JP6178277B2 - Influence factor information acquisition method and influence factor information acquisition apparatus in failure analysis - Google Patents

Description

  The present invention relates to an influence factor information acquisition method and an influence factor information acquisition apparatus in failure analysis, and in particular, when performing analysis similar to Weibull analysis on articles, equipment, etc. whose deterioration factors are unknown, deterioration factors, etc. The present invention relates to an influence factor information acquisition method and an influence factor information acquisition apparatus in failure analysis, which obtain information indicating the existence probability of the fault.

  Conventionally, when analyzing deterioration or failure of an article or equipment (hereinafter referred to as failure analysis), so-called Weibull analysis, cumulative hazard analysis, or the like is performed (see, for example, Patent Documents 1 to 3). This is an analysis of the relationship between an index that represents the degradation of the target that is a factor such as the occurrence of a failure and an index that represents the usage time that is considered to be related to the degradation of the target.

  These analysis results can be expressed as a plot graph on Weibull probability paper or the like. If the plot graph as a whole follows linear regression, it can be said that the analysis object can be analyzed by this analysis method, and parameters, function expressions, etc. useful for failure analysis can be derived therefrom.

  On the other hand, if the plot graph does not follow the linear regression, it is interpreted that (1) the analysis method is inappropriate, or (2) or the analysis data contains failure / deterioration due to different causes, different influence factors, etc. Is done. In particular, in the case of a plot graph that can be represented by a combination of a plurality of straight lines instead of a single linear regression, the possibility of (2) is considered.

International Publication Number WO2003 / 085548 Japanese Patent Laid-Open No. 10-034122 JP 2003-331087 A

Satoshi Makabe, New Edition Introduction to Reliability Engineering, Japanese Standards Association, pp.115-118, 2010.

  However, in the case of (2) described above, depending on the analysis target, it may be difficult to identify or estimate the cause of failure / deterioration and the influencing factors. For example, in the case of a subject used outdoors, it was difficult to identify the environmental conditions that affected the failure or deterioration of the subject and to sort the analysis data based on these environmental conditions. .

  In addition, depending on the shape of the plot, it is also considered to perform analysis using a composite distribution instead of analyzing the whole with one distribution (see, for example, Non-Patent Document 1). In order to perform analysis using a composite distribution, it is necessary to set the age of inflection points where the trend of the plot becomes discontinuous, but the reason for the composite distribution (for example, different failure patterns mixed in the analysis data) If the inflection point that is discontinuous is not known, the years of use etc. could not be set in the first place.

  The conventional technique related to Weibull analysis and the like is how to use the result of Weibull analysis and the like (see Patent Documents 1 to 3), and as in this proposal, linear regression of plots found in the results of Weibull analysis and the like. Regarding the divergence from, there is no technical proposal to solve the above-mentioned problem through the assumption of the influencing factors behind it.

  The present invention has been made to solve the above problems, and in order to proceed with failure analysis, information representing an influence factor that changes the tendency of the plot (hereinafter referred to as influence factor information). An object of the present invention is to propose an influence factor information acquisition method and influence factor information acquisition apparatus in failure analysis to obtain the above.

The method for acquiring influence factor information in failure analysis according to the present invention includes a probability density function defining unit that defines a probability density function that represents the existence probability of a plurality of influence factors that cause a failure, and a relative presence probability of the plurality of influence factors. A relative ratio defining section for defining a ratio, a parameter calculating section for calculating a parameter of a probability density function, and influence factor information for calculating the existence probability of the plurality of influence factors as influence factor information in failure analysis using the parameters An influence factor information acquisition method in failure analysis, executed by an influence factor information acquisition device comprising a calculation unit, wherein the plot between the period of use and the frequency of occurrence of failure has at least one inflection point to the probability density function defining portion is a plurality of impact factors causing the failure corresponding to each period of use before and after the inflection point of the plot A probability density function defined step of defining a probability density function representing the existence probability of the relative proportions specified section, the presence of the plurality of influencing factors in the use period including before and after the inflection point and the inflection point the relative proportions specified step of defining the relative proportions of the probability, the parameter calculator sets a function of information entropy based on the relative proportions of the existence probability of the plurality of influencing factors, based on the information of the plot wherein a parameter calculation step of information satisfying the functions of the entropy calculating the parameters of the probability density function, the influence factor information calculation unit, the influence factor information in the failure analysis of the existence probability of the plurality of influencing factors by using the parameter Te And an influencing factor information calculating step.

