Inspection assistance method and inspection assistance device

The inspection support device optimizes inspection intervals for social infrastructure equipment by calculating flaw detection probabilities and using Bayesian estimation to determine when the risk of flaw detection exceeds allowable limits, addressing the inefficiencies in current non-destructive inspection technologies.

WO2024237025A1PCT designated stage expired Publication Date: 2024-11-21HITACHI LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2024/015755
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-12
Filing Date
2024-04-22
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Current non-destructive inspection technologies for social infrastructure equipment, such as power plants and railway bogies, face challenges in determining optimal inspection intervals due to the rarity of flaw detection, leading to inefficient economic and risk management.

Method used

An inspection support device calculates the flaw detection probability and initial distribution of flaw size, determining inspection timing when the overlap between the flaw size probability distribution and failure probability exceeds a preset allowable risk, optimizing inspection intervals through Bayesian estimation and Fokker-Planck equation calculations.

Benefits of technology

This approach allows for the rational determination of inspection intervals, even for equipment where flaws are rarely detected, ensuring economic efficiency and effective risk management by utilizing the probability of non-detection and flaw growth analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2024015755_21112024_PF_FP_ABST
    Figure JP2024015755_21112024_PF_FP_ABST
Patent Text Reader

Abstract

In order to optimize the timing of an inspection, this inspection assistance device is characterized by executing: a step (S2) for calculating a flaw detection probability; a step (S3) for calculating a flaw size probability distribution with the lapse of time by using inspection variation input via a data input unit and the initial distribution of the flaw size probability distribution; a step (S4) for determining the timing of the inspection if the area of a region where the flaw size probability distribution with the lapse of time and a destruction probability overlap with each other exceeds a preset allowable risk; and a step (S7) for outputting the determined timing of the inspection.
Need to check novelty before this filing date? Find Prior Art

Description

Inspection support method and inspection support device

[0001] The present invention relates to a technique for an inspection support method and an inspection support device.

[0002] In order to maintain and manage social infrastructure equipment such as power plants and railway bogies, there is a need for greater efficiency through the advancement of non-destructive testing technology. The integrity of such social infrastructure equipment is guaranteed through periodic inspections carried out at set intervals. In general, non-destructive testing is required to be carried out at intervals that take into maximum consideration the risk to each piece of equipment. However, in power plants and other facilities, defects are rarely detected, and excessive inspections result in poor economic efficiency. Therefore, there is a need to determine the timing of inspections using rational indicators.

[0003] In response to such a problem, for example, Patent Document 1 discloses a method for analyzing a crack propagation, which includes: "an input unit that receives information regarding the number of times a crack propagation analysis is repeatedly executed, and also receives information regarding the interval between inspection periods for non-destructive testing within an evaluation period for the crack propagation analysis, the cost of non-destructive testing, the cost of repair, the cost of responding to an accident, and a formula for calculating a crack detection probability; a stress calculation unit that calculates stress near the tip of a crack that occurs in a structure that is irradiated with neutrons; a stress intensity factor calculation unit that calculates a stress intensity factor based on the stress calculated by the stress calculation unit and crack propagation information regarding cracks that occur in the structure and the crack propagation rate; a fracture toughness value derivation unit that calculates a fracture toughness value based on structure information including the Poisson's ratio and longitudinal elastic modulus of the structure and information including the neutron irradiation dose of the structure; and a method for calculating a fracture toughness value based on the crack propagation information and the crack propagation information. a crack growth amount calculation unit that calculates a crack growth amount based on an evaluation period for crack growth analysis and time information related to a time increment during the evaluation period; a fracture determination unit that determines the presence or absence of fracture by comparing the stress intensity factor calculated by the stress intensity factor calculation unit with the fracture toughness value calculated by the fracture toughness value derivation unit; an evaluation period determination unit that determines whether the sum of the time information and the time increments due to the number of repetitions of the crack growth analysis reaches the evaluation period; a crack detection determination unit that determines whether a crack is detected when the sum of the time increments reaches a time corresponding to the interval of the inspection period; and a repetition determination unit that determines whether the number of times the presence or absence of a crack determined by the crack detection determination unit during the period until the non-destructive inspection and during the period from the end of the last non-destructive inspection to the end of the evaluation period reaches the number of repetitions of the crack growth analysis. The document describes a destruction evaluation analysis device, a destruction evaluation analysis system, and a destruction evaluation analysis method, each of which is characterized by comprising: a destruction probability calculation unit that calculates the destruction probability of the structure from the ratio of the number of times destruction is determined by the destruction determination unit to the number of times the crack propagation analysis is repeated; and a cost calculation unit that calculates repair costs according to the ratio of cracks detected by the crack detection determination unit, inspection costs according to the number of non-destructive inspections and the inspection range, and accident response costs according to the destruction probability (see claim 1).

[0004] Patent No. 6746512

[0005] In the technology described in Patent Document 1, accident response costs are calculated using probabilistic destructive analysis, with the probability of flaw detection by non-destructive testing and the inspection interval as input. However, although the technology described in Patent Document 1 makes it possible to evaluate the inspection interval taking economic rationality into consideration, the inspection interval is input by the user and is not determined by the system.

[0006] Generally, to determine inspection intervals, it is necessary to predict the time when a defect may become a serious risk by predicting its progression using the initial distribution of defects. However, because defects are rarely detected, it is difficult to calculate risk using information on defect detection.

