Fault prediction device, fault prediction method, and program

Abstract

To provide a failure prediction device configured to correct prediction results of a physical model and a statistical model to accurately predict a risk on a failure of a prediction target.SOLUTION: A failure predicting device comprises: a first evaluating unit for employing a statistical model generated on the basis of data relating to a failure of a prediction target, to evaluate a failure rate representing the probability of a random failure of the prediction target; a second evaluating unit for employing a physical model generated on the basis of design data of the prediction target, to evaluate a failure probability representing the probability of a failure due to degradation of the prediction target over time; an acquiring unit for acquiring a qualitative evaluation result obtained by qualitatively evaluating the prediction target for each predetermined evaluation item; and a correcting unit for calculating a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by a correction factor based on the qualitative evaluation result.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to a failure prediction device, a failure prediction method, and a program. [Background technology]

[0002] Quantitative failure physics models and failure statistics data are generally used to evaluate the failure probability and failure rate of equipment and structures, which are important in product risk management. The accuracy of quantitative predictions is greatly affected by the accuracy of the equipment and structure physical models, the completeness of measurement data, and the presence or absence of experience in the product or business field. Therefore, for new products that do not have sufficient experience or measurement data, the prediction accuracy of failure probability and failure rate may be low.

[0003] For example, Patent Document 1 describes a technology that uses a statistical deterioration model to predict the useful life of equipment for which sufficient inspection and maintenance information, plant data, monitoring data, etc. have been accumulated, and that performs deterioration diagnosis based on the statutory useful life for equipment for which sufficient data cannot be obtained. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 5802619 Summary of the Invention [Problem to be solved by the invention]

[0005] However, some factors that cause product failures are difficult to quantitatively evaluate and are difficult to predict using mathematical models such as physical models that predict the probability of failure due to aging and statistical models that predict random failures. For this reason, in conventional technologies, there is a possibility that the prediction accuracy of failure probability and failure rate using physical models and statistical models will decrease.

[0006] The present disclosure has been made in consideration of such problems, and provides a failure prediction device, a failure prediction method, and a program that can correct individual prediction results of a physical model and a statistical model to accurately predict the risk of a failure of the predicted target. [Means for solving the problem]

[0007] According to one aspect of the present disclosure, a failure prediction device includes a first evaluation unit that evaluates a failure rate representing the probability of a random failure of the prediction object using a statistical model generated based on data regarding a failure of the prediction object, a second evaluation unit that evaluates a failure probability representing the probability of a failure due to aging of the prediction object using a physical model generated based on design data of the prediction object, an acquisition unit that acquires a qualitative evaluation result obtained by qualitatively evaluating the prediction object for each predetermined evaluation item, and a correction unit that calculates a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability, respectively, by a correction coefficient based on the qualitative evaluation result.

[0008] According to one aspect of the present disclosure, a failure prediction method includes the steps of: evaluating a failure rate representing a probability of random failure of a prediction target using a statistical model generated based on data regarding failures of the prediction target; evaluating a failure probability representing a probability of failure due to aging of the prediction target using a physical model generated based on design, manufacturing data, and operation and maintenance history data of the prediction target; obtaining a qualitative evaluation result by qualitatively evaluating the prediction target for each predetermined evaluation item; and calculating a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability, respectively, by a correction coefficient based on the qualitative evaluation result.

[0009] According to one aspect of the present disclosure, the program causes a computer of a failure prediction device to execute the following steps: evaluating a failure rate representing the probability of random failure of the prediction object using a statistical model generated based on data regarding failures of the prediction object; evaluating a failure probability representing the probability of failure due to aging of the prediction object using a physical model generated based on design data of the prediction object; obtaining a qualitative evaluation result by qualitatively evaluating the prediction object for each predetermined evaluation item; and calculating a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability, respectively, by a correction coefficient based on the qualitative evaluation result. Effect of the Invention

[0010] According to the failure prediction device, failure prediction method, and program disclosed herein, it is possible to accurately predict the risk of a failure of the prediction target. [Brief description of the drawings]

[0011] [Figure 1] 1 is a diagram illustrating an overall configuration of a failure prediction device according to a first embodiment of the present disclosure. [Diagram 2] 1 is a diagram illustrating a functional configuration of a failure prediction device according to a first embodiment of the present disclosure. [Diagram 3] FIG. 2 is a diagram for explaining qualitative evaluation according to the first embodiment of the present disclosure. [Figure 4] 5 is a flowchart showing an example of processing of the failure prediction device according to the first embodiment of the present disclosure. [Diagram 5] FIG. 2 is a first diagram showing an example of a corrected failure rate according to the first embodiment of the present disclosure. [Figure 6] FIG. 13 is a second diagram showing an example of a corrected failure rate according to the first embodiment of the present disclosure. [Figure 7] FIG. 2 is a first diagram showing an example of corrected failure probability according to the first embodiment of the present disclosure. [Figure 8] FIG. 13 is a second diagram showing an example of corrected failure probability according to the first embodiment of the present disclosure. [Figure 9]FIG. 13 is a third diagram showing an example of corrected failure probability according to the first embodiment of the present disclosure. [Figure 10] FIG. 11 is a diagram illustrating a functional configuration of a failure prediction device according to a second embodiment of the present disclosure. [Figure 11] 10 is a flowchart showing an example of processing of a failure prediction device according to a second embodiment of the present disclosure. [Figure 12] FIG. 11 is a first diagram for explaining a function of a first adjustment unit according to the second embodiment of the present disclosure. [Figure 13] FIG. 11 is a second diagram for explaining a function of the first adjustment unit according to the second embodiment of the present disclosure. [Figure 14] FIG. 11 is a first diagram for explaining a function of a second adjustment section according to a second embodiment of the present disclosure. [Figure 15] FIG. 11 is a second diagram for explaining a function of the second adjustment section according to the second embodiment of the present disclosure. [Figure 16] FIG. 11 is a third diagram for explaining a function of the second adjustment section according to the second embodiment of the present disclosure. [Figure 17] FIG. 11 is a diagram illustrating a functional configuration of a failure prediction device according to a third embodiment of the present disclosure. [Figure 18] 13 is a flowchart illustrating an example of a process of the failure prediction device according to the third embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] <First embodiment> Hereinafter, a failure prediction device according to a first embodiment of the present disclosure will be described with reference to FIGS.

[0013] (Overall composition) FIG. 1 is a diagram showing an overall configuration of a failure prediction device according to a first embodiment of the present disclosure. The failure prediction device 1 according to this embodiment predicts the risk of failure of a prediction target 30. The prediction target 30 is, for example, an industrial plant (power plant, chemical plant, etc.), a ship, a refrigerator, or other equipment or device constituting a product 3 (including facilities, equipment, etc.).

[0014] As described above, in the quantitative evaluation of failure rate and failure probability using a failure prediction model, for example, when there is a lack of data for generating the model, or when there are uncertain factors that are difficult to quantitatively evaluate, the prediction accuracy may decrease. In order to suppress such a decrease in prediction accuracy, in this embodiment, as shown in FIG. 1, a user A performs a qualitative evaluation of each prediction target 30 and inputs the evaluation result to the failure prediction device 1. In addition, the failure prediction device 1 corrects the prediction result using the failure prediction model based on the evaluation result by the user.

[0015] (Functional configuration) FIG. 2 is a diagram illustrating a functional configuration of the failure prediction device according to the first embodiment of the present disclosure. As shown in FIG. 2, the failure prediction device 1 includes a CPU 10, a memory 11, a storage 12, a connection interface 13, and a communication interface .

[0016] The memory 11 is a so-called main storage device, in which instructions and data for the CPU 10 to operate based on a program are developed.

[0017] The storage 12 is a so-called auxiliary storage device, and is, for example, a hard disk drive (HDD), a solid state drive (SSD), etc. The storage 12 stores data acquired and generated by the failure prediction device 1, data used for failure prediction, etc.

[0018] The connection interface 13 is an interface for connecting to the display device 2A, the input device 2B, etc. The display device 2A is a display device that displays information in a visibly recognizable manner, such as a liquid crystal display or an organic EL display. The input device 2B is an input device that accepts an operation by a user of the failure prediction device 1, such as a general mouse, keyboard, touch sensor, etc.

