Fatigue accumulation prediction method and system

Abstract

A method for identifying at least one significant location on a physical structure, comprising the steps of: receiving operating information, the operating information including, as information relating to the operation of the physical structure, the operation of the physical structure at different points in time and the operation of the physical structure at different operating levels, the operating level relating to the output level from the physical structure, and at least a portion of the operating information being received from a sensor sensing the state of the physical structure; predicting damage to the physical structure based on the operating information, predicted operation of the physical structure at at least one of the different operating levels, and at least one model of the physical structure, the occurrence of damage at multiple locations on the physical structure being predicted regardless of the proximity of sensors to each of the multiple locations; and identifying at least one significant location on the physical structure based on the predicted damage.

Description

[Technical Field]

[0001] Cross-citation of related applications This application claims priority under Section 119 of the United States Patent Act to U.S. Provisional Application No. 63 / 449,730 filed on 3 March 2023 and U.S. Non-Provisional Application No. 18 / 399,016 filed on 28 December 2023, the contents of each application being incorporated in their entirety.

[0002] Government licensing rights This invention was made possible with government support under DE-AR0001295, issued by the Agency for Advanced Research Projects (ARPA-E) of the Department of Energy. The government has certain rights to this invention.

[0003] This disclosure relates to a physical structure that accumulates fatigue and damage with use, and / or a system for monitoring the use of the physical structure and estimating and predicting the accumulation of fatigue and damage. [Background technology]

[0004] For example, physical structures such as energy plants and other equipment and subsystems accumulate fatigue and damage during use. This fatigue and damage accumulate due to factors such as pressure, vibration, and heat. To prevent (or avoid) the physical structure becoming unusable, repairs to eliminate the accumulated fatigue and damage may ultimately be necessary. [Overview of the Initiative]

[0005] At least some exemplary embodiments relate to a method for identifying at least one important location on a physical structure.

[0006] In some exemplary embodiments, the method includes the steps of: receiving operating information, which, as information relating to the operation of a physical structure, includes the operation of the physical structure at different points in time and the operation of the physical structure at different operating levels, the operating levels relating to output levels from the physical structure, and at least a portion of the operating information being received from a sensor sensing the state of the physical structure; predicting damage to the physical structure based on the operating information, predicting the operation of the physical structure at at least one of the different operating levels, and at least one model of the physical structure, predicting the occurrence of damage at multiple locations on the physical structure, regardless of the proximity of the sensor to each of the multiple locations; and identifying at least one significant location on the physical structure based on the predicted damage.

[0007] In some exemplary embodiments, the method further includes the step of generating a work order or alarm based on at least one important location and at least one model.

[0008] In some exemplary embodiments, at least one model includes one or more machine learning regression models or machine learning time series models.

[0009] In some exemplary embodiments, the operating information includes available time-series data indicating stress or material temperature within the physical structure and unavailable time-series data where the stress or material temperature within the physical structure is unknown, and the step of predicting damage to the physical structure includes predicting initial damage to the physical structure by interpolating the operating information to predict additional operating information at the same location or angle on the physical structure across different operating levels of the physical structure associated with the unavailable time-series data, and linking the operating information with the additional operating information over a specified prediction period to generate complete operating information.

[0010] In some exemplary embodiments, the step of predicting initial damage to a physical structure involves performing a rainflow count (RC) using a simplified RC algorithm to count the number of cycles in complete operating information; estimating the number of cycles until damage occurs at a location or angle of the physical structure where operating information or additional operating information is available; predicting the number of cycles until damage occurs at a different location or angle where complete operating information is unavailable, by quantifying the level of uncertainty in the number of cycles until damage occurs using a machine learning model; and determining the damage rate in each cycle of the number of cycles in complete operating information. The steps to be calculated according to TIFF2026510735000002.tif19150 further include, where "D" is the damage rate, "k" is the number of stress levels, and "n" i " is the cumulative number of cycles, "N i " represents the number of cycles until damage occurs at the i-th stress.

[0011] In some exemplary embodiments, the initial damage includes at least one of fatigue damage, creep damage, oxidation damage, or wear damage of the physical structure, and the fatigue damage includes surface cracks or subsurface cracks of the physical structure.

[0012] In some exemplary embodiments, the step of predicting damage to the physical structure is: Based on the formula in TIFF2026510735000003.tif9150, this further includes predicting the time-dependent growth of initial damage to the physical structure, where "a n " is the crack length in cycle n, and includes the level of uncertainty regarding the number of cycles until damage occurs, and "c" and "m" are material coefficients that include the level of uncertainty regarding the physical structure, and "K eff " is the stress level (σ n ) and based on the crack shape, This is the effective stress intensity factor calculated using the formula in TIFF2026510735000004.tif18150, where "σ n,min " is the minimum stress, "σ n,max " is the maximum stress, JPEG2026510735000005.jpg7150 is the stress intensity factor that changes based on the shape of the initial damage to the physical structure and the physical structure.

[0013] In some exemplary embodiments, the step of predicting damage to the physical structure is further based on operating a digital twin of the physical structure, where the digital twin is an electronically generated model of the physical structure.

[0014] In some exemplary embodiments, the state of the physical structure includes temperature data and flow rate data, and the method further includes updating the state of the reactor at at least one critical location and updating at least one model based on the updated state of the reactor at at least one critical location.

[0015] In some exemplary embodiments, the physical structure is included in a nuclear power plant that further includes a reactor, and the operating level is the output level of the reactor.

[0016] In some exemplary embodiments, the operating information includes details of repairs and installations of the physical structure.

[0017] In some exemplary embodiments, it further includes controlling the device to change the state in the physical structure based on at least one critical location and at least one model.

[0018] Some exemplary embodiments relate to a device configured to identify at least one critical location on a physical structure.

[0019] In some exemplary embodiments, the apparatus includes a processing circuit which receives operating information, which includes, as information relating to the operation of a physical structure, the operation of the physical structure at different points in time and the operation of the physical structure at different operating levels, the operating levels relating to output levels from the physical structure, and at least a portion of the operating information is received from sensors that sense the state of the physical structure, and is configured to predict damage to the physical structure based on the operating information, predicted operation of the physical structure at at least one of the different operating levels, and at least one model of the physical structure, the occurrence of damage at multiple locations of the physical structure is predicted regardless of the proximity of sensors to each of the multiple locations, and to identify at least one critical location on the physical structure based on the predicted damage.