[0007] The present invention has been made in view of the above background, and an object of the present invention is to optimize the inspection interval.

[0008] In order to solve the above-mentioned problems, the present invention is characterized in that an inspection support device executes the following steps: a first step of calculating a flaw detection probability; a second step of calculating a flaw size probability distribution over time using inspection variability and an initial flaw size probability distribution input via an input unit; a third step of determining an inspection time if the area of ​​the region where the flaw size probability distribution over time and the fracture probability overlap exceeds a predetermined allowable risk; and a fourth step of outputting the determined inspection time. Other solutions will be described as appropriate in the embodiments.

[0009] According to the present invention, the inspection interval can be optimized.

[0010] FIG. 1 is a functional block diagram showing an example of the configuration of an inspection support device according to a first embodiment. FIG. 2 is a hardware configuration diagram of the inspection support device according to the first embodiment. FIG. 3 is a flowchart showing the procedure of an inspection support method according to the first embodiment. FIG. 4 is a conceptual diagram showing an example of a model for calculating a flaw detection probability in the first embodiment. FIG. 5 is a conceptual diagram showing a flaw detection probability according to the present embodiment. FIG. 6 is a flowchart showing details of calculation of a flaw size probability distribution over time and determination of an inspection interval in the first embodiment. FIG. 7 is a conceptual diagram showing a flaw size probability distribution over time in the first embodiment. FIG. 8 is a conceptual diagram showing an example of Bayesian estimation of a flaw size probability distribution after inspection according to the first embodiment. FIG. 9 is a conceptual diagram showing an example of Bayesian estimation of a flaw size probability distribution after inspection according to the first embodiment. FIG. 10 is a schematic diagram showing the procedure of determining an inspection interval. FIG. 11 is a conceptual diagram showing a flaw size probability distribution over time in the second embodiment. FIG. 12 is a flowchart showing the procedure of an inspection support method according to the second embodiment. FIG. 13 is a flowchart showing details of calculation of a flaw size probability distribution over time and determination of an inspection interval in the second embodiment. FIG. 14 is a flowchart showing a detailed calculation procedure for calculating a flaw size probability distribution over time in the third embodiment. FIG. 15 is a diagram showing an example of an output screen.

[0011] Next, a mode for carrying out the present invention (referred to as an "embodiment") will be described in detail with reference to the drawings as appropriate. Note that the same components in each drawing are designated by the same reference numerals, and detailed descriptions of overlapping parts will be omitted.

[0012] First Embodiment An inspection support method according to a first embodiment of the present invention will be described with reference to FIGS.

[0013] <Inspection Support Device 1> FIG. 1 is a functional block diagram showing an example configuration of an inspection support device 1 according to a first embodiment. FIG. 1 shows an overview of the processing performed by the data input unit 101 to the Bayesian estimation unit 105. Details of the processing performed by the data input unit 101 to the Bayesian estimation unit 105 will be described later. The data input unit 101, which is an input unit, is configured with an interface that accepts input of inspection variability, material variability, initial flaw size probability distribution, and tolerable risk. The input inspection variability and material variability are sent to the detection probability calculation unit 102. The detection probability calculation unit 102 calculates the flaw detection probability (curve 301 in FIG. 5 ) and flaw non-detection probability (curve 302 in FIG. 5 ) by non-destructive inspection. Hereinafter, the flaw detection probability will be referred to as the detection probability, and the flaw non-detection probability will be referred to as the non-detection probability. The detection probability calculation unit 102 may calculate only the detection probability. The detection probability of non-destructive testing calculated by the detection probability calculation unit 102 (curve 301 in FIG. 5 ) is sent to the Bayesian estimation unit 105. The detection probability is the probability that a flaw will be detected by the test, and the non-detection probability is the probability that a flaw will not be detected by the test.

[0014] Furthermore, the material variation and initial flaw size probability distribution input to the data input unit 101 are sent to the progression analysis unit 103. The progression analysis unit 103 then calculates the flaw size probability distribution over time (curve 312 in FIG. 7). The flaw size probability distribution over time calculated by the progression analysis unit 103 (curve 312 in FIG. 7) is sent to the inspection interval determination unit 104. The flaw size probability distribution is a distribution of the probability related to the size of flaws that will occur.

[0015] The inspection interval determination unit 104, which is an inspection timing determination unit, an output processing unit, and an update unit, determines the inspection interval, which is the inspection timing. The inspection interval is the interval between one inspection and the next inspection. If the inspection being performed is the first inspection since the start of inspection, the inspection interval is the interval between the start of inspection and the first inspection. At this time, the inspection interval determination unit 104 calculates an estimated failure probability (area 304 in FIG. 7 ) based on the detection probability of the flaw 203 (curve 301 in FIG. 5 ) and the flaw size probability distribution over time (curve 312 in FIG. 7 ). The inspection interval determination unit 104 then compares the estimated failure probability (area 304 in FIG. 7 ) with the tolerable risk input via the data input unit 101 to determine the inspection interval. The estimated failure probability is the probability that the object currently being inspected will be destroyed due to a flaw.

[0016] The Bayesian estimation unit 105 performs Bayesian estimation of the flaw size probability distribution after inspection (curve 321 in FIG. 8B ) based on the non-detection probability (curve 302 in FIG. 5 ) and the flaw size probability distribution over time (curve 312 in FIG. 7 ). The flaw size probability distribution after inspection (curve 321) is sent to the progression analysis unit 103.

