected at predetermined time intervals when a good product cap was tightened in advance, and a plurality of defective torque waveforms obtained by plotting torque values detected at predetermined time intervals when a defective cap was tightened in the past, selects the occurrence frequency at a predetermined plurality of torque values from the histogram as a feature quantity, and stores a good / bad determination unit storing an algorithm created based on the plurality of feature quantities. When newly tightening a cap onto a container, it includes a feature quantity calculation unit for calculating the occurrence frequency of the plurality of torque values from a torque waveform obtained by plotting torque values detected at predetermined time intervals. It is characterized in that the quality of the cap tightened onto the container is determined by analyzing the occurrence frequency calculated by the feature quantity calculation unit according to the algorithm of the good / bad determination unit.

Advantages of the Invention

[0006] According to the above invention, a histogram showing the occurrence frequency for each torque value is created from the good torque waveform detected when tightening a good cap and the bad torque waveform detected when tightening a bad cap, and the occurrence frequencies at a plurality of predetermined torque values are selected as feature quantities from the above histogram, and a pass / fail determination algorithm is created in advance based on at least the above plurality of feature quantities. When newly tightening the cap onto the container, the occurrence frequencies at the above plurality of torque values are calculated from the detected torque waveform, and by analyzing the calculated occurrence frequencies with the above algorithm, it becomes possible to highly accurately detect capping defects such as brake caps and brake release caps.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Embodiments for Carrying Out the Invention

[0008] Regarding the following illustrated embodiments, FIG. 1 shows a side view of a capping device 3 for tightening and attaching a cap 2 to a container 1. In the present embodiment, the container 1 is made of resin such as PET, and although not shown, a threaded portion is formed on the outer periphery of the mouth portion 1a formed at the upper part. Also, the cap 2 is also made of resin, and although not shown, a threaded portion is formed on the inner periphery. With such a configuration, when attaching (capping) the cap 2 to the container 1, the cap 2 is positioned above the mouth portion 1a of the container 1, and then the cap 2 is lowered while rotating the container 1 and the cap 2 relative to each other, so that the threaded portion of the container 1 and the threaded portion of the cap 2 are screwed together, and the cap 2 is tightened around the container 1 to perform capping.

[0009] The capping device 3 includes a container holding means 4 for holding the container 1, a capping head 5 for holding the cap 2, a servo motor 6 for driving the capping head 5, an encoder 7 connected to the rotating portion of the servo motor 6 for outputting a pulse signal, and an ammeter 8 for detecting the current value output from the servo motor 6, and the servo motor 6 is controlled by a control means 9. The container holding means 4 includes a table 4a on which the container 1 is placed and a gripper 4b provided on the table 4a for gripping the neck portion formed below the mouth portion 1a of the container 1, and the container holding means 4 is provided at equal intervals on the outer periphery of a rotating table (not shown). The capping head 5 includes a chuck 5a for holding the cap 2 and a rotating shaft 5b for transmitting the drive of the servo motor 6 to the chuck 5a, and the capping head 5 is rotationally driven by the servo motor 6. Further, the capping head 5 can be moved up and down by a lifting means (not shown). The servo motor 6 is controlled by the control means 9, and the cap 2 is tightened around the container 1 by rotating the capping head 5 at a predetermined torque and rotational speed.

[0010] Here, when the container 1 and the cap 2 are rotated relative to each other to tighten the cap 2 on the container 1, the threaded portion of the cap 2 may ride up on the threaded portion of the container 1. If capping is completed with the threaded portion riding up like this, a mounting defect called a so-called bellows cap occurs, in which the cap 2 is mounted obliquely with respect to the container 1. Regarding this brake cap, although the screw part rides up at the beginning of capping, the screwing state of the screw part returns to normal while the cap 2 is being rotated, and capping may be completed as it is. In this case, at first glance, it is indistinguishable from a properly capped good product. However, when the screw part rides up, there is a risk that a part of the container 1 or the cap 2 may be damaged and debris may fall. Therefore, it is desirable to handle as defective products even the brake release caps in which the screw part that has once ridden up returns to a normal screwing state and then is properly capped. Conventionally, it has been difficult to distinguish such brake release caps. However, by using the method for determining the quality of the tightening cap according to the present invention, it has become possible to determine even brake release caps with high accuracy.