[0020] In some exemplary embodiments, the processing circuit is further configured to generate work instructions or alarms based on at least one critical location and at least one model.

[0021] In some exemplary embodiments, at least one model includes one or more machine learning regression models or machine learning time series models.

[0022] In some exemplary embodiments, the operating information includes available time-series data indicating stress or material temperature within the physical structure and unavailable time-series data where the stress or material temperature within the physical structure is unknown, and the processing circuit is configured to predict damage to the physical structure by interpolating the operating information to predict additional operating information at the same location or angle on the physical structure across different operating levels of the physical structure associated with the unavailable time-series data, and by linking the operating information with the additional operating information over a specified prediction period to generate complete operating information, thereby predicting at least initial damage to the physical structure.

[0023] In some exemplary embodiments, when predicting initial damage to a physical structure, the processing circuit further executes a rainflow count (RC) using a simplified RC algorithm, counts the number of cycles in the complete operation information, estimates the number of cycles until damage occurs at the position or angle of the physical structure where the operation information or additional operation information is available, and predicts the number of cycles until damage occurs at different positions or angles where the complete operation information is not available by quantifying the uncertainty level of the number of cycles until damage occurs using a machine learning model. The damage rate for each cycle of the number of cycles in the complete operation information is calculated as configured to be calculated by TIFF2026510735000006.tif19150, where "D" is the damage rate, "k" is the number of stress levels, "n i " is the cumulative number of cycles, and "N i " is the number of cycles until damage occurs at the i-th stress.

[0024] In some exemplary embodiments, the prediction of damage to the physical structure is further based on operating a digital twin of the physical structure, where the digital twin is an electronically generated model of the physical structure, and the initial damage includes at least one of fatigue damage, creep damage, oxidation damage, or wear damage to the physical structure, and the fatigue damage includes surface cracks or subsurface cracks in the physical structure.

[0025] In some exemplary embodiments, the processing circuit is further configured to control the device to change the state in the physical structure based on at least one critical position and at least one model.

[0026] Some exemplary embodiments relate to a non-temporary computer-readable medium which, when executed by a processor, the processor receives operating information, which includes, as information relating to the operation of a physical structure, the operation of the physical structure at different points in time and the operation of the physical structure at different operating levels, the operating levels relating to output levels from the physical structure, and at least a portion of the operating information received from sensors sensing the state of the physical structure, and which includes instructions that predict damage to the physical structure based on the operating information, a predicted operation of the physical structure at at least one of the different operating levels, and at least one model of the physical structure, and that, based on the predicted damage, the occurrence of damage at multiple locations on the physical structure is predicted regardless of the proximity of sensors to each of the multiple locations, and that, based on the predicted damage, at least one critical location on the physical structure is identified. [Brief explanation of the drawing]

[0027] The various features and advantages of the non-limiting embodiments described herein can be better understood by considering them together with the detailed description of the invention and the accompanying drawings. The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of the claims. Unless expressly stated otherwise, the accompanying drawings should not be considered to be drawn to actual scale. Dimensions in the drawings may be exaggerated for illustrative purposes.

[0028] [Figure 1] Figure 1 is a block diagram of the apparatus according to one embodiment. [Figure 2] Figure 2 is a schematic diagram of an energy plant according to one embodiment. [Figure 3] Figure 3 is a model diagram of a physical reactor. [Figure 4] Figure 4 is a flowchart of the operation performed by the device. [Figure 5] Figure 5 shows an example of linking time series of stress signals over an exemplary period to create a daily or annual operating routine. [Figure 6] Figure 6 shows an example of historical data at different power levels. [Figure 7]Figure 7 shows examples of simulated stress signals at different power levels. [Figure 8] Figure 8 shows an example of rainflow count and detected cycles. [Figure 9] Figure 9 shows an example of quantifying uncertainty to identify critical locations for crack initiation. [Figure 10] Figure 10 shows an example of crack propagation over a time cycle. [Modes for carrying out the invention]

[0029] This application discloses some detailed embodiments. However, the specific structural and functional details disclosed herein are merely examples to illustrate the embodiments. The embodiments can be implemented in various alternative forms and should not be construed as being limited to the embodiments described herein.

[0030] Accordingly, while the embodiments can be modified or replaced in various ways, these embodiments are shown as examples in the drawings and are described in detail in this application. However, it should be understood that the embodiments are not limited to the specific forms disclosed, but rather encompass all modifications, equivalents, and alternative forms. Throughout the description of the drawings, the same numbers refer to the same elements.

[0031] When an element or layer is described as “on,” “connected to,” “coupled to,” “attached to,” “adjacent to,” or “covering” another element or layer, that element or layer is described as being on, connected to, coupled to, attached to, adjacent to, or covering the other element or layer, either directly or through the other element or layer. On the other hand, when an element is described as being “directly on,” “directly connected,” or “directly coupled” to another element or layer, it should be understood that there is no intervening element or layer. Throughout this specification, the same number refers to the same element. In this application, the term “and / or” includes any combination or partial combination of one or more of the relevant descriptions.

[0032] In this application, terms such as “first,” “second,” and “third” may be used to describe various elements, regions, layers, and / or parts, but these elements, regions, layers, and / or parts are not limited by these terms. These terms are intended solely to distinguish one element, region, layer, or part from another. Accordingly, the first element, region, layer, or part described below may also be called the second element, region, layer, or part, without departing from the teaching of the exemplary embodiments.

[0033] Terms relating to spatial relative relationships (e.g., “down,” “below,” “underside,” “up,” “top,” etc.) may be used herein to facilitate understanding of the relationship between one element or feature and another, as shown in the figures. These terms are intended to include different orientations of the device during use or operation, in addition to the orientations shown in the figures. For example, if the device in the figure is turned over, an element described as “below” or “directly below” another element or feature will be oriented “above” that other element or feature. Thus, the term “down” may encompass both up and down directions. Furthermore, the device may be oriented in other directions (90-degree rotation or other directions), and the terms relating to spatial relative relationships used herein shall be interpreted accordingly.

