Inspection device, inspection method, and inspection program for forklift lift chains

Abstract

By understanding the condition of the lift chain and determining the degree of breakage, it becomes possible to inspect the lift chain at the appropriate time. [Solution] The inspection device 1 includes a learning model generation unit 101 that generates a learning model by machine learning by performing machine learning on a dataset based on the relationship between state data relating to the total driving time of the lift chain 42 of the forklift 2 and a fracture score relating to the degree of fracture of the lift chain 42 of the forklift 2, using the dataset 106 as training data; an acquisition unit 105 that acquires state data of the lift chain 42 of the forklift 2; and a prediction unit 103 that acquires a fracture score from the learning model by inputting the state data acquired from the acquisition unit 105 into the learning model generated by the learning model generation unit 101.

Description

Technical Field

[0001] The present invention relates to an inspection device, an inspection method, and an inspection program for inspecting a lift chain of a forklift.

Background Art

[0002] As is well known, a forklift is configured to lift and transport a pallet on which a load is placed with forks. The forklift is configured to lift and lower a lift bracket provided with forks via a lift chain by lifting and lowering a mast with a lift cylinder.

[0003] By the way, since the breakage of the lift chain directly leads to a load drop accident, it is very important to inspect the lift chain. Here, if the inspection frequency is increased, the abnormality of the lift chain can be quickly detected and it is safe. However, on the other hand, there is a problem that the more the inspection frequency increases, the more the cost required for inspection increases.

[0004] As a prior art, there is a chain detector that detects that a lift chain has extended by a predetermined amount (for example, Patent Document 1). However, with this chain detector, although it can detect that the lift chain has extended by a predetermined amount, it is impossible to know the risk of breakage (breakage degree) of the lift chain from the state of the lift chain.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] Therefore, the problem that this invention aims to solve is to enable the lift chain to be inspected at an appropriate time by knowing the degree of risk of breakage (degree of breakage) of the lift chain from its condition. [Means for solving the problem]

[0007] To solve the above problems, the inspection device according to the present invention is An inspection device for the lift chain of a forklift, A learning model generation unit generates a learning model by performing machine learning on a dataset based on the relationship between state data regarding the total drive time of a forklift's lift chain and a fracture score regarding the degree of breakage of the forklift's lift chain, using the dataset as training data. A data acquisition unit that acquires the status data of the forklift's lift chain, The system includes a prediction unit that obtains a fracture degree score from a learning model by inputting state data obtained from an acquisition unit into a learning model generated by a learning model generation unit.

[0008] Preferably, The inspection device is equipped with a total drive time detection sensor that detects the total drive time of the forklift's lift chain. The acquisition unit is configured to acquire status data based on the total operating time detected by the total operating time detection sensor.

[0009] Preferably, The acquisition unit is configured to acquire state data each time the lift chain is driven.

[0010] Preferably, The inspection device includes a control unit that, when the fracture degree score obtained by the prediction unit exceeds a preset first fracture degree level, guides the user to inspect the lift chain within a predetermined period, and further, when it exceeds a preset second fracture degree level, guides the user to inspect the lift chain immediately.

[0011] Furthermore, the inspection method according to the present invention is A method for inspecting the lift chain of a forklift, Computers Using a dataset based on the relationship between state data regarding the total drive time of a forklift's lift chain and a fracture score regarding the degree of breakage of the forklift's lift chain as training data, machine learning is performed on the training data, and a learning model is generated through machine learning. Obtain the condition data of the forklift's lift chain, The generated learning model is then used to input the acquired state data, and the process of obtaining a fracture score from the learning model is executed.

[0012] Furthermore, the inspection program according to the present invention, An inspection program for the lift chain of a forklift, On the computer, Using a dataset based on the relationship between state data regarding the total drive time of a forklift's lift chain and a fracture score regarding the degree of breakage of the forklift's lift chain as training data, machine learning is performed on the training data, and a learning model is generated through machine learning. Obtain the condition data of the forklift's lift chain, The generated learning model is then used to input the acquired state data, and the process of obtaining a fracture score from the learning model is executed. [Effects of the Invention]

[0013] According to the present invention, by knowing the degree of risk of lift chain breakage (degree of breakage) from the condition of the lift chain, the lift chain can be inspected at an appropriate time. [Brief explanation of the drawing]

[0014] [Figure 1] The image shows a forklift, with (A) being a top view and (B) being a side view. [Figure 2]Side view showing a forklift with the lift cylinder extended. [Figure 3] Block diagram showing the functional configuration of a computer. [Figure 4] Showing the screen of the display unit, where (A) is the first guidance screen and (B) is the second guidance screen. [Figure 5] Flowchart diagram for explaining the inspection program and inspection method. [Figure 6] Side view showing a forklift for explaining the second embodiment. [Figure 7] Side view showing a forklift for explaining the third embodiment. [Figure 8] Side view showing a forklift for explaining the fourth embodiment. [Figure 9] Side view showing a forklift for explaining the fifth embodiment.

