JP5074216B2 - Sensor network server and sensor network system - Google Patents

Description

  In order to save power, the present invention converts the structure of the time series data based on the user's instruction, especially for the observation value sent from the sensor node that performs intermittent operation that repeats the start state and the stop state at regular time intervals. , Relating to analysis technology.

  In recent years, development of a sensor network system including a small wireless sensor node (hereinafter referred to as a sensor node), a repeater, a base station, and a sensor network server (hereinafter referred to as a management server) equipped with a sensor function has been promoted. The sensor node observes the state of a person or a place (sensor data), relays the observed sensor data in a multi-hop by a relay, and transmits it to the management server via the base station. The management server executes various processes based on the received sensor data.

  A key device in a sensor network is a sensor node characterized by small size and low power. Because it is small in size, it can be attached to everything including the environment and people, and because of its low power, it can be operated for several years with a battery without external power supply. Furthermore, since it communicates by radio | wireless, it is possible to arrange | position to a wide range via a base station or a relay machine.

  As a characteristic operation in the sensor node, there is an intermittent operation. This drives the necessary hardware only when executing tasks such as sensing and data transmission. When there are no tasks to be executed, peripheral hardware such as sensors and RF are completely stopped, and the microcomputer is also low power. It is an operation to sleep in the mode. By performing the intermittent operation, the sensor node can operate for a long time under a limited battery.

  By using sensor nodes, not only a system that collects a wide range of observations, but also analysis of time-series observations collected over a long period of time makes sense and visualization of changes in the real world more value-added. High-value-added systems that generate high-level information are demanded and are beginning to be applied to various applications such as hygiene management, preventive maintenance, and visualization of organizational activities.

  Patent Document 1 discloses a technique for storing measurement data in a database and storing setting values such as a measurement mode as metadata in order to inspect a medical inspection apparatus with the same settings. Document 2 discloses a technique for collecting, integrating, analyzing, and meaning data related to a current event, and assigning an analysis result as a tag. However, the data targeted in these patent documents is data that does not have a time-series pattern.

JP 2006-162484 A Special Table 2004-533034

  Unlike the above-described prior art, the sensor network analyzes time-series observation values and has a wide variety of applications. Therefore, the type of sensor used, the number of sensor nodes, and the observation interval vary depending on the application. Therefore, there are problems described below.

  As a first problem, since there are various operation patterns for sensor node observation and observation value transmission, there are various time-series observation values obtained from the sensor node. For example, when observing temperature and humidity in hygiene management, it may be observed at a low frequency of once every 10 minutes. At this time, the sensor node starts once every 10 minutes, observes, and repeats the operation of sending the observation value to the server. On the other hand, when observing speech in order to recognize human utterances, high-frequency observation of 8000 Hz is necessary. However, if transmission is performed at each observation, the number of communications increases, and the wireless band is tight. Therefore, it is necessary to send the observation values for the amount of data that can be transmitted in one communication as a single packet.

  As described above, there are various types of time-series observation values obtained from sensor nodes, so it is necessary to create an analysis program for each observation and transmission operation pattern even for analysis common to many applications. is there.

  As a second problem, since the types of observation values in the time series are various, the structures of databases for storing the observation values are also various.

  As a third problem, when storing the observation values in the database, if the observation values are created for each observation time and the observation values are stored, the capacity increases to create the same number of records as the number of observations, There is an issue that takes time to search.

  As a fourth problem, the sensor node transmits the observation value wirelessly to the sensor network server, but there is a problem that the wireless communication is temporarily interrupted and the observation value is lost.

  As a fifth problem, when observation is performed using a plurality of sensor nodes, there is a problem that the observation timing is shifted for each sensor node.

  In the present invention, as means for solving the first problem, a data part for holding a time series observation value obtained from a sensor node, and a value for explaining time series information such as an observation time and an observation interval of the data part are provided. Define a general-purpose data structure consisting of meta parts to be retained. It also provides a means for creating this general data structure from observations and delivering it to the application.

  As a means for solving the second problem, as a means for creating the data structure, a time-series meta part including a sensor node ID, an observation time, an observation time, an observation interval, and a sleep time is stored together with the observation value. The user designates an identifier for identifying a time-series pattern of the data structure, a sensor node ID, an observation start time, and an observation time, and requests the sensor network server, so that the server There is provided a data structure creation means for creating a data structure corresponding to the identifier by referring to a table in which a rule for searching and converting a search result into a data structure identified by the identifier is described.

  Further, as a means for solving the third problem, the data structure holds a plurality of observation values in which the observation times transmitted from the sensor node in the data portion are continuous in a binary array state. Then, the observed values are stored in the database as the binary array. Thus, there is provided a database storage means that saves capacity because the number of records is reduced as compared with a method of storing in a database divided at each observation time, and enables high-speed storage because the number of accesses can be reduced.

  Furthermore, as a means for solving the fourth problem, as a supplement means when there is a missing value in the observation value when the data structure is created, when the user requests the data structure from the database, NULL completion, linear By requesting the sensor network server by specifying the missing / complementing means for complementing and nearest value complementing, the server searches the database, and if there is a defect in the search result, the missing complementing means is used to detect the missing data. Provided is a defect complement method selection means for complementing.