[0019] The communication interface 14 is an interface for communication connection between the failure prediction device 1 and an external device. The failure prediction device 1 according to this embodiment is communicatively connected to the prediction target 30 via the communication interface 14, and receives operation data (sensor measurement values, control command values, maintenance history, etc.) from the prediction target 30. In addition, the failure prediction device 1 may be communicatively connected to various databases (failure statistics DB 50, design / manufacturing / operation DB 51, evaluation DB 52, maintenance plan DB 53, cost DB 54, etc.) via the communication interface 14.

[0020] The CPU 10 is a processor that controls the overall operation of the failure prediction device 1. The CPU 10 operates according to a predetermined program to fulfill the functions of a setting unit 100, a first evaluation unit 101, a second evaluation unit 102, an acquisition unit 103, a correction unit 104, a first judgment unit 105, a second judgment unit 106, and a proposal unit 107.

[0021] The setting unit 100 selects a failure prediction model to be used for predicting failures of each of the multiple prediction targets 30. Furthermore, the setting unit 100 refers to the preventive maintenance plan for the prediction target 30 recorded in the maintenance plan DB 53 and the maintenance cost recorded in the cost DB 54, and sets objective functions such as the reliability, availability, and maintenance cost of the prediction target 30, and constraint conditions such as the reliability, availability, and total cost of the prediction target 30.

[0022] The failure prediction model includes a statistical model and a physical model. The statistical model outputs a failure rate indicating how many accidental failures of the prediction target 30 will occur during an evaluation period (for example, 10 years). The statistical model is generated in advance based on data on failures of the prediction target 30 accumulated in the failure statistics DB 50. The data on failures is, for example, operation history data, failure history data, maintenance history data, etc. of the prediction target 30 or its similar products. The failure prediction model may also be generated by directly referring to the failure rate database. The failure rate database is a database that stores data on the failure rate of each prediction target 30, and includes data on the failure risk of the prediction target 30 obtained by FMEA (Failure Mode and Effect Analysis) performed at the time of design. The failure rate database may also include data obtained from a public reliability database, a private database prepared independently, a combination of public data and private data weighted, etc.

[0023] The physical model simulates the aging deterioration of the prediction target 30, and outputs a time series of the failure probability of the prediction target 30 during an evaluation period (e.g., 10 years). The physical model is generated in advance based on the design data, manufacturing data, operation data, maintenance history data, etc. of the prediction target 30 accumulated in the design, manufacturing, and operation DB 51.

[0024] The failure rate and failure probability are related by the following formula (1): In the formula, λ(t) is the failure rate at time t, f(t) is the probability density function of the failure life, R(t) is the reliability at time t, and F(t) is the failure probability at time t. The failure probability F(t) is the complementary event of the reliability R(t).

[0025]

number

[0026] The first evaluation unit 101 evaluates a failure rate that indicates the probability of a random failure of the prediction target 30 using a statistical model.

[0027] The second evaluation unit 102 evaluates the failure probability, which represents the probability of failure of the prediction target 30 due to aging degradation, using a physical model.

[0028] The acquisition unit 103 acquires a qualitative evaluation result obtained by qualitatively evaluating the prediction target 30 for each predetermined evaluation item.

[0029] The correction unit 104 calculates a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by a correction coefficient based on the evaluation result of the qualitative evaluation of the prediction target 30 by the user A.

[0030] The first determination unit 105 determines, based on the corrected failure rate, whether the spare parts conditions for the prediction target 30 satisfy constraint conditions previously set for the prediction target 30. The maintenance plan DB 53 records in advance the spare parts conditions for each prediction target 30. The spare parts conditions include, for example, the quantity of spare parts (such as replacement parts in the event of a breakdown).

[0031] The second judgment unit 106 judges, based on the corrected failure probability, whether the preventive maintenance plan for the prediction target 30 satisfies a constraint condition previously set for the prediction target 30. The maintenance plan DB 53 records in advance the preventive maintenance plan for each prediction target 30. The preventive maintenance plan includes, for example, the number of times, the scope, and the timing of inspection, repair, replacement, etc. of the prediction target 30.

[0032] The proposing unit 107 proposes review of a countermeasure against a failure corresponding to at least one of the evaluation items when at least one of the first judging unit 105 and the second judging unit 106 judges that the constraint condition is not satisfied.

[0033] (Qualitative evaluation of the prediction target) FIG. 3 is a diagram for explaining the qualitative evaluation according to the first embodiment of the present disclosure. FIG. 3 shows an example of a score table for the qualitative evaluation of the prediction target 30. In the score table, evaluation items and evaluation points are predefined for each factor (subfactor) of the failure frequency correction coefficient (hereinafter also referred to as the "correction coefficient") of the prediction target 30. The factor (subfactor) links the "anxiety and uncertainty" related to the failure to the qualitative evaluation. The evaluation items may be common to all of the multiple prediction targets 30. The user checks the evaluation items shown in FIG. 3 for each prediction target 30, and inputs the evaluation results (evaluation points for each evaluation item) according to the check results into the failure prediction device 1 or records them in the evaluation DB 52. In the example of FIG. 3, the higher the evaluation point, the more likely the failure is to occur.

[0034] Next, the contents of the qualitative evaluation will be explained. In the example of Fig. 3, five factors of the failure frequency correction coefficient of the prediction target 30 are set: "Universal Subfactor", "Design Subfactor", "Stability and Matureness Subfactor", "Structural Subfactor", and "Monitoring Subfactor".

[0035] (1) Universal Subfactor The Universal Subfactor is a general failure frequency correction factor that is common to all forecast targets 30. For example, the Universal Subfactor includes evaluation items such as "Appearance," "Natural Environment," and "Uncertain Natural Load."

[0036] Appearance is an item for evaluating the appearance (dirt, peeling paint, etc.) of the prediction target 30 by comparing it with the standard state in the business field in which the prediction target 30 is operated. If dirt, etc. is left unattended, it is considered that the management state of the prediction target 30 is poor and that the prediction target 30 is in a state where it is likely to break down. For this reason, user A evaluates the prediction target 30 so that the worse the state of its appearance is, the higher the evaluation score. For example, as shown in FIG. 3, the user evaluates the appearance of the prediction target 30 in three stages: "better than standard (evaluation score: -1)", "standard (evaluation score: 0)", and "worse than standard (evaluation score: +1)". Note that the value of the evaluation score is an example. In this way, factors for which quantitative data is difficult to collect, such as the management state of the prediction target 30, can be obtained as quantitative evaluation results by user A's visual confirmation. In addition, since the appearance of the prediction target 30 is only evaluated in three stages by comparing it with other standard states, the evaluation time by user A can be reduced. Note that, for appearance, in order to prevent variation in evaluation results due to user subjectivity, the evaluation results may be determined by consensus of multiple users. Furthermore, a plurality of images showing standard conditions may be prepared in advance, and the failure prediction device 1 may compare and evaluate the images of the prediction target 30 taken by the user.

[0037] The natural environment is an item for evaluating the severity of the environment that the prediction target 30 is constantly exposed to in the region where it is operated. For example, if the prediction target 30 is operated in a cold or tropical region, failures due to temperature (such as breakage due to low temperature and overheating due to high temperature) are more likely to occur. In addition, if the prediction target 30 is a ship, a ship operated in an area with high wave height, wind strength, water temperature, and corrosive components (such as nitrate nitrogen) is more likely to have failures (fatigue damage, corrosion, etc.) than ships operated in other regions. The user refers to statistical data on the environment (air temperature, water temperature, etc.) recorded in the design, manufacturing, and operation DB51, etc., and evaluates the prediction target 30 so that the higher the severity of the operating region of the prediction target 30, the higher the evaluation score. In addition, the user may reduce the evaluation score if the prediction target 30 implements measures according to the environment (such as applying a thick coat of paint or periodically applying a highly durable paint, etc.).

[0038] Uncertain Natural Load is an item for evaluating maximum natural loads (natural disasters such as earthquakes, tsunamis, volcanoes, typhoons, hurricanes, and other abnormal weather) that occur stochastically in the area where the prediction target 30 is operated. For example, assume that the prediction target 30 is installed in an area at risk of volcanic eruption. Based on statistical data on the amount of volcanic ash fall during an eruption, the user evaluates the prediction target 30 so that the greater the amount of ash fall in the area where the prediction target 30 is operated, the higher the evaluation score. In addition to volcanic ash, the user may also evaluate the magnitude and frequency of huge wind loads such as typhoons and hurricanes, material deterioration due to increases in temperature and humidity, and the magnitude and frequency of huge fluid loads such as heavy rain.

