JP7127301B2 - ABNORMAL SENSOR DETECTION METHOD, ABNORMAL SENSOR DETECTION SYSTEM AND INFORMATION PROCESSING DEVICE - Google Patents

Description

The present invention relates to an abnormality sensor detection technology.

In recent years, the Internet of Things (IoT) has received increasing attention. In order to confirm the safety of patients in hospitals, there are cases where IoT sensors are installed at various locations in hospitals and data such as when patients get out of bed are collected. In the future, if a large number of sensors that use networks are introduced, maintenance costs will be required depending on the number of sensors installed, and maintenance costs are expected to rise sharply according to the number of sensors installed.

There is a technique for storing output signals of sensors provided in a controlled object in chronological order, extracting the correlations inherent in them, and quantifying the degree of deviation from the normal correlations. The technology determines whether there is an abnormality in the output of each sensor from the gap in the correlation quantified for each sensor pair, and identifies the sensor outputting an abnormal value (see, for example, Patent Document 1).

In the sensor soundness determination method, an accident event is estimated using the operation signal of the equipment, a sensor group is set for comparing correlation appraisal based on the accident event, and the sensor correlation is compared using the set sensor group. A technique for determining the soundness of a sensor is known (see Patent Literature 2, for example).

JP-A-7-110708 WO2015/040683

In a field where a large number of sensors are used, it is easier to introduce inexpensive sensors from the cost point of view. For example, depending on the sensor, in addition to the function of communicating sensor data to a management system (for example, an information processing device that collects sensor data), there is a communication function that detects an abnormality that has occurred in the sensor and notifies it to the management system. Some have separate items. However, inexpensive sensors do not necessarily have a sensor abnormality detection function or a communication function for notifying an abnormality to a management system. Therefore, even if a large number of such inexpensive sensors are installed, it is required to find an abnormal sensor. Furthermore, in addition to detecting anomalies in the management system, it is conceivable to manually detect anomalies in each sensor, but this method imposes a heavy workload.

Accordingly, in one aspect of the present invention, it is an object of an information processing apparatus to detect a sensor in which an abnormality has occurred, even if the sensor cannot notify the occurrence of an abnormality.

A computer acquires sensor data obtained by a plurality of sensors via a network. After that, the computer detects the detection target based on the order in which each sensor among the plurality of sensors detects the detection target, and the sensor that detects the detection target first in the order of the two sensors detects the detection target. the probability that the sensor whose order is later among the two sensors also detects the detection target when the sensor is detected corresponds to the identification information of each of the sensor whose order is first and the sensor whose order is later Then, relationship information indicating relationships between sensors is generated for each of a plurality of time ranges. The computer detects a sensor in which an abnormality has occurred among the plurality of sensors based on changes in the relationship information for each of the plurality of time ranges , When a plurality of abnormal sensors are detected based on a change, and in the relationship information after the change, one of the plurality of abnormal sensors in which the occurrence of the abnormality is detected as the subsequent sensor If the previous sensor indicated by the relationship between all the sensors that are present are all included in the plurality of abnormal sensors, the one of the plurality of abnormal sensors is the abnormal sensor is determined to have not occurred .

According to the present invention, an information processing apparatus can detect a sensor in which an abnormality has occurred, even if the sensor cannot notify the occurrence of an abnormality.

It is a figure explaining an example of the sensor installed in the hospital, and an operation|movement of a patient. It is a figure explaining the example of the method of detecting an abnormal sensor from the sensor installed in the hospital. It is an example of the block diagram explaining the information processing apparatus which performs the process which detects an abnormal sensor. It is a figure explaining the example of the relationship information table used for the process of a relationship information production|generation part. FIG. 10 is a diagram illustrating an example of processing by a relationship monitoring unit; It is a figure explaining the example which an abnormal sensor detection part detects an abnormal sensor. It is a figure explaining the example of the data processing of a relationship information production|generation part. It is a figure explaining the example of the data processing of a relationship information production|generation part. It is a figure explaining the example of the hardware constitutions of the information processing apparatus which concerns on this embodiment. 4 is a flowchart illustrating an example of processing of an information processing device that detects an abnormal sensor; It is a figure explaining the example of the threshold value set with respect to the variation|change_quantity of reliability. It is a figure explaining the example of the method of dividing into a segment and managing a sensor. FIG. 4 is a flow chart illustrating an example of a method for managing sensors by segmentation; FIG. FIG. 4 is a flow chart illustrating an example of a method for managing sensors by segmentation; FIG. It is a figure explaining the example of the process which determines an abnormal chain. FIG. 11 is a flowchart illustrating an example of processing for determining a chain of anomalies; FIG. FIG. 11 is a flowchart illustrating an example of processing for determining a chain of anomalies; FIG. It is a figure explaining the example of relationship collapse in flow line change. It is a figure explaining the example of the firing rate distribution in a flow line change. It is a figure explaining the example of installation of the sensor in embodiment corresponding to a flow line change.

Hereinafter, this embodiment will be described in detail with reference to the drawings.
FIG. 1 is a diagram illustrating an example of motions of a sensor and a patient installed in a hospital. In the example of FIG. 1, a door sensor 101, a human sensor 102, a sound sensor 103, and a bed sensor 104 are installed in order to acquire a series of motion data from when the patient passes through the corridor, opens the door, and enters the hospital bed. It is The door sensor 101 is installed above the door, for example, and detects opening and closing of the door. The human sensor 102 is installed with its detection direction directed toward the corridor, and detects the passage of a person in the corridor. The sound sensor 103 is installed around the door and detects the sound associated with the opening and closing of the door. The bed sensor 104 is installed in the mat of the bed or on the bed frame, and detects that the patient has entered the bed based on the load when the patient enters the bed. Therefore, each of the door sensor 101, the human sensor 102, the sound sensor 103, and the bed sensor 104 outputs a sensor value at the timing when the self-machine detects the detection target.

Here, a series of actions from a patient passing through a corridor, opening the door of his or her hospital room, to entering the bed, and the relationship between each sensor will be explained. First, when the patient passes through the corridor, the human sensor 102 detects that the patient has passed through the corridor. Next, when the patient opens the door, the door sensor 101 detects that the door has been opened. In addition, the sound sensor 103 detects the sound when the door is opened. When the patient is in bed, the bed sensor 104 detects that the patient is in bed based on the weight of the patient.

In this way, the patient passes through the corridor, opens the door of his/her hospital room, and enters the bed (hereinafter referred to as a "series of actions"). The sensor 103 and the bed sensor 104 sequentially detect motion. Therefore, each detection of the human sensor 102, the door sensor 101, the sound sensor 103, and the bed sensor 104 is related to the patient's motion.

