JP2018032229A - Equipment inspection order setting device and equipment inspection order setting method - Google Patents

Abstract

An object of the present invention is to extract a sign highly relevant to a failure when a plurality of signs occur, and to set the priority of equipment for performing maintenance and inspection based on the extracted signs with high accuracy.
A sign pattern extraction for extracting a sign related to a failure from a sign generated before a failure based on a sign and a failure accumulated in a failure history database and creating a sign pattern based on the extracted sign 11, a transition probability accumulation unit 12 that calculates and accumulates transition probabilities between predictors and predictor failures in the predictor pattern, and when the first predictor in the predictor pattern occurs, the predictor changes from the predictor to the failure based on the transition probability A cumulative failure probability calculation unit 13 that calculates the cumulative failure probability to reach, and a priority calculation unit 14 that calculates the priority of the facility 2 that performs maintenance inspection based on the cumulative failure probability with respect to the plurality of facilities 2 in which a sign has occurred It is characterized by providing.
[Selection] Figure 1

Description

  The present invention relates to an equipment inspection order setting device and an equipment inspection order setting method, and more particularly, to an equipment inspection order setting apparatus for setting a priority order of equipment for performing maintenance and inspection based on a failure history collected from equipment such as a cooling facility, and the like. The present invention relates to an equipment inspection order setting method.

  Conventionally, there is a remote monitoring service for monitoring a plurality of facilities such as a cooling / heating facility from a monitoring center. In this remote monitoring service, equipment status data indicating the operating status of each equipment such as the temperature and pressure of the equipment in operation is collected from each equipment. When an abnormality in the facility is detected by analyzing the collected facility state data, the maintenance staff performs an inspection on the facility in which the abnormality is detected according to the created inspection schedule.

  In recent years, there is a case where an inspection schedule is created so as to prevent the occurrence of an abnormality in advance by making an inspection target a facility where a failure sign is generated even if no abnormality is detected. In order to prevent the occurrence of anomalies more reliably, an inspection schedule is created to prioritize equipment that is predicted to fail early, in other words, equipment that has a high probability of failure. It is preferable to reduce the occurrence of failures.

  For example, in Patent Document 1, failure history is clustered according to a predetermined condition, a failure probability distribution is created for each clustered failure history group (hereinafter referred to as “failure group”), and any failure group is created. There has been proposed a technique for calculating the priority for performing maintenance and inspection of the equipment based on the failure probability distribution created for the failure group. Further, in Patent Document 1, a limit time (a dispatch limit time) at which a facility that further requires a maintenance worker to fail is calculated, and a maintenance staff is dispatched based on the dispatch limit time, thereby preventing a failure of the facility. Techniques for further reduction have also been proposed.

  In Patent Document 2, the state of the equipment after a predetermined period is simulated based on the DOMDD model (partially observable Markov decision process) using information such as equipment maintenance inspection information, equipment warnings and various measured values. Thus, a technique for determining the necessity for maintenance and inspection of equipment has been proposed.

  Patent Document 3 proposes a technique for automatically creating, adding, and updating an abnormality diagnosis model for diagnosing and responding to an abnormality of equipment using a graph network structure. In particular, the causal relationship of information such as maintenance work data, alarms, and operation events is expressed using a graph network structure.

  Patent Document 4 has a failure risk table showing a correspondence relationship between the elapsed time from the time when the failure sign state is detected and the magnitude of the failure risk, and within a predetermined period from the time when the failure sign state is detected. In addition, a technique has been proposed in which the magnitude of a failure risk for a device failure is determined based on a failure risk table.

JP 2014-167667 A JP 2014-229001 A JP 2010-122847 A JP 2010-091840 A

  When a failure occurs in a facility, the user cannot use the facility, so it is important to prevent the failure in advance. Therefore, it is necessary to preferentially check facilities that fail with a high probability at an early stage with respect to facilities in which a sign has occurred, and reduce the failures.

  For this reason, in Patent Documents 1-4, the probability of failure is calculated based on the occurrence of a sign, but when multiple signs occur in time series, which of these signs is related to the failure. This is not taken into account. Since there are several signs that are highly related to a failure and signs that are not related to a failure, even if the probability of failure is calculated based on a sign that is not related to a failure, the accuracy of the failure probability is high. It will be lower.

  Therefore, in the present invention, when a plurality of signs occur, a sign highly relevant to the failure is extracted, and the priority of the equipment that performs maintenance inspection based on the extracted sign is set with high accuracy. Objective.

  The equipment inspection order setting device according to the present invention includes a failure history storage unit that stores information on signs collected from a plurality of facilities and information on failures that have occurred after the signs, and the failure history storage unit stores the failure history storage unit. A sign pattern extraction unit that extracts a sign related to the failure from a plurality of the signs generated before the failure based on the sign and the failure, and creates a sign pattern based on the extracted sign; A transition probability accumulation unit that calculates and accumulates transition probabilities between predictors and predictor failures in the predictor pattern, and when the first predictor in the predictor pattern occurs, the failure from the predictor based on the transition probability A cumulative failure probability calculation unit that calculates a cumulative failure probability to reach a plurality of the facilities in which the sign has occurred before performing a maintenance check based on the cumulative failure probability Characterized in that it and a priority calculating unit for calculating a priority of the equipment.