  Furthermore, as a means for solving the fifth problem, when creating the data structure from observation values obtained from a plurality of sensor nodes, as a means for correcting a shift in observation timing, the user can By specifying a plurality of node IDs and identifiers describing observation times, observation start times, and data structure patterns, and correcting means for either the latest value correction or the latest value correction, and requesting the server, the server An observation time correction method selection unit that performs a correction by a designated method when a database is searched and observation times between a plurality of nodes are shifted is provided.

  By describing the selection and execution procedure of the analysis program, missing value supplement program, and observation time correction program created according to the above data structure, the program can be freely reconfigured and used in common between different observation values Provide program execution designation means that can.

  In the present specification, “value describing time-series information” means a value indicating a state at a certain time point and a change in the state over a certain period from that point.

  According to the present invention, the user can create an application and an analysis program regardless of the operation pattern of sensor node observation and observation value transmission. In addition, since it can be used in common even when the observation target is different, such as a program for correcting defects or correcting a deviation in observation timing, it is possible to reduce development man-hours.

  Embodiments of the present invention will be described below with reference to the accompanying drawings.

  FIG. 1 is a block diagram of a configuration of a sensor network system according to a first embodiment of the present invention.

  The sensor network system consists of sensor-net server 101, gateway (GW) 102, router node (Router Node, RT) 103, sensor node (Sensor Node, SS) 104, management computer (Management Computer) 105, a client computer 106, a wired sensor 107, an RFID reader (Radio Frequency Identification Reader) 108, and a LAN (Local Area Network) 109.

  The LAN 109 connects the sensor network server 101, the gateway 102, the management computer 105, the client computer 106, and the wired sensor 107 to each other.

  The sensor nodes 104 are distributed and installed in the sensor network system. In the sensor network system of the present embodiment, nine sensor nodes 104 are installed, but any number may be installed. Details regarding the sensor node will be described with reference to FIGS. 4 and 5. FIG.

  The router node 103 includes a ZigBee (registered trademark) communication unit (ZigBee Communicator) 122 and a routing manager 121. The ZigBee communication unit 122 communicates with the gateway 102, the sensor node 104, or another router node 103 using the ZigBee protocol. The routing manager 121 determines a transfer destination of information received from the outside.

  The router node 103 receives environmental information observed by the sensor node 104 or a request issued from the sensor net server 101 and transfers it to the gateway 102, the sensor node 104, or another router node 103.

  The gateway 102 includes a LAN communication unit (LAN Communicator) 120, a routing manager 121, a ZigBee communication unit 122, and a PAN control unit (PAN Controller) 123.

  The LAN communication unit 120 communicates with the sensor network server 101 via the LAN 109. The routing manager 121 determines a transfer destination of information received from the outside. The ZigBee communication unit 122 communicates with the router node 103 or the sensor node 104 using the ZigBee protocol. The PAN control unit 123 controls a PAN (Personal Area Network) configured from the gateway 102.

  The gateway 102 receives environmental information observed by the sensor node 104 or a request issued from the sensor network server 101, and transfers it to the sensor network server 101, the gateway 102, the router node 103, or the sensor node 104.

  The sensor network server 101 communicates with the base station 102, the management computer 105, the client computer 106, and the wired sensor 107 via the LAN 109. In addition, it receives observation values from the sensor node via the base station and controls the sensor node. Details of the sensor net server will be described with reference to FIGS.

  The management computer 105 is operated by an administrator of the sensor network system. The management computer 105 transmits various requests to the sensor network server 101 in response to an operation by the administrator.

  The client computer 106 is operated by a client of the sensor network system. The client computer 106 executes various applications. In addition, the client computer 106 receives environment information observed by the sensor node 104 from the sensor network server 101. The client computer 106 performs various processes based on the received environment information.

  The wired sensor 107 observes environmental information. Then, the wired sensor 107 transmits the observed environment information to the sensor network server 101 via the LAN 109.

  4 is a block diagram illustrating a hardware configuration of the sensor node 104, and FIG. 5 is a block diagram illustrating a functional configuration of the sensor node 104. As illustrated in FIG.

  The sensor node is an RF transceiver 401 that performs wireless transmission / reception, a display 402, an operation button (Button) 403, a sensor (Sensor (s)) 404, a microcomputer (MicroProcessor) 405, and a real-time clock with absolute time (Real-time Clock) 406, volatile memory (Volatile Memory) 407, non-volatile memory (Nonvolatile Memory) 408 such as flash memory (Flash Memory), read-only memory (Read-Only Memory) 409, power to each part of the node A battery 410 to be supplied.

  The sensor node 104 uses two operations, ie, a sleep state in which the power other than the Real-Time Clock 406 is turned off in hardware and a start-up state in which the power of all the circuits is turned on at regular intervals. The sensor node 104 performs sensing using various sensors 404 in the activated state. The sensed information is stored in a packet by the MicroProcessor 405 together with the time information of the Real-Time Clock 406, transmitted wirelessly from the RF Transceiver 401 to the base station and the relay station, and stored in the Nonvolatile Memory 408.

  The MicroProcessor 405 includes a central processing unit (CPU) 414 and a flash memory 413 that is a RAM, a ROM, and a nonvolatile memory. The CPU 414 performs various processes by storing the program recorded in the nonvolatile memory 408 in the memory 413 and executing it.

  An operation button (Button) 403 is an input device that receives a user operation, and can activate a specific operation of the sensor node 104 or set an operation parameter by a specific button operation.