FIG. 2 is a diagram illustrating an example of a method for detecting an abnormal sensor from sensors installed in a hospital. A normal state 210 is the human sensor 102 and the door sensor 102 in which an abnormality such as a failure or a dead battery has not occurred in a series of operations from the patient in FIG. 101, the sound sensor 103, and the bed sensor 104 are illustrated.

A graph 211 in the normal state 210 shows sensor values on the vertical axis and time (t) on the horizontal axis. A graph 211 shows that the human detection sensor 102, the door sensor 101, the sound sensor 103, and the bed sensor 104 detect the detection targets for a series of actions of the patient passing through the corridor, opening the door of his hospital room, and getting into the bed. The sensor values indicating that are output in order.

A relationship diagram 212 in the normal state 210 is a diagram showing the relationship between the sound sensor 103, the human sensor 102, the door sensor 101, and the bed sensor 104 in a series of motions of the patient. In the relationship diagram 212, an arrow from the door sensor 101 to the sound sensor 103 indicates the relationship that the generation of the sound is detected after detecting that the door is opened. An arrow from the human sensor 102 to the sound sensor 103 indicates the relationship in which the generation of sound is detected after the passage of a person is detected. An arrow from the sound sensor 103 to the bed sensor 104 indicates a relationship in which it is detected that the patient has entered the bed due to the weight applied to the bed after the door opening/closing sound is detected. In this way, the example of the relationship diagram 212 shows the detection order (time series) and the relationship of each sensor, with the sound sensor 103 at the center.

A failure state 220 is an example of a state in which the sound sensor 103 fails after a lapse of time from the normal state 210 . That is, the failure state 220 is a diagram showing the relationship between the failed sound sensor 103 and the motion sensor 102, door sensor 101, and bed sensor 104 in a series of patient actions.

A graph 221 in the fault state 220 shows the sensor value on the vertical axis and the time (t) on the horizontal axis. Since the sound sensor 103 is out of order in the failure state 220, the graph 221 does not output a sensor value indicating that the sound sensor 103 is detecting the sound of the door. Therefore, the graph 221 sequentially outputs sensor values indicating that the three sensors, ie, the human sensor 102, the door sensor 101, and the bed sensor 104, detect the detection target in response to a series of movements of the patient.

In this embodiment, the sensor value is "1" when the detection target is detected and "0" when the detection target is not detected, based on the result of comparison between the sensor data, which is the value detected by the sensor, and a preset threshold value. result. Note that the threshold is set to a different value depending on the type of each sensor.

For example, if the sensor is a sound sensor, the sensor data is a value (decibels) corresponding to the loudness of the sound. In the case of the sound sensor, when the sensor data is greater than a predetermined threshold, the sensor value becomes a value (for example, 1) indicating that the detection target is detected, and when the sensor data is equal to or less than the threshold, the detection target is detected. It becomes a value (eg, 0) indicating that there was not. Moreover, when the sensor is a door sensor, the sensor data may be a value indicating the degree of opening of the door, or whether or not the door is opened (1 or 0). In the latter case, the sensor data and sensor values for the door sensor will be the same.

A relationship diagram 222 in the failure state 220 is a diagram showing the relationship between the sound sensor 103 and the human sensor 102, the door sensor 101, and the bed sensor 104 in a series of operations when the sound sensor 103 fails. In the fault state 220, the human sensor 102 detects that the patient has passed through the corridor, the door sensor 101 detects that the door has been opened, and the bed sensor 104 detects that the patient has entered the bed.

On the other hand, in the failure state 220, since the sound sensor 103 is out of order, the door opening sound is not detected. Therefore, in the failure state 220, the relationship between the sound sensor 103 and other sensors is lost. Hereinafter, a state in which the relationship in the normal state 210 is no longer related to the fault state 220 will be referred to as a "lost relationship" state. Since the relationship diagram 222 represents a state in which the relationship between the sound sensor 103 and other sensors is broken, arrows representing the relationships between the sound sensor 103 and other sensors are indicated by dotted lines.

The abnormal sensor detection device according to the present embodiment obtains the relationship with other sensors for all sensors to be communicated, and changes the relationship from the relationship in the normal state 210 to the state in which the relationship is broken like the failure state 220. The abnormal sensor detection device can detect the abnormal sensor by detecting the abnormal sensor.

FIG. 3 is an example of a block diagram illustrating an information processing device that performs processing for detecting an abnormal sensor. The information processing device 300 includes a data acquisition unit 310 , a relationship information generation unit 320 , a relationship monitoring unit 330 , an abnormal sensor detection unit 340 , a notification unit 350 and a storage unit 360 .

The data acquisition unit 310 acquires sensor response time information, sensor values, and sensor identification information from a plurality of sensors installed on site via a network. The storage unit 360 stores the acquired information.

The relationship information generation unit 320 uses the data acquired by the data acquisition unit 310 to create association rules for each sensor with other sensors. The correlation rule relates to the tendency of the detection order of the detection target in each sensor, but this is the detection order of the detection target by the sensor from the nature of the target that can be detected by each sensor (for example, sound and vibration) and the positional relationship. It focuses on the fact that regularity can be found in the tendency of For example, an association rule refers to a rule of combination of sensors such that when sensor A reacts, sensor B reacts. The relationship information generation unit 320 calculates the reliability (relationship) for each association rule, and stores the reliability value for the association rule in the storage unit 360 as relationship information. For each sensor, the relationship monitor 330 periodically monitors the reliability of all association rules in which that sensor is included.

The abnormal sensor detection unit 340 detects a sensor as an abnormal sensor when the amount of change in the reliability of a certain sensor and all association rules including the sensor is larger than a predetermined threshold. The notification unit 350 notifies an administrator or the like of information related to the abnormal sensor. Information related to the abnormal sensor is notified to registrants pre-registered in the information processing apparatus 300, for example. The registrant may be, for example, a system administrator or, in the example of a hospital, a nurse station. Information related to the abnormal sensor to be notified includes identification information for identifying the abnormal sensor, location information where the sensor is installed, and the like.

As described above, the information processing apparatus 300 according to the present embodiment detects an abnormality occurring in the sensor based on the information transmitted to the information processing apparatus 300 using the communication function for communicating the detection result of the sensor to the information processing apparatus 300. can do. That is, even if the sensor does not have a dedicated communication function for notifying the sensor information processing device 300 (an example of a management system) of an abnormality, the information processing device 300 can detect that the sensor is abnormal. . Therefore, the information processing apparatus 300 can detect the presence or absence of an abnormal sensor more efficiently than manually discovering an abnormal sensor.

Hereinafter, the processing of the relationship information generating unit 320, the relationship monitoring unit 330, and the abnormal sensor detecting unit 340 will be explained more specifically with reference to FIGS. 4 to 6. FIG.