  Further, the predictor pattern extraction unit compares the change rate of the failure probability between the predictors with a threshold value for determining the relevance with the failure, thereby calculating the predictor associated with the failure from a plurality of the predictors. It is characterized by extracting.

  In addition, the predictive pattern extracting unit creates the predictive pattern by using the same type of the predictor as one predictor when the same type of the predictor is continuously present in the predictor pattern. To do.

  The cumulative failure probability calculation unit calculates the cumulative failure probability by Monte Carlo simulation, and sets an upper limit number of times of the Monte Carlo simulation according to the number of indications in the indication pattern.

  In addition, based on the priority calculated by the priority calculation unit, a dispatch limit time calculation unit is provided that calculates a dispatch limit time that is the limit time for which the facility fails unless the maintenance staff is dispatched. It is characterized by.

  Furthermore, the equipment inspection order setting method of the present invention is an equipment inspection order setting method for setting a priority order of the equipment for performing maintenance inspection based on failure histories collected from a plurality of equipment, The information on the signs collected from each of the information and the information on the failure that occurred after the signs are accumulated, and based on the signs and the failure, the signs related to the failure from the plurality of signs that occurred before the failure. When the first predictor in the predictor pattern is generated by creating a predictor pattern based on the extracted predictor, calculating and storing transition probabilities between predictors and predictor failures in the predictor pattern Based on the transition probability, a cumulative failure probability from the predictor to the failure is calculated, and a plurality of the facilities where the predictor has occurred are calculated based on the cumulative failure probability. And calculates the priority of the facility to perform maintenance and are.

  According to the present invention, when a plurality of signs are generated, a sign highly relevant to a failure is extracted, and the priority of equipment for performing maintenance inspection is set with high accuracy based on the extracted sign. be able to.

It is a block block diagram of the refrigeration equipment system containing an equipment inspection order setting device. It is the figure which showed the data structural example of the failure history information which shows the failure history accumulate | stored in a failure history database. It is a hardware block diagram of the computer which comprises the equipment inspection order setting apparatus. It is a schematic diagram of the sign leading to a failure. It is a flowchart which shows the process of an equipment inspection order setting apparatus. It is a schematic diagram which shows the some predictive pattern regarding the predictive leading to a failure. FIG. 7 is a schematic diagram of predictive patterns respectively showing a plurality of predictive patterns shown in FIG. 6. It is a schematic diagram of the sign explaining the case where the same kind of sign is thinned out. It is a figure explaining the transition state from a sign occurrence to a failure, (a) shows a continuous sign relationship when a sign occurrence occurs to a failure, and (b) calculates a transition probability over time in the continuous sign relationship. Shows the table. It is a figure which shows the path | route which leads to a failure when a precursor generate | occur | produces, (a) shows the path | route from the precursor Y2 to the failure K2, (b) shows the path | route from the precursor Y3 to the failure K2, (c) is A path from the predictor Y4 to the failure K2 is shown. It is the flowchart which showed the process which calculates a cumulative failure probability by Monte Carlo simulation. It is the flowchart which showed the process which calculates a cumulative failure probability by Monte Carlo simulation.

  FIG. 1 is a block configuration diagram showing an embodiment of an equipment inspection order setting device according to the present invention. FIG. 1 shows a cooling facility 2, an operation history receiver 3, a failure history database (DB) 4, and an equipment inspection order setting device 10. As shown in FIG. 1, the facility inspection order setting device 10 is connected to a plurality of refrigeration facilities 2 installed in a plurality of contracted facilities via an operation history receiving unit 3 and a failure history database 4. In the present embodiment, the cooling facility 2 is illustrated as a facility to be inspected, but other facilities that require inspection may be used.

  The operation history receiving unit 3, the failure history database 4, and the equipment inspection order setting device 10 are installed, for example, in a monitoring center. The operation history receiving unit 3 receives equipment state data indicating the equipment state from the cooling / heating equipment (hereinafter simply referred to as “facility”) 2, equipment state data when a sign or abnormality occurs, and the received equipment state data The fault history information is generated and accumulated in the fault history database 4 by, for example, summing up as necessary. The operation history receiving unit 3 is configured by a computer having a communication function, and the failure history database 4 is configured by a database server. As a method of receiving equipment state data and storing it in the failure history database 4, a method of transmitting online via a telephone line and an Internet line by means of communication software installed in a control device (not shown) of the equipment 2, etc. There is.

  FIG. 2 is a diagram illustrating a data configuration example of failure history information indicating a failure history stored in the failure history database 4 according to the present embodiment. The failure history information is configured by associating a failure time interval, installation environment information, and occurrence state information with a sign occurrence date and time indicating a date and time when a sign of failure occurred. The sign occurrence date and time is the date and time when the occurrence of the sign is detected by the operation history receiving unit 3 referring to the equipment state data transmitted from the equipment 2. The failure time interval is a time interval from the occurrence of the sign until the equipment 2 breaks down, and is a value indicating the difference between the date and time of occurrence of the abnormality and the date and time of occurrence of the sign. In this embodiment, the time is shown, but it may be time. Here, “abnormal” refers to a case where one piece of equipment state data deviates from a normal state indicating a value within a normal range. “Failure” means that the equipment 2 is stopped due to the occurrence of abnormality in the equipment status data. Further, the “sign” means a case where the threshold value indicating that the equipment state data is close to an abnormal state even if the equipment state data is within a normal range.

