JPH1055344A - Diagnostic system for distributed system and storage medium storing diagnostic program of distributed system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the traffic load to be applied on a center from nodes by diagnosing the state of every object based on the data received from a sensor that monitors and senses the distributed objects to be diagnosed and diagnosing every object singly or in cooperation with other objects. SOLUTION: The nodes 100A to 100C are connected to a monitor (center) 200 via a LAN/WAN 500. Then the nodes 100A to 100C monitor the objects 400A to 400C and transfer the diagnostic results among the nodes 100A to 100C. The sensor input holding parts 110 of the nodes 100A to 100C monitor the objects 400A to 400C and acquire the numerical data on the objects 400A to 400C from the sensors 300A to 300C to hold these numerical data. Then every data analysis part 120 analyzes the numerical data acquired from every part 110 by making reference to every knowledge base 140, and every diagnostic agent 130 diagnoses the object abnormality, etc., based on the analysis result and by referring to every diagnostic model 150.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a diagnostic system for a distributed system and a storage medium storing a diagnostic program for the distributed system, and more particularly, to a plant or the like having a target diagnostic model and being a complex of various jobs. The present invention relates to a diagnostic system for a distributed system for diagnosing abnormalities occurring at each node in a distributed system such as operations at a plurality of nodes of a network, and a storage medium storing a diagnostic program for the distributed system.

The diagnosis target of a distributed system is one in which nodes (components) such as a plant, a distributed operating system, a communication system, and a transportation system are dispersed, and as a whole, constitute one system.

[0003]

2. Description of the Related Art FIG. 9 shows a configuration of a conventional diagnostic system. As shown in the drawing, conventionally, a plurality of nodes 10 ₁ -
10 _n and the center 20 are connected via a line. For example, when an error due to operation is detected at the node 10 ₁ , an alarm signal is collected at the center 20 via the line. Thus, when the center 20 receives the alarm signal, the center 20 recognizes the abnormal node and instructs a recovery process or the like for the node. Further, the center 20 monitors the operation states of the plurality of nodes, samples the alarm signals transmitted from the plurality of nodes, and extracts the node causing the abnormality.

[0004]

However, the conventional diagnostic system described above processes alarm signals collected from a plurality of nodes in one center, so that it is a small-scale system composed of a small number of nodes. Although there is no problem if an alarm signal is received from many nodes in a large-scale system, the traffic volume increases, the line becomes congested, and the processing load on the center increases. Further, there is a problem that the sampling interval for the diagnostic processing on the center side becomes longer in proportion to the traffic amount, and delays in dealing with nodes that need to take prompt recovery processing and the like.

Further, when the sensor itself is stopped due to some abnormality, there is a problem that an abnormality of the object cannot be detected. Further, when the specification or configuration of the diagnosis target (object) is changed, not only the specification of the node is changed, but also the center needs to process the change of the program or the like in accordance with the change of the object. Become. Similarly, when a new node and a new object are installed, a new system generation process corresponding to the node and the object is required, and there is a problem in flexibility and expandability.

The present invention has been made in view of the above points, and provides a diagnostic system in a distributed system capable of reducing a traffic load from a node to a center, and a storage medium storing a diagnostic program in the distributed system. The purpose is to: Further, an object of the present invention is to provide a diagnostic system in a distributed system capable of diagnosing a node even when a sensor is suspended due to a failure or maintenance, and a storage medium storing a diagnostic program in the distributed system. That is.

It is a further object of the present invention to provide a diagnostic system in a distributed system and a storage medium storing a diagnostic program in the distributed system that can flexibly deal with a change in the specification or configuration of an object.

[0008]

According to a first aspect of the present invention, an object to be diagnosed which is distributed and arranged, a sensor which monitors the object and senses the object at an arbitrary or predetermined cycle, and a state of the object based on data from the sensor are provided. And a plurality of nodes that perform diagnosis independently or in cooperation with others.

In a second aspect of the present invention, in each of the above nodes,
A sensor input information holding means for holding sensing information passed from the sensor, a knowledge base for storing knowledge for analyzing the sensing information, and an analysis for analyzing the sensing information by referring to the knowledge base based on the sensing information Means, an abnormal cause and at least one diagnostic model having an element of a phenomenon and a symptom of the abnormal cause, and collating with the diagnostic model based on a result analyzed by the analyzing means,
And inference means for generating a hypothesis of the cause of the abnormality, making an inquiry about the hypothesis to another node, and inferring the cause of the abnormality based on a notification from the other node.

