Chain elongation detection device

Abstract

A chain elongation detection device. An embodiment relates to a chain elongation detection device capable of detecting partial elongation of a chain. A chain elongation detection device that can reliably detect partial elongation of a chain is provided. The chain elongation detection device includes: a measurement unit (15) that continuously measures the total elongation amount of a plurality of chain links within a specified section of a chain (10) meshing with a pair of sprockets (12, 13); and an estimation unit (22) that estimates the elongation amount of each chain link by inputting the total elongation amount measured by the measurement unit (15) into a system that has been trained using a large number of data sets combined with elongation amount data of each chain link in one turn of the chain and data corresponding to the total elongation amount measured by the measurement unit (15).

Description

[0001] Cross-reference to Related Applications

[0002] This application is based on Japanese Patent Application No. 2021-118280 filed on Jul. 16, 2021, and claims the benefit of priority therefrom. This application incorporates the entire contents of that application by reference thereto. Technical Field

[0003] Embodiments of the present invention relate to a chain elongation detection device capable of detecting partial elongation of a chain. Background Art

[0004] Generally, in transportation devices such as escalators and automatic roads, there are provided steps for passengers and objects to ride on and a moving handrail for passengers to hold, which move in a cycle synchronously with a ring-shaped chain rotated and driven by a driving device. As is well known, a chain is formed by arranging shaft portions composed of pins, sleeves, and rollers at a prescribed pitch and connecting between these shaft portions alternately using inner links and outer links.

[0005] A chain having such a structure undergoes elongation over time due to wear of the above-mentioned shaft portions. If this elongation becomes large, the chain cannot mesh well with a sprocket, and sometimes a chain skipping phenomenon where the chain passes over the teeth of the sprocket may occur. In order not to cause such a problem, a technique for measuring the elongation of a chain and monitoring it has been proposed (for example, refer to Patent Document 1).

[0006] The technique for measuring the elongation of a chain disclosed in Patent Document 1 provides optical sensors at two positions separated from each other along the traveling direction of the chain whose elongation amount is to be measured, and connects their signals to a microcomputer for measurement. The distance between the two optical sensors is set to N times the standard pitch of the chain. Each optical sensor is arranged such that when each shaft portion of the chain passes through the optical axis, the optical axis is blocked. When the chain is not elongated, the rollers pass through the optical axes of the two optical sensors simultaneously, so sensor signals are emitted simultaneously. However, if the chain elongates due to long-term use, the emission time of the sensor signals will have a time difference proportional to its elongation amount. This time difference is proportional to the elongation amount of the chain, and the total elongation amount of N consecutive links located between the two optical sensors is measured. In addition, during measurement, the chain speed is set to be constant.

[0007] Between the two optical sensors, there are always N (for example, 15) links along with the traveling of the chain, and the total elongation amount of these N links is continuously measured. Since the chain elongates substantially uniformly, the total elongation amount of the chain over time can generally be monitored without problems by this measurement technique, and it can be determined whether the chain needs to be replaced.

[0008] However, in reality, although the frequency is low, there are still the following situations: partial elongation occurs where only one link is significantly elongated, thus requiring chain replacement. In this case, in the aforementioned measurement technique, since the total elongation of N links that enter a specified range (between two photoelectric sensors) through the advancement of the chain is measured, the part where only one link is significantly elongated will be buried in the measurement data, resulting in the elongation amount being determined not to have reached the threshold for chain replacement. Therefore, it is impossible to correctly determine the situation where the chain needs to be replaced due to partial elongation, and there is a high possibility of missing it. In the worst case of this situation, the chain link in this part may be damaged. Summary of the Invention

[0009] In the measurement techniques to date, when partial elongation occurs randomly in one link, or a small number of links less than a specified number, or separated links in one loop of the chain, it cannot be detected.

[0010] The present invention provides a chain elongation detection device capable of reliably detecting partial elongation of a chain.

[0011] The chain elongation detection device according to an embodiment of the present invention detects the elongation of a chain formed by connecting a plurality of links in a ring shape and spanned between a pair of sprockets, and is characterized by comprising: a measurement unit that continuously measures the total elongation of a plurality of links in a specified section of the chain that meshes with the pair of sprockets and travels in the length direction; and a prediction unit that predicts the elongation of each link by inputting the total elongation measured by the measurement unit into a system that has been trained using a large number of data sets formed by combining the elongation data of each link in one loop of the chain and data corresponding to the total elongation measured by the measurement unit.