  2. The influence factor information acquisition method according to claim 1, wherein in the probability density function defining step, the existence probability is defined using a residual between the plot information and a regression line value with respect to the plot information. To do.

  3. The influencing factor information acquisition method according to claim 2, wherein in the influencing factor information calculation step, the inflection point is specified based on a plurality of probability density functions respectively corresponding to the existence probabilities of the influencing factors. .

  4. The influence factor information acquisition method according to claim 3, wherein when there are two or more inflection points, the analysis range dividing unit analyzes the analysis range for the plot information so that one inflection point is always included. Is further provided with an analysis range dividing step for dividing.

  In the influence factor information acquisition apparatus in failure analysis according to the present invention, when the plot between the period of use and the frequency of occurrence of the failure has at least one inflection point, each before and after the inflection point of the plot A probability density function defining part that defines a probability density function that represents the probability of existence of a plurality of influencing factors that cause the failure corresponding to the use period, and the inflection point and the use period including before and after the inflection point. Based on the information in the plot, a relative ratio defining unit that defines the relative ratio of the existence probabilities of a plurality of influence factors, and a function of information entropy based on the relative ratio of the existence probabilities of the plurality of influence factors A parameter calculating unit that calculates a parameter of the probability density function that satisfies the information entropy function, and the presence of the plurality of influencing factors using the parameter. So that and an influence factor information calculating unit that calculates a probability as influencing factor information in the failure analysis.

  According to the present invention, even when the influence factor related to failure analysis is not clear, the probability density function representing the existence probability of a plurality of influence factors from the information of the plot between the period of use and the occurrence frequency of the failure is highly accurate. By using the probability density function, it is possible to estimate the usable period of use due to a plurality of influence factors with high accuracy.

It is a block diagram which shows the structure of the influence factor information acquisition apparatus in the 1st Embodiment of this invention. It is a plot which shows the result of a cumulative hazard analysis. It is a plot of the residual e with respect to the years of service t. It is a plot of the adjustment residual Oe with respect to the years of service t. It is a graph with which it uses for description of the relationship between adjustment residual Oe and the influence of influence factors A and B. FIG. It is a graph which shows the calculation result (respective probability density distribution) of existence probability A and B. It is a graph which shows the calculation result of influence ratio A and B. It is a graph which shows the calculation result of adjustment information entropy CH (t). It is a graph which shows the relationship between adjustment residual Oe and adjustment information entropy CH (t). It is a flowchart which shows the influence factor information acquisition processing procedure. It is a block diagram which shows the structure of the influence factor information acquisition apparatus in the 2nd Embodiment of this invention. It is a graph which shows the example of the plot in which the residual e has a some inflection point, and the example of a division | segmentation of an analysis range. It is a graph which shows the example of an analysis of existence probability in case residual e has a plurality of inflection points.

<Outline of influence factor information acquisition device>
The outline of the influence factor information acquisition apparatus in the embodiment of the present invention will be described below. This influencing factor information acquisition device has a plurality of influencing factors regarding the occurrence of a failure of an object such as an article or facility, and a plot between the operating time (the number of years of use) that causes the failure and the frequency of occurrence of the failure is When having at least one or more inflection points, a probability density function representing the existence probability of influencing factors corresponding to the respective areas before and after the inflection points with respect to the information obtained from the plot, and the inflection points After specifying the relative ratio (influence ratio) of existence probabilities for multiple influencing factors at each point (year of use) in the included plot graph, the probability density of existence probabilities for multiple influencing factors using information entropy theory Parameters for obtaining the function are calculated, and the existence probabilities and the influence ratios of the plurality of influence factors that best reflect the tendency of the plot are calculated with high accuracy.

