Learning and inference devices

Abstract

The standard values ​​for evaluation of ride comfort in elevator installation are fixed thresholds set independently by elevator manufacturers. As a result, the standards are often stricter than market needs, and on-site installations have had to spend a great deal of time and effort adjusting the elevator to meet these strict standards. [Solution] The learning device is equipped with a data acquisition unit that acquires learning data including elevator vibration noise information and elevator quality judgment information by elevator passengers, and a model generation unit that uses the learning data to generate a trained model for inferring quality judgment results from the passenger's perspective from the elevator vibration noise information.

Description

[Technical Field]

[0001] The present disclosure relates to a learning device and an inference device related to elevator quality judgment. [Background technology]

[0002] When installing and adjusting an elevator, it is necessary to consider the ride comfort. In Patent Document 1, the acceleration in the acceleration waveform of the elevator car vibration measured by an accelerometer is calculated or extracted. The acceleration is then compared with pre-stored specified values ​​for evaluation items to evaluate the ride comfort of the elevator. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 9-77406 Summary of the Invention [Problem to be solved by the invention]

[0004] Typically, the specified values ​​for evaluation items are fixed thresholds independently set by elevator manufacturers. Because these specified values ​​are uniformly applied, they are often stricter than market needs. In the case of Patent Document 1, the elevator is compared with the standard value to determine whether it passes or fails. This has led to the problem that on-site users spend a great deal of time and effort making adjustments to meet the strict standards and obtain a pass rating.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to enable quality assessment that allows for optimal installation and adjustment based on the ride comfort actually felt by passengers, etc. [Means for solving the problem]

[0006] The learning device of the present disclosure includes a learning data acquisition unit that acquires learning data including elevator vibration noise information and elevator quality judgment information by elevator passengers, and a model generation unit that uses the learning data to generate a trained model for inferring quality judgment results from the passenger's perspective from the elevator vibration noise information. [Effects of the Invention]

[0007] This disclosure makes it possible to infer quality assessment results that allow for more optimal installation and adjustment in practice, compared to assessments that are made by comparing with excessive specified values ​​specified by the manufacturer. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a schematic diagram showing the general configuration of an elevator in a first embodiment. [Figure 2] FIG. 1 is a configuration diagram of a learning device according to a first embodiment. [Figure 3] FIG. 2 is a configuration diagram of a neural network according to the first embodiment. [Figure 4] 4 is a flowchart showing a learning process of the learning device according to the first embodiment. [Figure 5] 1 is a configuration diagram of an inference device according to a first embodiment. [Figure 6] 4 is a flowchart showing quality evaluation processing of the inference device in embodiment 1. [Figure 7] FIG. 2 is a diagram illustrating an example of hardware resources of the learning device according to the first embodiment. [Figure 8] FIG. 4 is a diagram illustrating another example of hardware resources of the learning device in the first embodiment. [Figure 9] FIG. 10 is a configuration diagram of a learning device according to a second embodiment. [Figure 10] FIG. 10 is a configuration diagram of an inference device according to a second embodiment. [Figure 11] FIG. 11 is a configuration diagram of a learning device according to a third embodiment. [Figure 12] FIG. 11 is a configuration diagram of an inference device according to a third embodiment. [Figure 13] FIG. 10 is a configuration diagram of a learning device according to a fourth embodiment. [Figure 14] FIG. 10 is a configuration diagram of an inference device according to a fourth embodiment. [Figure 15] FIG. 13 is a configuration diagram of a learning device according to a fifth embodiment. [Figure 16] FIG. 10 is a configuration diagram of an inference device according to a fifth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present disclosure will be described with reference to the accompanying drawings. In the drawings, the same or corresponding parts are designated by the same reference numerals, and redundant explanations will be appropriately simplified or omitted.

[0010] Embodiment 1 FIG. 1 is a schematic diagram showing the general configuration of an elevator, and FIG. 2 is a configuration diagram of a learning device for determining the quality of an elevator.

[0011] In Fig. 1, elevator car 1 is installed in hoistway 2 provided in a building. Car 1 moves within hoistway 2 between multiple floors.

[0012] A machine room 3 is provided directly above the elevator shaft 2. In the machine room 3, a traction machine 4, a traction machine control panel 5, a deflector sheave 6, and an elevator control device 7 are provided.

[0013] The car 1 and the counterweight 8 are each connected to both ends of a rope 9. The rope 9 is hung between the hoist 4 and the deflector sheave 6. Thus, the car 1 and the counterweight 8 are suspended. An acceleration sensor 10 that measures the vibration of the car 1 is installed on top of the car 1. The acceleration sensor 10 has a communication function and can transmit the measured acceleration wirelessly.