[0017] The above process is then repeated to repeatedly determine the inspection intervals until the sum of the inspection intervals exceeds the operation period of the inspection object 202 (see FIG. 4). The operation period is, for example, the period from the inspection object 202 to the completion of the warranty of the inspection object 202.

[0018] 2 is a hardware configuration diagram of the test support device 100 according to the first embodiment. The test support device 100 has a memory 111 configured with a RAM or the like, and an arithmetic device 112 configured with a CPU, a GPU or the like. The test support device 100 also has a storage device 113 configured with a HD, an SSD or the like. Furthermore, the test support device 100 has an input device 114 such as a keyboard or a mouse, and an output device 115 such as a display.

[0019] Then, the program stored in the storage device 113 is loaded into the memory 111, and the loaded program is executed by the arithmetic unit 112. This embodies the detection probability calculation unit 102 to the Bayesian estimation unit 105 shown in Fig. 1. Incidentally, the data input unit 101 in Fig. 1 corresponds to the input device 114 in Fig. 2.

[0020] <Flowchart> Fig. 3 is a flowchart showing the steps of the inspection assistance method according to the first embodiment. The inspection assistance method according to this embodiment will be described with reference to Fig. 1 and Fig. 3. As shown in Fig. 3, first, a user inputs inspection variability, initial flaw size probability distribution, and tolerable risk via data input unit 101 (S1). These data are stored in memory 111 (see Fig. 2) or on a hard disk, and are used in subsequent steps. Details of inspection variability and initial flaw size probability distribution will be described later.

[0021] Next, the detection probability calculation unit 102 calculates the detection probability of the target flaw 203 (S2: first step).

[0022] Step S2 will be described with reference to Figures 4 and 5. Figure 4 is a conceptual diagram showing an example of a model for calculating the probability of flaw detection in the first embodiment. Figure 4 shows the case of ultrasonic flaw detection for crack-like defects as a specific example of flaw inspection. Other non-destructive inspection methods, such as eddy current flaw detection and radiographic testing, can also be modeled using a similar technique.

[0023] In the example shown in Fig. 4, an ultrasonic probe 201 is placed on an inspection object 202, and ultrasonic waves 204 are transmitted toward a flaw 203. The ultrasonic waves 204 are reflected by the flaw 203, and the reflected ultrasonic waves 204 are received by the ultrasonic probe 201. At this time, whether or not the flaw 203 is detected depends on the size of the flaw 203 and inspection variations. As shown in the example of Fig. 4, inspection variations include variations 205 in the installation position of the ultrasonic probe 201 and variations 206 in the ultrasonic transmission direction. Further, examples of inspection variations include variations in material constants such as the speed of sound inside the inspection object 202.

[0024] 5 is a conceptual diagram showing the flaw detection probability according to this embodiment. In the graph shown in FIG. 5, the horizontal axis represents the size of the flaw 203 (see FIG. 4), and the vertical axis represents the probability. The detection probability is represented by the curve shown in curve 301 for flaw size "a." The detection probability is expressed as "POD(a)." POD stands for Probability of Detection. The non-detection probability is represented by the curve shown in curve 302, and is expressed as "1-POD(a)."

[0025] The detection probability curve 301 can be obtained using a physical simulation of nondestructive testing or experimental results. For example, in the physical simulation illustrated in FIG. 4 , the detection probability calculation unit 102 randomly samples one value for each of the size of the flaw 203 and the inspection variability. The detection probability calculation unit 102 then performs a physical simulation under those conditions. The inspection variability includes, for example, the variability 205 in the installation position of the ultrasonic probe 201 shown in FIG. 4 , the variability 206 in the ultrasonic transmission direction, and the variability of material constants such as the speed of sound of the inspection target 202. The detection probability calculation unit 102 performs random sampling and simulation multiple times and calculates the detection probability from the calculated detection signal strength. A maximum likelihood estimation method, such as a hit / miss method, is known as a method for estimating the detection probability curve 301 from the detection strength of the ultrasonic waves 204.

[0026] Returning to the explanation of Figure 3, after step S2, the propagation analysis unit 103 calculates the flaw size probability distribution over time (S3: second step). Specifically, in step S3, the propagation analysis unit 103 calculates the flaw size probability distribution over time using inspection variability and the initial flaw size probability distribution (initial flaw size probability distribution). Step S3 will be described using a fatigue crack as an example of the flaw 203. It is generally known that in the stable crack growth stage (stage Iib of fatigue crack growth), a crack grows in response to repeated stress according to a law known as the Paris law, expressed in equation (1).

[0027] da / dt=C(ΔK) m / τ (1)

[0028] In equation (1), a is the crack size, t is the elapsed time, τ is the stress repetition interval, C and m are material constants, and ΔK is the stress intensity factor range. ΔK is calculated using a calculation method such as the finite element method (FEM). In this manner, in this embodiment, the time change in the flaw size probability distribution is assumed to be the time change in the flaw size probability distribution for crack-like defects that follow the Paris law. In this manner, the inspection support method of this embodiment can be applied to a general crack as the flaw 203.

