Machine fault diagnosis device

Abstract

To infer necessary handling for a fault of a device with high accuracy immediately after shipping the device and in a service period, etc.SOLUTION: A device fault diagnosis apparatus D includes: a data generation unit 3 which converts flow chart data 1 for diagnosing each handling for a device fault to basic case example data indicating results of virtually diagnosing specific handling for the device fault in accordance with a specific symptom of the device fault, according to the presence or absence of each of symptoms of the device fault; a model construction unit 5 which constructs a fault inference model 6 for inferring necessary handling for the device fault, in accordance with a new symptom of the device fault, on the basis of the basic case example data; and a handling inference unit 8 which infers necessary handling for the device fault, in accordance with the new symptom of the device fault, on the basis of the fault inference model 6.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a technique for inferring necessary measures for equipment failures.

Background Art

[0002] Techniques for inferring necessary measures for equipment failures are disclosed in Patent Documents 1 and 2. In Patent Documents 1 and 2, in accordance with the actual symptoms of equipment failures, accumulated case data indicating that the actual measures for equipment failures have been diagnosed in reality is created. Then, based on the accumulated case data, necessary measures for equipment failures are inferred according to new symptoms of equipment failures.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] In Patent Documents 1 and 2, it is impossible to infer necessary measures for equipment failures immediately after the shipment of equipment for which there is no accumulated case data. And during the service period of equipment for which there is accumulated case data, when new symptoms of equipment failures that are not included in the accumulated case data appear, it is impossible to infer necessary measures for equipment failures, and when the actual symptoms of equipment failures included in the accumulated case data are biased, the influence on the inference of necessary measures for equipment failures is great.

[0005] Therefore, in order to solve the above problems, an object of the present disclosure is to accurately infer necessary measures for equipment failures immediately after the shipment of equipment and during the service period and the like.

Means for Solving the Problems

[0006] To solve the above problems, based on the knowledge of developers, service staff, etc., flowchart data for diagnosing each treatment for a device failure is created according to the presence or absence of each symptom of the device failure. However, the flowchart data has a different data format compared to the accumulated case data of the prior art. Therefore, the flowchart data is converted into basic case data indicating that a specific treatment for the device failure has been virtually diagnosed according to a specific symptom of the device failure. Then, the basic case data has the same data format as the accumulated case data of the prior art. And based on the basic case data, a failure inference model for inferring the necessary treatment for the device failure according to a new symptom of the device failure is constructed.

[0007] Specifically, the present disclosure includes a data creation unit that converts flowchart data for diagnosing each treatment for a device failure according to the presence or absence of each symptom of the device failure into basic case data indicating that a specific treatment for the device failure has been virtually diagnosed according to a specific symptom of the device failure, a model construction unit that constructs a failure inference model for inferring the necessary treatment for the device failure according to a new symptom of the device failure based on the basic case data, and a treatment inference unit that infers the necessary treatment for the device failure according to a new symptom of the device failure based on the failure inference model. The device failure diagnosis device is characterized by comprising these components.

[0008] According to this configuration, immediately after the device is shipped, basic case data exists, and the basic case data includes symptoms of device failures with little bias of various kinds and treatments for device failures. Therefore, immediately after the device is shipped, the necessary treatment for the device failure can be inferred with high accuracy.

[0009] In addition, in the present disclosure, the data creation unit creates, in duplicate, the basic case data in a number proportional to the occurrence probability of the basic case data based on the branching probability of the presence or absence of each symptom of the failure of the device, and the model construction unit constructs the failure inference model based on the basic case data created in duplicate in a number proportional to the occurrence probability of the basic case data. The device for diagnosing equipment failures is characterized in that.

[0010] According to this configuration, in the basic case data and the failure inference model, the branching probability of the presence or absence of each symptom of the failure of the device in the flowchart data can be accurately reflected.

[0011] In addition, in the present disclosure, the data creation unit adds basic interpolation data for specifying other symptoms of the device failure or other treatments that may occur for the device failure to be considered when the presence or absence of each symptom of the device failure or each treatment for the device failure occurs in the basic case data, and the model construction unit constructs the failure inference model based on the basic case data and the basic interpolation data. The device for diagnosing equipment failures is characterized in that.

[0012] According to this configuration, basic interpolation data can be added to the basic case data to increase the learning data. Then, in the basic interpolation data, other symptoms of the device failure to be considered or other treatments that may occur for the device failure are specified, and in the failure inference model, the causal relationship between the symptoms of the device failure and the treatments for the device failure can be significantly expressed.

[0013] In addition, in the present disclosure, the data creation unit converts the accumulated case data indicating that the actual treatment for the device failure has been realistically diagnosed according to the actual symptoms of the device failure into additional case data having the same data format as the basic case data, and the model construction unit updates the failure inference model based on the basic case data and the additional case data. The device for diagnosing equipment failures is characterized in that.

[0014] According to this configuration, during the service period of the device and the like, additional case data is added, and the additional case data includes various symptoms of device failures and treatments for device failures. Therefore, during the service period of the device and the like, the necessary treatment for the device failure can be inferred with high accuracy.