<First Embodiment>
<Configuration of influence factor information acquisition device>
As shown in FIG. 1, the influence factor information acquisition device 1 uses the calculation function unit 10 to calculate data (hereinafter referred to as “this”) that is a source of calculation for acquiring influence factor information (after-operation) in failure analysis. Is referred to as “facility data”), data relating to the specifications of the equipment (hereinafter referred to as “facility specification data”), and data relating to inspection of the equipment (hereinafter referred to as “facility inspection data”). ) Etc. are stored in the data storage unit 100, the calculation function unit 10, all calculation processes calculated by the calculation function unit 10, and all calculation processes and result recording functions for recording the results in the data storage unit 100. And a display / output function unit 80 that displays or outputs the influence factor information obtained by the calculation function unit 10.

  The influence factor information acquisition apparatus 1 is realized by installing a computer program (software) on a computer (hardware) including a CPU (Central Processing Unit), a memory, an interface, and the like. Each unit is realized by cooperation of various hardware resources of a computer and a computer program. Further, the computer program may be provided in a state of being stored in a computer-readable recording medium or a storage device, or may be provided via an electric communication line.

  The calculation function unit 10 analyzes the facility data to be analyzed by an analysis method such as Weibull analysis, the residual e between the analysis result plot and the regression line, and the adjusted residual Oe. Is a residual for determining the shape of whether the plot of the adjustment residual Oe, which is the result of the Weibull analysis by the residual calculation function unit 30 and the analysis execution function unit 20, is convex upward or downward. The difference plot shape discrimination function unit 40 is set based on the existence probability of the influence factors A and B that cause the failure in the analysis range, the relative ratio (influence ratio) of the existence probability, and the influence ratio of the influence factors A and B. Setting function unit 50 for setting adjustment information entropy multiplied by the adjustment coefficient α so that the adjustment information entropy and the plot of the adjustment residual Oe are best matched. Are constituted by the parameter such calculating function unit 60 calculates various parameters defining the to the probability distribution (normal distribution).

  The analysis execution function unit 20 does not clearly know the influencing factors affecting the failure and deterioration, and the cumulative hazards related to equipment failures that are considered to be mixed by the influences of these different influencing factors. Perform analysis.

Here, FIG. 2 shows an analysis result by cumulative hazard analysis. In FIG. 2, the horizontal axis represents the logarithmic value L gt of the equipment usage years t, and the vertical axis represents the cumulative hazard value H (t) in the usage years t representing the failure occurrence frequency, and the regression line Y (t) of the plot graph Is obtained by the least square method or the like. In this case, the linear regression equation of the regression line is represented by Y (t) = 6.740 ^* Lotegt-23.81. In addition, it is also possible to use the number of months of use or the number of days of use instead of the number of years of use t as the usage period.

  Here, the cumulative hazard analysis may be performed by, for example, a known Weibull analysis that matches the nature of the data to be analyzed, and each axis of the plot in FIG. 2 is limited to the cumulative hazard value H (t) or the number of years of use t. Instead, it may be an index representing the deterioration of the object and an index considered to be related to the deterioration of the object, and may follow various known probability sheets.

  In this case, the plot is bent gently with respect to the regression line Y (t) as a whole, and it cannot be said that the plot is along linear regression. In the present embodiment, it is assumed that two different influence factors A and B that affect a failure lead to such a result. From FIG. 2, it is assumed that there is an inflection point (inflection point) in the vicinity of 3.50 <L <gt <3.57, but it is not clear on the plot (in the conventional method, the inflection point (inflection point is invisible)). The point of use (t) may be read and the analysis data may be separated using the use year t as a threshold value).

