JPH02103602A - Diagnosis supporting device - Google Patents

Abstract

PURPOSE:To prevent incidence of an accident similar to a past accident by preparing an appropriate inspection chart reflecting past examples of accidents. CONSTITUTION:An inference arithmetic section 6 selects past accidents occurred at plants having a structure or operating conditions similar to an object plant to be diagnosed from examples of plan accidents contained in a data base section 5 and statistically finds relations between operating hours and stating and stopping frequencies and accident occurring probabilites. Then the section 6 searches parts where the accident occurring probabilities are high from the object plant and prepares an inspection chart based on the searched results. Accordingly, the inspection chart which convers parts having higher accident occurring probabilities without omission and is less in fatile inspection can be prepared. Therefore, incidence of an accident similar to a past accident can be prevented.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention provides an appropriate method for periodic inspections and remaining life diagnosis of plants such as boilers, buildings, mechanical products, etc. that reflects accident examples of other plants, etc. The present invention relates to a diagnostic support device that creates a medical chart for inspection.

[Conventional technology I]

Conventionally, when conducting periodic inspections and remaining life diagnosis for plants such as boilers, experts in inspection and diagnosis inspected and diagnosed the target plant based on the standard inspection sheet that indicated the plant inspection method and its cycle. They determined the necessary areas and created an inspection chart.

Generally, plants such as boilers are designed and constructed for each plant according to its use. For example, in the case of a boiler, the structure differs slightly depending on its output, fuel, operating conditions, etc. However, although some of the contents of standard inspection procedures are for specific types of plants, most of them are general contents that cover all types of plants.

Therefore, in order to conduct an inspection that takes into account accidents and damage specific to the plant being diagnosed, a person familiar with the structure and inspection methods of the plant should research past accident records and create a chart for inspection. There was a need.

However, this process is very complicated and involves selecting records of plants with similar structures to the plant to be diagnosed from the accident records of many plants and examining each individual part, and it is difficult to avoid oversights. Ta.

[Problem to be solved by the invention■]

As mentioned above, in the past, medical records for inspection were created by relying on the judgment of experts, and there was no consideration given to the rational use of past accident records, and the work was complicated. There were problems such as overlooking parts that needed inspection and unnecessary inspections.

[Conventional technology■]

Structural metal materials used in high-temperature, high-pressure equipment such as boilers for power generation plants are exposed to high temperatures and high stress for long periods of time, resulting in creep damage and corrosive wear damage during operation.
Furthermore, during startup and shutdown, they suffer from low-cycle fatigue damage due to thermal stress. Metal materials exposed to such environments are usually designed based on creep rupture strength after 100,000 hours. However, about half of the boilers currently in use have already been used for more than 100,000 hours, and more are planned for use in the future. In addition, due to the recent base load operation of nuclear power plants, thermal power plants are exposed to harsh operating conditions where they are repeatedly started and stopped every day.

Under these circumstances, it is important to properly diagnose the remaining life of high-temperature, high-pressure equipment from the viewpoint of safety and economy, and highly accurate remaining life diagnosis technology is being studied in various fields.

Conventional methods for diagnosing the remaining life of metal materials are broadly classified into analytical methods using numerical calculations, destructive methods, and non-destructive methods. The analytical method calculates the damage rate based on the creep rupture life diagram from the stress and strain of the member determined by FEM etc. and the operating time. The destructive method is a method in which a test piece is cut out from an actual machine member and subjected to tensile, creep rupture, fatigue tests, etc. to determine the damage rate.

Non-destructive methods include a method of copying the metal structure using a replica and estimating the damage rate from the state of the micro-metal structure, and a method of estimating the damage rate from changes in physical quantities caused by heating, creep, fatigue, etc. There is. The replica method involves measuring the amount of deformation of crystal grains in the metal structure (
JP-A No. 62-258087), cavity (
Creep damage), method of measuring microcracks, Cr-M
A method of estimation based on the degree of decomposition of pearlite in O steel.

There are methods for qualitatively evaluating the state of precipitation of the sigma phase in stainless steel.