[0012] According to the above structure, it is possible to reliably detect partial elongation of a chain that has been difficult to detect in the past, and it is possible to effectively replace a chain with partial elongation. Brief Description of the Drawings

[0013] Figure 1 It is a configuration diagram of a chain elongation detection device according to an embodiment.

[0014] Figure 2A It is an exploded perspective view of a chain that is an object of elongation detection according to an embodiment.

[0015] Figure 2B It is a diagram illustrating an example of the sensor configuration of a measurement unit for chain elongation according to an embodiment.

[0016] Figure 3In the figure, (a) is a graph showing the elongation of each link of the chain measured by the chain elongation measuring unit, and (b) is a waveform graph showing the elongation of each link in one turn of the chain.

[0017] Figure 4 In the figure, (a) is a diagram showing the elongation measurement state of a plurality of chain links within a predetermined interval defined by the setting position of the sensor, and (b ... Figure 3 (b) is a waveform diagram showing changes in the total elongation of a plurality of chain links when the elongation of the partially extended chain is measured.

[0018] Figure 5 This is a diagram showing a deep learning model configured as an example of a chain elongation estimation unit in one embodiment.

[0019] Figure 6 Shown Figure 5 An example of a training dataset for a deep learning model shown in FIG. 1 , where (a) shows target data on the output side and (b) shows data on the input side.

[0020] Figure 7 Shown Figure 5 An example of a large training dataset for the entire deep learning model shown in Figure 1, where (a) shows the target data on the output side and (b) shows the data on the input side.

[0021] Figure 8 Shown Figure 5 Another example of a large training dataset for a deep learning model is shown, where (a) shows the target data on the output side and (b) shows the data on the input side.

[0022] Figure 9 Shown Figure 5 Another example of a large amount of training data sets for a deep learning model is shown, where (a) shows target data on the output side and (b) shows data on the input side.

[0023] Figure 10 This is a configuration diagram showing an example in which a plurality of estimation units are provided in the chain extension detection device according to one embodiment.

[0024] Figure 11 This is a structural diagram showing an example in which the estimation unit in the chain extension detection device according to one embodiment is provided on the escalator microcomputer side.

[0025] Figure 12 It is a diagram illustrating a chain extension measuring unit in a chain extension detection device according to another embodiment.

[0026] Figure 13In (a), it is a diagram showing a state where only one link in a loop of the chain is greatly elongated, and in (b), it is a waveform diagram showing the measurement result of the total elongation of a plurality of links in a specified measurement interval obtained by measuring the chain in (a) using an elongation measurement unit Figure 12 of the elongation.

[0027] Figure 14 In, (a) is a diagram showing a partial elongation waveform which is the target data on the output side in the case where the measured result of the partially elongated chain shown Figure 13 is shifted to form a training dataset 1 for a deep learning model, and (b) is a diagram showing the measured waveform on the input side in this case.

[0028] Figure 15 In, (a) is a diagram showing a partial elongation waveform which is the target data on the output side in the case where the measured result of the partially elongated chain shown Figure 13 is shifted to form a training dataset 2 for a deep learning model, and (b) is a diagram showing the measured waveform on the input side in this case.

[0029] Figure 16 In, (a) is a diagram showing a partial elongation waveform which is the target data on the output side in the case where the measured result of the partially elongated chain shown Figure 13 is shifted to form a training dataset 3 for a deep learning model, and (b) is a diagram showing the measured waveform on the input side in this case.

[0030] Figure 17 In, (a) is a diagram showing a partial elongation waveform which is the target data on the output side in the case where the measured result of the partially elongated chain shown Figure 13 is shifted to form a training dataset 4 for a deep learning model, and (b) is a diagram showing the measured waveform on the input side in this case. Detailed implementation mode

[0031] Hereinafter, this implementation mode will be described in detail with reference to the accompanying drawings.

[0032] In Figure 1 , a chain elongation detection device according to an implementation mode is shown. Figure 1 The drive chain part of an escalator is shown. Here, the escalator has a drive chain, a handrail drive chain, a step chain, etc., and in this implementation mode, the drive chain is exemplified as the chain 10 to be detected for elongation.