[0014] In this elevator, a traction machine control panel 5 controls the rotation and stopping of a traction machine 4, thereby causing a car 1 installed in a hoistway 2 to travel upward or downward.

[0015] At each floor, there is provided a landing 11 for boarding the elevator car 1. When the car 1 arrives at each floor, the doors open and passengers 12 can board the car 1.

[0016] <Learning Phase> 2, learning device 20 includes a learning data acquisition unit 21, a model generation unit 22, and a trained model storage unit 23. It is assumed that learning device 20 is installed in, for example, a management center that manages multiple elevators.

[0017] The learning data acquisition unit 21 acquires car vibration measurement information as input 1. The car vibration measurement information is one type of elevator vibration and noise information. Note that the elevator vibration and noise information may include only vibration information, only noise information, or both vibration and noise information.

[0018] The car vibration measurement information is acceleration time-series data that is a time series of acceleration measured by an acceleration sensor 10, which is a vibration measurement device. For example, the elevator control device 7 receives the acceleration transmitted from the acceleration sensor 10, accumulates it, and converts it into acceleration time-series data. The elevator control device 7 then transmits this to a management center via the public line network 14.

[0019] Furthermore, the learning data acquisition unit 21 acquires, as input 2, ride comfort judgment information of the passenger 12 associated with the acceleration time series data of input 1. The ride comfort judgment information is an evaluation of the ride comfort by the passenger 12 who is in the car 1, and is one type of elevator quality judgment information by the passenger. The ride comfort judgment information is input by the passenger 12 using the mobile terminal 13 that he or she carries. For example, dedicated software is stored in the mobile terminal 13 of the passenger 12. Then, the passenger 12 inputs the elevator ride comfort while riding or when getting off. This is done using a selection method in which options such as excellent, good, fair, and poor are displayed and one is selected.

[0020] This ride comfort evaluation information is transmitted from the mobile terminal 13, received by the elevator control device 7, and linked to the acceleration time-series data. In other words, it is the ride comfort evaluated at a point in time in the acceleration time-series data or between certain points in time. Note that a large amount of car vibration measurement information and ride comfort evaluation information is transmitted from multiple elevators to the learning data acquisition unit 21. Elevators to be acquired can also be added later.

[0021] The learning data acquisition unit 21 acquires learning data consisting of input 1: car vibration measurement information and input 2 (correct answer): ride comfort judgment information, and sends it to the model generation unit 22.

[0022] The model generation unit 22 learns the ride comfort judgment result from the passenger's perspective, which is one of the quality judgment results from the passenger's perspective, based on the learning data sent from the learning data acquisition unit 21. That is, the model generation unit 22 infers the optimal ride comfort judgment result from the passenger's perspective from the car vibration measurement information and the ride comfort judgment information, and generates a learned model that outputs it.

[0023] The learning algorithm used by the model generation unit 22 may be a known algorithm such as supervised learning, unsupervised learning, reinforcement learning, etc. As an example, a case where the model generation unit 22 is configured by a neural network will be described.

[0024] The model generation unit 22 learns the passenger-perspective ride comfort assessment results to be output, for example, according to a neural network model by so-called supervised learning. Here, supervised learning refers to a method in which a learning device is provided with pairs of input and result (label) data, and the device learns the features of the learning data and infers the result from the input.

[0025] A neural network consists of an input layer consisting of multiple neurons, an intermediate layer (hidden layer) consisting of multiple neurons, and an output layer consisting of multiple neurons. The intermediate layer may be one layer, or two or more layers.

[0026] For example, in a three-layer neural network configuration like the one shown in Figure 3, multiple inputs are input to the input layer (X1-X3). These values ​​are multiplied by weight W1 (w11-w16) and input to the middle layer (Y1-Y2). The result is further multiplied by weight W2 (w21-w26) and output from the output layer (Z1-Z3). This output result changes depending on the values ​​of weights W1 and W2.

[0027] The model generation unit 22 learns the ride comfort judgment results from the passenger's perspective by so-called supervised learning in accordance with the learning data sent from the learning data acquisition unit 21. Note that Input 2: ride comfort judgment information is the teacher result (label), that is, the correct answer.

[0028] The neural network of the model generation unit 22 learns by inputting input 1: cage vibration measurement information into the input layer and adjusting weights W1 and W2 so that the result output from the output layer approaches input 2 (correct answer): passenger ride comfort judgment information.

[0029] The trained model storage unit 23 stores the trained model 24 output from the model generation unit 22.

[0030] Next, the learning process in the learning device 20 of FIG. 2 will be described. FIG. 4 is a flowchart of the learning process of the learning device 20.