  A display unit (Display) 402 is an output device for displaying information to the user. For example, when the sensor node is used indoors or outdoors for environmental measurement, the latest measured value measured by the sensor unit 404 can be displayed. At this time, in order to reduce power consumption, it is preferable to display nothing at normal time and display the latest measured value only when a specific button operation is performed. When the sensor node is a portable terminal such as a name tag type or a wristwatch type, the time information is normally displayed, and when a text message is received from the sensor network server 101, the message is displayed or a voice message is displayed. When receiving a message, it is preferable to display the incoming call information. A hierarchical operation menu can be displayed in conjunction with the user's button operation. By performing button operations according to the operation menu, an application user (Application User) or a system administrator (System Manager) may set operation parameters of the sensor node or check error information when communication fails. it can.

  Note that the router node 103 and the gateway 102 have the same hardware configuration as that of the sensor node 104, and are configured by ordinary computers having a communication function and the like. Therefore, the description of the configuration of the router node and the gateway is omitted.

  As shown in FIG. 5, the function of the sensor node 104 is a main unit 503 that manages observation, observation value accumulation, and event issuance, an observation unit 502 that performs observation using sensors, and issues events such as observation values. An event issuing unit 504, a communication unit 505 that transmits an event wirelessly or receives a command wirelessly, an observation value accumulation management unit 507 that manages accumulation of observation values in the external storage unit 408, and a parent relay device There is also a command processing unit 511 that processes commands received from the parent base station. These functions are stored in the nonvolatile memory 408, and the CPU 414 performs various processes by storing the program recorded in the nonvolatile memory 408 in the memory 413 and executing it.

FIG. 3 is a block diagram showing the hardware of the sensor network server 101. As shown in FIG. The sensor network server 101 includes a central processing unit (CPU) 301, an external communication unit 302, a power source 303, a hard disk drive (HDD) 304, a keyboard 305 as an input unit, a display 306 as an output unit, and a memory 307.
The sensor network server 101 receives data from the base station via the external communication unit 302 and transmits commands to the base station. In addition, a request from the client computer 106 is received through the external communication unit 302. The CPU 301 interprets the user command input from the keyboard 305 and distributes it to the base station through the external communication unit 302. In addition, the CPU 301 stores the program recorded in the external storage device such as the HDD 304 in the memory 307 and executes it, thereby processing the data acquired through the external communication unit 302 in accordance with the instructions of the program and transferring the data to the HDD 304. Accumulated or displayed on the display 306.

  FIG. 2 is a block diagram illustrating functions executed by the CPU 301 of the sensor network server 101. The functions of the sensor network server 101 include an external request receiving unit 201, an analysis management unit 204, an observation value processing unit 208, an observation time correction unit 209, an analysis manager 127 including a defect complementing unit 210, an analysis module group 202, and the like. Further, the observation value database 116 and the processing value database 217 of the database 207 stored in the HDD 304 are accessed via the database writing unit 205 and the database reading unit 206 in the database access function 113. In addition, a timer event 213 and an observation event 214 from the sensor node are received via the event reception unit 203. Note that these functions are usually configured by programs executed by the CPU 301.

  The event reception unit 114 receives the observation event 214 from the sensor node 104. The received event can be handled in common among the functions of the sensor network server by using a general-purpose data structure called Information212. Details of Information 212 will be described with reference to FIG. The analysis manager 127 performs analysis at the timing when the request 211 from the user, the event 214 from the sensor node, or the timer event 213 arrives. The user instructs the information required for analysis and the analysis procedure using the analysis request script 211.

  The instruction from the user is received by the external request reception unit 201 and interpreted by the analysis management unit 204. The analysis management unit 204 confirms the information necessary for the analysis described in the script and waits until a necessary observation value is obtained from the sensor node 104 or acquires the necessary data from the database reading unit 206. The database reading unit 206 searches the database 207 on the HDD 304, acquires necessary data, and after the missing value correction and correction of the observation time are performed in the observation value processing unit 208, it is converted into the information requested by the user, It returns to the analysis management unit 204. Upon receiving the return, the analysis management unit 204 selects an analysis module described in the script from the analysis module group 202 and delivers the information to the analysis module. If the script has an instruction to store the data in Information in the machining value database 117, the analysis management unit 204 delivers the Information to the database writing unit 205, and the database writing unit 205 stores the database 207 described in the script. Data is written in the machining value table 1400 in the table.

  FIG. 8 shows a data holding format in the Information memory described in FIG. Information 212 includes a data part 802 that holds an observed value and a meta part 801 that holds a value that describes time-series information of the data part. The data unit 802 holds the observed value from the sensor node obtained by the event receiving unit 203 as binary. The meta part 801 has a data name (name) 809, a time series type (type) 810, an axis (axis) 811, and a unit (uom) 812. The meta part 801 further describes time series information of the data part 802. The observation start time (startTime) 1402, the observation time (totalTime) 1403, the sampling rate (SamplingRate) 1404, the duration (Duration) 1405 indicating how many observations exist in one observation value group, and sleep An interval 1406 indicating time is held. By combining information describing these time series in the meta unit 801, time series data with various observation patterns can be handled in common. The correspondence between this observation pattern and Information will be described in a specific example described later.

  FIG. 14 is a diagram illustrating a table definition of a database that stores Information. In the database, ID1401 for identifying the observed value at which node, Type810 indicating the time series type, StartTime1402 indicating the observation start time, TotalTime1403 for describing the time of the data stored in one record, Sampling Rate 1404 to indicate how many data per second, Duration 1405 to indicate how many observations exist in one observation value group, Interval 1406 to indicate sleep time, these are handled as meta part 801 Is called. For the column indicating the meta part 801, only the column attached to the stored information is substituted. Further, the database has a data column (Data) 1407 for storing the data portion 802, and stores data as binary. When there are a plurality of elements in the data part 802, meta information for each element is calculated, and one element is stored as one record.

