Rail-guided vehicle system
The track guided bogie system diagnoses bogie conditions using fixed-position measurement units and machine learning, reducing sensor requirements and maintenance complexity while ensuring accurate condition assessment.
Patent Information
- Application Number
- JP2024013320
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2024-01-31
- Publication Date
- 2024-11-21
- Estimated Expiration
- 2040-05-12
AI Technical Summary
Existing track guided bogie systems require encoders and linear sensors on each vehicle for wheel deterioration detection, increasing costs and complexity.
A track guided bogie system that diagnoses conditions without special sensors on the bogie, using a measurement unit at a fixed position to measure vibration and sound, extracting learning features, forming a learning model, and diagnosing bogie states based on machine learning.
Reduces costs by eliminating the need for sensors on each vehicle and allows for accurate condition diagnosis of bogies, simplifying maintenance of overhead transport vehicles.
Smart Images

Figure 0007587768000002 
Figure 0007587768000003 
Figure 0007587768000004
Abstract
Description
[Technical field]
[0001] The present invention relates to a configuration for diagnosing the condition of a bogie in a rail-guided bogie system. [Background technology]
[0002] Conventionally, in rail-guided vehicle systems, the condition of the vehicle has been diagnosed by various methods in consideration of the possibility of failure. Patent Document 1 discloses a method for detecting deterioration of the running wheels of a traveling vehicle.
[0003] The traveling bogie disclosed in Patent Document 1 includes a drive wheel unit. The drive wheel unit includes a traveling wheel, a reducer, a traveling motor, etc. The traveling motor has an encoder. This encoder detects the rotation speed of the traveling wheel. An output torque signal and an encoder signal are extracted from the traveling motor, and the slip speed is calculated from the difference between the encoder signal and a signal from a linear sensor provided on the traveling bogie. Deterioration of the traveling wheel is detected based on the fact that the output torque and slip speed of the traveling motor pass through a predetermined abnormal area in a space of traveling torque and slip speed (e.g., a two-dimensional plane). [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Patent No. 6337528 Summary of the Invention [Problem to be solved by the invention]
[0005] However, the technology of Patent Document 1 requires that an encoder and a linear sensor be provided for each target overhead traveling vehicle in order to detect deterioration of the traveling wheels, and in this respect there is room for improvement.
[0006] The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a rail-guided vehicle system that diagnoses the state of the vehicle without providing a special sensor on the vehicle side.
[0007] The problem to be solved by the present invention has been described above. Next, the means for solving this problem and the effects thereof will be described.
[0008] The present invention 1. According to an aspect, there is provided a rail guided vehicle system having the following configuration. That is, the rail guided vehicle system includes a track and a vehicle that travels along the track. The rail guided vehicle system includes a measurement unit, a diagnosis device, and A feature extraction unit, a learning control unit, and a learning partial data extraction unit, The measurement unit is provided at a fixed position with respect to the track and measures at least one of vibration and sound. The diagnosis device diagnoses a condition of a carriage that has passed a measurement point on the track that corresponds to an installation position of the measurement unit, based on measurement data measured by the measurement unit. The feature extraction unit extracts learning features included in learning partial data, which is partial data corresponding to data when the cart passes the measurement point, from among the measurement data measured by the measurement unit. The learning control unit forms a learning model for diagnosing the state of the cart, using a learning data set, which is a collection of learning features based on the measurement data measured by the measurement unit for a plurality of carts. The learning partial data cut-out unit cuts out the learning partial data from the measurement data measured by the measurement unit. While the measurement data is offset in various ways in the time axis direction, a correlation is found between reference data serving as a basis and the measurement data to be cut out. With the measurement data offset in the time axis direction so as to maximize the correlation, the partial data is cut out from the measurement data in a time interval determined based on the reference data.
[0009] This makes it possible to diagnose the state of the carriage without providing a special sensor on the carriage side. Since there is no need to provide a measuring unit on each traveling carriage, costs can be reduced. The partial learning data and the learning model can be obtained satisfactorily. The partial data (partial learning data) can be obtained accurately.