In the method of estimating physical quantities, hardness (Japanese Patent Application Laid-Open No. 57-104
No. 837, JP-A-58-92952), electrical resistance (JP-A-58-60248), ultrasonic sound velocity (
JP-A-53-120585), misorientation due to X-rays and coil impedance due to eddy current (JP-A-53-88781), etc. have been proposed.

However, the conventional technology has the following problems. The analytical method can be easily applied to any member as long as the material, shape, temperature, and operating history of the diagnostic part are known.

However, due to variations in the lifespan and stress distribution of the material itself,
In reality, it is difficult to accurately capture temperature distribution, wall thinning rate, etc., and the input data is often an extrapolated value or estimated value, which poses a problem in terms of accuracy.

Although destructive methods can estimate the damage rate with relatively high accuracy, they require a great deal of cost and time for sample collection, repair, and evaluation testing, so they are not suitable for periodic remaining life diagnosis.

Non-destructive methods allow diagnosis to be made in a relatively short period of time and without damaging the component. However, the material of the parts.

The application conditions are limited, such as the manufacturing history such as heat treatment conditions and the type of damage sustained by the component, such as creep or fatigue, resulting in decreased accuracy or making evaluation impossible. There's a problem.

Therefore, when the same location is diagnosed using the above diagnostic technology,
The results rarely coincided, and these results needed to be comprehensively judged by a diagnostic expert.

[Problem to be solved by the invention■]

As mentioned above, in the conventional remaining life diagnosis system for metal materials, it is difficult to incorporate the inherent life and local stress distribution of the diagnostic material itself using analytical methods, so estimated values are based on input data. In many cases, there is a problem with accuracy.

Destructive methods are used to take samples from diagnostic materials;
Repairs and evaluation tests require time, and they are not suitable for periodic remaining life diagnosis.

A problem with non-destructive methods is that the application conditions are limited depending on the material of the diagnostic member and the type of damage.

A first object of the present invention is to create an appropriate inspection chart that reflects past accident cases.

The second object of the present invention is to comprehensively and rationally judge and infer the diagnostic results to determine the remaining life evaluation value and its reliability, and to provide a metal material that can point out inaccurate diagnostic results. The purpose of this invention is to provide a diagnostic support device.

[Means I to solve the problem]

The above first purpose consists of a database section that stores data such as plant accident records, design conditions, and operation history, and an inference/calculation section that searches the database based on input information and performs statistical calculations and creates inspection charts. This invention is characterized by providing a processing device (first invention).

[Means to solve the problem■]

In order to achieve the above-mentioned second objective, there is provided a metal material diagnosis support device that evaluates the damage rate of actual machine parts using analytical, destructive and non-destructive diagnostic means and diagnoses the remaining life. It has a calculation section that calculates the damage rate of the actual machine parts, and a judgment section that is equipped with fuzzy reasoning that comprehensively judges the damage rates of each component, and is configured to diagnose the remaining life through this fuzzy reasoning. (Second invention).

[Effect I]

The above inference/calculation unit selects accidents that occurred in plants with similar structures or similar operating conditions to those to be diagnosed from plant accident examples stored in the database, and statistically calculates the relationship between operating time, number of starts and stops, and probability of accident occurrence. Based on the results, an inspection chart is created based on the results.

This makes it possible to create an inspection chart that does not overlook parts with a high probability of causing an accident and with fewer unnecessary inspections, making it possible to prevent accidents similar to those that have occurred in the past.

[Effect■]

According to the present invention, by providing fuzzy inference in the remaining life diagnosis system for metal materials, damage rates estimated by multiple diagnostic methods can be comprehensively evaluated, and damage rates, remaining life, and their reliability can be quantitatively evaluated. It is calculated as follows.

[Embodiments of the invention] (Example I) Embodiments of the first invention will be described below with reference to the drawings.

FIG. 1 is a configuration diagram of an embodiment of a plant diagnosis support device according to the present invention.

This device consists of a processing device 1, a display device 2, an input device 3, and an output device 4. The processing device 1 includes a database unit 5 and an inference/calculation unit 6, and the database unit 5 receives necessary data from a database 7 in the host computer and an operation recording device 8 installed in the plant. The inference/calculation unit 6 searches the data beta 7 and creates an inspection chart based on the information input from the input device 3 according to the dialog screen output to the display device 2 Output to.

