Failure probability evaluation apparatus
Patent Information
- Application Number
- JP2023148400
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-09-13
- Publication Date
- 2026-02-12
AI Technical Summary
【0011】 本発明によれば、部品の故障確率関数の同定精度を向上して、部品の故障確率の評価精度を向上することができる。
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a failure probability evaluation device that targets parts used in a plurality of machines and evaluates the failure probability of the parts for each machine. [Background technology]
[0002] In power generation, transportation, or other industrial machinery, in order to perform the desired function, it is important to understand the failure risk of each part and perform maintenance (more specifically, repair, replacement, etc.) of each part at an appropriate time. Patent Document 1 discloses a failure probability evaluation system that targets parts used in multiple machines and evaluates the failure probability of each part for each machine.
[0003] The failure probability evaluation system in Patent Document 1 includes a failure history database that stores failure history data of parts in a plurality of machines, an operation database that stores operation data acquired in time series by sensors of the plurality of machines, and a calculation unit that identifies a failure probability function of the parts using the failure history database and the operation database, and calculates the failure probability of the parts for each machine using the identified failure probability function.
[0004] The calculation unit obtains the failure time of the part (more specifically, the operating time from the initial or previous machine failure to the current failure) or survival time (the operating time from the initial or previous machine failure to the present) from the failure history data of the part.Then, by substituting the operation data obtained in the period corresponding to the failure time or survival time of the part into a damage model using the operation data as a parameter, the calculation unit calculates the cumulative damage of the part and identifies a failure probability function using the cumulative damage as an explanatory variable.
[0005] In the failure probability assessment system of Patent Document 1, by using the cumulative damage of parts obtained from operational data, it is possible to improve the accuracy of the assessment of the failure probability of parts by taking into account the load on parts, which differs from machine to machine. [Prior art documents] [Patent documents]
[0006] [Patent Document 1] JP 2019-160128 A Summary of the Invention [Problem to be solved by the invention]
[0007] However, in a machine, a sensor that was not initially installed may be added for some reason. In this case, the operation data acquired by the additional sensor exists after the sensor is installed, but does not exist before the sensor is installed.
[0008] For example, if a part failure occurs before the installation of a sensor, and the part is maintained thereafter and no failure occurs to date, the time from the beginning of the machine to the time of the part failure is obtained as the part failure time, and the time from the time of the part failure to the present is obtained as the part survival time. The operation data acquired by the additional sensor does not correspond to the above-mentioned part failure time, nor does it correspond to a part of the above-mentioned part survival time. Therefore, the operation data acquired by the additional sensor cannot be utilized. Therefore, there is room for improvement in terms of the identification accuracy of the failure probability function.
[0009] The present invention has been made in consideration of the above-mentioned circumstances, and an object of the present invention is to provide a failure probability evaluation device that can improve the identification accuracy of a failure probability function of a part and thereby improve the evaluation accuracy of the failure probability of the part. [Means for solving the problem]
[0010] In order to achieve the above object, the present invention provides a failure probability evaluation device that targets parts used in a plurality of machines and evaluates the failure probability of the parts for each of the machines, the failure probability evaluation device comprising: a maintenance history database that stores maintenance history data of the parts in the plurality of machines; an operation database that stores operation data acquired in a chronological order by a sensor of the plurality of machines; and a calculation device that identifies a failure probability function of the parts using the maintenance history database and the operation database, and calculates the failure probability of the parts for each of the machines using the identified failure probability function, the calculation device uses a damage model having the operation data as a parameter to calculate a progression of damage to the parts after installation of the sensor, learns the progression of damage to the parts after installation of the sensor and estimates the progression of damage to the parts before installation of the sensor, calculates accumulated damage to the parts based on at least one of the progression of damage to the parts after installation of the sensor and the progression of damage to the parts before installation of the sensor, and identifies the failure probability function using the accumulated damage. Effect of the Invention
[0011] According to the present invention, it is possible to improve the accuracy of identifying the failure probability function of a part, and thereby improve the accuracy of evaluating the failure probability of the part.