[0011] FIG. 2 shows a configuration diagram of the control means 9, and includes a torque command unit 11 that controls the torque of the servo motor 6, a rotational speed command unit 12 that controls the rotational speed of the servo motor 6, a torque detection unit 13 that detects the torque value related to the chuck 5a, a torque waveform creation unit 14 that forms a torque waveform, a rotational speed waveform creation unit 15 that calculates the rotational speed and rotational angle of the servo motor 6 based on the pulse signal detected by the encoder 7 every predetermined time, plots the rotational speed in time series to create a waveform of the rotational speed, a feature amount calculation unit 16 that calculates a feature amount described later from the created torque waveform, and a pass / fail determination unit 17 that determines the pass / fail of the tightened cap based on the calculated feature amount. Torque and rotational speed are registered in the torque command unit 11 and the rotational speed command unit 12 for each combination of the container 1 and the cap 2. When capping is performed, the capping head 5 is rotated at the torque and rotational speed registered in these torque command unit 11 and rotational speed command unit 12. In the torque detection unit 13 described above, a relational expression between a current value and a torque value is registered in advance, and based on the current value detected by the ammeter 8 every predetermined time (for example, 25 msec), the torque value related to the chuck 5a is detected when capping is performed. The torque waveform creation unit 14 plots the torque values detected by the torque detection unit 13 in time series to form a torque waveform as shown in Fig. 3(a). The rotational speed waveform creation unit 15 calculates the rotational speed and rotational angle of the servo motor 6 based on the pulse signal detected by the encoder 7 every predetermined time, and further plots the rotational speed in time series to create a rotational speed waveform as shown in Fig. 3(b). Specifically, the torque waveform shown in Fig. 3(a) shows an example of the torque waveform generated by the torque waveform creation unit 14 from the torque detected by the torque detection unit 13. The vertical axis represents torque, and the horizontal axis represents the time required for capping. Also, the rotational speed waveform shown in Fig. 3(b) shows an example of the rotational speed waveform generated by the rotational speed waveform creation unit 15. The vertical axis represents rotational speed, and the horizontal axis represents the time required for capping.

[0012] Subsequently, the feature quantity calculation unit 16 and the pass / fail determination unit 17 will be described. The pass / fail determination unit 17 of the present embodiment is equipped with an SVM (Support Vector Machine), and by using the algorithm generated by the SVM, the pass / fail of the tightened cap is determined. Here, the SVM is a machine that obtains the maximum distance between discrimination boundaries based on the input feature quantities and outputs the result, and is known as a type of machine learning model. The feature quantity is a value that is likely to distinguish the data to be classified. In the present embodiment, it is a value for classifying good caps, brake caps, and brake release caps by the SVM.

[0013] Next, the generation procedure of the algorithm by the SVM will be described. First, the cap 2 for performing the pass / fail determination is tightened around the container 1 by the capping head. Prepare approximately 100 torque waveforms each created when tightening the cap determined to be a good-quality cap, the torque waveform created when tightening the cap determined to be a brake cap, and the torque waveform created when tightening the cap determined to be a brake release cap. FIG. 4 shows a display obtained by selecting and overlapping one of each of the above three types of torque waveforms, showing the case where capping is good (good product: solid line), the case where a brake cap occurs (dashed line), and the case where a brake release cap occurs (two-dot chain line). In this torque waveform, although a clear difference can be seen in the brake cap compared to the good product and the brake release cap, it can be understood that when comparing the good product and the brake release cap, the clear difference is difficult to distinguish. Note that what is shown in FIG. 4 is merely an example, and the time taken for tightening the three torque waveforms does not vary significantly. However, in reality, there are torque waveforms with short tightening times and long tightening times for good-quality caps, brake caps, and brake release caps, and there are also various shapes of torque waveforms such as torque waveforms that rise smoothly from the initial rise to the final torque and torque waveforms that rise rapidly.