[0039] In addition, the failure prediction device 1 may read statistical data on the environment and natural load of each region recorded in the design, manufacturing, and operation DB 51, etc., and data showing the implementation status of countermeasures according to the environment of the prediction target 30, and automatically evaluate the Natural Environment and Uncertain Natural Load.

[0040] (2) Design Subfactor The Design Subfactor is a factor of the failure frequency correction coefficient related to the design of each prediction target 30. For example, the Design Subfactor includes evaluation items such as "Complexity," "Standard and Code," and "Life Consumption."

[0041] Complexity is an item for evaluating whether the prediction target 30 has a complex structure. In the example of Fig. 3, the user evaluates the prediction target 30 so that the more complex the prediction target 30 is, the higher the evaluation score will be. The complexity of the structure is evaluated based on, for example, the number of nozzles of the prediction target 30, the number of pipe branches, etc.

[0042] Standard and Code is an item that evaluates whether the prediction target 30 complies with the design standard, from the viewpoint that equipment and structures that comply with the design standard are highly reliable and less likely to break down. The user evaluates the prediction target 30 that complies with the design standard so as to give it a low evaluation point. In addition, the user evaluates the prediction target 30 for which there is no design standard or which does not comply with the design standard so as to give it a high evaluation point.

[0043] Life Consumption is an item that evaluates the consumption of the design life of the prediction target 30. In the example of Figure 3, the consumption of the design life of the prediction target 30 (use time relative to the design life) is divided into four intervals (interval 1: less than n1%, interval 2: n1% to less than n2%, interval 3: n2% to less than n3%, interval 4: n3% or more), and an evaluation point is set that is weighted by the failure frequency corresponding to each interval. The failure frequency of each interval is obtained from data (bathtub curve) that indicates the failure rate of the prediction target 30 according to the passage of time, which is recorded in the design, manufacturing, and operation DB 51. The user specifies the interval of the design life from the use time of the prediction target 30 and determines the evaluation point.

[0044] The failure prediction device 1 may read out the design data and operation data of the prediction target 30 recorded in the design / manufacturing / operation DB 51, and automatically evaluate each evaluation item of the Design Subfactor.

[0045] (3) Stability and Matureness Subfactor The Stability and Matureness Subfactor is a factor of the failure frequency correction coefficient related to the stability and maturity of the prediction target 30. For example, the Stability and Matureness Subfactor includes evaluation items such as "Unexpected Damage Detection", "Unplanned Failure Detection", "Specialty and Matureness", "Similarity", and "Measures".

[0046] Unexpected Damage Detection is an item for evaluating the frequency of detecting unexpected damage that does not lead to the shutdown of the prediction target 30 during regular inspections, etc. User A evaluates the item so that the lower the detection frequency, the lower the evaluation point.

[0047] Unplanned Failure Detection is an item for evaluating the frequency of detection of a fatal failure that requires the operation of the prediction target 30 to be stopped during a regular inspection, etc. User A evaluates the item so that the lower the detection frequency, the lower the evaluation point.

[0048] Specialty and Matureness is an item that evaluates the specialness and maturity of the prediction target 30. When the prediction target 30 is a mass-produced product that has been used for a long time in multiple business fields, or when a certain number or more of equipment or structures with a similar design to the prediction target 30 are in use, statistical data on the failure of the prediction target 30 is sufficiently accumulated, and the uncertainty of the failure prediction is small. For this reason, User A evaluates such prediction targets 30 with low speciality and high maturity so that they are given a lower evaluation score than non-mass-produced products (high speciality) that are used only in a specific industry, or products with a similar design that have less than a certain number of operating records (low maturity). In addition, the evaluation score may be increased or decreased depending on the experience of the designer or manufacturer of the prediction target 30, the operator, etc. When the equipment or structures used in the product are newly developed or special, defects due to initial failures are likely to occur. In this embodiment, in the evaluation item of Specialty and Matureness, the evaluation results of the specialness and maturity of the prediction target 30 are reflected in the correction coefficient, so that the failure prediction device 1 can evaluate the failure rate and failure probability taking into account the risk of such initial failures.

[0049] Similarity is an item that evaluates the similarity of the cause or condition of damage or failure that occurred in the past for the prediction target 30. If a structure adjacent to the prediction target 30 is damaged by a load, the prediction target 30 may also be damaged by a similar load. In addition, if a part similar to the prediction target 30 (for example, a corresponding structure on the opposite side of the ship) or the same model is damaged, the prediction target 30 may also be damaged in the same way. For this reason, user A evaluates prediction targets 30 that have causes or conditions similar to equipment, etc. that have been damaged or broken in the past so that the evaluation score of the prediction target 30 for which inspection or countermeasures have not been completed is higher than that for which inspection or countermeasures have been completed. In addition, the evaluation score may be different between a case where the prediction target 30 is similar to equipment, etc. that has been damaged in the past and a case where the prediction target 30 is similar to equipment, etc. that has been broken in the past.

[0050] Measures is an item for evaluating whether the cause of past damage or failure of the prediction target 30 has been identified and whether measures have been taken. User A evaluates the prediction target 30 for which no measures have been taken against damage or failure so that the evaluation point is higher than that for which measures have been taken. In addition, the evaluation point may be different depending on whether the measures taken are permanent or temporary.

[0051] The failure prediction device 1 may read out the failure history data and maintenance history data of the prediction target 30 recorded in the failure statistics DB 50, and automatically evaluate each evaluation item of the Stability and Matureness Subfactor.

[0052] In the conventional technology, for example, when a countermeasure against a failure occurs in a prediction target, it may be evaluated as "the failure will not recur because the countermeasure was taken," and may not be reflected in the prediction model. However, in reality, depending on the location of the failure and the contents of the countermeasure, the failure may recur in the same location or an adjacent location, which is one of the factors that reduce the prediction accuracy of the prediction model. In contrast, in this embodiment, the failure status of the location where a failure such as a failure or damage occurs (equipment, structure), the occurrence frequency and the countermeasure, and the inspection and strength examination status of the similar location and adjacent location are evaluated in each evaluation item of the Stability and Matureness Subfactor (especially Unexpected Damage Detection, Unplanned Failure Detection, Similarity, Measures). This makes it possible to appropriately evaluate the location where a failure occurred in the past, the failure content, the countermeasure content, and the status of the adjacent or similar location in the qualitative evaluation. The failure prediction device 1 applies a correction coefficient based on the qualitative evaluation result to the failure rate and failure probability, thereby making it possible to accurately predict the failure risk taking into account past failures and the countermeasure content.

[0053] (4) Structural Subfactor Structural Subfactor is a factor of the failure frequency correction coefficient related to events specific to the structure. For example, Structural Subfactor includes evaluation items such as "Structural Criticality" and "Damage Rate".

[0054] Structural Criticality is an item that evaluates the impact of damage to the prediction target 30 (structure) on other structures. For example, if a crack occurs in the prediction target 30 and affects adjacent structures, User A evaluates the prediction target 30 so that its evaluation score is higher than if it does not affect adjacent structures.

[0055] Damage Rate is an item that evaluates the progress of progressive damage such as fatigue, corrosion, wear, creep, etc. of the prediction target 30 (structure). For example, if the prediction target 30 does not use a structure or components that slow down the progress of cracks, User A evaluates the score of this prediction target 30 to be higher than that of a prediction target that uses a structure or components that slow down the progress of cracks.

[0056] The failure prediction device 1 may read out the design data of the prediction target 30 recorded in the design / manufacturing / operation DB 51 and automatically evaluate each evaluation item of the structural subfactor.

[0057] For example, when the prediction target 30 is a structure, the risk of damage such as cracks occurring and the propagation of the cracks varies from part to part. For example, when a crack occurs in a certain part, the magnitude of the risk changes depending on whether a force acts in the direction of opening the crack or in the direction of closing the crack at the part to which the crack propagates. For this reason, in this embodiment, the risk of damage specific to the structure is evaluated for each part using each evaluation item of the Structural Subfactor. By reflecting this evaluation result in the correction coefficient, the failure prediction device 1 becomes able to accurately predict the failure risk taking into account the structure of the prediction target 30, the impact on adjacent structures, and the like.