  The residual calculation function unit 30 calculates the difference between the accumulated hazard value H (t) associated with the service year t and the calculated value Y (t) based on the linear regression equation for the service year t. The residual et at t is obtained according to (Equation 1).

et = LogH (t) -Y (t) ......... (Equation 1)
et: Residual error when using years t H (t): Cumulative hazard value when using years t Y (t): Linear regression equation when using years t (Y (t) = 6.740 ^* L gt− Calculated value according to 23.81)

  Subsequently, as shown in FIG. 3, the residual calculation function unit 30 plots the horizontal axis as the service life t and the vertical axis as the residual e. At this time, when the residual plot shape discrimination function unit 40 determines that the shape of the plot graph of the residual e is convex downward, the residual calculation function unit 30 calculates the residual e according to the following (Equation 2). Recalculate. As a result, the residual calculation function unit 30 can interpret the analysis related to the deviation from the linear regression of the plot as a maximum value problem having a convex maximum value using the deviation width (residual e).

  et = − [LogH (t) −Y (t)] ………………………………………… (Formula 2)

  When the residual e has a negative value, the residual calculation function unit 30 calculates the floor function of the minimum value emin of the residual e (maximum less than the real number for a certain real number) by the following (Equation 3). For example, when emin is “−1.2”, the floor function is “−2”) is subtracted from the residual e to derive an adjusted residual Oe.

Oet = et− [emin] ……………………………………… (Formula 3)
Oet: Adjustment residual for years of service t

  As a result, as shown in FIG. 4, the adjustment residual Oe is always a positive value and has an upwardly convex distribution when the horizontal axis is the service life t and the vertical axis is the adjustment residual Oe.

  Here, the adjustment residual Oe is divided into two regions of monotonic increase and monotonic decrease with the maximum value (inflection point) as a boundary. Assuming that the deviation from the linear approximation in FIG. 2 is due to failure / deterioration due to different causes and two different influencing factors (influencing factors) A and B in the analysis data, as shown in FIG. The distribution of the adjustment residual Oe is a region due to a failure where the influence of the influence factor A is dominant on the left side of the maximum value point Pmax, and a region due to the failure where the influence of the influence factor B is dominant on the right side of the maximum value point Pmax. And the maximum value point Pmax can be interpreted as a point where the influence of the influence factors A and B antagonize (ie, interchange).

  Here, it is assumed that the presence of the influence factors A and B can be expressed by a probability. In this embodiment, this probability follows a normal distribution, the existence probability of the influencing factor A is called the existence probability A, the existence probability of the influencing factor B is called the existence probability B, and the setting function unit 50 uses the existence probability A and the existence probability. The calculation formula of the probability B is set as shown in (Formula 4).

・・・（式４）
ただし、Ｐｊｔ：影響因子ｊの使用年数ｔのときの存在確率
N(t, μ_j, σ_j ²)：使用年数ｔを変数とする、平均μｊ、標準偏差σｊの正規分布
ｊ：ＡあるいはＢ
μＡ≠μＢ

... (Formula 4)
However, Pjt: Existence probability when the influence factor j is the number of years of use t
N (t, μ _j , σ _j ² ): Normal distribution of mean μ _j and standard deviation σ _{j with} years of service t as variables j: A or B
μA ≠ μB

  In this case, FIG. 6 shows the probability density functions PDF-A and PDF-B of the existence probabilities A (PAt) and B (PBt). That is, the intersection between the probability density function PDF-A of the existence probability A (PAt) and the probability density function PDF-B of the existence probability B (PBt) is the point where the influence of the influence factors A and B antagonizes (ie, interchanges). (Year of use t). However, since parameters such as the average μj and the standard deviation σj in the existence probability Pjt are unknown, the intersection point is directly determined from only the probability density functions PDF-A and PDF-B of the existence probabilities A (PAt) and B (PBt). I can't ask for it.

  Therefore, as described above, in the present embodiment, since the relationship between the existence probabilities Pjt is considered to be reflected in the distribution of the adjustment residual Oe, the setting function unit 50 determines the existence probability PAt and the existence probability in the usage years t. A relative ratio with PBt (hereinafter referred to as influence ratios RAt and RBt) is defined as the following (Equation 5). FIG. 7 shows the influence ratio RAt of the existence probability PAt and the influence ratio RBt of the existence probability PBt. That is, also in this case, the intersection of the influence ratio RAt and the influence ratio RBt is the point (the number of years of use t) where the influence of the influence factors A and B antagonize (that is, interchange).

・・・・・・・・・・・・・・・・・・・・・・・・・・・（式５）
Ｒｊｔ：存在確率ｊの使用年数ｔのときの相対割合
ｋ：Ａ、Ｂ、…

・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ (Formula 5)
Rjt: Relative ratio when the existence probability j is the number of years of use t: A, B,...