[0012] Problems, configurations and effects other than those described above will become apparent from the following description. [Brief description of the drawings]
[0013] [Figure 1] 1 is a block diagram showing a configuration of a failure probability evaluation device in one embodiment of the present invention. [Diagram 2] 11 is a diagram showing a specific example of maintenance history data in one embodiment of the present invention. FIG. [Diagram 3] 1 is a flowchart showing a procedure for identifying a failure probability function in one embodiment of the present invention. [Figure 4] 1A to 1C are diagrams showing specific examples of operational data in one embodiment of the present invention and specific examples of the progression of damage to parts. [Diagram 5] FIG. 13 is a diagram illustrating a specific example of the variation of the failure probability function in one embodiment of the present invention. [Figure 6] FIG. 11 is a diagram showing a specific example of a display form of failure probability in one embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0014] An embodiment of the present invention will be described with reference to the drawings.
[0015] FIG. 1 is a block diagram showing the configuration of a failure probability evaluation device in this embodiment.
[0016] The failure probability evaluation device 100 of this embodiment targets parts used in multiple machines 1 (wind power generators in this embodiment), and evaluates the failure probability of the parts for each machine 1. The failure probability evaluation device 100 includes a maintenance history database 11, an operation database 12, a calculation device 13, an input device 14, and a communication device 15.
[0017] The maintenance history database 11 and the operation database 12 are composed of storage devices such as hard disks. The calculation device 13 has a processor that executes processing based on a program, and a memory that temporarily stores intermediate or final results of the processing. The input device 14 is composed of an input / output interface such as a keyboard and a display. The communication device 15 is composed of a communication interface that connects to a communication network with multiple machines 1 (more specifically, a satellite communication network, the Internet, an intranet, etc.).
[0018] The maintenance history database 11 stores maintenance history data of parts in the multiple machines 1 that is input via the input device 14. If the machine 1 has a function of detecting a failure in a part and transmitting information relating to the failure as maintenance history data, the maintenance history database 11 may store the maintenance history data of parts in the multiple machines 1 that is received by the communication device 15.
[0019] As shown in FIG. 2, the maintenance history data has the following data items: date and time of maintenance, the installation site (site name) of the machine 1, the identification number (machine number) of the machine 1, the part to be maintained (part name), the reason for maintenance (event), and the maintenance content, and is composed of records that are combinations of this information. The maintenance content of "reactive maintenance" means that maintenance was performed after a part failure or abnormality occurred, and the maintenance content of "proactive maintenance" means that maintenance was performed even if no part failure or abnormality occurred. The maintenance history data includes not only records with "reactive maintenance" as the maintenance content (in other words, failure history data) but also records with "proactive maintenance" as the maintenance content, and by utilizing them, it is possible to improve the accuracy of identifying the failure probability function described later.
[0020] The machines 1 acquire, for example, temperature, wind speed, and power generation amount in time series using sensors and transmit them as operation data. The operation database 12 stores the operation data of the machines 1 received by the communication device 15.
[0021] The operational data may be composed of sensor measurement values, but may also be composed of statistical values (such as maximum values, minimum values, average values, or standard deviations) for each predetermined time period (such as one day) in order to reduce the amount of data. Alternatively, the operational data may be composed of calculated values calculated based on the sensor measurement values so that they can be easily used as parameters for a damage model, which will be described later.
[0022] The calculation device 13 has a function of identifying a failure probability function of a part by using the above-mentioned maintenance history database 11 and operation database 12. The calculation device 13 has, as configurations related to the above-mentioned functions, a maintenance time / survival time calculation unit 16, a damage model identification unit 17, a damage calculation / estimation unit 18, a cumulative damage calculation unit 19, and a failure probability function identification unit 20. The calculation device 13 also has a function of calculating the failure probability of a part for each machine 1 by using the identified failure probability function. The calculation device 13 has, as configurations related to the above-mentioned functions, a failure probability calculation unit 21 and a cumulative damage prediction unit 22.
[0023] First, the function of identifying the failure probability function will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the procedure of a process for identifying the failure probability function in this embodiment.
[0024] In step S1, the maintenance time / survival time calculation unit 16 sets the target part. Then, from among the maintenance history data stored in the maintenance history database 11, records including the set part name are extracted, and the extracted records are classified by machine 1 (specifically, by combination of site name and machine number). Then, based on the maintenance implementation date and time included in the records classified by machine 1, the corrective maintenance time of the part (specifically, the operation time from the initial or previous maintenance of the machine 1 to the current corrective maintenance), proactive maintenance time (specifically, the operation time from the initial or previous maintenance of the machine 1 to the current proactive maintenance), or survival time (the operation time from the previous maintenance to the present) are calculated. If no record (in other words, maintenance history) exists for any of the machines 1, the operation time from the initial to the present of the machine 1 is calculated as the survival time of the part. In the following description, the corrective maintenance time is referred to as the maintenance time, and the survival time is assumed to include the proactive maintenance time.