FIG. 4 is a diagram illustrating an example of a relationship information table used for processing by the relationship information generation unit. The relationship information generation unit 320 uses the data acquired by the data acquisition unit 310 to create association rules for each sensor with other sensors. After that, the relationship information generation unit 320 calculates the reliability of the association rule, and stores the relationship information in which the association rule and the calculated reliability value are associated with each other in the storage unit 360 .

The relationship information table 410 is a table in which association rules and reliability values calculated by the relationship information generation unit 320 are associated with each other. The reliability is a probabilistic relationship between items, for example, "the probability that sensor B will react when sensor A reacts". A correlation rule that if sensor A responds, sensor B also responds is expressed as, for example, {A⇒B}. The degree of support (supp) of an association rule can be calculated using Equation 1 below. Support is the importance of data in association rules.

In Equation 1, σ represents standard deviation and M represents all transactions in the sensor data stored in storage unit 360 . A transaction is a unit of processing for a set of related data. Formula 1 is a formula for calculating the support (supp) from the sensor X to the sensor Y. The degree of support (supp) from sensor X to sensor Y can be calculated by σ(X∪Y)/M. That is, σ(X∪Y)/M, which is the support (supp), is the probability of (σ(X∪Y)) transactions including sensor X and sensor Y among all transactions (M).

Next, the confidence (conf) of the association rule can be calculated using Equation 2 below.

In Equation 2, σ represents standard deviation. Equation 2 is an equation for calculating the reliability (conf) from sensor X to sensor Y. The reliability (conf) from sensor X to sensor Y can be calculated by σ(X∪Y)/σ(X). That is, the reliability (conf) from sensor X to sensor Y is the probability of a transaction that includes sensor X and sensor Y (σ(X∪Y)) among all transactions, and the probability that sensor X is included in all transactions. It can be calculated by dividing by the probability of including (σ(X)) transactions.

The relationship information table 410 includes items of condition part, conclusion part, support, and reliability. The condition part and the conclusion part represent a combination of sensors corresponding to the association rule among all the sensors managed by the information processing apparatus 300, and hold sensor identification information. The condition part and the conclusion part have a correlation between sensors such as {condition part⇒conclusion part}. Therefore, for example, the first row of the relationship information table 410 states that "the support level of the case detected by the sound sensor 103 (conclusion part) after the detection by the door sensor 101 (condition part) is 0.6, and the reliability is 1 is.

FIG. 5 is a diagram illustrating an example of processing by the relationship monitoring unit. The relationship information table 410a is a table containing the same values as the relationship information table 410 in FIG. The relational diagram 420a represents the relation with the sound sensor 103 of the relational information table 410a at time _T0 whose reliability is equal to or higher than a predetermined value. In addition, below, the case where reliability is more than a predetermined value is called "there is relationship." Also, depending on the reliability calculation method, it may be determined that there is a "relationship" when the reliability is equal to or less than a threshold.
The arrow from the door sensor 101 to the sound sensor 103 in the relationship diagram 420a indicates that the reliability of the case detected by the sound sensor 103 (conclusion part) after the detection by the door sensor 101 (condition part) in the first row of the relationship information table 410a is 1”, indicating that there is a relationship. The arrow from the human sensor 102 to the sound sensor 103 in the relationship diagram 420a indicates the case where the sound sensor 103 (conclusion part) detects after the human sensor 102 (condition part) detects it in the third row of the relationship information table 410a. is 0.8”, indicating that there is a relationship. The arrow from the bed sensor 104 to the sound sensor 103 in the relationship diagram 420a indicates the reliability of the case detected by the sound sensor 103 (conclusion part) after the detection by the bed sensor 104 (condition part) in the fourth row of the relationship information table 410a. degree is 0.8", indicating that there is a relationship.

The relationship information table 410b shows the reliability corresponding to the same combination of sensors (combination of sensors in the condition part and the conclusion part) as in the relationship information table 410a at time T1 after _a predetermined time Δt has elapsed from time _T0 . It is a table that holds In the relationship information table 410b, since the sound sensor 103 is in a failed state, the reliability from the door sensor 101 to the sound sensor 103 (first row), the reliability from the human sensor 102 to the sound sensor 103 (3 line), and the reliability from the bed sensor 104 to the sound sensor 103 (fourth line) is zero. In the relationship diagram 420b, each arrow in the relationship diagram 410a is a dotted line representing that the relationship has collapsed.

Here, the relationship monitoring unit 330 periodically monitors the reliability of each sensor with other sensors between time T ₀ and time T ₁ .

The abnormal sensor detection unit 340 detects that the difference in reliability (Δconf) between the relationship information table 410b and the relationship information table 410a changes per unit time (Δconf/Δt) when it is larger than a predetermined threshold. 103 is detected as an abnormal sensor.

FIG. 6 is a diagram illustrating an example in which an abnormal sensor detection unit detects an abnormal sensor. A relationship diagram 510 represents a state in which the relationships between the sound sensor 103, the door sensor 101, the motion sensor 102, and the bed sensor 104 are broken. Therefore, the arrows between the sound sensor 103, the door sensor 101, the human sensor 102, and the bed sensor 104 in the relationship diagram 510 are indicated by dotted lines. The abnormal sensor detection unit 340 detects the sound sensor 103 as an abnormal sensor when the amount of change per hour in the reliability of all related sensors is greater than a predetermined threshold.

A relationship diagram 520 represents a state in which the relationship between the sound sensor 103 and the door sensor 101 is broken. Therefore, the arrow between the sound sensor 103 and the door sensor 101 in the relationship diagram 510 is represented by a dotted line. On the other hand, the sound sensor 103, the human sensor 102, and the bed sensor 104 are related. The abnormal sensor detection unit 340 does not determine the sensor as an abnormal sensor when there is a sensor whose reliability per unit time change amount for related sensors is smaller than a predetermined threshold value.

A relationship diagram 530 represents a state in which the relationships among the sound sensor 103, the door sensor 101, and the human sensor 102 are broken. Therefore, the arrows between the sound sensor 103, the door sensor 101, and the human sensor 102 in the relationship diagram 510 are indicated by dotted lines. On the other hand, the sound sensor 103 and the bed sensor 104 are in a state of being related. The abnormal sensor detection unit 340 does not determine the sensor as an abnormal sensor when there is a sensor whose reliability per unit time change amount for related sensors is smaller than a predetermined threshold value.

7A and 7B are diagrams for explaining an example of data processing by the relationship information generation unit. A table 600 in FIG. 7A shows data from 10:00 to 10:00:41 on July 28, 2017 acquired by the data acquisition unit 310 of the information processing device 300 from the door sensor 101, the motion sensor 102, the sound sensor 103, and the bed sensor 104. This is an example of data between seconds. In the table 600, a value "1" indicating that the detection target is detected at the time when the detection target is detected by each of the door sensor 101, the human sensor 102, the sound sensor 103, and the bed sensor 104 is input.