[0031] First, the learning data acquisition unit 21 acquires input 1: car vibration measurement information and input 2 (correct answer): ride comfort judgment information as learning data (step S001). Note that here, the learning data acquisition unit 21 simultaneously acquires input 1: car vibration measurement information and input 2 (correct answer): ride comfort judgment information. However, as long as input 1 and input 2 can be associated with each other, the learning data acquisition unit 21 may acquire them at different times.

[0032] Next, the model generation unit 22 learns the ride comfort judgment results from the passenger's perspective by so-called supervised learning in accordance with the learning data sent from the learning data acquisition unit 21, and generates a trained model 24 (step S002). Note that if a trained model 24 already exists, it will be re-trained by adding the current learning data, and the trained model 24 will be generated.

[0033] Next, this trained model 24 is stored in the trained model storage unit 23 (step S003).

[0034] <Utilization phase> Next, a quality evaluation of ride comfort using the trained model 24 generated by the learning device 20 will be described.

[0035] FIG. 5 is a configuration diagram of an inference device 30 in which the target product is an elevator. The inference device 30 includes an inference data acquisition unit 31 and an inference unit 32. The inference device 30 is installed in a management center, similar to the learning device 20. The learning device 20 and the inference device 30 are interconnected to form a quality evaluation device or a quality evaluation system.

[0036] The inference data acquisition unit 31 acquires car vibration measurement information during travel as input 3. The inference unit 32 uses the trained model 24 to infer and output a passenger-perspective ride comfort assessment result 40 from the car vibration measurement information. The passenger-perspective ride comfort assessment result 40 is one of the quality assessment results from the passenger's perspective.

[0037] Next, the quality evaluation process in the inference device 30 of FIG. 5 will be described. FIG. 6 is a flowchart showing the quality evaluation process of the inference device 30.

[0038] First, the inference data acquisition unit 31 acquires input 3: car vibration measurement information (step S101). For example, after installing and adjusting a new elevator in a building, an installation worker performs a test run of car 1. While the elevator is running, the acceleration sensor 10 of car 1 transmits measured acceleration wirelessly. During installation and adjustment, this acceleration is received and stored on a work terminal carried by the worker. After the test run, the installation worker sends the accumulated car vibration measurement information, which is acceleration time series data, to the inference device 30. Then, the inference data acquisition unit 31 acquires the car vibration measurement information.

[0039] Next, the inference unit 32 acquires the trained model 24 from the trained model storage unit 23. Then, the inference unit 32 inputs input 3: car vibration measurement information to the trained model 24 (step S102).

[0040] Next, the trained model 24 outputs the passenger-perspective ride comfort judgment result 40 as an output for the input 3: car vibration measurement information (step S103).

[0041] This result is sent to the work terminal carried by the installer and displayed (step S104). The elevator installer determines whether or not adjustments to the elevator ride comfort are necessary based on the displayed ride comfort assessment result 40 from the passenger's perspective. This ride comfort assessment result 40 may be excellent, good, fair, or poor, for example. Note that this assessment result may relate to the overall ride comfort. Alternatively, the travel time of car 1 from the lowest to the highest floor may be divided into sections, and the assessment result may relate to the ride comfort for each time section.

[0042] This passenger-perspective ride quality assessment result 40 is based on the evaluations of many passengers who have actually used the elevator. Therefore, the inference device 30 can infer a quality that is more appropriate for the actual situation than when comparing with excessive specified values ​​specified by the manufacturer. In this way, in the first embodiment, adjustments can be made based on appropriate passenger-perspective quality assessment results, thereby reducing the time required for adjustment work.

[0043] It should be noted that passengers may input the ride comfort assessment information not only from the mobile terminal 13 but also from a terminal (touch panel, push button, etc.) installed in the car.

[0044] The learning device 20 and the inference device 30 may be built into an elevator. In this case, a trained model dedicated to the elevator is generated based on car vibration measurement information acquired by the elevator and passenger ride comfort evaluation information. For example, using a dedicated trained model to assess the quality during elevator maintenance improves work efficiency.

[0045] The learning device 20 and the inference device 30 may be configured to exist as a single function on a cloud server.

[0046] The elevator control device 7 may transmit the car vibration measurement information to the inference device 30. The results of the inference device 30 may also be sent to the elevator control device 7. In this case, the installation worker will obtain the passenger-perspective ride comfort evaluation result 40 from the elevator control device 7.

[0047] The results of the inference device 30 may be sent to a cloud server. In this case, the installation worker accesses the cloud server via a browser and checks the results of the ride comfort assessment from the passenger's perspective.

[0048] Although the above description has been given of a case where supervised learning is applied to the learning algorithm used by the model generation unit 22, the present invention is not limited to this. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can also be applied in addition to supervised learning.