  As a specific example of the first embodiment described above, an example is shown in which an observation value obtained by a system using sensor nodes and sensor network servers is processed into Information, and this Information is stored in a database.

  The sensor node 104 takes various observation and observation value transmission operation patterns in consideration of the difference in the observation target, the limitation of the communication band, and the battery life. FIG. 6 shows a sequence diagram representing an operation pattern of observation and observation value transmission in the sensor node 104.

  601 in FIG. 6 is a diagram showing an operation of performing observation at a constant interval but transmitting an observation value only when an event occurs. For example, a human sensor is used to detect whether a person is present at 10-second intervals. This is an operation of performing transmission only when the state changes from a state where there is no person to a state where there is no person or when the state changes from a state where there is no person to a state where there is no person. The sensor node performs observation after activation (605), determines the presence or absence of an event (606), and when an event occurs, transmits the event to the base station (607) and sleeps. It takes about 10 milliseconds to activate the sensor and perform observations, and even about 50 milliseconds for transmission. Accordingly, since the sensor node is almost in a dormant state, it can operate with a long life.

  Reference numeral 602 represents a sequence for transmitting a single observation value for each observation. The sensor node performs observation after activation (609), transmits the observation value to the base station (610), and sleeps. It takes about 10 to 50 milliseconds from when the sensor is activated until it enters sleep again.

  Reference numeral 603 represents a sequence for repeating the operation of combining observation values observed continuously for a certain period and transmitting them as one packet. The sensor node always observes during activation (611), and at the same time accumulates the observation value in the nonvolatile memory. The observed values are combined as one packet (612) and transmitted to the base station (613). As an example, consider the case of sending 10 observations. If transmission is performed for each observation, the number of communication is 10 times, and the total communication time is 500 milliseconds. On the other hand, in the case of transmission by this method, the number of communication is one and the communication time is 50 milliseconds, so the communication band can be saved. However, since the sensor node never sleeps, it has the shortest life.

  Reference numeral 604 represents data that repeats a sleep operation after combining (615) the observation values observed (614) continuously for a certain period and transmitting (616) as one packet. In this method, high-frequency observation can be performed, communication bandwidth can be saved, and battery life can be extended.

  Fig. 7 shows the data format such as the observation values that can be received by the server in the above observation pattern, the storage format of the observation values in the database, and the retention format in the memory when the values are acquired from the database. It is shown in the observation value expression in the database and the database search result. Information of each observation pattern is shown in FIG.

  In FIG. 7, a pattern indicated by 601 in which an observation value is transmitted only when an event occurs is shown as a data structure 701 that lasts a certain observation value 709 for a certain period Δd from the event occurrence to the end. As meta information describing this observation value group, an observation start time T and an observation time Δd are required. Information of this operation is shown at 805 in FIG. Node ID <id> 1401, observation start time <startTime> 1402, and observation time <totalTime> 1403 are required as information describing the data part time series. Each observation value 803 is stored in the data unit 802 and the observation time <time> is stored in the meta unit 801.

  A pattern that repeats the operation of transmitting a single observation value 710 together with the observation and sleeping as shown in 602 is shown in a data structure 702 expressed by the observation value and the observation time. The observation time T is required as meta-information explaining this observation value. The result of obtaining this observation value from the database is a form in which the observation time T is given to each observation value, as indicated by 706. Information of this operation is shown at 806. As information describing the data part time series, there are a node ID <id> 1401 and an observation start time <startTime> 1402, and each observation value 803 is held in the data part 802 and the observation time <time> is held in the meta part 801. .

  A pattern 603 shown in Fig. 603 is a data structure 703 represented by an observation value group, an observation interval, and an observation period. As meta information describing the data portion 703, observation start time T, observation time ΔT, and observation interval Δd are required. The observation value group is stored in the database as it is as a single record of the payload acquired wirelessly. Thus, for example, when storing 1-axis 3-axis acceleration observed at 50 Hz in the database, if the column is divided and stored for each observation time, access to the database is required 20 times, but the payload is 1 When storing as one record, it can be stored in the database with one access. As a result, the observation value group can be stored and acquired at high speed. Information 807 of this operation is shown. The meta unit 801 has a node ID <id> 1401, an observation start time <startTime> 1402, an observation time <totalTime> 1403, and an observation interval <observationInterval> 1404 as values that describe time-series information in the data part. In the data part, a plurality of observed values are collectively stored as an observed value group 804 as a binary array.

  The data structure 704 shows the pattern shown in 604, which is continuously observed, transmitted after a certain time, and then sleeping. The data holding format in the database is the same as the data structure shown in 703, and it is possible to save the capacity by not holding the data of the sleep period. As meta information describing this observation value group, an observation time T, an observation interval Δd, an observation duration Δd2, and a sleep interval Δt are required. Information of this operation is shown at 808. Information that describes the data part time series, node ID <id> 1401, observation start time <startTime> 1402, observation time <totalTime> 1403, observation interval <observationInterval> 1404, observation duration <duration> 1405, sleep interval < The meta section 801 has sleepInterval> 1406. A plurality of observation values are collectively stored in the data portion 802 below <Data> as an observation value group 804.