[0010] According to a second aspect of the present invention, there is provided a track guided vehicle system having the following configuration. That is, the track guided vehicle system includes a track and a vehicle that travels along the track. The track guided vehicle system includes a measurement unit, a diagnosis device, a feature extraction unit, and a learning control unit. The measurement unit is provided at a fixed position relative to the track, and measures at least one of vibration and sound. The diagnosis device diagnoses the condition of the vehicle that has passed a measurement point on the track that corresponds to the installation position of the measurement unit, based on the measurement data measured by the measurement unit. The feature extraction unit extracts learning features included in learning partial data, which is partial data corresponding to data when the cart passes through the measurement point, from among the measurement data measured by the measurement unit. The learning control unit forms a learning model for diagnosing the state of the cart using a learning dataset, which is a collection of learning features based on the measurement data measured by the measurement unit for a plurality of carts. The feature extraction unit extracts diagnostic features included in diagnostic partial data, which is partial data corresponding to data when the vehicle passes the measurement point, from among the measurement data measured by the measurement unit. The diagnostic device calculates a machine learning evaluation value corresponding to the diagnostic features extracted by the feature extraction unit, based on the learning model trained by the learning control unit. The rail-guided vehicle system includes a diagnostic partial data cut-out unit. The diagnostic partial data cut-out unit cuts out the diagnostic partial data from the measurement data measured by the measurement unit. While the measurement data is offset in various ways in the time axis direction, a correlation is found between reference data serving as a basis and the measurement data to be cut out. With the measurement data offset in the time axis direction so as to maximize the correlation, the partial data is cut out from the measurement data in a time section determined based on the reference data.
[0011] This means: The state of the carriage can be diagnosed without providing a special sensor on the carriage side. Since there is no need to provide a measuring unit on each traveling carriage, costs can be reduced. A learning model can be obtained well. The machine learning evaluation value can be used to satisfactorily diagnose the state of the bogie. The diagnostic partial data can be satisfactorily acquired. The diagnostic partial data can be accurately acquired.
[0012] The rail guided vehicle system is preferably configured as follows: The rail guided vehicle system includes a learning partial data extraction unit that extracts the learning partial data from measurement data measured by the measurement unit.
[0013] This makes it possible to obtain good partial learning data.
[0022] The rail guided vehicle system is preferably configured as follows: the rail is suspended from a ceiling of a building or a platform installed on the ground, and the vehicle is a ceiling transport vehicle that travels along the rail.
[0023] Since overhead transport vehicles are installed at high altitudes, their maintenance work is generally difficult. However, with this configuration, the condition of the overhead transport vehicle can be checked without lowering it from the track, making the maintenance work less complicated. [Brief description of the drawings]
[0024] [Figure 1] 1 is a schematic diagram showing the configuration of a rail guided vehicle system according to an embodiment of the present invention; [Diagram 2] FIG. 2 is a diagram showing a traveling rail provided in the rail guided vehicle system. [Diagram 3] FIG. 2 is a block diagram showing the configuration of a diagnostic device. [Figure 4] 13 is a flowchart showing a process performed in the rail-guided vehicle system to obtain a learning dataset. [Diagram 5] 1 is a flowchart showing a process for performing machine learning. [Figure 6] 5 is a flowchart showing a process performed in the rail guided vehicle system so that a state determination unit makes a determination. [Figure 7] 13 is a graph showing the relationship between the operating time of the traveling vehicle and the evaluation value. [Figure 8] 6 is a graph illustrating offset processing of a detection signal. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0025] Next, an embodiment of the present invention will be described with reference to the drawings. Fig. 1 is a schematic diagram showing the configuration of a rail guided vehicle system 3 according to an embodiment of the present invention. Fig. 2 is a diagram showing a traveling rail 9 provided in the rail guided vehicle system 3. Fig. 3 is a block diagram showing the configuration of a diagnosis device 17.
[0026] 1 is an automatic transport system installed in, for example, a semiconductor manufacturing factory. The rail guided vehicle system 3 can transport a transport object 5 such as a FOUP. FOUP is an abbreviation for Front Opening Unified Pod.
[0027] The rail guided vehicle system 3 includes a plurality of traveling vehicles 7 and a traveling rail (track) 9.
[0028] The traveling carriage 7 is a rail-guided carriage that travels along a traveling rail 9. In this embodiment, an overhead transport vehicle called an OHT is used as the traveling carriage 7. OHT is an abbreviation for Overhead Hoist Transfer. However, the traveling carriage 7 is not limited to this. In the rail-guided carriage system 3 of this embodiment, a plurality of traveling carriages 7 having substantially the same configuration travel simultaneously. This can improve the efficiency of transportation.
[0029] 1, each of the traveling carriages 7 has a carriage control unit 11. The carriage control unit 11 is configured as a known computer having a CPU, a ROM, a RAM, a HDD, and the like.
[0030] The carriage control unit 11 controls the transfer operation of the transport target object 5 by a transfer mechanism such as a multi-joint robot arm or a hoist provided on the traveling carriage 7, and automatic traveling by a traveling mechanism (not shown) provided on the traveling carriage 7. The traveling carriage 7 controlled by the carriage control unit 11 travels along the traveling rails 9 with an appropriate interval between it and other traveling carriages 7.