[0015] In addition, in the present disclosure, the data creation unit adds additional interpolation data for specifying the presence or absence of each symptom of the device failure and the presence or absence of the actual symptom, or other symptoms of the device failure to be considered or other treatments for the device failure that may occur when each treatment for the device failure and the actual treatment occur, in the combined processing data of the basic case data and the additional case data, and the model construction unit updates the failure inference model based on the basic case data, the additional case data, and the additional interpolation data. The device failure diagnosis device is characterized in that.

[0016] According to this configuration, additional interpolation data can be added to the additional case data to increase the learning data. Then, in the additional interpolation data, other symptoms of the device failure to be considered or other treatments for the device failure that may occur are specified, and in the failure inference model, the causal relationship between the symptoms of the device failure and the treatment for the device failure can be significantly expressed.

Effect of the Invention

[0017] In this way, the present disclosure can accurately infer the necessary treatment for the device failure immediately after the device is shipped and during the service period and the like.

Brief Description of the Drawings

[0018]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Mode for Carrying Out the Invention

[0019] Embodiments of the present disclosure will be described with reference to the accompanying drawings. The embodiments described below are examples of the implementation of the present disclosure, and the present disclosure is not limited to the following embodiments.

[0020] (Configuration of the equipment fault diagnosis device of the present disclosure) The configuration of the equipment failure diagnosis device of the present disclosure is shown in FIG. 1. The equipment failure diagnosis device D includes flowchart data 1, accumulated case data 2, data creation unit 3, learning data 4, model construction unit 5, failure inference model 6, new case data 7, and treatment inference unit 8. The flowchart data 1, accumulated case data 2, learning data 4, failure inference model 6, and new case data 7 are databases. The data creation unit 3, model construction unit 5, and treatment inference unit 8 are computers on which the equipment failure diagnosis programs shown in FIGS. 2 to 9 and 15 to 18 are installed.

[0021] In the initial learning process shown in FIGS. 2 to 9, based on the knowledge of developers and service staff, etc., according to the presence or absence of each symptom of the equipment failure, flowchart data 1 for diagnosing each treatment for the equipment failure is created. However, the flowchart data 1 has a different data format compared to the additional case data described later. Therefore, the flowchart data 1 is converted into basic case data indicating that a specific treatment for the equipment failure has been virtually diagnosed according to a specific symptom of the equipment failure. Then, the basic case data has the same data format compared to the additional case data described later. And based on the basic case data, a failure inference model 6 for inferring the necessary treatment for the equipment failure according to a new symptom of the equipment failure is constructed.

[0022] In the additional learning process shown in FIGS. 15 to 18, based on the inquiries of consumers, etc. and the treatments of service staff, etc., according to the actual symptoms of the equipment failure, accumulated case data 2 indicating that the actual treatment for the equipment failure has been actually diagnosed according to the actual symptoms of the equipment failure is created. However, the accumulated case data 2 has a different data format compared to the above-mentioned basic case data. Therefore, the accumulated case data 2 is converted into additional case data indicating that the actual treatment for the equipment failure has been actually diagnosed according to the actual symptoms of the equipment failure. Then, the additional case data has the same data format compared to the above-mentioned basic case data. And based on the additional case data, the failure inference model 6 for inferring the necessary treatment for the equipment failure according to a new symptom of the equipment failure is updated.

[0023] (Initial learning process of the present disclosure) The processing outline of the initial learning of the present disclosure is shown in FIG. 2. The data creation unit 3 creates the basic case data FC, the branch probability data BE, and the basic interpolation data FI as the learning data 4. The model construction unit 5 constructs a failure inference model 6 for inferring the necessary measures for the device failure according to the new symptoms of the device failure based on these learning data 4. The treatment inference unit 8 infers the necessary measures for the device failure according to the new symptoms of the device failure based on the failure inference model 6 and the new case data 7 (which describes the new symptoms of the device failure).

[0024] The creation process of the basic case data of the present disclosure is shown in FIGS. 3 to 5. The data creation unit 3 converts the flowchart data 1 for diagnosing each measure for the device failure according to the presence or absence of each symptom of the device failure into the basic case data FC indicating that a specific measure for the device failure has been virtually diagnosed according to a specific symptom of the device failure. Note that the data creation unit 3 arranges the differences in character types and the fluctuations in expressions (with the same meaning) in the flowchart data 1, and registers each symptom S1 to S5 of the device failure and each measure T1 to T4 for the device failure as the basic case ID.

[0025] In the flowchart data 1, the flow from the following symptoms to the measures is expressed. If symptom S1 does not exist, measure T1 is executed. If symptom S1 exists, the presence or absence of symptom S2 is judged. If symptom S2 does not exist, the presence or absence of symptom S4 is judged. If symptom S2 exists, the presence or absence of symptom S3 is judged. If symptom S3 does not exist, the presence or absence of symptom S5 is judged. If symptom S3 exists, measure T2 is executed. If symptom S4 does not exist, measure T4 is executed. If symptom S4 exists, the presence or absence of symptom S5 is judged. If symptom S5 does not exist, measure T2 is executed. If symptom S5 exists, measure T3 is executed.

[0026] The data creation unit 3 extracts the following scenarios SC1 to SC7 from the flowchart data 1. Then, in scenarios SC1 to SC7, the following basic case data FC1 to FC7 are created. Here, in the basic case data FC1 to FC7, other symptoms of equipment failures that cannot occur or other measures for equipment failures that cannot occur are specified.