Embodiments for Carrying out the Invention

[0015] Hereinafter, based on the drawings, embodiments of an inspection device, inspection method, and inspection program for a lift chain of a forklift according to the present invention will be described.

[0016] <First Embodiment> <Forklift 2> As shown in FIGS. 1 and 2, the forklift 2 has a pair of left and right outer mast 40 and inner mast 41 provided in the vertical direction Z in front of the vehicle body 21. Each mast 40, 41 is arranged at intervals in the left - right direction Y of the vehicle body 21. And the outer mast 40 and the inner mast 41 are configured to be able to slide in a stacked state.

[0017] The forklift 2 includes a lift bracket 51, and the lift bracket 51 is supported so as to be able to move up and down in the vertical direction Z of the vehicle body 21 along the inner mast 41.

[0018] The forklift 2 is equipped with a pair of left and right forks 52 for inserting into pallets on which loads are placed. The forks 52 are mounted on the left and right sides of the lift bracket 51 and are spaced apart in the left-right direction Y of the vehicle body 21. Therefore, the forks 52 move up and down together with the lift bracket 51.

[0019] The forklift 2 has a pair of lift cylinders 44 mounted on the front of the vehicle body 21, positioned vertically in the Z direction. The lift cylinders 44 are arranged parallel to the masts 40 and 41 and are connected to the outer mast 40. Furthermore, as shown in Figure 2, the upper ends of the lift cylinders 44 are connected to the inner mast 41, so when the lift cylinders 44 extend, the inner mast 41 rises, and as shown in Figure 1, when the lift cylinders 44 retract, the inner mast 41 lowers.

[0020] The forklift 2 is equipped with a chain wheel 43 located on the upper part of the inner mast 41 and a lift chain 42 attached to the chain wheel 43. One end 42a of the lift chain 42 is connected to the outer mast 40, and the other end 42b is connected to the lift bracket 51.

[0021] As shown in Figure 1, when the lift cylinder 44 retracts, the inner mast 41 overlaps with the outer mast 40 and is positioned downward, and the other end 42b of the lift chain 42 is positioned downward, so the lift bracket 51 and fork 52 are positioned downward.

[0022] As shown in Figure 2, when the lift cylinder 44 extends, the inner mast 41 protrudes upward from the outer mast 40, and the other end 42b of the lift chain 42 is positioned upward, so that the lift bracket 51 and fork 52 are positioned upward.

[0023] Furthermore, the forklift 2 is equipped with a pair of straddle arms 53 on the left and right sides of the vehicle body 21. Each straddle arm 53 is spaced apart in the left-right direction Y of the vehicle body. The straddle arms 53 are configured to guide the masts 40 and 41 to move back and forth in the front-rear direction X of the vehicle body 21.

[0024] Forklift 2 has a driver's seat 20 at the rear of the vehicle body 21. A brake pedal 50 is located below the driver's seat 20, and the operator applies and releases the brakes by operating the brake pedal 50 with their foot. Forklift 2 has an operating unit 33 in front of the driver's seat 20.

[0025] The control unit 33 is equipped with multiple hydraulic levers 54, and by operating the hydraulic levers 54, the operator drives the lift cylinder 44, tilt cylinder and reach cylinder (not shown) to raise, lower, tilt and move the fork 52 forward and backward.

[0026] The control unit 33 is equipped with an accelerator lever 55, which the operator can use to move the vehicle body 21 forward or backward by tilting the accelerator lever 55. The forklift 2 is equipped with a steering wheel 56 on the left side of the driving space 20, which the operator can use to change the direction of travel of the vehicle body 21 by operating the steering wheel 56.

[0027] The forklift 2 includes a head guard 57 that covers the area above the operating space 20 to protect the operator from falling objects, a front wheel 58 located at the front of the straddle arm 53, and a rear wheel 59 located at the rear of the vehicle body 21. The rear wheel 59 is connected to a travel motor and a slewing motor (not shown) and serves as the drive wheel for moving the vehicle body 21 forward and backward and turning.

[0028] <Inspection device 1> As shown in Figure 2, the forklift 2 is equipped with an inspection device 1 for inspecting the lift chain 42. The inspection device 1 comprises a computer 10 mounted on the vehicle body 21 and a total drive time detection sensor 11 provided on the chain wheel 43.