[0034] The terminology used in this application is for the sole purpose of describing various embodiments and is not intended to limit the embodiments. The singular forms "a," "an," and "the" used in this application include the plural form unless otherwise specified. Furthermore, the terms "includes," "including," "comprises," and / or "comprising" used in this application are intended to identify the presence of the described features, integers, steps, operations, and / or elements, and should not be used to exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0035] In this specification, when the terms “about” or “substantially” are used in relation to a numerical value, that value includes a manufacturing or operational tolerance (e.g., ±10%). Furthermore, when the terms “generally” or “substantially” are used in relation to a geometric shape, precision of the geometric shape is not required, and the tolerance of the shape is intended to be within the scope of this disclosure. In addition, whether a numerical value or shape is modified with “about,” “generally,” or “substantially,” that numerical value or shape should be interpreted as including a manufacturing or operational tolerance (e.g., ±10%).

[0036] Examples of embodiments can be illustrated by referring to symbolic representations of actions and operations (e.g., flowcharts, flow diagrams, data flow diagrams, structural diagrams, block diagrams, etc.) that may be implemented in combination with the units and / or devices described in more detail below. Although described in a particular way, functions or operations specified in a particular block may be performed in a different manner than the flow specified in the flowchart, flow diagram, etc. For example, functions or operations shown to be performed sequentially in two consecutive blocks may actually be performed simultaneously, or in some cases in the reverse order.

[0037] Unless otherwise defined, all terms used in this Application (including technical and scientific terms) have the same meaning as those generally understood by those skilled in the art in the field to which the exemplary embodiments belong. Furthermore, terms should be interpreted to be consistent with their meaning in the context of the relevant art, including terms defined in commonly used dictionaries, and should not be interpreted in an idealized or overly formal sense unless expressly defined herein.

[0038] While this description is based on specific examples and drawings, those skilled in the art should be able to modify, add to, and substitute the exemplary embodiments as described. For example, the described techniques may be performed in a different order than described, and / or the components of the described systems, architectures, devices, circuits, etc., may be connected or combined in a different manner than described above, or the results may be adequately achieved by other components or equivalents.

[0039] Figure 1 is a block diagram of the apparatus according to one embodiment.

[0040] There is a need to reduce lifecycle maintenance costs, and consequently, reduce costs in the design, construction, and operation of various structures, by reducing the size of components used in various structures, increasing the modularity of components included in a structure to reduce the costs that may arise regarding the repair and replacement of components during the lifecycle, and / or reducing the number of on-site personnel required to operate the structure.

[0041] Such structures include a variety of engineering structures, such as industrial power turbines, aircraft engines, wind turbines, and nuclear reactors. In nuclear reactors, one structure being considered to achieve the above objectives is the Small Modular Reactor (SMR). SMRs are designed to reduce the levelized cost of power generation by nuclear reactors through design features that include reduced internal compartmentation, a smaller overall size of reactor components, and improved modularity. These design features are implemented to reduce the operating and construction costs of SMRs compared to conventional boiling water reactors. Furthermore, SMRs are designed to allow the use of commercially available off-the-shelf equipment (turbines, generators, etc.) outside the reactor core and chimney. Maintenance is facilitated by the excellent modularity of SMRs. For example, in some SMRs, the fuel assemblies, chimneys, control rods, steam separators, steam dryers, in-core instrumentation assemblies, and much of the internal reactor structure are removable, making maintenance easier. In addition, the number of permanent on-site personnel is reduced along with the cost savings. For example, during downtime, a temporary support team will be needed in addition to the regular staff.

[0042] Given these unique features of SMRs (specifically, the reduced core / chimney size and personnel requirements), and the fact that SMRs are designed with cost and maintenance reductions in mind, it is clear that procedures for more rigorous monitoring of the reactor state are needed to further optimize maintenance and shutdown schedules.

[0043] However, structures such as industrial power turbines, aircraft engines, wind turbines, and nuclear reactors including SMRs can be subjected to various forms of damage, including fatigue damage such as cracks and subsurface cracks, creep damage, oxidation damage, and wear damage. When such damage occurs, the physical structure is compromised, and the damage progresses over time. This increases lifecycle maintenance costs, for example, by incurring costs associated with monitoring the occurrence of damage and repairing or replacing damaged parts.

[0044] As will be explained in more detail below, in order to achieve the goal of reducing operating and maintenance costs, the exemplary embodiment focuses on constructing a digital twin simulation of the physical structure and predicting both initial damage to one or more components of the physical structure and the propagation of such damage where it is predicted that such damage is likely to occur.

[0045] According to an exemplary embodiment, the digital twin enables maintenance of the structure as needed, rather than based on a schedule regardless of the structure's state, thereby reducing the cost of unnecessary maintenance work. Furthermore, according to the exemplary embodiment, the digital twin also plays a role in automating the monitoring of the structure, which helps to alleviate labor shortages.

[0046] In the following, we will describe digital twins using nuclear reactors such as SMRs as an example, but digital twins according to exemplary embodiments can be applied to analyze damage to any physical structure that exhibits damage in the form of cracks caused by fatigue, creep, wear, oxidation, or other physical phenomena.

[0047] Referring to Figure 1, in some exemplary embodiments, the apparatus 900 (which may be an electronic device, computer, arithmetic unit, and / or instrument according to the exemplary embodiment) may be configured to perform any of the methods, steps, operations, etc., described herein according to the exemplary embodiment. The apparatus 900 may include a processor 920, memory 930, and interface 940 electrically connected via bus 910. Interface 940 may be a communication interface (e.g., a wired or wireless transceiver). Interface 940 may be connected to communicate with one or more external devices such as other processing units, sensors, and memory.

