Abnormality prediction device and escalator system

Abstract

The purpose of the present disclosure is to provide an abnormality prediction device for predicting abnormality in a moving handrail of an escalator in order to improve the efficiency of maintenance and inspection and prevent accidents, and an escalator system. An abnormality prediction device (10) according to one aspect of the present disclosure predicts abnormality in a moving handrail (12) of an escalator (11). The abnormality prediction device (10) is provided with: a data acquisition unit that acquires use information pertaining to the use of the escalator (11); a calculation unit (15) that outputs a characteristic value of the moving handrail (12) from the use information input from the data acquisition unit, using an estimation model (16) trained in advance by machine learning to infer the characteristic value of the moving handrail (12) from the use information of the escalator (11); and a determination unit that, when the characteristic value of the moving handrail (12) output from the calculation unit (15) falls below a preset threshold value, determines that replacement of the moving handrail (12) is necessary.

Description

Abnormality prediction device and escalator system

[0001] The present disclosure relates to an abnormality prediction device that predicts an abnormality in a moving handrail of an escalator, and an escalator system.

[0002] The components used in escalators deteriorate over long periods of operation. For example, the handrails used in escalators are molded resin parts made from multiple materials, including metal steel wire, synthetic resin, and canvas. As the escalator operates, the handrails are continuously bent and driven by being squeezed between multiple rollers. As a result, the synthetic resin, canvas, and other components deteriorate over time due to fatigue caused by heat, humidity, and continuous mechanical stress, and the handrails of escalators deteriorate over long periods of operation.

[0003] Therefore, it is necessary to predict various defects and abnormalities caused by deterioration of moving handrails in advance and set up replacement plans, and to replace moving handrails in advance before defects or abnormalities occur, as part of maintenance inspections and accident prevention.Patent Documents 1 to 3 propose devices that predict or detect abnormalities in elevators such as escalators.

[0004] Japanese Patent No. 5513456 Japanese Patent Application Laid-Open No. 2023-76024 Japanese Patent Application Laid-Open No. 2003-238041

[0005] Patent Documents 1 and 2 disclose devices that use machine learning to predict or detect elevator abnormalities based on inputs of acoustic signals and temperature distribution images of parts. However, when diagnosing elevator abnormalities using these devices, if an abnormality occurs in an elevator part and the sound or vibration does not change, no change will appear in the acoustic signals or temperature distribution images of the parts, and the elevator abnormality cannot be detected. Patent Document 3 also discloses a device that determines the remaining lifespan of a part based on data on part replacement due to breakdowns or accidents in the field. However, this device cannot determine the remaining lifespan of a moving handrail if data on serious accidents, which are rare causes of replacement, is not available.

[0006] The present disclosure has been made to solve the above-mentioned problems, and its purpose is to provide an abnormality prediction device and an escalator system that predict abnormalities in the moving handrail of an escalator in order to improve the efficiency of maintenance and inspection and prevent accidents.

[0007] An abnormality prediction device according to one aspect of the present disclosure predicts an abnormality in a moving handrail of an escalator. The abnormality prediction device includes a data acquisition unit that acquires usage information related to escalator use, a calculation unit that uses an estimation model previously trained through machine learning to infer a characteristic value of the moving handrail from the escalator usage information and outputs the characteristic value of the moving handrail from the usage information input from the data acquisition unit, and a determination unit that determines that the moving handrail needs to be replaced if the characteristic value of the moving handrail output from the calculation unit falls below a preset threshold. The characteristic value of the moving handrail includes at least one value of the bond strength between the synthetic resin and metal steel wire that make up the moving handrail, the physical property value of the synthetic resin that makes up the moving handrail, and the physical property value of the canvas that makes up the moving handrail.

[0008] According to the abnormality prediction device of the present disclosure, a characteristic value of the moving handrail is output from usage information using an estimation model trained by machine learning, and if the characteristic value of the moving handrail falls below a preset threshold, it is determined that the moving handrail needs to be replaced.This makes it possible to predict abnormalities in the moving handrail of an escalator in order to improve the efficiency of maintenance and inspection and prevent accidents.

[0009] Fig. 1 is a schematic diagram showing the configuration of an escalator system equipped with an abnormality prediction device according to embodiment 1. Fig. 2 is a cross-sectional view of a moving handrail. Fig. 3 is a functional block diagram for explaining the hardware configuration of the abnormality prediction device. Fig. 4 is a schematic diagram showing the configuration of an escalator system equipped with an abnormality prediction device according to embodiment 2. Fig. 5 is a schematic diagram showing the configuration of an escalator system equipped with an abnormality prediction device according to embodiment 3. Fig. 6 is a schematic diagram showing the configuration of an escalator system equipped with an abnormality prediction device according to embodiment 4.

[0010] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are designated by the same reference numerals, and their description will not be repeated in principle.

[0011] Embodiment 1. Fig. 1 is a schematic diagram showing the configuration of an escalator system 100 equipped with an abnormality prediction device 10 according to embodiment 1. Fig. 2 is a cross-sectional view of a moving handrail 12. Fig. 3 is a functional block diagram for explaining the hardware configuration of the abnormality prediction device 10. The escalator system 100 according to embodiment 1 includes an escalator 11 and an abnormality prediction device 10 that predicts an abnormality in the moving handrail 12 of the escalator 11.