[0049] Deep learning, which learns to extract feature quantities themselves, can also be used as the learning algorithm used in the model generation unit 22. The model generation unit 22 may also perform machine learning according to other known methods, such as genetic programming, inductive logic programming, or support vector machines.

[0050] Furthermore, although the passenger's ride comfort assessment information is used to output the ride comfort assessment result from the passenger's perspective, other factors besides ride comfort may also be used. In short, as long as an evaluation from the passenger's perspective is possible, other quality assessment information by passengers may also be used as learning data. As a result, the inference device can output the quality assessment result from the passenger's perspective.

[0051] 7 is a diagram illustrating an example of hardware resources of the learning device 20. The learning device 20 has a processor 20a, a memory 20b, and a transmission / reception circuit 20c as hardware resources. Note that there may be multiple processors 20a, memories 20b, and transmission / reception circuits 20c.

[0052] In the first embodiment, the processing in the learning data acquisition unit 21 and the model generation unit 22 is executed by the processor 20a using software, firmware, or a combination of software and firmware written as a program stored in the memory 20b. The transmission / reception circuit 20c is responsible for acquiring information in the learning data acquisition unit 21. The memory 20b temporarily stores information during processing and functions as the trained model storage unit 23.

[0053] The processor 20a is also called a CPU (Central Processing Unit), central processing unit, arithmetic unit, microprocessor, microcomputer, or DSP. The memory 20b may be a semiconductor memory, a magnetic disk, a flexible disk, an optical disk, a compact disk, a minidisk, or a DVD. Usable semiconductor memories include RAM, ROM, flash memory, EPROM, and EEPROM.

[0054] FIG. 8 is a diagram showing another example of hardware resources of the learning device 20. In the example of FIG. 8, the learning device 20 has a processing circuit including a processor 20a, a memory 20b, a transceiver circuit 20c, and dedicated hardware 20d. Some of the functions of the learning device 20 are realized by the dedicated hardware 20d. Note that all of the functions of the learning device 20 may also be realized by the dedicated hardware 20d. The dedicated hardware 20d may be a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

[0055] The hardware resources of the inference device 30 are also similar to those shown in Figures 7 and 8. That is, the processing in the inference data acquisition unit 31 and the inference unit 32 is executed by a processor using software, firmware, or a combination of software and firmware written as a program stored in memory.

[0056] Embodiment 2 In the second embodiment, in addition to input 1 and input 2 (correct answer), car interior noise measurement information is input and used as learning data. This is because passengers 12 may feel uncomfortable riding in elevator car 1 due to noise generated inside the car while the car is moving.

[0057] Fig. 9 is a configuration diagram of a learning device 20 in embodiment 2. Fig. 10 is a configuration diagram of an inference device 30 in embodiment 2. Note that the same reference numerals are used to designate the same or similar configurations as those of the learning device in Fig. 2 and the inference device in Fig. 4, respectively, and descriptions thereof will be omitted.

[0058] First, a sound measuring device (not shown) for measuring noise is installed inside elevator car 1. The sound measuring device has a communication function and can wirelessly transmit the measured values ​​of noise frequency and the like.

[0059] The learning data acquisition unit 21a acquires noise measurement information inside the car as input 4. The noise measurement information inside the car is noise time-series data that shows noise values ​​measured by a sound measurement device in time series. This is synchronized with input 1: car vibration measurement information. For example, the elevator control device 7 receives and stores the noise values ​​transmitted from the sound measurement device at the same time as it receives the acceleration, and converts this into noise time-series data. The elevator control device 7 then transmits this to the management center via the public line network 14. Note that both the car vibration measurement information and the noise measurement information inside the car are types of elevator vibration noise information.

[0060] The learning data acquisition unit 21a acquires, as learning data, input 1: car vibration measurement information, input 2 (correct answer): ride comfort judgment information, and input 4: car in-noise measurement information.The learning data acquisition unit 21a then sends input 1: car vibration measurement information, input 2 (correct answer): ride comfort judgment information, and input 4: car in-noise measurement information to the model generation unit 22a.

[0061] The model generation unit 22a learns the results of the ride comfort assessment from the passenger's perspective by so-called supervised learning in accordance with this learning data, and generates a learned model 24a.

[0062] In the inference device 30, the inference data acquisition unit 31a acquires input 3: car vibration measurement information and input 5: car in-car noise measurement information. Note that input 5: car in-car noise measurement information was obtained in synchronization with the car vibration measurement information during a test run. Then, the inference unit 32a outputs a passenger-perspective ride comfort judgment result 40a from input 3: car vibration measurement information and input 5: car in-car noise measurement information.