  Then, the setting function unit 50 uses the information entropy theory to obtain the average μj and the standard deviation σj that are parameters of the existence probability Pjt, and uses the information entropy calculation formula as the influence ratio of the existence probability Pjt in (Expression 5). Using Rjt, the following (formula 6) is set.

・・・・・・・・・・・・・・・・・・・・・・・・・（式６）
Ｈ（ｔ）：使用年数ｔにおける情報エントロピー

・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ ・ (Formula 6)
H (t): Information entropy in years of use t

  Here, information entropy can be understood as meaning the informational value of the information. This will be described with reference to FIG. 5 in accordance with the situation of the embodiment. The maximum value Pmax (convex apex) in FIG. 5 is a place where the influence of the influence factor A and the influence factor B on the failure is antagonized (replaced), and it is determined beforehand by which factor the failure occurs. Most difficult to predict. This means that the value of information is the highest, that is, the information entropy is the largest.

  Conversely, the further away from the maximum value Pmax, the more likely the failure due to the influence of the influence factor A occurs. That is, since it becomes easier to predict that a failure due to the influencing factor A will occur in advance toward the left side, the value of information becomes lower and the information entropy becomes lower. Further, the further away from the maximum value Pmax on the right side, the easier it is to predict a failure due to the influence factor B, so the value of information becomes lower and the information entropy becomes lower.

  That is, the shape of the plot showing the relationship between the adjustment residual Oe and the usage time t shown in FIG. 5 is the relationship between the information entropy H (t) derived from the influence ratio RAt and the influence ratio RBt and the usage time t. It can be considered similar. As described above, the probability distribution (normal distribution in this case) that best matches the plot of the adjustment residual Oe and the locus of the information entropy H (t) is the existence probability A or the existence probability B to be obtained.

  Here, since the plot of the adjustment residual Oe and the value taken by the information entropy H (t) do not necessarily match, the setting function unit 40 multiplies the above (Equation 6) by the adjustment coefficient α, The adjustment information entropy CH (t) of 7) is specified. FIG. 8 shows a plot of adjustment information entropy CH (t), and FIG. 9 shows a plot of residual Oe in FIG. 5 representing the relationship between adjustment residual Oe and usage time t and adjustment information entropy in FIG. It shows that the plot of CH (t) is similar.

・・・・・・・・・・・・・・・・・・・・・・・・（式７）
ＣＨ（ｔ）：使用年数ｙにおける調整情報エントロピー
α：調整係数

... (Equation 7)
CH (t): Adjustment information entropy α in the service life y: Adjustment coefficient

  The parameter calculation function unit 60 derives various parameters that define a probability distribution (normal distribution) satisfying (Expression 7) that best matches the plot of the adjustment residual Oe by the least square method or the like. That is, the parameter calculation function unit 60 calculates the probability density representing the existence probability Pjt (normal distribution) when the calculated value based on the adjustment information entropy CH (t) in (Equation 7) and the plot of the adjustment residual Oe are the best match. Various parameters that define the function are obtained by the method of least squares.

  In the present embodiment, since the existence probability A and the existence probability B are set to follow a normal distribution, the various parameters required by the parameter calculation function unit 60 are the average μA, standard deviation σA, and average μB in (Expression 4). Standard deviation σB. The parameter calculation function unit 60 can obtain various parameters by the least square method and simultaneously obtain the adjustment coefficient α.

  Since the parameter calculation function unit 60 calculates various parameters (average μA, standard deviation σA, average μB, standard deviation σB), the existence probability A (PAt) and the existence probability B (PBt) according to (Equation 4). ) And an influence ratio RAt and an influence ratio RBt based on the existence probability A (PAt) and the existence probability B (PBt) according to (Expression 5).