[0014] Next, feature quantities to be input to the SVM of the pass / fail determination unit 17 are selected from the torque waveforms of the good-quality caps, brake caps, and brake release caps prepared in this way. In the present embodiment, a histogram showing the occurrence frequency of torque values as shown in FIG. 5 is created from the prepared torque waveforms, and feature quantities are selected from the histogram. The histogram shown in FIG. 5 shows an example of a histogram created from the torque waveforms of the prepared good-quality caps, brake caps, and brake release caps, and the occurrence frequency is obtained for each torque value magnitude from the torque waveforms. In this embodiment, it is divided into nine gradations for every 30 torques, that is, in the ranges of torque values of 0 to 30, 31 to 60, 61 to 90, 91 to 120, 121 to 150, 151 to 180, 181 to 210, 211 to 240, and 241 to 270, and the number of occurrences in the range of the torque value is calculated.

[0015] In FIG. 5, it can be understood that in the range of the torque value of 31 to 60, good products occur more significantly than the brake cap and the brake release cap. Similarly, in the range of the torque value of 91 to 120, it can be understood that the brake release cap occurs more significantly than the good product and the brake cap, and in the range of the torque value of 151 to 180, it can be understood that the brake cap occurs more significantly than the good product and the brake release cap. Therefore, in this embodiment, the number of occurrences in the range of the torque value of 31 to 60 is extracted as feature quantity 1, the number of occurrences in the range of the torque value of 91 to 120 is extracted as feature quantity 2, and the number of occurrences in the range of the torque value of 151 to 180 is extracted as feature quantity 3, and these feature quantities 1 to 3 are learned by the SVM. Specifically, a histogram as shown in FIG. 5 is created from all the prepared torque waveforms, and the feature quantities 1 to 3 obtained from the torque waveform of the good product cap, the feature quantities 1 to 3 obtained from the torque waveform of the brake cap, and the feature quantities 1 to 3 obtained from the brake release cap are input to the SVM for learning in a state where the determination results of the caps are linked, and an algorithm for determining the quality of the cap is generated.

[0016] On the other hand, the above-mentioned feature quantity calculation unit 15 calculates the feature quantities 1 to 3 based on the torque waveform detected when the capping device 3 newly winds the cap 2 onto the container 1. That is, when the cap 2 is wound onto the container 1, the torque acting on the capping head 5 is detected by the above-mentioned torque detection unit 13, and based on the detected torque value, the torque waveform creation unit 14 creates the torque waveform shown in FIG. 3(a). Then, the feature quantity calculation unit 15 creates a histogram showing the occurrence frequency of the torque values shown in Fig. 5 from the torque waveform, and from within the histogram, the number of occurrences (feature quantity 1) in the range of torque values 31 to 60, the number of occurrences (feature quantity 2) in the range of torque values 91 to 120, and the number of occurrences (feature quantity 3) in the range of torque values 151 to 180, which are the same as when the algorithm was created using the SVM of the pass / fail determination unit 17, are calculated.