[0063] In this way, by adding the noise inside the car that is generated when car 1 is running, which affects the ride comfort felt by passengers, to the learning process, it becomes possible to infer more accurate quality assessment results.

[0064] Embodiment 3 In the third embodiment, in addition to input 1 and input 2 (correct answer), product information about the elevator is used as learning data. Fig. 11 is a configuration diagram of a learning device 20 in the third embodiment. Fig. 12 is a configuration diagram of an inference device 30 in the third embodiment. Note that the same reference numerals are used to designate the same or similar components as those in the learning device in Fig. 2 and the inference device in Fig. 4, respectively, and descriptions thereof will be omitted.

[0065] The learning data acquisition unit 21b acquires input 1: car vibration measurement information, input 2 (correct answer): ride comfort assessment information, and product information for the elevator. Here, car size is used as an example of product information. This is because even if the vibration is the same, the ride comfort felt by passengers will differ if the size of the car 1 they are riding in is different. For example, if the car 1 is large, the vibration will not be noticeable, but if the car 1 is small, it will be noticeable.

[0066] The learning data acquisition unit 21b sends the learning data of input 1: car vibration measurement information, input 2 (correct answer): ride comfort judgment information, and product information to the model generation unit 22b.

[0067] The model generation unit 22b learns the results of the ride comfort assessment from the passenger's perspective by so-called supervised learning in accordance with this learning data, and generates a learned model 24b.

[0068] In the inference device 30, the inference data acquisition unit 31b acquires input 3: car vibration measurement information and product information of the elevator after installation. The inference unit 32b acquires the trained model 24b from the trained model storage unit 23b. Then, the inference unit 32b outputs a passenger-perspective ride comfort assessment result 40b based on input 3: car vibration measurement information and the elevator product information.

[0069] In this way, by adding product information that affects the ride comfort perceived by passengers and learning it, it becomes possible to infer more accurate quality assessment results based on information measured in other types of elevators.

[0070] Although the product information used here is the car size, it may also be, for example, the speed, the load capacity, the weight of the car, the lifting stroke, etc. This is because differences in these factors also affect the riding comfort of passengers.

[0071] Furthermore, instead of or in addition to product information, information about the installation environment in which the elevator is installed may be used as learning data. Examples of installation environment information include the region (city, prefecture, country, etc.) in which the elevator is located. Even with the same vibration, passengers may experience different riding comfort depending on the region.

[0072] Furthermore, installation environment information includes, for example, the construction company of the building in which the elevator is installed and the owner of the building. The installation conditions of an elevator may vary depending on the construction method used by the construction company. Furthermore, depending on the owner of the building, such as whether the elevator is installed in a department store, a mall, or a hotel, the passengers who ride in the elevator and the riding comfort they experience will also vary.

[0073] In this way, by adding and learning information about the installation environment that affects the ride comfort perceived by passengers, it becomes possible to infer more accurate quality assessment results.

[0074] Naturally, the functions of the third embodiment can also be applied to the second embodiment.

[0075] Embodiment 4 In the fourth embodiment, in addition to input 1: car vibration measurement information, input 6: car position information is also used as learning data. This is because even if the vibration is the same, the ride comfort felt by passengers may differ depending on the height position of the car when the vibration occurs.

[0076] Fig. 13 is a configuration diagram of learning device 20 in embodiment 4. Fig. 14 is a configuration diagram of inference device 30 in embodiment 4. Note that the same reference numerals are used to designate the same or similar configurations as learning device 20 in Fig. 2 and inference device 30 in Fig. 4, respectively, and descriptions thereof will be omitted.

[0077] The learning data acquisition unit 21c acquires input 1: car vibration measurement information, input 2 (correct answer): ride comfort judgment information, and input 6: car position information. Here, the car position information can be acquired, for example, from an altimeter installed in the car 1. It can also be acquired from the hoist control panel 5, which grasps the rotation speed of the hoist 4.

[0078] The learning data acquisition unit 21c sends the learning data of input 1: car vibration measurement information, input 2 (correct answer): ride comfort judgment information, and input 6: car position information to the model generation unit 22c.

[0079] The model generation unit 22c generates a trained model 24c for inferring the results of the passenger's perspective ride comfort assessment and elevator adjustment locations by so-called supervised learning in accordance with this training data.

[0080] In the inference device 30, the inference data acquisition unit 31c acquires input 3: car vibration measurement information and input 7: position information of car 1 while the elevator is running. The inference unit 32c acquires the trained model 24c from the trained model storage unit 23c. Then, the inference unit 32c outputs a passenger-perspective ride comfort assessment result 40c including elevator adjustment points from input 3: car vibration measurement information and input 7: position information of car 1.