[0048] Memory 930 is a non-temporary computer-readable medium capable of storing instruction programs and / or other information. Memory 930 may be a non-volatile memory such as flash memory, phase-change random access memory (PRAM), magnetoresistive RAM (MRAM), resistive random access RAM (ReRAM), or ferroelectric RAM (FRAM), or a volatile memory such as static RAM (SRAM), dynamic RAM (DRAM), or synchronous DRAM (SDRAM). The processor 920 may be configured to execute the stored instruction programs to perform one or more functions. For example, the processor 920 may execute the instruction programs stored in memory 930 to control various equipment, operations, etc., of an energy plant, including a nuclear power plant (also referred to herein as a nuclear power plant, but not limited to it). The processor 920 may execute the instruction programs stored in memory 930 to perform methods, operations, functions, etc., of any exemplary embodiment, including, for example, generating, maintaining, and / or operating virtual twins (e.g., digital twins) of one or more physical structures, equipment, processes, etc., of various structures or systems, including an energy plant.

[0049] One or more of the processor 920, memory 930, and / or interface 940 may be, may include, and / or implement one or more instances of processing circuits, such as hardware including logic circuits, a combination of hardware / software such as a processor that runs software, or a combination thereof. In some exemplary embodiments, one or more instances of processing circuits may include, but are not limited to, a central processing unit (CPU), an application processor (AP), an arithmetic logic unit (ALU), a graphics processing unit (GPU), a digital signal processor, a microcomputer, a field-programmable gate array (FPGA), a system on a chip (SoC), a programmable logic unit, a microprocessor, or an application-specific integrated circuit (ASIC). In some exemplary embodiments, the memory, memory unit, etc. described herein may include a non-temporary computer-readable storage device for storing a program of instructions, such as a solid-state drive (SSD), and one or more instances of the processing circuit may be configured to execute a program of instructions for implementing some or all of the functions of the processor 920, memory 930, interface 940, etc., according to any exemplary embodiment described herein, such as performing any of the operations of the method according to the exemplary embodiment.

[0050] The apparatus 900 may be configured to perform various methods according to any embodiment (for example, based on the execution of a program of instructions stored in memory 930 by the processor 920), for example, receiving stress time-series history data including stress values ​​over different time points in different first operating levels of the physical structure; the stress time-series history data is generated based on sensors that sense the state of the physical structure at a first plurality of locations of the physical structure; and applying Gaussian process regression and autoregressive time-series models to the stress time-series history data to interpolate and predict stress values ​​over various operating levels and locations of the physical structure, and the state of the physical structure. If stress data is not available between different operating levels and locations, the steps include: predicting at least one stress time series for damage analysis across operating levels and / or locations of the physical structure; linking the stress time series history data with the predicted stress time series to generate predicted stress time series data that matches the prediction period; performing a rainflow count (RC) on the predicted stress time series data using a simplified RC algorithm and a selected operating level from the first and second operating levels to count the number of cycles in the predicted stress time series data; and determining the number of cycles (N) until crack initiation for each cycle of the predicted stress time series data. i ) calculates, and then uses Minor's law to calculate the accumulation of damage at any location on the physical structure in each cycle and to quantify the uncertainty, and uses a Gaussian process regression model to calculate N at any location on the physical structure I Predict and include locations where no sensor exists, and N in each cycle. I The method includes the steps of: identifying the most important location among any locations of the physical structure for which the lowest predicted value is predicted.

[0051] Digital twin simulations of physical structures can be performed by applying at least autoregressive time series models and Gaussian process regression to stress time series history data. Autoregressive time series models describe specific time-varying processes that are proportional to past values ​​and probability terms (imperfectly predictable terms). Gaussian process regression, on the other hand, is a nonparametric representation of a stochastic process that generalizes to spatial variables and characterizes the state of a quantity (e.g., stress) as a function of location based on observations at different locations.

[0052] For example, the digital twin applies an autoregressive time series model to interpolate and predict stress values ​​across different operating levels and locations of the physical structure, predicts at least one stress time series for damage analysis across operating levels and / or locations of the physical structure if no available stress data exists between different operating levels and locations of the physical structure, links the stress time series history data with the predicted stress time series to generate predicted stress time series data that matches the prediction period requested by the operator, performs a rainflow count (RC) on the predicted stress time series data using a simplified RC algorithm and a selected operating level from the first and second operating levels to count the number of cycles in the predicted stress time series data, and for each cycle of the predicted stress time series data, calculates the number of cycles (N) until crack initiation. i The system can calculate the stress level and then use Minor's law to calculate the accumulation of damage at any point in the physical structure during each cycle and quantify the uncertainty. The digital twin applies a Gaussian process regression model to determine the minimum number of operating cycles N required for damage analysis. I By identifying locations where scenarios demonstrating crack formation can be created, it is also possible to select critical locations within the target component.

[0053] For example, an operator might request the execution of a digital twin to simulate daily operating conditions that combine multiple power level shifts over a certain period. The digital twin uses existing stress / temperature time series while predicting any missing ones, linking the appropriate time series, and then performs a simulation of daily operating conditions over the period requested by the operator.

[0054] The physical model encompassed by a digital twin includes inherently uncertain parameters (material-related parameters, noise terms, initial crack size). While Gaussian process regression models are used to quantify these uncertainties using the best values ​​obtained from initial stress data, the digital twin can be calibrated using real-world data. Furthermore, Bayesian calibration algorithms within the digital twin can adjust the parameters within the digital twin, thereby fitting predictions to actual operating data. For example, additional information from sensors can be input as new historical data, and the digital twin simulation can be rerun based on this new historical data. It is also possible to compare the new historical data with the calculated damage accumulation and adjust the parameters of the digital twin simulation accordingly. Additionally, the digital twin simulation can be used to perform simulations of crack and other damage propagation based on Paris's law.

[0055] The device 900 may be mounted on or incorporated into a control system of an energy plant, such as a nuclear power plant, according to exemplary embodiments. For example, the device 900 may be incorporated into the EMP control system for a nuclear power plant described in U.S. Patent Application No. 17 / 490,052, incorporated herein (shown in Figure 1). The device 900 may also be directly or indirectly connected to various sensors located in various equipment and physical structures, etc. (e.g., on and / or inside equipment) in a nuclear power plant. The device 900 may directly or indirectly receive sensor data generated by the sensors (e.g., as historical data as described herein).