[0058] (5) Monitoring Subfactor Monitoring Subfactor is a factor of the failure frequency correction coefficient related to inspection and monitoring. For example, Monitoring Subfactor includes "Discrepancy Factor", "Monitoring and Inspection (Time)", "Monitoring and Inspection (Space)", "Detectability (Visibility)", "Detectability (Accessibility)", etc.

[0059] The discrepancy factor is an item that evaluates the degree of discrepancy between the state quantity assumed in the design data and the state quantity measured during actual operation. For example, among multiple state quantities, a state quantity (e.g., temperature, etc.) that contributes most to damage or failure of the prediction target 30 is selected. Then, User A evaluates the prediction target 30 so that the evaluation score is higher as the ratio between the design conditions and the operating conditions of the prediction target 30 is larger.

[0060] Monitoring and Inspection (Time) is an item that evaluates the interval between inspections and monitoring. User A evaluates the longer the monitoring interval of the prediction target 30, the higher the evaluation score, and the shorter the monitoring interval, the lower the evaluation score. By expanding the scope of inspection and monitoring of the prediction target 30 or shortening the interval, it is possible to reduce the uncertainty of failure prediction.

[0061] Monitoring and Inspection (Space) is an item that evaluates the comprehensiveness of inspection and monitoring. User A divides the prediction target 30 into multiple parts and evaluates it so that the more inspections there are, the lower the evaluation score, and the fewer the inspections there are, the higher the evaluation score.

[0062] Detectability (Visibility) is an item for evaluating the ease of inspection of the prediction target 30. User A evaluates the prediction target 30 so that the more difficult it is to visually inspect it, the higher the evaluation score.

[0063] Detectability (Accessibility) is an item for evaluating the ease of inspecting the predicted target 30. User A evaluates the predicted target 30 so that the more difficult it is to approach and inspect it, the higher the evaluation score.

[0064] In addition, the failure prediction device 1 may automatically evaluate each evaluation item of the Monitoring Subfactor based on the design data and operation data recorded in the design, manufacturing, and operation DB 51, the preventive maintenance plan recorded in the maintenance plan DB 53, etc.

[0065] The evaluation items, evaluation stages, and points shown in FIG. 3 are merely examples and are not limited to these. These contents may be changed as desired depending on the model of the prediction target 30, the operational purpose (business field) of the prediction target 30, and the like. Furthermore, user A may exclude evaluation items that have already been incorporated into the statistical model or physical model from the targets of the qualitative evaluation. Alternatively, the correction unit 104 of the failure prediction device 1 may exclude the evaluation points of evaluation items that have already been incorporated into the statistical model or physical model when calculating the correction coefficient. For example, user A may arbitrarily set which evaluation items the correction unit 104 uses when calculating the correction coefficient.

[0066] (Processing flow) FIG. 4 is a flowchart showing an example of processing of the failure prediction device according to the first embodiment of the present disclosure. Hereinafter, a method for evaluating a failure rate and a failure probability using the failure prediction device 1, as well as a method for evaluating spare part conditions and a preventive maintenance plan, will be described with reference to FIG.

[0067] First, the setting unit 100 selects a failure prediction model to be used for predicting the failure rate and failure probability for each of all prediction targets 30 of the product 3 (step S100). For example, the setting unit 100 selects the failure prediction model according to an operation of the user A via the input device 2B.

[0068] The setting unit 100 also sets an objective function and constraint conditions related to the maintenance of the prediction target 30 (step S101). The objective function is, for example, the reliability, availability, and maintenance cost of the product 3. The constraint conditions are, for example, lower limits of the reliability and availability of the product 3, and an upper limit of the total cost. For example, the setting unit 100 sets the objective function and constraint conditions according to an operation of the user A via the input device 2B.

[0069] Next, the failure prediction device 1 evaluates and corrects the failure rate and failure probability of the prediction target 30.

[0070] First, a method for evaluating and correcting the failure rate will be described. The first evaluation unit 101 calculates the failure rate of the prediction target 30 based on the statistical model selected by the setting unit 100 (step S102). When there are multiple prediction targets 30, the first evaluation unit 101 calculates the failure rates of all the prediction targets 30.

[0071] Next, the acquisition unit 103 acquires the qualitative evaluation result of the prediction target 30 by the user A (step S103). For example, the acquisition unit 103 acquires the qualitative evaluation result input by the user A through the input device 2B. In addition, when the user A records the qualitative evaluation result in the evaluation DB 52, the acquisition unit 103 may read out the qualitative evaluation result of the prediction target 30 from the evaluation DB 52.

[0072] The correction unit 104 multiplies the failure rate calculated in step S102 by a correction coefficient based on the qualitative evaluation result to calculate a corrected failure probability (step S104). The correction unit 104 according to this embodiment uses the total value of the evaluation points of each evaluation item in the qualitative evaluation as the correction coefficient. Note that in other embodiments, when the total value of the evaluation points is a negative value, the correction unit 104 may use the reciprocal or exponent of the absolute value of the evaluation points as the correction coefficient.

[0073] FIG. 5 is a first diagram illustrating an example of a corrected failure rate according to the first embodiment of the present disclosure. Graph G10 shown in Fig. 5 is an example of a failure rate and a corrected failure rate in a certain evaluation period (for example, 10 years from the present). The solid line in graph G10 indicates the failure rate (uncorrected failure rate) calculated by the first evaluation unit 101. The dashed line in graph G10 indicates the corrected failure rate calculated by the correction unit 104 by multiplying the uncorrected failure rate by a correction coefficient. In this way, the correction unit 104 can calculate a corrected failure rate that takes into account factors that cannot be considered in a statistical model, such as the management status, operational performance, and monitoring interval of each prediction target 30.

[0074] FIG. 6 is a second diagram showing an example of the corrected failure rate according to the first embodiment of the present disclosure. The correction unit 104 may change the correction coefficient according to the consumption of the design life by simulating the bathtub curve of the prediction target 30 as shown in graph G11 of FIG. 6. For example, as shown in FIG. 3, the evaluation score of Life Consumption is set to different values ​​such as {1, 0, 1, 2} according to the section (section 1 to section 4) of the usage time of the prediction target 30. The correction unit 104 sums up the evaluation score of Life Consumption and other evaluation points for each section of Life Consumption to obtain four correction coefficients corresponding to each section. The correction unit 104 sums up the evaluation score "1" of Section 1 of Life Consumption and the evaluation points of other evaluation items to obtain the correction coefficient of Section 1. The correction unit 104 also sums up the evaluation score "0" of Section 2 of Life Consumption and the evaluation points of other evaluation items to obtain the correction coefficient of Section 2. The correction unit 104 similarly obtains correction coefficients for Sections 3 to 4. The dashed line in the graph G11 indicates the corrected failure rate calculated by the correction unit 104 by multiplying the uncorrected failure rate by the correction coefficient for each of the sections 1 to 4. In this way, the correction unit 104 can obtain a more accurate corrected failure rate according to the consumption (usage time) of the design life of the prediction target 30.

[0075] Furthermore, when there are a plurality of prediction targets 30, the correction unit 104 calculates the corrected failure rates of all the prediction targets 30 respectively.

[0076] Next, the first judgment unit 105 evaluates the reliability, availability, and total cost including the spare parts holding cost of the prediction target 30 based on the corrected failure rate, the current spare parts conditions (number of spare parts) of the prediction target 30, and the evaluation period (step S105).

[0077] The first judging unit 105 also judges whether the reliability, availability, and total cost evaluated in step S105 satisfy the constraint conditions set in step S101 (step S106).

[0078] If the first judging unit 105 judges that all of the reliability, the availability, and the total cost satisfy the constraint conditions (step S106: YES), the failure prediction device 1 ends the process.

[0079] On the other hand, if the first judgment unit 105 judges that at least one of the reliability, availability, and total cost does not satisfy the constraint conditions (step S106: NO), the proposal unit 107 proposes a review of the failure countermeasures for the prediction object 30 (step S107).