  As a result, the existence probabilities PAt and PBt, which are important information when specifying the specific influence factors A and B, and the influence ratios RAt and RBt for the failure are determined. Even if specific influential factors A and B are not specified, it can be used when estimating the same equipment / failure. Further, since the parameters (average μA, standard deviation σA, average μB, standard deviation σB) are obtained, the probability density function of the existence probabilities PAt and PBt can be derived from (Equation 4). It becomes possible to clarify the usage time t that is the intersection of −A and PDF-B, that is, the usage time t that is the inflection point in the plot result of Weibull analysis or the like.

  The total calculation process / result recording function unit 70 records the calculation process and calculation results by the parameter calculation function unit 60 in the data storage unit 100 and outputs the calculation results to the display / output function unit 80.

  The display / output function unit 80, based on the calculation result by the parameter calculation function unit 60, the probability density functions PDF-A, PDF-B, the influence ratio RAt and the influence ratio RBt as shown in FIGS. Is visualized and displayed as a graph or the like.

<Operation of influence factor information acquisition device>
The operation of the influence factor information acquisition apparatus 1 having such a configuration will be described with reference to the flowchart of FIG. In step SP1, the residual calculation function unit 30 of the calculation function unit 10 calculates the cumulative hazard value associated with the service life t based on the result of the cumulative hazard analysis related to the equipment failure by the analysis execution function unit 20 (FIG. 2). By calculating the difference between H (t) and the calculated value Y (t) based on the linear regression equation at the time of use t, the residual et at the time of use t is calculated according to (Expression 1).

  In step SP2, the residual plot shape discriminating function unit 40 plots the plot with two or more inflection points when the horizontal axis is the number of years used t and the vertical axis is the residual e as shown in FIG. If there is only one inflection point, the process proceeds to the next step SP3 (step SP2: No).

  In step SP3, if the residual plot shape determination function unit 40 determines that the plot shape of the residual e is convex downward (step SP3: No), the process returns to step SP1 again, and the residual calculation function unit 30 ( Recalculate the residual e according to equation 2). On the other hand, if the plot of the residual e becomes convex upward, if the residual plot shape discriminating function unit 40 discriminates (step SP3: Yes), the process proceeds to step SP4, and the residual calculation function unit 30 determines the residual. The floor function of the minimum value emin of e is obtained, and the process proceeds to the next step SP5.

  In step SP5, the residual calculation function unit 30 subtracts the floor function of the minimum value emin of the residual e from the residual e by (Equation 3), calculates the adjusted residual Oe, and then proceeds to the next step SP6. Move.

  In step SP6, the setting function unit 50 divides the adjustment residual Oe into two regions, monotonically increasing and monotonically decreasing, with the maximum value (inflection point) as a boundary, and the monotonically increasing influencing factor A and the monotonically decreasing influencing factor B. Can be expressed by presence probabilities PAt and PBt based on (Equation 4), and a normal distribution as shown in (Equation 4) is set as a probability distribution candidate of the existence probabilities PAt and PBt, and the next step SP7 is performed. Move.

  In step SP7, the setting function unit 50 defines a function for deriving a relative ratio (influence ratio Rjt) between the existence probability PAt and the existence probability PBt in the usage years t according to (Equation 5), and proceeds to the next step SP8.

  In step SP8, the setting function unit 50 utilizes the information entropy theory and uses the influence ratio Rjt of the existence probability Pjt in (Equation 5) and the adjustment coefficient α to obtain the adjustment information entropy CH (t) as in (Equation 7). The function is set, and the process proceeds to the next step SP9.

  In step SP9, the parameter calculation function unit 60 determines various parameters (averages) that define a probability distribution (normal distribution) that satisfies the adjustment information entropy CH (t) of (Expression 7) that best matches the plot of the adjustment residual Oe. (μA, standard deviation σA, average μB, standard deviation σB) and the optimum value of the adjustment coefficient α are calculated, and the process proceeds to the next step SP10.

  In step SP10, the parameter etc. calculation unit 60 uses the various parameters (average μA, standard deviation σA, average μB, standard deviation σB) calculated in step SP9 to calculate the existence probabilities PAt and PBt of the influence factors A and B. Determine and move to next step SP11.