[0056] Figure 2 is a schematic diagram of an energy plant according to some exemplary embodiments. Such an energy plant may be a nuclear power plant equipped with a reactor and configured to operate the reactor to generate electricity (e.g., heat, power, etc.). The energy plant 1000 includes a reactor housing system 100 (e.g., a reactor module including a reactor, a reactor module including a reactor pressure vessel further comprising a reactor), a turbine 200, a condenser 400, and may further include a control system 700, but exemplary embodiments are not limited to these. A first conduit 110 leads from the reactor housing system 100 to the turbine, and a second conduit 120 leads from the condenser 400 to the reactor housing system 100. The turbine 200 and the condenser 400 may be connected. A heated working fluid (e.g., including a coolant such as steam, water, liquid metal coolant, or gaseous working fluid) delivered through the first conduit 110 passes through the turbine 200, which rotates the turbine 200, which in turn rotates the generator 300 to generate electricity (e.g., produce electricity). After passing through the turbine 200, the working fluid is cooled by the condenser 400. Coolant conduits 410 supply another working fluid (e.g., a second coolant), which is pumped up (and then returned) from a local reservoir (e.g., a body of water such as a river) to perform the necessary heat exchange to cool the working fluid within the condenser 400. The working fluid cooled in the condenser 400 is supplied (e.g., pumped up) to the reactor housing system 100 via the supply conduit 120, and the cycle is repeated. This constitutes a coolant loop. The working fluid within the reactor housing system 100 can be recirculated via recirculation conduits 130 (e.g., via a recirculation pump outside the reactor housing system 100 and a jet pump inside the reactor housing system 100) to control the power level of the reactor within the reactor housing system 100 or to cool the reactor housing system 100 during abnormal conditions, but the exemplary embodiments are not limited thereto, and such recirculation conduits may be omitted.

[0057] In some exemplary embodiments, multiple sensors, such as a first sensor 500 and a second sensor 600, can be combined, arranged, etc., to measure the physical state of one or more parts, physical structures, equipment, etc., of the energy plant 1000. For example, the first sensor 500 and the second sensor 600 may be any known temperature sensor, vibration sensor, radiation sensor, accelerometer, etc., capable of measuring moisture carryover (MCO), temperature, radiation level, acceleration (e.g., vibration), translation (e.g., translation from deformation or displacement), etc., within the energy plant 1000. In some exemplary embodiments, the first sensor 500 and the second sensor 600 may be any known stress measuring sensor. The first sensor 500 and the second sensor 600 can generate sensor data that is provided (e.g., transmitted via a communication link) as historical data to a computing device such as a control system 700.

[0058] Although the first sensor 500 and the second sensor 600 are shown as being implemented connected to the first conduit 110 and the recirculation conduit 130, respectively, the exemplary embodiments are not limited thereto. The accompanying drawings are intended to convey the comprehensive concept of the moisture carryover measurement method and system of the present invention and are not limiting to the present invention. Accordingly, the first sensor 500 and the second sensor 600 may be implemented connected to other suitable conduits / lines, tanks, or other structures consistent with the teachings of this application.

[0059] The energy plant 1000 may include a control system 700 that receives information (e.g., sensor data) from the first sensor 500 and the second sensor 600. The control system 700 may include a computing device (e.g., the computing device 900 shown in Figure 1) configured to perform methods, operations, etc., according to any exemplary embodiment. The control system 700 may be configured to communicate with one or more devices of the energy plant 1000 and control their operation to operate the energy plant 1000, for example, to generate power in a reactor included in the reactor housing system 100.

[0060] The first sensor 500 and the second sensor 600 are located on a conduit in an energy plant 1000, including a steam power generation system, but the exemplary embodiments are not limited thereto. The first sensor 500 and the second sensor 600 may be located elsewhere in the steam power generation system (e.g., in fittings, boilers, turbines, pipes, support structures), or in other forms of systems such as pipelines, molten salt furnaces, hydroelectric power plants, geothermal power plants, and non-condensing power plants. The principles disclosed herein can be applied to any physical structure and are not limited to the exemplary embodiments disclosed.

[0061] The control system 700 includes the device 900 shown in Figure 1, and can perform one or more operations, etc., in a manner according to an exemplary embodiment, based on the processor 920 of the device 900 executing instructions stored in the memory 930 of the device 900. Historical data is input (e.g., received) to the device 900 via the interface 940 from one or more sensors 500, 600, etc. of the energy plant 1000 and stored in the memory 930.

[0062] The control system 700 may, based on the digital twin simulation and method described herein, use the apparatus 900 or other hardware in accordance with exemplary embodiments to provide any of several forms of output. For example, the control system 700 may output (e.g., transmitted via alarms, display interfaces, light-emitting diode (LED) displays, etc.) a display of predicted damaged areas or areas requiring inspection, including locations most important for inspection or repair. The control system 700 may output warnings or alarms, including visual and / or audible information, via displays and speakers (not shown), based on the digital twin simulation and historical data, and / or any method according to exemplary embodiments. The control system 700 may output (e.g., transmitted) recommendations regarding maintenance or inspection. The control system 700 may, based on the digital twin simulation, modify the operation of the energy plant 1000 (e.g., adjust the operation of the nuclear power plant, including the reactor in the reactor housing system 100) (e.g., reduce the reactor output until inspection is complete to reduce the risk of damage). The control system 700 may, based on the digital twin simulation, modify the reactor schedule, including cycles and output levels. The control system 700 can implement a computer learning protocol trained using historical data to perform a digital twin simulation and provide various outputs of the control system 700. The control system 700 may adjust the digital twin simulation based on new historical data acquired in real time during reactor operation.

[0063] In some exemplary embodiments, the energy plant 1000 may be a nuclear power plant that includes a reactor, which is a boiling water reactor (BWR), within a reactor housing system 100, but exemplary embodiments are not limited thereto. For example, the energy plant 1000, which includes a control system 700 and a device 900 configured to perform the method according to the exemplary embodiment, may include, but is not limited to, any type of reactor, including boiling water reactors (BWRs), pressurized water reactors (PWRs), liquid metal-cooled reactors (e.g., sodium-cooled fast reactors (SFRs)), molten salt reactors (MSRs), improved boiling water reactors (ABWRs), economically efficient simplified boiling water reactors (ESBWRs), small modular reactors (SMRs), BWRX-300 reactors, and others.