[0080] Then, User A reviews the failure countermeasures so that the evaluation score of at least one evaluation item changes, and then performs the qualitative evaluation again. For example, if the reliability of the prediction target 30 is below the constraint condition, User A reviews the failure countermeasures so that the reliability satisfies the constraint condition, that is, so that the correction coefficient (evaluation score of the qualitative evaluation) becomes smaller. Specifically, if the prediction target 30 is a newly developed device and there is room for design changes, the evaluation score of Specialty and Matureness can be lowered by taking measures to change the parts that make up the prediction target 30 to mass-produced products with high maturity. In addition, the evaluation score of Monitoring and Inspection can be lowered by taking measures to shorten the monitoring interval of the prediction target 30 and to increase the number of monitored points. Furthermore, if it is before design and manufacturing, the evaluation score of Design can be lowered by making the prediction target 30 comply with the design standard, or the original failure probability can be lowered by changing the design conditions to reduce the load or improving the material strength.

[0081] Next, the acquisition unit 103 reacquires the qualitative evaluation result of the prediction target 30 by the user A (step S108). Then, the first evaluation unit 101 recalculates the corrected failure rate using the correction coefficient based on the reacquired qualitative evaluation result (step S109). These processes are the same as steps S103 to S104.

[0082] Furthermore, user A may review the constraint conditions (reliability, availability, total cost). When user A inputs the changed constraint conditions, the setting unit 100 resets the constraint conditions (step S110).

[0083] The first judgment unit 105 reevaluates the reliability, availability, and total cost of the product 3 based on the corrected failure rate of each prediction target 30 recalculated in step S108 and the spare part conditions of the current prediction target 30 (step S111). After that, the process returns to step S106, and the first judgment unit 105 judges whether the reevaluated reliability, availability, and total cost satisfy the constraint conditions. The failure prediction device 1 repeatedly executes the processes of steps S107 to S111 until the constraint conditions are satisfied.

[0084] Next, a method for evaluating and correcting the failure probability will be described. The second evaluation unit 102 calculates the failure probability of the prediction target 30 based on the physical model selected by the setting unit 100 (step S112). When there are multiple prediction targets 30, the second evaluation unit 102 calculates the failure probabilities of all the prediction targets 30.

[0085] Next, the acquisition unit 103 acquires the qualitative evaluation result of the prediction target 30 by the user A (step S113). This process is similar to step S103.

[0086] The correction unit 104 multiplies the failure probability calculated in step S111 by a correction coefficient based on the qualitative evaluation result to calculate a corrected failure probability (step S114). The correction unit 104 according to the present embodiment obtains the total value of the evaluation points as the correction coefficient, similarly to step S104. In addition, in other embodiments, when the total value of the evaluation points is a negative value, the correction unit 104 may use the reciprocal or exponent of the absolute value of the evaluation points as the correction coefficient.

[0087] FIG. 7 is a first diagram illustrating an example of corrected failure probability according to the first embodiment of the present disclosure. Graph G20 shown in FIG. 7 is an example of a time series of failure probability and corrected failure probability in a certain evaluation period (for example, 10 years from the present). The solid line in graph G20 indicates the time series of failure probability (uncorrected failure probability) calculated by the second evaluation unit 102. The dashed line in graph G20 indicates the time series of corrected failure probability calculated by the correction unit 104 by multiplying the uncorrected failure probability by a correction coefficient. In this way, the correction unit 104 can calculate a corrected failure probability that takes into account factors that are difficult to consider in a physical model, such as the management status, operational performance, and monitoring interval of each prediction target 30.

[0088] FIG. 8 is a second diagram illustrating an example of the corrected failure probability according to the first embodiment of the present disclosure. The prediction target 30 may be replaced with a new one during the evaluation period. In the example of FIG. 8, it is assumed that the prediction target 30 is replaced at time t1. At this time, the failure probability predicted from the physical model (uncorrected failure probability) drops once at the time of replacement (time t1) as shown by the solid line in graph G21, but rises again as time passes. The correction unit 104 calculates a time series of corrected failure probability as shown by the dashed line in graph G21 by multiplying this uncorrected failure probability by the correction coefficient. In this way, the correction unit 104 can also use the correction coefficient to calculate the corrected failure probability of the prediction target 30 that is replaced, repaired, or the like during the evaluation period.

[0089] FIG. 9 is a third diagram illustrating an example of the corrected failure probability according to the first embodiment of the present disclosure. The correction unit 104 may change the correction coefficient according to the consumption of the design life by simulating the bathtub curve of the prediction target 30 as shown in graph G22 of FIG. 9. For example, the correction unit 104 sums up the evaluation points of Life Consumption and other evaluation points for each section of Life Consumption to obtain four correction coefficients corresponding to each section. The dashed line of graph G22 indicates the corrected failure probability calculated by the correction unit 104 by multiplying the pre-correction failure probability by the correction coefficient for each section. As in the example of FIG. 9, when the prediction target 30 is replaced during the evaluation period, the correction unit 104 obtains the corrected failure probability by multiplying the correction coefficient for each section for each period before and after the replacement (the period before time t1 and the period after time t1). Note that, when replacement is not performed as in the example of FIG. 7, the correction unit 104 divides the entire evaluation period into four sections and obtains the corrected failure probability by multiplying the correction coefficient for each section. In a physical model that predicts the failure probability due to aging deterioration, it is difficult to simulate an early failure (failure probability in section 1) represented by a bathtub curve. However, by taking into account the bathtub curve of the prediction target 30 in this way, the correction unit 104 can calculate a corrected failure probability that reflects early failures that are difficult to predict with a physical model.

[0090] Furthermore, when there are multiple prediction targets 30, the correction unit 104 calculates the corrected failure probabilities for all the prediction targets 30 respectively.

[0091] Next, the second judgment unit 106 evaluates the reliability, availability, and total cost of the prediction object 30 based on the corrected failure probability and the current preventive maintenance plan for the prediction object 30 (a preventive maintenance plan including the number of inspections, repairs, replacements, etc., and the timing of such inspections, repairs, replacements, etc., for the prediction object 30) (step S115).

[0092] Furthermore, the second judging unit 106 judges whether the reliability, availability, and total cost evaluated in step S115 satisfy the constraint conditions set in step S101 (step S116).

[0093] If the second judging unit 106 judges that all of the reliability, the availability, and the total cost satisfy the constraint conditions (step S116: YES), the failure prediction device 1 ends the process.

[0094] On the other hand, if the second judgment unit 106 judges that at least one of the reliability, the availability rate, and the total cost does not satisfy the constraint condition (step S116: NO), the proposal unit 107 proposes reviewing the failure countermeasures for the prediction target 30 (step S117). This process is the same as step S107. User A reviews the failure countermeasures so as to change the evaluation score of at least one evaluation item, and then performs the qualitative evaluation again.

[0095] Next, the acquisition unit 103 reacquires the qualitative evaluation result of the prediction target 30 by the user A (step S118). Then, the second evaluation unit 102 recalculates the corrected failure probability using the correction coefficient based on the reacquired qualitative evaluation result (step S119). These processes are the same as steps S113 to S114.

[0096] Furthermore, user A may review the constraint conditions (reliability, availability, total cost). When user A inputs the changed constraint conditions, the setting unit 100 resets the constraint conditions (step S120).

[0097] The second judgment unit 106 reevaluates the reliability, availability, and total cost of the product 3 based on the corrected failure probability of each prediction target 30 recalculated in step S119 and the current preventive maintenance plan for the prediction target 30 (step S121). After that, returning to step S116, the second judgment unit 106 judges whether the reevaluated reliability, availability, and total cost satisfy the constraint conditions. The failure prediction device 1 repeatedly executes the processes of steps S117 to S121 until the constraint conditions are satisfied.

[0098] The correction unit 104 may calculate a common correction coefficient for the failure rate and the failure probability from the qualitative evaluation result, or may calculate different correction coefficients. For example, when there is an evaluation item already incorporated in the physical model or statistical model among the evaluation items of the qualitative evaluation (FIG. 3), the correction unit 104 may exclude the evaluation point of this evaluation item from the correction coefficient. Specifically, when the failure rate is calculated from failure statistics of equipment and facilities having the same degree of complexity as the prediction target 30, the correction unit 104 calculates the correction coefficient for the failure rate excluding the evaluation point of Complexity. Alternatively, when the structural complexity of the prediction target 30 is taken into account in a structural analysis model (physical model) such as FEM (Finite Element Model), the correction unit 104 calculates the correction coefficient for the failure probability excluding the evaluation point of Complexity. Also, for example, when the evaluation content of the Natural Environment (such as wave height) is taken into account in the physical model, the correction coefficient for the failure probability may be calculated excluding the evaluation point of the Natural Environment. This allows the correction unit 104 to more accurately correct the failure rate and the failure probability.