  In step SP11, since the parameter calculation function unit 60 has determined the existence probabilities PAt and PBt, the usage period t (33 years in this case) that is the intersection of the probability density functions PDF-A and PDF-B is determined as the influence factor A. And the influence of the influence factor B on the failure is calculated to be at the time of antagonism (replacement), and the process proceeds to the next step SP12.

  In step SP12, the parameter calculation function unit 60 calculates the service life t (33 years in this case) that is the intersection of the probability density functions PDF-A and PDF-B in step SP11. A function of the influence ratio Rjt is determined based on the probabilities PAt and PBt, and the process proceeds to the next step SP13.

  In step SP13, the entire calculation process / result recording function unit 70 records all the calculation processes and results from step SP1 to step SP12 in the data storage unit 100, and proceeds to the next step SP14.

  In step SP14, the display / output function unit 80 displays the probability density functions PDF-A and PDF-B of the existence probabilities A and B (FIG. 6), and if necessary, the influence ratio RAt of the existence probability PAt and the existence probability PBt. The influence ratio RBt is displayed, and the series of processes is terminated (FIG. 7).

<Second Embodiment>
<Configuration of influence factor information acquisition device>
As shown in FIG. 11 in which the same reference numerals are assigned to the corresponding parts in FIG. 1, the influence factor information acquisition device 200 is equipment data that is the basis of calculation for acquiring influence factor information in failure analysis in the calculation function unit 110, Data storage unit 100 storing equipment specification data, equipment inspection data, etc., calculation function unit 110, all calculation processes calculated by calculation function unit 110, and all calculation processes / results recorded in data storage unit 100 The recording function unit 70 and the display / output function unit 80 that displays or outputs the influence factor information obtained by the calculation function unit 110 are configured.

  The calculation function unit 110 is the same as the calculation function unit 10 in the first embodiment except that the analysis range dividing function unit 90 is provided between the residual plot shape determination function unit 40 and the setting function unit 50. It has the same configuration.

  As shown in FIG. 12, the residual plot shape discriminating function unit 40 may have a plurality of inflection points in the plot of the residual e depending on the result of the Weibull analysis performed by the analysis execution function unit 20 or the like. In this case, the shape of whether the residual e plot is convex upward or downward is determined, and the determination result is output to the analysis range dividing function unit 90.

  When the plot of the residual e has a plurality of inflection points, the analysis range dividing function unit 90 performs analysis by dividing the analysis range of the residual e. Here, the dividing point may be the service life t near the inflection point or the service life t near the middle of the inflection point. However, each division range must include one inflection point. In this case, it is divided into a divided range t1 having an upwardly convex inflection point and a divided range t2 having a downwardly convex inflection point.

  Then, the analysis range dividing function unit 90 makes the plot of the residual e convex upward for each of the divided ranges t1 and t2 according to (Expression 1) or (Expression 2) described above. Analysis is performed in the same manner as in the embodiment. However, at this time, the probability density function indicating the existence probability of the influence factors that straddle the adjacent divided ranges t1 and t2 is analyzed using the same parameters as a matter of course.

  Then, the parameter calculation function unit 60, in each divided range, the existence probability Pjt (normal distribution) when the calculated value based on the adjustment information entropy CH (t) in (Equation 7) and the plot of the adjustment residual Oe best match. ) Are determined by the method of least squares. FIG. 13 shows the existence probabilities Pjt (j = A, B, C in a total of three influence factors A, B, C when dividing into two regions of division ranges t1, t2, that is, one influence factor in common. ) Probability density functions PDF3, PDF4, and PDF5.

  As a result, the parameter calculation function unit 60 can calculate the intersection of the probability density functions PDF3 and PDF4, thereby obtaining a point where the influences of the influence factors A and B antagonize (that is, change) (year of use t). At the same time, by calculating the intersection of the probability density functions PDF4 and PDF5, it is possible to obtain a point (year of use t) at which the influences of the influence factors B and C antagonize (that is, change).

<Operation of influence factor information acquisition device>
The operation of the influence factor information acquisition apparatus 200 having such a configuration will be described with reference to the flowchart of FIG. In this case, the processing from step SP1 to step SP14 is common, and in step SP2, the residual plot shape discrimination function unit 40 discriminates that the plot has two or more inflection points. Move (step SP2: Yes).