[0031] As shown in FIG. 2, the traveling rail 9 is provided in a building 13 of a semiconductor manufacturing factory or the like in which the rail guided vehicle system 3 is used. Specifically, the traveling rail 9 is provided suspended from the ceiling of the building 13. Note that the traveling rail 9 may be provided suspended from a stand installed on the ground. The traveling rail 9 has a shape capable of guiding the travel path of the traveling vehicle 7, for example, as shown in FIG. 1. Note that the shape of the traveling rail 9 is not particularly limited.
[0032] The rail guided vehicle system 3 includes a learning system 1. The learning system 1 includes a measuring device 15, a diagnostic device 17, and a notification unit 45, as shown in FIG.
[0033] The measuring device 15 can measure at least one of vibrations and sounds generated in association with the traveling of the traveling carriage 7. In this embodiment, the measuring device 15 is composed of a sensor (measurement unit) 21. The sensor 21 detects vibrations or sounds of the traveling rail 9 when the traveling carriage 7 is traveling, and outputs a detection signal according to the detection result of the sensor 21 to the diagnosis device 17. This detection signal corresponds to measurement data measured by the sensor 21.
[0034] As described above, the sensor 21 of this embodiment detects vibration or sound of the traveling rail 9. In this embodiment, the sensor 21 is provided on the traveling rail 9 side, and is not provided on the traveling carriage 7. In other words, the sensor 21 is provided at a fixed position relative to the traveling rail 9. When the sensor 21 is a vibration sensor that detects vibration, the sensor is composed of, for example, an acceleration sensor, an AE (Acoustic Emission) sensor, an ultrasonic sensor, or a shock pulse sensor. When the sensor 21 is an acoustic sensor that detects sound, the sensor is composed of, for example, a microphone. Hereinafter, the position where the sensor 21 is disposed may be referred to as a measurement point P1. The measurement point P1 is determined at an appropriate point on the traveling rail 9, as shown in FIG. 2. Although not shown, a special area called a diagnosis area is set near the measurement point P1.
[0035] The learning system 1 is provided with a position sensor 25. The position sensor 25 detects the traveling carriage 7 passing through the measurement point P1. The position sensor 25 is provided at a fixed position with respect to the traveling rail 9.
[0036] The position sensor 25 is disposed in the vicinity of a portion of the traveling rail 9 that corresponds to the measurement point P1. The position sensor 25 is composed of a photoelectric sensor having a light emitting / receiving unit. When the light emitting / receiving unit detects that light emitted by the light emitting / receiving unit has been reflected by a reflecting unit or the like of the traveling carriage 7, the position sensor 25 outputs a carriage detection signal to the diagnosis device 17. However, the configuration of the position sensor 25 is not limited to this.
[0037] The diagnostic device 17 can diagnose the condition of the traveling carriage 7 that has passed the measurement point P1 on the traveling rail 9, based on the detection signal detected by the sensor 21. The diagnostic device 17 is electrically connected to each of the sensor 21 and the position sensor 25.
[0038] The diagnostic device 17 is configured as a known computer having a CPU, a ROM, a RAM, a HDD, etc. The diagnostic device 17 executes a program stored in a storage unit configured from the ROM, the RAM, the HDD, etc., to perform various processes (including the diagnosis described below).
[0039] The notification unit 45 is composed of a display such as a liquid crystal monitor, a light, a buzzer, etc. The notification unit 45 is electrically connected to the diagnosis device 17. The notification unit 45 notifies an operator, etc. of the result of the diagnosis of the state of the traveling cart 7 by the diagnosis device 17 by displaying information, and by announcing sound and / or light. This allows the operator, etc. to quickly know the state of the traveling cart 7.
[0040] As shown in FIGS. 1 and 3, the diagnostic device 17 includes an AD conversion unit 37, a data extraction unit 31, a data conversion unit 33, and an analysis unit .
[0041] The AD conversion unit 37 converts the analog signal output by the sensor 21 into a digital signal. The AD conversion unit 37 outputs the converted digital signal to the data extraction unit 31.
[0042] The data extraction unit 31 extracts only a portion of a predetermined time range including the timing at which vibration or sound is generated on the traveling rail 9 by the passage of the traveling bogie 7 from the detection signal output by the sensor 21. This time range is determined based on the timing at which the position sensor 25 outputs the bogie detection signal. Hereinafter, the detection signal extracted in this manner may be referred to as a partial signal (partial data). The partial signal is used for learning or diagnosis depending on the situation. Therefore, the data extraction unit 31 functions as a learning partial data extraction unit and a diagnosis partial data extraction unit.