[0099] (Action and effect) As described above, the failure prediction device 1 according to this embodiment includes a first evaluation unit 101 that evaluates the failure rate of the prediction object 30 using a statistical model, a second evaluation unit 102 that evaluates the failure probability of the prediction object 30 using a physical model, an acquisition unit 103 that acquires qualitative evaluation results obtained by qualitatively evaluating the prediction object 30 for each predetermined evaluation item, and a correction unit 104 that calculates a corrected failure rate and a corrected failure probability by multiplying the failure rate and failure probability by correction coefficients based on the qualitative evaluation results, respectively.

[0100] In this way, the failure prediction device 1 can correct the prediction results of the failure rate and the failure probability, and can accurately predict the risk of failure of the prediction target 30. In addition, the failure prediction device 1 can evaluate both accidental failures and failures due to aging from the shared perspective of the probability of failure by applying a correction coefficient based on the qualitative evaluation result of the prediction target 30 to both the failure rate and the failure probability.

[0101] In addition, the failure prediction device 1 further includes a first judgment unit 105 that judges whether the spare part conditions of the prediction object 30 satisfy the constraint conditions based on the corrected failure rate, a second judgment unit 106 that judges whether the preventive maintenance plan of the prediction object 30 satisfies the constraint conditions based on the corrected failure probability, and a proposal unit 107 that proposes a review of failure countermeasures corresponding to at least one evaluation item when at least one of the first judgment unit 105 and the second judgment unit 106 judges that the constraint conditions are not satisfied.

[0102] In this way, the failure prediction device 1 can more accurately evaluate whether the prediction target 30 satisfies the constraint conditions by using the failure rate and failure correction rate after correction based on the qualitative evaluation result. Furthermore, if the prediction target 30 does not satisfy the constraint conditions, the failure prediction device 1 can suggest to the user A to review the failure countermeasures for the prediction target 30.

[0103] <Second embodiment> Next, a failure prediction device 1 according to a second embodiment of the present disclosure will be described with reference to Fig. 10 to Fig. 16. Components common to the first embodiment will be given the same reference numerals and detailed description will be omitted.

[0104] (Functional configuration) FIG. 10 is a diagram illustrating a functional configuration of a failure prediction device according to the second embodiment of the present disclosure. As shown in FIG. 10, the CPU 10 of the failure prediction device 1 according to this embodiment further has the functions of a first adjustment unit 108 and a second adjustment unit 109.

[0105] The first adjustment unit 108 adjusts the spare part conditions based on the corrected failure rate.

[0106] The second adjustment unit 109 adjusts the preventive maintenance plan based on the corrected failure probability.

[0107] (Processing flow) FIG. 11 is a flowchart illustrating an example of processing of the failure prediction device according to the second embodiment of the present disclosure. Hereinafter, a method for evaluating a failure rate and a failure probability using the failure prediction device 1, as well as a method for evaluating spare part conditions and a preventive maintenance plan, will be described with reference to FIG.

[0108] First, the setting unit 100 selects a failure prediction model (step S200) and sets an objective function and constraint conditions (step S201). These processes are the same as those in the first embodiment (steps S100 to S101 in FIG. 4).

[0109] Next, in steps S202 to S204, the failure prediction device 1 evaluates and corrects the failure rate of the prediction object 30. Note that the processes in steps S202 to S204 are the same as those in the first embodiment (steps S102 to S104 in FIG. 4).

[0110] Next, the first adjustment unit 108 adjusts the spare parts conditions for the prediction target 30 (step S205). For example, assume that the product 3 is equipment used in production. When the prediction target 30 breaks down, if no spare spare parts are prepared, the period during which the product 3 is out of operation will be extended, and the profit and loss amount due to production loss will increase. If a sufficient number of spare parts are prepared, the production loss can be reduced. On the other hand, the more spare parts there are, the higher the cost associated with the spare parts will be. Therefore, the first adjustment unit 108 determines the number of spare parts that can maximize the amount of production loss reduction (i.e., maximize the profit from the spare parts) while suppressing the increase in the cost of the spare parts.

[0111] FIG. 12 is a first diagram for explaining a function of the first adjustment unit according to the second embodiment of the present disclosure. As shown in FIG. 12, the first adjustment section 108 obtains a probability distribution G30 of the number of failures in an evaluation period (for example, 10 years) of the prediction target 30 based on the corrected failure rate.

[0112] FIG. 13 is a second diagram for explaining the function of the first adjustment unit according to the second embodiment of the present disclosure. Furthermore, the first adjustment unit 108 obtains functions representing the amount of production loss reduction, the cost of spare parts, and the profit from spare parts as an index (KPI) for evaluating the number of spare parts based on the probability distribution G30 of the number of failures. Graph G31 in Fig. 13 represents the predicted amounts of the amount of production loss reduction, the cost of spare parts, and the profit from spare parts according to the number of spare parts. In the example of Fig. 13, since the profit amount is maximized when the number of spare parts is 12, the first adjustment unit 108 sets the spare parts condition (number of spare parts) of the prediction target 30 to "12 pieces."

[0113] Next, the first judgment unit 105 evaluates the reliability, availability, and total cost of maintenance of the prediction target 30 based on the corrected failure rate and the spare parts conditions (number of spare parts) of the prediction target 30 adjusted in step S205 (step S206). The subsequent processes of steps S207 to S212 are the same as those of the first embodiment (steps S106 to S111 in FIG. 4).

[0114] Furthermore, in steps S213 to S215, the failure prediction device 1 evaluates and corrects the failure probability of the prediction object 30. Note that the processes in steps S213 to S215 are the same as those in the first embodiment (steps S112 to S114 in FIG. 4).

[0115] Next, the second adjustment unit 109 adjusts the preventive maintenance plan (the number of times and timing of inspection, repair, replacement, etc. of the prediction target 30) (step S216). The second adjustment unit 109 calculates the failure risk of the prediction target 30 using, for example, a Risk-Based Maintenance (RBM) method, and adjusts the maintenance period (for example, replacement period).

[0116] FIG. 14 is a first diagram for explaining a function of the second adjustment unit according to the second embodiment of the present disclosure. Graph G40 shown in FIG. 14 is a simulation of the amount of damage to the prediction target 30, with the uncertainty from the worst case to the best case assumed for the degree of deterioration occurring in the prediction target 30 set as a probability distribution. In the example of FIG. 14, the prediction target 30 is replaced twice at a predetermined interval during the evaluation period (service provision period). In addition, the probability Pf that the amount of damage to the prediction target 30 exceeds the damage limit in each replacement period represents the failure probability of the prediction target. The second adjustment unit 109 uses the corrected failure probability obtained by multiplying this failure probability Pf by a correction coefficient to obtain the expected value of the number of failures of the prediction target 30 during the evaluation period.

[0117] FIG. 15 is a second diagram for explaining the function of the second adjustment unit according to the second embodiment of the present disclosure. The second adjustment unit 109 obtains the preventive maintenance cost and the expected value of the corrective maintenance cost of the prediction target 30 based on the expected value of the number of failures of the prediction target 30. The preventive maintenance cost is the cost required for replacing the prediction target 30. The corrective maintenance cost is obtained as an expected value by multiplying the cost required for actual corrective maintenance by the corrected failure probability. Graph G41 in FIG. 15 shows an example of the relationship between the length of the replacement cycle of the prediction target 30 and the preventive maintenance cost, corrective maintenance cost, and total maintenance cost, which is the sum of these, of the prediction target 30. The longer the replacement cycle, the smaller the preventive maintenance cost. On the other hand, the longer the replacement cycle of the prediction target 30, the larger the corrective maintenance cost. The second adjustment unit 109 sets the replacement cycle at which the total maintenance cost, which is the sum of the preventive maintenance cost and the corrective maintenance cost, is the smallest, as the optimal replacement cycle.

[0118] Furthermore, for example, when the prediction target 30 is a structure made up of a plurality of components, maintenance may be performed to replace a part of the prediction target 30 at each replacement cycle. In this case, the second adjustment unit 109 may set the replacement parts (maintenance range) at each replacement cycle. For example, the second adjustment unit 109 sets that, among the components constituting the prediction target 30, the top X% of components with the highest failure probability are replaced at each replacement cycle. Note that the replaced components have a lower failure probability, so the components to be replaced at each replacement cycle change. This allows the second adjustment unit 109 to set a preventive maintenance plan that can reduce preventive maintenance costs compared to replacing all components at each replacement cycle while preventing the prediction target 30 from exceeding the damage limit.