[0012] The escalator 11 is a known device used in the market, and although not shown, a truss is installed between the upper and lower floors of a building. Although not shown, the truss is provided with a drive unit, an upper sprocket, a drive chain, and an emergency brake at its upper floor end, and a lower sprocket at its lower floor end. Although not shown, the escalator 11 has an endless step chain hung between the upper and lower sprockets, and a plurality of steps connected to the step chain. In the escalator 11, the step chain and the plurality of steps move in a circular motion as the upper sprocket rotates.

[0013] Furthermore, a pair of balustrades is provided on the truss of the escalator 11, and each balustrade is provided with an endless moving handrail 12. The escalator 11 employs, for example, a roller drive type drive device, in which the moving handrail 12 is sandwiched between a pressure roller 13 and a drive roller 14, and the moving handrail 12 is circulated in synchronization with the movement of the steps by rotating the drive roller 14.

[0014] As can be seen from the cross-sectional view of FIG. 2 , the moving handrail 12 is made of, for example, synthetic resin 22, metal steel wire 23, and canvas 24, and is a resin molded product with the metal steel wire 23 disposed inside the synthetic resin 22. The moving handrail 12 is integrated by fusing the components together using adhesives, coupling agents, etc. When the escalator 11 is in operation, the moving handrail 12 is continuously bent and compressed between the pressure roller 13 and the drive roller 14. Furthermore, the synthetic resin 22 and canvas 24 deteriorate over time due to heat and humidity. Therefore, various defects and abnormalities occur in the moving handrail 12 when the escalator 11 is operated over a long period of time.

[0015] For example, in the case of a moving handrail 12 using metal steel wire 23, operating the escalator 11 over a long period of time can reduce the bond strength between the metal steel wire 23 and the synthetic resin 22, causing the metal steel wire 23 to be unable to be held in the synthetic resin 22 and resulting in a problem of being exposed from the surface of the moving handrail 12. Furthermore, operating the escalator 11 over a long period of time can cause the canvas 24 of the moving handrail 12 to wear, reducing the frictional force between the canvas 24 and the drive roller 14 and causing slippage, which can cause the moving speed of the moving handrail 12 to be slower than the moving speed of the steps.

[0016] Therefore, the escalator system 100 according to the first embodiment is equipped with an abnormality prediction device 10 that predicts in advance various defects and abnormalities that may occur in the moving handrail 12. As shown in Fig. 1, the abnormality prediction device 10 includes a calculation unit 15 that outputs characteristic values ​​of the moving handrail 12 from usage information related to the use of the escalator 11, and a determination unit 17 that determines whether or not the moving handrail 12 needs to be replaced based on the characteristic values ​​of the moving handrail 12 output from the calculation unit 15.

[0017] The calculation unit 15 has an estimation model 16 that has been trained in advance by machine learning based on usage information regarding the use of the escalator 11 and data on characteristic values ​​of the moving handrail 12. The estimation model 16 includes a network structure, such as a known neural network, support vector machine (SVM), or Bayesian network, and internal parameters used by the network structure, and the internal parameters are optimized (adjusted) by training based on data on the characteristic values ​​of the moving handrail 12 estimated from the usage information regarding the use of the escalator 11 and the characteristic values ​​of the moving handrail 12 measured in response to the usage information. Furthermore, various algorithms can be used to train the estimation model 16, and can be selected from known algorithms such as multiple regression, Ridge regression, Lasso regression, support vector regression, ElasticNet regression, decision tree, Gaussian process regression, random forest, and gradient boosting decision tree.

[0018] Here, synthetic resin 22 generally undergoes oxidation and hydrolysis reactions due to temperature and humidity, resulting in the severing of molecular chains over time, resulting in a decrease in strength and rigidity. Synthetic resins such as vulcanized rubber may harden as the severing of molecular chains recombines, increasing the crosslink density. Furthermore, coupling agents and adhesives also deteriorate over time due to oxidation and hydrolysis reactions caused by temperature and humidity. Furthermore, canvas 24 is made of synthetic fibers such as polyester, and like synthetic resins, it also deteriorates due to oxidation and hydrolysis caused by temperature and humidity during use.

[0019] Deterioration of the moving handrail 12 is not only caused by deterioration of the synthetic resin 22 and canvas 24 themselves, but also by fatigue-induced deterioration of the synthetic resin 22, the interface between the synthetic resin 22 and the metal steel wire 23, and the canvas 24 due to continuous bending during operation and repeated application of pressure loads from the pressure roller 13 and the like.

[0020] Therefore, the information input to the estimation model 16 to estimate the characteristic values ​​of the moving handrail 12 is usage information related to the use of the escalator 11. The usage information related to the use of the escalator 11 includes, for example, the temperature and humidity of the installation location of the escalator 11, the travel distance of the moving handrail 12, the number of days the escalator 11 has been installed, and information on the load that the moving handrail 12 receives from the pressure roller 13 and the drive roller 14. Note that the information included in the usage information related to the use of the escalator 11 is not limited to this, and further information may be added as long as it improves the accuracy of the characteristic values ​​of the moving handrail 12 estimated by the estimation model 16. Conversely, the information included in the usage information related to the use of the escalator 11 may be reduced, although this will decrease the accuracy of the characteristic values ​​of the moving handrail 12 estimated by the estimation model 16. For example, the usage information regarding the use of the escalator 11 may include at least one of the following information: the temperature and humidity of the installation location of the escalator 11, the distance traveled by the moving handrail 12, the number of days since the escalator 11 was installed, and the load that the moving handrail 12 receives from the pressure roller 13 and the drive roller 14.