[0025] In step S2, the damage model identification unit 17 sets the coefficients of the damage model d(Xt) for the set part. The damage model d(Xt) calculates damage per unit time using the operation data Xt at time t as a parameter. If it is assumed that the operation data Xt at time t is composed of values x1, x2, ..., xm at time t, the damage model d(Xt) may be expressed by an equation that linearly combines the values x1, x2, ..., xm, as shown in the following equation (1). In this case, the damage model identification unit 17 sets the coefficients c1, c2, ..., cm. Note that the damage model d(Xt) is not limited to equation (1) and may be expressed by other equations (in detail, if the operation data includes temperature, for example, an equation incorporating the Arrhenius equation).
[0026]
number
[0027] Proceeding to step S3, the damage calculation / estimation unit 18 calculates the damage transition of the parts of each machine 1 by substituting the operation data of each machine 1 stored in the operation database 12 into the above-mentioned damage model. Here, it is assumed that a sensor that was not initially installed is added to any of the machines 1 for some reason. In this case, as shown in FIG. 4(a), the values x1, x2, and x3 acquired by the additional sensor exist after the installation time t2 of the sensor, but do not exist before that. Therefore, as shown in FIG. 4(b), the damage transition A of the parts obtained by substituting the operation data including the values x1, x2, and x3 into the damage model exists after the installation time t2 of the sensor, but does not exist before that.
[0028] Proceeding to step S4, the damage calculation / estimation unit 18 learns the transition A of the damage to the component after the sensor installation time t2 by regression analysis, time series analysis, or the like, and estimates the transition B of the damage to the component before the sensor installation time t2 (see FIG. 4(b)).
[0029] Proceeding to step S5, the cumulative damage calculation unit 19 calculates the cumulative damage of the part corresponding to the maintenance time or survival time of the part based on at least one of the transition of the damage of the part after the installation of the sensor and the transition of the damage of the part before the installation of the sensor. A specific description will be given with reference to Fig. 4(b).
[0030] If the corrective maintenance of the part is performed before the installation of the sensor, the time from the initial time t0 of the machine 1 to the time t1 of the corrective maintenance of the part is obtained as the maintenance time of the part, and the time from the time t1 of the corrective maintenance of the part to the present time t3 is obtained as the survival time of the part. The cumulative damage calculation unit 19 calculates the cumulative damage D01 from the initial time t0 of the machine 1 to the time t1 of the corrective maintenance of the part based on the transition B of the damage of the part before the installation time t2 of the sensor, and sets it as the cumulative damage corresponding to the maintenance time of the part described above. In addition, the cumulative damage calculation unit 19 calculates the cumulative damage D12 from the time t1 of the corrective maintenance of the part to the installation time t2 of the sensor based on the transition B of the damage of the part before the installation time t2 of the sensor, and calculates the cumulative damage D23 from the installation time t2 of the sensor to the present time t3 based on the transition A of the damage of the part after the installation time t2 of the sensor, and sets the sum of the cumulative damage D12 and the cumulative damage D23 as the cumulative damage corresponding to the survival time of the part described above.
[0031] Proceeding to step S6, the failure probability function identification unit 20 identifies a failure probability function F(D) with the cumulative damage D as a parameter by using the cumulative damage corresponding to the maintenance time of the part and the cumulative damage corresponding to the survival time of the part, using a known maximum likelihood estimation method or Bayesian estimation method. The maximum likelihood estimation method searches for a parameter of the failure probability function so as to maximize the log-likelihood sum L defined by the following equation (2). In the equation, f is a failure probability density function obtained by differentiating the failure probability function F(D). The first term on the right side of the equation represents the likelihood of the cumulative damage corresponding to the maintenance time of the part, and the second term represents the likelihood of the cumulative damage corresponding to the survival time of the part.
[0032]
number
[0033] Proceeding to step S7, the failure probability function identification unit 20 judges whether the variation of the failure probability density function f has been minimized based on whether the variation of the failure probability density function f is equal to or less than a predetermined value. If the variation of the failure probability density function f has not been minimized, the process returns to step S2. That is, the failure probability function identification unit 20 outputs a command to change the coefficient of the damage model d(Xt) to the damage model identification unit 17. The damage model identification unit 17 changes the coefficient of the damage model d(Xt) in response to the command.