[0029] Computer 10 consists of a control unit, an arithmetic unit, a main memory unit, an auxiliary memory unit, a communication unit, and the like. The control unit and arithmetic unit include a CPU or MPU and perform various calculations and control operations. The main memory unit includes RAM (DRAM) and ROM and stores programs and various data. The auxiliary memory unit includes an HDD, SSD, flash memory, EEPROM, and the like and stores programs and various data. The communication unit includes a wired or wireless communication circuit and communicates data with the total drive time detection sensor 11, etc. Note that computer 10 may not be mounted on the vehicle body 21 and may be a server computer or the like located away from the forklift 2.

[0030] The total drive time detection sensor 11 is composed of, for example, a rotation speed sensor attached to the chain wheel 43. Since the chain wheel 43 rotates at a constant speed, the total drive time of the lift chain 42 (the cumulative time the chain wheel 43 is driven) can be detected by measuring the total number of rotations of the chain wheel 43 with the rotation speed sensor.

[0031] The computer 10 is connected to the hydraulic lever 54 and is configured to acquire the total drive time of the lift chain 42 using the total drive time detection sensor 11 each time the hydraulic lever 54, which operates the lift cylinder 44, is driven. Therefore, the computer 10 can acquire state data regarding the state of the total drive time of the lift chain 42 each time the lift chain 42 is driven together with the lift cylinder 44.

[0032] As shown in Figure 3, the computer 10 includes a data collection unit 100 for collecting training data 106. The training data 106 includes state data relating to the total driving time of the lift chain 42. The state data relating to the total driving time of the lift chain 42 includes the total driving time of the lift chain 42 as measured by the total driving time detection sensor 11.

[0033] The computer 10 includes a learning model generation unit 101 that performs machine learning on training data 106 collected by the collection unit 100, and generates and stores a learning model through machine learning. In this embodiment, the learning model generation unit 101 performs supervised learning. In supervised learning, a large amount of training data 106, that is, pairs of input data ID and output data OD, is input to the learning model generation unit 101.

[0034] The input data ID contains the total drive time of the lift chain 42. The output data OD is the fracture degree score S. The fracture degree score S is determined by evaluating the input data ID, and a numerical parameter from 0 to 10 is set as the fracture degree score S, which is used to determine when to inspect the lift chain 42 based on its fracture degree.

[0035] For example, if the fracture score S is high, meaning the degree of fracture of the lift chain 42 is high and it is time to inspect it, and the total operating time of the lift chain 42 is long, then the degree of fracture (deterioration, fatigue) of the lift chain 42 is advanced, and it is time to inspect the lift chain 42.

[0036] On the other hand, if the fracture score S is low, that is, if the degree of fracture of the lift chain 42 is low and the inspection time is far off, then the total driving time of the lift chain 42 is short, and the degree of fracture (deterioration, fatigue) of the lift chain 42 has not progressed, so the inspection time for the lift chain 42 is far off.

[0037] In fact, the degree of breakage of the lift chain 42 can be determined by the total operating time of the lift chain 42. Therefore, it can be inferred that there is a certain relationship, such as a correlation, between the total operating time of the lift chain 42 and the timing of inspection of the lift chain 42.

[0038] The learning model generation unit 101 uses a general machine learning algorithm such as a neural network. The learning model generation unit 101 performs machine learning using correlated input data ID and output data OD as training data 106 to generate a model that estimates output from input (learning model), that is, a model that takes input data ID consisting of state data related to the total driving time of the lift chain 42 as input and outputs a fracture score S related to the fracture degree of the lift chain 42.

[0039] The computer 10 includes an acquisition unit 105 that acquires an input data ID. In this embodiment, the acquisition unit 105 is connected to the total drive time detection sensor 11. As described above, the input data ID includes the total drive time of the lift chain 42. The input data ID is acquired by the acquisition unit 105 each time the lift chain 42 is driven, or at predetermined intervals.

[0040] The computer 10 includes a prediction unit 102 that predicts the fracture score S of the lift chain 42 by applying the learning model generated by the learning model generation unit 101 to the input data ID acquired from the acquisition unit 105.

[0041] To determine when to inspect the lift chain 42, a first fracture level L1 (medium urgency for inspection) and a second fracture level L2 (high urgency for inspection) are set and stored in the computer 10.

