JP7848005B2 - Paper handling apparatus and its maintenance method - Google Patents

Description

This invention relates to a paper sheet handling device and a method for maintaining the same.

Conventional automated cash transaction machines, as paper handling devices, separate accumulated banknotes one by one using rollers, feed them out, and then process them while they are being transported by a conveyor belt or rollers. During this process, the transport status of each banknote is detected by a passage sensor located in the transport section, and based on this detection signal, control of the transport direction and fault detection are performed.

Incidentally, in paper handling devices such as automated cash machines, the rollers, conveyor belts, and drive units deteriorate after prolonged feeding and transport of banknotes and other paper materials. This prevents the normal feeding, stacking, and transport of paper materials. As a result, frequent rejections (defective banknotes) due to transport problems and stacking failures occur, causing problems with subsequent transactions. If the deterioration is left untreated, the paper handling device will eventually malfunction, rendering it unable to perform transactions.

To prevent such problems, Patent Document 1 measures the time signal of paper passing through the transport path and diagnoses a malfunction based on the degree of deviation of that time from the normal range.

Furthermore, in ATMs and photocopiers, maintenance personnel visually inspect rollers and conveyor belts at predetermined intervals to determine whether replacement is necessary, and replace them if required.

Japanese Patent Publication No. 2007-210718

Maintenance work on paper handling equipment is carried out at predetermined intervals, according to a prescribed manual. This is because there is currently no established method for determining whether replacement is necessary, making it impossible to identify rollers and conveyor belts that actually require replacement. Therefore, even rollers and conveyor belts with low transaction volumes during the maintenance period that do not require inspection cannot be omitted, hindering improvements in maintenance efficiency and cost reduction.

If a regular maintenance service system is not established, maintenance should only be performed when replacement is truly necessary. However, for the reasons mentioned above, we have not yet been able to perform such maintenance work.

To perform efficient and low-cost maintenance work, it is necessary to accurately measure the degree of deterioration of rollers and conveyor belts. However, fault diagnosis methods using specific sensors, such as those described in Patent Document 1, are difficult to use accurately because their measurement ranges tend to be unevenly distributed.

This study considers the case where the transported paper sheets have various creases and stiffnesses, similar to banknotes in an automated cash exchange machine. In this case, diagnostic methods based on signals during paper sheet transport are difficult to use accurately because the transport characteristics change due to differences in creases, stiffness, and friction coefficients on the paper sheet surface. Therefore, it is unclear what the correct criteria are, making accurate deterioration assessment challenging.

Therefore, the object of the present invention is to provide a paper sheet handling device and a maintenance method therefor that enable efficient maintenance work.

To solve the above problems, the paper sheet handling device according to the present invention is a paper sheet handling device for handling paper sheets, comprising: a paper sheet processing mechanism for transporting and processing paper sheets; a control unit for controlling the paper sheet processing mechanism; a sensor unit for detecting sound or vibration generated from the paper sheet processing mechanism; and an abnormality determination unit that determines an abnormality in the paper sheet processing mechanism using a machine learning model based on the control state of the paper sheet processing mechanism by the control unit and the sound or vibration generated from the paper sheet processing mechanism detected by the sensor unit.

According to the present invention, an abnormality in the paper sheet processing mechanism can be determined by a machine learning model based on the control state of the paper sheet processing mechanism and the sound or vibration generated from the paper sheet processing mechanism.

This is an overall schematic diagram showing the maintenance system for automated cash transaction machines. This is a diagram showing the configuration of the banknote processing mechanism. This is a diagram illustrating the configuration of an automated cash transaction system. This is an explanatory diagram of the flow of banknotes in a banknote processing facility. This diagram shows the configuration of the paper feed and storage unit. This is an explanatory diagram showing an example of how to install an anomaly detection device. This is a block diagram of the anomaly detection device. This is a block diagram of the sound collection device. This is an explanatory diagram showing an example of the operation information for each part. This is an explanatory diagram showing an example of a method for extracting sound data. This is an explanatory diagram showing an example of a method for extracting sound data. This is an explanatory diagram showing another example of a method for extracting sound data. This is a flowchart of the deposit counting process. This is a flowchart of the deposit collection process. This is a flowchart of the withdrawal process.