  In the system of the present embodiment, by providing logic that mutually converts the various data formats described above, the user can acquire any information, and therefore be aware of the difference in information depending on the type of sensor. It is possible to create an analysis module and an application without any problem.

  Next, in this embodiment, a specific example of mutual conversion of the information data format described above will be shown.

  FIG. 9 shows a sequence for creating information from the observation value database 116 of the database 207 in which observation events or observation values transmitted from the sensor node 104 are stored.

  910 is a diagram showing a sequence for creating Information at the timing when an observation event arrives from the sensor. The management server 101 receives, from the user, the script 1802 “convert the observed value for 10 minutes coming from the sensor # 1001 and perform Fourier transform” in the external request reception unit 201 (S901). The analysis management unit 204 enters an observation value waiting state S902 that waits for an observation value from the sensor. When the observation value arrives (S903), the analysis management unit 204 confirms the identifier described in the script (S904), and confirms whether 10-minute data necessary for information creation is available (S905). If the 10-minute data is not available, the observation value standby state is entered again. When the 10-minute data is available, the observation start time, observation time, and observation interval are extracted according to the definition of Information (S906), attached to the data as a meta part, and Information is created (S907). Distribute to the analysis module described (S908).

  920 in FIG. 9 is a diagram illustrating an information creation sequence in a case where analysis is periodically performed by a timer. The management server 101 receives a script 1803 from the user, “acquire data from the database at 2 o'clock every day, convert it to information for analysis, perform Fourier transform, and store the result in the database” at the external request reception unit 201 (S921). The analysis management unit 204 starts a timer (S921). Upon receipt of the observation value conversion command from the timer (S922), the analysis management unit 204 searches the observation value database 116 of the database 207 and acquires observation values for the time specified by the script (S924). It is confirmed whether or not the observation values are aligned (S925). If they are not aligned, the data is complemented by the method designated by the user from the types of missing data interpolation shown in FIG. 16 (S926). When the data is available, meta information such as observation start time, observation time, observation interval, etc. is extracted according to the definition of Information shown in FIG. 7 (S927), attached to the data part as a meta part, and information is created ( S928), and distributes to the analysis module described in the received script (S929).

  As an example, processing for converting from 706 to 707 is performed as follows.

  The external request reception unit 201 receives the script 1802 shown in FIG. 1802 is a script that uses the observation event of node ID # 1001 for 600 seconds, generates it as Information of ID # 1, and performs Fast Fourier Transform (FFT). The observation value processing unit 208 confirms the identifier described in the script and instructs the event reception unit 114 to wait for the observation value. Each time the event reception unit 114 receives an observation event sent from the sensor node 104, the event reception unit 114 checks the identifier described in the script to check whether the requested data is available. When the data is available, the observation start time T and the observation interval ΔT are calculated from the received event, and the observation values are collected into a binary array and distributed to the analysis management unit 204.

  FIG. 10 shows a functional block diagram of the observation value processing unit 208 of FIG. The observation value processing unit 208 includes an observation value processing management unit 1001, a processing request reception unit 1002, and a processing result output unit 1003. Needless to say, these are configured as programs executed by the CPU 301, like the observed value processing unit 208. The processing request receiving unit 1002 processes the observation value at the timing when the conversion instruction from the user or the communication from the sensor node 104 arrives. The observation value processing unit 208 receives the processing request at the processing request receiving unit 1002. The observation value processing unit 1001 confirms the information described in the analysis request script 211, and processes the observation value into information by the method 910 or 920 described in the script 211. The created information is output from the processing result output unit 1003 to the analysis module described in the script 211. Information is output from the database writing unit 205 to the processing value database 117 of the database 207 in accordance with the instructions of the script 211.

  FIG. 11 is a functional block diagram of the analysis management unit 204 in FIG. The analysis management unit 204 includes an analysis execution management unit 1101, an analysis module selection unit 1102, an analysis procedure interpretation unit 1103, an analysis result output unit 1104, and an analysis data input unit 1105. The analysis execution management unit 1101 analyzes the analysis request script 1802 at the external request reception unit 201 or at the timing at which the timer reception or observation event described in the script 1803 is received by the event reception unit 203. Needless to say, these are configured as programs executed by the CPU 301.

  The analysis procedure interpretation unit 1103 interprets the analysis procedure described in the script. The analysis data input unit 1105 communicates with the database reading unit 206 and acquires data necessary for analysis from the database 207. The analysis module selection unit 1102 selects the analysis module described in 1802 or 1803 in FIG. 18, distributes the data acquired from the analysis data input unit 1105 to the analysis module, and acquires the analysis result. The acquired analysis result is output from the analysis result output unit 1104 to the database writing unit 205 and stored in the database 207.

  FIG. 12 is a diagram illustrating a function of the database reading unit 206 that reads data for creating information for analysis from the database 207. The database reading unit 206 includes a database reading management unit 1201, a reading request receiving unit 1202, and a database selecting unit 1203. Needless to say, these are also configured as programs executed by the CPU 301.

  The database reading unit 206 receives an observation value or processing value reading request from the analysis management unit 204 at the reading request receiving unit 1202. The database read management unit 1201 interprets the request 1802 received by the read request reception unit 1202, selects the corresponding database 207 by the database selection unit 1203, and acquires data. The acquired data is sent to the observation value processing unit 208 in order to perform the processing described in the request 1802. The observation value processing unit 208 is supplemented by the defect complementation unit 210 in accordance with the request 1802 and is output from the reading result reception unit 1202 to the observation value processing unit 208.