[0021] The characteristic values ​​of the moving handrail 12 estimated by the estimation model 16 include at least one value of the bond strength between the synthetic resin 22 and the metal steel wire 23 that make up the moving handrail 12, the physical property values ​​of the synthetic resin 22 that makes up the moving handrail 12, and the physical property values ​​of the canvas 24 that makes up the moving handrail 12. Therefore, by inputting information about the temperature and humidity of the installation location of the escalator 11, the travel distance of the moving handrail 12, the number of days the escalator 11 has been installed, and the load that the moving handrail 12 receives from the pressure roller 13 and the drive roller 14 into the estimation model 16, the calculation unit 15 can estimate at least one value of the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24.

[0022] The determination unit 17 determines that the moving handrail 12 needs to be replaced when at least one of the values ​​of the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24 output from the calculation unit 15 falls below a preset threshold. The bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24 are each evaluated when the escalator is designed or manufactured, and are set in advance as the respective threshold values.

[0023] Specifically, the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24, which are measured when various defects or abnormalities occur, such as the metal steel wire 23 becoming exposed from the moving handrail 12, cracks occurring in the synthetic resin 22, breakage of the synthetic resin 22, or slippage between the canvas 24 and the drive roller 14, are specified in advance and set as thresholds. In this way, the abnormality prediction device 10 calculates at least one value from the estimation model 16 of the bond strength between the synthetic resin 22 and the metal steel wire 23 of moving handrails 12 used in the market, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24, and by comparing this with the threshold value, it is possible to predict when defects or abnormalities in the moving handrail 12 will occur.

[0024] The threshold values ​​preset for the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24 may not necessarily be one, but may be multiple. For example, a threshold value may be set for a probability of 80% or more that a defect or abnormality will occur within six months, and a threshold value may be set for a probability of 80% or more that a defect or abnormality will occur within three months. This makes it possible to determine when an abnormality will occur in the moving handrail 12 of the escalator 11, and to efficiently replace the moving handrail 12.

[0025] 3, the hardware configuration of the abnormality prediction device 10 includes a processor 110, a memory 120, a storage 130, an input / output interface 140, a media reader 150, and a communication interface 160. These components are connected via a processor bus 170.

[0026] The processor 110 is a computer that reads programs (for example, an operating system (OS) 131, an estimation program 132, and a judgment program 133) stored in the storage 130, and deploys the read programs in the memory 120 to execute them. The processor 110 is configured, for example, by a central processing unit (CPU), a field programmable gate array (FPGA), a graphics processing unit (GPU), or a multi-processing unit (MPU). The processor 110 may also be configured by a processing circuitry.

[0027] The memory 120 is configured by a volatile memory such as a dynamic random access memory (DRAM) or a static random access memory (SRAM), or a non-volatile memory such as a read only memory (ROM) or a flash memory.

[0028] The storage 130 is configured by a non-volatile storage device such as a hard disk drive (HDD) or a solid state drive (SSD), etc. The storage 130 stores an OS 131, an estimation program 132, a determination program 133, a threshold 134, an estimation model 16, etc.

[0029] The estimation program 132 is a program for estimating at least one value of the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24 from information on the temperature and humidity of the installation location of the escalator 11, the travel distance of the moving handrail 12, the number of days since the escalator 11 was installed, and the load that the moving handrail 12 receives from the pressure roller 13 and the drive roller 14, using the estimation model 16 through machine learning training. The judgment program 133 is a program for predicting when a defect or abnormality will occur in the moving handrail 12 by comparing at least one value of the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24 estimated by the estimation program 132 with a threshold value 134. The abnormality prediction device 10 functions as the calculation unit 15 by executing the estimation program 132, and functions as the judgment unit 17 by executing the judgment program 133.

[0030] The input / output interface 140 receives input operations from input devices such as a keyboard, a mouse, and a touch device, and outputs information to output devices such as a display and a speaker.

[0031] The communication interface 160 acquires data from a management device (not shown) for the escalator 11 and transmits and receives data to and from other devices via wired or wireless communication. The communication interface 160 functions as a data acquisition unit that acquires temperature data 18a, humidity data 18b, mileage data 18c, installation date data 18d, and load data 18e from the pressure roller 13 and drive roller 14 shown in FIG. 1 . The communication interface 160 may communicate with other devices to output information such as predicted timing of defects or abnormalities in the moving handrail 12 to the other devices. Here, the load data 18e corresponds to the pressing forces of the pressure roller 13 and drive roller 14 adjusted when the moving handrail 12 was installed. This is because it is assumed that the pressing forces of the pressure roller 13 and drive roller 14 adjusted at the time of installation will not change during use.

[0032] The media reading device 150 accepts a storage medium such as a removable disk 180, a memory chip, or a USB memory, and acquires data and programs (e.g., the estimation program 132 and the determination program 133) stored in the removable disk 180, the memory chip, the USB memory, etc. The media reading device 150 may read data stored in the removable disk 180, or store predicted information in the removable disk 180 or the like and output the information.

[0033] Note that the anomaly prediction device 10 receives the trained estimation model 16 from a server or the like via the communication interface 160, for example, since the trained estimation model 16 is prepared in advance. The anomaly prediction device 10 may also read the estimation model 16 stored on a removable disk 180 or the like using the media reading device 150. Furthermore, the anomaly prediction device 10 may perform training by machine learning using training data and generate or update the estimation model 16 itself.