In a third aspect, in the above inference means,
Including a notifying means for notifying the inference result to the center. According to a fourth aspect of the present invention, in the above-mentioned notifying means, the center is notified as the certainty of how many of the causes of the abnormality and the phenomena and symptoms of the diagnostic model were confirmed. According to a fifth aspect of the present invention, the diagnostic model has a tree structure in which a root of an abnormal cause and a phenomenon and a symptom of the abnormal cause are leaves.

In a sixth aspect, in the above inference means,
According to the inquiry passed from the inference means of another node, the sensing information input to the own node is analyzed,
Confirmation means to confirm the symptoms or symptoms of the inquiry,
When the phenomenon or symptom can be confirmed by the confirmation means,
Inquiring other nodes for input of sensing information that matches the inquiry, means for inferring,
Includes inquiry response means for returning an inference result to the inquiry source node.

According to a seventh aspect, the diagnostic model is updated as needed by changing or adding an object. The eighth invention is
In the knowledge base, it is updated as needed by changing or adding objects. According to a ninth aspect, based on the sensing information passed from the sensor, an analysis process of analyzing the sensing information with reference to a knowledge base storing knowledge for analyzing the sensing information, and an analysis process performed by the analysis process. Based on the result, the abnormal cause is compared with at least one diagnostic model having elements of the phenomena and symptoms of the abnormal cause, a hypothesis of the abnormal cause is generated, and an inquiry about the hypothesis is sent to another node. Done against
An inference process of inferring the cause of the abnormality based on the notification from the other node.

According to the first aspect of the present invention, at each node corresponding to each object, based on sensing data of the object sensed by the sensor,
By diagnosing the state of the object and cooperating with other nodes, it is possible to diagnose the state of the target object of the own node. Etc. can be grasped. That is, in each node, the state of the object corresponding to the node can be managed in the local node without managing the center, and the traffic load on the center is reduced.

According to the second and ninth inventions, at each node, sensing information is analyzed based on the knowledge base, and a diagnosis model (cause of abnormality, phenomenon / symptom) based on the analysis result (phenomenon / symptom). By generating a hypothesis of the cause, making an inquiry about the hypothesis to another node, and inferring the cause of the abnormality based on a notification from the other node, thereby depending on the center. Since the state of the object corresponding to the node can be grasped without doing so, it becomes possible for the administrator of the node to perform a treatment or the like on the object.

According to the third and fourth aspects of the present invention, at a certain node, by transmitting the inferred final conclusion to the center, the data acquired from the sensor of each node is not directly collected at the center, At the center, it is possible to collect only information on the certainty factor of the cause analyzed for each node. According to the fifth aspect, by adopting a tree structure with the root of the cause of the abnormality and the phenomena and symptoms of the cause of the abnormality as leaves, it becomes possible to perform event-driven backward inference of the cause. By tracing to the root, the cause of the abnormality can be ascertained.

According to the sixth aspect, in response to the inquiry passed from the inference means of another node, the sensing information input to the own node is analyzed to confirm the phenomenon or symptom of the inquiry. By inquiring of other nodes whether there is any input of sensing information that matches the inquiry, it becomes possible to search for a state relationship with the inquiring node and respond to the inquiring node. As a result, the inquiring node can grasp the relevance of the cause of abnormality with another node based on the phenomenon or symptom of the inquiring node.

According to the seventh and eighth aspects, even when an object is changed or added, the diagnostic model and the knowledge base can be updated on a node basis.

[0018]

FIG. 1 shows the configuration of a diagnostic system according to the present invention. In the diagnostic system shown in FIG. 1, a plurality of nodes 100 (hereinafter simply referred to as agents) are connected to a LAN / W.
Between the nodes 100 and the supervisor (center) by the AN500
200. Each node 100 monitors the object 400 via the sensor 300.