[0033] The chain 10 is formed into a loop and is installed between a pair of sprockets (the driving sprocket 12 and the driven sprocket 13). The driving sprocket 12 is provided on the speed reducer 11 of the escalator driving motor. The driven sprocket 13 is used to drive the looped steps and handrails (not shown). Further, the chain 10 is driven by meshing with the pair of sprockets 12 and 13 and travels in a loop along the length direction.

[0034] As Figure 2A shown, the chain 10 arranges the shaft portions 101 at a prescribed pitch and connects between the both ends of these adjacent shaft portions 101 by the chain links 102. Further, the shaft portion 101 is composed of a pin shaft 101c, a sleeve 101b, and a roller 101a. In addition, the chain link 102 has an outer chain link 102a and an inner chain link 102b, and these outer chain links 102a and inner chain links 102b are alternately connected between the shaft portions 101.

[0035] Returning to Figure 1 , the chain elongation detection device of this embodiment has: a measurement unit 15 for measuring the elongation of the chain; a transmission device 20 that transmits the measurement data measured by the measurement unit 15 to the monitoring center 21; and an estimation unit 22 that is provided in the monitoring center 21 and estimates the elongation amount of each chain link of the chain 10 based on the transmitted measurement data.

[0036] The measurement unit 15 for measuring the elongation of the chain is basically the same as the measurement unit described in the prior art, and has two sensors (hereinafter referred to as photoelectric sensors) 151, 152 and an elongation amount calculation unit 153 configured in the escalator microcomputer 19. As Figure 2B shown, the measurement unit 15 for measuring the elongation of the chain continuously measures the total elongation amount of a plurality of chain links 102 of the chain 10 traveling in a prescribed section defined by the installation positions of the two photoelectric sensors 151, 152. Further, in Figure 2B , only the outer chain link of the chain link 102 is shown, and the illustration of the inner chain link is omitted.

[0037] The two photoelectric sensors 151, 152 are arranged at a length interval that is an integral multiple of the pitch of the chain 10 along the moving direction of the chain 10. For example, as Figure 1 and Figure 2B shown, the first and second photoelectric sensors 151, 152 are arranged at positions on the left and right of the chain break detection device 18 arranged on the chain 10. The setting interval of the two photoelectric sensors 151, 152, that is, the distance between the respective optical axes, is set to a pitch interval corresponding to the chain link of the chain 10, for example, 15 × pitch.

[0038] As Figure 2BAs shown, the photoelectric sensors 151 and 152 are composed of a light projecting part and a light receiving part, and output a signal to the elongation calculation part 153 whenever the shaft part 101 of the chain 10 passes through the optical axis between them. The elongation calculation part 153 calculates the elongation of the chain 10 based on the magnitude of the time deviation between the two detection signals input from the two photoelectric sensors 151 and 152.

[0039] The estimation part 22 estimates the elongation of each link of the chain 10 from the measurement data sent from the measurement part 15 as described above. The estimation part 22 has an estimation system, such as a deep learning model, that has been trained using a large number of data sets, and uses this system to estimate the elongation of each link of the chain 10. The training data set is a combination of the elongation data of each link in one turn of the chain 10 and data equivalent to the total elongation measured by the measurement part 15, and a large number (for example, 1000 or 2000) of such data sets are prepared for training. That is, the data equivalent to the aforementioned total elongation is used as the input, and the elongation data of each link is used as the target for training. The elongation of each link is estimated by inputting the total elongation measured by the measurement part 15 into the system trained in this way.

[0040] Figure 3 (a) of shows the elongation measurement part of one link of the chain, Figure 3 (b) of shows the elongation of each link of one turn of the chain 10 formed by connecting a plurality of such links, for each link. As described above, the link 102 of the chain 10 has an outer link 102a and an inner link 102b, and they are alternately connected between the shaft parts 101. As is well known, the elongation between the inner links 102b is substantially zero, and elongation occurs between the outer links 102a. Therefore, as shown in Figure 3 (b) of, the non-elongated side and the elongated side are represented as an alternately repeating zigzag elongation.