[0064] Some exemplary embodiments, including the exemplary embodiment shown in Figure 2, include energy plants, including nuclear reactors (e.g., nuclear power plants), but the exemplary embodiments are not limited thereto, and devices configured to perform the methods according to the exemplary embodiment may be placed in various systems, processes, etc. (e.g., factories, fossil fuel power plants, etc.).

[0065] According to some exemplary embodiments, the exemplary nuclear power plant may be operated to generate electricity (e.g., heat, power, etc.) from the reactor. Such a nuclear power plant may include energy plant 1000 as shown in Figure 2, or other exemplary nuclear power plants. Therefore, it should be understood that the methods according to some exemplary embodiments may include methods for operating the exemplary nuclear power plant, which involves using the reactor of the exemplary nuclear power plant to generate electricity (e.g., heat, power, etc.).

[0066] Figure 3 illustrates the modeling of the physical structure of energy reactor 1000, or other physical structures. In other words, the equipment 900 can model the characteristics and features of a physical reactor. The physical structure can be modeled using a digital twin. A digital twin is a digital simulation using a model, which reduces operating and maintenance costs by supporting predictive maintenance while replacing costly and time-consuming physical testing.

[0067] The physical reactor model may include physical properties such as the size, material, service life, and estimated cumulative fatigue of various elements. Examples of elements in a physical reactor (or other physical structure) include the reactor, piping, support structures, connecting fittings, mounts, compressor blades, gears, chambers, walls, containment vessel, pumps, sensors, wiring, storage tanks, control devices, gates, valves, and generators. The sub-components of these elements can also be simulated. Elements can be simulated in great detail. For example, in the case of a pipe, the length, diameter, thickness, material, service life, and installation method can be simulated. In this way, the device 900 can generate a high-fidelity model of a physical reactor (or other physical structure). It is possible to simulate operating modes of the physical reactor, i.e., both (planned) operating modes and (transient) abnormal modes. Examples of operating modes include operation at different power levels (e.g., 10% power, 20% power, etc., in the case of a nuclear reactor) and transitions between power levels. Extreme transitions between power levels (e.g., from 20% to 100%, or vice versa) can result in abnormal modes being observed, such as brief anomalies in the observed signals, as the fluid flow takes time to reach a steady state at the new operating level. Rapid and unplanned transitions are also likely to cause similar effects.

[0068] The device 900 can be divided into modules including high-fidelity virtual models and diagnostic / predictive models, which are implemented by processor 920 by executing code that transforms processor 920 into a dedicated processor for performing the functions of these models. The purpose of the high-fidelity models is to generate a rich database with various operating modes and / or design parameters.

[0069] High-fidelity virtual models may include operation and prediction modules that can simulate the physical reactor in operating and abnormal modes. For example, processor 920 can execute code that transforms it into a special-purpose processor to perform the functions of the operation and prediction module, which includes generating a multiphysics model of the physical reactor, including a structural model of the physical reactor and a fluid flow model of the liquid moving within the physical reactor. This also makes it possible to launch a computational fluid dynamics (CFD) model that simulates multiphase flow through feedpipes, etc. In other words, device 900 can model the physical reactor in operating and abnormal modes according to the structural model of the physical reactor and the fluid flow model of the physical reactor. The operation and prediction module can also improve the modeling by incorporating model bias and uncertainty. For example, model bias and uncertainty can be quantified by comparison with experimental data and analyzed large-eddy simulation (LES) solutions. The operation and prediction module may also be a structure-based turbulence analysis (STRUCT) model.

[0070] Thermal mixing within reactor piping involves moderate to strong turbulent motion. Modeling turbulence is often computationally expensive. LES can construct accurate scale models, but at the cost of computation is very high. STRUCT can capture the physical phenomena in question by applying controlled resolution within a specific flow area, reducing operating costs by approximately half compared to LES. STRUCT excels at handling low-frequency data, which is a major cause of damage in real reactors such as power reactors. The discrepancies between models using LES and STRUCT and actual measurement results can introduce bias and uncertainty into the models.

[0071] The thermal load in the piping of a physical reactor can be calculated using a coupled computational fluid dynamics and finite element analysis. The finite element analysis can be performed under the assumption of isotropic linear elasticity and thermal expansion. Gaussian process regression models can quantify, propagate, and update uncertainties associated with the intrinsic model (e.g., a physical twin) by performing Bayesian inference.

[0072] A diagnostic and forecasting model may include measurement modules, state transition modules, and risk modules. For example, processor 920 can execute code that transforms it into a special-purpose processor to perform functions of the diagnostic and forecasting model, such as measurement modules, state transition modules, and risk modules.

[0073] Based on the results of the operation and prediction modes output from the high-fidelity virtual model, a generative model can be generated that includes stresses at model points in the digital twin of the physical reactor. The model points may be completely independent of the positions of sensors on the physical reactor. The model points can be placed at any location in the digital twin of the physical reactor and may be concentrated around areas where fatigue and damage are known to occur more quickly (e.g., joints where fluids mix, bends in piping, heat exchange areas, interfaces between different materials, etc.).

[0074] The generated model can be used in a state evolution module to simulate fatigue and damage to the physical structure at a model point based on the stress on the physical structure at that point. The state evolution module can also receive actual measurements from the physical reactor from a measurement module so that the simulation can be updated (for example, in real time). The measurement module can receive information directly or indirectly from sensors such as Sensor 500, 600, or controllers such as Controller 700.

[0075] The state change module can predict physical structural damage, or the possibility of physical structural damage, based on the planned operation of the physical reactor. Damage includes fatigue damage such as cracks, creep damage such as ruptures, oxidation damage such as chemical erosion, and wear damage such as friction damage due to the movement of parts. Fatigue damage is predicted based on the number of operating cycles that occur over time, while creep damage, oxidation damage, and wear damage are predicted based on the length of operating time.