The relationship information generation unit 320 identifies sensors having a value indicating detection within, for example, three seconds from the time when the sound sensor 103 detected the detection target, and other sensors (correlation rules) that are related to the sound sensor 103. do. In this way, a processing unit of a series of related data used when specifying a sensor related to a certain sensor is called a transaction. In the table 600, the relationship information generator 320 identifies other sensors that are related to the sound sensor 103 in the following transactions (1) to (7).

(1) {102⇒103}, {101⇒103}, {104⇒103}
(2) {101⇒103}, {102⇒103}, {104⇒103}
(3) {102⇒103}, {101⇒103}, {104⇒103}
(4) {103⇒104}
(5) {101⇒103}, {102⇒103}
(6) {104⇒103}, {101⇒103}
(7) {103⇒102}

Next, the relationship information generation unit 320 calculates reliability for each combination of the sound sensor 103 and other related sensors. The reliability of {101⇒103} can be calculated by the number of transactions including both the door sensor 101 and the sound sensor 103/the number of transactions including the door sensor 101 . In table 600, it is 5/5 and the reliability is 1. The reliability of {102⇒103} can be calculated by the number of transactions including both the human sensor 102 and the sound sensor 103/the number of transactions including the human sensor 102 . In table 600, it is 4/5 and the reliability is 0.8. The reliability of {104⇒103} can be calculated by the number of transactions including both the bed sensor 104 and the sound sensor 103/the number of transactions including the bed sensor 104 . In table 600, it is 4/5 and the reliability is 0.8.

Table 610 in FIG. 7B is an example of data from 10:00 to 10:00:41 on July 29, 2017, the next day of table 600 in FIG. 7A. In the table 610, the sound sensor 103 did not detect anything between 10:00 and 10:00:41 on July 29, 2017. Therefore, the relationship information generation unit 320 cannot specify any of the relationships {101⇒103}, {102⇒103}, and {104⇒103} from the table 610 . Then, the relationship information generation unit 320 calculates the reliability of {101⇒103}, {102⇒103}, and {104⇒103} as zero.

The abnormal sensor detection unit 340 determines whether or not the amount of change in reliability per time of {101⇒103}, {102⇒103}, and {104⇒103} is greater than a predetermined threshold. The amount of change in reliability per hour {101⇒103} is obtained by (Δconf/Δt=1/1 day) when the timing of monitoring the reliability of the relationship monitoring unit 330 is every day. The amount of change in reliability per time {102⇒103} is obtained by (Δconf/Δt=0.8/1day) when the timing of monitoring the reliability of the relationship monitoring unit 330 is every day. The amount of change in reliability per time {104⇒103} is obtained by (Δconf/Δt=0.8/1day) when the timing of monitoring the reliability of the relationship monitoring unit 330 is every day. The abnormal sensor detection unit 340 detects the sound sensor 103 as an abnormal sensor because the amount of change in the reliability of the combination of the sound sensor 103 and all other related sensors is greater than a predetermined threshold.

As described above, the information processing apparatus 300 according to the present embodiment detects an abnormality occurring in the sensor based on the information transmitted to the information processing apparatus 300 using the communication function for communicating the detection result of the sensor to the information processing apparatus 300. can do. That is, even if the sensor does not have a dedicated communication function for notifying the sensor information processing device 300 (an example of a management system) of an abnormality, the information processing device 300 can detect that the sensor is abnormal. . Therefore, the information processing apparatus 300 can detect the presence or absence of an abnormal sensor more efficiently than manually discovering an abnormal sensor.

FIG. 8 is a diagram illustrating an example of the hardware configuration of the information processing apparatus according to this embodiment. The information processing device 300 includes a processor 11 , a memory 12 , a bus 15 , an external storage device 16 and a network connection device 19 . Furthermore, as an option, the information processing device 300 may include an input device 13 , an output device 14 and a media drive device 17 . The information processing device 300 may be realized by, for example, a computer.

Processor 11 may be any processing circuit including a Central Processing Unit (CPU). Note that the processor 11 implements each function of the relationship information generation unit 320 , the relationship monitoring unit 330 , and the abnormal sensor detection unit 340 by executing programs stored in the external storage device 16 , for example. The memory 12 operates as the storage unit 360 and stores the relationship information table 410 and the data acquired by the data acquisition unit 310 . Furthermore, the memory 12 appropriately stores data obtained by the operation of the processor 11 and data used for processing by the processor 11 . The network connection device 19 is used for communication with other devices, operates as a data acquisition unit 310 and a notification unit 350, and is used for communication with other devices and mobile communication terminals.

The input device 13 is implemented as, for example, a button, keyboard, mouse, etc., and the output device 14 is implemented as a display or the like. The bus 15 connects the processor 11, the memory 12, the input device 13, the output device 14, the external storage device 16, the media drive device 17, and the network connection device 19 so that data can be exchanged between them. The external storage device 16 stores programs, data, and the like, and appropriately provides the stored information to the processor 11 and the like. The media drive device 17 can output data in the memory 12 and the external storage device 16 to the portable storage medium 18 and can read programs, data, and the like from the portable storage medium 18 . Here, the portable storage medium 18 is any portable storage medium including a floppy disk, a Magnet-Optical (MO) disk, a Compact Disk Recordable (CD-R), and a Digital Versatile Disk Recordable (DVD-R). It can be used as a medium.

FIG. 9 is a flowchart illustrating an example of processing of an information processing device that detects an abnormal sensor. The data acquisition unit 310 periodically acquires sensor data from a sensor to be managed (step S101). Further, sensor values are generated based on the sensor data.

The storage unit 360 stores the acquired sensor data. The relationship information generator 320 identifies relationship rules using the acquired sensor data (step S102).

The relationship information generator 320 selects one sensor _Si from the sensors to be managed (step S103). i is an index, which is a number assigned to the sensor to be managed. The relationship information generator 320 calculates the reliability for each association rule including the sensor _Si (step S104). The relationship monitoring unit 330 compares the amount of change in reliability per hour with a predetermined threshold (step S105). Based on the comparison in step S105, the abnormal sensor detection unit 340 determines whether the relationship is broken (step S106). If the relationship is broken (YES in step S106), the abnormal sensor detection unit 340 detects sensor _Si as an abnormal sensor (step S110). The notification unit 350 notifies the registrant pre-registered in the information processing apparatus 300 of the information related to the abnormal sensor (step S111). The registrant may be, for example, a system administrator or, in the example of a hospital, a nurse station. Information related to the abnormal sensor to be notified includes identification information for identifying the abnormal sensor, location information where the sensor is installed, and the like.

After the process of step S111 or when there is no relationship collapse (NO in step S106), the relationship information generation unit 320 determines whether or not all sensors have been selected in the process of step S103 (step S107).