  The installation environment information is information related to the specifications and installation location of the equipment 2, the equipment number that identifies the equipment 2, the building number that identifies the building in which the equipment 2 is installed, the installation destination that indicates the address where the equipment 2 is installed, It includes various information related to the installation start indicating the year of installation, the model name of the facility 2 and the installation direction indicating the direction in which the facility 2 is installed in the building.

  Occurrence status information is information related to the occurrence status of signs and abnormalities. Predictive contents indicating the contents of the generated signs, temperature and humidity at the time of occurrence of the sign, temperature and set temperature at the time of occurrence of the sign The temperature difference between the start and stop counts indicating the number of times the thermostat of the equipment 2 has been turned on / off from the start of the count until the occurrence of the predictor, and the facility between the start of the count and the occurrence of the predictor Operating time when the time when 2 is in operation is integrated, Thermo ON time when the time when the thermostat of the facility 2 is ON is accumulated from the start of counting until the occurrence of a sign, per operating time The ratio of the number of starts / stops indicating the number of starts / stops, the ratio of the thermo-ON time indicating the thermo-ON time per operation time, and the room in the facility 2 between the start of the count and the occurrence of the sign Degrees is the target temperature reaching time indicating the average time to reach the set temperature, including various types of information about the abnormal content indicating an abnormality occurrence time and generated abnormality contents indicating the date and time when the equipment 2 is stopped by the occurrence of abnormality. FIG. 2 shows a part of the above described occurrence status information.

  FIG. 3 is a hardware configuration diagram of a computer forming the equipment inspection order setting device 10 in the present embodiment. In this embodiment, the computer forming the equipment inspection order setting device 10 can be realized by a hardware configuration of a general-purpose personal computer (PC) that has existed. That is, as shown in FIG. 3, the computer is provided as a CPU 21, ROM 22, RAM 23, HDD controller 25 to which a hard disk drive (HDD) 24 is connected, a mouse 26 and keyboard 27 provided as input means, and a display device. An input / output controller 29 for connecting each display 28 and a network controller 30 provided as a communication means are connected to an internal bus 31.

  FIG. 4 shows the signs Y1, Y2, Y3, Y4 in the time series occurring between the failure K1 and the failure K2. That is, the signs Y1, Y2, Y3, Y4 leading to the failure K2 are shown. In FIG. 4, the outline content of setting the priority order of equipment inspection by the equipment inspection order setting device 10 will be described, and then each configuration of the equipment inspection order setting device 10 will be described. In FIG. 4, after the occurrence of the failure K1, the signs Y1, Y2, Y3, and Y4 occur, and then the failure K2 occurs. Of the signs Y1, Y2, Y3, and Y4, only the sign type of the sign Y3 is the B sign, and the other sign types of the signs Y1, Y2, and Y4 are the A sign.

  The equipment inspection order setting device 10 extracts a sign highly relevant to the failure K2 out of the signs Y1, Y2, Y3, Y4 leading to the failure K2 based on the failure history data, and based on the extracted sign Pattern D is extracted. Then, the equipment inspection order setting device 10 calculates the cumulative failure probability that leads to the failure K2 when the first predictor Y2 in the predictive pattern D occurs in the equipment operating state, and based on this cumulative failure probability, The priority order of the facility 2 that performs maintenance inspection on the facility 2 is set.

  Returning to FIG. 1, the equipment inspection order setting device 10 includes a predictive pattern extraction unit 11, a transition probability accumulation unit 12, a cumulative failure probability calculation unit 13, a priority calculation unit 14, and a dispatch limit time calculation unit 15. The sign pattern extraction unit 11 extracts a sign that is highly related to a failure from among a plurality of signs leading to a failure. As shown in FIG. 4, for example, a failure K1 occurs, and after the failure K1 is restored, a plurality of Y1, Y2, Y3, and Y4 occur while the next failure K2 occurs. The failure K2 may occur due to, for example, a single sign Y4, but usually occurs frequently when a plurality of signs Y1, Y2, Y3, and Y4 continue. Even if it is said that a plurality of signs Y1, Y2, Y3, Y4 are consecutive, these signs Y1, Y2, Y3, Y4 are separated in time series from the occurrence of a sign Y1 having a low relevance to the occurrence of a failure or the occurrence of a failure K2. Are also included (not shown). For this reason, the predictive pattern extracting unit 11 is closely related to, for example, the signs Y2, Y3, and Y4 that are highly related to the failure K2, among the multiple signs Y1, Y2, Y3, and Y4 leading to the failure K2, that is, the occurrence of the failure. Related signs Y2, Y3, Y4 are extracted. The sign pattern extraction unit 11 extracts signs Y2, Y3, and Y4 that are closely related to the occurrence of the failure, excludes the sign Y1 that is not highly related to the occurrence of the failure, and indicates signs Y2, Y3, and Y4 leading to the failure K2. A predictive pattern D consisting of The predictive pattern D is a collection of one or a plurality of predictors (predictors Y2, Y3, Y4 in FIG. 4) immediately before the failure. When a plurality of predictors are involved, the predictor Y2, which is likely to reach the failure K2. The generation order of Y3 and Y4 is also defined. In FIG. 4, the sign pattern D is defined in the order of the signs Y2, Y3, and Y4.