[0029] After step S3, the inspection interval determination unit 104 determines the inspection interval (S4: third step). The area of ​​region 304 enclosed by curve 312 (flaw size probability distribution over time) and curve 305 (failure probability) in FIG. 7 represents the estimated failure probability of the inspection object 202 after time t has elapsed. The failure probability indicates the probability that the inspection object 202 (see FIG. 4 ) will be destroyed given a given flaw size. t indicates the elapsed time since the previous inspection. The inspection interval determination unit 104 updates t at predetermined intervals to update the flaw size probability distribution over time indicated by curve 312. Then, the inspection interval determination unit 104 calculates the estimated failure probability indicated by region 304 in FIG. 7 each time the flaw size probability distribution indicated by curve 312 is updated. The failure probability is the probability that the inspection object 202 will be destroyed given the flaw size. The estimated fracture probability shown in region 304 is the area of ​​the region where the fracture probability and the flaw size probability distribution over time overlap. Details of FIG. 7 will be described later.

[0030] 6 is a flowchart showing the details of step S3 (calculating the flaw size probability distribution over time) and step S4 (determining the inspection interval) in FIG. 3. First, the user inputs the calculation time Tmax as an evaluation parameter (S301). Next, the evolution analysis unit 103 solves the Fokker-Planck equation shown below to calculate the flaw size probability distribution "p(a, t)" after the calculation time (t) has elapsed (S302).

[0031] The progression analysis unit 103 repeats this process until t>Tmax. As a result, the progression analysis unit 103 solves the Fokker-Planck equation and calculates the flaw size probability distribution up to the time Tmax. In this embodiment, C or ΔK of the Paris law shown in equation (1) includes noise components of the ultrasonic waves 204 that reflect variations in material constants or variations in repeated stress. Taking this fact into consideration and assuming the Markov property for the noise of the ultrasonic waves 204, the Fokker-Planck equation, which is a stochastic differential equation corresponding to the Paris law, is obtained (equation (2)).

[0032] da=μ(a,t)dt+σ(a,t)dWt...(2)

[0033] In equation (2), μ(a, t) is the mean value of the right-hand side of the Paris law (equation (1)), σ(a, t) is the standard deviation of the right-hand side of the Paris law, and Wt is the standard Wiener process. When the Itō integral is used, the Fokker-Planck equation is expressed as the following equation (3).

[0034]

[0035] When the Stratonovich integral is used, the Fokker-Planck equation is expressed as the following equation (4).

[0036]

[0037] In this embodiment, either integration method may be used.

[0038] Then, the progression analysis unit 103 outputs the flaw size probability distribution after t (calculation time) as a solution to the Fokker-Planck equation (S303).

[0039] In this way, in step S3, the progression analysis unit 103 calculates the flaw size probability distribution over time by solving the Fokker-Planck equation, which is a flaw progression equation in which the flaw size is a continuous random variable. In this way, the flaw propagation can be analytically solved.

[0040] Next, the inspection interval determination unit 104 determines whether the estimated failure probability is greater than the allowable risk (S401). The estimated failure probability will be described later. If the estimated failure probability is equal to or less than the allowable risk (S401 → No), the inspection interval determination unit 104 determines whether the calculation time (t) is greater than Tmax (S402).

[0041] If the calculation time (t) is equal to or less than Tmax (S402→No), the test support device 100 updates the calculation time (t) and returns the process to step S302. If the calculation time (t) is greater than Tmax (S402→Yes), the test interval determination unit 104 outputs an error (S403) and stops the process.

[0042] On the other hand, if the estimated failure probability is greater than the tolerable risk (S401 → Yes), the inspection interval determination unit 104 determines the inspection interval (S404). The calculation time (t) is set to "0" when the processing of FIG. 6 is started. Then, the calculation time (t) is updated each time the determination result in step S402 is "No." The update interval varies depending on the inspection target 202, but is, for example, in units of one day. In step S404, the inspection interval determination unit 104 subtracts 1 from the value of the calculation time (t) when step S404 is executed (t-1=T) and determines the calculation time as the inspection interval. Thereafter, the inspection support device 100 proceeds to step S5 of FIG. 3.

[0043] In the processing of FIG. 6, steps S301 to S303 correspond to the processing of step S3 in FIG. 3, and steps S401 to S404 correspond to the processing of step S4 in FIG.

[0044] Next, the processing of FIG. 6 will be described with reference to FIG. 7 . FIG. 7 is a conceptual diagram showing the flaw size probability distribution over time according to the first embodiment. In the graph shown in FIG. 7 , the horizontal axis represents flaw size, and the vertical axis represents probability. Curve 311 is the initial flaw size probability distribution (prior probability distribution). When step S3 is executed for the first time, the prior probability distribution represented by curve 311 is the initial flaw size probability distribution input in step S1. When step S3 is executed for the second or subsequent time, the prior probability distribution represented by curve 311 is the flaw size probability distribution after inspection, which is Bayesian-estimated in step S5 (described later). Curve 312 is the flaw size probability distribution after the calculation time "T = t - 1" (≦Tmax), which is the inspection interval in step S404 of FIG. 6, has elapsed. Incidentally, curve 312 is a curve obtained by time-integrating the Fokker-Planck equation described above, using the initial flaw size probability distribution represented by curve 311 as the initial value. In other words, curve 312 is the flaw size probability distribution output in step S302 of FIG. 6. Curve 305 is the failure probability of the inspection object 202 for flaw size "a." The failure probability shown by curve 305 is the probability that failure will occur in the inspection object 202 for the size of the flaw 203. Curve 305 is obtained by a calculation method such as FEM. Region 304 will be described later. Incidentally, the failure probability (curve 305) is different from the detection probability (curve 301) shown in FIG. 5.