[0017] Hereinafter, the operation of the capping device 3 having the above configuration will be described centering on the method for determining the quality of the screwing cap according to the present invention. First, when the container 1 is supplied to the container holding means 4 and the gripper 4b grips the neck of the container 1, the capping head 5 holding the cap 2 is positioned above it. Thereafter, the capping head 5 lowers the chuck 5a, and the servo motor 6 rotates the capping head 5 under the control of the control means 9, whereby the screw portion formed on the mouth portion 1a of the container 1 and the screw portion formed on the cap 2 are screwed together. The control means 9 rotates the servo motor 6 at a predetermined torque and rotational speed, whereby the cap 2 is tightened onto the container 1. During the tightening, the current value output from the servo motor is detected by the ammeter 8 at predetermined time intervals, and the detected current value is calculated by the torque detection unit 13 as a torque value corresponding to the current value. The torque waveform creation unit 14 plots the calculated torque values in time series to create a torque waveform as shown in Fig. 3(a). Also, the rotational speed waveform creation unit 15 calculates the rotational speed of the servo motor 6 from the pulse signal output by the encoder 7 at predetermined time intervals, and plots the calculated rotational speeds in time series to create a rotational speed waveform as shown in Fig. 3(b). In such a tightening operation of the cap 2, when the tightening is completed with the threaded portion of the cap 2 and the threaded portion of the container 1 properly screwed together, the cap 2 becomes a good-quality cap. When the tightening is completed with the threaded portion of the cap 2 riding on the threaded portion of the container 1, the cap 2 becomes a bell-mouth cap. When the threaded portions of the cap 2 and the container 1 ride on each other and then are properly screwed together to complete the tightening, the cap 2 becomes a bell-mouth release cap.

[0018] Next, the feature quantity calculation unit 16 provided in the control means 9 creates a histogram as shown in FIG. 5 based on the torque waveform created by the torque waveform creation unit 14, and extracts the occurrence frequencies in the three torque value ranges of the torque value range of 31 to 60, the torque value range of 91 to 120, and the torque value range of 151 to 180 as feature quantities 1 to 3.

[0019] When the feature quantity calculation unit 16 extracts the feature quantities 1 to 3 in this way, these feature quantities 1 to 3 are input to the SVM of the pass / fail determination unit 17. The SVM analyzes based on an algorithm generated in advance by learning the input feature quantities 1 to 3, determines whether the tightened cap 2 is any one of a good-quality cap, a bell-mouth cap, and a bell-mouth release cap, and performs a pass / fail determination of the tightened cap. Based on the determination result of the pass / fail determination unit 17, the control means 8 determines the container 1 with the cap 2 determined to be a bell-mouth cap or a bell-mouth release cap as a defective container, and outputs a reject command to reject means (not shown) arranged downstream.

[0020] Next, the second embodiment will be described. When the material of the cap 2 and the container 1 to be tightened or the shape of their threaded portions changes, the torque acting on the capping head 5 changes. Therefore, there may be cases where the pass / fail determination cannot be accurately made only by the feature quantities 1 to 3 selected from the histogram used in the first embodiment. Therefore, in this embodiment, in addition to the feature amounts 1 to 3 selected from the histogram used in the first embodiment, percentiles shown in FIG. 6 are created, and second feature amounts 1 to 3 are selected from the percentiles. Based on the feature amounts 1 to 3 and the second feature amounts 1 to 3, the SVM algorithm of the pass / fail determination unit 17 is generated. Also in the second embodiment, similar to the first embodiment, the cap 2 for performing pass / fail determination is tightened around the container 1 by the capping head 5. When the cap 2 determined to be a good cap is tightened, the torque waveform created at that time, the torque waveform created when the cap 2 determined to be a brake cap is tightened, and the torque waveform created when the cap 2 determined to be a brake release cap is tightened are each prepared in a quantity of about 100. Subsequently, percentiles as shown in FIG. 6 are created from each torque waveform. A percentile is a representation in which the torque waveforms obtained in FIG. 3 are arranged in ascending order of torque value, and the rank of the torque value is expressed as a percentage (percent display). Also in FIG. 6, the cases where capping is good (good product), the case where a brake cap occurs, and the case where a brake release cap occurs are shown superimposed. In FIG. 6, when comparing the torque values at 60 percent, 70 percent, and 80 percent, a large difference can be seen between the good product and the brake release cap, and between the brake release cap and the brake cap, respectively. Therefore, in the second embodiment, the torque value at 60 percent is selected as the second feature amount 1, the torque value at 70 percent is selected as the second feature amount 2, and the torque value at 80 percent is selected as the second feature amount 3, and these second feature amounts 1 to 3 are learned by the SVM of the pass / fail determination unit 17. In the pass / fail determination unit 17 of this embodiment, together with the feature amounts 1 to 3 selected from the histogram shown in FIG. 5, which is the same as in the first embodiment, the second feature amounts 1 to 3 selected from the percentiles shown in FIG. 6 are input to the SVM in a state where they are associated with the determination results of good product, brake cap, and brake release cap, respectively, for learning, and an algorithm for determining the pass / fail of the cap is generated.