  In step SP15, the analysis range dividing function unit 90 includes a divided range t1 having an upward inflection point and a downward inflection point so that each division range always includes one inflection point. After dividing into the range t2, it returns to step SP1 again and the process after step SP2 may be performed.

<Other embodiments>
In the above-described embodiment, the case where the existence probabilities PAt and PBt of the influencing factors A and B are assumed to follow a normal distribution has been described. However, the present invention is not limited to this, and various other types are possible. You may make it follow a probability distribution. For example, various probability distribution functions are stored in a server or the like, and these are combined and calculated by brute force using the least squares method or the maximum likelihood method, and the probability distribution that best matches the adjustment residual Oe is adopted. It may be. In addition, probability distribution function candidates may be displayed, and an analyst may select which probability distribution is used for calculation.

  DESCRIPTION OF SYMBOLS 1,200 ... Influence factor information acquisition apparatus 10, 110 ... Calculation function part, 20 ... Analysis execution function part, 30 ... Residual calculation function part, 40 ... Setting function part, 50 ... Parameter calculation function part, 60 ... Previous Calculation process / result recording function unit, 70 ... display / output function unit, 80 ... residual plot shape discrimination function unit, 90 ... analysis range division function unit, 100 ... data storage unit.

Claims (5)

A probability density function defining part for defining a probability density function representing the existence probability of a plurality of influence factors causing a failure; a relative ratio defining part for defining a relative ratio of the existence probabilities of the plurality of influence factors; and a probability density function Executed by an influence factor information acquisition apparatus comprising: a parameter calculation unit that calculates a parameter of the calculation unit; and an influence factor information calculation unit that calculates the existence probability of the plurality of influence factors as influence factor information in failure analysis using the parameters. Influencing factor information acquisition method in failure analysis,
If the plot between the frequency of occurrence of the fault and use period having at least one inflection point, the probability density function defined portion, corresponding to each period of use before and after the inflection point of the plot A probability density function defining step defining a probability density function representing the probability of existence of a plurality of influencing factors causing the failure;
The relative proportions specified section, the relative proportions specified step of defining the relative proportions of the existence probability of the plurality of influencing factors in the use period including before and after the inflection point and the inflection point,
The parameter calculation unit sets an information entropy function based on a relative ratio of the existence probabilities of the plurality of influencing factors, and sets a parameter of the probability density function that satisfies the information entropy function based on the plot information. A parameter calculating step to calculate,
The influence factor information in failure analysis, wherein the influence factor information calculation unit has an influence factor information calculation step of calculating the existence probability of the plurality of influence factors as influence factor information in failure analysis using the parameter Acquisition method. In the influence factor information acquisition method in the failure analysis according to claim 1,
In the probability density function defining step, the existence probability is defined using a residual between the plot information and the value of the regression line corresponding to the plot information. In the influence factor information acquisition method in the failure analysis according to claim 2,
In the influence factor information calculation step, the inflection point is specified based on a plurality of probability density functions respectively corresponding to the existence probabilities of the plurality of influence factors. In the influence factor information acquisition method in the failure analysis according to claim 3,
The analysis range dividing step of dividing the analysis range for the information of the plot by the analysis range dividing unit so that one inflection point is always included when two or more inflection points exist. Influencing factor information acquisition method in failure analysis. A plurality of influencing factors that cause the failure corresponding to respective use periods before and after the inflection point of the plot when the plot between the use period and the frequency of occurrence of the failure has at least one or more inflection points; A probability density function defining part for defining a probability density function representing the existence probability of
A relative proportion defining section for defining a relative proportion of the existence probability of the plurality of influencing factors in the period of use including the inflection point and before and after the inflection point;
A parameter calculation unit that sets an information entropy function based on a relative ratio of the existence probabilities of the plurality of influencing factors, and calculates a parameter of the probability density function that satisfies the information entropy function based on the information of the plot; ,
An influence factor information acquisition device in failure analysis, comprising: an influence factor information calculation unit that calculates the existence probability of the plurality of influence factors as influence factor information in failure analysis using the parameter.

Images