[0041] Each link 102 of the chain 10 has a substantially uniform elongation as shown in Figure 3 (b) of, but rarely, as shown in the figure, there is a case where only one link has a large elongation. If the part where only one link is elongated exceeds the replacement value as shown in the figure, the chain needs to be replaced. Figure 4 is the measured value obtained by measuring the elongation of the chain 10 using the measurement part 15 shown in Figure 1 . The photoelectric sensors 151 and 152 are arranged at length intervals that are an integer multiple (set to 15 times the pitch of the chain 10 as described above in this embodiment) of the pitch of the chain 10 in the moving direction of the chain 10, and sequentially measure, as the chain 10 advances, as shown in Figure 3 . Figure 4The total elongation of 15 consecutive links that have entered the interval defined by the installation positions of the photoelectric sensors 151 and 152 as shown in (a). Therefore, as shown in Figure 4 (b), Figure 3 The variation in the minute elongation of each link as shown can be measured smoothly, and a relatively flat measured value can be obtained. In addition, for the part where only one link is greatly elongated, due to the smoothing of the measured value, it also tends to be buried in the overall measured value and is judged to have an elongation smaller than the replacement line of the chain.

[0042] Figure 1 The estimation unit 22 provided in the monitoring center 21 estimates the elongation of each link based on the measurement data, which is the total elongation of 15 smoothed links as shown in (b) of Figure 4 , and is sent from the measurement unit 15 provided on the escalator side. As the system for estimation by the estimation unit 22, a trained deep learning model as shown in Figure 5 is used. This deep learning model is configured as a three-layer neural network, and the input side has nodes equivalent to the number of links in one circle of the chain 10 (for example, 112 nodes). The input data is the chain elongation measurement data measured by the measurement unit 15, that is, the measurement data obtained by measuring the total elongation of 15 consecutive links by changing the measurement range one link at a time for one circle of the chain as the chain 10 advances.

[0043] The measurement data input to the trained deep learning model passes through the first layer (336 nodes), the second layer (224 nodes), and the third layer (112 nodes) and is output. That is, the output side outputs the estimated value of one circle of the chain converted into the elongation of each link from 112 nodes equivalent to the number of links in one circle of the chain 10 on the input side. The neural network constituting this deep learning model is pre-trained using a training dataset to have the above-mentioned estimation function.

[0044] An example of the training dataset is described using Figure 6 (a) and (b). Figure 6 (a) is simulated elongation data that defines the elongation of each link in one circle of the chain, and it is the training data (target data: the data to be obtained) as the output. This simulated elongation data simulates the elongation mode of the actual chain and is set as a zigzag elongation amount in which the non-elongated side and the elongated side alternate. For the elongated links, their elongation amounts are set as random values. That is, the simulated elongation data is set with random values such that the elongation amount of the inner links that generally do not elongate is 0 and the outer link sides have random elongation amounts respectively.

[0045] Figure 6 (b) is obtained by calculation inFigure 1 The measurement unit 15 measures in groups of every 15 links Figure 6 When measuring the chain simulated elongation shown in (a) of Figure 6 what kind of simulated measurement data will be measured, which is the data on the input side of the training data. Prepare a large number, for example, 1000, of data sets composed of such simulated measurement data on the input side and

[0046] Figure 7 The examples of the data sets for this training are shown in (a) and (b) of . Use these large number of data sets from the 1st to the Mth (in this case, the 1000th) to train the deep learning model. When the training is carried out in a manner with a sufficiently small error, if the measurement data of every 15 links measured by the measurement unit 15 is provided as described above, the elongation of each link will be output with high precision.

[0047] In the above structure, using the chain elongation detection device described in Figure 1 , for example, measure the elongation of the chain 10 once a day. That is, rotate and drive a pair of sprockets 12 and 13 to make the chain 10 travel around in the length direction. Use the elongation calculation unit 153 in the escalator microcomputer 19 to calculate the total elongation of 15 links that have entered the section defined by the installation positions of the photoelectric sensors 151 and 152 through this travel. Measure the total elongation of every consecutive 15 links of the chain 10 in one circle. The measurement data is temporarily stored in the memory in the escalator microcomputer 19.