  FIG. 13 is a diagram showing the function of the database writing unit 205 for outputting the analysis result to the database 207. The database writing unit 205 selects a database writing management unit 1301, an analysis result acquisition unit 1302 that acquires an analysis result from the analysis management unit 204, a data structure determination unit 1303 that determines information of the analysis result, and a data storage destination A database selection unit 1304 is included.

  The analysis result is received by the analysis result acquisition unit 1302 from the analysis management unit 204 and delivered to the database write management unit 1301. The data structure determination unit 1303 compares the received identifier of the analysis result with the information identifier described in the database write request script, and if they match, returns the analysis result to the database write management unit 1301. If the identifiers do not match, the data is delivered to the data processing unit 208, receives the processing result 800, and returns the processing result to the database write management unit 1301. The database write management unit 1301 delivers the processing result 800 to the database selection unit 1304. The database selection unit 1304 that has received the processing result outputs the processing result to the database 207 described in the script.

  FIG. 15 is a sequence diagram showing complementation at the time of data loss in the loss complementation unit 210.

  Reference numeral 1501 shown in the left part of FIG. 15 is a sequence diagram in the case of performing defect complementation every time an observation value is received. When the server receives the observation value at the event reception unit 203 (S1501), the server confirms the data part and confirms whether there is any missing (S1502). Thereafter, a complementing method is selected from FIG. 16 according to a request 1800 from the user (S1503), missing value compensation is performed (S1504), information is created, and this information is transmitted to the observation value input unit (S1505).

  1502 shown in the right part of FIG. 15 is a diagram showing a sequence for complementing missing data when acquiring observation values from the database periodically or at a timing designated by the user. The missing complement unit 210 receives the observation value request (S1521) and searches the database (S1522). The observation value acquired from the database is confirmed to be missing (S1523). If there is a defect, the complementing method described in the script is confirmed (S1524), and the value is complemented (S1525). The complement result is processed into Information, and the Information is transmitted to the observation value input unit.

  FIG. 16 shows a missing value complementing method in the missing complement unit 210. 1600 represents a missing observation value and a missing state, 1604 indicates that the observation value exists, and 1605 indicates that the observation value is missing.

  1601 indicates that a missing portion is complemented with an empty value. Information users need to respond according to missing values when using data. 1602 indicates that the missing part is complemented by the last value acquired before the missing part. Reference numeral 1603 indicates that a missing portion is estimated and interpolated from the preceding and following relationships. The estimation method includes linear interpolation and moving average interpolation.

  FIG. 17 is a diagram illustrating an operation for correcting a shift in observation time when a plurality of nodes are simultaneously observing. 1701 in the upper part of FIG. 17 indicates that correction is performed using an observed value (latest value of the time series value) that arrived before the request time. 1702 in the lower part of the figure indicates that correction is performed using the closest observed value (the latest value of the time series value) before and after the request time.

  Next, a second embodiment of the analysis management unit, which is a part of the functional configuration of the sensor network server, will be described.

  19A, 19B, and 19C are diagrams illustrating a specific example of the analysis management unit in the case where the activity of the organization is visualized.

  In order to visualize the activity of an organization, it is necessary to detect the activity status of members such as position, walking, telephone, and presence, and communication relationships such as conversation, influence, and intimacy in business. The method is to make use of a sensor network. The physical quantity acquired by the sensor to detect this communication includes infrared rays for detecting the face-to-face state, voice for detecting speech and environment, and acceleration for detecting human movement.

  A business microscope system is a system for detecting the movement of a person and communication between people from the physical quantities obtained by these sensors and visualizing the state of the organization to help improve the organization.

  As shown in FIG. 19A, two people wear sensor # 1 and sensor # 2. In order to detect communication, the sensor is equipped with an acceleration sensor 1901, an audio sensor 1902, and a face-to-face sensor 1903 by infrared transmission / reception, each detecting human movement in the acceleration sensor and communication in the voice sensor and face-to-face sensor. To do.

  In the upper part of FIG. 19B, the observation timing of the acceleration sensor 1901, the voice sensor 1902, and the face-to-face sensor 1903 and the transmission timing of the observation value are shown. The acceleration sensor 1901 repeats observation and transmission in the form of 602. The voice sensor 1902 needs to be observed at a high frequency of 8000 Hz in order to recognize a human voice. Therefore, observation and transmission are performed in the form of 604 with high frequency and low consumption capacity. The face-to-face sensor 1903 transmits the ID of the other party that communicated with infrared rays in the format 602. However, if there is no infrared communication, a null value is transmitted.

  The result of recognizing motion, face-to-face, and speech using the above observed values is shown as 1904 in the lower part of FIG. Also, the analysis result of sensor # 1 is shown in 1905, and the analysis result of sensor # 2 is shown in 1906. 1904 indicates that the owner of sensor # 1 walked to the owner of sensor # 2 and had a conversation. The data structure of the analysis result 1904 can be expressed by 601 and 602 in FIG.

  FIG. 19C shows a flowchart for performing motion analysis using the observation values of the acceleration sensor. The observation value analysis of the acceleration sensor acquires the observation value for one hour from the database every other hour and outputs it as the operation analysis result every other minute.