Each node 100 monitors an agent-specific object 400. In addition, communication between the agents is possible via the LAN / WAN 500, and the result of diagnosis (abnormal symptom information) at each node 100 can be exchanged between the nodes 100. Each node 100 monitors the object 400,
A sensor holding unit 110 that acquires numerical data of the object 400 from the sensor 300 and holds the numerical data;
A data analysis unit 120 for analyzing the numerical data obtained from the sensor input holding unit 110 with reference to the knowledge base 140;
A diagnostic agent 120 that refers to a diagnostic model 150 based on an analysis result obtained from the data analysis unit 120 and diagnoses an abnormality of an object, a knowledge base 140 that stores inference rules for analyzing numerical data, and has a tree structure. It comprises a diagnostic model 150.

The sensor 300 monitors the object 400, acquires partial information of the object 400 as numerical data at a predetermined time interval, or according to an instruction from the sensor input holding unit 110, and obtains the sensor information holding unit of the node 100. Transfer to 110. The data analysis unit 120 may use any data other than the numerical data obtained from the sensor input holding unit 110 as long as it is data for diagnosis, and uses data obtained from another agent. For example, a memory dump or the like passed from each agent at the time of abnormality may be used. Such numerical data is acquired, and converted into phenomena and symptoms with reference to the knowledge base 140.

The diagnostic agent 130 refers to a diagnostic model 150 as shown in FIG. The diagnostic model 150 shown in the figure has a tree structure composed of three layers,
The root is the cause of the abnormality, and the leaves show phenomena and symptoms. Diagnostic model 15 referenced by this diagnostic agent 130
0 may be the same or different for each agent, and n types of models capable of analyzing the object 400 are held as necessary. When a change or the like occurs in the object 400 to be analyzed, the diagnostic model 150 corresponding to the object 400 is additionally updated.

Here, the diagnostic model 150 will be described in detail. In the example of the diagnostic model 150 in FIG. 2, the root causes the failure cause (A), the intermediate phenomena (B, C) caused by the cause, and the phenomena (D, E, F, G, H). The relationship between the tree branches is only “AND”.
Although a relationship of “OR” or “EXOR” is also conceivable, these can be generally expressed using only “AND”.
For "OR", the tree may be divided, and "EXO"
"R" can be expressed by using "AND" and negation, or "EXOR" can be expressed even if a constraint condition is added to the model.

Instead of placing the numerical data transferred from the sensor input holding unit 110 on the leaves of the diagnostic model 150, the data base 120
A converted phenomenon or symptom is set by referring to 0. For example, when the sensor 110 is a temperature sensor,
The phenomenon that the input data from the temperature sensor rises with the passage of time is a phenomenon called “temperature rise” using the knowledge base 140.

The knowledge base 140, as shown in FIG.
The rule referred when analyzing the symptom information (numerical data) passed from the data analysis unit 120 in the diagnosis unit 120 is stored. The knowledge base 140 is created based on the design knowledge of the diagnosis target, and derives the relationship between the cause of the event occurrence and the result. For example, if the cause is A, the cause and the analysis result are such that the conclusion is B. Generate knowledge indicating the causal relationship of. Therefore, the knowledge base 140 is
As with the diagnostic model 150, the object 4 to be analyzed
When 00 is changed, it is possible to appropriately update the numerical data, which is the monitoring result of the object, so that it can be analyzed.

With such a configuration, when the diagnostic agent 130 acquires the phenomena / symptoms converted by the data analysis unit 120, the diagnostic agent 130 collates the phenomena / symptoms with a leaf (phenomenon / symptom) of one diagnostic model 150, and the diagnostic model 150 The cause of the root abnormality is assumed. A virtual child agent is created for each of the assumed collation results, messages are exchanged with other agents at the level of the child agent, and leaves to be collated by the other agent are examined. And
The likelihood of the hypothesis is notified to the monitor 200 based on the number of collated leaves of the diagnostic model 150. The data analysis unit 120
, The converted input from the sensor 300 (phenomenon
If the symptom) matches a plurality of diagnostic models 150, a plurality of child agents will be generated.

Next, a diagnosis method in the above configuration will be described. FIG. 4 is a sequence chart showing a linkage between two diagnostic agents according to the present invention. The example shown in the figure shows the linkage between the nodes 100A and 100B. Step 101) The sensor 300A is the object 40
The state of 0A is detected, and the state is transferred as numerical data to the sensor input holding unit 110 of the node 100A.