[0043] The data conversion unit 33 converts the partial signal extracted by the data extraction unit 31 to obtain a frequency spectrum. The data conversion unit 33 includes a feature extraction unit 39.
[0044] The feature extraction unit 39 acquires a frequency spectrum contained in the partial signal extracted by the data extraction unit 31. Specifically, the feature extraction unit 39 executes a discrete Fourier transform process on the partial signal output by the data extraction unit 31. This results in a frequency spectrum that represents the partial signal as the relationship between the frequency and the intensity of the signal oscillating at that frequency. In this embodiment, this frequency spectrum is used as a feature (learning feature or diagnostic feature).
[0045] The analysis unit 35 forms a learning model by machine learning a large number of frequency spectra input from the data conversion unit 33. In this embodiment, the learning model is a function that inputs a frequency spectrum and outputs an evaluation value. When the learning model is formed, the analysis unit 35 uses this learning model to calculate an evaluation value from the frequency spectrum input from the data conversion unit 33 for any traveling vehicle 7. This evaluation value is based on machine learning and can therefore be called a machine learning evaluation value. The evaluation value obtained from the learning model is used to diagnose the state of the traveling vehicle 7.
[0046] The analysis unit 35 includes a machine learning unit (learning control unit) 41 and a state determination unit (diagnosis unit) 43.
[0047] The machine learning unit 41 uses frequency spectra of partial signals corresponding to the multiple traveling vehicles 7 as a group of learning data to form a learning model for diagnosing the state of the traveling vehicles 7. In this embodiment, a well-known one-class SVM is used as the learning model. SVM is an abbreviation for Support Vector Machine. In this embodiment, outlier detection (in other words, abnormality detection) is performed by unsupervised learning of the one-class SVM.
[0048] The construction of the learning model will be described in detail below. Figures 4 and 5 are flowcharts showing the process in the learning system 1 for the machine learning unit 41 to form the learning model.
[0049] First, the learning system 1 collects data for the machine learning unit 41 to form a learning model through the process shown in FIG.
[0050] First, for the traveling carriage 7 traveling along the traveling rail 9, a detection signal regarding vibration or sound caused by traveling is acquired by the sensor 21 (step S101). Here, all the traveling carriages 7 passing the measurement point on the traveling rail 9 are normal traveling carriages 7. Hereinafter, data obtained by measuring the normal traveling carriages 7 may be referred to as normal data.
[0051] Next, the data extraction unit 31 of the diagnosis device 17 extracts a part of the detection signal acquired by the sensor 21 (step S102). In this way, the partial signal described above is obtained. The time range in which the partial signal is extracted from the detection signal is set to be a predetermined time before the timing at which the position sensor 25 detects that the traveling cart 7 passes through the measurement point P1, and is a predetermined time after the timing.
[0052] Next, in the data conversion unit 33 of the diagnosis device 17, the feature extraction unit 39 executes a discrete Fourier transform process on the partial signal from the data extraction unit 31 to extract a frequency spectrum (step S103). In this embodiment, the feature input to the learning model is this frequency spectrum. Strictly speaking, the feature is the signal strength of each frequency component contained in the frequency spectrum. Appropriate smoothing is performed on the frequency spectrum. There are various smoothing methods, and one example is a method of convolution integration of a Gaussian filter.
[0053] Next, a storage unit (not shown) included in the diagnostic device 17 stores the frequency spectrum acquired by the feature amount extraction unit 39 as learning data constituting a learning data set (step S104).
[0054] The processes of steps S101 to S104 are repeated until the traveling vehicle 7 passes the measurement point P1 a sufficient number of times to obtain a large number (e.g., several thousand) of frequency spectra (step S105). In order to obtain frequency spectra under the same conditions, the traveling speed at which the traveling vehicle 7 passes the measurement point P1 for data collection is controlled to be constant at a predetermined data acquisition speed.
[0055] Not only one but multiple traveling vehicles 7 can pass through the measurement point P1 set on the traveling rail 9. In this embodiment, when, for example, several hundred frequency spectra are obtained for one normal traveling vehicle 7, the traveling vehicle 7 is replaced with another normal traveling vehicle 7 (step S106), and an operation of collecting the same number of frequency spectra is performed.
[0056] A learning data set is obtained by the above process. In this embodiment, the learning data set is substantially a collection of a large number of frequency spectra. This learning data set is stored in the storage unit of the diagnostic device 17. Note that the learning data set stored in the storage unit can be changed (updated) as appropriate.