[0027] In scenario SC1, symptom S1 exists, symptom S2 exists, symptom S3 exists, and treatment T2 is executed. Therefore, in the basic case data FC1, "YES" is described for symptom S1, "YES" is described for symptom S2, "YES" is described for symptom S3, "1" is described for treatment T2, and "-(hyphen)" is described for other symptoms S4, S5 and other treatments T1, T3, T4.

[0028] In scenario SC2, symptom S1 exists, symptom S2 exists, symptom S3 does not exist, symptom S5 does not exist, and treatment T2 is executed. Therefore, in the basic case data FC2, "YES" is described for symptom S1, "YES" is described for symptom S2, "NO" is described for symptom S3, "NO" is described for symptom S5, "1" is described for treatment T2, and "-(hyphen)" is described for other symptoms S4 and other treatments T1, T3, T4.

[0029] In scenario SC3, symptom S1 exists, symptom S2 exists, symptom S3 does not exist, symptom S5 exists, and treatment T3 is executed. Therefore, in the basic case data FC3, "YES" is described for symptom S1, "YES" is described for symptom S2, "NO" is described for symptom S3, "YES" is described for symptom S5, "1" is described for treatment T3, and "-(hyphen)" is described for other symptoms S4 and other treatments T1, T2, T4.

[0030] In scenario SC4, symptom S1 is present, symptom S2 is not present, symptom S4 is present, symptom S5 is not present, and treatment T2 is executed. Therefore, in the basic case data FC4, "YES" is described for symptom S1, "NO" is described for symptom S2, "YES" is described for symptom S4, "NO" is described for symptom S5, "1" is described for treatment T2, and "- (hyphen)" is described for other symptoms S3 and other treatments T1, T3, T4.

[0031] In scenario SC5, symptom S1 is present, symptom S2 is not present, symptom S4 is present, symptom S5 is present, and treatment T3 is executed. Therefore, in the basic case data FC5, "YES" is described for symptom S1, "NO" is described for symptom S2, "YES" is described for symptom S4, "YES" is described for symptom S5, "1" is described for treatment T3, and "- (hyphen)" is described for other symptoms S3 and other treatments T1, T2, T4.

[0032] In scenario SC6, symptom S1 is present, symptom S2 is not present, symptom S4 is not present, and treatment T4 is executed. Therefore, in the basic case data FC6, "YES" is described for symptom S1, "NO" is described for symptom S2, "NO" is described for symptom S4, "1" is described for treatment T4, and "- (hyphen)" is described for other symptoms S3, S5 and other treatments T1, T2, T3.

[0033] In scenario SC7, symptom S1 is not present, and treatment T1 is executed. Therefore, in the basic case data FC7, "NO" is described for symptom S1, "1" is described for treatment T1, and "- (hyphen)" is described for other symptoms S2 - S5 and other treatments T2, T3, T4.

[0034] Thus, immediately after the shipment of the device, the basic case data FC exists, and the basic case data FC includes the symptoms of the device failure with little bias and the treatment for the device failure. Therefore, immediately after the shipment of the device, the necessary treatment for the device failure can be inferred with high accuracy. Then, in the basic case data FC, data specifying other symptoms of the device failure that cannot occur or other treatments for the device failure that cannot occur is added, and in the failure inference model 6, the causal relationship between the symptoms of the device failure and the treatment for the device failure can be significantly expressed. The causal relationship between the symptoms of the device failure and the treatment for the device failure will be described later with reference to FIGS. 10 to 14.

[0035] The additional processing of the branch probability data of the present disclosure is shown in FIGS. 6 and 7. The data creation unit 3 creates, based on the branch probability of the presence or absence of each symptom of the device failure, a number of basic case data FC proportional to the occurrence probability of the basic case data FC in duplicate, and creates it as the branch equal probability data BE.

[0036] In the flowchart data 1, the branch probabilities of the symptoms S1 to S5 are each set to 1 / 2 in FIGS. 6 and 7, but may be different from 1 / 2 as a modification. In the flowchart data 1, the number of branches of the symptoms S1 to S5 is each set to 2 in FIGS. 6 and 7, but may be determined to be 3 or more as a modification.

[0037] In the scenario SC1, the presence or absence of three types of symptoms S1, S2, and S3 is determined, and the occurrence probability is (1 / 2) 3 = 1 / 8. In the scenario SC2, the presence or absence of four types of symptoms S1, S2, S3, and S5 is determined, and the occurrence probability is (1 / 2) 4 = 1 / 16. In the scenario SC3, the presence or absence of four types of symptoms S1, S2, S3, and S5 is determined, and the occurrence probability is (1 / 2) 4 = 1 / 16. In the scenario SC4, the presence or absence of four types of symptoms S1, S2, S4, and S5 is determined, and the occurrence probability is (1 / 2) 4 = 1 / 16. In the scenario SC5, the presence or absence of four types of symptoms S1, S2, S4, and S5 is determined, and the occurrence probability is (1 / 2) 4= 1 / 16. In scenario SC6, the presence or absence of three types of symptoms S1, S2, and S4 is determined, and the occurrence probability is (1 / 2) 3 = 1 / 8. In scenario SC7, the presence or absence of one type of symptom S1 is determined, and the occurrence probability is (1 / 2) 1 = 1 / 2