[0042] The computer 10 includes a control unit 103 that can determine whether the fracture score S is equal to or greater than the first and second fracture levels L1 and L2. If the lift chain 42 is determined to be equal to the first fracture level L1 or higher (for example, fracture score S = 5 or higher), it may break if it is not inspected within a predetermined period (for example, within one month if the lift chain 42 is used at a normal frequency). Also, if it is determined to be equal to the second fracture level L2 or higher (for example, fracture score S = 8 or higher), it may break if the lift chain 42 is not inspected urgently (for example, within one week if the lift chain 42 is used at a normal frequency).

[0043] If the acquired fracture score S meets the condition of being at or above the first fracture level L1, it is determined that the lift chain 42 should be inspected within a specified period. If the acquired fracture score S meets the condition of being at or above the second fracture level L2, it is determined that the lift chain 42 should be inspected immediately.

[0044] The forklift 2 is equipped with a display unit 60 in the control panel 33 of the operating space 20 (Figure 1(A)), and the display unit 60 is connected to the computer 10.

[0045] As shown in Figure 4(A), when the acquired fracture score S satisfies the condition that it is equal to or greater than the first fracture level L1, the computer 10 displays a first guidance screen on the display unit 60 via the control unit 103. The first guidance screen displays instructions such as, "Please inspect this lift chain within one month. Do not use this lift chain for more than one month," guiding the user to inspect the lift chain 42 within a predetermined period.

[0046] As shown in Figure 4(B), if the acquired fracture score S satisfies the condition that it is at or above the second fracture level L2, the computer 10 displays a second guidance screen on the display unit 60 via the control unit 103. The second guidance screen displays a message such as, "Please inspect this lift chain immediately. Do not use this lift chain," instructing the user to inspect the lift chain 42 immediately.

[0047] By following the guidance screen displayed on the display unit 60, the operator can inspect the lift chain 42 at the appropriate time. This allows the operator to determine the degree of breakage of the lift chain 42 from its condition, enabling timely inspection. As a result, the lift chain 42 can be inspected at an appropriate frequency, minimizing the costs associated with inspections.

[0048] <Inspection Program and Inspection Methods> As shown in Figure 5, computer 10 executes the following inspection program and inspection method.

[0049] The data collection unit 100 collects training data 106 (collection step: S1). Then, the learning model generation unit 101 performs machine learning on the training data 106 collected by the data collection unit 100 in the collection step S1, and generates and stores a learning model through machine learning (learning model generation step: S2). The acquisition unit 105 acquires the input data ID (acquisition step: S3).

[0050] The prediction unit 102 predicts the fracture degree score S (inspection timing) by applying the learning model generated in the learning model generation step S2 to the input data ID acquired in the acquisition step S3 (prediction step: S4). The control unit 103 controls the display unit 60 to display a guidance screen based on the output data OD predicted in the prediction step S4 (control step: S5).

[0051] Although preferred embodiments of the present invention have been described above, the configuration of the present invention is not limited to these embodiments. For example, the total drive time detection sensor 11 is not limited to a rotation speed sensor attached to the chain wheel 43, but may also consist of an optical sensor or a magnetic sensor attached to the inner mast 41, and the total drive time may be detected by detecting the passage of the rotating part of the lift chain 42. Alternatively, the total drive time detection sensor 11 may consist of a camera attached to the inner mast 41, and the total drive time may be detected by changes in the image of the lift chain 42 captured by the camera. Alternatively, the total drive time detection sensor 11 may consist of an acceleration sensor attached to the lift chain 42, and the total drive time may be detected by the change in the acceleration of the lift chain 42.

[0052] <Second Embodiment> Next, a second embodiment will be described. Note that parts that overlap with the first embodiment may be omitted from the explanation.

[0053] The inspection device 1 may further include a sagging detection sensor 1101. As shown in Figure 6, the sag detection sensor 1101 consists of a first detection sensor 11101 located at the top of the inner mast 41, a second detection sensor 11201 located in the center of the inner mast 41, and a third detection sensor 11301 located at the bottom of the inner mast 41.

[0054] The sag detection sensor 1101 is configured, for example, as a non-contact distance sensor, and is set up to detect the amount of sag in the lift chain 42 by measuring the distance from a set position to the lift chain 42 using laser light.

[0055] The slack detection sensor 1101 is configured to detect the amount of slack in the lift chain 42 based on the difference between a set distance A from a preset position to the lift chain 42 and a measured distance B from the preset position to the lift chain 42'. In other words, when the lift chain 42 sags, the difference between the set distance A and the measured distance B increases, and the amount of slack can be detected from this difference.

[0056] The computer 10 is connected to the hydraulic lever 54 and is configured to acquire the amount of slack in the lift chain 42 from the slack detection sensor 1101 each time the hydraulic lever 54, which operates the lift cylinder 44, is driven. Therefore, the computer 10 can acquire state data regarding the slack state of the lift chain 42 each time the lift chain 42 is driven together with the lift cylinder 44.