[0057] Because of the bogie replacement work in step S106, the learning data set obtained by the process in Fig. 4 includes frequency spectra obtained from a plurality of traveling bogies 7 (different individuals). By performing machine learning using such a learning data set, it is possible to prevent excessive learning of characteristics specific to each individual traveling bogie 7. The number of traveling bogies 7 from which frequency spectra are collected is arbitrary as long as it is two or more, but it is preferable to set a relatively large number so that the individual differences between the traveling bogies 7 are not noticeable.
[0058] This learning data set does not include any frequency spectra obtained from abnormal traveling vehicles 7, and all frequency spectra are obtained from normal traveling vehicles 7.
[0059] Next, the learning system 1 uses the obtained learning dataset to perform machine learning by the machine learning unit 41 through the process shown in Fig. 5. The process in Fig. 5 corresponds to the training phase of machine learning.
[0060] First, in the analysis unit 35 of the diagnosis device 17, the machine learning unit 41 appropriately initializes parameters of a discriminant for discriminating an abnormality in the traveling carriage 7 (step S201). In this embodiment, the discriminant is expressed as follows.
number
[0061] In the above formula, f(x) is the discriminant, which corresponds to the learning model of this embodiment. x, which is the input to the discriminant f, is an N-dimensional vector representing the input frequency spectrum. In this embodiment, N means the number of frequency components when the frequency spectrum is represented by the signal strength of each of a plurality of frequency components. The output of the discriminant f is a scalar value called an evaluation value, which will be described in detail later.
[0062] K(xi,x) is a known kernel function of one-class SVM. The kernel function is a mapping function to a high-dimensional space such that the closer the frequency spectrum is to the origin, the more outliers it becomes. In one-class SVM, a hyperplane that is the maximum distance from the origin is determined in the mapping to this high-dimensional space. This hyperplane is the criterion for determining outliers. In this embodiment, a Gaussian kernel is used as the kernel function, but the kernel function is not limited to this.
[0063] αi, xi, and ρ are parameters to be machine-learned. In this embodiment, forming a learning model corresponds to obtaining optimal values of αi, xi, and ρ to determine a discriminant. In step S201, these parameters are initialized. The values used for initialization (initial values of parameters) can be, for example, random values. The parameters other than the above, σ, are called hyperparameters and are appropriately determined by the designer.
[0064] Next, the machine learning unit 41 extracts one frequency spectrum included in the training data set, and substitutes it into the discriminant f in the form of an N-dimensional vector x to obtain a value f(x) (step S202). Then, αi, xi, and ρ are adjusted so that the obtained value f(x) approaches 0 (step S203). The processes of steps S202 and S203 are repeated for all frequency spectra in the training data set (step S204). This completes one epoch of learning. An epoch refers to learning the training data included in the training data set once.
[0065] The processes of steps S202 to S204 are further repeated an appropriate number of times (step S205). This realizes learning for a plurality of epochs. By repeating learning an appropriate number of times for one learning data set, a learning model with good performance can be obtained.
[0066] 5, the values of the parameters αi, xi, and ρ are obtained, and the parameter values are stored in the storage unit of the diagnostic device 17. The above-mentioned discriminant f into which the values of the parameters αi, xi, and ρ are substituted represents a trained model. Therefore, the obtained values of the parameters αi, xi, and ρ themselves can be considered to be essentially a trained model.
[0067] In outlier detection using one-class SVM, if the value of f(x) is 0 or positive, it means that x is not an outlier. If the value of f(x) is negative, it means that x is an outlier, and the smaller the value of f(x), the greater the degree of deviation of the value. The state discrimination unit 43 of this embodiment diagnoses whether the traveling vehicle 7 is normal or abnormal by utilizing the fact that the frequency spectrum when an abnormality occurs in the traveling vehicle 7 is an outlier relative to the frequency spectrum when it is normal.
[0068] When actually carrying out a diagnosis, the process shown in Figure 6 is carried out. The process in Figure 6 corresponds to the inference phase of machine learning.
[0069] First, vibration or sound caused by the traveling of the traveling carriage 7 is acquired by the sensor 21 (step S301). At this time, the traveling carriage 7 is controlled so as to pass the measurement point P1 at the same speed as the data acquisition speed described above. A partial signal is extracted from this detection signal (step S302). Hereinafter, this partial signal may be referred to as a diagnostic partial signal. Furthermore, a frequency spectrum is extracted from this diagnostic partial signal (step S303).
[0070] The processing in steps S301 to S303 is substantially similar to that in steps S101 to S103 in Fig. 4. However, in the case of Fig. 6, it is unclear whether or not an abnormality has occurred in the traveling vehicle 7 passing through the measurement point P1.