The embodiments of the present invention will be described below with reference to the drawings. In this embodiment, the paper sheet handling apparatus and its maintenance method will be described from the following perspectives.

(1) When an abnormality occurs during normal operation, sound data from the abnormality site is extracted based on the timing of the abnormality and the timing of the operation of each part. The cause of the abnormality in the abnormality site is then identified from this sound data using machine learning analysis.

(2) The deterioration state of a specific part is trained using machine learning based on the operating sounds of that part only. Then, in maintenance mode, only that specific part is operated, and the deterioration state of that part is analyzed using machine learning.

(3) By combining (1) and (2) above, signs of malfunctions and other abnormalities can be detected, leading to early maintenance.

A paper handling device based on the above perspective, for example, a paper handling device that transports paper sheets using conveying means such as rollers or belts, comprises: detection means for detecting the deterioration and malfunction status of rollers, conveying belts, and drive units; determination means for determining the deterioration and malfunction status of rollers, conveying belts, and drive units; and notification means for notifying maintenance timing of rollers, conveying belts, and drive units based on the determination means.

The detection means may detect the deterioration and malfunction state of the drive unit (and the transport means driven by the drive unit) from sound.

The detection means may detect the deterioration and failure state of the drive unit by using the sound generated from the drive unit and the control signal that controls the drive unit.

The decision-making mechanism may use machine learning to determine the sound detected by the detection mechanism.

Among the sounds detected by the detection means, machine learning may be used to determine which sounds correspond to the control signals that control the drive unit within a given time range.

According to this embodiment, the condition of the drive unit can be determined more accurately, resulting in improved efficiency and reduced costs for maintenance work.

The first embodiment will be explained using Figures 1 to 15. Figure 1 is an overall schematic diagram showing the maintenance system of the automated cash transaction device 1. Details of the automated cash transaction device 1, as an example of a "paper handling device," will be described later.

The automated cash transaction machine 1 incorporates a safe 14. At least a portion of the banknote processing mechanism 2 is located within the safe 14. The automated cash transaction machine 1 may be equipped with one or both of the multiple types of anomaly detection devices 110 and 120.

The safe 14 in the automated cash transaction system 1 is designed to prevent the theft of banknotes stored inside. In addition to its theft prevention function, the safe 14 in the maintenance system of the automated cash transaction system 1 also has a function to block or attenuate noise. As described later, by installing sensor units 111 and 121 inside the safe 14, it is possible to detect sound or vibration inside the safe 14, thereby suppressing the influence of sound or vibration from outside the safe 14.

One anomaly detection device is a fixed-type anomaly detection device 110, which is built into the ATM 1. The other anomaly detection device is a portable anomaly detection device 120, which can be retrofitted to the ATM 1. An example of the first anomaly detection device 110 is the anomaly detection device 15, which will be described later. An example of the second anomaly detection device 120 is the sound collection device 16, which will be described later.

The first anomaly detection device 110 comprises a sensor unit 111 and a determination unit 112. The determination unit 112 uses a machine learning model to determine whether an anomaly has occurred in the target part of the banknote processing mechanism 2, based on the detection signal from the sensor unit 111 and the control information from the control device 100.

The second anomaly detection device 120 comprises a sensor unit 121 and a data storage device 122. The data storage device 122 stores the detection signal from the sensor unit 121 and transmits the stored detection signal to the management server 120. The management server 120 determines whether an anomaly has occurred in the target part of the banknote processing mechanism 2 based on the detection signal received from the second anomaly detection device 120. The second anomaly detection device 120 may transmit the detection signal to the management server 130 while attached to the ATM 1, or it may transmit the detection signal to the management server 130 from a location away from the ATM 1 (e.g., a maintenance center).

The sensor units 111 and 121 detect sound or vibration, or both sound and vibration, and output them as electrical signals. The detectable frequency range is not limited. Any frequency range that can provide data to determine whether or not an abnormality has occurred in the target part of the banknote processing mechanism 2 is sufficient. In this specification, abnormality detection includes not only cases where an abnormality has occurred, but also cases where there are warnings that an abnormality may occur.

The first anomaly detection device 110 may include not only a sensor unit 111 installed inside the safe 14, but also other sensor units 111 (not shown) installed outside the safe 14. Similarly, the second anomaly detection device 120 may also include multiple sensor units 121, and at least one of these sensor units 121 may be installed outside the safe 14.