[0034] After that, the above-mentioned steps S3 to S6 are performed, and the process proceeds to step S7. In step S7, the failure probability function identification unit 20 judges whether the variation of the failure probability density function f has been minimized depending on whether the variation of the failure probability density function f is equal to or less than a predetermined value. This is because the fact that the variation of the failure probability density function f is large (as shown in FIG. 5, the variation of the failure probability function is large) means that the failure occurrence prediction interval has a width, and it is necessary to reduce the variation in order to minimize the prediction interval width and accurately estimate the next failure occurrence date and time. The variation of the failure probability function can be evaluated by the coefficient of variation (the ratio of the standard deviation and the average value of the failure probability function) (see Patent Document 1). Furthermore, even if the variation of the failure probability density function f is equal to or more than a predetermined value, it is judged whether the variation of the failure probability density function f has been minimized depending on whether the rate of change of the variation is equal to or less than a predetermined value. When the variation in the failure probability density function f is minimized, the identification (setting of coefficients) of the damage model d(Xt) by the damage model identification unit 17 and the identification of the failure probability function F(D) by the failure probability function identification unit 20 are completed.
[0035] In this embodiment, when the damage model d(Xt) is defined, the cumulative damage (corresponding to the cumulative damage D01 and cumulative damage D12 in FIG. 4(b)) estimated by the damage calculation / estimation unit 18 through extrapolation also changes accordingly. Generally, extrapolation has lower estimation accuracy than interpolation, but in the present invention, the extrapolation result is used to identify the failure probability function, and the variation in the failure probability density function f is reduced to bring it closer to a cumulative damage model based on the true mechanism that leads the target to failure, and a correction function for the extrapolation result is activated, making it possible to ensure the estimation accuracy.
[0036] Next, the function of calculating the component failure probability for each machine 1 using the identified failure probability function F(D) will be described in detail.
[0037] The failure probability calculation unit 21 sets the target machine 1 and parts. Then, when calculating the current failure probability, it outputs a command to the cumulative damage calculation unit 19. The cumulative damage calculation unit 19 calculates the cumulative damage Da up to the present for the set machine 1 and parts in response to the command. In detail, if there is no maintenance history for the parts, it calculates the cumulative damage Da from the beginning of the machine 1 to the present, and if there is a maintenance history for the parts, it calculates the cumulative damage Da from the time of the last maintenance to the present. The failure probability calculation unit 21 calculates the current failure probability Pa by substituting the cumulative damage Da up to the present calculated by the cumulative damage calculation unit 19 into the failure probability function F(D) identified by the failure probability function identification unit 20.
[0038] The failure probability calculation unit 21 outputs a command to the cumulative damage calculation unit 19 and the cumulative damage prediction unit 22 when calculating the failure probability in the future (for example, after the time period Δt set by the user has elapsed). The cumulative damage calculation unit 19 calculates the cumulative damage Da up to the present for the set machine 1 and parts in response to the command. The cumulative damage prediction unit 22 predicts the cumulative damage Db for the time period Δt for the set machine 1 and parts in response to the command. In detail, for example, the operation data stored in the operation database 12 is learned by regression analysis or time series analysis, etc., to predict the operation data for the time period Δt. Then, the operation data for the time period Δt is substituted for the damage model identified by the damage model identification unit 17, thereby predicting the progress of damage to the parts in the time period Δt. Then, the cumulative damage Db for the time period Δt is predicted based on the progress of damage to the parts in the time period Δt. The failure probability calculation unit 21 calculates the future failure probability Pb using the failure probability function F(D) identified by the failure probability function identification unit 20, the cumulative damage Da up to the present calculated by the cumulative damage calculation unit 19, and the cumulative damage Db for the period Δt predicted by the cumulative damage prediction unit 22 (see equation (3) below).
[0039]
number
[0040] As described above, the failure probability evaluation device 100 of this embodiment can identify the failure probability function of a part by utilizing operation data acquired by a sensor added to the machine 1. Therefore, it is possible to improve the accuracy of identifying the failure probability function of a part and improve the evaluation accuracy of the failure probability of the part.