[0081] In this way, by learning the position of car 1, which affects the ride comfort perceived by passengers, it becomes possible to infer more accurate quality assessment results. Also, when the assessment results include a poor evaluation such as "fail," it becomes possible to estimate areas that need adjustment.

[0082] The inputs 6 and 7 may be air pressure, which changes depending on the position of the car 1. This is because the same vibration may cause the passenger to feel different riding comfort depending on the air pressure. In this case, it is possible to provide the portable terminal 13 carried by the passenger 12 with an air pressure measurement function and use the information obtained. Alternatively, a barometer may be installed in the car 1.

[0083] Furthermore, rail displacement information such as rail bends, steps, and curves at the position of car 1 may be included in inputs 6 and 7. This allows for more accurate estimation of the areas requiring adjustment.

[0084] Naturally, the functions of the fourth embodiment can also be applied to the second and third embodiments.

[0085] Embodiment 5. In the fifth embodiment, information on the measurement of the sound of the door opening and closing of the elevator car 1 is evaluated based on the results of the ride comfort assessment from the passenger's perspective.

[0086] Fig. 15 is a configuration diagram of learning device 20 in embodiment 5. Fig. 16 is a configuration diagram of inference device 30 in embodiment 5. Note that the same reference numerals are used to designate the same or similar configurations as learning device 20 in Fig. 2 and inference device 30 in Fig. 4, respectively, and descriptions thereof will be omitted.

[0087] The learning data acquisition unit 21d acquires input 8: car 1 door opening / closing sound measurement information, and input 9 (correct answer): passenger hearing comfort judgment information. The car 1 door opening / closing sound measurement information is one type of elevator vibration noise information.

[0088] The car 1 has a sound measuring device (not shown) above the car door (not shown). Measurement begins when the car door opens and ends when the car door closes. Passengers 12 also input hearing comfort evaluation information regarding the opening and closing of the car door using their own mobile terminals 13. This hearing comfort evaluation information is one of the elevator quality evaluation information provided by passengers. For example, if the passenger finds the door opening and closing sound extremely unpleasant, they can select "poor," and if they do not find it unpleasant, they can select "excellent."

[0089] The learning data acquisition unit 21d sends the learning data of input 8: car 1 door opening / closing sound measurement information and input 9 (correct answer): passenger listening comfort judgment information to the model generation unit 22d.

[0090] The model generation unit 22d learns the results of the passenger-perspective listening comfort assessment using the learning data through so-called supervised learning, and generates a learned model 24d.

[0091] In the inference device 30, the inference data acquisition unit 31d acquires input 10: door opening / closing sound measurement information of car 1. This door opening / closing sound measurement information was measured when an installation worker performed a trial run of the door opening and closing of car 1 after installing and adjusting a new elevator in a building.

[0092] The inference unit 32d acquires the trained model 24d from the trained model storage unit 23d. Then, the inference unit 32d outputs a passenger-perspective listening comfort judgment result 40d from the input 10: car 1 door opening / closing sound measurement information.

[0093] This passenger-perspective hearing comfort assessment result 40d is based on the evaluations of many passengers who have actually used the elevator. Therefore, it is a more appropriate assessment that is suited to the actual situation than an assessment that compares with excessive specified values ​​set by the manufacturer. In this way, in the fifth embodiment, it is possible to infer a more appropriate assessment result for the quality of the car doors that is suited to the actual situation.

[0094] Although the preferred embodiments have been described in detail above, the present invention is not limited to these embodiments, and various modifications and substitutions can be made to the above-described embodiments without departing from the scope of the disclosure.

[0095] Furthermore, when the number, quantity, amount, range, etc. of each element is mentioned in the embodiments, the device of this disclosure is not limited to the mentioned number unless otherwise specified or clearly specified in principle. Furthermore, the structures, etc. described in the embodiments are not necessarily essential unless otherwise specified or clearly specified in principle.