  The transition probability accumulation unit 12 calculates the transition probability between each predictor in the extracted predictor pattern D and accumulates the transition probability of the predictor pattern D. The transition probability is a probability at the time of transition from a certain sign to the next sign or failure. The transition probability accumulating unit 12 calculates transition probabilities over time between predictors and predictor failures, and stores the calculation results. For example, as shown in FIG. 9A, in the predictive pattern D3 including the predictors Y2, Y3, and Y4, when the predictor Y2 leads to the failure K2, a portion where the predictor Y2 and the failure K2 are continuous (continuous portion L1), The part where the sign Y2 and the sign Y3 continue (continuous part L2), the part where the sign Y3 and the failure K2 continue (continuous part L3), the part where the sign Y3 and the sign Y4 continue (continuous part L4), the sign Y4 And a failure between the failure K2 (continuous portion L5). The transition probability accumulation unit 12 calculates transition probabilities for the continuous portions L1 to L5, respectively, and accumulates the transition probabilities. That is, in order to calculate the cumulative failure probability leading to the failure K2 in the cumulative failure probability calculation unit 13, the transition probability of the predictive pattern D is calculated and accumulated in advance and stored as a table.

  When the first sign in the sign pattern D occurs, the cumulative failure probability calculation unit 13 calculates the cumulative failure probability from the first sign to the failure based on the transition probability calculated by the transition probability accumulation unit 12. For example, as shown in FIG. 9A, when the first sign Y2 in the sign pattern D3 occurs, the cumulative failure probability that leads to the failure K2 is calculated. Since the signs Y2, Y3, Y4 do not always occur in the order of signs of the sign pattern D3 during actual operation of the equipment, the cumulative failure probability that leads to the failure K2 is calculated for the sign pattern D3.

  The priority calculation unit 14 calculates the priority for performing maintenance and inspection on each facility where a sign has occurred based on the cumulative failure probability calculated by the cumulative failure probability calculation unit 13. The dispatch limit time calculation unit 15 calculates the dispatch limit time based on the priority calculated by the priority calculation unit 14. The maintenance staff stands by at a branch office or the like, and heads for the facility 2 as a maintenance inspection destination if necessary. Basically, when a sign of a failure occurs, it is preferable to perform a maintenance check before the time when the failure is predicted to occur in order to prevent the occurrence of the failure in the facility 2 where the sign has occurred. . Here, the dispatch limit time is a limit time when the facility 2 fails unless the maintenance staff is dispatched.

  In addition, about the specific process in the predictive pattern extraction part 11, the transition probability accumulation | storage part 12, the cumulative failure probability calculation part 13, the priority calculation part 14, and the dispatch limit time calculation part 15, it mentions the below-mentioned equipment inspection order setting apparatus 10 This will be described in detail in the operation description.

  Further, the program for executing the equipment inspection order setting method according to the present embodiment can be provided not only by communication means but also by being stored in a computer-readable recording medium such as a CD-ROM or DVD-ROM. It is. The program provided from the communication means or the recording medium is installed in the computer, and the equipment inspection order setting method is realized by the CPU of the computer sequentially executing the program.

  Next, the operation of the equipment inspection order setting device 10 will be described. While the equipment 2 is in operation, equipment status data including information necessary for generating failure history information is transmitted from the equipment 2 in which a sign or failure has occurred to the operation history receiving unit 3. When the operation history receiving unit 3 detects the occurrence of the sign by analyzing the received equipment state data, the sign occurrence date and time, the installation environment information, the sign content included in the occurrence state information, temperature, humidity, Failure history information is generated by setting a temperature difference, start / stop count, operation time, thermo-ON time, start / stop count ratio, thermo-ON time ratio, and target temperature arrival time. In addition, when the operation history receiving unit 3 detects the occurrence of an abnormality by analyzing the received equipment state data, the abnormality occurs with reference to the installation environment information from the failure history information generated by the occurrence of the sign. The failure history information corresponding to the facility 2 is identified, and the date and time of occurrence of the abnormality and the details of the abnormality are set and registered based on the received equipment state data. Further, the operation history receiving unit 3 calculates a time interval from the occurrence of the sign to the occurrence of the abnormality, and sets and registers the failure time interval. In this way, the failure history information is generated by the operation history receiving unit 3 with the occurrence of the sign and failure, and the failure history information shown in FIG.

  Then, the process of the equipment inspection order setting apparatus 10 is demonstrated with reference to the flowchart shown in FIG. In step S <b> 101, the predictive pattern extraction unit 11 extracts a predictor that is highly related to a failure from a plurality of predictors leading to a failure, and creates a predictive pattern based on the extracted predictor. The sign extraction and the sign pattern creation will be described with reference to FIGS. In FIG. 6, as in FIG. 4, after the occurrence of the failure K1, the signs Y1, Y2, Y3, Y4 occur, and then the failure K2 is reached. In FIG. 6, a symbol T indicates a time axis. Of the signs Y1, Y2, Y3, and Y4, only the sign type of the sign Y3 is the B sign, and the other sign types of the signs Y1, Y2, and Y4 are the A sign.

  First, creation of the predictive pattern D will be described. As shown in FIG. 6, it is determined using the rate of change in the failure probability when the number of predictors is changed, which predictor Y1, Y2, Y3, Y4 is closely related to the failure K2, retroactively from the failure K2. To do. In other words, whether the cause of the failure K2 is only the predictor Y4 (predictor pattern D1), the predictor Y3, Y4 (predictor pattern D2), or the predictor Y2, Y3, Y4 (predictor pattern D3), Alternatively, it is determined whether the signs are Y1, Y2, Y3, Y4 (predictive pattern D4).

