JPH02284538A - Generation system for network fault diagnostic data - Google Patents

Abstract

PURPOSE:To attain accurate fault diagnosis when a fault occurs by clarifying the cause of an error and updating data used for fault diagnosis including the cause so that the error of diagnosis is reduced when the error of diagnosis occurs. CONSTITUTION:An inference part FM1 refers to a diagnostic knowledge base DB, indicates the inspection method of a fault state to a state inspection part FM2 and causes it to execute the method. FM2 collects network information in accordance with the indication of FM1, refers collection data to the diagnostic knowledge base DB, recognizes the right or wrong of data and transmits the result to FM1. The cause of the fault can be obtained by the repetition of the transfer of processings in FM1 and FM2. The history of the processings is stored in a storage part 2. A correct answer rate inspection part 3 and an error cause detection part 5 calculate the correct answer rate of the result obtained by fault clarification by the causes of the fault at the rate of right or wrong, and a certainty updating part 6 detects data being the cause for deciding it to be other fault from the history of the processings of FM1 and FM2 although it is the fault concerned when the correct answer rate is less than a previously given value, and changes the calculating method of certainty which the data supplies.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a network fault diagnosis data generation method [Prior Art] Recent trends in the construction of networks for computers and communication equipment include wide-area decentralization and large-scale expansion. I get kicked. Accordingly, when a failure occurs, the loss inevitably tends to increase. At the same time, network operation and maintenance become more complex and costs increase. Therefore, there is a need for a method for effectively diagnosing network failures based on their history.

However, until now, there has been no example of a network failure analysis system that has a function of updating the first and second data used for diagnosing a failure occurring in a network based on the failure history. In the device or system for detecting and diagnosing failures occurring in networks such as communications and computers, which is the object of the present invention, the diagnostic method and the content of lateral distortion for that purpose are simple. For example, if an incorrect diagnosis is made, All that was required was the addition of one new item in the form of deterioration, which was a new data linking the symptoms of a disorder and its cause on a one-to-one basis.

There is no diagnostic mechanism that can analyze even complex failures like the present invention, and it is necessary to calculate the correct answer rate to update the information used in the mid-diagnosis process and check the one difference factor. , there is no example of a mechanism for modifying parameters related to eye difference factors.

[Problem to be solved by the invention]

The conventional network fault diagnosis data generation method described above does not update and generate data used for diagnosing faults that occur in the network based on the fault history, so it is not possible to perform accurate fault diagnosis for complex faults. The problem is that there is no.

[Failure to decide on issues]

The system of the present invention is designed to prevent physical failures that occur in the devices and transmission paths that make up the network, and to We will respond to network failures caused by misoperation by the user of the device, malfunction of network equipment due to a defect in the computer program created to use the network device, or malfunction of equipment due to incorrect setting of parameter values for operating various devices. The name of the fault state that the network is in, the name of the relevant fault state,
Failure information for determining that the relevant failure state has occurred;
Calculation method for calculating the degree of certainty that indicates the possibility of falling into the relevant failure state, name of the next stage failure state that can be searched by detecting other detailed failure information, 1 of the relevant failure state
The name of the fault state before the stage, the method for recovering from the fault in the own fault state, the question for the fault state, and the content of the report.
A network fault diagnostic data generation method that updates fault diagnostic data that describes the overall fault structure of a network fault by a collection of unit data that indicates one fault state and that describes the entire fault structure of a network fault. Use the data to formulate a hypothesis about the cause of the failure, instruct the status inspection unit to inspect network information to verify that hypothesis, and use the inspection results obtained by the status inspection unit to investigate the cause in more detail. The inference section repeats the process of forming a hypothesis and instructing the inspection of information, and as a result, determines the cause and presents an old or new method depending on the cause.The inference section collects network information according to the instructions of the inference section, and the collected information is Processing by the inference unit and the status inspection unit obtained in a fault analysis using the network fault diagnosis method, which consists of a status inspection unit that determines whether the usage conditions of each device match the predetermined usage conditions and sends the determination result to the inference unit. means for storing the history of failure analysis; and means for calculating the correct answer rate of results obtained by fault analysis as a percentage of correct answers for each cause of the fault;
If the correct answer rate for each cause of each failure is less than a predetermined value, the data that causes the failure to be determined as another failure regardless of the failure is detected from the processing history of the inference unit and the status inspection unit; The present invention is characterized by comprising means for changing the calculation method of the certainty factor contributed by the data.