As described above, the sensor units 111 and 121 may consist of at least one or more sensors for detecting sound, or at least one or both of the sensors for detecting vibration.

An example of the control device 100 shown in Figure 1 is the control unit 11 and the main control device 13, which will be described later. The management server 130, the control device 100, and the second anomaly detection device 120 are connected via a communication network CN to enable bidirectional communication. The communication network CN may be a wide-area public communication network such as the Internet, or a dedicated communication network. The communication network CN may be wired or wireless.

Figure 2 will be used to explain the recirculating ATM 1. Figure 2 is a diagram showing the configuration of the banknote processing mechanism 2 in the recirculating ATM 1.

The recirculating automated cash transaction device 1 is an automated cash transaction device that reuses banknotes deposited by one user as disbursed banknotes to other users. The banknote processing mechanism 2 is a mechanism for processing banknotes deposited by users and banknotes disbursed to users.

The banknote processing mechanism 2 is equipped with a deposit/withdrawal slot 3 for users to deposit banknotes and receive withdrawn banknotes. The authentication unit 4 is located in the middle of the transport mechanism 9 and performs authentication and denomination determination of banknotes transported one by one from the deposit/withdrawal slot 3.

The temporary stacker 5 is a storage unit for temporarily holding deposited banknotes. The return unit 6 is a storage unit for storing deposited banknotes by denomination. The reject unit 7 is a storage unit for banknotes that are not to be used as disbursed banknotes. The loading and retrieval unit 8 is a storage unit used for loading and retrieving banknotes to the banknote processing mechanism 2. The transport mechanism 9 is a mechanism for transporting banknotes between these units 3, 4, 5, 6, 7, and 8.

In each of the above-mentioned units 2-8 and transport mechanism 9, a passage sensor 10 is provided at a predetermined location to detect whether or not the banknotes have been transported correctly. For example, an optical sensor is used as the passage sensor 10. The optical passage sensor 10 detects the passage of banknotes when the banknotes pass between the light-emitting side and the light-receiving side, and the transmission of light from the light-emitting side to the light-receiving side is blocked. It detects this blocking signal and thus detects the passage of the banknotes.

The banknote processing mechanism 2 includes a control unit 11 that controls the operation of each unit 2-8 and the transport mechanism 9 in the banknote processing mechanism in order to perform banknote deposit and withdrawal transactions. Signals from each passage sensor 10 are sent to the control unit 11 via the signal line 12.

As shown in Figure 3, the ATM 1 includes devices other than the banknote processing mechanism 2. These other devices include, for example, a passbook printer 1P, a display device 1D, and a card processing mechanism 1C. Furthermore, a main control device 13 is provided inside the ATM 1 to control the operation of the other devices 1P, 1D, and 1C. The main control device 13 controls the banknote processing mechanism 2 by communicating with the control unit 11 via a signal line 12.

The operation of the banknote processing mechanism 2 will be explained using Figure 4. Figure 4 is a cross-sectional view of the banknote processing mechanism 2 showing the flow of banknotes in sequence. The left side of Figure 4 shows the process during deposit, the center shows the process during deposit collection, and the right side shows the process during withdrawal. Details of the deposit, deposit collection, and withdrawal operations will be described later with flowcharts.

In Figure 4, banknotes inserted by the user during deposit are separated one by one at the deposit/withdrawal slot 3 and fed to the authentication unit 4. Banknotes that have been authenticated and their denomination determined in the authentication unit 4 are accumulated in the temporary stacker 5. Banknotes that cannot be authenticated or their denomination determined in the authentication unit 4 are transported back to the deposit/withdrawal slot 3 and returned to the user. At the branching point of the transport mechanism 9, a guide mechanism 9G (shown in Figure 2), called a gate, is provided to switch the direction of banknote transport (the above describes the deposit counting operation).

The amount counted by the authentication unit 4 is displayed on the display device 1D (shown in Figure 3) of the automated cash transaction device 1. When the user presses the confirmation button (not shown), the transaction is completed, and the banknotes accumulated in the temporary stacker 5 are dispensed one by one again, pass through the authentication unit 4 again, and are accumulated in the return storage unit 6 by denomination. At this time, banknotes of denominations that are not returned, and banknotes deemed unsuitable for withdrawal due to their poor condition, are accumulated in the reject storage unit 7 (shown in Figure 2) (this completes the deposit and storage operation).