[0071] Next, the state discrimination unit 43 inputs the frequency spectrum obtained in step S303 into a discriminant to calculate an evaluation value (step S304). At this time, the parameters αi, xi, and ρ of the discriminant are the parameters obtained by the process in Fig. 5. Therefore, the discriminant used in step S304 is synonymous with the learned learning model.
[0072] Next, the state discrimination unit 43 judges whether the calculated evaluation value is less than the threshold value v1 (step S305). Fig. 7 shows a graph of the transition of the evaluation value when the same traveling vehicle 7 is continuously operated. This graph can be obtained by actually conducting a durability test of the traveling vehicle 7 after completing the training phase of Fig. 5 in the learning model (discriminant).
[0073] In this embodiment, the threshold value v1 is set to a value at which it is assumed that the traveling vehicle 7 will soon break down, as shown in the graph of FIG.
[0074] Specifically, in the graph of FIG. 7, the traveling vehicle 7 breaks down at t2 and becomes inoperable. In this case, it is preferable to generate some kind of warning at, for example, t1, a time slightly before t2. Therefore, the threshold value v1, which is the boundary between normality and abnormality, is determined so as to have a certain margin with respect to the evaluation value (hereinafter referred to as failure evaluation value v2) corresponding to a state in which the traveling vehicle 7 is clearly broken. The threshold value v1 and the failure evaluation value v2 are negative values. The failure evaluation value v2 can be determined with reference to the results of the durability test described above. The threshold value v1 is larger than the failure evaluation value v2 by the margin described above. The size of this margin is determined appropriately taking into consideration the importance of continued operation of the rail-guided vehicle system 3 and the cost of replacing parts.
[0075] If it is determined in step S305 that the evaluation value is equal to or less than the threshold value v1, the state determination unit 43 determines that the state of the traveling bogie 7 is abnormal (step S306). In this case, the state determination unit 43 transmits the determination result (that the state of the traveling bogie 7 is abnormal) to the notification unit 45 (step S307). As a result, the notification unit 45 notifies, for example, one or both of "that the state of the traveling bogie 7 is abnormal" and "that a replacement traveling bogie or part needs to be prepared for the traveling bogie 7". The remaining operable time is the length of time until the timing when the evaluation value is expected to become equal to or less than the failure evaluation value v2. The remaining operable time can be estimated, for example, from the graph in FIG. 7 and the current evaluation value.
[0076] If it is determined in step S305 that the evaluation value is equal to or greater than the threshold value v1, the state determination unit 43 determines that the state of the traveling carriage 7 is normal (step S308).
[0077] Regardless of whether the discrimination result is normal or abnormal, the process returns to step S301, and diagnosis is performed again in response to the passage of the traveling carriage 7.
[0078] The notification by the notification unit 45 preferably includes information that a part for the traveling carriage 7 or a replacement traveling carriage needs to be prepared. In this case, it is possible to prompt a worker or the like to make appropriate preparations.
[0079] As described above, the rail guided vehicle system 3 of this embodiment includes the traveling rail 9 and the traveling vehicle 7 that travels along the traveling rail 9. The rail guided vehicle system 3 also includes the sensor 21 and the diagnostic device 17. The sensor 21 is provided at a fixed position relative to the traveling rail 9, and measures vibrations or sound. Based on measurement data measured by the sensor 21, the diagnostic device 17 diagnoses the condition of the traveling vehicle 7 that has passed a measurement point P1 that corresponds to the installation position of the sensor 21 on the traveling rail 9.
[0080] This makes it possible to diagnose the state of the traveling carriage 7 without providing a special sensor on the traveling carriage 7 side. Since it is not necessary to provide the sensor 21 on each traveling carriage 7, costs can be reduced.
[0081] The rail guided vehicle system 3 of this embodiment includes a feature extraction unit 39 and a machine learning unit 41. The feature extraction unit 39 extracts a frequency spectrum included in a learning partial signal, which is a partial signal corresponding to a detection signal when the traveling vehicle 7 passes through the measurement point P1, among detection signals detected by the sensor 21. The machine learning unit 41 forms a learning model for diagnosing the state of the traveling vehicle 7 by using a learning dataset, which is a collection of frequency spectra based on detection signals measured by the sensor 21 for a plurality of traveling vehicles 7.
[0082] This makes it possible to obtain a good learning model.
[0083] The rail guided vehicle system 3 of the present embodiment includes a data extraction unit 31. The data extraction unit 31 extracts a learning partial signal from the detection signal detected by the sensor 21.
[0084] This makes it possible to obtain the learning partial signals satisfactorily.