〔Example〕

Next, embodiments of the present invention will be described in detail using the drawings.

FIG. 1 is a block diagram showing one embodiment of the present invention.

In FIG. 1, FM is a diagnostic section, which includes an inference section FMI and a status inspection section FM2, and performs failure diagnosis.

The DB is a diagnostic knowledge base and stores fault diagnostic data that is an update target of the present invention.

The inference unit FMI uses the knowledge used for fault diagnosis in the inference knowledge base DB to formulate a strategy for investigating the cause. The diagnostic knowledge base DB records knowledge about network failures, and the state in which a certain device has fallen, or the test result of network information obtained as a test result in the condition inspection unit FM2, can be used to determine the failure state of each device. are associated.

The failure states of each device are hierarchically related according to the types and details of confirmed data and inspection results, and when a certain failure state is confirmed, more detailed It presents several possible failure conditions.

In order to check the possible fault states presented in the diagnostic knowledge base DB, the inference unit FMI refers to the diagnostic knowledge base DB to find out how to test the fault state.
Instruct M2 to execute.

In other words, the diagnostic knowledge base DB may be considered as a collection of data cards that define failure states according to the content of collected information. These cards are organized in a hierarchical manner, with the collected information going from a rough stage to a detailed stage, and as the data changes, one fault state will move from one fault state to the next. are related. When the cause of a failure is identified, it is linked to the recovery power method.

The state inspection unit FM2 collects network information according to instructions from the inference unit FMI. The collection method is to ask the Ministry of Utilization about the situation, I! 2I? Examples include inspecting fM traces and confirming parameters necessary for using equipment such as system configuration definitions. The usage conditions (parameters,
(line trace logic regulations, operation methods, etc.)
It confirms whether it is correct or not and sends the result to the inference unit FMI.

This diagnostic section F M (1) Inference section FMI and status inspection section)
The cause of the failure is finally determined by repeating the exchange of M2 processing.

In FIG. 1, reference numeral 1 denotes a diagnosis history recording section. When the diagnosis history recording section 1 receives a message indicating data transfer from the diagnosis section FM, it receives diagnostic processing contents from the inference section FMI and the condition inspection section FM2, and stores the contents in the diagnosis history storage section 2.

The method of receiving from the inference part FMI and the state inspection part FM2 is as follows.
Even if the reception is received sequentially each time the processing is performed, or when the diagnosis is completed, the inference section 1 of the diagnosis section FM and the state inspection section FM
Uke may take the contents that 2 has memorized all at once. The contents and format of the data stored by the diagnosis history recording section 1 in the diagnosis history storage section 2K will be explained using FIG.

The second factor is a diagram showing the inscription and format of data that the diagnosis history recording section 1 stores in the diagnosis history storage section 2.

In FIG. 2, the main data recorded by the diagnostic history recording section 1 is the processing section that is performing the processing (either the inference section FMI or the condition inspection section FM2), the name of the fault state that is being processed, or number, if the processing unit is FM2, the contents of the inspection to be processed, the inspection results, the confidence factor used in the inspection, the confidence factor obtained as the inspection result, and the cause of the failure determined by 1.

In FIG. 2, 30 is a processing unit, 31 is a failure state name or a check mark, 32 is a set of processing introduction data when the processing unit is a condition inspection unit FM20, 33 is an inspection item entry field, and 34 is a The test result is U-pi, 35 is the confirmation coefficient entry field, 36 is the confirmed J degree, 37 is the cause of the disorder determined by the diagnosis, and 38 is the true cause of the disorder. Here, cause of failure 3
8 is recorded in the correct answer rate survey g153 in FIG.