[0048] This measurement data is sent by the measurement data sending device 20 to the monitoring center 21 at a certain specified time. In the monitoring center 21, this measurement data is input into the trained deep learning model that becomes the estimation unit 22. Figure 5 The trained deep learning model shown in converts the input measurement data of every 15 links into the measurement data (estimated value) of the elongation of each link. The data after this conversion is used for the determination of whether the chain needs to be replaced.

[0049] In this way, the measurement data (estimated value) of the elongation of each link of the chain 10 can be obtained, so even if only one link elongates, it can be reliably detected. That is, in the prior art, regarding the elongation data of the chain, it is judged whether the chain needs to be replaced based on the waveform obtained in the form of Figure 4 in (b) of . Therefore, in the case where the actual elongation of the chain is in a state where only one link elongates significantly as Figure 3 , sometimes it will be judged that the chain does not need to be replaced.

[0050] In contrast, in the present embodiment, the estimation unit 21 is used to obtain data that converts the measurement data of Figure 4 into the measurement value of each link of Figure 3 . Then, this data is used for judgment. Therefore, even a part where only one link is greatly elongated can be correctly detected without making a wrong judgment. Thus, it is possible to correctly detect the situation where only one link is partially elongated, automatically prompt the inspector to conduct on-site investigation and issue an instruction to replace the chain, and further improve safety.

[0051] As another example of the training dataset for the prediction unit 22, there are Figure 8 (a), (b) of Figure 9 and the training datasets shown in (a), (b) of

[0052] Figure 8 is training data that mainly assumes partial elongation where a part of the chain 10 is elongated, showing the case of preparing 1000 training datasets. Figure 8 (a) of Figure 8 is simulated elongation data representing the elongation of each link of the chain 10, which is the output-side data (target). Figure 8 (b) of Figure 5 is simulated measurement data that simulates the measurement data of every 15 links measured by the measurement unit 15. It is calculated based on the elongation amount of

[0053] (a) of Figure 8 and is the training data on the input side of the deep learning model of Figure 8 (a) of

[0054] In the simulated elongation data of Figure 8 (a) mentioned above, the second one from the top is that the elongation amount of the inner link that does not elongate is 0, and only one outer link randomly elongates in one turn of the chain. In addition,

[0055] For the first, third, and 1000th ones from the top in Figure 8 (a) of

[0054] it is assumed that the elongation amount of the inner link that does not elongate is 0, and only a few consecutive outer links with a maximum of about 5 in one part of the outer link side of the chain elongate in one turn of the chain. The elongation amount of each link is randomly set. Figure 8 The above-mentioned training dataset of

[0055] assumes that partial elongation occurs concentratedly in one part of one turn of the chain. In one turn of the chain, the elongation amount of the partially elongated part is a random value, and the link number where elongation occurs is also random. In addition, the number of partially elongated links appearing in one part is set to a random number within a relatively small range of 5 links or less. According to the deep learning model trained using such a dataset, it is possible to reliably detect partial elongation of the chain generated in various ways.

[0055] In addition, in Figure 8In this case, it is assumed that partial elongation of one or a smaller number (less than 5) of links occurs at one part in one turn of the chain. However, it is also possible to assume that the above partial elongation occurs at a plurality of mutually separated parts in one turn of the chain to form Figure 8 the simulated elongation data of (a), and based on this, calculate by theoretical calculation Figure 8 the simulated measurement data of (b).

[0056] When using a deep learning model trained with the training data using this Figure 8 , the training of the deep learning model is easy and the training accuracy is improved accordingly. That is, when using Figure 6 , Figure 7 as the training data, since it is the training data in which all the links of one turn of the chain are randomly elongated, the amount of information that the deep learning model has to estimate is very large. Therefore, it is quite difficult to obtain a model with a relatively high estimation accuracy, and it is not guaranteed that a high-precision model will necessarily be obtained.

[0057] In contrast, when using a deep learning model trained with the training data using Figure 8 shown, in one turn of the chain, the maximum number of parts to be estimated is about 5 links, and the information estimated by the deep learning model is greatly reduced. Therefore, the training of the deep learning model is relatively easy, and the estimation accuracy of the obtained model can be expected to be improved, and the elongation amount of each link can be estimated with higher accuracy. However, for the chain data in which the whole is elongated, since its structure is different from that of the training model, the estimation accuracy may instead decrease.