[0045] The inspection interval determination unit 104 then compares the tolerable risk input in step S1 with the estimated fracture probability after the calculation time "T" has elapsed (step S401 in FIG. 6). This comparison is performed each time the flaw size probability distribution shown by curve 312 is updated. The inspection interval determination unit 104 then determines the calculation time at which the tolerable risk and the estimated fracture probability are equal, or the maximum calculation time at which the fracture probability does not exceed the tolerable risk, as the inspection interval between the nth inspection and the (n-1)th inspection (step S404 in FIG. 6). n is the number of times step S3 is processed.

[0046] In other words, if the estimated failure probability shown in area 304 exceeds the allowable risk entered and set in step S1, the previous calculation time "T" is output as the inspection interval (i.e., the time when the next inspection will be performed).

[0047] Returning to the explanation of Fig. 3, the Bayesian estimation unit 105 then performs Bayesian estimation of the flaw size probability distribution after inspection (S5: fifth step).

[0048] 8A and 8B are conceptual diagrams showing an example of Bayesian estimation of the flaw size probability distribution after inspection according to the first embodiment. The flaw size probability distribution before inspection is p(a|before inspection) = p(a, T). Note that the flaw size probability distribution before inspection is the flaw size probability distribution over time (curve 312) at the calculation time "T = t-1" described above.

[0049] As shown in Fig. 8B, the flaw size probability distribution after inspection (curve 321) when the flaw 203 is not detected during inspection is expressed by the following equation (5) according to Bayes' theorem. In equation (5), the flaw size probability distribution after inspection (curve 321) is calculated based on the flaw 203 non-detection probability (curve 302) and the flaw size probability distribution before inspection (curve 312) shown in Fig. 8A. The flaw 203 non-detection probability (curve 302) is calculated in step S1 of Fig. 3.

[0050] p(a|after test)=A(1-POD(a))p(a|before test) (5)

[0051] In equation (5), POD(a) is the detection probability of the flaw 203 when the flaw size is "a" (curve 301 in FIG. 5). In other words, "1-POD(a)" indicates the non-detection probability of the flaw 203 (curve 302 in FIG. 5). Furthermore, "p(a|before inspection)" indicates the flaw size probability distribution before inspection. The flaw size probability distribution before inspection is the flaw size probability distribution over time immediately before inspection. Furthermore, A is a normalization constant that makes the total probability of "p(a|after inspection)" "1." As a result, the flaw size probability distribution after inspection becomes like curve 321 in FIG. 8B.

[0052] In this way, in step S5, the Bayesian estimation unit 105 estimates the flaw size probability distribution after inspection by Bayesian estimation, using the flaw detection probability (POD(a)) calculated in step S2 and the flaw size probability distribution over time calculated in step S3.

[0053] Returning to the explanation of FIG. 3 , the inspection support device 1 then determines whether the sum of the inspection intervals has reached the end of the target operation period (S6). If the sum of the inspection intervals has not exceeded the target operation period (S6 → No), the inspection support device 1 sets the post-inspection flaw size probability distribution (curve 321 in FIG. 8B ) as the initial flaw size probability distribution (curve 311 in FIG. 7 ). The inspection support device 1 then returns the process to step S3. The initial flaw size probability distribution is the initial distribution of the flaw size probability distribution over time used in step S3. In this way, the inspection support device 1 sequentially determines the inspection interval by setting the post-inspection flaw size probability distribution calculated in step S3 as the initial distribution of the flaw size probability distribution over time.

[0054] The inspection support device 1 then executes steps S3, S4, and S5 using the post-inspection flaw size probability distribution (curve 321) as the initial flaw size probability distribution (curve 311 in FIG. 7 ). This allows the inspection interval determination unit 104 to determine the next inspection interval. If the total inspection interval exceeds the target operation period (S6 → Yes), the inspection interval determination unit 104 outputs all of the determined inspection intervals (S7: fourth step), completing the procedure.

[0055] FIG. 9 is a schematic diagram showing the procedure for determining the inspection interval. In the following, step numbers correspond to those in FIG. 3. The horizontal axis of the graph shown in FIG. 9 represents the elapsed time since the start of operation of the object, and the vertical axis represents the estimated failure probability at each elapsed time. The dashed line 401 represents the tolerable risk input in step S1. The curves 402 to 404 represent the estimated failure probability calculated in step S4. Points 405 to 407 represent the estimated failure probability calculated using the flaw size probability distribution p(a|before inspection) before the inspection. Points 408 to 410 represent the estimated failure probability calculated using the flaw size probability distribution p(a|before inspection) immediately after each inspection. The reason why the estimated failure probability is low at points 408 to 410 is as follows. That is, the flaw size probability distribution after Bayesian estimation is reset as the initial flaw size probability distribution, and the estimated failure probability shown in region 304 in FIG. 7 becomes a low value.

[0056] According to the first embodiment, even for an object in which a flaw 203 is rarely detected, a post-inspection flaw size probability distribution is calculated utilizing the probability of non-detection of the flaw 203. Then, based on the post-inspection flaw size probability distribution, an inspection interval is determined using a rational index. In other words, according to the first embodiment, for an inspection object in which flaw detection is rare, it is possible to rationally determine the next inspection interval by utilizing information that no flaws were detected in the regular inspection. Furthermore, in step S5, the post-inspection flaw size probability distribution is Bayesian-estimated, and the Bayesian-estimated post-inspection flaw size probability distribution becomes a new initial flaw size probability distribution. In this way, it is possible to determine multiple inspection intervals. Furthermore, by using Bayesian estimation, the event in which a flaw was not detected as a result of the inspection can be reflected in the next inspection.