  Which one of the predictive patterns D1 to D4 is selected depends on the following concept. For example, when the predictive pattern D1 and the predictive pattern D2 are compared, if the change rate of the failure probability is small, the change in the failure tendency due to the increase in the number of predictors is small. It is judged that the number is low, and the predictive pattern D1 having a small number of predictors is adopted. In addition, when the predictive pattern D1 and the predictive pattern D2 are compared, if the rate of change in failure probability is large, the change in failure tendency due to an increase in the number of predictors is large. Judging that it becomes high, the sign pattern D2 having a large number of signs is adopted.

  Note that if the number of predictors in the predictive pattern is increased without limitation because it is determined that the relevance with the failure is high, the probability of failure is reduced due to the increase of the predictor number, and the data amount of the predictor pattern also increases. Therefore, it is preferable that the number of signs in the sign pattern is as small as possible. In order to estimate the occurrence of a failure, it is preferable that there are many types of predictive patterns. In addition, when there is a predetermined time interval between the signs, that is, when the signs are separated in time series, it is determined that the sign that is apart from the occurrence of the failure is unlikely to be related to the failure, and the sign Exclude from pattern.

Next, a specific method for determining the predictive pattern will be described. FIG. 7 shows separately the predictive patterns D1, D2, D3, and D4 leading to the failure K2 shown in FIG. Here, as indicated by reference symbol F1, when the predictor Y4 occurs and reaches the failure K2 (predictor pattern D1), the failure probability P ₁ is 0.2, and as indicated by reference symbol F2, the predictor Y4 follows the predictor Y3. There failure probability P ₂ 0.35 when reaching the fault occurred K2 (predictor pattern D2), as shown at F3, when following the sign Y2 reaches the sign Y3, Y4 occurs failure K2 ( predictor pattern D3) of the failure probability P ₃ 0.23, as indicated by reference numeral F4, the failure probability P when leading to failure following the sign Y1 is sign Y2, Y3, Y4 occurs K2 (predictor pattern D4) ₄ is 0.2.

The change rate PC _x of the failure probability P _x can be calculated by the following formula 1.
Rate of change PC _x = (P _x −P _{x + 1} ) × 100 / P _x ...
Further, a threshold value for determining the relevance with the failure is set in advance based on the change rate PC _x of P _x of the predictive pattern D. When the change rate PC _x of P _x of the predictive pattern D exceeds the threshold value, it is determined that the relevance with the failure is high. Therefore, if the threshold value is set small, the level for determining the relevance with the failure decreases, the number of predictors in the predictive pattern D increases. Conversely, if the threshold value is set large, the level for determining the relevance with the failure increases, The number of signs in the sign pattern D decreases. Here, the threshold is set to 30%, for example. Then, the change rate PC _x of the failure probability P _x of predictor pattern D is, until the threshold value 30% or less, comparing the change rate PC _x of the failure probability P _x of predictor pattern D.

When the rate of change PC _x of the failure probability P _x of the predictor patterns D1, D2, D3, and D4 is calculated based on Equation 1, the predictor pattern D1 and the predictor pattern D2 are:
The rate of change PC _x (D1 · D2) = (0.2−0.35) × 100 / 0.2 = 75%, which is higher than the threshold 30% compared to the threshold 30%. adopt.

Similarly, in the predictive pattern D2 and the predictive pattern D3,
The rate of change PC _x (D2 · D3) = (0.35-0.23) × 100 / 0.35 = 34%, which is higher than the threshold value of 30%, which is also an indication in this case. The pattern D3 is adopted.

Similarly, in the predictive pattern D3 and the predictive pattern D4,
Change rate PC _x (D3 · D4) = (0.23-0.2) × 100 / 0.23 = 13%, which is 30% or less of the threshold value compared to the threshold value of 30%. The sign pattern D3 with a small amount is employed. Thus, by comparing the rate of change PC _x of the failure probability P _x respect predictor pattern D1, D2, D3, D4, employing a sign pattern D3 as predictor pattern D leading to failure K2.

  Further, in FIG. 8, after the same type (A predictor) of the signs Y10, Y11, Y12 are continuously generated, the B predictor Y13 is generated, and thereafter, the A predictor Y14 is generated, resulting in the failure K2. Show the case. As shown in FIG. 8, when the same type (A predictor) of the signs Y10, Y11, Y12 occurs continuously, the same type of sign is regarded as one predictor, and the same type of predictor is thinned out. Pattern D is created. That is, the same type of predictor often occurs continuously (for example, in FIG. 8, A predictor is continuously generated), and when the predictor pattern D including all these same types of predictors is created, Since many similar predictive patterns D are created, when the same type of predictor is generated continuously, the predictor pattern D is generated by regarding the same type of predictor as one predictor.

  Returning to the flowchart of FIG. 5, in step S101, after the predictive pattern D3 is created, the process proceeds to step S102. In step S102, the transition probabilities are calculated for each predictor in the predictor pattern D3, and the transition probabilities over time between predictor failures are accumulated. Regarding the predictive pattern D3, as shown in FIG. 9A, there are five continuous portions L1, L2, L3, L4, and L5 in the path from the predictor Y2 to the failure K2. The transition probability accumulation unit 12 calculates transition probabilities in the passage of time between the signs Y2, Y3, Y4 in the continuous portions L1, L2, L3, L4, L5, and between the signs Y2, Y3, Y4 and the failure K2. Accumulate the calculated transition probability. FIG. 9B shows a table in which transition probabilities in the passage of time for the continuous portions L1, L2, L3, L4, and L5 are calculated. The calculated transition probability table is stored in the RAM 23 or the HDD 24. After calculating the transition probability, the process proceeds to step S103.