When a user initiates a withdrawal, the number of banknotes corresponding to the amount specified by the user are separated one by one from the return storage unit 6, passed through the authentication unit 5, and transported to the deposit/withdrawal slot 3 for accumulation (this constitutes the withdrawal operation).

Figure 5 shows an example of the configuration of the return bank 6. The return bank 6 consists of a banknote storage section 6A for storing banknotes and a banknote dispensing section 6B for collecting and dispensing the stored banknotes.

The banknote dispensing unit 6B includes, for example, a stack/feed roller 61, a backup roller 62, a gate roller 63, a brush roller 64, a separation/stack guide 65, and a pickup roller 66.

As an example of a maintenance method for the automated cash transaction device 1, a method for determining the deterioration of the brush roller 64 of the return storage unit 6 will be described.

The brush roller 64 is mounted on the same axis as the gate roller 63, and multiple sheets are arranged radially as brushes on its outer circumference. By rotating the brush roller 64, banknotes stored continuously in the banknote storage section 6A can be accumulated without interference.

In the banknote processing mechanism 2, the elasticity of the brushes on the brush roller 64 deteriorates over time due to the handling, stacking, and transport of large quantities of banknotes. This decrease in brush elasticity reduces the banknote transport force, resulting in undesirable performance. Therefore, the brush roller 64 is treated as a regularly replaced part and is replaced periodically by maintenance personnel. However, since the replacement of the brush roller 64 is performed independently of the banknote transaction volume, it may be replaced even when unnecessary. Conversely, sometimes the brushes are not replaced until the scheduled replacement time, even if their elasticity has decreased, which can lead to poor banknote stacking.

As mentioned above, performing maintenance work regardless of the actual condition of the parts to be replaced (such as whether a malfunction is occurring or whether a malfunction is likely to occur in the near future) makes it difficult to improve the efficiency and reduce the cost of maintenance work.

Furthermore, overseas locations lack a robust maintenance service network and sufficient technical skills among their maintenance personnel. Therefore, there is a need to provide maintenance personnel with some form of work instruction when replacement work is truly necessary.

Therefore, this embodiment proposes an automated cash transaction device 1 that solves the above-mentioned problems, improves the efficiency and reduces the cost of maintenance work, and enables the issuance of work instructions to maintenance personnel only when replacement work is truly necessary.

In this embodiment, an anomaly detection device 15 is installed inside the automated cash transaction machine 1 shown in Figures 2 and 3. This device collects the operating sounds of the target parts of the automated cash transaction machine 1 and uses a machine learning model to predict anomalies and detect malfunctions.

Alternatively, a maintenance worker may bring a portable sound collection device 16 to the installation location of the ATM 1, and use a separate analysis device to analyze the data collected by the sound collection device 16 during maintenance work for predictive analysis and fault detection.

Figure 6 schematically shows an example of the installation of the anomaly detection device 15. By installing the anomaly detection device 15 inside the automated cash transaction machine 1, the influence of ambient noise around the ATM 1 can be reduced, and the characteristics of the sound inside the ATM 1 can be acquired more clearly.

The return compartment 6, reject compartment 7, and loading/recovery compartment 8 are each housed within the safe 14, along with the transport mechanisms 9 connected to these units 6-8, for storing banknotes. When collecting sound from the return compartment 6, reject compartment 7, and loading/recovery compartment 8, installing an abnormality detection device 15 inside the safe 14 reduces the impact of noise from external mechanisms within the automated cash transaction system 1.

The control unit 11 of the banknote processing mechanism 2, as described in Figure 2, continuously records the operation type of the banknote processing mechanism 2, whether or not an abnormality has occurred, and the operation information of each actuator and sensor, along with time information, in the recording area within the control unit 11. Figure 9 is an example of this operation information, showing the timing of operation of each part over time. Figure 9 shows the operation timing of multiple motors M1-M3, multiple solenoids SL1, SL2, and multiple sensors SR1-SR4. Therefore, if an abnormality is detected by the control unit 11, it is possible to immediately confirm which part was operating at the time of abnormality detection (immediately before or immediately after abnormality detection).