[0057] The training data 106 includes state data regarding the slack state of the lift chain 42. The state data regarding the slack state of the lift chain 42 includes the amount of slack of the lift chain 42 as measured by the slack detection sensor 1101.

[0058] The input data ID contains the amount of slack in the lift chain 42. The output data OD is the fracture degree score S.

[0059] For example, if the fracture score S is high, meaning the degree of fracture of the lift chain 42 is high and it is time for inspection, and the amount of slack in the lift chain 42 is large, then the degree of fracture (deterioration, fatigue) of the lift chain 42 is advanced, and it is time for inspection of the lift chain 42.

[0060] On the other hand, if the fracture score S is low, that is, if the degree of fracture of the lift chain 42 is low and the inspection time is far off, then when the amount of slack in the lift chain 42 is small, the degree of fracture (deterioration, fatigue) of the lift chain 42 has not progressed, so the inspection time for the lift chain 42 is far off.

[0061] The fracture score S may be set using any of the numerical parameters of the state data obtained from the amount of slack in the lift chain 42 detected by the first sensor 11101, the amount of slack in the lift chain 42 detected by the second sensor 11201, or the amount of slack in the lift chain 42 detected by the third sensor 11301, or it may be set using numerical parameters weighted by a weighting coefficient.

[0062] If the amount of slack in the lift chain 42 exceeds a predetermined amount at multiple locations, rather than just one, the degree of breakage of the lift chain 42 is likely to increase. Also, if the amount of slack in the upper part of the lift chain 42 is greater than in other parts, the likelihood of it detaching from the chain wheel 43 and breaking increases, so the weighting coefficient for the upper part may be set to be larger than that for the amount of slack in other parts.

[0063] Furthermore, the degree of breakage of the lift chain 42 can actually be determined by the amount of slack in the lift chain 42. Therefore, it can be inferred that there is a certain relationship, such as a correlation, between the amount of slack in the lift chain 42 and the timing of inspection of the lift chain 42.

[0064] The acquisition unit 105 is connected to the slack detection sensor 1101. As described above, the input data ID includes the amount of slack in the lift chain 42. The input data ID is acquired by the acquisition unit 105 each time the lift chain 42 is driven, or at predetermined intervals.

[0065] The computer 10 predicts the fracture score S of the lift chain 42 by applying the learning model generated by the learning model generation unit 101 to the input data ID acquired from the acquisition unit 105.

[0066] For example, the slack detection sensor 1101 consists of multiple non-contact distance sensors, but it may also be a single non-contact distance sensor. By providing multiple sensors, the amount of slack in the lift chain 42 can be detected in more detail. Alternatively, the sagging detection sensor 1101 may be a CCD camera, and the amount of sagging may be detected by comparing a pre-stored image of a lift chain 42 without sagging with an image of the current lift chain 42. Alternatively, the slack detection sensor 1101 may be a contact-type gauge, and the amount of slack may be detected by applying the gauge to the lift chain 42. Furthermore, the slack in the lift chain 42 may be measured not only by the amount of slack, but also by the angle of the slack portion of the lift chain 42, etc.

[0067] <Third Embodiment> Next, a third embodiment will be described. Note that parts that overlap with the first embodiment may be omitted from the explanation.

[0068] The inspection device 1 may further include a rust detection sensor 1102. As shown in Figure 7, the rust detection sensor 1102 consists of a first detection sensor 11102 located at the top of the inner mast 41, a second detection sensor 11202 located in the center of the inner mast 41, and a third detection sensor 11302 located at the bottom of the inner mast 41.

[0069] The rust detection sensor 1102 is configured, for example, as a CCD (charge-coupled device) camera, and is set up to detect the amount of rust on the lift chain 42 by capturing images of the lift chain 42 with the CCD camera and performing image analysis.

[0070] The rust detection sensor 1102 measures the amount of rust (area of ​​rust) in a predetermined area of ​​the lift chain 42 by illuminating the lift chain 42 with a light and taking a picture of the captured image, and then analyzing the captured image. More specifically, the rust detection sensor 1102 is configured to enhance the characteristics of rust by removing noise from the captured image and adjusting the contrast through image preprocessing, then to detect the boundaries of the rust using an edge detection algorithm, and further to identify the rusted areas using a region segmentation algorithm and to measure the area of ​​rust by counting the number of rust pixels.