  1905 in the left part of the figure is a flowchart in which the analysis management unit 204 acquires observation values from the database and delivers them to the analysis module 202. The user transmits a script “acquire acceleration observation values for one hour” to the analysis management unit 204 via the external communication unit 302. The analysis management unit 204 receives the script (S1901), checks the data type and analysis time described in the script, and starts a timer that starts every other hour (S1902). When receiving the observation value acquisition command from the timer (S1903), the analysis management unit acquires the acceleration for one hour from the database (S1905). Thereafter, it is confirmed whether or not the acquired observation values are aligned for one hour (S1906). If they are not aligned, the blank is filled with NULL (S1907). When all the data is collected, meta information such as observation intervals is extracted (S1908), information is created (S1909), and delivered to the analysis module 202 (S1910).

1906 in the right part of the figure is a flowchart in which the analysis module 202 receives information and analyzes the operation. When receiving the information (S1911), the analysis module extracts a data part according to the type definition described in the script (S1912), and performs a discriminant analysis using the data every other minute (S1913). Discriminant analysis is performed on all observations for one hour. It is confirmed whether the analysis for one hour has been completed (S1914), and if it has not been completed, the next observation value analyzed last time is analyzed. When the analysis of all data is completed, information with the analysis result as the data part is created (S1915) and distributed to the analysis management part (S1906)
According to the above procedure, the motion can be analyzed from the observed acceleration value. The discriminant analysis used in this procedure is a general technique. Therefore, even when the input data is a voice observation value, it is possible to perform analysis by unifying the acceleration observation value and Information.

  Next, a third embodiment of the analysis management unit 204, which is a function executed by the CPU 301 of the sensor network server 101, will be described with reference to FIGS.

  FIG. 20 is an embodiment for vibration preventive maintenance of a pipe joint and a pipe support part.

  FIG. 20 shows a schematic diagram of a piping system in a nuclear power plant. The piping system includes a pump 2001 and a valve 2002, a piping 2003 connecting them, and a support unit 2007 that supports the piping 2003. The failure of piping is thought to occur for the following reasons. The first cause is that the number of repetitions of stress applied to the pipe joint due to vibration exceeds the fatigue limit of the pipe, causing cracks in the pipe joint 2006 and water leakage. The second cause is that the pipe support part 2007 is destroyed when the number of repeated stresses applied to the pipe support part 2007 exceeds the fatigue limit number of the pipe support part 2007, thereby causing the pipe 2003 itself to be assumed. External stress is applied, pipe breakage occurs, and water leaks. The third cause is that the impact of the fluid is repeatedly applied to the inside of the pipe due to the effects of pump cavitation, etc., thereby reducing the wall thickness inside the pipe. Water leakage occurs when the pipe wall thickness becomes smaller than the threshold value.

  In order to predict the breakage of pipe joints and pipe support parts, vibration sensors 2008 are periodically installed at multiple locations in the piping system to observe piping vibrations, thereby conducting stress analysis of the entire piping system and piping. Calculate the stress applied to seam 2006 or pipe support 2007. The frequency of vibration that causes pipe breakage is about 200 Hz at maximum.

  The total number of pipes in the plant is 50,000. When nodes are installed in all pipes and 200Hz acceleration is sent, there is a problem that the transfer bandwidth becomes tight. In order to solve the above problem, as shown in FIG. 21, the node does not transmit all the observed values, but stores the observed value 2100 in the nonvolatile memory 408 in the node, and then transmits the resized data. As the resizing method, there are a method 2101 for resizing by dividing by a time frame, a method 2102 for resizing by thinning observation values, and a method 2103 for reducing the resolution and resizing.

  The user creates a higher-order program for stress analysis and vibration waveform display executed by the analysis management unit 204 and transmits an observation value acquisition request to the external request reception unit 201. After receiving the request, the sensor network server 101 receives the resized observation value from the node at the event reception unit 114, performs the processing described in the request at the observation value processing unit 208, and creates the Information 800, Delivered to top-level applications 2016 for stress analysis 2014 and vibration waveform display 2015. Thereby, it is possible to handle the resized data without modifying the upper program.

  While various embodiments of the present invention have been described in detail above, it goes without saying that the present invention is not limited to these embodiments.

1 is a diagram illustrating a system configuration of a sensor network in which the present invention is implemented. FIG. 2 is a block diagram illustrating a functional configuration of a sensor network server according to the first embodiment. FIG. 2 is a block diagram illustrating a hardware configuration of a sensor network server according to the first embodiment. FIG. 3 is a block diagram illustrating a hardware configuration of a sensor node according to the first embodiment. FIG. 3 is a block diagram illustrating a functional configuration of a sensor node according to the first embodiment. FIG. 6 is a diagram illustrating the types of operation of the sensor node according to the first embodiment. It is the figure which showed the kind of information of the observed value of Example 1. FIG. It is the figure which expressed Information of Example 1 with XML. 3 is a flowchart for creating information of the first embodiment. FIG. 3 is a block diagram illustrating a functional configuration of an observation value processing unit according to the first embodiment. FIG. 3 is a block diagram illustrating a functional configuration of an analysis management unit according to the first embodiment. FIG. 3 is a block diagram illustrating a functional configuration of a database reading unit according to the first embodiment. FIG. 3 is a block diagram illustrating a functional configuration of a database writing unit according to the first embodiment. It is the figure which showed the data holding | maintenance format in the database of Example 1. FIG. It is a flowchart of the process which complements the missing value of Example 1. It is a figure which shows the kind of complement method of the missing value of Example 1. It is a figure which shows the correction method of the observation time in case the some node of Example 1 exists. FIG. 6 is a diagram illustrating a script example for creating Information of the first embodiment. FIG. 10 is a diagram (part 1) for explaining the visualization of the activity of the organization according to the second embodiment. FIG. 10 is a diagram (part 2) for explaining the visualization of the activity of the organization according to the second embodiment. FIG. 13 is a third diagram illustrating visualization of an activity of an organization according to the second embodiment. It is a figure for demonstrating the preventive maintenance of piping of Example 3. FIG. It is a figure showing the function which converts the observation value resized by the node of Example 3 into Information.