[0119] FIG. 16 is a third diagram for explaining the function of the second adjustment unit according to the second embodiment of the present disclosure. The second adjustment unit 109 may receive a selection of a maintenance menu (preventive maintenance cycle, maintenance range) by the user A. At this time, the second adjustment unit 109 presents a graph G42 showing the Pareto efficiency for each maintenance menu to the user A, as shown in FIG. 16, for example. The graph G42 is an example showing the relationship between preventive maintenance cost and reliability for each maintenance menu. The reliability threshold and the maintenance cost threshold show the constraint conditions set in step S101. In another embodiment, the graph G42 may show the relationship between reliability and total cost, and the total cost and preventive maintenance cost. The total cost is the sum of the preventive maintenance cost and the impact of the accident, and the impact of the accident includes the restoration work cost, the accident investigation cost, and the production loss during the breakdown stop period. The user A refers to this graph G42 and selects a maintenance menu that satisfies both the constraint conditions of reliability and the maintenance cost. The user A may also check the contents of the maintenance menu that satisfies the constraint conditions and select the optimal one. The second adjustment unit 109 sets a preventive maintenance plan based on the maintenance menu selected by the user A.

[0120] Next, the second judgment unit 106 evaluates the reliability, availability, and total cost of maintenance of the prediction target 30 based on the corrected failure probability and the preventive maintenance plan of the prediction target 30 adjusted in step S217 (step S217). The subsequent processes of steps S218 to S223 are the same as those of the first embodiment (steps S116 to S121 in FIG. 4).

[0121] (Action and effect) As described above, the failure prediction device 1 according to this embodiment further includes the first adjustment unit 108 that adjusts the spare part conditions based on the corrected failure rate.

[0122] In this way, the failure prediction device 1 can accurately set optimal spare part conditions based on the corrected failure rate.

[0123] Moreover, the failure prediction device 1 further includes a second adjustment unit 109 that adjusts the preventive maintenance plan based on the corrected failure probability.

[0124] In this way, the failure prediction device 1 can accurately set an optimal preventive maintenance plan based on the corrected failure probability.

[0125] <Third embodiment> Next, a failure prediction device 1 according to a second embodiment of the present disclosure will be described with reference to Fig. 17 to Fig. 18. Components common to the first and second embodiments are given the same reference numerals and detailed description thereof will be omitted.

[0126] (Functional configuration) FIG. 17 is a diagram illustrating a functional configuration of the failure prediction device according to the third embodiment of the present disclosure. As shown in FIG. As shown in FIG. 17, the CPU 10 of the failure prediction device 1 according to this embodiment further has the function of a third adjustment section 110.

[0127] The third adjustment unit 110 adjusts the spare part conditions and preventive maintenance plan for each of the multiple prediction targets 30 based on the corrected failure rate, corrected failure probability, spare part conditions, and preventive maintenance plan for each of the multiple prediction targets 30 so as to satisfy the predetermined constraint conditions for the entire product 3.

[0128] (Processing flow) FIG. 18 is a flowchart illustrating an example of processing of the failure prediction device according to the third embodiment of the present disclosure. Hereinafter, a method for evaluating a failure rate and a failure probability using the failure prediction device 1, as well as a method for evaluating spare part conditions and a preventive maintenance plan, will be described with reference to FIG.

[0129] First, the setting unit 100 selects a failure prediction model for each of the multiple prediction targets 30 (step S300), and sets an objective function and constraint conditions for the entire product 3 (step S301). These processes are the same as those in the first embodiment (steps S100 to S101 in FIG. 4).

[0130] Next, the failure prediction device 1 evaluates and corrects the failure rate of the prediction target 30 in steps S302 to S304, and adjusts the spare part conditions in step S305. The processes in steps S302 to S305 are the same as those in the second embodiment (steps S202 to S205 in FIG. 11).

[0131] Furthermore, the failure prediction device 1 evaluates and corrects the failure probability of the prediction object 30 in steps S306 to S308, and adjusts the preventive maintenance plan in step S309. The processes in steps S306 to S309 are the same as those in the second embodiment (steps S213 to S216 in FIG. 11).

[0132] Next, the third adjustment unit 110 calculates the sensitivity of the objective variable (reliability or total cost) of all prediction targets 30 of the product 3 based on the various parameters (Figures 12 to 16) calculated by the first adjustment unit 108 and the second adjustment unit 109 when adjusting the spare part conditions and the preventive maintenance plan (step S310).

[0133] For example, the third adjustment unit 110 creates a Pareto efficiency graph G42 shown in FIG. 16 for each prediction target 30. In addition, the third adjustment unit 110 sets, as the objective variable, one of the reliability of the prediction target 30 or the total maintenance cost, whichever is set in advance or specified by the user A. Then, the third adjustment unit 110 sets the preventive maintenance cost of the prediction target 30 as the explanatory variable M i As a result, the objective variable of the 30 prediction targets (reliability R i , or total cost T i ) to create a regression equation for the product 3, where i is an integer between 1 and n, where product 3 has n prediction targets 30.

[0134] In addition, the third adjustment unit 110 adjusts the preventive maintenance cost M i,j The reliability sensitivity and total cost sensitivity of the prediction target 30 at (j=0) are calculated. The initial maintenance plan is a maintenance plan set in advance by user A. In addition, multiple initial maintenance plans may be set. The reliability sensitivity is expressed by the following formula (2). In addition, the total cost sensitivity is expressed by the following formula (3). Here, i represents the equipment ID, and j represents the maintenance condition ID for the equipment or facility.

[0135]

number

[0136]

number

[0137] The third adjustment unit 110 calculates the reliability sensitivity and the total cost sensitivity for all prediction targets 30 that constitute the product 3.

[0138] Next, the third adjustment unit 110 sets a maintenance menu for the entire product 3 based on the calculated reliability sensitivity and total cost sensitivity (step S311).

[0139] Specifically, the third adjustment unit 110 adjusts the explanatory variable (preventive maintenance cost M i ), and low sensitivity variables, which are explanatory variables that have a small effect on the objective variable. Then, the third adjustment unit 110 sets a maintenance menu for the product 3 so as to increase the high sensitivity variables (increase preventive maintenance costs). The third adjustment unit 110 may also set a maintenance menu for the product 3 so as to maintain the current state of the low sensitivity variables or to decrease them.

[0140] For example, when the objective variable is reliability, the third adjustment unit 110 selects an explanatory variable (preventive maintenance cost M i ) as high sensitivity variables, and the top Y% (for example, 10%) of explanatory variables that have a small effect on reliability are extracted as low sensitivity variables. Note that X and Y are arbitrary values, and different values ​​may be set for each. The third adjustment unit 110 extracts the preventive maintenance cost M i In other words, preventive maintenance costs M i The maintenance menu is set so as to prioritize preventive maintenance of the prediction target 30 related to the third adjustment unit 110. i For the prediction target 30 related to the above, the priority of preventive maintenance may be maintained or may be lowered.

[0141] In addition, for example, when the objective variable is the total cost, the third adjustment unit 110 selects an explanatory variable (preventive maintenance cost M i ) as high sensitivity variables, and the top Y% of explanatory variables with a small effect on the total cost are extracted as low sensitivity variables. The third adjustment unit 110 extracts the preventive maintenance cost M i In other words, preventive maintenance costs M i The maintenance menu is set so as to prioritize preventive maintenance of the prediction target 30 related to the maintenance cost M i For the prediction target 30 related to the above, the priority of preventive maintenance may be maintained or may be lowered.

[0142] Next, the first judgment unit 105 (or the second judgment unit 106) evaluates the reliability, availability, and total cost of the entire product 3 based on the adjusted failure rate and adjusted failure probability of each prediction target 30 and the maintenance menu set in step S311 (step S312). This process is the same as that of the second embodiment (steps S206 and S217 in FIG. 11). Moreover, the subsequent processes of steps S313 to S318 are the same as those of the second embodiment (steps S207 to S212 and S218 to S223 in FIG. 11).

[0143] In addition, the third adjustment unit 110 adjusts the preventive maintenance cost M i,j While calculating the sensitivity for (j=1, 2, . . . ), a maintenance menu with maximum reliability or minimum total cost may be searched for within a range that satisfies the constraint conditions.