FIG. 2 shows the detailed configuration of this device. Database part 5
includes design/structural data 51, repair/modification data 52,
Data such as operation history data 53, inspection history data 54, plant accident/damage data 55, standard inspection criteria data 56, standard inspection method data 57, etc. are stored. on the other hand,
The inference/calculation unit 6 performs diagnostic information manpower 61, target plant information search 62, similar plant accident data search 63, statistical processing 64, accident probability calculation 65, inspection location determination 66, inspection chart creation 67, inspection chart output 68, etc. Consists of routines.

This device first inputs information such as the name of the plant to be diagnosed and inspection period according to the dialog screen output on the display device 2, and then inputs information such as the target plant's structure, repair/modification history, operation history, inspection oil layer, etc. is searched from the database 7. Then, we search for plant accident data that has a similar structure or similar operation history to the target plant, and calculate the relationship between accident probability and operating time or number of starts and stops by location and cause, as shown in Figure 3, which will be described later. etc., calculate the probability of an accident in each part of the target plant, and find parts with a higher probability of an accident than the standard value. Based on this result, the inspection points are determined in conjunction with the standard inspection standard data 56, and an inspection chart is created with reference to the standard inspection method data 57. Output the medical record (Figures 4 to 8). Note that FIGS. 4 to 8 will be described later.

As an example of organizing plant accident data, FIG. 3 shows the relationship between the probability of an accident and the operating time at part A of the plant. Accidents that occur in plants have different occurrence times and probability distributions depending on the location and type of cause. For example, accidents that show a distribution like cause 1 that occurs intensively at the beginning of operation due to defects in plant creation, etc., accidents like cause 2 that occur periodically during the operation period although the probability of occurrence is low, There are accidents like cause 3 that suddenly become more likely to occur after a certain period of time due to deterioration of components over time.

By sorting out accidents that have occurred in plants with similar structures and similar operation histories to those to be diagnosed as described above, inspection locations and methods can be determined based on the probability that an accident has occurred in each part of the target plant.

Examples of output to the display device 2 are shown in FIGS. 4 and 5.

The dialog screen 2a displays the points that need to be inspected and their methods in writing, and the diagnostic point display screen 2b shows a schematic diagram of the diagnosed points (
Fig. 4) and a detailed diagram (Fig. 5) are shown.

In addition, by selecting the command displayed on the command screen 2C using an input device such as an X and Y direction input device (for example, a mouse) or a keyboard, the inspection chart shown in FIGS. 6 to 8 can be output to the output device. It can be output from 4.

FIGS. 6 to 8 are examples of inspection charts.

FIG. 6 is a chart summarizing the inspection items of the target plant, and FIG. 7 is a diagram showing the content of the output explaining in detail inspection points and inspection methods. FIG. 8 is a record chart for recording results during inspection work, and items that can be inspected at the same time are summarized to facilitate recording.

In the embodiments of the present invention described above, inspection of plants is mentioned, but it is not necessarily limited to plants (if the contents of the database are replaced, the basics of the present invention can also be used for inspection of buildings, mechanical products, etc.). It is clear that it can be implemented without changing the principle.

(Example 2) Next, an example of the second invention will be described with reference to FIGS. 9 to 17.

FIG. 9 shows the overall configuration of an expert system 1000 for diagnosing the remaining life of a boiler for thermal power generation.

This system 1000 has a database 2000 consisting of data for remaining life diagnosis, a knowledge base 3000 consisting of knowledge of diagnosis methods, etc., and calculates the damage rate and its reliability by comprehensively judging multiple diagnosis results. fuzzy inference 400
0 and the remaining life calculation based on the result 50
Display device 7000 where 00 and inspector 5ooo input measurement data and display necessary information regarding remaining life.
and a user interface 600 that rewrites the information of the remaining life diagnosis system and the measured information and performs mutual input/output.
Consists of 0.

The database 2000 includes a structure database 100, a material database 200, and an operation history database 300.
It consists of. Knowledge base 3000 includes analysis method rule base 400, grain deformation method rule base 500, cavity method rule pace 600, and carbide method rule base 7.
It consists of 00.