Explanation of symbols

  101 ... Sensor network server, 102 ... Gateway, 103 ... Router node, 104 ... Sensor node, 105 ... Management computer, 106 ... Client computer, 107 ... Wired sensor, 108 ... RFID reader, 109 ... LAN, 114 ... Event reception unit, 116 ... Observation value database, 117 ... Processing value database, 127 ... Analysis manager, 201 ... External request reception unit, 202 ... Analysis module group, 204 ... Analysis management unit, 205 ... Database writing unit, 206 ... Database reading unit, 208 ... Observation value processing unit, 209... Observation time correction unit, 210... Defect missing unit, 211... Analysis request reception unit, 212.

Claims (13)

A sensor network server connected via a network to a plurality of sensor nodes that perform intermittent observation,
A communication unit, a processing unit, and a storage unit;
The processing unit is configured to receive the time series value as time series information based on a plurality of time series values observed at the same sensor node received via the communication unit, and the time series information. Create a data structure that consists of a meta part that holds the values that describe
Based on the value describing the time-series information in the meta part, the time-series information in the data part is interpreted, time correction is performed on the time-series information received in a different expression form from the plurality of sensor nodes, and Perform time interpolation and convert to time-series information in the data format specified by the user .
A sensor net server characterized by that. The sensor network server according to claim 1,
The data portion holds the time series values in a plurality of types of data formats,
The meta unit holds a set of an identifier indicating the type of the data format of the time series value and a value describing the time series information .
A sensor net server characterized by that. The sensor network server according to claim 1,
The processor is
The time correction is performed by correcting the time-series deviation of the time-series values for each of the plurality of sensor nodes on the basis of designation from the user based on the latest value or the latest value of the time-series values .
A sensor net server characterized by that. The sensor network server according to claim 1,
The processor is
When the time series value is missing, the time series value is complemented by a null value, the immediately preceding time series value, or the predicted value from the context based on designation from the user. Performing the time completion ,
A sensor net server characterized by that. The sensor network server according to claim 1,
The storage unit has a table describing a conversion method for converting a plurality of types of the time series values,
The processing unit compares the type of the time series value specified by the user with the type of the time series value stored in the storage unit, and performs the time series by the corresponding conversion method described in the table. Convert value types ,
A sensor net server characterized by that. The sensor network server according to claim 1,
The processor is
Accepting a script that specifies the data format conversion procedure of the data structure or the data structure analysis procedure, and based on the script, the conversion procedure specified by the user can be rearranged.
A sensor net server characterized by that. The sensor net server according to claim 2,
The storage unit
The data part that holds the time series value in a plurality of types of data formats, and the meta part that holds a set of the identifier that indicates the type of the data format of the time series value and a value that describes the time series information Storing the data structure consisting of :
A sensor net server characterized by that. A sensor network system for analyzing observation values from sensors ,
A plurality of sensor nodes that perform intermittent observations;
A sensor network server connected to the plurality of sensor nodes via a network;
The sensor net server is
Based on a plurality of time series values observed at the same sensor node received via the previous period network, a data part for holding the time series values as time series information, and a value for explaining the time series information are held. Manage the data structure consisting of meta parts
Based on the value describing the time-series information in the meta part, the time-series information in the data part is interpreted, time correction is performed on the time-series information received in a different expression form from the plurality of sensor nodes, and Perform time interpolation and convert to time-series information in the data format specified by the user .
A sensor network system characterized by this. The sensor network system according to claim 8 ,
The sensor net server is
In the data portion, the time series values are held in a plurality of types of data formats,
The meta part holds a set of an identifier indicating the type of the data format of the time series value and a value describing the time series information .
A sensor network system characterized by this. The sensor network system according to claim 8 ,
The sensor net server is
It has an input part where a request from the user is input,
Based on the request input via the input unit, the time series value of each of the plurality of sensor nodes is corrected based on either the latest value or the latest value of the time series value. To correct the time ,
A sensor network system characterized by this. The sensor network system according to claim 8 ,
The sensor net server is
It has an input part where a request from the user is input,
When the time series value is missing, based on the request input through the input unit, the null value, the immediately preceding time series value, or the predicted value from the context, the method Complementing the time series value and performing the time completion ,
A sensor network system characterized by this. The sensor network system according to claim 8 ,
The sensor net server is
An input unit for inputting a request from the user ;
A storage unit for storing a table describing a conversion method for converting a plurality of types of time series values;
The type of the time series value according to the request from the input unit is compared with the type of the time series value stored in the storage unit, and the time series value is converted by the corresponding conversion method described in the table. To
A sensor network system characterized by this. The sensor network system according to claim 8 ,
The sensor net server is
It has an external request reception unit that receives request scripts from users ,
According to the request script, the conversion procedure of the data format of the data structure or the analysis procedure of the data structure is rearranged .
A sensor network system characterized by that.

Images