  In step S103, the cumulative failure probability of the predictive pattern D3 is calculated. A known Monte Carlo method is used to calculate the cumulative failure probability of the predictive pattern D3 in step S103. The Monte Carlo method is a method of performing numerical calculation using random numbers to obtain an approximate solution of a required solution. In this embodiment, the metropolis method is used among the Monte Carlo methods.

  The predictive pattern D3 in FIG. 10A will be described as an example. As shown in FIG. 10A, there are three paths G1, G2, and G3 when the sign Y2 occurs and reaches the failure K2. That is, when the sign Y2 occurs, the path G1 directly from the sign Y2 to the fault K2, the path G2 from the sign Y2 via the sign Y3 to the fault K2, the fault from the sign Y2 via the signs Y3, Y4 There are three routes G3 to K2. By using the Monte Carlo method for these three paths G1, G2, and G3, one cumulative failure probability is calculated.

  Further, as shown in FIG. 10B, when the sign Y3 occurs, that is, the path G4 that directly leads to the failure K2 after the sign Y3 occurs, and the path that leads from the sign Y3 to the fault K2 via the sign Y4. Similarly, the two types of G5 are calculated as one cumulative failure probability using the Monte Carlo method. As shown in FIG. 10C, when the sign Y4 occurs, there is only one route G6 that leads to the failure K2, so there is no need to use the Monte Carlo method.

  11 and 12 show flowcharts of processing by Monte Carlo simulation. First, the outline of the Monte Carlo simulation will be described. In the Monte Carlo simulation, (1) the sign that occurs at the beginning of the sign pattern D is set as an initial value x0, and (2) the current state is xt, and the next state candidate, that is, the next sign or failure that occurs is x (3) calculate the ratio r = P (x ′) / P (xt) between the probability of the next state candidate and the probability of the current state, and (4) generate a uniform random number R, (5) If R <r, the next state candidate (next sign or failure) is set to the next state, and if R <r, the current state is set to the next state. (6) (2) to (2) 5) is repeated until time t. Further, the processes (1) to (6) are repeated until the distribution of cumulative failure probabilities converges. If it does not converge, this simulation is repeated up to a preset upper limit.

  In FIG. 11, in step S201, a sign generated at the beginning of the sign pattern D3 is set. That is, the sign Y2 of the sign pattern D3 is set as an initial value, and the process proceeds to step S202.

  In step S202, the initial value (predictor Y2) set in step S201 is selected as the current state, and the process proceeds to step S203.

  In step S203, a transition candidate that occurs next to the sign Y2 is selected. As shown in FIG. 10A, as a transition candidate that occurs after the sign Y2, the sign Y3 or the failure K2 can be considered. Therefore, the sign Y3 or the failure K2 is selected as a transition candidate, and the process proceeds to step S204.

  In step S204, the transition probability at time t (here, 1 hour) is calculated for the transition candidate selected in step S203, and the process proceeds to step S205. Based on the above-described P (x ′) / P (xt), the transition probability between the predictor Y3 and the failure K2 at time t is P (predictor Y3) / P (predictor Y2) = 0.2, P (failure K2 ) / P (predictor Y2) = 0.4. Although the transition probability of the predictor Y3 and the transition probability of the failure K2 may be calculated for each process of the flowchart, the table in FIG. 9B is used here. With reference to the table of FIG. 9B, the transition probability between the predictor Y3 and the failure K2 occurring within one hour from the occurrence of the predictor Y2 is 0.2 for the predictor Y3 and 0.4 for the failure K2. is there.

  In step S205, a uniform random number R is generated, and the process proceeds to step S206. This uniform random number R is used to estimate a transition candidate that occurs after the sign Y2. For example, the generated uniform random number R is 0.7.

  In step S206, the uniform random number R is compared with the transition probability of the sign Y3 calculated in step S204 and the transition probability of the failure K2, and the next state is determined. As the next state, as shown in FIG. 10A, there is a state of either maintaining the current state, transitioning to the next predictor, or transitioning to a failure. When maintaining, it progresses to step S207, when changing to the next precursor, it progresses to step S209, and when changing to a failure, it progresses to step S211.

  This time, the next state of the predictor Y2 is determined. When the uniform random number R is smaller than the transition probability of the predictor Y3, the transition is made to the predictor Y3. When the uniform random number R is smaller than the sum of the transition probability of the predictor Y3 and the transition probability of the fault K2, When the transition is made to the failure K2 and the uniform random number R exceeds the total value, the current sign Y2 is maintained.

  For example, judging from the above-mentioned numerical values, the uniform random number R (0), the transition probability (0.2) of the predictor Y3, and the transition probability (0.4) of the failure K2 are obtained. .7) exceeds the total value (0.2 + 0.4), and the process proceeds to step S207 to maintain the current sign Y2.