Figure 7 is an example block diagram showing the internal configuration of the anomaly detection device 15. The anomaly detection device 15 acquires sound during the operation of the banknote processing mechanism 2 using a microphone 15A and stores it in the data storage unit 15B. The arithmetic unit 15C has a data processing unit 15D and a machine learning model 15E. The machine learning model 15E is pre-trained by collecting various normal sound data and sound data for various failure causes for each actuator (motor, solenoid, etc.). Actuators are examples of "drive units," "target drive units," and "target parts." Sensors are examples of "target parts."

In degradation detection, sound data collected under normal conditions is trained using machine learning, and those with significant differences in characteristics are identified as being in a degraded state. In failure detection, since the sound characteristics differ depending on the cause of the failure, the sound for each cause is trained using machine learning, and the cause is classified accordingly.

Figure 8 is a block diagram showing the internal configuration of the sound collection device 16. The sound collection device 16 has a configuration similar to the anomaly detection device 15, but without the arithmetic unit 15C. Specifically, it comprises a microphone 16A and a data storage unit 16B.

The anomaly detection device 15 of this embodiment can perform the following multiple determination processes. In the first determination process, if an anomaly occurs during normal operation, the actuator operating at the time of the anomaly is identified from the time-series actuator operation information, and the cause of the failure is determined using a machine learning model corresponding to the identified actuator and collected sound data. In the second determination process, a dedicated degradation determination operation is performed on a specific actuator during maintenance work, and the degradation state is determined using a machine learning model for the degradation determination operation and collected sound data. These processes are described in detail below.

(First determination process) Failure cause determination during normal operation

Using the time information of the operation data of each actuator and sensor acquired by the banknote processing mechanism 2, the data processing unit 15D extracts the sound collection data corresponding to the operation time of the target actuator.

Figure 10 is a schematic diagram of the sound collection data extraction process. The upper part of Figure 10 shows the time-series information of the sound collection data. The lower part of Figure 10 shows the time-series information of the operation information of each actuator and the anomaly detection signal, corresponding to Figure 9.

In this example, there were no abnormalities after the operation of motors M2 and M3 and solenoids SL1 and SL2. Since an abnormality detection signal was output immediately after motor M1 operated, this shows that the sound data collected when only motor M1 was operating has been extracted.

The data processing unit 15D preprocesses the extracted data into a format suitable for machine learning. For example, it performs an FFT (Fast Fourier Transform) transformation on the data. The machine learning model 15E then determines the cause of failure for motor M1 based on the preprocessed data. Alternatively, the data preprocessed by the data processing unit 15D may be sent to the management server 130 shown in Figure 1, where the machine learning model may perform the failure determination.

If the machine learning model 15E determines that replacement is necessary, the display device 1D of the ATM 1 will display the part to be replaced, a message prompting replacement, and the replacement procedure. Alternatively, an electronic message indicating that replacement is necessary (including information identifying the part to be replaced and information identifying the ATM 1) may be sent to the management server 130 or the maintenance center (not shown). The anomaly detection device 15 may also send the above electronic message to an information terminal (not shown) held by a maintenance worker. The electronic message may be sent via email, SMS (Short Message Service), SNS (Social Networking Service), etc.

For ATMs 1 that do not have a built-in anomaly detection device 15, the first determination process can be performed using a sound collection device 16 carried by a maintenance worker. The maintenance worker manually operates the banknote processing mechanism 2 during maintenance. The sound of this operation is acquired by the sound collection device 16. The operation information of each actuator and sensor is transferred from the memory area of the control unit 11 to an external memory (not shown). The maintenance worker takes the sound collection device 16 and the external memory back to the maintenance center and performs the first determination process using an analysis device (not shown) at the maintenance center. Alternatively, the maintenance worker may perform the first determination process by connecting the analysis device, which is connected to the sound collection device 16 and the external memory, to the management server 130.

(Second Judgment Process) Deterioration Judgment During Maintenance Work

The second determination process involves determining the deterioration state of a specific part within the automated cash transaction machine 1. Here, the brush roller 64 is used as an example of the specific part, but the process is similar for other parts.

In determining the deterioration of the brush roller 64, the banknote processing mechanism 2 performs a dedicated deterioration detection operation in which only the actuator (brush roller motor) connected to the brush roller is rotated at a constant speed and in a constant direction.