Note that fuzzy inference is a method of estimating quantitative results from ambiguous information such as human judgment.

Next, the functions of each component will be explained.

The structure database 100 within the database 2000 is
It is used to memorize the structure of the target boiler.
The material database 200 includes creep rupture data, fatigue life data, etc. (10th data) of materials used in the boiler.
(See the creep rupture curve in the figure and the stress relaxation curve in FIG. 11). The operation history database 300 is for storing operation time, number of starts/stops, output conditions, etc. (see FIG. 12).

The analysis method rule base 400 in the knowledge base 3000 is
Receives and receives data 20 such as the shape and material of the diagnostic member from the structure database 100, data 30 such as creep rupture life diagram and thinning rate from the material database 200, and data 40 such as operating time from the operation history database 300. , the damage rate 50 is determined by the method shown in FIG.

The grain deformation method rule base 500 calculates the damage rate 60 according to the broken line arrow shown in FIG. 11 from the master curve 31 for the material of the diagnostic member and the measured deformation coefficient 13.
seek.

Similar to the grain deformation method rule base 500, the cavity method rule base 600 is based on the master curve 32 and cavity area ratio 1 of the diagnostic member from the material database 200.
The damage rate 70 is determined from the measured values of No. 4 in accordance with the broken line arrow shown in FIG.

In the carbide method rule base 700, the damage rate 80 is similarly determined from the master data 33 and the measured distance between carbides. Here, the grain deformation method rule base 500 functions only when the diagnostic member is a Cr-Mo steel base material, and the cavity method rule base 600 functions only when the diagnostic member is stainless steel or a welded part.

The fuzzy inference 4000 uses the damage rates 50, 60, and 70.80 found in the knowledge base 3000 to calculate the overall damage rate (dispersion E is shown on the horizontal axis) and its reliability level 90 using the method shown in Figure 16. , if inaccurate data exists, output 16 prompts the inspector to re-measure, etc. On this occasion,
As mentioned above, in the case of Cr-Mo steel base material, the damage rate is 50.
60°80 is used, and in the case of stainless steel and welded parts, a damage rate of 50.70.80 is used for comprehensive evaluation.

The remaining life calculation 5000 calculates the remaining life using the analytical rule base 400 based on the damage rate 90 determined by the fuzzy inference 4000 (outputs the result).

Next, the fuzzy inference method will be explained.

The fuzzy inference 4000 has a function of calculating a more reliable damage rate by comprehensively determining the damage rates determined by a plurality of diagnostic methods. The inference method will be explained below with reference to FIG.

First, in the case of Cr-Mo @ base metal, analysis method, grain deformation method (in the case of stainless steel, welded part, cavity method)
and membership functions (μ4 to μC) that take into account the estimation accuracy of each method for the damage rate obtained by the carbide method.
Replace with The obtained membership function (μ, ~μ,
) are superimposed to find their common set using the following formula.

, cr=min(μ4. μ, l μC) −・(
1) This is to use the minimum value of the three membership functions for each damage as the membership function of the comprehensive judgment result. The centroid position of the obtained membership function μ is the overall damage rate, and the height of the membership function is the reliability of the result. With this evaluation method, when the difference between the respective diagnostic results is large, the reliability decreases, and when the difference is small, the reliability increases.

Furthermore, if there is a difference in the estimated value of the damage rate due to a measurement error, etc., and a common set of the three membership functions does not exist (μ = 0) as shown in Figure 17, the difference between the two membership functions Based on the high reliability of the common set, incorrect estimated damage rates are pointed out, and in the case of analytical methods, instructions are given to review inaccurate input data such as thinning rate, and non-destructive methods such as grain deformation methods are indicated. In the case of specific measures, re-measurement is recommended. Furthermore, if re-measurement is difficult, the damage rate is determined using only valid membership functions.

Through the fuzzy inference described above, it is possible to obtain a comprehensive damage rate that takes into account the reliability of each damage rate estimation result, and to point out inaccurate damage rate estimates on behalf of experts, and to re-measure, etc. The remaining life can be calculated with higher accuracy.