[0034] The estimation model 16 has been trained in advance by machine learning using training data. The training data is data that correlates data on at least one of the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical properties of the synthetic resin 22, and the physical properties of the canvas 24, which are measured for a plurality of moving handrails 12 collected from the market, with collected data on the temperature, humidity, number of days since installation, running distance, and loads received from the pressure roller 13 and drive roller 14 at the installation location of the escalator 11.

[0035] The training data does not necessarily need to be data on at least one of the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical properties of the synthetic resin 22, and the physical properties of the canvas 24, measured for multiple moving handrails 12 collected from the market. For example, the training data may be data on at least one of the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical properties of the synthetic resin 22, and the physical properties of the canvas 24, measured for moving handrails 12 that have been subjected to accelerated deterioration in a high-temperature, high-humidity chamber, or for moving handrails 12 that have been subjected to accelerated fatigue by repeated bending using a fatigue testing machine or the like. The training data may also include both data measured on moving handrails 12 collected from the market and data measured on moving handrails 12 that have been subjected to accelerated deterioration using a testing machine or the like.

[0036] Next, the bond strength between the synthetic resin 22 and the metal steel wire 23 estimated by the calculation unit 15 will be described in detail. The bond strength between the synthetic resin 22 and the metal steel wire 23 can be evaluated, for example, by the load (pull-out strength) when the metal steel wire 23 is pulled out from a cross section of a cut moving handrail 12 in a direction perpendicular to the cross section. Furthermore, because the metal steel wire 23 has a central wire and multiple strands, the load (pull-out strength) when the central wire is pulled out from a cross section of a cut moving handrail 12 in a direction perpendicular to the cross section may also be evaluated as the bond strength between the synthetic resin 22 and the metal steel wire 23. Note that there are no particular limitations on the method, device, or conditions for measuring the pull-out strength of the metal steel wire 23 or the pull-out strength of the central wire; however, it is preferable to use the bond strength between the synthetic resin 22 and the metal steel wire 23 evaluated using the same pull-out strength measurement method, device, and conditions when generating the estimation model 16.

[0037] The pull-out strength of the metal steel wire 23 or the pull-out strength of the central wire changes depending on the oxidation reaction and hydrolysis reaction occurring in the synthetic resin 22, the coupling agent at the interface between the synthetic resin 22 and the metal steel wire 23, and the central wire, and the adhesive, repeated bending, and fatigue due to the load received from the pressure roller 13, etc. Therefore, the calculation unit 15 can estimate the pull-out strength of the metal steel wire 23 or the pull-out strength of the central wire using the estimation model 16, using as input information the temperature and humidity of the installation location of the escalator 11, the travel distance of the moving handrail 12, the number of days since the escalator 11 was installed, and the load received by the moving handrail 12 from the pressure roller 13 and drive roller 14.

[0038] The threshold value 134 used in the judgment unit 17 is set in advance based on the pull-out strength of the metal steel wire 23 of the moving handrail 12 or the pull-out strength of the central wire, which was evaluated when the escalator was designed or manufactured. As a result, the pull-out strength of the metal steel wire or the pull-out strength of the central wire of the moving handrail 12 used in the market can be estimated by the estimation model 16 trained by machine learning and compared with the threshold value 134, making it possible to predict when an abnormality such as the metal steel wire 23 protruding from the moving handrail 12 will occur.

[0039] The estimation model 16 is generated by performing machine learning training using training data that correlates data on the pull-out strength of the metal steel wire 23 or the pull-out strength of the central wire measured for multiple moving handrails 12 collected from the market with data collected on the temperature, humidity, number of days in installation, running distance, and loads received from the pressure roller 13 and drive roller 14 at the installation location of the escalator 11.

[0040] The training data does not necessarily need to be data measured on multiple moving handrails 12 collected from the market. For example, the training data may be data on the pull-out strength of the metal steel wires 23 measured on moving handrails 12 that have been subjected to accelerated deterioration in a high-temperature, high-humidity chamber, or on moving handrails 12 that have been subjected to accelerated fatigue by repeatedly bending using a fatigue testing machine or the like, or data on the pull-out strength of the central wire. The training data may include both data measured on moving handrails 12 collected from the market and data measured on moving handrails 12 that have been subjected to accelerated deterioration using a testing machine or the like.

[0041] Next, the physical property values ​​of the canvas 24 estimated by the calculation unit 15 will be described in detail. The physical property values ​​of the canvas 24 can be evaluated, for example, by the kinetic or static friction coefficient of the canvas 24. If the kinetic or static friction coefficient of the canvas 24 decreases, the canvas 24 and the drive roller 14 that drives the moving handrail 12 slip, causing a defect in which the speed of the moving handrail 12 becomes slower than the speed of the steps. The kinetic or static friction coefficient of the canvas 24 at which slip occurs is identified, and the identified value is set in advance as the threshold value 134. The calculation unit 15 estimates the kinetic or static friction coefficient of canvas 24 currently in use on the market using an estimation model 16 trained by machine learning. The determination unit 17 compares the estimated kinetic or static friction coefficient of the canvas 24 with the threshold value 134 to predict when slippage of the moving handrail 12 will occur.

[0042] Canvas 24 is made of synthetic fibers such as polyester. As a result, canvas 24 deteriorates during use due to oxidation and hydrolysis caused by temperature and humidity, and the unevenness of its surface wears away when it comes into contact with pressure roller 13, etc. Therefore, calculation unit 15 can estimate the dynamic friction coefficient or static friction coefficient of canvas 24 using estimation model 16, using as input information such as the temperature and humidity of the installation location of escalator 11, the travel distance of moving handrail 12, the number of days since escalator 11 was installed, and the loads that moving handrail 12 receives from pressure roller 13 and drive roller 14.