Step 102) Data analysis unit 120
Obtains numerical data from the sensor input holding unit 110 and refers to the knowledge base 140 to convert the numerical data into a phenomenon
Get symptoms. Step 103) The diagnostic agent 130 checks whether a phenomenon or symptom that matches the diagnostic model 150 has occurred based on abnormal symptom information such as a phenomenon or symptom that is an analysis result of the data analysis unit 120. The checking method is performed by the data analysis unit 12 in the diagnostic agent 130.
The abnormal symptom information acquired from 0 is analyzed and collated with the leaves of the diagnostic model, and the cause of the abnormality of the diagnostic model is assumed as a hypothesis.

Step 104) As a result of the collation in step 103, it is assumed that an abnormal cause of the root of the diagnostic model having the matched leaf has occurred, and whether a leaf other than the leaf confirmed by the own agent cannot be confirmed by another agent. To the agent 130B of another node 100B. Step 105) The diagnostic agent 130B of the node 100B generates a child agent every time an inquiry is received, and refers to the diagnostic model 150 based on the numerical information acquired from the sensor input holding unit 110 in the own agent 130B, and from the diagnostic agent 130A. Check whether there is any inquiry or data.

More specifically, the sensor input holding unit 110 of the own agent determines whether the phenomenon or symptom of the inquiry has occurred.
Investigate by acquiring and analyzing sensor information from. Alternatively, it checks whether there is any input from the sensor 300 that matches the inquiry. If there is, the message is returned to the diagnostic agent 130A. At this time,
Time consistency of information input from the sensor 300 is required.

The diagnosis agent 130B of the node 100A can operate even if a phenomenon contrary to the inquiry from the diagnosis agent 130A has occurred.
A shall be notified. Step 106) The diagnostic agent 130B further receives a sensor 300C from another diagnostic agent 130C.
If the diagnostic agent 130
If you can confirm the phenomenon / symptom of the inquiry from A,
It inquires whether the input from the sensor 300C is to another agent, and waits for a response.

Further, an element satisfying the inquiry phenomenon from the diagnostic agent 130A is acquired from the knowledge base 140B, or it is checked whether an element satisfying the inquiry phenomenon has been input from the own sensor input holding unit 110B. Step 107) The diagnostic agent 130B confirms the result checked with reference to the self-diagnosis model 150B and the result obtained from the other diagnostic agent 130C,
Notify the diagnostic agent 130A.

Step 108) Upon receiving the reply to the inquiry, the diagnostic agent 130A confirms the cause of the abnormality and the percentage of leaves established in the diagnostic model 150A (how many of the abnormal causes and the leaves of the diagnostic model 150A (phenomenon / symptom) could be confirmed). (Certainty)) to the monitoring person (center) 200 to be diagnosed. At this time, if node 100
When a phenomenon contrary to the inquiry phenomenon is observed in the result acquired from the diagnostic agent 130B of B, the observation result is acquired. Diagnostic model 1
Since 50 is an "AND tree", the assumed cause is wrong, and in that case, the process is terminated.

Here, the diagnostic agent 130A
The phenomenon (counterexample) that is contrary to the phenomenon of the inquiry to the diagnostic agent 130 </ b> B indicates that a phenomenon that denies the phenomenon explaining the assumed cause is confirmed. Further, the phenomenon that could not be confirmed when notifying the monitor 200 is a phenomenon whose establishment is unknown, and is distinguished from the counterexample.

[0034]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 5 shows a configuration of a diagnostic system according to one embodiment of the present invention. The configuration shown in the figure will be described using three agents for simplification of the description. The configuration of each agent has the configuration shown in FIG.
Will be described with reference numerals of A, B, and C, respectively.

[0035] Between the nodes 100 shown in the figure, the diagnostic agent 130A makes an inquiry to the diagnostic agent 130B to the diagnostic agent 130B about the phenomenon occurring in the own node 100A. It makes an inquiry to the diagnostic agent 130C of the other node 100C, compares the result obtained from the diagnostic agent 130C with the collation result of the self-diagnostic agent 130B, and compares the result with the diagnostic agent 13 of the node 100A.
Notify 0A. Thus, an example will be described in which the agent 130A notifies the monitor 200 of the final collation result.