[0021] Next, when the capping device 3 winds a new cap 2 around the container 1, when the torque waveform creating unit 14 creates a torque waveform based on the torque value detected by the torque detection unit 13, the feature amount calculation unit 15 of the present embodiment creates a histogram showing the occurrence frequency of the torque values shown in FIG. 5 and percentiles as shown in FIG. 6. Furthermore, the feature amount calculation unit 15 calculates the number of occurrences in the range of torque values from 31 to 60, the number of occurrences in the range of torque values from 91 to 120, and the number of occurrences in the range of torque values from 151 to 180 from the histogram, that is, feature amounts 1 to 3, and calculates the torque values at 60 percentiles, 70 percentiles, and 80 percentiles from the percentiles, that is, the second feature amounts 1 to 3.

[0022] Then, the pass / fail determination unit 17 inputs to the SVM of the pass / fail determination unit 17 based on the feature amounts 1 to 3 and the second feature amounts 1 to 3 calculated by the feature amount calculation unit 15. The SVM analyzes based on the algorithm that generated the input feature amounts 1 to 3 and the second feature amounts 1 to 3, determines whether the wound cap 2 is any of a non-defective cap, a brake cap, and a brake release cap, and performs a pass / fail determination of the wound cap.

[0023] Note that a first algorithm using only the histogram according to the first embodiment and a second algorithm using the histogram and percentiles according to the second embodiment may be registered in the SVM of the pass / fail determination unit 17 and selectively used. For example, when capping the first combination of cap and container, analysis may be performed using the first algorithm, and when capping the second combination of cap and container, analysis may be performed using the second algorithm.

[0024] Also, as the feature amounts input to the SVM, it is necessary to select the optimal ones according to the type of defective cap for which pass / fail is determined, the materials of the cap and container, and conditions such as the shape and dimensions of the threaded portion. For example, as defective caps, in addition to the brake caps and brake release caps determined in the above embodiment, it is also possible to determine other defective caps such as inner biting caps and band biting caps. The above inner biting cap occurs when the annular protrusion formed on the top surface of the cap is not inserted into the opening of the mouth of the container, and the band biting cap occurs when the ring-shaped band formed at the lower end of the cap bites into the screw portion of the container. To determine these defective caps, in the first embodiment, feature quantities were selected using a histogram, and in the second embodiment, feature quantities were selected using a histogram and percentiles. In addition to these, any value of the maximum value, minimum value, sum, or standard deviation of the torque waveform may be used for selecting feature quantities. Specifically, in addition to the number of occurrences in the torque range using a histogram, the sum or standard deviation of the torque waveform may be added as feature quantities. Also, in addition to the number of occurrences in the torque range using a histogram and the torque value at a predetermined percentage value using percentiles, the maximum value, minimum value, sum, and standard deviation values of the torque waveform may be selected as feature quantities. Also, a predetermined range of the created torque waveform, for example, the range from 300 degrees to 150 degrees back from the end of tightening, may be continuously differentiated, and the maximum value and minimum value may be obtained from the differentiated waveform and selected as feature quantities.

[0025] In the above embodiment, the pass / fail determination unit 17 generates an algorithm using SVM (Support Vector Machine), but other methods may be used, for example, it is possible to use random forest or the like.