[0038] In the basic case data FC1, in proportion to the occurrence probability of scenario SC1, which is 1 / 8, the number of created cases is 16 × 1 / 8 = 2. In the basic case data FC2, in proportion to the occurrence probability of scenario SC2, which is 1 / 16, the number of created cases is 16 × 1 / 16 = 1. In the basic case data FC3, in proportion to the occurrence probability of scenario SC3, which is 1 / 16, the number of created cases is 16 × 1 / 16 = 1. In the basic case data FC4, in proportion to the occurrence probability of scenario SC4, which is 1 / 16, the number of created cases is 16 × 1 / 16 = 1. In the basic case data FC5, in proportion to the occurrence probability of scenario SC5, which is 1 / 16, the number of created cases is 16 × 1 / 16 = 1. In the basic case data FC6, in proportion to the occurrence probability of scenario SC6, which is 1 / 8, the number of created cases is 16 × 1 / 8 = 2. In the basic case data FC7, in proportion to the occurrence probability of scenario SC7, which is 1 / 2, the number of created cases is 16 × 1 / 2 = 8

[0039] Here, when the basic case data FC with the number proportional to the occurrence probability of the basic case data FC is "not created repeatedly", it is considered that the occurrence probabilities of scenarios SC1 to SC7 are equal, and it is considered that the branch probabilities of symptoms S1 to S5 are different from the intended 1 / 2

[0040] On the other hand, when the basic case data FC with the number proportional to the occurrence probability of the basic case data FC is "created repeatedly", it is considered that the branch probabilities of symptoms S1 to S5 are equal to the intended 1 / 2, and it is considered that the occurrence probabilities of scenarios SC1 to SC7 are different

[0041] In this way, in the basic case data FC and the fault inference model 6, the branch probabilities of the presence or absence of each symptom of the device fault in the flowchart data 1 can be accurately reflected

[0042] The additional processing of the basic interpolation data of the present disclosure is shown in FIGS. 8 and 9. The data creation unit 3 adds basic interpolation data FI1 to FI14 that identify the following items when there is or is not each symptom of the device failure or when each treatment for the device failure occurs in the basic case data FC1 to FC7. That is, in the basic interpolation data FI1 to FI14, other symptoms of the device failure to be considered or other treatments for the possible device failure are identified, and other symptoms of the device failure that do not need to be considered or other treatments for the impossible device failure are identified. Here, the basic interpolation data FI1 to FI14 is added based on the basic case data FC1 to FC7 in which there is or is not each symptom of the device failure or each treatment for the device failure has occurred. Below, a part of the basic interpolation data FI will be described in detail.

[0043] In the basic interpolation data FI3, assuming that "YES" of symptom S2 has occurred, the flowchart data 1 flow passing through "YES" of symptom S2, or the basic case data FC1, FC2, and FC3 in which "YES" of symptom S2 has occurred are extracted. Then, symptoms S1, S3, and S5 are identified as symptoms to be considered, treatments T2 and T3 are identified as possible treatments, symptom S4 is identified as a symptom that does not need to be considered, and treatments T1 and T4 are identified as impossible treatments. Further, "YES" is described for symptom S2, "(blank) (YES, NO, 1 is also possible)" is described for symptoms S1, S3, S5 and treatments T2, T3, and "- (hyphen)" is described for symptom S4 and treatments T1, T4.

[0044] In the basic interpolation data FI4, assuming that "NO" of symptom S2 occurs, the flowchart data 1 flow passing through "NO" of symptom S2, or the basic case data FC4, FC5, FC6 where "NO" of symptom S2 occurs is extracted. Then, symptoms S1, S4, S5 are identified as symptoms to be considered, treatments T2, T3 are identified as possible treatments, symptom S3 is identified as an unnecessary symptom, and treatment T1 is identified as an impossible treatment. Further, "NO" is described for symptom S2, "(blank) (YES, NO, 1 is also acceptable)" is described for symptoms S1, S4, S5 and treatments T2, T3, and "- (hyphen)" is described for symptom S3 and treatment T1.

[0045] In the basic interpolation data FI7, assuming that "YES" of symptom S4 occurs, the flowchart data 1 flow passing through "YES" of symptom S4, or the basic case data FC4, FC5 where "YES" of symptom S4 occurs is extracted. Then, symptoms S1, S2, S5 are identified as symptoms to be considered, treatments T2, T3 are identified as possible treatments, symptom S3 is identified as an unnecessary symptom, and treatments T1, T4 are identified as impossible treatments. Further, "YES" is described for symptom S4, "(blank) (YES, NO, 1 is also acceptable)" is described for symptoms S1, S2, S5 and treatments T2, T3, and "- (hyphen)" is described for symptom S3 and treatments T1, T4.

[0046] In the basic interpolation data FI8, assuming that "NO" of symptom S4 occurs, the flowchart data 1 flow passing through "NO" of symptom S4, or the basic case data FC6 where "NO" of symptom S4 occurs is extracted. Then, symptoms S1, S2 are identified as symptoms to be considered, treatment T4 is identified as a possible treatment, symptoms S3, S5 are identified as unnecessary symptoms, and treatments T1, T2, T3 are identified as impossible treatments. Further, "NO" is described for symptom S4, "(blank) (YES, NO, 1 is also acceptable)" is described for symptoms S1, S2 and treatment T4, and "- (hyphen)" is described for symptoms S3, S5 and treatments T1, T2, T3.