[0043] The estimation model 16 is generated by performing machine learning training using training data that correlates data on the dynamic friction coefficient or static friction coefficient of the canvas 24 measured for multiple moving handrails 12 collected from the market with data collected on the temperature, humidity, number of days since installation, running distance, and information on the load received from the pressure roller 13 and drive roller 14 at the installation location of the escalator 11.

[0044] The training data does not necessarily need to be data measured on multiple moving handrails 12 collected from the market. For example, the training data may be data on the dynamic friction coefficient or static friction coefficient of the canvas 24 measured on a moving handrail 12 that has been subjected to accelerated deterioration in a high-temperature, high-humidity chamber or a moving handrail 12 that has been subjected to accelerated fatigue by repeated bending using a fatigue testing machine or the like. The training data may include both data measured on moving handrails 12 collected from the market and data measured on moving handrails 12 that have been subjected to accelerated deterioration using a testing machine or the like. Note that there are no particular limitations on the measurement method, measurement device, or measurement conditions for the dynamic friction coefficient or static friction coefficient of the canvas 24, but it is preferable to use the dynamic friction coefficient or static friction coefficient of the canvas 24 evaluated using the same measurement method, measurement device, and measurement conditions when generating the estimation model 16.

[0045] Next, a detailed description will be given of the physical property values ​​of the synthetic resin 22 estimated by the calculation unit 15. The physical property values ​​of the synthetic resin 22 can be evaluated, for example, by the tensile strength of the synthetic resin 22. Because the moving handrail 12 is subjected to repeated bending and loads from the pressing load of the pressure roller 13, if the tensile strength of the synthetic resin 22 decreases, there is a risk that cracks will occur in the moving handrail 12 or the moving handrail 12 will break.

[0046] The tensile strength of synthetic resin 22 deteriorates due to oxidation reactions and hydrolysis caused by temperature and humidity, and also changes due to fatigue caused by repeated bending and loads from pressure roller 13, etc. Therefore, calculation unit 15 can estimate the tensile strength of synthetic resin 22 using estimation model 16, using as input information the temperature and humidity of the installation location of escalator 11, the travel distance of moving handrail 12, the number of days since escalator 11 was installed, and the loads that moving handrail 12 receives from pressure roller 13 and drive roller 14.

[0047] The estimation model 16 is generated by performing machine learning training using training data that correlates data on the tensile strength of the synthetic resin 22 measured for multiple moving handrails 12 collected from the market with data collected from information on the temperature, humidity, number of days in installation, running distance, and loads received from the pressure roller 13 and drive roller 14 at the installation location of the escalator 11.

[0048] The training data does not necessarily need to be data measured on multiple moving handrails 12 collected from the market. For example, the training data may be data on the tensile strength of the synthetic resin 22 measured on moving handrails 12 that have been subjected to accelerated deterioration in a high-temperature, high-humidity chamber or on moving handrails 12 that have been subjected to accelerated fatigue by repeated bending using a fatigue testing machine or the like. The training data may include both data measured on moving handrails 12 collected from the market and data measured on moving handrails 12 that have been subjected to accelerated deterioration using a testing machine or the like. Note that there are no particular limitations on the measurement method, measurement device, or measurement conditions for the tensile strength of the synthetic resin 22, but it is preferable to use the tensile strength of synthetic resin 22 evaluated using the same measurement method, measurement device, and measurement conditions when generating the estimation model 16.

[0049] For example, a thermoplastic elastomer or vulcanized rubber is used for the synthetic resin 22 used in the moving handrail 12. Thermoplastic elastomers include polyurethane elastomers, polyamide elastomers, polyolefin elastomers, polystyrene elastomers, and polyester elastomers, but polyurethane elastomers are preferred from the standpoint of abrasion resistance and strength, and among polyurethane elastomers, ether-based polyurethane elastomers are more preferred from the standpoint of hydrolysis resistance.

[0050] When a thermoplastic elastomer is used for the synthetic resin 22 of the moving handrail 12, the melt flow rate of the thermoplastic elastomer may be used as a physical property value of the synthetic resin 22 estimated by the calculation unit 15. The melt flow rate is an indicator of the fluidity of thermoplastic synthetic resins, including thermoplastic elastomers, when molten, and is affected by the length of the synthetic resin's molecular chain. Therefore, since the molecular chains of thermoplastic elastomers are broken during commercial use due to fatigue caused by oxidation reactions, hydrolysis reactions, repeated bending, and loads from the pressure roller 13, the melt flow rate of the synthetic resin 22 can be used as an indicator of the degree of deterioration of the moving handrail 12.

[0051] The melt flow rate of synthetic resin 22 deteriorates during use in the market due to oxidation reactions and hydrolysis caused by temperature and humidity, and also changes due to fatigue caused by repeated bending and loads from pressure roller 13, etc. Therefore, calculation unit 15 can estimate the melt flow rate of synthetic resin 22 using estimation model 16, using as input information the temperature and humidity of the installation location of escalator 11, the travel distance of moving handrail 12, the number of days since escalator 11 was installed, and the loads that moving handrail 12 receives from pressure roller 13 and drive roller 14.