In FIG. 1, 3 is a correct answer rate testing section. The correct answer rate testing section 3 starts the control/receiving process when the storage of the diagnostic processing results from the diagnostic history recording section 1 is completed. The correct answer rate inspection unit 3 refers to the contents of the diagnosis history storage unit 2, displays the cause of the failure determined by the diagnosis on the user interface unit 4, and inquires the operator Vζ whether the result is correct or not.

The operator inputs whether it is correct or not, and if it is incorrect, inputs it by selecting the code number corresponding to the true cause of the failure or the nonny displayed by the user interface unit 4.

The user interface section 4 can be a general terminal device equipped with a display and a keyboard.

Correct N rate +・' (spine 3 is user interface gL+
4, the correctness of the diagnosis result and the true cause of the failure are received, and the correct answer rate corresponding to the true cause of the failure is updated.

Further, the true cause of the failure is written at the end of the data already recorded in the diagnosis history storage section 2 by the diagnosis history recording section 1.

The method of updating the correct answer rate is: if the cause of the failure has already occurred KN times and there are correct answers, if the current diagnosis is correct (n + 1 / N + 1)% if it is wrong (n / N
+1). Further, in the case of an error, the item of the cause of the fault determined incorrectly is correlated with the true cause of the fault and counted.

If the correct answer rate for the corresponding failure cause is lower than the pre-specified correct answer rate, control is passed to the error factor detection unit 5,
Reports the causes of failures for which the accuracy rate is less than the specified value, and the incidence of causes of failures that are incorrectly determined.

The incidence rate of the incorrectly determined cause of failure is counted up in association with the true cause of failure in the correct answer rate checking unit 3, and is therefore calculated as a ratio to the number of occurrences of the true cause of failure. Generally, there are a plurality of erroneously determined failure causes for each true failure cause.

In FIG. 1, when the error factor detection unit 5 determines the failure cause with the highest occurrence rate among the incorrectly determined failure causes received from the correct answer rate checking unit 3, the error factor detection unit 5 selects the corresponding true failure cause. Search for a fault condition immediately before branching off from the path leading to the path. In other words, an error occurs when selecting the next fault state from the fault state immediately before the branch.

Therefore, the error factor detection unit 5 performs backward work from the lowest fault state in the hierarchy, that is, the cause of the fault that was incorrectly determined to be the true fault cause, and finds the first common fault state reached. Check the subordinate failure status.

The error factor detection unit 5 detects a common fault state (
(hereinafter referred to as a branch point) and each failure state (hereinafter referred to as a branch failure state) that is one step closer to the fault cause at that branch point (immediately after the occurrence of one fault cause that was incorrectly determined to be the corresponding true fault cause) ) is detected, control is passed to the confidence update unit 6, which incorrectly determines that it is the true cause of the failure and notifies the cause of the failure, the branch point, and the branch failure state.

The certainty update unit 6 in FIG. 1 refers to the diagnosis history storage unit 2, searches for diagnostic results that result in the corresponding fault cause that has been erroneously determined using data that indicates the true fault cause, and makes inferences. Search for a branch point regarding the processing of the partial FMI. The branch fault state immediately after that is detected, the test results are checked for each test content of the branch fault state that branched to the erroneously determined cause of the fault, and the correctness or failure of each test result is counted.

For a diagnosis result that results in a corresponding fault cause that was incorrectly determined based on data that has a corresponding true fault cause recorded in all diagnostic history storage units 2, a corresponding fault cause that was incorrectly determined as a result Counts whether the test result of the branch fault state branched to is correct or incorrect.

As a result, for example, if there are test contents for a branch fault condition that branches to the corresponding fault cause that has been determined from time to time, from C1 to Cn, the number of positive results and the number of negative results for each test content are superior. . At this time, the common property of the valley test contents is that when a positive test result is obtained, the confidence level tends to increase, and when a negative test result is obtained, the confidence level tends to decrease.

Therefore, in order to prevent the error of transitioning to a branch fault state due to the erroneously determined cause of the fault, it is sufficient to perform processing to lower the confidence level. This can be achieved by lowering the confidence coefficient of test contents that have many positive test results, or by increasing the reliability coefficient of test contents that have many negative test results. Here, a method is used in which the reliability coefficient of the test content with the most positive test results is lowered.