[0057] Second Embodiment Next, an examination support method according to a second embodiment of the present invention will be described with reference to Fig. 10. In Fig. 10, the same processes as those in Fig. 7 will not be described.

[0058] 10 is a conceptual diagram showing the flaw size probability distribution over time according to the second embodiment. In the first embodiment, in step S5, the Bayesian estimation unit 105 Bayesian-estimates the flaw size probability distribution after inspection according to equation (5). In contrast, in the second embodiment, if the flaw 203 is not detected even after a predetermined number of inspections are performed at each inspection timing, the inspection support device 1 calculates the flaw size probability distribution after inspection by changing the scale parameter "a0" in equation (6).

[0059] In the second embodiment, it is assumed that the flaw size probability distribution follows a Weibull distribution. Curve 331 in FIG. 10 is the initial flaw size probability distribution, which is a Weibull distribution, and is expressed by Equation (6). The Weibull distribution is used because it is easy to update the flaw size probability distribution. If the flaw size probability distribution can be updated, the flaw size probability distribution does not have to be limited to a Weibull distribution.

[0060]

[0061] In equation (6), "β" is a predetermined shape parameter, and "a0" is a predetermined scale parameter.

[0062] In addition, in FIG. 10, curve 332 is the flaw size probability distribution after the calculation time "T" has elapsed (the flaw size probability distribution over time) with the initial flaw size probability distribution shown by curve 331 as the initial value.

[0063] That is, in the second embodiment, if the calculation result indicates that the flaw 203 remains undetected even after a predetermined number of inspections, the initial flaw size probability distribution is deemed to be erroneous. The fact that the flaw 203 remains undetected even after a predetermined number of inspections means that the inspection detection probability, which will be described later, does not reach the upper limit detection probability "α." Therefore, the progression analysis unit 103 changes the scale parameter "a0" in equation (6) and processes the flaw size probability distribution over time again. The hollow arrows in FIG. 10 indicate that the initial flaw size probability distribution (curve 331) is changed if the inspection detection probability does not reach the upper limit detection probability "α." The inspection detection probability is the probability that a flaw is detected during inspection.

[0064] (Flowchart) The above processing will be explained using Fig. 11. Fig. 11 is a flowchart showing the procedure of the inspection support method according to the second embodiment. In Fig. 11, the same processing as in Fig. 3 is assigned the same step number, and the explanation thereof will be omitted.

[0065] First, in step S1A, the shape parameter "β", the scale parameter "a0", and the upper limit detection probability "α" are input instead of the initial flaw size probability distribution. Then, steps S3A and S4A are performed.

[0066] FIG. 12 shows details of steps S3A and S4A in FIG. 11. The processes in steps S301 to S303 are the same as those shown in FIG. 6, except that the initial flaw size probability distribution is a Weibull distribution with a shape parameter "β" and a scale parameter "α0." After step S303, the inspection interval determination unit 104 calculates the inspection detection probability (S411). The detection probability calculated in step S2 (curve 301 in FIG. 5) is POD(a), and the flaw size probability distribution over time calculated in steps S3 and S4 before inspection (curve 332) is p(a|before inspection). Then, the inspection detection probability of the flaw 203 is expressed by the following equation (7): The inspection detection probability is the probability that a flaw of a certain size will be detected when an inspection is performed.

[0067]

[0068] Next, the inspection interval determination unit 104 determines whether the inspection detection probability shown in formula (7) is greater than the upper limit detection probability "α" (S412: sixth step). If the inspection detection probability is less than the upper limit detection probability "α" (S412 → No), the inspection interval determination unit 104 determines whether the calculation time (t) is greater than Tmax (S413). In the example shown in FIG. 12, Tmax input in step S301 is used, but this is not limiting. If the calculation time (t) is equal to or less than Tmax (S413 → No), the inspection support device 100 returns the process to step S302.

[0069] If the calculation time (t) is greater than Tmax (S413 → No), the inspection interval determination unit 104 updates the scale parameter "α0" (S414). Then, the inspection support device 100 performs the processes from step S301 onwards, using the Weibull distribution with the updated scale parameter "α0" as the initial flaw size probability distribution.

[0070] On the other hand, if the determination in step S412 is "Yes," the test interval determination unit 104 performs the process of step S404. The process of step S404 is the same as the process shown in FIG.

[0071] If the flaw 203 is not detected even after a predetermined number of times, it is considered that the actual detection probability of the flaw 203 was lower. In other words, it is considered that the initial distribution of the flaw size probability distribution was incorrect. Therefore, the inspection interval determination unit 104 estimates the actual detection probability of the flaw 203 as the upper limit detection probability "α". Then, the inspection interval determination unit 104 repeatedly calculates p(a|before inspection) while changing the scale parameter "a0" so that the value of equation (7) is equal to α or greater than α. In other words, the inspection interval determination unit 104 changes the initial flaw size probability distribution by changing the scale parameter "a0" of the Weibull distribution, and performs the processes from step S3 onwards again. In this way, the progression analysis unit 103 updates the initial distribution of the flaw size probability distribution. Using this updated initial flaw size probability distribution, the progression analysis unit 103 calculates the flaw size probability distribution (curve 332 in FIG. 10) after the calculation time "T" has elapsed in step S404, and estimates p(a|after inspection).