In FIG. 5, "q" indicates an inquiry to the diagnostic agent of another node.
“A” indicates a response from another diagnostic agent to the inquiry source. (1) The data analysis unit 120A of the node 100A refers to the knowledge base 140A and checks whether the phenomenon of collating with the leaf illustrated in FIG. 2 has occurred. In the example shown in FIG. 5, two leaves of the diagnostic model 150A are collated from x → H, y → G. If matched, create a child agent.

(2) Assuming that an abnormal cause of a model having a leaf to be collated, that is, root A, has occurred, and whether another leaf (phenomenon / symptom) has occurred at another node is determined. Inquires the diagnostic agent 130B (q1: D,
E, F). In the example of FIG. 2, they are D, E, and F. (3) Diagnostic agent 130 of another node 100B
In B, the inquiry is received from the diagnostic agent 130A of the node 100A, and it is checked whether or not the inquiry is established.
In the example of FIG. 5, in the diagnostic node 100C, the sensor 3
Since the input of 00C is “z”, the knowledge base 140C
It can be seen from FIG. 2 that “D” in FIG. 2 holds. Thereby, the diagnostic agent 130C notifies the node 100B of a reply (a1: D) to the inquiry (q1) of the node 100B from the diagnostic node 100B.

(4) The diagnostic agent 130B
Then, assuming the input x from the sensor 300B, "E"
Holds, the diagnosis agents 130A and 13A of the other node 100A determine whether there is an input x from the sensor 300B.
Inquire to 0C (q2: x). Here, the input x
Is input from the sensor 300A to the diagnostic agent 130A, the diagnostic agent 130A notifies the diagnostic agent 130B of the node 100B that the input x is input from the sensor 300A of the own node 100A (a2: x).

(5) If the diagnostic node 1 of another node
In the case of having information (counterexample) contrary to the inquiry at 00, by notifying that fact, it becomes possible to cut branches. Branch pruning uses counterexamples to narrow the search range. In the knowledge base 140, for example, u → 〜G
Such an inference rule is set, and if the input of the sensor 300 is u, the subtree having G as a leaf can be excluded from the search range. Thereby, the search time of the structural tree is reduced.

(6) The diagnostic agent 130A of the node 100A notifies the monitor 200 that 4/5 of the leaves of the model of the root A in the abnormal state in FIG. 2 are satisfied. Next, an example in which the present invention is applied to a transportation system will be described. FIG. 6 shows an application example of a traffic system according to an embodiment of the present invention, FIG. 7 shows an example of a knowledge base of the traffic system according to an embodiment of the present invention, and FIG. 5 is an example of an example diagnostic model.

The knowledge base 140 stores each of the nodes 100A,
It is assumed that common knowledge is stored in B and C. Here, the content of the knowledge base 140 indicates that the numerical data acquired from the sensor 300 is 0% to 20% as “empty state (= QUA
T) ”, 21 to 40% are“ satisfactory (= FAV) ”, 41
~ 60% are "Slightly congested (= LJAM)", 61-80
% Is "congestion (= JAM)", and 81 to 100% is "considerable congestion (= MJAM)".

The nodes 100A, 100B, 100
C sets a monitoring target point on a certain road as an object 400. (1) In the node 100A, the traffic sensor 300
From A, the sensor input holding unit 110 stores 75 as numerical data.
% Has been acquired. That is, the traffic sensor 300
In A, as a result of monitoring the traffic volume, the traffic volume at a predetermined time is investigated, the ratio with the maximum traffic volume is obtained, and the node 1
It notifies the sensor input data holding unit 110A of 00A.

(2) Data analyzer 1 of node 100A
20A refers to the knowledge base 140A and checks whether the phenomenon of collating with the leaf shown in FIG. 8 has occurred. In the example of FIG. 7, since 75% corresponds to the thresholds 61 to 80%, the two models of the diagnostic model are collated from 75% → JAM. When matching,
Create a child agent.