[0058] Figure 9 The training data shown in Figure 9 assumes that the elongation amounts of all the outer links are elongated by a uniform fixed value. That is, in this case, it is assumed that the whole chain is uniformly elongated rather than partially elongated. The elongated side and the non-elongated side appear alternately, and each elongation amount is equal throughout one turn of the chain. This is similar to the most common elongation method in an actual chain.

[0059] In Figure 9 's training data, the elongation amounts of all the links are made constant. Therefore, compared with Figure 6 , Figure 7 where the elongation amounts of all the links are set randomly, it is easier to train, and the estimation accuracy after training is also higher. In addition, Figure 9 assumes the most common elongation method of a general chain. Therefore, for the influence of sensor noise and the like mixed in the measurement data, by using this Figure 9The training data is provided to the trained deep learning model with the measurement data and output, and the sensor noise is also removed. Therefore, compared with the prior art that does not use the deep learning model, it is expected to be able to detect the average elongation of the chain with higher accuracy.

[0060] In each of the above training data, except for the case where all data is 0, the input-side simulated measurement data was standardized in such a way that the maximum value became a specific value of 1 or less than 1.

[0061] In addition, the output side of the training data may also include the actual chain elongation data actually measured by the measurement system, rather than just the simulated elongation data generated by simulation.

[0062] Figure 10 The chain elongation detection device has multiple (22A, 22B) trained deep learning models as the estimation unit 22 in the monitoring center 21, and a comprehensive determination unit 23 that receives their output values and comprehensively judges the chain state. In this case, the characteristics of the multiple deep learning models 22A and 22B are different. For example, the deep learning model 22A uses Figure 8 the training data for training, and the deep learning model 22B uses Figure 9 the training data for training. If configured in this way, in the deep learning model 22A, mainly the parts that are greatly elongated partially are extracted, and in the deep learning model 22B, the average elongation amount is extracted. The comprehensive determination unit 23 determines whether the chain needs to be updated based on the extraction results of the two.

[0063] In this Figure 10 chain elongation detection device, both the case where the average elongation amount increases and the case where there are large elongations in parts can be effectively captured. Therefore, the elongation state and abnormalities of the chain will not be missed, the safety of passengers can be further ensured, and maintenance can be performed with higher accuracy.

[0064] Figure 11 The chain elongation detection device constitutes a trained deep learning model as the estimation unit 22 inside the escalator microcomputer 19, rather than on the monitoring center 21 side. If configured in this way, the elongation amount of each link can be estimated inside the escalator microcomputer 19. Therefore, the estimated data of the elongation amount of each link is extracted on each escalator side and output to the monitoring center 21 side. Since the monitoring center 21 side does not perform data processing on the data sent from each escalator one by one, it only needs to judge whether there are abnormalities in the chain and whether it needs to be replaced, which has the effect of reducing the processing on the monitoring center 21 side.

[0065] In any of the above embodiments, the two sensors 151 and 152 for detecting the elongation of the chain are photoelectric sensors, but they are not necessarily limited to photoelectric sensors. As long as the same measurement principle is used, the type of sensor does not matter. For example, they can be laser displacement sensors, proximity sensors, etc., and are applicable to all systems for measuring the total elongation of a continuous chain section.

[0066] In the foregoing embodiment, Figure 1 and Figure 10 、 Figure 11 the chain elongation measuring units 15 shown in the figure are each configured with a first sensor 151 and a second sensor 152 at intervals corresponding to a plurality of chain links (15 chain links are illustrated) along the traveling direction of the chain, and detect the total elongation of the 15 chain links that enter between the sensors 151 and 152 as the chain 10 moves. However, a chain elongation measuring unit other than this structure can also be used. In the following embodiments, the chain elongation measuring unit 25 having the structure shown in Figure 12 the figure is used.

[0067] The chain 10 to be measured is mounted between a pair of sprockets 12 and 13. The first sensor 251 and the second sensor 252 constituting the chain elongation measuring unit 25 are arranged near the corresponding sprockets 12 and 13, detect the passing moments of the teeth of the sprockets 12 and 13, and output detection signals. The elongation of the chain 10 is measured based on the generation moments of these detection signals. The measurement principle has been applied for by the applicant of the present application as Japanese Patent Application No. 2020-131993. In addition, Figure 12 is a diagram for explaining the measurement principle. The sprockets 12 and 13 are drawn with the same diameter, but this is for ease of understanding of the explanation. Even if they are of different diameters as shown in Figure 1 etc., it is the same.