[0041] The failure probability evaluation device 100 of this embodiment outputs data including the failure probability of a part calculated by the calculation device 13 to, for example, a user interface 23, an operation planning system 24, and a part inventory management system 25. The user interface 23 is owned, for example, by the owner of the machine 1, an operating company, or an insurance company.
[0042] The user interface 23 is, for example, a mobile terminal, and operates in cooperation with the computing device 13. The user interface 23 displays, for example, the screen shown in FIG. 6. This screen has a part setting unit 31, a period setting unit 32, and a failure probability display unit 33. The part setting unit 31 shows a schematic representation of the configuration of the machine 1, and allows the user to set a target part (e.g., a gearbox). The period setting unit 32 allows the user to set a period Δt from the present. The failure probability display unit 33 displays the failure probability after the period Δt set by the period setting unit 32 has elapsed, for the part set by the part setting unit 31.
[0043] The operation planning system 24 can change the operation plan of the machine 1 according to the failure probability of the parts. For example, if the failure probability of the parts at the next regular inspection is higher than expected, the machine 1 is actively stopped or the output of the machine 1 is suppressed in order to extend the life of the parts. When the operation planning system 24 changes the operation plan, it outputs the information to the failure probability evaluation device 100. The cumulative damage prediction unit 22 of the calculation device 13 changes the prediction of the cumulative damage of the parts based on the above-mentioned information. Accordingly, the failure probability calculation unit 21 of the calculation device 13 changes and outputs the future failure probability.
[0044] In the above embodiment, the failure probability evaluation device 100 has been described as including one calculation device 13, but the present invention is not limited to this and may include a plurality of calculation devices. That is, the maintenance time / survival time calculation unit 16, the damage model identification unit 17, the damage calculation / estimation unit 18, the cumulative damage calculation unit 19, the failure probability function identification unit 20, the failure probability calculation unit 21, and the cumulative damage prediction unit 22 may be configured with a plurality of calculation devices.
[0045] Furthermore, in the above embodiment, the machine 1 has been described as a wind power generator, but it goes without saying that the machine 1 is not limited to this. [Explanation of symbols]
[0046] 1 machine 11 Maintenance History Database 12. Working Database 13 Arithmetic unit 100 Failure Probability Evaluation Device
Claims
1. 1. A failure probability evaluation device for evaluating the failure probability of a part used in a plurality of machines, for each of the machines, a maintenance history database storing maintenance history data for the parts in the plurality of machines; an operation database that stores operation data acquired in time series by sensors of the plurality of machines; a calculation device that identifies a failure probability function of the part using the maintenance history database and the operation database, and calculates a failure probability of the part for each machine using the identified failure probability function, The computing device calculating a progression of damage to the component after the sensor is installed using a damage model that uses the operational data as a parameter; learning a change in damage to the component after the sensor is installed, and estimating a change in damage to the component before the sensor is installed; calculating cumulative damage to the component based on at least one of a change in damage to the component after the installation of the sensor and a change in damage to the component before the installation of the sensor; A failure probability evaluation device, characterized in that the failure probability function is identified using the cumulative damage.
2. 2. The failure probability evaluation device according to claim 1, The failure probability evaluation device is characterized in that the calculation device identifies the damage model.
3. 3. The failure probability evaluation device according to claim 2, The failure probability evaluation device is characterized in that the calculation device identifies the damage model so as to minimize the variation in a failure probability density function obtained by differentiating the failure probability function.
4. 2. The failure probability evaluation device according to claim 1, It has a user interface that allows you to set the period from the present, The failure probability evaluation device is characterized in that the calculation device calculates the failure probability of the part after the period has elapsed using the identified failure probability function and displays the calculated failure probability on the user interface.
5. 5. The failure probability evaluation device according to claim 4, The computing device Calculating cumulative damage to the part from the time of initial maintenance of the machine or the part to the present; learning the operation data to predict operation data for the period, and predicting cumulative damage to the part for the period using the damage model and the operation data for the period; A failure probability evaluation device characterized by calculating the failure probability of the part after the period has elapsed using the identified failure probability function, the cumulative damage of the part from the initial stage of the machine or the time of maintenance of the part to the present, and the cumulative damage of the part during the period.
6. 6. The failure probability evaluation device according to claim 5, A failure probability evaluation device characterized in that the calculation device changes the prediction of cumulative damage to the part for the period based on information from an operation planning system that changes the operation plan of the machine.