[0071] The computer 10 is connected to the hydraulic lever 54 and is configured to acquire the amount of rust on the lift chain 42 from the rust detection sensor 1102 each time the hydraulic lever 54, which operates the lift cylinder 44, is driven. Therefore, the computer 10 can acquire state data regarding the rust condition of the lift chain 42 each time the lift chain 42 is driven together with the lift cylinder 44.

[0072] The training data 106 includes state data regarding the rust condition of the lift chain 42. The state data regarding the rust condition of the lift chain 42 includes the amount of rust on the lift chain 42 as measured by the rust detection sensor 1102.

[0073] The input data ID contains the amount of rust on the lift chain 42. The output data OD is the fracture degree score S.

[0074] For example, if the fracture score S is high, meaning the degree of fracture of the lift chain 42 is high and it is time for inspection, and if the amount of rust on the lift chain 42 is large, it means that the degree of fracture (deterioration, fatigue) of the lift chain 42 is advanced, and it is time for inspection of the lift chain 42.

[0075] On the other hand, if the fracture score S is low, that is, if the degree of fracture of the lift chain 42 is low and the inspection time is far off, then if the amount of rust on the lift chain 42 is small, the degree of fracture (deterioration, fatigue) of the lift chain 42 has not progressed, and therefore the inspection time for the lift chain 42 is far off.

[0076] The fracture score S may be set using any of the numerical parameters of the condition data obtained from the amount of rust on the lift chain 42 detected by the first sensor 11102, the amount of rust on the lift chain 42 detected by the second sensor 11202, or the amount of rust on the lift chain 42 detected by the third sensor 11302, or it may be set using numerical parameters weighted by a weighting coefficient.

[0077] If the amount of rust on the lift chain 42 exceeds a predetermined amount in multiple locations rather than just one, it is thought that the degree of breakage of the lift chain 42 will increase. Also, if the amount of rust on the upper part of the lift chain 42 is greater than in other parts, the likelihood of it detaching from the chain wheel 43 and breaking increases, so the weighting coefficient for the upper part may be set to be larger than that for the amount of rust in other parts.

[0078] Furthermore, the degree of breakage of the lift chain 42 can actually be determined by the amount of rust on the lift chain 42. Therefore, it can be inferred that there is a certain relationship, such as a correlation, between the amount of rust on the lift chain 42 and the timing of the lift chain 42 inspection.

[0079] The acquisition unit 105 is connected to the rust detection sensor 1102. As described above, the input data ID includes the amount of rust on the lift chain 42. The input data ID is acquired by the acquisition unit 105 each time the lift chain 42 is driven, or at predetermined intervals.

[0080] The computer 10 predicts the fracture score S of the lift chain 42 by applying the learning model generated by the learning model generation unit 101 to the input data ID acquired from the acquisition unit 105.

[0081] For example, the rust condition of the lift chain 42 may be determined not only by the amount of rust (area of ​​rust), but also by a combination of rust intensity and rust pattern. Rust intensity may be determined by identifying the color of the rust from images captured by the rust detection sensor 1102 and measuring the intensity from the color. Rust pattern may be determined by identifying the arrangement of rust from images captured by the rust detection sensor 1102 and selecting a rust pattern. Alternatively, weighting coefficients may be set for the area, intensity, and pattern of the rust, and the state of the rust may be defined as a numerical parameter based on these weighting coefficients. Furthermore, the rust detection sensor 1102 does not have to be a CCD camera; for example, it may be an infrared camera, ultraviolet camera, X-ray camera, potentiometer, current meter, etc.

[0082] <Fourth Embodiment> Next, a fourth embodiment will be described. Note that parts that overlap with the first embodiment may be omitted from the explanation.

[0083] The inspection device 1 may further include a total volume detection sensor 1104. As shown in Figure 8, the total volume detection sensor 1104 is composed of, for example, a load sensor (load cell) attached to the other end 42b of the lift chain 42. The load sensor measures the load of the cargo placed on the lift bracket 51 and the fork 52, thereby enabling the detection of the total volume load (cumulative weight of the loads applied to the lift chain 42).

[0084] The computer 10 is connected to the hydraulic lever 54 and is configured to acquire the total load on the lift chain 42 using the total load detection sensor 11 each time the hydraulic lever 54, which operates the lift cylinder 44, is driven. Therefore, the computer 10 can acquire state data regarding the state of the total load on the lift chain 42 each time the lift chain 42 is driven together with the lift cylinder 44.

[0085] The training data 106 includes state data regarding the state of the total load applied to the lift chain 42. The state data regarding the state of the total load applied to the lift chain 42 includes the total load applied to the lift chain 42 as detected by the total load detection sensor 1104.

[0086] The input data ID contains the total load applied to the lift chain 42. The output data OD is the fracture degree score S.