  In step S207, since there is no transition to either the predictor Y3 or the failure K2, the state of the predictor Y2 is maintained, and the process proceeds to step S208. In step S208, after counting up the time t, the time t is compared with the upper limit time N to determine whether the time t has reached the upper limit time N. That is, when the time t has not reached the upper limit time N (Yes), the process returns to Step S202, and Steps S202 and after are repeated. If the time t has reached the upper limit time N (No), the process proceeds to step S212 in FIG.

  Next, the case of returning from step S208 to step S202 will be described. When step S202 and subsequent steps are repeated, the transition candidate from the predictor Y2 becomes the predictor Y3 or the failure K2 as in the previous time. Since the time t is now 2 hours due to counting up, the transition probability of the predictor Y3 and the transition probability of the failure K2 are 0.3 for the predictor Y3 and 0 for the failure K2, referring to the table of FIG. 9B. .2. Then, as a result of generating the uniform random number R in step S205, for example, when the uniform random number R is 0.25, the uniform random number R (0.25), the transition probability (0.3) of the sign Y3, The transition probability (0.2) of the failure K2. Under this condition, step S206 corresponds to a case where the uniform random number R (0.25) is smaller than the transition probability (0.3) of the predictor Y3, and the process proceeds to step S209 to transition to the next predictor Y3 state. .

  In step S209, it is determined that transition to the next sign (predictor Y3) is made, and the process proceeds to step S210. In step S210, the next predictor (predictor Y3) is changed to the current state, that is, the current state is changed from the predictor Y2 to the predictor Y3, the process returns to step S202, and step S202 and subsequent steps are repeated. Returning to step S202, when step S202 and subsequent steps are repeated, the transition candidate from the predictor Y3 becomes the predictor Y4 or the failure K2, as shown in FIG. 10B. The transition probability of the predictor Y4 and the transition probability of the failure K2 are 0.5 for the predictor Y4 and 0.5 for the failure K2 from the table of FIG. 9B. As a result of generating the uniform random number R in step S205, for example, when the uniform random number R is 0.6, the uniform random number R (0.6), the transition probability (0.5) of the sign Y4, The transition probability (0.5) of the failure K2. Under this condition, in step S206, when the uniform random number R (0.6) is smaller than the sum of the transition probability (0.5) of the predictor Y4 and the transition probability (0.5) of the failure K2. Correspondingly, the process proceeds to step S211 and transitions to the state of the failure K2.

  In step S211, it is determined that the transition is to the failure K2, and the process proceeds to step S212. In step S212, one simulation is completed, the number of simulations is counted up, and the process proceeds to step S213.

  In step S213, it is determined whether the number of simulations has reached a preset upper limit number. If the upper limit number has been reached (Yes), the series of processes is terminated, and if the upper limit number has not been reached (No), the process proceeds to step S214.

  In step S214, a series of processing is repeated until the cumulative failure probability distribution converges. That is, in step S214, when the cumulative failure probability distribution has converged (Yes), a series of processing ends, and when the cumulative failure probability distribution has not converged (No), the process returns to step S201, and steps S201 and thereafter are performed. repeat.

  Here, the setting of the upper limit number of simulations will be described. As the number of predictors in the predictive pattern D increases, the number of branches leading to a failure increases. For this reason, the upper limit number of simulations is set according to the number of signs in the sign pattern D. For example, when the number of predictors in the predictor pattern D is 1, the upper limit number of simulations is set to 10,000, and when the number of predictors in the predictor pattern D is two, the upper limit number of simulations is set to 20000. Further, when the number of signs in the sign pattern D is three like the sign pattern D3, the upper limit number of simulations is set to 30000 times.

  Returning to the flowchart of FIG. 5, in step S <b> 104, the priority for performing maintenance and inspection on each facility where a sign has occurred is calculated based on the cumulative failure probability calculated in step S <b> 103. In other words, the equipment that requires the most maintenance inspection, that is, the equipment that has the highest probability of failure, is selected based on the cumulative failure probability among the equipments where the sign has occurred. In step S104, when a sign is generated in each of the plurality of facilities in the facility operating state, by comparing the cumulative failure probabilities of the sign pattern D related to the signs, the probability of the most likely failure from the plurality of facilities in which the sign has occurred is obtained. A high facility is selected, and the process proceeds to step S105.

  As a specific example of the processing in step S104, for example, when a sign occurs in a plurality of facilities at a certain timing, the cumulative failure probabilities of the predictive patterns related to these signs are compared, and the cumulative failure probability is high among these facilities. Select equipment with predictive pattern. Then, maintenance inspection of the selected equipment is performed with priority. Thus, based on the cumulative failure probability, it is possible to grasp the equipment that requires the most maintenance inspection (equipment with the highest probability of failure).

  In step S105, the dispatch limit time is calculated for the equipment selected as requiring the most maintenance in step S104. The dispatch limit time is calculated by comparing the failure prediction time that is predicted to result in a failure after the occurrence of a sign and the maintenance time until the maintenance staff reaches the facility and completes the maintenance inspection. .

  The failure prediction time is calculated on the basis of a sign pattern related to the sign that has occurred. That is, the calculation is performed based on the predicted time from occurrence of a sign to failure. In addition, the maintenance time is calculated based on the time required for a maintenance person waiting at a branch office or the like close to the facility where the sign is generated to go to the facility and complete the maintenance inspection. Then, the dispatching limit time is calculated so that the maintenance time is shorter than the predicted failure time, and the maintenance staff performs maintenance inspection based on the dispatching limit time.