[0072] In this way, in the second embodiment, if the inspection detection probability does not reach the upper limit detection probability, which is a predetermined probability, even after a predetermined number of inspections, the initial flaw size probability distribution is changed in step S4. In particular, by using a Weibull distribution as the flaw size probability distribution, it is possible to easily change the initial flaw size probability distribution simply by changing the shape parameter "a0".

[0073] Steps other than step S1 and step S5 are the same as those in the first embodiment. As an application of this embodiment, instead of the Weibull distribution, it is also possible to assume another probability distribution in which the initial flaw size probability distribution has a scale parameter.

[0074] According to the second embodiment, if a flaw 203 is not detected even after a predetermined number of inspections, the initial flaw size probability distribution is determined to be erroneous and is changed. In this way, the initial flaw size probability distribution can be generated to reflect prior knowledge. This makes it possible to determine a more realistic inspection interval than in the first embodiment.

[0075] Third Embodiment Next, an inspection support method according to a third embodiment of the present invention will be described with reference to Fig. 13. In the first embodiment, in step S3, the evolution analysis unit 103 calculates a solution to the Fokker-Planck equation, which is a stochastic differential equation. The evolution analysis unit 103 then calculates a flaw size probability distribution over time. In this embodiment, the detailed calculation step of the stochastic differential equation in step S3 is replaced with a Monte Carlo calculation.

[0076] 13 is a flowchart showing the detailed calculation procedure of step S3 according to the third embodiment. First, the progression analysis unit 103 inputs the calculation time "Tmax", the evaluation time interval "ΔT", and the number of samples for the Monte Carlo calculation via the data input unit 101 (S311). Next, the progression analysis unit 103 randomly samples flaw sizes according to the flaw size probability distribution (S312). The flaw size probability distribution used is set in advance. In this embodiment, a normal distribution is used, but the Weibull distribution described above may also be used.

[0077] Next, the progression analysis unit 103 randomly samples the flaw growth rate according to a probability distribution (S313). As described above, the progression analysis unit 103 uses a normal distribution with a mean μ(a, t) and a standard deviation σ(a, t), but this is not limited to a normal distribution. Note that t is the calculation time (t) used in the first embodiment.

[0078] Then, the progression analysis unit 103 calculates the progression of the flaw 203 based on the progression speed of the flaw size distribution randomly sampled in step S313, and calculates the flaw size after time ΔT (S314).

[0079] Next, the progress analysis unit 103 determines whether the total elapsed time (ΔT x number of executions of S314) exceeds the specified calculation time "Tmax" (S315). If the total elapsed time does not exceed the specified calculation time "Tmax" (S315 → No), the progress analysis unit 103 returns to the process of S313 and sequentially calculates the flaw size after the next ΔT time. If the total elapsed time exceeds the specified calculation time "Tmax" (S315 → Yes), the progress analysis unit 103 completes one flaw sampling calculation.

[0080] Next, the progress analysis unit 103 determines whether or not the flaw sampling calculations (S313 to S315) for the number of samplings specified in step S311 have been completed (S316). If not completed (S316 → No), the progress analysis unit 103 returns to the processing of step S312 and repeatedly executes steps S312 to S316. If the number of calculations reaches the number of samplings and the progress analysis unit 103 determines that the sampling calculations have been completed (S316 → Yes), the progress analysis unit 103 proceeds to the processing of step S317.

[0081] In step S317, the progression analysis unit 103 outputs the flaw size probability distribution (empirical distribution) for each time ΔT. In this way, the progression analysis unit 103 calculates the flaw size probability distribution after the lapse of time T, as indicated by curve 312 in FIG. 7. Note that T corresponds to the calculation time (t-1) in the first and second embodiments. Then, the inspection support device 1 completes step S3 in FIG. 3 and proceeds to step S4. The flaw size probability distribution for each time ΔT output in step S317 may be output as a histogram, or may be output after being fitted to an appropriate probability distribution.

[0082] The steps other than step S3 are the same as those in the first embodiment. It is also possible to combine the second embodiment and the third embodiment.

[0083] In this way, in the third embodiment, the flaw size and flaw growth rate are randomly sampled to calculate the flaw size probability distribution over time.

[0084] According to the third embodiment, the flaw size probability distribution after the calculation time "T" shown by the curve 312 in Figure 7 is calculated by a Monte Carlo simulation, rather than analytically as in the first embodiment. Note that the calculation time "T" can be expressed as the total value of ΔT. This allows the flaw size probability distribution after the calculation time "T" to be calculated without advanced mathematical knowledge.

[0085] [Example of Output Screen] Fig. 14 is a diagram showing an example of an output screen in this embodiment. The output screen 500 shown in Fig. 14 is a screen output to the output device 115 (see Fig. 2) in step S7 of Fig. 3 or Fig. 11. The output screen 500 shown in Fig. 14 has an examination interval output section 501. The examination interval output section 501 outputs the interval from the previous examination (or the start of the examination) to the next examination.

[0086] The present invention is not limited to the above-described embodiments and includes various modifications. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those having all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

[0087] Furthermore, in this embodiment, a crack is used as an example of the flaw 203, but the flaw 203 is not limited to a crack. For example, this embodiment is applicable to general flaws that grow over time, such as holes. When a flaw other than a crack is used as the flaw 203, the methods and formulas used in this embodiment can be the same as those described in this embodiment, except that the Paris law shown in formula (1) is not used.