[0087] For example, if the fracture score S is high, meaning the degree of fracture of the lift chain 42 is high and it is time for inspection, this is because the total load on the lift chain 42 is large, indicating that the degree of fracture (deterioration, fatigue) of the lift chain 42 is advanced, and it is time for inspection of the lift chain 42.

[0088] On the other hand, if the fracture score S is low, that is, if the degree of fracture of the lift chain 42 is low and the inspection time is far off, then when the total volume load on the lift chain 42 is small, the degree of fracture (deterioration, fatigue) of the lift chain 42 has not progressed, and therefore the inspection time for the lift chain 42 is far off.

[0089] Furthermore, the degree of breakage of the lift chain 42 can be determined by the total load applied to it. Therefore, it can be inferred that there is a certain relationship, such as a correlation, between the total load applied to the lift chain 42 and the timing of inspection of the lift chain 42.

[0090] The acquisition unit 105 is connected to the total volume detection sensor 1104. As described above, the input data ID includes the total volume loaded onto the lift chain 42. The input data ID is acquired by the acquisition unit 105 each time the lift chain 42 is driven, or at predetermined intervals.

[0091] The computer 10 predicts the fracture score S of the lift chain 42 by applying the learning model generated by the learning model generation unit 101 to the input data ID acquired from the acquisition unit 105.

[0092] For example, the total volume detection sensor 1104 is not limited to a load sensor attached to the chain wheel 43, but may also consist of a balance weight sensor or a stress sensor. Alternatively, the total volume detection sensor 1104 may consist of a camera mounted on the lift bracket 51, which identifies the load from the captured image and detects its weight.

[0093] <Fifth Embodiment> Next, a fifth embodiment will be described. Note that parts that overlap with the first embodiment may be omitted from the description.

[0094] The inspection device 1 may further include a lifting / lowering count detection sensor 1105. As shown in Figure 9, the lifting / lowering count detection sensor 1105 is composed of, for example, a rotation sensor attached to the chain wheel 43. When the lift chain 42 moves up and down, the chain wheel 43 rotates, so the rotation sensor detects the start and stop of the rotation of the chain wheel 43, thereby enabling the detection of the number of times the lift chain 42 moves up and down (the cumulative number of times the chain wheel 43 starts rotating after it has stopped rotating).

[0095] The computer 10 is connected to the hydraulic lever 54 and is configured to acquire the number of times the lift chain 42 is raised and lowered by the lift count detection sensor 1105 each time the hydraulic lever 54, which operates the lift cylinder 44, is driven. Therefore, each time the lift chain 42 is driven together with the lift cylinder 44, the computer 10 can acquire state data regarding the number of times the lift chain 42 has been raised and lowered.

[0096] The training data 106 includes state data regarding the number of times the lift chain 42 has been raised and lowered. The state data regarding the number of times the lift chain 42 has been raised and lowered includes the number of times the lift chain 42 has been raised and lowered as detected by the lift count detection sensor 11.

[0097] The input data ID contains the number of times the lift chain 42 has been raised or lowered. The output data OD is the fracture score S.

[0098] For example, if the fracture score S is high, meaning the degree of fracture of the lift chain 42 is high and it is time for inspection, and the lift chain 42 is used for many lifting and lowering cycles, then the degree of fracture (deterioration, fatigue) of the lift chain 42 is advanced, and it is time for inspection of the lift chain 42.

[0099] On the other hand, if the fracture score S is low, that is, if the degree of fracture of the lift chain 42 is low and the inspection time is far off, then when the number of times the lift chain 42 is raised and lowered is low, the degree of fracture (deterioration, fatigue) of the lift chain 42 has not progressed, and therefore the inspection time for the lift chain 42 is far off.

[0100] Furthermore, the lifting / lowering count detection sensor 1105 may detect the number of times the lift chain 42 has risen and the number of times the lift chain 42 has lowered separately. That is, the lifting / lowering count detection sensor 1105 can detect the number of times the lift chain 42 has risen based on the number of times the chain wheel 43 has started rotating in the forward direction, and can detect the number of times the lift chain 42 has lowered based on the number of times the chain wheel 43 has started rotating in the reverse direction. Since the wear and breakage of the lift chain 42 is greater when it is lifting a load than when it is lowering, in the breakage score S, the weighting coefficient for the number of times the lift chain 42 has risen may be set to be larger than the weighting coefficient for the number of times the lift chain 42 has lowered, and the numerical parameter may be set using a weighted average based on the weighting coefficients.