[0144] (Action and effect) As described above, the failure prediction device 1 according to this embodiment further includes a third adjustment unit 110 that adjusts the spare part conditions and the preventive maintenance plan for the entire product 3 based on the corrected failure rate, corrected failure probability, spare part conditions, and preventive maintenance plan for each of the multiple prediction targets 30 so as to satisfy the constraint conditions set in advance for the entire product 3.

[0145] In this way, the failure prediction device 1 can optimize the maintenance plan for the entire product 3.

[0146] Although the embodiments of the present disclosure have been described in detail above, the present invention is not limited to these embodiments and some design changes and the like are possible without departing from the technical spirit of the present invention.

[0147] <Additional Notes> The failure prediction device, the failure prediction method, and the program described in the above-described embodiments can be understood, for example, as follows.

[0148] (1) According to a first aspect of the present disclosure, a failure prediction device (1) includes a first evaluation unit (101) that evaluates a failure rate representing the probability of an accidental failure of the prediction target (30) using a statistical model generated based on data regarding failures of the prediction target (30); a second evaluation unit (102) that evaluates a failure probability representing the probability of a failure due to aging of the prediction target (30) using a physical model generated based on design data of the prediction target (30); an acquisition unit (103) that acquires a qualitative evaluation result obtained by qualitatively evaluating the prediction target (30) for each predetermined evaluation item; and a correction unit (104) that calculates a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by correction coefficients based on the qualitative evaluation results, respectively.

[0149] In this way, the failure prediction device can correct the prediction results of the failure rate and failure probability, and can accurately predict the risk of failure of the prediction target. In addition, the failure prediction device can evaluate both random failures and failures due to aging from the shared perspective of the probability of failure by applying a correction coefficient based on the qualitative evaluation result of the prediction target to both the failure rate and failure probability.

[0150] (2) According to a second aspect of the present disclosure, the failure prediction device (1) according to the first aspect further includes a first determination unit (105) that determines, based on a corrected failure rate, whether a spare part condition of the prediction target (30) satisfies a constraint condition previously set for the prediction target (30); a second determination unit (106) that determines, based on a corrected failure probability, whether a preventive maintenance plan for the prediction target (30) satisfies the constraint condition previously set for the prediction target (30); and a proposal unit (107) that proposes a review of a failure countermeasure corresponding to at least one evaluation item when at least one of the first determination unit (105) and the second determination unit (106) determines that the constraint condition is not satisfied.

[0151] In this way, the failure prediction device can more accurately evaluate whether the prediction target satisfies the constraint conditions by using the failure rate and failure correction rate after correction based on the qualitative evaluation result. Furthermore, if the prediction target does not satisfy the constraint conditions, the failure prediction device can suggest to the user to review the failure countermeasures for the prediction target.

[0152] (3) According to a third aspect of the present disclosure, the failure prediction device (1) according to the second aspect further includes a first adjustment unit (108) that adjusts spare part conditions based on the corrected failure rate.

[0153] In this way, the failure prediction device can accurately set optimal spare part conditions based on the corrected failure rate.

[0154] (4) According to a fourth aspect of the present disclosure, the failure prediction device (1) according to the second or third aspect further includes a second adjustment unit (109) that adjusts the preventive maintenance plan based on the corrected failure probability.

[0155] In this way, the failure prediction device can accurately set an optimal preventive maintenance plan based on the corrected failure probability.

[0156] (5) According to a fifth aspect of the present disclosure, the failure prediction device (1) according to any one of the first to fourth aspects further includes a third adjustment unit (110) that adjusts spare part conditions and a preventive maintenance plan for the entire product (3) consisting of a plurality of prediction targets (30) based on the corrected failure rate, the corrected failure probability, the spare part conditions, and the preventive maintenance plan for each of the plurality of prediction targets (30) so as to satisfy predetermined constraint conditions for the entire product (3).

[0157] In this way, the failure prediction device can optimize the maintenance plan for the entire product.

[0158] (6) According to a sixth aspect of the present disclosure, a failure prediction method includes the steps of: evaluating a failure rate representing a probability of random failure of the prediction target (30) using a statistical model generated based on data regarding failures of the prediction target (30); evaluating a failure probability representing a probability of failure due to aging of the prediction target (30) using a physical model generated based on design data of the prediction target (30); acquiring a qualitative evaluation result obtained by qualitatively evaluating the prediction target (30) for each predetermined evaluation item; and calculating a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by a correction coefficient based on the qualitative evaluation result, respectively.

[0159] (7) According to a seventh aspect of the present disclosure, the program causes a computer of the failure prediction device (1) to execute the following steps: evaluating a failure rate representing the probability of an accidental failure of the prediction target (30) using a statistical model generated based on data regarding failures of the prediction target (30); evaluating a failure probability representing the probability of a failure due to aging of the prediction target (30) using a physical model generated based on design data of the prediction target (30); acquiring a qualitative evaluation result obtained by qualitatively evaluating the prediction target (30) for each predetermined evaluation item; and calculating a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by a correction coefficient based on the qualitative evaluation result, respectively. [Explanation of symbols]

[0160] 1. Failure prediction device 10 CPU 100 Setting section 101 First Evaluation Section 102 Second Evaluation Section 103 Acquisition Department 104 Correction section 105 1st Judgment Section 106 Second Judgment Section 107 Proposal Department 108 1st adjustment section 109 2nd adjustment section 110 Third adjustment section 11. Memory 12. Storage 13 Connection Interface 14 Communication Interface 2A display device 2B Input device 3 Products 30 Prediction Target

Claims

1. a first evaluation unit that evaluates a failure rate representing a probability of a random failure of the prediction target by using a statistical model generated based on data regarding the failure of the prediction target; a second evaluation unit that evaluates a failure probability representing a probability of failure due to aging of the prediction object by using a physical model generated based on design data of the prediction object; An acquisition unit that acquires a qualitative evaluation result obtained by qualitatively evaluating the prediction target for each predetermined evaluation item; a correction unit that calculates a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by a correction coefficient based on the qualitative evaluation result, respectively; A failure prediction device comprising:

2. a first determination unit that determines whether a spare part condition of the prediction object satisfies a constraint condition preset for the prediction object based on the corrected failure rate; a second determination unit that determines whether a preventive maintenance plan for the prediction object satisfies a constraint condition set in advance for the prediction object based on the corrected failure probability; a suggestion unit that suggests reviewing a countermeasure against a failure corresponding to at least one of the evaluation items when at least one of the first judgment unit and the second judgment unit judges that the constraint condition is not satisfied; and The failure prediction device according to claim 1 , further comprising:

3. a first adjustment unit that adjusts the spare part conditions based on the corrected failure rate; The failure prediction device according to claim 2 .

4. A second adjustment unit adjusts the preventive maintenance plan based on the corrected failure probability. The failure prediction device according to claim 2 or 3.

5. a third adjustment unit that adjusts spare part conditions and a preventive maintenance plan for the entire product based on the corrected failure rate, the corrected failure probability, the spare part conditions, and the preventive maintenance plan for each of the plurality of prediction objects so as to satisfy a constraint condition that is preset for the entire product consisting of the plurality of prediction objects; The failure prediction device according to claim 2 .

6. A step in which a failure prediction device evaluates a failure rate representing a probability of a random failure of the predicted object using a statistical model generated based on data regarding the failure of the predicted object; a step in which the failure prediction device evaluates a failure probability representing a probability of failure due to aging deterioration of the prediction target by using a physical model generated based on design data of the prediction target; The failure prediction device acquires a qualitative evaluation result obtained by qualitatively evaluating the prediction target for each predetermined evaluation item; a step in which the failure prediction device calculates a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by a correction coefficient based on a result of the qualitative evaluation; The failure prediction method includes:

7. A step of estimating a failure rate representing a probability of a random failure of the predicted object using a statistical model generated based on data regarding the predicted failure; A step of evaluating a failure probability representing a probability of failure due to aging of the prediction object using a physical model generated based on design data of the prediction object; A step of acquiring a qualitative evaluation result obtained by qualitatively evaluating the prediction target for each predetermined evaluation item; calculating a corrected failure rate and a corrected failure probability by multiplying the failure rate and the failure probability by a correction coefficient based on the qualitative evaluation result, respectively; A program for causing a computer of a failure prediction device to execute the above.

Images