As another embodiment of the present invention, in the above-mentioned embodiment, analysis method 2 crystal grain deformation method is used as a method for diagnosing remaining life.

Although the cavity method and carbide method are listed, it is not necessarily limited to these, but is a destructive method.

Any method can be used without changing the basic principle of the present invention as long as it is related to damage rate estimation such as hardness, methods using ultrasonic waves, methods using X-rays or eddy currents, and crack length measurements. It is clear that it can be done.

Furthermore, although fuzzy inference is applied to three types of diagnostic methods in the embodiments of the present invention, it is not necessarily limited to three types, and any number of types as long as there are at least two or more types are applicable. It is clear that this can be done without changing the

〔Effect of the invention〕

As explained above, according to the first invention, it is possible to create an inspection chart without overlooking areas with a high possibility of an accident occurring, and with fewer unnecessary inspections, thereby preventing accidents similar to those that have occurred in the past. can be prevented.

Further, according to the second invention, as a remaining life diagnosis system for metal materials, by providing fuzzy inference for remaining life evaluation, comprehensive judgment can be made using a plurality of diagnostic methods. In this case, it is possible to obtain highly accurate remaining life diagnosis results that can take into account the reliability of individual diagnostic methods and instruct reexamination of inaccurate measurement results.

[Brief explanation of drawings]

Fig. 1 is an overall configuration diagram of a plant diagnosis support device according to an embodiment of the first invention, Fig. 2 is a detailed configuration diagram thereof, Fig. 3 is an explanatory diagram showing an example of statistical arrangement of plant accident data, and Fig. 4 5 is an explanatory diagram of a display screen including a schematic diagram of a diagnosed location, FIG. 5 is an explanatory diagram of a display screen including a detailed diagram of a diagnosed location, and FIG. 6 is an inspection item list outputted to an output device. Figure, 7th
Figure 8 is a detailed explanatory diagram of the inspection method, Figure 8 is a diagram showing a record table of inspection results, Figure 9 is a flowchart showing an embodiment of the second invention, Figures 10, 11, and 12 are A characteristic diagram showing an example of data stored in the database, Fig. 13 is a flowchart showing a damage rate estimation method using the analysis method shown in Fig. 9, and Fig. 14 shows a damage rate estimation method using the crystal grain deformation method shown in Fig. 9. FIG. 15 is a characteristic diagram showing the damage rate estimation method using the cavity method in FIG.
FIG. 17 is a characteristic diagram illustrating inaccurate damage rate judgment. 1... Processing device, 5... Database section, 6...
Inference/calculation section, 1000...Remaining life diagnosis expert system, 2000...Database, 3000...
・Knowledge base, 4000...Fuzzy reasoning, 5000
...Remaining life calculation, 7000...Display device. Fig. Fig. Fig. 21 / I Island Ki 7゜

Claims (4)

[Claims] (1) A diagnostic support device that creates inspection charts for plants, etc. includes a database section consisting of data on accidents and damage to plants, etc., data on structures of plants, etc., and searches and performs statistical processing on the data. A diagnostic support device comprising a processing device comprising an inference/calculation unit that creates an inspection chart reflecting the results. (2) In claim (1), the accident data of similar plants is statistically organized using Weibull distribution, etc., and the relationship between operating time or the number of starts and stops and the probability of accident occurrence is determined, and the probability of occurrence of an accident within the plant to be diagnosed is determined. A diagnostic support device characterized in that the diagnostic support device is configured to search for parts that have a value equal to or higher than a reference value and create the inspection chart. (3) In claim (1), the inspection chart is created by searching a database section for accident data of plants that have similar structures, operating histories, repair/modification histories, etc. Diagnostic support equipment. (4) In a metal material diagnosis support device that evaluates the damage rate of actual machine parts using analytical, destructive and non-destructive diagnostic means and diagnoses the remaining life, the damage rate of the actual machine parts is evaluated by a plurality of said diagnostic means. A diagnostic support device comprising: a calculation unit that performs calculation; and a determination unit that is provided with fuzzy inference that comprehensively determines each damage rate, and diagnoses remaining life through the fuzzy inference.