[0101] Furthermore, the degree of breakage of the lift chain 42 can actually be determined by the number of times the lift chain 42 is raised and lowered. That is, since the lift chain 42 wears down and gets damaged each time it is raised and lowered, it can be inferred that there is a certain relationship, such as a correlation, between the number of times the lift chain 42 is raised and lowered and the timing of inspection of the lift chain 42.

[0102] The acquisition unit 105 is connected to the lifting / lowering count detection sensor 1105. As described above, the input data ID includes the number of times the lift chain 42 is lifted or lowered. The input data ID is acquired by the acquisition unit 105 each time the lift chain 42 is driven, or at predetermined intervals.

[0103] The computer 10 predicts the fracture score S of the lift chain 42 by applying the learning model generated by the learning model generation unit 101 to the input data ID acquired from the acquisition unit 105.

[0104] For example, the lifting / lowering count detection sensor 1105 is not limited to a rotation speed sensor attached to the chain wheel 43, but may also consist of an optical sensor or a magnetic sensor attached to the inner mast 41, which detects the passage of the rotating part of the lift chain 42 to detect the number of lifting / lowering counts. Alternatively, the lifting / lowering count detection sensor 1105 may consist of a camera attached to the inner mast 41, and the lifting / lowering count may be detected by changes in the image of the lift chain 42 captured by the camera. Alternatively, the lifting / lowering count detection sensor 1105 may consist of an acceleration sensor attached to the lift chain 42, and the lifting / lowering count may be detected by the change in the acceleration of the lift chain 42.

[0105] The effects of the present invention will be explained. The inspection device 1 includes a learning model generation unit 101 that generates a learning model by machine learning using a dataset based on the relationship between state data regarding the total driving time of the lift chain 42 of the forklift 2 and the degree of breakage score S regarding the degree of breakage of the lift chain 32 of the forklift 2 as training data, performs machine learning on the training data, an acquisition unit 105 that acquires state data of the lift chain 42 of the forklift 2, and a prediction unit 102 that obtains the degree of breakage score S from the learning model by inputting the state data acquired from the acquisition unit 105 into the learning model generated by the learning model generation unit 101. This allows the lift chain 42 to be inspected at an appropriate time by knowing the risk of breakage (degree of breakage) of the lift chain 42 from the state of the lift chain 42. [Explanation of symbols]

[0106] 2 Forklifts 41 Inner Mast 42 Lift Chains 1. Inspection device 10 Computers 11. Total operating time detection sensor 101 Learning Model Generation Unit 102 Prediction Section 105 Acquisition Department 103 Control Unit 106 Training Data S Fracture score

Claims

1. An inspection device for the lift chain of a forklift, A learning model generation unit generates a learning model by performing machine learning on a dataset based on the relationship between state data relating to the total drive time of the forklift's lift chain and a fracture score relating to the degree of fracture of the forklift's lift chain, using the dataset as training data. The acquisition unit acquires the state data of the forklift's lift chain, The system includes a prediction unit that obtains the fracture degree score from the learning model by inputting the state data obtained from the acquisition unit into the learning model generated by the learning model generation unit. An inspection device characterized by the following features.

2. The inspection device includes a total drive time detection sensor that detects the total drive time of the lift chain of the forklift, The acquisition unit is configured to acquire the status data based on the total driving time detected by the total driving time detection sensor. The inspection device according to feature 1.

3. The acquisition unit is configured to acquire the state data each time the lift chain is driven. The inspection device according to feature 1.

4. The inspection device includes a control unit that, when the fracture degree score obtained by the prediction unit exceeds a preset first fracture degree level, guides the user to inspect the lift chain within a predetermined period, and further, when the fracture degree score exceeds a preset second fracture degree level, guides the user to inspect the lift chain immediately. The inspection device according to feature 1.

5. A method for inspecting the lift chain of a forklift, Computers Using a dataset based on the relationship between state data relating to the total drive time of the forklift's lift chain and a fracture score relating to the degree of breakage of the forklift's lift chain as training data, machine learning is performed on the training data, and a learning model is generated by the machine learning. The state data of the lift chain of the forklift is acquired, The process involves inputting the acquired state data into the generated learning model, thereby obtaining the fracture degree score from the learning model. An inspection method characterized by the following features.

6. An inspection program for the lift chain of a forklift, On the computer, Using a dataset based on the relationship between state data relating to the total drive time of the forklift's lift chain and a fracture score relating to the degree of breakage of the forklift's lift chain as training data, machine learning is performed on the training data, and a learning model is generated by the machine learning. The state data of the lift chain of the forklift is acquired, The generated learning model is then used to input the acquired state data, thereby executing a process to obtain the fracture degree score from the learning model. An inspection program characterized by the following features.

Images