[0096] Various aspects of the present disclosure are summarized below as appendices. (Appendix 1) a learning data acquisition unit that acquires learning data including elevator vibration noise information and elevator quality assessment information by passengers of the elevator; a model generation unit that generates a trained model for inferring a quality judgment result from a passenger's perspective from the elevator vibration and noise information using the training data; A learning device comprising: (Appendix 2) The learning device described in Appendix 1 is characterized in that the elevator vibration noise information is elevator car vibration measurement information while the elevator car is moving, measured by a vibration measuring device installed in the car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the car. (Appendix 3) The learning device described in Appendix 1 is characterized in that the elevator vibration noise information is car vibration measurement information while the elevator is moving, measured by a vibration measuring device installed in the car, and car interior noise measurement information while the elevator is moving, measured by a sound measuring device installed in the car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the car. (Appendix 4) 4. The learning device according to claim 2 or 3, wherein the ride comfort assessment information is input by a selection method from a portable terminal carried by the passenger. (Appendix 5) 4. The learning device according to claim 2 or 3, wherein the ride comfort assessment information is input from a terminal installed in the car. (Appendix 6) 6. The learning device according to any one of claims 2 to 5, wherein the model generation unit is configured by a neural network. (Appendix 7) 7. A learning device according to any one of claims 2 to 6, wherein the learning data includes product information of the elevator. (Appendix 8) The learning device described in Appendix 7, characterized in that the product information includes at least one of the size of the car, the speed of the car, the load capacity of the car, the weight of the car, and the elevator's ascent and descent distance. (Appendix 9) 7. The learning device according to any one of claims 2 to 6, wherein the learning data includes information about the installation environment of the elevator. (Appendix 10) The learning device described in Appendix 9, characterized in that the installation environment information includes at least one of the location of the elevator, information about the construction company of the building in which the elevator is installed, and information about the owner of the building in which the elevator is installed. (Appendix 11) 7. The learning device according to any one of claims 2 to 6, wherein the learning data includes position information of the car. (Appendix 12) 12. The learning device according to claim 11, wherein the learning data includes displacement information of the rails of the car. (Appendix 13) The learning device described in Appendix 11, characterized in that the model generation unit generates the trained model for inferring the quality judgment result from the passenger's perspective and the adjustment points of the elevator from the car vibration measurement information. (Appendix 14) 7. The learning device according to claim 2, wherein the learning data includes air pressure information inside the car. (Appendix 15) The learning device described in Appendix 1 is characterized in that the elevator vibration noise information is door opening / closing sound measurement information measured by a sound measuring device inside the car, and the elevator quality judgment information is information judging the audibility of the sound when the door opens and closes by passengers riding in the car. (Appendix 16) an inference data acquisition unit that acquires elevator vibration and noise information; an inference unit that uses a trained model generated by a learning device to generate a trained model for inferring a quality judgment result from a passenger's perspective from the elevator vibration noise information, using learning data including the elevator vibration noise information and elevator quality judgment information by passengers of the elevator, and outputs a quality judgment result from the passenger's perspective from the elevator vibration noise information acquired by the inference data acquisition unit, (Appendix 17) The inference device described in Appendix 16, characterized in that the elevator vibration noise information is elevator car vibration measurement information while the elevator car is moving, measured by a vibration measuring device installed in the car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the car. (Appendix 18) The inference device described in Appendix 16, characterized in that the elevator vibration noise information is car vibration measurement information while the elevator is moving, measured by a vibration measuring device installed in the car, and car interior noise measurement information while the elevator is moving, measured by a sound measuring device installed in the car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the car. (Appendix 19) the learning data includes product information of the elevator, the inference data acquisition unit acquires product information of the elevator, The inference device described in Appendix 17 or 18, characterized in that the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and elevator product information acquired by the inference data acquisition unit. (Appendix 20) The learning data includes installation environment information of the elevator, the inference data acquisition unit acquires installation environment information of the elevator, The inference device described in Appendix 17 or 18, wherein the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and elevator installation environment information acquired by the inference data acquisition unit. (Appendix 21) The learning data includes position information of the car, the inference data acquisition unit acquires the car's position information, The inference device described in Appendix 17 or 18, characterized in that the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and the car position information acquired by the inference data acquisition unit. (Appendix 22) the learning data includes air pressure information inside the car, the inference data acquisition unit acquires air pressure information inside the car, The inference device described in Appendix 17 or 18, characterized in that the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and air pressure information inside the car acquired by the inference data acquisition unit. (Appendix 22) The inference device described in Appendix 16, characterized in that the elevator vibration noise information is door opening / closing sound measurement information measured by a sound measuring device inside the car, and the elevator quality judgment information is information judging the audibility of the sound when the door opens and closes by passengers riding in the car. [Explanation of symbols]

[0097] 1. Cage, 2. Hoistway, 3. Machine room, 4. Hoisting machine, 5. Hoisting machine control panel, 6 deflector, 7 elevator control device, 8 counterweight, 9 Rope, 10 Accelerometer, 11 Platform, 12 Passenger, 13 mobile devices, 20 Learning Devices, 20a processor, 20b memory, 20c transmitting and receiving circuit, 20d dedicated hardware, 21, 21a, 21b, 21c, 21d learning data acquisition unit, 22, 22a, 22b, 22c, 22d model generation unit; 23, 23a, 23b, 23c, 23d learning model memory unit, 24, 24a, 24b, 24c, 24d Trained model, 30 reasoning device, 31, 31a, 31b, 31c, 31d inference data acquisition unit, 32, 32a, 32b, 32c, 32d reasoning section, 40, 40a, 40b, 40c: Passenger-perspective riding comfort evaluation results, 40d Listening comfort evaluation results from passenger's perspective

Claims

1. a learning data acquisition unit that acquires learning data including elevator vibration noise information and elevator quality assessment information by passengers of the elevator; a model generation unit that generates a trained model for inferring a quality judgment result from a passenger's perspective from the elevator vibration and noise information using the training data; A learning device comprising:

2. The learning device described in claim 1, characterized in that the elevator vibration noise information is elevator car vibration measurement information while the elevator is running, measured by a vibration measuring device installed in the elevator car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the elevator car.