  As described above, the equipment inspection order setting device 10 creates a predictive pattern D by extracting a predictor highly related to a failure from a plurality of predictors leading to a failure based on the failure history data. When the first sign in the sign pattern D occurs, the cumulative failure probability leading to the failure is calculated. Then, the equipment inspection order setting device 10 calculates the priority of the equipment 2 that performs maintenance inspection based on the cumulative failure probability. For this reason, if there are multiple signs before the failure, exclude those signs that are not related to the failure from the signs and lead to the failure based on the signs that are highly related to the failure. The failure probability can be calculated, and the accuracy of the failure probability can be improved.

  As a result, the priority of the equipment 2 that performs maintenance and inspection can be calculated with high accuracy, and the priority of the equipment 2 that performs maintenance and inspection in consideration of the failure time and the probability of reaching the failure can be determined. In addition, by setting the maintenance inspection order from the equipment 2 that is likely to fail, the maintenance inspection can be performed efficiently, the failure can be reduced, and the equipment 2 can be used for a long time.

  In addition, a threshold for determining the relationship between a sign and a failure is included, and the number of signs related to a failure in the sign pattern D can be adjusted by adjusting this threshold. In addition, the data amount of the predictive pattern D can be adjusted.

  In addition, when the same type of predictor is continuously present in the predictor pattern D, the predictor pattern D is created with the same type of predictor as one predictor, and therefore, the similar predictor pattern D can be excluded. Data amounts of different types of predictive patterns D can be ensured.

  In addition, the cumulative failure probability is calculated by Monte Carlo simulation, and the upper limit of the number of Monte Carlo simulations is set according to the number of signs in the sign pattern D, so that simulation suitable for the sign pattern D can be performed. The load of calculation by simulation can be reduced. In particular, when the number of facilities and the amount of failure history data are large, the amount of computation for Monte Carlo simulation is enormous, but by setting the upper limit number of simulations, the necessary computation results can be obtained and the computation load is increased. Can be reduced.

  In addition, the dispatch limit time calculation unit 15 calculates the dispatch limit time based on the priority calculated by the priority calculation unit 14, so that the maintenance staff inspects the equipment 2 based on the calculated dispatch limit time. By doing so, it becomes possible to further reduce the failure of the equipment 2.

  2 equipment, 3 operation history receiving unit, 4 failure history database, 10 equipment inspection rank setting device, 11 predictive pattern extraction unit, 12 transition probability accumulation unit, 13 cumulative failure probability calculation unit, 14 priority calculation unit, 15 dispatch limit time Calculation unit, D, D1, D2, D3, D4 predictive pattern, G1, G2, G3, G4, G5, G6 path, K1, K2 failure, L1, L2, L3, L4, L5 continuous part, Y1, Y2, Y3 , Y4, Y10, Y11, Y12, Y13Y, 14

Claims (6)

A failure history storage unit that stores information on the signs collected from a plurality of facilities and information on failures that occurred after the signs;
Based on the sign stored in the failure history storage unit and the failure, the sign related to the failure is extracted from the plurality of signs generated before the failure, and the sign based on the extracted sign A predictive pattern extraction unit for creating a pattern;
A transition probability accumulation unit that calculates and accumulates transition probabilities between predictors and predictive failures in the predictor pattern;
When the first sign in the sign pattern occurs, a cumulative failure probability calculation unit that calculates a cumulative failure probability from the sign to the failure based on the transition probability;
A priority calculation unit that calculates the priority of the facility that performs maintenance and inspection based on the cumulative failure probability for the plurality of facilities in which the sign has occurred,
An equipment inspection order setting device comprising: The equipment inspection rank setting device according to claim 1,
The sign pattern extraction unit extracts the sign related to the failure from a plurality of the signs by comparing a change rate of a failure probability between the signs and a threshold value for determining an association with the failure. Equipment inspection order setting device characterized by the above. The equipment inspection rank setting device according to claim 1 or 2,
The sign pattern extraction unit creates the sign pattern with the same kind of the same sign as the one sign when the same kind of the sign is continuously present in the sign pattern. Inspection order setting device. The equipment inspection rank setting device according to any one of claims 1 to 3,
The cumulative failure probability calculation unit calculates the cumulative failure probability by Monte Carlo simulation, and sets an upper limit number of times of the Monte Carlo simulation according to the number of signs in the sign pattern. . The equipment inspection rank setting device according to any one of claims 1 to 4,
Based on the priority calculated by the priority calculation unit, a dispatch limit time calculation unit that calculates a dispatch limit time, which is a limit time for the equipment to fail, unless a maintenance worker is dispatched. Equipment inspection order setting device. A facility inspection order setting method for setting a priority order of the facilities for performing maintenance inspection based on failure histories collected from a plurality of facilities,
Accumulate information on signs collected from multiple facilities and information on failures that occurred after the signs,
Based on the sign and the failure, extract the sign associated with the failure from a plurality of the signs that occurred before the failure, and create a sign pattern based on the extracted sign
Calculate and accumulate transition probabilities between predictors and predictive failures in the predictor pattern,
When the first sign in the sign pattern occurs, the cumulative failure probability from the sign to the failure is calculated based on the transition probability,
A facility inspection order setting method, comprising: calculating a priority of the facility that performs maintenance inspection on the plurality of facilities in which the sign has occurred based on the cumulative failure probability.

Images