Explanation of Reference Numerals

[0026] 1 Container 2 Cap 3 Capping device 7 Torque detection means 9 Control means 13 Torque detection unit 16 Feature quantity calculation unit 17 Pass / fail determination unit

Claims

1. A method for determining the quality of a cap is provided, comprising the steps of: tightening a cap having a threaded portion on a container having a threaded portion at its mouth; and determining whether the tightened cap is a good cap tightened in a normal state or a defective cap tightened in an abnormal state. A plurality of good torque waveforms are prepared in advance, in which torque values ​​detected at predetermined times when good caps are fastened are plotted, and a plurality of defective torque waveforms are prepared in advance, in which torque values ​​detected at predetermined times when defective caps are fastened are plotted, and a histogram showing the occurrence frequency of each torque value is created from the good torque waveforms and the defective torque waveforms. Selecting occurrence frequencies at a plurality of predetermined torque values ​​from the histogram as feature quantities, and creating an algorithm for determining pass / fail based on the plurality of feature quantities; This method for determining whether a cap is good or bad is characterized in that, when a new cap is tightened onto a container, the occurrence frequency at multiple torque values ​​is calculated from a torque waveform obtained by plotting the torque values ​​detected at specified times, and the calculated occurrence frequency is analyzed using the above algorithm to determine whether the cap is good or bad.

2. creating percentiles in which torque values ​​are arranged in ascending order from the non-defective torque waveform and the defective torque waveform, and selecting a torque value at a predetermined percentage value of the percentile as a second feature value; and creating an algorithm for determining whether the torque is good or bad based on the feature value selected from the histogram and the second feature value selected from the percentile; The method for determining whether a cap is good or bad, as described in claim 1, characterized in that when a new cap is tightened onto a container, the occurrence frequency at the multiple torque values ​​and the torque value at the specified percentage value are calculated from a torque waveform plotted of torque values ​​detected at specified time intervals, and the calculated occurrence frequency and torque value are analyzed using the algorithm to determine whether the cap is good or bad.

3. A capping device comprising: a container holding means for holding a container having a threaded portion formed at the mouth thereof; a capping head for gripping a cap having a threaded portion formed thereon; a servo motor for rotating the capping head; a torque detection means for detecting a torque acting on the capping head at predetermined time intervals; a torque waveform creation means for plotting torque values ​​detected by the torque detection means to create a torque waveform; and a control means for controlling the servo motor, wherein the capping device tightens the cap held by the capping head around the container with a predetermined torque, The control means creates a histogram showing the occurrence frequency of each torque value from a plurality of good torque waveforms in which torque values ​​detected at predetermined times when a good cap was previously tightened are plotted, and a plurality of defective torque waveforms in which torque values ​​detected at predetermined times when a defective cap was previously tightened are plotted, selects occurrence frequencies at a plurality of predetermined torque values ​​from the histogram as feature values, and stores an algorithm created based on the plurality of feature values; and a feature value calculation unit that calculates occurrence frequencies of the plurality of torque values ​​from a torque waveform obtained by plotting torque values ​​detected at predetermined time intervals when a new cap is fastened to a container, A capping device characterized in that the occurrence frequency calculated by the feature calculation unit is analyzed using an algorithm of the quality determination unit to determine whether a cap tightened around a container is quality or not.

4. creating percentiles in which torque values ​​are arranged in ascending order from the non-defective torque waveform and the defective torque waveform, and selecting a torque value at a predetermined percentage value of the percentile as a second feature amount; the algorithm of the quality determination unit is created based on the feature amount selected from the histogram and the second feature amount selected from the percentile; the feature calculation unit calculates a torque value at a predetermined percentage value in addition to an occurrence frequency at the plurality of torque values ​​from a torque waveform obtained by plotting torque values ​​detected at predetermined time intervals when a new cap is fastened to a container; The capping device described in claim 3, characterized in that the quality determination unit determines whether the cap tightened onto the container is quality or not by analyzing the occurrence frequency and torque value calculated by the feature calculation unit using the algorithm.

Images