(3) Assuming that an abnormal cause of the model having a leaf to be collated, that is, “MJAM” of the root has occurred, it is determined whether another leaf (phenomenon / symptom) has occurred at another node. Inquiry to the node's diagnostic agent 130B (q
1: QUT, LJAM, LJAM). (4) Diagnostic agent 130 of another node 100B
In B, the inquiry is received from the diagnostic agent 130A of the node 100A, and it is checked whether or not the inquiry is established.
In the diagnostic node 100C, since the input of the sensor 300C is “59%”, referring to the knowledge base 140C, it can be seen that “LJAM” in FIG. 7 is established. As a result, the diagnostic agent 130C makes an inquiry from the diagnostic node 100B of the node 100B (q
The response (a1: LJAM) to 1) is sent to the node 100B.
Notify.

(5) The diagnostic agent 130B
Then, if the input from the sensor 300B is assumed to be 61 to 80%, “LJAM” is established. Therefore, it is determined whether there is an input value between 61 and 80% from the sensor 300B by another node 100.
An inquiry is made to the diagnostic agents 130A and 130C of A (q2: JAM). Here, since the input 80% is input to the diagnostic agent 130A from the sensor 300A, the diagnostic agent 130A
(A2: JAM) that 80% of the input is input from the sensor 300A of the own node 100A.

(6) Thus, the diagnostic agent 130B of the node 100B notifies the diagnostic agent 130A of the node 100A that LJAM and JAM are established. The diagnostic agent 130A
Based on the data acquired from the diagnostic agent 130B, the cause of the congestion in the own node 100A is not only the point of the own node 100A but also the other nodes 100B and 10B.
It is presumed that the cause is congestion at each object (point) of 0C, and the result is notified to the traffic control center 200.

As described above, in the traffic system, by extracting the traffic congestion information from the traffic volume at a plurality of points, it is also possible to locally control the switching time of the traffic light and the like at each node. Further, in the above example, although there is a slight level difference between the inquired nodes, the traffic congestion state is shown. However, the present invention is not limited to this example, and a contradictory phenomenon (anti-example: smooth, empty) is detected. In this case, without using the state of the node that detected the counterexample without rejecting it, it is used for controlling the signal switching time, so that the node that has acquired the phenomenon of “free” or “good” can be used. At the point, control such as lengthening the signal waiting time may be considered.

In the above-described embodiment, an example corresponding to a traffic system has been described. However, the present invention is not limited to this example, and can be applied to process management in a plant. When applied to a plant, it is possible to investigate the progress of a job in each node and find out where a process is delayed from a predetermined time limit by an inquiry / response between nodes.

In the present invention, the function of each node in the above embodiment is constructed by a program and stored in a disk device, or a floppy disk or CD-R
It is also possible to store in a portable storage medium such as OM. It should be noted that the present invention is not limited to the above-described embodiment, but can be variously modified and applied within the scope of the claims.

[0050]

As described above, according to the first aspect, in the distributed nodes, the object to be diagnosed of each node is monitored, and an adjacent node or another node is accessed. In addition, the information of the object can be grasped by the node alone. Therefore,
It is not necessary to transmit all the alarms from the diagnostic nodes to one center as in the related art, and the communication load between the node centers can be reduced.

According to the second and ninth aspects of the present invention, each node makes an inquiry to an adjacent node or a node whose association is defined in advance, and it is possible to grasp the cause of the abnormality of the own node. Therefore, unlike the related art, it is not necessary to sample information on all nodes or all related nodes in the center, and it is possible to take countermeasures for the cause of an abnormality on a local node basis, and to perform recovery processing for the cause of the abnormality. Etc. can be performed quickly. Further, the node that has received the inquiry can also grasp the state of the own node based on the sensing information coming into the own node and the phenomena / symptoms acquired from the inquiry node. If there is a reflected cause, it is also possible to take countermeasures at the own node.
Therefore, even if the sensor of a certain sensor or the agent of a certain node is out of order or in maintenance, the remaining sensors and the agent can execute the diagnosis with constant accuracy.

As a result, the robustness and survivability of the entire diagnostic system are greatly improved. Similarly, even when the center is suspended for some reason, it is possible to grasp the object state by inquiring the agents of the nodes in a range where inquiry can be made in each locally arranged node.

Since a plurality of different diagnostic models can be provided for each agent, it is possible to create a diagnostic model based on partial design information. Third and fourth
According to the invention, since it is possible to collect only information on the certainty factor of the cause analyzed for each node, it is not necessary to collect and process all the sensing information in the center. Is reduced.