[0047] In the basic interpolation data FI12, assuming that "1" of treatment T2 has occurred, the flow of flowchart data 1 passing through "1" of treatment T2, or the basic case data FC1, FC2, FC4 in which "1" of treatment T2 has occurred is extracted. Then, symptoms S1, S2, S3, S4, S5 are specified as the symptoms to be considered, and treatments T1, T3, T4 are specified as treatments that cannot occur. Furthermore, "1" is described for treatment T2, "(blank) (YES or NO is also possible)" is described for symptoms S1, S2, S3, S4, S5, and "- (hyphen)" is described for treatments T1, T3, T4.

[0048] In the basic interpolation data FI13, assuming that "1" of treatment T3 has occurred, the flow of flowchart data 1 passing through "1" of treatment T3, or the basic case data FC3, FC5 in which "1" of treatment T3 has occurred is extracted. Then, symptoms S1, S2, S3, S4, S5 are specified as the symptoms to be considered, and treatments T1, T2, T4 are specified as treatments that cannot occur. Furthermore, "1" is described for treatment T3, "(blank) (YES or NO is also possible)" is described for symptoms S1, S2, S3, S4, S5, and "- (hyphen)" is described for treatments T1, T2, T4.

[0049] In the basic interpolation data FI1 to FI14, it is assumed that "YES or NO" of one type of symptom or "1" of one type of treatment has occurred. As a variant, it may be assumed that "YES or NO" of two or more types of symptoms and / or "1" of two or more types of treatments have occurred.

[0050] In this way, by adding the basic interpolation data FI to the basic case data FC, the learning data 4 can be increased. In the basic interpolation data FI, other symptoms of equipment failures to be considered or other measures for possible equipment failures are specified, and in the failure inference model 6, the causal relationship between the symptoms of equipment failures and the measures for equipment failures can be significantly expressed. Furthermore, in the basic interpolation data FI, other symptoms of equipment failures that do not need to be considered or other measures for equipment failures that cannot occur are specified, and in the failure inference model 6, the causal relationship between the symptoms of equipment failures and the measures for equipment failures can be expressed even more significantly. The causal relationship between the symptoms of equipment failures and the measures for equipment failures will be described later with reference to FIGS. 10 to 14.

[0051] A specific example of the flowchart data of the present disclosure is shown in FIG. 10. Specific examples of the failure inference model of the present disclosure are shown in FIGS. 11 to 14. The failure inference model 6 shown in FIGS. 11 to 14 is a Bayesian network or the like that probabilistically expresses the causal relationship between the symptoms of equipment failures and the measures for equipment failures for the flowchart data 1 shown in FIG. 10.

[0052] In the failure inference model 6 shown in FIG. 11, the basic case data FC and the branch probability data BE without a "- (hyphen)" are created, but the basic interpolation data FI is not added. Since there is no "- (hyphen)" in the basic case data FC and the branch probability data BE and the basic interpolation data FI is not added, there are many shortage link locations. Since there is no "- (hyphen)" in the basic case data FC and the branch probability data BE, there are also many excess link locations.

[0053] In the failure inference model 6 shown in FIG. 12, the basic case data FC and the branch probability data BE with a "- (hyphen)" are created, but the basic interpolation data FI is not added. Since there is a "- (hyphen)" in the basic case data FC and the branch probability data BE, the number of excess link locations decreases. Since the basic interpolation data FI is not added, shortage link locations remain.

[0054] In the failure inference model 6 shown in FIG. 13, the basic case data FC and the branch probability data BE with "- (hyphen)" described are created, the basic interpolation data FI is added, but the number of data in the learning data 4 is small. Since there is an additional amount of the basic interpolation data FI, the number of missing link locations is reduced. Although the number of data in the learning data 4 is small, it is approaching the flowchart data 1.

[0055] In the failure inference model 6 shown in FIG. 14, the basic case data FC and the branch probability data BE with "- (hyphen)" described are created, the basic interpolation data FI is added, and the number of data in the learning data 4 is large. Since there is an additional amount of the basic interpolation data FI, the number of missing link locations is reduced. Since the number of data in the learning data 4 is large, it is approaching the flowchart data 1 considerably.

[0056] (Additional learning process of the present disclosure) The outline of the additional learning process of the present disclosure is shown in FIG. 15. The data creation unit 3 creates additional case data AC, combination process data CP, and additional interpolation data AI as learning data 4 in addition to the basic case data FC and the branch probability data BE. The model construction unit 5 updates the failure inference model 6 for inferring the necessary measures for the device failure according to the new symptoms of the device failure based on these learning data 4. The treatment inference unit 8 infers the necessary measures for the device failure according to the new symptoms of the device failure based on the failure inference model 6 and the new case data 7 (the new symptoms of the device failure are described).

[0057] The creation process of additional case data of the present disclosure is shown in FIG. 16. The data creation unit 3 converts the accumulated case data 2 indicating that the actual treatment for the device failure has been realistically diagnosed according to the actual symptoms of the device failure into additional case data AC having the same data format as the basic case data FC, and identifies other symptoms of the device failure that cannot occur or other treatments for the device failure. Note that the data creation unit 3 compares the flowchart data 1 and the accumulated case data 2, arranges the differences in character types and the fluctuations in expressions (same meaning), maintains the existing symptoms S1 to S5 of the device failure and the existing treatments T1 to T4 for the device failure as the basic case IDs, and registers the new symptom S6 of the device failure and the new treatment T5 for the device failure as the additional case IDs.