3. The learning device described in claim 1, characterized in that the elevator vibration noise information is car vibration measurement information while the elevator is moving, measured by a vibration measuring device installed in the car, and car interior noise measurement information while the elevator is moving, measured by a sound measuring device installed in the car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the car.

4. 3. The learning device according to claim 2, wherein the ride comfort evaluation information is input by a selection method from a portable terminal carried by the passenger.

5. 3. The learning device according to claim 2, wherein the ride comfort judgment information is input from a terminal installed in the car.

6. 3. The learning device according to claim 2, wherein the model generation unit is configured by a neural network.

7. 3. The learning device according to claim 2, wherein the learning data includes product information of the elevator.

8. The learning device according to claim 7, characterized in that the product information includes at least one of the size of the car, the speed of the car, the load capacity of the car, the weight of the car, and the ascent and descent distance of the elevator.

9. 3. The learning device according to claim 2, wherein the learning data includes information about the installation environment of the elevator.

10. The learning device according to claim 9, characterized in that the installation environment information includes at least one of the location of the elevator, information about the construction company of the building in which the elevator is installed, and information about the owner of the building in which the elevator is installed.

11. 3. The learning device according to claim 2, wherein the learning data includes position information of the car.

12. 12. The learning device according to claim 11, wherein the learning data includes displacement information of rails of the car.

13. The learning device according to claim 11, characterized in that the model generation unit generates the trained model for inferring the quality judgment result from the passenger's perspective and the adjustment points of the elevator from the car vibration measurement information.

14. 3. The learning device according to claim 2, wherein the learning data includes information about atmospheric pressure inside the car.

15. The learning device described in claim 1, characterized in that the elevator vibration noise information is door opening / closing sound measurement information measured by a sound measuring device inside the car, and the elevator quality judgment information is information judging the audibility of the sound when the door is opened and closed by passengers riding in the car.

16. an inference data acquisition unit that acquires elevator vibration and noise information; an inference unit that uses learning data including the elevator vibration noise information and elevator quality judgment information by passengers of the elevator to generate a trained model for inferring a quality judgment result from a passenger's perspective from the elevator vibration noise information, and outputs a quality judgment result from the passenger's perspective from the elevator vibration noise information acquired by the inference data acquisition unit, using the trained model generated by a learning device.

17. The inference device described in claim 16, characterized in that the elevator vibration noise information is elevator car vibration measurement information while the elevator car is moving, measured by a vibration measuring device installed in the car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the car.

18. 17. The inference device according to claim 16, wherein the elevator vibration noise information is elevator car vibration measurement information while the elevator is moving, measured by a vibration measuring device installed in the elevator car, and elevator car interior noise measurement information while the elevator is moving, measured by a sound measuring device installed in the elevator car, and the elevator quality judgment information is ride comfort judgment information by passengers riding in the elevator car.

19. the learning data includes product information of the elevator, the inference data acquisition unit acquires the product information, The inference device according to claim 17 or 18, characterized in that the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and the product information acquired by the inference data acquisition unit.

20. The learning data includes installation environment information of the elevator, the inference data acquisition unit acquires the installation environment information, The inference device according to claim 17 or 18, characterized in that the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and the installation environment information acquired by the inference data acquisition unit.

21. The learning data includes position information of the car, the inference data acquisition unit acquires the car's position information, The inference device according to claim 17 or 18, characterized in that the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and the car position information acquired by the inference data acquisition unit.

22. the learning data includes air pressure information inside the car, the inference data acquisition unit acquires air pressure information inside the car, The inference device according to claim 17 or 18, characterized in that the inference unit outputs a quality judgment result from the passenger's perspective based on the car vibration measurement information and the air pressure information inside the car acquired by the inference data acquisition unit.

23. The inference device described in claim 16, characterized in that the elevator vibration noise information is door opening / closing sound measurement information measured by a sound measuring device inside the car, and the elevator quality judgment information is information judging the audibility of the sound when the door is opened and closed by a passenger riding in the car.

Images