[0052] Furthermore, the threshold value 134 used in the judgment unit 17 is set in advance based on the correlation between the melt flow rate of the synthetic resin 22 and the abnormality of the metal steel wire 23 protruding from the moving handrail 12, the occurrence of cracks in the synthetic resin 22, and breakage of the synthetic resin 22. As a result, the melt flow rate of the synthetic resin 22 of the moving handrail 12 used in the market can be estimated by the estimation model 16 trained by machine learning, and compared with the threshold value 134 to predict the time when the abnormality of the metal steel wire 23 protruding from the moving handrail 12, the occurrence of cracks in the synthetic resin 22, or breakage of the synthetic resin 22 will occur.

[0053] For example, vulcanized rubber may be used as the synthetic resin 22 used in the moving handrail 12. Examples of vulcanized rubber include natural rubber, butadiene rubber, styrene rubber, butyl rubber, chloroprene rubber, ethylene propylene rubber, ethylene propylene diene rubber, isoprene rubber, chlorosulfonated polyethylene rubber, urethane rubber, and nitrile rubber.

[0054] When vulcanized rubber is used for the synthetic resin 22 of the moving handrail 12, the hardness or crosslink density of the vulcanized rubber may be used as the physical property value of the synthetic resin 22 estimated by the calculation unit 15. Vulcanized rubber may harden due to the severing and recombination of molecular chains caused by heat during use in the market. Therefore, the hardness and crosslink density of vulcanized rubber change over time during use in the market, and can be used as indicators of the degree of deterioration of the moving handrail 12.

[0055] The hardness or crosslink density of vulcanized rubber deteriorates during use in the market due to oxidation reactions and hydrolysis caused by temperature and humidity, and also changes due to fatigue caused by repeated bending and loads from the pressure roller 13, etc. Therefore, the calculation unit 15 can estimate the hardness or crosslink density of the vulcanized rubber using the estimation model 16, using as input information the temperature and humidity of the installation location of the escalator 11, the travel distance of the moving handrail 12, the number of days since the escalator 11 was installed, and the loads that the moving handrail 12 receives from the pressure roller 13 and drive roller 14.

[0056] Furthermore, the threshold value 134 used in the judgment unit 17 is set in advance based on the correlation between the hardness or crosslink density of the vulcanized rubber and the abnormality of the metal steel wire 23 protruding from the moving handrail 12, the occurrence of cracks in the synthetic resin 22, and breakage of the synthetic resin 22. In this way, the hardness or crosslink density of the vulcanized rubber of the moving handrail 12 used in the market can be estimated by the estimation model 16 trained by machine learning, and compared with the threshold value 134 to predict the time when the abnormality of the metal steel wire 23 protruding from the moving handrail 12, the occurrence of cracks in the synthetic resin 22, or breakage of the synthetic resin 22 will occur.

[0057] The training data is prepared by correlating data on the melt flow rate of the synthetic resin or the hardness or crosslink density of the vulcanized rubber measured on multiple moving handrails 12 collected from the market with data collected on information related to the temperature, humidity, number of days installed, running distance, and loads from the pressure roller 13 and drive roller 14 at the installation location of the escalator 11. Note that the training data does not necessarily need to be data on the melt flow rate of the synthetic resin or the hardness or crosslink density of the vulcanized rubber measured on multiple moving handrails 12 collected from the market. For example, the training data may be data on the melt flow rate of the synthetic resin or the hardness or crosslink density of the vulcanized rubber measured on moving handrails 12 that have been subjected to accelerated deterioration in a high-temperature, high-humidity chamber, or data on moving handrails 12 that have been subjected to accelerated fatigue by repeated bending using a fatigue testing machine or the like. The training data may include both data measured on moving handrails 12 collected from the market and data measured on moving handrails 12 that have been subjected to accelerated deterioration using a testing machine or the like.

[0058] Embodiment 2 In the abnormality prediction device 10 according to the first embodiment, by inputting usage information relating to the use of the escalator 11 into the estimation model 16, it is possible to estimate at least one value from among the bond strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24. Also, in the first embodiment, it was explained that the usage information relating to the use of the escalator 11 includes information on the temperature and humidity of the installation location of the escalator 11, the travel distance of the moving handrail 12, the number of days since the escalator 11 was installed, and the load that the moving handrail 12 receives from the pressure roller 13 and the drive roller 14.

[0059] However, the usage information regarding the use of the escalator 11 is not limited to information regarding the temperature and humidity of the installation location of the escalator 11, the travel distance of the moving handrail 12, the number of days since the escalator 11 was installed, and the load that the moving handrail 12 receives from the pressure roller 13 and the drive roller 14. For example, if polyurethane elastomer is used for the synthetic resin 22 of the moving handrail 12, it is preferable to add the moisture content of the synthetic resin 22 (polyurethane elastomer) to the usage information regarding the use of the escalator 11.

[0060] Figure 4 is a schematic diagram showing the configuration of an escalator system 200 equipped with an abnormality prediction device 10a according to embodiment 2. In escalator system 200 shown in Figure 4, the same components as those in escalator system 100 shown in Figure 1 are designated by the same reference numerals, and detailed description thereof will not be repeated. Furthermore, moving handrail 12 shown in Figure 4 has the configuration shown in the cross-sectional view of Figure 2. Escalator system 200 according to embodiment 2 includes escalator 11 and abnormality prediction device 10a that predicts an abnormality in moving handrail 12 of escalator 11.