The method of giving the reduced confidence coefficient is to use a diagnosis result that results in a corresponding fault cause that has been erroneously determined based on data that has the corresponding true fault cause recorded in all the diagnosis history storage units 2. Then, the target is set so that the certainty of a branch fault state that branches to the incorrectly determined fault cause is slightly lower (5% or 10%) than the confidence of a branch fault state that leads to the true fault cause. Give the confidence coefficient of the test content, and give the confidence coefficient whose confidence coefficient decreases the least (or whose correct answer rate exceeds a pre-given rate) among all new confidence coefficients. This makes it possible to select the fault condition for the correct diagnosis, improving the accuracy rate.

As described above, in the network fault diagnosis data update method of the present invention, when a diagnosis error occurs, it is possible to clarify the cause of the error and update data used for fault diagnosis including the cause so as to reduce the diagnosis error. can.

〔Effect of the invention〕

As explained above, in the network fault diagnosis data update method of the present invention, when a diagnosis error occurs, the cause of the error is clarified, and data used for fault diagnosis including the cause is updated to reduce the diagnosis error. This has the effect of providing data that enables accurate fault diagnosis when fault light is severe.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the network fault diagnosis data updating method of the present invention, and FIG. 2 shows the content and format of data stored in the diagnosis history storage section by the diagnosis history recording section of this embodiment. FIG. FM is a diagnosis section, FMI is the above reasoning section, FM2 is a state inspection section, DB is a diagnostic knowledge base, 1 is a diagnosis history recording section, 2 is a diagnosis history storage section, 3 is an accuracy rate inspection section, 4 is a user interface section, 5 is an error factor detection unit, 6 is a certainty update unit, 30 is a processing unit, 31 is a fault state name or number, 32
is the data, 33 is the test item entry field, 34 is the test result entry field, 35 is the confidence coefficient entry field, 36 is the confidence level, 37゜3
8 is the cause of the failure. Agent: Susumu Uchihara, patent attorney = 171

Claims (1)

[Claims] In a network for the purpose of electrically transmitting information consisting of computers, terminal equipment, exchanges, and communication equipment, physical failures that occur in the equipment or transmission paths that make up the network, and We will respond to network failures caused by misoperation by the user of the device, malfunction of network equipment due to a defect in the computer program created to use the network device, or malfunction of equipment due to incorrect setting of parameter values for operating various devices. the name of the fault condition the network is in,
Failure information for determining that the relevant failure state has occurred;
Calculation method for calculating the degree of certainty that indicates the possibility of falling into the relevant failure state, name of the next stage failure state that can be searched by detecting other detailed failure information, 1 of the relevant failure state
The name of the fault state before the stage, the method for recovering from the fault in the relevant fault state, the question for the relevant fault state, and the content of the report.
A network fault diagnostic data generation method that updates fault diagnostic data that describes the entire fault structure of a network fault by a set of unit data that indicates one fault state and that describes the entire fault structure of a network fault, the method comprising: Use the data to formulate a hypothesis about the cause of the failure, instruct the status inspection unit to inspect network information in order to verify that hypothesis, and based on the inspection results obtained by the status inspection unit,
The inference section repeats the process of forming a hypothesis to investigate the cause in more detail and instructing the inspection of information, and as a result determines the cause and presents a recovery method according to the cause.The inference section collects network information according to the instructions of the inference section. , an inference unit and status obtained through failure analysis using a network fault diagnosis method, which includes a status inspection unit that determines whether the collected information matches predetermined usage conditions for each device and sends the determination result to the inference unit. A means for storing the processing history of the inspection department, a means for calculating the correct answer rate of the results obtained by failure analysis as a percentage of correct answers for each cause of the failure, and a means for calculating the correct answer rate for each cause of failure, and a means for calculating the correct answer rate for each cause of each failure given in advance. If it is less than the value, the data that causes the failure to be determined as another failure is detected from the processing history of the inference unit and the status inspection unit, and the method of calculating the confidence that the data contributes is changed. 1. A network failure diagnosis data generation method, comprising: means for generating network fault diagnostic data.