[0076] For example, suppose that among the model points of the physical structure, model point 1 is a point on a pipe at a T-shaped joint of the physical reactor. Based on the stress at model point 1 in the generative model, the state evolution module predicts that a crack will occur at model point 1 after the physical reactor has been operating at 100% power for 10,000 hours. The state evolution module can also predict crack propagation and the timing at which the crack will impair the function of the physical structure.

[0077] Damage can accumulate due to a variety of causes, including temperature and temperature changes (as well as the expansion and compression of materials associated with temperature changes), pressure and pressure changes, motion (such as the rotation of compressor fans and generators), erosion (due to fluid movement), vibration, and radiation (from external sources such as nuclear fuel, heat, and sunlight). A physical reactor may incorporate different sensors to measure each of these damage sources depending on the location. A digital twin of a physical reactor can simulate damage at many additional locations on the digital twin, including locations where it is not practical to install sensors. The digital twin can be used to simulate damage at additional points based on detection data at detection points, and it is also possible to predict the accumulation of damage at detection points and additional points based on the planned operation of the physical reactor.

[0078] The risk module of the device 900 can determine possible damage and solutions to that damage based on the state assessment by the state assessment module, observable quantities, and unobservable quantities (which can be inferred from the observable quantities). Observable quantities may include measured values ​​and historical data. Unobservable quantities may include stress concentration factors or damage accumulation indices in critical areas. Risk values ​​are inferred by Bayesian inference based on observable and unobservable characteristics. Decisions can then be made based on the risk. For example, if the risk of a crack propagating to pipe failure exceeds a threshold, a decision is made to reduce the output of the physical reactor. This decision is fed back to the state evolution module to determine whether the risk factor falls below the threshold. In another embodiment, the decision becomes a work order or a recommendation for a work order to repair the physical structure. The decision of the diagnostic and forecasting model may be transmitted to the controller 700 of the energy plant 1000.

[0079] Figure 4 is a flowchart of the operation performed by the device 900. In S110, the device 900 can receive operational information regarding the operation of the physical structure (e.g., the physical structure of the energy plant 1000). The operational information may include different time instances of the operation of the physical structure. These time instances may be continuous (e.g., operation of the physical structure throughout the year or over other periods) or discontinuous (e.g., operation of the physical structure over discontinuous sampling periods). The operational information may also include different operating levels of the physical structure. For example, in the case of an energy plant, the operating levels may be different power output levels (including special operating levels such as power output levels for maintenance, testing, etc.). The power output levels may be spaced evenly, such as every 10% (e.g., 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 100% of the maximum power output level). The power output levels may be distributed at uneven intervals. For example, if the normal power output is approximately 75%, the power levels in the operating information may be concentrated around the normal power output to obtain more detailed information (e.g., 5%, 25%, 50%, 60%, 65%, 70%, 75%, 80%, 85%, 100%). When using continuous operating information, the operating levels can be divided based on the output levels in sub-periods of the operating period. In this way, the device 900 can acquire operating information for each operating level.

[0080] Operating information may be received in the device 900 from controllers and sensors of the physical structure. For example, the transmitter 900 receives sensor data from sensors 500 and 600 and control information from the controller 700. Sensor data may be received directly or indirectly from sensors 500 and 600. Sensor data may include temperature, vibration, fluid pressure, fluid velocity, repetition rate per minute, radiation level, or other measurable physical properties of the physical structure. Sensor data may be received directly or indirectly by the device 900. For example, if the device 900 is separate from the controller 700, the device 900 may receive sensor data directly from sensors 500 and 600 or indirectly via the controller 700. In some embodiments, the controller 700 and the device 900 may be the same device (for example, the controller 700 controls the operation of the energy plant 1000 and performs the operation described in relation to Figure 4). Control information may include control of the physical structure (e.g., the energy plant 1000). The control information may include the output level of the physical structure, and the sensor data may include the physical characteristics of the physical structure corresponding to the output level.

[0081] In S130, the device 900 predicts the accumulation of damage to the physical structure. Damage prediction can be made based on operating information, predicted operation of the physical structure at at least one of different operating levels, and at least one model of the physical structure. The at least one model may include a digital twin of the physical structure. The digital twin may be an electronic model of the physical structure. The digital twin may include the physical components of the physical structure. For example, in the case of an energy plant 1000, the digital twin may include digital representations of the reactor 100, pipes 110, 120, 130, turbine 200, generator 300, condenser 400, and other structural elements of the energy plant 1000.

[0082] Predicting damage to physical structures may involve assessing the condition of those structures. For example, if historical data includes the entire operating history of a physical structure and information on its repair, installation, and maintenance, the current state of the physical structure can be modeled. Knowing when specific repairs were performed, or when parts were scanned or inspected, allows for a more accurate prediction of its condition. The state of the physical structure can be updated whenever additional historical data is acquired, for example, by resetting the cumulative fatigue model for a part when that part is replaced. This additional historical data can be categorized into multiple categories, ranging from "measurements" from direct damage inspections to general operational information such as parts being replaced, reactor shutdowns, and changes in operating power levels. By inputting all of this into a digital twin, the digital twin can refine its predictions to better fit reality.

[0083] At least one of the models may include a Gaussian regression model or an autoregressive time series model.

[0084] As part of the modeling process, future predictions may be made by interpolating operating information of a physical structure across some, all, or combinations of different operating levels of the physical structure. For example, stresses associated with the operation of a physical structure at different operating levels can be predicted. For combinations of operating levels, the process may include linking the operating information to match the prediction period.

[0085] For example, if there is a plan to operate energy plant 1000 at 75% power during the day and 50% power at night for one month, it is possible to predict the stresses that will occur during that month due to these operating levels by linking the operating information for 75% and 50% power during the forecast period using at least one model. For example, stresses caused by thermal changes, liquid pressure, liquid movement, and radiation can be calculated. The predicted stresses for the month can be used to predict the damage that will accumulate in the physical structure at various points in the physical structure.

[0086] Figure 5 shows an example of linking time series of stress signals over an exemplary period to create a daily or annual operating routine.