[0088] In addition, in this embodiment, the test interval, which is the interval between tests, is output, but this is not limiting. For example, test timing (test period) such as the number of days after the previous test (or the start of testing) to perform a test, the number of days after the start of testing to perform each test, or the test date and time calculated based on the test interval may be output.

[0089] Furthermore, some or all of the above-described configurations, functions, data input unit 101 to Bayesian estimation unit 105, storage device 113, etc. may be implemented in hardware, for example, by designing them as integrated circuits. Furthermore, as shown in FIG. 2 , the above-described configurations, functions, etc. may be implemented in software by a processor such as a CPU interpreting and executing a program that implements each function. Information such as programs, tables, and files that implement each function can be stored in a recording device such as memory 111 or an SSD (Solid State Drive), or in a recording medium such as an IC (Integrated Circuit) card, an SD (Secure Digital) card, or a DVD (Digital Versatile Disc). Furthermore, in each embodiment, control lines and information lines are shown that are considered necessary for explanation, and not all control lines and information lines are necessarily shown in the product. In reality, it is safe to assume that almost all configurations are interconnected.

[0090] 100 Inspection support device 101 Data input unit (input unit) 102 Detection probability calculation unit 103 Progression analysis unit 104 Inspection interval determination unit (inspection timing determination unit, output processing unit) 105 Bayesian estimation unit 115 Output device 202 Inspection object 203 Flaw 205 Variation in installation position (inspection variation) 206 Variation in ultrasonic transmission direction (inspection variation) 301 Curve (flaw detection probability) 302 Curve 304 Region (region where flaw size probability distribution over time and fracture probability overlap) 305 Curve (fracture probability) 311 Curve (initial distribution of flaw size probability distribution) 312 Curve (flaw size probability distribution over time) 321 Curve (flaw size probability distribution after inspection) 401 Dashed line (tolerable risk) 331 Curve (initial distribution of flaw size probability distribution in Weibull distribution) 332 Curve (flaw size probability distribution over time in Weibull distribution) S2 Calculate flaw detection probability (first step) S3 Calculate flaw size probability distribution over time (second step) S4 Determine inspection interval (third step) S5 Bayesian estimation of flaw size probability distribution after inspection (fifth step) S7 Output all inspection intervals (fourth step) S411 Compare detection probability during inspection with upper limit detection probability (sixth step)

Claims

1. An inspection support method characterized in that an inspection support device executes the following steps: a first step of calculating the probability of flaw detection; a second step of calculating a flaw size probability distribution over time using the inspection variability and an initial flaw size probability distribution input via an input unit; a third step of determining an inspection timing if the area of ​​the region where the flaw size probability distribution over time and the fracture probability overlap exceeds a predetermined tolerable risk; and a fourth step of outputting the determined inspection timing.

2. The inspection support method described in claim 1, further comprising: a fifth step of estimating a post-inspection flaw size probability distribution by Bayesian estimation using the flaw detection probability calculated in the first step and the flaw size probability distribution over time calculated in the second step; wherein the inspection support device sequentially determines the inspection timing by setting the post-inspection flaw size probability distribution calculated in the fifth step as the initial distribution of the flaw size probability distribution over time used in the second step.

3. The inspection support method according to claim 1, characterized in that the Bayesian estimation performed in the fifth step is calculated as the following formula (1): p(a|after inspection) = A(1-POD(a))p(a|before inspection) ... (1) where, in formula (1), POD(a) is the probability of detecting a flaw when the flaw size is "a", p(a|before inspection) is the flaw size probability distribution over time immediately before inspection, and A is a normalization constant set so that the total probability is 1.

4. The inspection support method according to claim 1, characterized in that in the second step, the time change in the flaw size probability distribution is the time change in the flaw size probability distribution for crack-like defects that follow the Paris law.

5. The inspection support method as described in claim 1, characterized in that in the third step, if the inspection detection probability does not reach a predetermined upper limit detection probability even after the inspection is performed a predetermined number of times, the inspection support device performs a sixth step of changing the initial distribution of the flaw size probability distribution.

6. The inspection support method described in claim 5, characterized in that the flaw size probability distribution is a Weibull distribution having a predetermined shape parameter and a scale parameter, and the inspection support device updates the initial distribution of the flaw size probability distribution by changing the scale parameter in the sixth step, and re-performs processing from the second step onwards.

7. The inspection support method according to claim 5, characterized in that the inspection detection probability is in accordance with the following formula (2): In equation (2), POD(a) is the probability of detection of the flaw calculated in the first step, and p(a|before inspection) is the flaw size probability distribution over time before inspection.

8. The inspection support method according to claim 1, characterized in that in the second step, the inspection support device calculates the flaw size probability distribution over time by solving the Fokker-Planck equation, which is a flaw progression equation in which the flaw size is a continuous random variable.

9. The inspection support method according to claim 1, characterized in that in the third step, the flaw size and the flaw growth rate are randomly sampled to calculate the flaw size probability distribution over time.

10. An inspection support device comprising: a detection probability calculation unit which calculates the probability of flaw detection; a progression analysis unit which calculates the flaw size probability distribution over time using the inspection variability and the initial flaw size probability distribution input via the input unit; an inspection timing determination unit which determines the inspection timing when the area of ​​the region where the flaw size probability distribution over time and the fracture probability overlap exceeds a predetermined allowable risk; and an output processing unit which outputs the determined inspection timing.

Citation Information

Patent Citations

  • Method for diagnosing structural member

    JP1986140838A