If all the sensors have not been selected (NO in step S107), the relationship information generator 320 adds 1 to the index i and returns the process to step S103. In the process of step S103, the relationship information generation unit 320 selects the sensor _Si+1 that has not yet been selected. If all sensors have been selected (YES in step S107), the information processing apparatus 300 ends the abnormal sensor detection process.

As described above, the information processing apparatus 300 according to the present embodiment detects an abnormality occurring in the sensor based on the information transmitted to the information processing apparatus 300 using the communication function for communicating the detection result of the sensor to the information processing apparatus 300. can do. That is, even if the sensor does not have a dedicated communication function for notifying the sensor information processing device 300 (an example of a management system) of an abnormality, the information processing device 300 can detect that the sensor is abnormal. . Therefore, the information processing apparatus 300 can detect the presence or absence of an abnormal sensor more efficiently than manually discovering an abnormal sensor.

In step S102, the relationship information generation unit 320 identifies association rules using all of the acquired sensor data. Here, the data processing amount of the relationship information generation unit 320 may be reduced by not processing sensor data such as hours when the hospital is not open due to business hours or days of the week, holidays, and the like.

When the sensor managed by the information processing device 300 is worn by a person, the relationship information generation unit 320 may specify the association rule on a person-by-person basis. By specifying the association rule for each person in this way, the data processing amount of the relationship information generation unit 320 may be reduced.

When the relationship information generation unit 320 specifies the association rule, the data processing amount of the relationship information generation unit 320 is reduced by classifying the sensor combinations for each weather-related variable such as weather, temperature, and humidity. may

<Others>
FIG. 10 is a diagram illustrating an example of threshold values set for the amount of change in reliability. The relationship monitoring unit 330 according to this embodiment may change the threshold used for comparison with the amount of change in reliability using part of the amount of change in relationship information. A relationship diagram 710 is a diagram showing an example of the relationship between the sound sensor 103, the door sensor 101, the human sensor 102, and the bed sensor 104 at time t-2. In the relationship diagram 710, the reliability from the door sensor 101 to the sound sensor 103 is 1, the reliability from the human sensor 102 to the sound sensor 103 is 0.8, and the reliability from the bed sensor 104 to the sound sensor 103 is 0.8. is 0.8. Here, the threshold Th _{t-2 used by the relationship monitoring unit 330 for comparison with the amount of change in reliability at time t-2} is 0.5.

Next, a relationship diagram 720 is a diagram showing an example of the relationship between the sound sensor 103, the door sensor 101, the human sensor 102, and the bed sensor 104 at time t-1. In the relationship diagram 720, the reliability from the door sensor 101 to the sound sensor 103 is 0.8, the reliability from the human sensor 102 to the sound sensor 103 is 0.7, and the reliability from the bed sensor 104 to the sound sensor 103 is 0.7. Reliability is 0.6. Here, the threshold Th _{t-1 used by the relationship monitoring unit 330 for comparison with the amount of change in reliability at time t-1} is 0.5.

The relationship monitoring unit 330 changes a part of the amount of change in the relationship information between the time t-2 and the time t-1, that is, the threshold. Specifically, first, the relationship monitoring unit 330 calculates the average value (Th _ADD ) of the amount of change in reliability between the time t−2 and the time t−1 based on Equation (3). After that, the relationship monitoring unit 330 subtracts a portion (for example, 10%) of the calculated average value of the reliability change amount from the previous threshold Th _t−1 as shown in Equation 4, thereby obtaining the threshold is updated from time to time.

In this manner, by updating the threshold as needed and using the changed threshold thereafter, it is possible to improve the detection accuracy of the abnormal sensors of the relationship monitoring unit 330 and the abnormal sensor detection unit 340 .

Further, the information processing device 300 may handle sensors within a predetermined distance as a group. FIG. 11 is a diagram illustrating an example of a method of managing sensors for each segment. A segment is a concept for handling sensors within a predetermined distance as a group. In the following, an example of forming a group simply by treating two adjacent areas as one segment and sensors existing close to each other as a group will be shown, but the present invention is not limited to this.

In FIG. 11, common area A, hospital room a, common area B, hospital room b, common area C, and treatment room c are illustrated. The information processing apparatus 300 divides and manages each area such as a common area A, a hospital room a, a common area B, a hospital room b, a common area C, and a treatment room c into segments including one or more. The information processing device 300 stores segments including one or more areas. In FIG. 11, for example, an adjacency information table 1010 shows a case where adjacent areas are managed as segments. In the adjacency information table 1010, for example, when the common area A and the hospital room a are in an adjacency relationship, they are expressed as {A, a}. The information processing device 300 also stores a sensor location table 1020 that is information on areas where sensors are installed.

In the method of managing sensors for each segment, the relationship information generation unit 320 identifies association rules from a group of sensors belonging to a segment. Specifically, the sensor 1001, the sensor 1002, the sensor 1003, and the sensor 1004 belong to the segment of the adjacency relationship {A, a}. Then, the relationship information generation unit 320 identifies an association rule including the sensor 1001, the sensor 1002, the sensor 1003, and the sensor 1004 for the segment of the adjacency relationship {A, a}. A sensor 1001, a sensor 1002, and a sensor 1005 belong to the segment of the adjacency relationship {A, B}. Then, the relationship information generation unit 320 identifies association rules from the sensors 1001, 1002, and 1005 for the segments of the adjacency relationship {A, B}.

In this way, when managing sensors by dividing them into segments, the relationship information generation unit 320 identifies an association rule from a group of sensors belonging to a segment. Therefore, the amount of data processing is reduced and the processing speed is improved compared to the case where the relationship information generation unit 320 identifies the combination of related sensors from all the association rules.

12A and 12B are flowcharts illustrating an example method for managing sensors in segments. The data acquisition unit 310 periodically acquires sensor data from the sensor to be managed (step S201). The storage unit 360 stores the acquired sensor data. The relationship information generator 320 acquires segment information to which each sensor belongs from the adjacent information table 1010 and the sensor location table 1020 (step S202).

The relationship information generating unit 320 selects one segment _Ak from the adjacent information table 1010 (step S203). k is an index, which is a number assigned to each segment. The relationship information generator 320 identifies association rules for the sensors belonging to the segment _Ak (step S204).

The relationship information generation unit 320 selects one sensor B _j from among the sensors belonging to the segment A _k (step S205). j is an index, which is a number assigned to the sensor to be managed. The relationship information generator 320 calculates the reliability for each association rule including the sensor _Bj (step S206). The relationship monitoring unit 330 compares the amount of change in reliability per hour with a predetermined threshold (step S207). Based on the comparison in step S207, the abnormal sensor detection unit 340 determines whether the relationship is broken (step S208). If the relationship is broken (YES in step S208), the abnormal sensor detection unit 340 detects sensor _Bj as an abnormal sensor (step S209). The notification unit 350 notifies the registrant pre-registered in the information processing apparatus 300 of the information related to the abnormal sensor (step S214). After the process of step S214 or when there is no relationship collapse (NO in step S208), the relationship information generation unit 320 determines whether all sensors included in the segment to be processed have been selected in the process of step S205. is determined (step S210).