[0085] In the rail guided vehicle system 3 of this embodiment, the feature extraction unit 39 extracts a frequency spectrum included in a diagnostic partial signal, which is a partial signal corresponding to a detection signal when the traveling vehicle 7 passes through the measurement point P1, among the detection signals detected by the sensor 21. The diagnosis device 17 calculates an evaluation value corresponding to the frequency spectrum extracted by the feature extraction unit 39, based on a learning model trained by the machine learning unit 41.
[0086] This makes it possible to effectively diagnose the state of the traveling carriage 7 using the evaluation value.
[0087] The rail guided vehicle system 3 of the present embodiment includes a data extraction unit 31. The data extraction unit 31 extracts a partial signal for diagnosis from the detection signal detected by the sensor 21.
[0088] This makes it possible to obtain a good diagnostic partial signal.
[0089] The rail guided vehicle system 3 of this embodiment includes a position sensor 25. The position sensor 25 is provided at a fixed position with respect to the traveling rail 9, and detects the traveling vehicle 7 passing through the measurement point P1. The data extraction unit 31 extracts partial signals (a learning partial signal and a diagnosis partial signal) based on the timing at which the position sensor 25 detects the traveling vehicle 7.
[0090] This makes it possible to obtain the partial signal better.
[0091] In the rail guided vehicle system 3 of this embodiment, the traveling rail 9 is suspended from the ceiling of a building 13. The traveling vehicle 7 is a ceiling transport vehicle that travels along the traveling rail 9.
[0092] Since overhead transport vehicles are installed at high places, their maintenance work is generally cumbersome. However, according to this embodiment, the condition of the overhead transport vehicle, the traveling cart 7, can be checked without lowering it from the traveling rail 9, which reduces the complexity of the maintenance work.
[0093] Next, a modified example regarding the cutting out by the data cutting unit 31 will be described.
[0094] In the above embodiment, the range in which the partial signal is extracted from the detection signal is determined based on the timing when the position sensor 25 outputs the carriage detection signal. However, in this configuration, for example, due to individual differences in the mounting positions of the wheels of the traveling carriage 7, there is a concern that the timing at which vibration or sound is generated in the traveling rail 9 when the traveling carriage 7 passes the measurement point P1 may differ, which may reduce the accuracy of the discrimination result. Therefore, in this modified example, a correction is performed to adjust the detection signal to be extracted in the direction of the time axis.
[0095] In the learning system 1 of the modified example, before collecting data for the learning data set, a reference signal (reference data) serving as a standard for determining a data extraction section by the data extraction unit 31 is stored in advance in the storage unit of the diagnosis device 17. This reference signal can be obtained by making the traveling cart 7 pass the measurement point P1 and based on the output result of the AD conversion unit 37 at that time. As the traveling cart 7 at this time, a traveling cart 7 in a normal state is used, and the traveling cart 7 passes the measurement point P1 at the above-mentioned data acquisition speed.
[0096] Thereafter, the process of FIG. 4 is started. The graph of FIG. 8 illustrates the relationship between the waveform of the detection signal obtained in step S101 and the waveform of the reference signal when the sensor 21 is a vibration sensor. However, for convenience, the detection signal is shown in a simplified form in FIG. 8. The data extractor 31 calculates the correlation between the reference signal 51 and the obtained detection signal 53 while offsetting the detection signal 53 in various ways in the time axis direction, thereby searching for the direction and amount of offset that maximizes the correlation. The data extractor 31 offsets the detection signal 53 in the time axis direction so that the correlation between the detection signal 53 and the reference signal 51 is maximized.
[0097] In this modification, the time interval during which the data extractor 31 extracts the detection signal is determined based on a reference signal, not on the timing at which the position sensor 25 outputs the cart detection signal. The data extractor 31 extracts the detection signal in this time interval to obtain a partial signal.
[0098] The offset of the detection signal is performed not only in the process of Fig. 4 but also in the process of Fig. 6. In this modification, the accuracy of diagnosis by the learning model can be improved by the process of eliminating the timing shift of the waveform acquired by the sensor 21.
[0099] As described above, in the rail guided vehicle system 3 of this modified example, the data extraction unit 31 obtains a correlation between the reference signal 51 serving as a reference and the detection signal 53 to be extracted, while offsetting the detection signal 53 in various ways in the time axis direction. The data extraction unit 31 extracts a partial signal from the detection signal 53 in a time section determined based on the reference signal 51, with the detection signal 53 offset in the time axis direction so as to maximize the correlation.
[0100] This makes it possible to acquire partial signals with the timing discrepancy eliminated.
[0101] Although the preferred embodiment of the present invention has been described above, the above configuration can be modified, for example, as follows.