Multiple recirculating chambers 6 are provided in a single banknote processing mechanism 2. A dedicated operation for deterioration detection is performed for each recirculating chamber 6. Using this dedicated operation results in data consisting only of the sound from one brush roller. By omitting the sounds from other actuators, it becomes easier to extract characteristic sounds.

Figure 11 shows the degradation detection process using a dedicated method. The upper part of Figure 11 shows the time series of the collected sound data. The lower part of Figure 11 shows the operating timing of one brush roller (the operating timing of one brush roller motor). Degradation detection is performed by extracting the sound data collected when the brush roller is operating.

In this example, sound frequency components are used to perform a determination using a machine learning model. When multiple actuators are operating simultaneously, the frequency components mix, making it difficult to distinguish which actuator's frequency components are being used. This is especially true when multiple brush rollers are operating, as the similar frequency components make identification even more difficult.

In this embodiment, high-precision degradation detection can be achieved by performing a dedicated operation in which each actuator is operated individually. As shown in Figure 12, since multiple identical actuators are operated one by one, it is possible to determine which actuator is producing the sound using a single abnormality detection device 15, without needing to provide an abnormality detection device 15 for each actuator.

If a malfunction occurs in any of the actuators in the banknote processing mechanism 2, maintenance personnel may not be able to determine whether the malfunction is in the actuator itself or due to other issues, such as a malfunction in the control unit, cables, or the presence of foreign matter. However, if information on the time of the malfunction is available, it is possible to extract sound data from before and after the malfunction occurred.

Figures 13 to 15 show flowcharts for each operation.

Figure 13 shows the processing of the deposit counting operation. In the deposit counting operation, when the control unit 11 receives an operation start signal from the main control unit 13, the transport mechanism motors M1/M2 each begin to rotate forward (step S101). Then, the solenoid of the guide mechanism is driven to the deposit counting route (step S102). The temporary stacker motor begins to rotate in the accumulation direction (step S103), and the identification unit 4 is activated (step S104).

The deposit/discharge port motor begins rotating in the separation direction, transporting the banknotes from the deposit/discharge port to the temporary stacker (step S105). Once the banknotes have been separated from the deposit/discharge port, the deposit/discharge port motor and the temporary stacker motor are stopped (steps S106, S107). Finally, the transport mechanism motors M1/M2 are stopped, ending the deposit counting operation (step S108).

Figure 14 shows the processing of the deposit and storage operation. During the deposit and storage operation, the transport mechanism motors M1/M3 begin to reverse (step S201). Then, the solenoid of the guide mechanism is driven to the deposit and storage route (step S202). The reject compartment motor and the return compartment motor begin to rotate in the accumulation direction (steps S203, S204), and the authentication unit 4 is activated (step S205). The temporary stacker motor begins to rotate in the separation direction, and the banknotes are transported from the temporary stacker 5 to the reject compartment 7 and the return compartment 6 (step S206).

Once the banknotes in temporary stacker 5 have been separated, the temporary stacker motor, reject box motor, and return box motor are stopped (steps S207, S208, S209). Finally, the transport mechanism motors M1/M3 are stopped, ending the deposit and storage operation (step S210).

Figure 15 shows the processing for the cash withdrawal operation. During the cash withdrawal operation, the transport mechanism motors M1/M2/M3 each begin rotating forward (step S301). Then, the solenoid of the guide mechanism is driven to the cash withdrawal route (step S302). The deposit/withdrawal port motor begins rotating in the accumulation direction (step S303), and the identification unit 4 is activated (step S304).

The recirculation compartment motor begins rotating in the separation direction, transporting the banknotes from the recirculation compartment 6 to the deposit/withdrawal slot 3 (step S305). Once the banknotes in the recirculation compartment 6 have been separated, the recirculation compartment motor and the deposit/withdrawal slot motor are stopped (steps S306, S307). Finally, the transport mechanism motors M1/M2/M3 are stopped, ending the withdrawal operation (step S308). Thus, in this banknote processing mechanism, the operation status, operating time, and operating pattern of each actuator differ depending on the type of operation.