If all the sensors have not been selected (NO in step S210), the relationship information generator 320 adds 1 to the index j and returns the process to step S205. In the process of step S205, the relationship information generation unit 320 selects the sensor B _j+1 that has not yet been selected.

If all sensors have been selected (YES in step S210), the relationship information generator 320 determines whether or not all segments have been selected in the process of step S203 (step S211). If all segments have not been selected (NO in step S211), the relationship information generator 320 adds 1 to the index k and returns the process to step S203. In the process of step S203, the relationship information generation unit 320 selects the segment A _k+1 that has not yet been selected. If all segments have been selected (YES in step S211), the information processing apparatus 300 ends the abnormal sensor detection process.

Here, the information related to the abnormal sensor notified in step S214 includes identification information for identifying the abnormal sensor, location information where the sensor is installed, and the like. The administrator or the like who has received the notification uses the notified terminal device or the like to display a map based on the positional relationship between the segments and the sensors in FIG. 11, and displays the position of the abnormal sensor on the map. .

In this way, when managing sensors by dividing them into segments, the relationship information generation unit 320 identifies an association rule from a group of sensors belonging to a segment. Therefore, the amount of data processing is reduced and the processing speed is improved compared to the case where the relationship information generation unit 320 identifies the combination of related sensors from all the association rules.

FIG. 13 is a diagram illustrating an example of processing for determining an abnormal chain. In _{FIG .} 13, there are sensor S _a , sensor S _b , sensor S _c , sensor S _d and sensor S _{( f1 ) .} shall be related.

Here, when sensor S _a , sensor S _b , sensor S _c and sensor S _d fail, the relationship R ₁ between sensor S _(f1) and sensor S _a , sensor S _b , sensor S _c and sensor S _d , R ₂ , R ₃ , and R ₄ are broken. Then, the information processing apparatus 300 detects the sensor S _(f1) as an abnormal sensor even if the sensor S (f1 ₎ is not out of order. This process of determining that a normal sensor is abnormal due to another abnormal sensor is called an abnormal chain.

The information processing apparatus 300 determines that the sensor S _{( f1)} determined to be abnormal has no relationship in all combinations held in the conclusion part of the relationship information table 410 of FIG. is judged not to be abnormal. With this processing, the information processing apparatus 300 can prevent a normal sensor from being determined to be abnormal due to a chain of abnormalities. That is, when all the sensors held in the condition part are abnormal, the sensor S _{( f1)} held in the conclusion part is erroneously detected as abnormal. ₎ is judged not to be abnormal. The information processing apparatus 300 performs processing for all sensors determined to be abnormal to prevent a normal sensor from being determined to be abnormal due to a chain of abnormalities.

14A and 14B are flowcharts illustrating an example of processing for determining an abnormal chain. FIGS. 14A and 14B are flowcharts of processing executed when it is determined in step S110 of FIG. 9 that there is an abnormal sensor.

The abnormal sensor detection unit 340 determines whether or not multiple abnormal sensors are detected (step S301). If a plurality of abnormal sensors are not detected (NO in step S301), the information processing apparatus 300 ends the process of determining a chain of abnormalities.

When a plurality of abnormal sensors are detected (YES in step S301), the relationship information generation unit 320 selects one abnormal sensor S _(fn) from the plurality of detected abnormal sensors (step S302). n is an index, which is a number assigned to each abnormal sensor. The relationship information generation unit 320 has the abnormal sensor S _{( fn)} in the conclusion part of the relationship information table 410 of FIG. S303). The abnormal sensor detection unit 340 confirms whether the sensor stored in the condition part of the identified association rule is an abnormal sensor (step S304). The abnormal sensor detection unit 340 determines whether or not relationship collapse has occurred in all of the identified correlation rules (step S305). If relationship collapse occurs in all of the identified association rules (YES in step S305), the abnormal sensor detection unit 340 determines sensor S _(fn) as an abnormal chain sensor (not abnormal sensor) (step S306). ). When the relationship collapse has not occurred in all the identified correlation rules (NO in step S305), or when the process of step S306 ends, the abnormal sensor detection unit 340 performs the process on all the sensors determined to be abnormal. It is determined whether or not it is completed (step S307). If the process has not been completed for all sensors determined to be abnormal sensors (NO in step S307), the abnormal sensor detection unit 340 adds 1 to the index n and returns the process to step S302. In the process of step S302, the relationship information generation unit 320 selects the abnormal sensor S _(fn+1) that has not yet been selected. If the processing has been completed for all the abnormal sensors (YES in step S307), the information processing apparatus 300 ends the processing for determining a chain of abnormalities.

By the process of determining the chain of anomalies, the information processing apparatus 300 can prevent a normal sensor from being determined to be abnormal due to the chain of anomalies.

<Embodiment corresponding to flow line change>
15A and 15B are diagrams for explaining an example of relationship collapse due to changes in flow lines. A change in flow line refers to a path or trajectory along which a person or an object moves. A relationship diagram 1500a is a diagram showing relationships among a plurality of sensors installed in a common area, a patient room a, and a patient room b in a hospital. In the shared area of the relationship diagram 1500a, the reliability value is greater than that of the hospital room a, the hospital room b, etc., so the edge (line) between the sensors indicating the reliability value is shown thick.

In the relationship diagram 1500a, the hospital room of the patient is assumed to be hospital room a. Then, while the patient is in the hospital room a, the sensor 1501 detects the patient's whereabouts. Then, the information processing apparatus 300 calculates the reliability between the sensors 1501 and the plurality of sensors present in the hospital room a.

Here, it is assumed that the patient's hospital room is moved from hospital room a to hospital room b. A relational diagram 1500b is a diagram showing the relationship between a plurality of sensors installed in each of the common area, sickroom a, and sickroom b in the hospital when the sickroom of the patient becomes sickroom b. Since the patient does not use the patient's room a, the sensor 1501 installed in the patient's room a has no object to be detected and does not output the sensor value. Then, the reliability between the sensors 1501 and the plurality of sensors existing in the hospital room a becomes 0, and the relationship collapses. Accordingly, the abnormal sensor detection unit 340 of the information processing device 300 detects the sensor 1501 as an abnormal sensor.