[0061] The polyurethane elastomer used in the synthetic resin 22 of the moving handrail 12 easily absorbs moisture, and the amount of moisture contained in the polyurethane elastomer affects the physical properties of the polyurethane elastomer. For example, if the moisture content of a polyurethane elastomer is high, the tensile strength decreases. Therefore, if a polyurethane elastomer is used for the moving handrail 12, a high moisture content may cause cracks or breakage in the moving handrail 12. Furthermore, the moisture in the polyurethane elastomer affects the adhesive strength at the interface between the elastomer and the metal steel wire 23. Therefore, if the moisture content of a polyurethane elastomer is high, the bonding strength between the synthetic resin 22 and the metal steel wire 23 decreases.

[0062] Therefore, information on the moisture content of the synthetic resin 22 is added to the training data for the estimation model 16. Furthermore, the abnormality prediction device 10a adds the information on the moisture content of the synthetic resin 22 to the information to be input into the trained estimation model 16. To acquire the moisture content data 18f of the synthetic resin 22, the abnormality prediction device 10a is equipped with a non-contact moisture content meter 19. The non-contact moisture content meter 19 is a device that can measure the moisture content of the synthetic resin 22 by irradiating the moving handrail 12 with, for example, near-infrared rays, infrared rays, or microwaves. Note that the moisture content of the synthetic resin 22 may be measured by providing the moisture content meter 19 even if polyurethane elastomer is not used for the synthetic resin 22.

[0063] By inputting information into estimation model 16, such as the temperature and humidity of the installation location of escalator 11, the travel distance of moving handrail 12, the number of days since escalator 11 was installed, information on the load that moving handrail 12 receives from pressure roller 13 and drive roller 14, and information on the moisture content of synthetic resin 22, calculation unit 15 can estimate with high accuracy at least one value among the bonding strength between synthetic resin 22 and metal steel wire 23, the physical property values ​​of synthetic resin 22, and the physical property values ​​of canvas 24. Note that the hardware configuration of abnormality prediction device 10a is the same as the hardware configuration of abnormality prediction device 10 shown in FIG. 3, and therefore detailed description thereof will not be repeated.

[0064] The moisture meter 19 is not limited to a non-contact type, and may be a contact type moisture meter that measures the moisture percentage by contacting the moving handrail 12. When a contact type moisture meter 19 is used, the moisture percentage of the synthetic resin 22 is measured by pressing the contact type moisture meter 19 against the moving handrail 12 when the escalator 11 is stopped. Therefore, the escalator system 200 needs to be provided with a mechanism that presses the contact type moisture meter 19 against the moving handrail 12 when the escalator 11 is temporarily stopped in order to measure the moisture percentage of the synthetic resin 22. Alternatively, an operator may measure the moisture percentage of the synthetic resin 22 with the contact type moisture meter during periodic inspection and input the result to the calculation unit 15.

[0065] Embodiment 3. In the abnormality prediction device 10 according to embodiment 1, it has been described that the temperature data 18a, humidity data 18b, mileage data 18c, number of days in installation data 18d, and load data 18e from the pressure roller 13 and drive roller 14 are acquired from a management device (not shown) for the escalator 11. However, this is not limiting, and the temperature data, humidity data, mileage data, number of days in installation data, and load data from the pressure roller and drive roller may be measured and managed by the abnormality prediction device.

[0066] Figure 5 is a schematic diagram showing the configuration of an escalator system 300 equipped with an abnormality prediction device 10b according to embodiment 3. In escalator system 300 shown in Figure 5, the same components as those in escalator system 100 shown in Figure 1 are designated by the same reference numerals, and detailed description thereof will not be repeated. Furthermore, moving handrail 12 shown in Figure 5 has the configuration shown in the cross-sectional view of Figure 2. Escalator system 300 according to embodiment 3 includes escalator 11 and abnormality prediction device 10b that predicts an abnormality in moving handrail 12 of escalator 11.

[0067] The abnormality prediction device 10b acquires temperature data 18a and humidity data 18b directly from a thermometer and hygrometer (not shown) that measure the temperature and humidity at the installation location of the escalator 11. Furthermore, the abnormality prediction device 10b calculates the travel distance of the moving handrail 12 from the drive time of the drive roller 14, thereby acquiring travel distance data 183. The abnormality prediction device 10b also stores the installation date of the escalator 11, calculates the number of days since the installation date, and acquires installation day number data 184. Furthermore, the abnormality prediction device 10b stores information on the pressing forces of the pressure roller 13 and drive roller 14 that were adjusted at the time of installation, and acquires load data 185 from the pressure roller 13 and drive roller 14 from the information. In other words, the abnormality prediction device 10b measures and manages the temperature data 181, humidity data 182, travel distance data 183, installation day number data 184, and load data 185 from the pressure roller 13 and drive roller 14.

[0068] The calculation unit 15 inputs temperature data 181, humidity data 182, mileage data 183, installation days data 184, and load data 185 from the pressure roller 13 and drive roller 14 within the abnormality prediction device 10b into the estimation model 16, and estimates at least one value of the bonding strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24.

[0069] The abnormality prediction device 10b stores temperature data 181, humidity data 182, mileage data 183, installation days data 184, and load data 185 from the pressure roller 13 and drive roller 14 in the storage 130 as necessary.

[0070] Embodiment 4 In the abnormality prediction device 10 according to embodiment 1, it has been explained that the management device (not shown) for the escalator 11 acquires the temperature data 18a and the humidity data 18b from a thermometer and a hygrometer (not shown) that measure the temperature and humidity at the installation location of the escalator 11. However, this is not limiting, and the temperature data may be acquired from a thermometer and the humidity data may be acquired from a hygrometer installed near the installation location of the escalator.