[0102] In the above embodiment, the machine learning unit 41 and the state discrimination unit 43 are provided in one diagnostic device 17, but may be provided in different devices. For example, only the machine learning unit 41 in FIG. 3 can be taken out and made into a device separate from the diagnostic device 17. Hereinafter, a computer having the machine learning unit 41 and different from the diagnostic device 17 is referred to as a training device. In this example, the place where the learning model is formed is different from the place where the diagnosis of the state of the traveling cart 7 is performed. A learning data set obtained by data collection in the diagnostic device 17 is provided to the training device. Data of the learning model formed in the training device (substantially, the parameters of the above discriminant) is provided to the diagnostic device 17. The provision of data between devices can be performed, for example, by known communication.
[0103] The state determination unit 43 can also determine that the state of the traveling carriage 7 is approaching an abnormality when a state in which the evaluation value continues to decrease by a large amount per unit time (for example, at the timing of t3 in FIG. 7). This determination result can be notified by the notification unit 45, in the same way as when the evaluation value falls below the threshold value v1.
[0104] Obviously many modifications and variations of the present invention are possible in light of the above teachings, and it is therefore to be understood that, within the scope of the appended claims, the invention may be practiced otherwise than as specifically described herein. [Explanation of symbols]
[0105] 3. Rail-guided vehicle system 7 Travelling cart (cart) 9 Running rail (track) 13 Building 17 Diagnostic Equipment 21 Sensor (measuring part) 25 Position Sensor 31 Data extraction unit (learning partial data extraction unit, diagnostic partial data extraction unit) 39 Feature Extraction Unit 41 Machine learning unit (learning control unit) 51 Reference signal (reference data) 53 Detection signal (measurement data)
Claims
1. A rail-guided vehicle system including a track and a vehicle that travels along the track, A measurement unit that is provided in a fixed position with respect to the track and measures at least one of vibration and sound; a diagnosis device that diagnoses a state of a carriage that has passed a measurement point on the track that corresponds to an installation position of the measurement unit, based on measurement data measured by the measurement unit; and a feature extraction unit that extracts learning features included in learning partial data, which is partial data corresponding to data when the cart passes through the measurement point, from among the measurement data measured by the measurement unit; and a learning control unit that uses a learning data set that is a set of learning features based on measurement data measured by the measurement unit for a plurality of carts to form a learning model for diagnosing the state of the cart; a learning partial data extracting unit that extracts the learning partial data from the measurement data measured by the measurement unit; Equipped with While the measurement data is offset in various ways along a time axis, a correlation is obtained between the reference data as a reference and the measurement data to be extracted; a track guided vehicle system, characterized in that the partial data is extracted from the measurement data in a time interval determined based on the reference data, with the measurement data being offset in the time axis direction so as to maximize the correlation.
2. A rail-guided vehicle system including a track and a vehicle that travels along the track, A measurement unit that is provided in a fixed position with respect to the track and measures at least one of vibration and sound; a diagnosis device that diagnoses a state of a carriage that has passed a measurement point on the track that corresponds to an installation position of the measurement unit, based on measurement data measured by the measurement unit; and a feature extraction unit that extracts learning features included in learning partial data, which is partial data corresponding to data when the cart passes through the measurement point, from among the measurement data measured by the measurement unit; and a learning control unit that uses a learning data set that is a set of learning features based on measurement data measured by the measurement unit for a plurality of carts to form a learning model for diagnosing the state of the cart; Equipped with the feature extraction unit extracts diagnostic features included in diagnostic partial data, which is partial data corresponding to data when the cart passes through the measurement point, from the measurement data measured by the measurement unit; the diagnostic device calculates a machine learning evaluation value corresponding to the diagnostic feature extracted by the feature extraction unit, based on the learning model trained by the learning control unit; a diagnostic partial data extraction unit that extracts the diagnostic partial data from the measurement data measured by the measurement unit, While the measurement data is offset in various ways along a time axis, a correlation is obtained between the reference data as a reference and the measurement data to be extracted; a track guided vehicle system, characterized in that the partial data is extracted from the measurement data in a time interval determined based on the reference data, with the measurement data being offset in the time axis direction so as to maximize the correlation.
3. The rail guided vehicle system according to claim 2, A rail guided vehicle system comprising: a learning partial data extraction unit that extracts the learning partial data from measurement data measured by the measurement unit.
4. The rail guided vehicle system according to any one of claims 1 to 3, The track is suspended from the ceiling of a building or from a platform installed on the ground, A rail-guided vehicle system, wherein the vehicle is a ceiling transport vehicle that travels along the rail.
Citation Information
Patent Citations
Button apparatus
JP1988037528A
Movable body diagnosis system
JP2006315813A
Abnormality diagnostic device and method
JP2017109650A
JPP6689479B
JPP7449489B