Therefore, with the operation and time information of each actuator and sensor, it is possible to identify the actuator and type of operation that was operating immediately before the malfunction. Knowing the actuator and type of operation that was operating immediately before the malfunction allows for the identification of the learning model to be used in machine learning, thus enabling highly accurate fault cause identification.

The sound collection device 16 is installed in the ATM 1 during maintenance work, but the abnormality detection device 15 can perform fault detection (self-diagnosis) even during normal operation outside of maintenance work. The deterioration detection operation can be performed periodically, for example, during the downtime of the ATM 1. The cause of the abnormality can be determined immediately upon occurrence. With this configuration, it is possible to appropriately determine whether or not replacement of the target part (unit replacement) is necessary, thereby preventing unnecessary maintenance replacements and improving the efficiency and cost of maintenance work.

Furthermore, the present invention is not limited to the embodiments described above. Those skilled in the art can make various additions and modifications within the scope of the present invention. The features described in the embodiments can also be combined as appropriate.

1: Automatic cash transaction machine, 2: Banknote processing mechanism, 3: Deposit/withdrawal slot, 4: Authentication unit, 5: Temporary stacker, 6: Return vault, 7: Reject vault, 8: Loading/recovery vault, 9: Transport mechanism, 10: Pass sensor, 11: Control unit, 13: Main control unit, 14: Safe,
15: Anomaly detection device, 16: Sound collection device, 100: Control device, 110: Anomaly detection device, 120: Anomaly detection device, 130: Management server

Claims (6)

A paper sheet handling device for handling paper sheets,
A paper sheet processing mechanism that transports and processes paper sheets,
A control unit for controlling the paper sheet processing mechanism,
A sensor unit for detecting sound or vibration generated from the paper sheet processing mechanism,
The system includes an abnormality determination unit that determines an abnormality in the paper processing mechanism using a machine learning model, based on the control state of the paper processing mechanism by the control unit and the sound or vibration generated from the paper processing mechanism detected by the sensor unit .
The control unit drives one target drive unit at a time from among the multiple drive units included in the paper sheet processing mechanism.
The sensor unit detects sound or vibration generated from the target drive unit,
The abnormality determination unit determines an abnormality in the target drive unit using the machine learning model, based on the control state of the target drive unit by the control unit and the sound or vibration generated from the target drive unit detected by the sensor unit.
Paper sheet handling device. The aforementioned paper sheet processing mechanism has a plurality of drive units, and some of these drive units are provided inside the safe as target drive units.
The control unit drives each of the target drive units one by one.
The sensor unit is installed inside the safe and detects sound or vibration generated from the target drive unit.
The paper handling apparatus according to claim 1, wherein the abnormality determination unit determines an abnormality in the target drive unit using the machine learning model based on the control state of the target drive unit by the control unit and the sound or vibration generated from the target drive unit detected by the sensor unit. The paper sheet handling device according to claim 1, wherein the paper sheet processing mechanism is a banknote processing mechanism for processing banknotes. The paper handling device according to claim 3, wherein the control unit uses the banknote processing mechanism to perform deposit counting, deposit storage, and withdrawal operations. A method for maintaining paper sheet handling equipment that handles paper sheets,
The sensor unit detects sounds or vibrations generated from the paper sheet processing mechanism that transports and processes paper sheets.
The control state of the paper sheet processing mechanism by the control unit is acquired.
Based on the sound or vibration generated from the paper processing mechanism detected by the sensor unit and the acquired control state, the abnormality determination unit uses a machine learning model to determine if there is an abnormality in the paper processing mechanism.
The control unit drives one target drive unit at a time from among the multiple drive units included in the paper sheet processing mechanism.
The sensor unit detects sound or vibration generated from the target drive unit,
The abnormality determination unit determines an abnormality in the target drive unit using the machine learning model, based on the control state of the target drive unit by the control unit and the sound or vibration generated from the target drive unit detected by the sensor unit.
Maintenance method for paper handling equipment. The aforementioned paper sheet processing mechanism has a plurality of drive units, and some of these drive units are provided inside the safe as target drive units.
The sensor unit is installed inside the safe and detects sound or vibration generated from the target drive unit.
The maintenance method for a paper handling device according to claim 5, wherein the abnormality determination unit determines an abnormality in the target drive unit using the machine learning model based on the control state of the target drive unit by the control unit and the sound or vibration generated from the target drive unit detected by the sensor unit.