[0071] Figure 6 is a schematic diagram showing the configuration of an escalator system 400 equipped with an abnormality prediction device 10c according to embodiment 4. In escalator system 400 shown in Figure 6, the same components as those in escalator system 100 shown in Figure 1 are designated by the same reference numerals, and detailed description thereof will not be repeated. Furthermore, moving handrail 12 shown in Figure 6 has the configuration shown in the cross-sectional view of Figure 2. Escalator system 400 according to embodiment 4 includes escalator 11 and abnormality prediction device 10c that predicts an abnormality in moving handrail 12 of escalator 11.

[0072] The abnormality prediction device 10c includes a temperature data receiving device 90 that receives temperature data 18a from a thermometer (not shown) installed near the installation location of the escalator, and a humidity data receiving device 91 that receives humidity data 18b from a hygrometer (not shown). By including the temperature data receiving device 90 and the humidity data receiving device 91 in the abnormality prediction device 10c, there is no need to provide a thermometer and a hygrometer on the escalator 11 itself.

[0073] The calculation unit 15 inputs the temperature data 18a and humidity data 18b received by the temperature data receiving device 90 and humidity data receiving device 91, as well as the mileage data 18c, the number of days of installation data 18d, and the load data 18e from the pressure roller 13 and the drive roller 14 into the estimation model 16, thereby estimating at least one value of the bonding strength between the synthetic resin 22 and the metal steel wire 23, the physical property values ​​of the synthetic resin 22, and the physical property values ​​of the canvas 24.

[0074] The embodiments disclosed herein are intended to be combined appropriately within the scope of compatibility. The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present disclosure is defined by the claims, not the above description, and is intended to include all modifications within the meaning and scope of the claims.

[0075] 10, 10a to 10c: abnormality prediction device; 11: escalator; 12: moving handrail; 13: pressure roller; 14: drive roller; 15: calculation unit; 16: estimation model; 17: judgment unit; 18a, 181: temperature data; 18b, 182: humidity data; 18c, 183: travel distance data; 18d, 184: installation days data; 18e, 185: load data; 18f: moisture content data; 19: moisture content meter; 22: synthetic resin; 23: metal steel wire; 24: canvas; 90: temperature data receiving device; 91: humidity data receiving device; 100, 200, 300, 400: escalator system.

Claims

1. An abnormality prediction device that predicts abnormalities in a moving handrail of an escalator, comprising: a data acquisition unit that acquires usage information regarding use of the escalator; a calculation unit that uses an estimation model that has been trained in advance by machine learning to infer characteristic values ​​of the moving handrail from the usage information of the escalator, and outputs the characteristic values ​​of the moving handrail from the usage information input from the data acquisition unit; and a judgment unit that determines that the moving handrail needs to be replaced if the characteristic value of the moving handrail output from the calculation unit falls below a predetermined threshold, wherein the characteristic values ​​of the moving handrail include at least one value of the bond strength between the synthetic resin and metal steel wire that make up the moving handrail, the physical property values ​​of the synthetic resin that makes up the moving handrail, and the physical property values ​​of the canvas that makes up the moving handrail.

2. The abnormality prediction device according to claim 1, wherein the joint strength is the pull-out strength of the metal steel wire of the moving handrail or the pull-out strength of a central wire of the metal steel wire.

3. The abnormality prediction device according to claim 1, wherein the physical property value of the canvas is the dynamic friction coefficient or static friction coefficient of the canvas provided on the back surface of the moving handrail.

4. The abnormality prediction device according to claim 1, wherein the physical property value of the synthetic resin is the tensile strength of the synthetic resin that constitutes the handrail.

5. The abnormality prediction device according to claim 1, wherein the synthetic resin constituting the moving handrail is a thermoplastic elastomer, and the physical property value of the synthetic resin is the melt flow rate of the thermoplastic elastomer.

6. The abnormality prediction device according to claim 1, wherein the synthetic resin constituting the moving handrail is vulcanized rubber, and the physical property value of the synthetic resin is the hardness of the vulcanized rubber or the crosslink density of the vulcanized rubber.

7. The abnormality prediction device according to claim 1, wherein the usage information of the escalator includes at least one of the following information: temperature and humidity at the installation location of the escalator, the distance traveled by the moving handrail, the number of days the escalator has been installed, and information on the load received by the moving handrail from the pressure roller and drive roller.

8. The abnormality prediction device according to claim 7, further comprising a recording device that records at least one of the following information: temperature and humidity at the installation location of the escalator, the distance traveled by the moving handrail, the number of days the escalator has been installed, and the load received by the moving handrail from the pressure roller and the drive roller.

9. An abnormality prediction device as described in claim 7 or claim 8, wherein the synthetic resin constituting the moving handrail is polyurethane elastomer, and the usage information of the escalator further includes information on the moisture content of the synthetic resin.

10. The abnormality prediction device according to claim 9, further comprising a non-contact moisture content meter that measures the moisture content of the moving handrail by irradiating it with near-infrared rays, infrared rays, or microwaves, or a contact moisture content meter that measures the moisture content by contacting the moving handrail.

11. The abnormality prediction device according to claim 7, further comprising a receiving device for receiving measurement data from a thermometer and a hygrometer installed near the location where the escalator is installed.

12. An escalator system comprising: the escalator; and the abnormality prediction device according to claim 1.

Images