[0058] In the additional case data AC1, based on the accumulated case data 2, "YES" is described for symptom S5, "NO" is described for symptom S6, "1" is described for treatment T3, and "- (hyphen)" is described for the other symptoms S1, S2, S3, S4 and the other treatments T1, T2, T4, T5.

[0059] In the additional case data AC2, based on the accumulated case data 2, "NO" is described for symptom S4, "YES" is described for symptom S6, "1" is described for treatment T4, and "- (hyphen)" is described for the other symptoms S1, S2, S3, S5 and the other treatments T1, T2, T3, T5.

[0060] In the additional case data AC3, based on the accumulated case data 2, "NO" is described for symptom S3, "NO" is described for symptom S5, "1" is described for treatment T5, and "- (hyphen)" is described for the other symptoms S1, S2, S4, S6 and the other treatments T1, T2, T3, T4.

[0061] Here, the accumulated case data 2 may be inconsistent with the flowchart data 1. However, the additional case data AC may be inconsistent with the basic case data FC.

[0062] In this way, during the service period of the device or the like, additional case data AC is added, and the additional case data AC includes symptoms of various device failures and treatments for device failures. Therefore, during the service period of the device or the like, the necessary treatment for the device failure can be inferred with high accuracy. Then, in the additional case data AC, data identifying other symptoms of device failures that cannot occur or other treatments for device failures that cannot occur is added, and in the failure inference model 6, the causal relationship between the symptoms of device failures and the treatments for device failures can be significantly expressed. The causal relationship between the symptoms of device failures and the treatments for device failures will be described later with reference to FIG. 19.

[0063] The creation process of the combined processing data of the present disclosure is shown in FIG. 17. The data creation unit 3 specifies that the actual symptoms S6 of device failures or the actual treatments T5 for device failures in the additional case data AC1 to AC3, which are not included in each symptom S1 to S5 of device failures or each treatment T1 to T4 for device failures in the basic case data FC1 to FC7 or the branch probability data BE, cannot occur in the basic case data FC1 to FC7 or the branch probability data BE with "-(hyphen)", and creates combined processing data CP of the basic case data FC1 to FC7 or the branch probability data BE and the additional case data AC1 to AC3.

[0064] In this way, the basic case data FC and the additional case data AC can be combined as data in the same format. Then, in the basic case data FC, data identifying other symptoms of device failures that cannot occur or other treatments for device failures that cannot occur (in the additional case data AC, they are included as actually occurring symptoms or treatments) is added, and in the failure inference model 6, the causal relationship between the symptoms of device failures and the treatments for device failures can be significantly expressed. The causal relationship between the symptoms of device failures and the treatments for device failures will be described later with reference to FIG. 19.

[0065] The addition process of the additional interpolation data of the present disclosure is shown in FIG. 18. In the combined processing data CP of the basic case data FC1 to FC7 and the additional case data AC1 to AC3, the data creation unit 3 adds additional interpolation data AI1 to AI17 that identify the following items when there is or is not each symptom of the device failure and the actual symptom, or when each treatment for the device failure and the treatment for the actual symptom occur. That is, in the additional interpolation data AI1 to AI17, other symptoms of the device failure to be considered or other treatments for the possible device failure are identified, and other symptoms of the device failure that do not need to be considered or other treatments for the impossible device failure are identified. Here, based on the combined processing data CP of the basic case data FC1 to FC7 and the additional case data AC1 to AC3 where there is or is not each symptom of the device failure and the actual symptom, or when each treatment for the device failure and the treatment for the actual symptom occur, the additional interpolation data AI1 to AI17 are added.

[0066] In FIG. 18, as the present disclosure, the data creation unit 3 adds the additional interpolation data AI1 to AI17 based on the basic case data FC1 to FC7 and the additional case data AC1 to AC3. In addition to FIG. 18, as a modification example, the data creation unit 3 may add the additional interpolation data AI1 to AI17 based on the basic interpolation data FI1 to FI14 and the additional case data AC1 to AC3. The data creation unit 3 can add the same additional interpolation data AI1 to AI17 regardless of whether it is the present disclosure in FIG. 18 or the modification example other than FIG. 18. Hereinafter, a part of the additional interpolation data AI will be described in detail using the present disclosure in FIG. 18.

[0067] In the additional interpolation data AI6, assuming that "NO" of symptom S3 occurs, the basic case data FC2, FC3 and the additional case data AC3 where "NO" of symptom S3 occurs are extracted. Then, symptoms S1, S2, and S5 are identified as symptoms to be considered, treatments T2, T3, and T5 are identified as possible treatments, symptoms S4 and S6 are identified as unnecessary symptoms to be considered, and treatments T1 and T4 are identified as impossible treatments. Furthermore, "NO" is described for symptom S3, " (blank) (YES, NO, or 1 is also acceptable)" is described for symptoms S1, S2, S5 and treatments T2, T3, T5, and "- (hyphen)" is described for symptoms S4, S6 and treatments T1, T4. That is, in the additional interpolation data AI6, in addition to the basic interpolation data FI6, "- (hyphen)" is described for symptom S6, and " (blank) (1 is also acceptable.)" is described for treatment T5.