FIG. 16 is a diagram for explaining an example of the firing frequency distribution in a flow line change. The firing frequency distribution 1600a is an example showing the daily change in the firing rate (sensor output rate) of the sensors installed in the common area before the patient moves from the hospital room a. The vertical axis of the firing frequency distribution 1600a represents the firing rate per hour of the sensors installed in the common area, and the horizontal axis represents time (0:00 to 24:00). In the example of the firing frequency distribution 1600a, the firing rate of the sensors installed in the common area is low during the night, and the firing rate of the sensors installed in the common area increases during the daytime. The firing frequency distribution 1600b is an example that represents the daily change in the firing rate of the sensors installed in the common area after the patient moves from the patient's room a to the patient's room b (after the flow line changes). The common area is unaffected by a patient's room move from room a to room b. Therefore, even after the flow line change, the firing frequency distribution 1600b does not change from the firing frequency distribution 1600a.

Firing frequency distribution 1610a is an example that represents a one-day change in the firing rate of sensors installed in hospital room a before the patient moves from hospital room a. The vertical axis of the firing frequency distribution 1610a represents the firing rate per hour of the sensor installed in the hospital room a, and the horizontal axis represents time (0:00 to 24:00). In the example of the firing frequency distribution 1610a, since the patient is in the hospital room, the firing rate of the sensor installed in the hospital room a is low during the night when the patient is not active, and the sensor installed in the hospital room a is active during the daytime. has an increased rate of fire. The firing frequency distribution 1610b is an example that represents the daily change in the firing rate of the sensors installed in the patient's room a after the patient moves from the patient's room a to the patient's room b (after the flow line changes). The hospital room a is in a state where the patient has disappeared. Therefore, in the example of the firing frequency distribution 1610b, the firing rate during the daytime set in the hospital room a is lower than in the firing frequency distribution 1610a.

Firing frequency distribution 1620a is an example that represents a one-day change in the firing rate of sensors installed in hospital room b before the patient moves from hospital room a. The vertical axis of the firing frequency distribution 1620a represents the firing rate per hour of the sensor installed in the hospital room b, and the horizontal axis represents time (0:00 to 24:00). In the example of the firing frequency distribution 1620a, since the patient is not in the hospital room, the firing rate is low both at night and during the day. The firing frequency distribution 1620b is an example that represents the daily change in the firing rate of the sensor installed in the patient's room b after the patient moves from the patient's room a to the patient's room b (after the flow line changes). A new patient has entered the hospital room b. Therefore, in the example of the firing frequency distribution 1620b, the firing rate during the daytime when the patient is active is higher than in the firing frequency distribution 1610a.

FIG. 17 is a diagram illustrating an installation example of sensors for preventing erroneous detection of an abnormal sensor when a flow line changes. The hospital in FIG. 17 has common areas A to C, hospital rooms a and b, and a treatment room c. A sensor (a) is installed in the common area A, a sensor (b) is installed near the door leading to the hospital room a, and a sensor (c) is installed near the boundary with the common area B. In patient room a, sensors (d) and (e) are installed near the door leading out of patient room a. In the common area B, a sensor (f) is installed near the boundary with the common area A, a sensor (g) is installed near the door leading to the hospital room b, and a sensor (h) is installed near the boundary with the common area C. In sickroom b, sensor (i) and sensor (j) are installed near the door leading out of sickroom b. In the common area C, a sensor (k) is installed near the boundary with the common area B, and a sensor (l) and a sensor (m) are installed near the door leading into the treatment room c. In the treatment room c, a sensor (n) and a sensor (o) are installed near the door leading out of the treatment room.

In the example of FIG. 17, it is assumed that the patient enters the patient's room a from the common area A before the flow line changes. Then, the sensor fires in the order of (a), (b), (d), and (e) in the operation before the flow line change. After that, when the patient moves from the hospital room a to the hospital room b, the sensors (a), (b), (d), and (e) stop firing in that order. Then, the abnormal sensor detection unit 340 of the information processing apparatus 300 determines that the relationship between the sensors in the hospital room a has been broken, and detects, for example, the sensor (d) as an abnormal sensor.

In order to prevent erroneous detection of an abnormal sensor due to a change in flow line, the abnormal sensor detection unit 340 of the information processing device 300 extracts sensors installed across rooms when an abnormal sensor is detected. A sensor installed at a location across rooms may be determined from, for example, an adjacent information table or a sensor location table. When the abnormality sensor detection unit 340 detects an abnormality in the sensor (d) installed in the hospital room a, it extracts the sensors (b) and (d) installed across the hospital room a and the common area A. do.

If the patient has not moved from room a to room b and sensor (d) is indeed abnormal, sensor (b) has a firing frequency distribution in which the firing rate rises during the daytime. The firing frequency distribution is such that the firing rate is low even in the middle band. On the other hand, when the patient moves from hospital room a to hospital room b and sensor (d) is not abnormal, both sensors (b) and (d) have a firing frequency distribution in which the firing rate is low even during the daytime. Therefore, even if the abnormal sensor detection unit 340 according to the present embodiment detects an abnormal sensor once, the timing of the time distribution of the firing frequency of the abnormal sensor installed in the straddling part of the room and the other sensor are the same. If they match, the sensor is determined not to be an abnormal sensor. As a result, it is possible to prevent erroneous detection of an abnormal sensor due to a change in flow line.

The information processing apparatus 300 for detecting an abnormal sensor according to the present embodiment obtains the relationship with other sensors for all sensors to be communicated, and determines the relationship from the relationship in the normal state 210 to the relationship in the fault state 220. An abnormal sensor can be efficiently detected by detecting a sensor that has moved to a collapsed state. In addition, the information processing apparatus 300 can prevent erroneous detection of an abnormal sensor by re-determining whether a sensor that has been detected as an abnormal sensor is really abnormal based on a chain of abnormalities or a change in flow line.