According to the fifth aspect of the present invention, by adopting a tree structure, it becomes possible to make an event-driven backward inference of the cause.
It is possible to easily find the cause of the abnormality. According to the sixth aspect, not only the inquiry node but also the inquiry destination node can grasp the phenomena and symptoms, and also can grasp the relevance of the cause of the abnormality to other nodes. Processing can be performed at a local level without performing.

According to the seventh and eighth aspects, the diagnostic model and the knowledge base can be updated on a node basis even when an object is changed or added, so that the diagnostic system has flexibility and expandability. Is improved. This allows
It is possible to perform only the update work at a local level without performing a large-scale program, diagnostic model or knowledge base update work at the center.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a diagnostic system of the present invention.

FIG. 2 is an example of a diagnostic model of the present invention.

FIG. 3 is an example of a knowledge base of the present invention.

FIG. 4 is a sequence chart showing a linkage between two diagnostic agents of the present invention.

FIG. 5 is a configuration diagram of a diagnostic system according to an embodiment of the present invention.

FIG. 6 is a diagram showing an application example of a traffic system according to an embodiment of the present invention.

FIG. 7 is an example of a knowledge base in a transportation system according to an embodiment of the present invention.

FIG. 8 is an example of a diagnostic model in the traffic system of the present invention.

FIG. 9 is a configuration diagram of a conventional diagnostic system.

[Explanation of symbols]

 Reference Signs List 100 node 110 sensor input holding unit 120 data analysis unit 130 diagnostic agent 140 knowledge base 150 diagnostic model 200 monitor (center) 300 sensor 400 object 500 LAN / WAN

Claims (9)

[Claims] 1. An object to be diagnosed distributed on a network, a sensor for monitoring the object and sensing at an arbitrary or predetermined cycle, and a plurality of objects for diagnosing a state of the object based on data from the sensor. A diagnostic system for a distributed system, comprising: a plurality of nodes that have a diagnosis function of (1) and perform diagnosis independently or in cooperation with others. 2. The apparatus according to claim 2, wherein the node is a sensor input information holding unit that holds sensing information passed from the sensor; a knowledge base that stores knowledge for analyzing the sensing information; Analysis means for analyzing the sensing information with reference to a knowledge base; at least one diagnostic model having an abnormality cause and elements of phenomena and symptoms of the abnormality cause; and the diagnosis based on a result analyzed by the analysis means Inference means for performing a comparison with the model to generate a hypothesis of the cause of the abnormality, making an inquiry about the hypothesis to another node, and inferring the cause of the abnormality based on a notification from the other node. The diagnostic system in the distributed system according to claim 1. 3. The diagnostic system according to claim 2, wherein the inference unit includes a notification unit that notifies a center of an inference result. 4. The diagnostic system according to claim 3, wherein the notifying unit notifies the center of how many of the cause of the abnormality and the phenomena and symptoms of the diagnostic model have been confirmed as a certainty factor. 5. The diagnostic system according to claim 2, wherein the diagnostic model has a tree structure in which the root of the abnormality is a root and the phenomenon and the symptom of the abnormality are leaves. 6. The inference means analyzes sensing information input to its own node in response to an inquiry passed from inference means of another node,
Confirmation means for confirming the phenomenon or symptom of the inquiry, and means for inquiring other nodes for input of sensing information matching the inquiry and inferring when the phenomenon or the symptom can be confirmed by the confirmation means. 3. The diagnostic system in the distributed system according to claim 2, further comprising: a query response unit that returns an inference result to a query source node. 7. The diagnostic system according to claim 2, wherein the diagnostic model is updated as needed by changing or adding the object. 8. The diagnostic system according to claim 2, wherein the knowledge base is updated as needed by changing or adding the object. 9. An analysis process for analyzing the sensing information with reference to a knowledge base for storing knowledge for analyzing the sensing information based on the sensing information passed from the sensor; Based on the result, the abnormal cause is compared with at least one diagnostic model having elements of the phenomena and symptoms of the abnormal cause, a hypothesis of the abnormal cause is generated, and an inquiry about the hypothesis is sent to another node. A diagnostic program in a distributed system, comprising: an inference process for inferring the cause of an abnormality based on a notification from the other node.