[0068] In the additional interpolation data AI8, assuming that "NO" of symptom S4 occurs, the basic case data FC6 and the additional case data AC2 where "NO" of symptom S4 occurs are extracted. Then, symptoms S1, S2, and S6 are identified as symptoms to be considered, treatment T4 is identified as a possible treatment, symptoms S3 and S5 are identified as unnecessary symptoms to be considered, and treatments T1, T2, T3, and T5 are identified as impossible treatments. Furthermore, "NO" is described for symptom S4, " (blank) (YES, NO, or 1 is also acceptable)" is described for symptoms S1, S2, S6 and treatment T4, and "- (hyphen)" is described for symptoms S3, S5 and treatments T1, T2, T3, T5. That is, in the additional interpolation data AI8, in addition to the basic interpolation data FI8, " (blank) (YES or NO is also acceptable.)" is described for symptom S6, and "- (hyphen)" is described for treatment T5.

[0069] In the additional interpolation data AI9, assuming that "YES" of symptom S5 occurs, the basic case data FC3, FC5 and the additional case data AC1 where "YES" of symptom S5 occurs are extracted. Then, symptoms S1, S2, S3, S4, S6 are identified as symptoms to be considered, treatment T3 is identified as a possible treatment, and treatments T1, T2, T4, T5 are identified as impossible treatments. Further, "YES" is described for symptom S5, "(blank) (YES, NO, 1 is also acceptable)" is described for symptoms S1, S2, S3, S4, S6 and treatment T3, and "- (hyphen)" is described for treatments T1, T2, T4, T5. That is, in the additional interpolation data AI9, in addition to the basic interpolation data FI9, "(blank) (either YES or NO is acceptable)" is described for symptom S6, and "- (hyphen)" is described for treatment T5.

[0070] In the additional interpolation data AI10, assuming that "NO" of symptom S5 occurs, the basic case data FC2, FC4 and the additional case data AC3 where "NO" of symptom S5 occurs are extracted. Then, symptoms S1, S2, S3, S4 are identified as symptoms to be considered, treatments T2, T5 are identified as possible treatments, symptom S6 is identified as an unnecessary symptom to be considered, and treatments T1, T3, T4 are identified as impossible treatments. Further, "NO" is described for symptom S5, "(blank) (YES, NO, 1 is also acceptable)" is described for symptoms S1, S2, S3, S4 and treatments T2, T5, and "- (hyphen)" is described for symptom S6 and treatments T1, T3, T4. That is, in the additional interpolation data AI10, in addition to the basic interpolation data FI10, "- (hyphen)" is described for symptom S6, and "(blank) (1 is also acceptable)" is described for treatment T5.

[0071] In the additional interpolation data AI13, assuming that "1" of treatment T3 has occurred, the basic case data FC3, FC5 and the additional case data AC1 where "1" of treatment T3 has occurred are extracted. Then, symptoms S1, S2, S3, S4, S5, S6 are identified as the symptoms to be considered, and treatments T1, T2, T4, T5 are identified as treatments that cannot occur. Furthermore, "1" is described for treatment T3, "(blank) (YES, NO, or 1 is also acceptable)" is described for symptoms S1, S2, S3, S4, S5, S6, and "- (hyphen)" is described for treatments T1, T2, T4, T5. That is, in the additional interpolation data AI13, in addition to the basic interpolation data FI13, "(blank) (either YES or NO is also acceptable)" is described for symptom S6, and "- (hyphen)" is described for treatment T5.

[0072] In the additional interpolation data AI14, assuming that "1" of treatment T4 has occurred, the basic case data FC6 and the additional case data AC2 where "1" of treatment T4 has occurred are extracted. Then, symptoms S1, S2, S4, S6 are identified as the symptoms to be considered, symptoms S3, S5 are identified as symptoms that do not need to be considered, and treatments T1, T2, T3, T5 are identified as treatments that cannot occur. Furthermore, "1" is described for treatment T4, "(blank) (YES, NO, or 1 is also acceptable)" is described for symptoms S1, S2, S4, S6, and "- (hyphen)" is described for symptoms S3, S5 and treatments T1, T2, T3, T5. That is, in the additional interpolation data AI14, in addition to the basic interpolation data FI14, "(blank) (either YES or NO is also acceptable)" is described for symptom S6, and "- (hyphen)" is described for treatment T5.

[0073] Only in the additional case data AC1, "NO" of symptom S6 has occurred, and in the additional interpolation data AI16, the content of the additional case data AC1 is reflected. Only in the additional case data AC2, "YES" of symptom S6 has occurred, and in the additional interpolation data AI15, the content of the additional case data AC2 is reflected. Only in the additional case data AC3, "1" of treatment T5 has occurred, and in the additional interpolation data AI17, the content of the additional case data AC3 is reflected. However, "(blank) (either YES, NO, or 1 is also acceptable)" is added.

[0074] In the additional interpolation data AI1 to AI17, it is assumed that "YES or NO" of one type of symptom or "1" of one type of treatment has occurred. As a modification, it may be assumed that "YES or NO" of two or more types of symptoms and / or "1" of two or more types of treatments have occurred.