The following supplementary remarks are further disclosed regarding the embodiments including the respective examples described above.
(Appendix 1)
the computer
Acquire sensor data obtained by multiple sensors via the network,
storing the acquired sensor data;
Based on the order in which each of the plurality of sensors detects the detection target, relationship information indicating the relationship between the sensors is generated for each of a plurality of time ranges,
monitoring changes in the relationship information generated for each of the plurality of time ranges for each of the plurality of sensors;
An abnormal sensor detection method, comprising detecting a sensor in which an abnormality has occurred among the plurality of sensors based on a change in the relationship information.
(Appendix 2)
The abnormal sensor detection method according to appendix 1, wherein the relationship information is generated for a sensor within a predetermined distance among the plurality of sensors.
(Appendix 3)
A plurality of abnormal sensors are detected based on the change in the relationship information, and when the detection of the first abnormal sensor is caused by the second abnormal sensor, the first abnormal sensor is not an abnormal sensor. The abnormal sensor detection method according to appendix 1, characterized by:
(Appendix 4)
The abnormal sensor detection method according to appendix 1, wherein a threshold used for detecting a change in the relationship information is changed using part of the amount of change in the relationship information.
(Appendix 5)
a plurality of sensors for obtaining sensor data;
The sensor data is acquired from the plurality of sensors via a network, the acquired sensor data is stored, and the relationship between the sensors is determined based on the order in which each sensor among the plurality of sensors detects a detection target. generating relationship information indicating each of a plurality of time ranges; monitoring changes in the relationship information generated for each of the plurality of time ranges for each of the plurality of sensors; and an information processing device that detects a sensor in which an abnormality has occurred among the plurality of sensors based on a change in the abnormal sensor detection system.
(Appendix 6)
a data acquisition unit that acquires sensor data obtained by a plurality of sensors via a network;
a storage unit that stores the acquired sensor data;
a relationship information generating unit that generates relationship information indicating relationships between sensors for each of a plurality of time ranges based on the order in which each sensor among the plurality of sensors detects a detection target;
a relationship monitoring unit that monitors, for each of the plurality of sensors, changes in the relationship information generated for each of the plurality of time ranges;
An information processing apparatus, comprising: an abnormal sensor detection unit that detects, among the plurality of sensors, a sensor in which an abnormality has occurred based on a change in the relationship information.
(Appendix 7)
to the computer,
Acquire sensor data detected by multiple sensors,
First information regarding the order in which at least two sensors among the plurality of sensors each detected a detection target in a first time range, which is calculated based on the sensor data, and the first time range Based on a comparison with second information about the order in which the at least two sensors each detect the detection target in a different second time range, an abnormality occurs in one of the at least two sensors. An abnormal sensor detection program characterized by executing a process of detecting.
(Appendix 8)
The first information is information about the frequency of occurrence of combinations of the order in which the at least two sensors each detect the detection target in the first time range,
The abnormality according to Supplementary Note 7, wherein the second information is information about the frequency of occurrence of combinations of orders in which the detection targets are detected by the at least two sensors in the second time range. Sensor detection program.
(Appendix 9)
The first time range is included in the time period before the abnormality occurs in the at least two sensors,
9. The abnormal sensor detection program according to appendix 7 or 8, wherein the second time range is included in a time period after the occurrence of an abnormality in the at least two sensors.

101 Door sensor 102 Human sensor 103 Sound sensor 104 Bed sensor 300 Information processing device 310 Data acquisition unit 320 Relationship information generation unit 330 Relationship monitoring unit 340 Abnormal sensor detection unit 350 Notification unit 360 Storage unit

Claims (5)

the computer
Acquire sensor data obtained by multiple sensors regarding a detection target via a network,
Based on the order in which each sensor among the plurality of sensors detected the detection target, when the first sensor among the two sensors that detected the detection target detects the detection target, the two sensors detect the detection target. The probability that the sensor whose order is later among the two sensors also detects the detection target is associated with the identification information of each of the sensor whose order is earlier and the sensor whose order is later, and between the sensors Generate relationship information shown as the relationship of each of multiple time ranges,
detecting a sensor in which an abnormality has occurred among the plurality of sensors based on changes in the relationship information for each of the plurality of time ranges ;
A case where a plurality of sensors in which an abnormality has occurred in the detection is detected based on a change in the relationship information, and the plurality of sensors in which the occurrence of the abnormality is detected in the relationship information after the change When all of the preceding sensors shown in the relationship between all the sensors in which one of the abnormal sensors is the subsequent sensor are included in the plurality of abnormal sensors, the A method of detecting an abnormal sensor , comprising: determining that the one of a plurality of abnormal sensors does not have an abnormality. The abnormal sensor detection method according to claim 1, wherein the relationship information is generated for a sensor within a predetermined distance among the plurality of sensors. a plurality of sensors for obtaining sensor data about a detected object;
The sensor data is acquired from the plurality of sensors via a network, and the order of two sensors that detect the detection target based on the order in which each sensor among the plurality of sensors detects the detection target When the sensor whose order is earlier detects the detection target, the sensor whose order is later among the two sensors also detects the detection target is associated with the respective identification information of the sensor, relationship information indicating a relationship between sensors is generated for each of a plurality of time ranges, and based on changes in the relationship information for each of the plurality of time ranges a sensor in which an abnormality has occurred is detected from among the plurality of sensors, and a plurality of sensors in which an abnormality has occurred in the detection is detected based on a change in the relationship information, and In the later relationship information, one of a plurality of abnormal sensors in which the occurrence of the abnormality is detected is the later sensor, and the previous is indicated by the relationship between all the sensors. and an information processing device that determines that the abnormality does not occur in the one of the plurality of abnormal sensors when all of the sensors are included in the plurality of abnormal sensors. Abnormal sensor detection system. a data acquisition unit that acquires, via a network, sensor data obtained by a plurality of sensors regarding a detection target;
Based on the order in which each sensor among the plurality of sensors detected the detection target, when the first sensor among the two sensors that detected the detection target detects the detection target, the two sensors detect the detection target. The probability that the sensor whose order is later among the two sensors also detects the detection target is associated with the identification information of each of the sensor whose order is earlier and the sensor whose order is later, and between the sensors a relationship information generating unit that generates relationship information indicated as the relationship of each of a plurality of time ranges;
an abnormal sensor detection unit that detects a sensor in which an abnormality has occurred among the plurality of sensors based on changes in the relationship information for each of the plurality of time ranges, wherein the sensor in which an abnormality has occurred in the detection is the sensor when a plurality of abnormal sensors are detected based on a change in the relationship information, and in the relationship information after the change, one of the plurality of abnormal sensors in which the occurrence of the abnormality is detected If all the preceding sensors indicated by the relationship between all the sensors that are the sensors are included in the plurality of abnormal sensors, the one of the plurality of abnormal sensors and the abnormality sensor detection unit that determines that the abnormality does not occur . to the computer,
Acquiring sensor data obtained by multiple sensors regarding a detection target,
Based on the order in which each sensor among the plurality of sensors detected the detection target, when the first sensor among the two sensors that detected the detection target detects the detection target, the two sensors detect the detection target. The probability that the sensor whose order is later among the two sensors also detects the detection target is associated with the identification information of each of the sensor whose order is earlier and the sensor whose order is later, and between the sensors Generate relationship information shown as the relationship of each of multiple time ranges,
detecting a sensor in which an abnormality has occurred among the plurality of sensors based on changes in the relationship information for each of the plurality of time ranges ;
A case where a plurality of sensors in which an abnormality has occurred in the detection is detected based on a change in the relationship information, and the plurality of sensors in which the occurrence of the abnormality is detected in the relationship information after the change When all of the preceding sensors shown in the relationship between all the sensors in which one of the abnormal sensors is the subsequent sensor are included in the plurality of abnormal sensors, the An abnormal sensor detection program characterized by executing a process of determining that the one of a plurality of abnormal sensors does not have an abnormality.

Images