[0068] The chain 10 is driven by the rotation of the sprockets 12 and 13 and moves in a circumferential manner along its length direction. At this time, the first sensor 251 detects the passing moment of the teeth of the sprocket 12 and outputs a first detection signal. The second sensor 252 detects the passing moment of the teeth of the sprocket 13 and outputs a second detection signal. These first and second detection signals are input to a calculation unit 253 for calculating the chain elongation, and the elongation of the chain 10 is measured based on the phase difference between these first and second detection signals.

[0069] This calculation unit 253 corresponds to the elongation calculation unit 153 shown in Figure 1 etc., and the calculation result is sent by the sending device 20 shown in Figure 1 to the monitoring center 21. In addition, as the first sensor 251 and the second sensor 252, transmissive photoelectric sensors, reflective photoelectric sensors, proximity sensors, etc. can be used.

[0070] The operation of the chain elongation measurement unit 25 will be described. As Figure 13 shown in (a) of [], it is assumed that only one link in one turn of the chain 10 is significantly elongated. During the period when the one significantly elongated link moves with the movement of the chain 10 from Figure 12 the starting point A of meshing with the sprocket 13 to the ending point B of meshing as shown in [], the measured value of the measurement unit 25, that is, the elongation amount detected by the first sensor 251 and the second sensor 252 and calculated by the calculation unit 253 changes as Figure 13 shown in (b) of []. This elongation amount is the total elongation amount of the number of links in a specified section from the above starting point A to the ending point B.

[0071] The measured elongation amount is input into the Figure 1 shown estimation unit 22, and the elongation amount of each link is estimated using the Figure 5 shown deep learning model. In this case, the training of the deep learning model constituting the estimation unit 22 is performed using the Figures 14 to 17 shown training data set. These Figures 14 to 17 shown training data sets are obtained by shifting the link positions of the Figure 13 shown measured data, and a large number of them are prepared and used as training data.

[0072] This is because the measurement result of the measurement unit 25 for the elongation of only one link as shown in (a) of [] is non-linear as shown in (b) of [], so it is difficult to generate the total elongation amount of multiple links (input side) based on the elongation amount of each link (output side) of (a) as in the data set of the foregoing embodiment shown in []. Figure 13 Figure 13 Figures 6 to 9

[0073] Figure 12 Figure 13 Figures 14 to 17

[0074]

[0075] shown measurement waveform. By following the measurement data of this test, a measurement waveform with the position of one link elongation shifted is formed as shown in []. Such a measurement waveform can be easily obtained through data processing, so a large number of training data sets are prepared using one measured data to train the deep learning model.

[0074] By inputting the total value of multiple elongations measured by the measurement unit 25 into the estimation unit 22 trained in this way, it is possible to reliably detect the partial elongation of the chain 10 in the same way as in the foregoing embodiment.

[0075] ​​Some embodiments of the present invention have been described, but these embodiments are presented as examples and are not intended to limit the scope of the invention. These new embodiments can be implemented in various other ways, and various omissions, substitutions, and changes can be made without departing from the gist of the invention. These embodiments and their variations are included in the scope and gist of the invention, and are included in the invention described in the claims and its equivalent scope.

Claims

1. A chain elongation detection device for detecting the elongation of a chain formed by connecting a plurality of chain links into a ring and spanned between a pair of sprockets, characterized in that, Comprising: A measurement unit that continuously measures the total elongation of a plurality of chain links traveling within a specified section of a chain that meshes with the pair of sprockets and travels in the longitudinal direction; and An estimation unit that estimates the elongation of each chain link by inputting the total elongation measured by the measurement unit into a system that has been trained with a large number of data sets formed by combining the elongation data of each chain link in one loop of the chain and data corresponding to the total elongation measured by the measurement unit. The data set for training the estimation unit is formed by generating a large number of data sets as training data sets, which are formed by using a data set that is pre-combined with the elongation data of each chain link in one loop of the chain that has undergone partial elongation and the measured value of the total elongation of a plurality of chain links within the section from the starting point to the ending point where the chain meshes with one sprocket, and shifting the chain link positions of these data sets.

Images