[0075] In this way, the additional interpolation data AI can be added to the additional case data AC, and the learning data 4 can be increased. And in the additional interpolation data AI, other symptoms of the device failure to be considered or other treatments for the possible device failure are specified, and in the failure inference model 6, the causal relationship between the symptom of the device failure and the treatment for the device failure can be significantly expressed. Furthermore, in the additional interpolation data AI, other symptoms of the device failure that do not need to be considered or other treatments for the device failure that cannot occur are specified, and in the failure inference model 6, the causal relationship between the symptom of the device failure and the treatment for the device failure can be expressed even more significantly. The causal relationship between the symptom of the device failure and the treatment for the device failure will be described later with reference to FIG. 19.

[0076] A specific example of the failure inference model of the present disclosure is shown in FIG. 19. The failure inference model 6 shown in FIG. 19 is a Bayesian network or the like that probabilistically represents the causal relationship between the symptom of the device failure and the treatment for the device failure with respect to the flowchart data 1 shown in FIG. 10 and the accumulated case data 2 shown in FIG.

[0077] In the failure inference model 6 shown in FIG. 19, the additional case data AC and the combined processing data CP in which "- (hyphen)" is described are created, the additional interpolation data AI is added, and the number of data in the learning data 4 is large. The number of symptoms of the device failure and the number of treatments for the device failure increase, and the number of links between the symptom of the device failure and the treatment for the device failure also increases. Since the flowchart data 1 is considered and the accumulated case data 2 is considered, the branch probability of the presence or absence of each symptom of the device failure is weighted from 1 / 2 to other values.

Industrial Applicability

[0078] The equipment failure diagnosis device of the present disclosure can accurately infer the necessary measures for equipment failures not only during the service period but also immediately after the equipment is shipped.

Explanation of symbols

[0079] D: Equipment failure diagnosis device 1: Flowchart data 2: Accumulated case data 3: Data creation unit 4: Learning data 5: Model construction unit 6: Failure inference model 7: New case data 8: Treatment inference unit FC: Basic case data BE: Branch probability data FI: Basic interpolation data AC: Additional case data CP: Combining process data AI: Additional interpolation data

Claims

1. A data creation unit that converts flowchart data for diagnosing each treatment for a failure of the device into basic case data indicating that a specific treatment for the failure of the device has been virtually diagnosed according to specific symptoms of the failure of the device according to the presence or absence of each symptom of the failure of the device; A model construction unit that constructs a failure inference model, which is a causal relationship network that probabilistically represents the causal relationship between the symptoms of the device failure and the treatment for the device failure, for inferring the necessary treatment for the device failure according to new symptoms of the device failure based on the basic case data; A treatment inference unit that infers the necessary treatment for the device failure according to new symptoms of the device failure based on the failure inference model, comprising: The data creation unit duplicates and creates the basic case data in a number proportional to the occurrence probability of the basic case data based on the branch probability of the presence or absence of each symptom of the device failure; The model construction unit constructs the failure inference model based on the basic case data duplicated in a number proportional to the occurrence probability of the basic case data; A device failure diagnosis apparatus characterized by the above.

2. The data creation unit adds basic interpolation data for specifying other symptoms of the device failure to be considered or other treatments for the possible device failure when each symptom of the device failure or each treatment for the device failure occurs in the basic case data; The model construction unit constructs the failure inference model based on the basic case data and the basic interpolation data; The device failure diagnosis apparatus according to claim 1, characterized by the above.

3. A data creation unit that converts flowchart data for diagnosing each treatment for a failure of the device into basic case data indicating that a specific treatment for the failure of the device has been virtually diagnosed according to specific symptoms of the failure of the device according to the presence or absence of each symptom of the failure of the device; A model construction unit that constructs a failure inference model, which is a causal relationship network that probabilistically represents the causal relationship between the symptoms of the device failure and the treatment for the device failure, for inferring the necessary treatment for the device failure according to new symptoms of the device failure based on the basic case data; A treatment inference unit that infers the necessary treatment for the device failure according to new symptoms of the device failure based on the failure inference model, comprising: The data creation unit adds basic interpolation data for specifying other symptoms of the device failure or other treatments that may occur for the device failure to be considered when the presence or absence of each symptom of the device failure or each treatment for the device failure occurs in the basic case data. The model construction unit constructs the failure inference model based on the basic case data and the basic interpolation data. A device failure diagnosis apparatus characterized by the above.

4. The data creation unit converts accumulation case data indicating that the actual treatment for the device failure has been actually diagnosed according to the actual symptoms of the device failure into additional case data having the same data format as the basic case data. The model construction unit updates the failure inference model based on the basic case data and the additional case data. The device failure diagnosis apparatus according to any one of claims 1 to 3, characterized by the above.

5. The data creation unit adds additional interpolation data for specifying other symptoms of the device failure or other treatments that may occur for the device failure to be considered when the presence or absence of each symptom of the device failure and the actual symptoms, or each treatment for the device failure and the actual treatment occur in the combined processing data of the basic case data and the additional case data. The model construction unit updates the failure inference model based on the basic case data, the additional case data, and the additional interpolation data. The device failure diagnosis apparatus according to claim 4, characterized by the above.

Images