[0087] Referring to Figure 5, the selection of the coupled operating level is related to the operation of the physical structure and can be changed on an hourly or daily basis to better accommodate the operation of the physical structure. Furthermore, the length of the coupled time can be adjusted to take into account downtime of the physical structure, such as maintenance work or power outages.

[0088] As shown in Figure 5, stress data is collected at various power levels (e.g., 50%, 80%, 90%, 100%) over short periods, such as 10 seconds. By linking this data together, it is possible to generate data over longer periods, such as one day or one year.

[0089] Returning to Figure 4, as mentioned earlier, the model points and the prediction of damage at each model point may be completely independent of the proximity of sensors to each model point. For example, regardless of the location of the flow sensor, at least one model can predict the flow rate at any point in the digital twin based on sensor data and the operating level of the physical structure. Similarly, regardless of the location of the temperature sensor, the temperature at any point in the physical structure can be modeled and predicted based on sensor data and the operating level of the physical structure. Predicted physical properties such as pressure, temperature, and flow rate may be used to model the accumulation of damage at any point on the digital twin of the physical structure.

[0090] The step of predicting damage to the physical structure may include performing a rainflow count (RC) using a simplified RC algorithm, counting the number of cycles in the operating information, analyzing the number of cycles until damage occurs, and calculating the accumulation of damage in each cycle using Equation 1 below. TIFF2026510735000007.tif19150(1)

[0091] In Equation 1, "D" is the damage rate, "k" is the number of stress levels, "ni" is the cumulative number of cycles at the i-th stress, and "N" is the number of stress levels. i " represents the average number of cycles until damage occurs at the i-th stress. i This may be obtained for each material being modeled. For example, if 316L stainless steel tubing is used, the N of 316L stainless steel may be obtained. i This can be obtained from the SN data of 316L stainless steel. The device 900 uses Equation 1 to predict the stress at the location where data is missing, and the N at each location. i Estimate the location where cracks / damage are most likely to occur (i.e., N i By selecting the position with the smallest value, you can identify the most important position.

[0092] Crack propagation can be described using Equation 2, which is based on Paris's law. TIFF2026510735000008.tif9150(2)

[0093] In equation 2, "a n " is the crack length in cycle n. The material factors "c" and "m" are obtained from data-calibrated material models, such as the fatigue crack growth (FCG) data calibration model for 316L stainless steel. Effective stress intensity factor K eff The stress level (σ n Based on the crack shape, it can be calculated using Equation 3. TIFF2026510735000009.tif18150(3)

[0094] In equation 3, "σ n,min " is the minimum stress, "σ n,max " refers to the maximum stress. First, the stress tensor of 675-point probes is collected from finite element analysis. Next, the stress level and cycle number are obtained from a simplified rainflow count algorithm. For each probe, the number of cycles required for crack initiation (N) iThe calculation is performed. To address data loss due to coarse probe settings, a Gaussian process surrogate is used to make predictions for positions at any position and angle on the pipe, for example. From given conditions (current cumulative cycle count and power level), the number of cycles required for crack initiation can be predicted.

[0095] In Equation 3, the stress intensity factor K(σ n ,a n The stress intensity factor K(σ) can be calculated in various ways, and the calculation may be based on the shape of the initial crack and the shape of the structure under study. For example, assuming that a surface crack with a size of 1 / 10 of the pipe thickness already exists, the stress intensity factor K(σ) can be calculated in various ways. n ,a n ) can be calculated, for example, using the following equations 4 to 6, but exemplary embodiments are not limited to these. TIFF2026510735000010.tif14150(4) TIFF2026510735000011.tif11150(5) TIFF2026510735000012.tif14150(6)

[0096] In equations 4-6, the membrane stress σm and bending stress σb can be obtained from the stress tensor of finite element analysis (FEA). Mm and Mb may also be derived from the American Society of Mechanical Engineers (ASME) Boiler and Pressure Vessel Standard (BPVC) BPVC.XI.1-2013. These modeling techniques can be applied to multiple model points within a digital twin of the physical structure to model the accumulation of fatigue and damage at these model points.

[0097] In S150, the apparatus 900 can identify at least one critical location on the physical structure based on at least one model. To determine whether a model point on the digital twin of the physical structure is a critical location, cumulative fatigue and / or damage thresholds may be used. For example, a fatigue prediction threshold where the probability of crack initiation is 80% may be used as the threshold. Alternatively, in another embodiment, a predicted damage threshold for cracks exceeding 10 cm may be used as the threshold.

[0098] Predicted damage may include surface cracks and / or subsurface cracks. It is possible to predict whether cracks will occur in pipes, fittings, seals, support structures, etc. Other types of damage, such as generator failure and sensor detachment, can also be predicted.

[0099] In S170, the device 900 can generate one or more outputs based on the expected accumulation of fatigue and damage at identified critical locations. For example, the device 900 can generate work instructions and / or alarms based on identified critical locations. Alarms may be display alarms, audible alarms, written alarms, and / or alarms transmitted by electronic means (e.g., email, warning messages, or other communications to other devices). Work instructions may be transmitted by display, written, or electronic means. Work instructions may indicate critical locations and necessary repairs (e.g., visual or sensory damage assessment, patching, replacement).

[0100] Alternatively, or additionally, the output may include controlling devices that change the state of the physical structure based on identified critical locations and at least one model. For example, if the physical structure continues to operate at 100% power and at least one model predicts a crack larger than 10 cm, a command is sent from the device 900 to the controller 700 to reduce the operating level to 70% until actual damage can be assessed and / or repaired. Thus, in addition to controlling the device, work instructions may be generated.

[0101] New operational information is received in real time or at regular intervals (such as daily or hourly), and at least one model, predicted fatigue and damage, and output may be updated based on the new operational information. For example, if the operational level of the physical structure decreases (from the planned or predicted operational level) due to lower-than-expected demand, the received operational information will reflect the lower operational level, and the predicted fatigue and damage will be adjusted to be lower than previously predicted based on that operational information. In another embodiment, if the operational level of the physical structure increases due to increased demand or the offlineing of other facilities, the received operational information will reflect the increased operational level, and the predicted fatigue and damage will be adjusted to be higher than previously predicted based on that operational information. In other words, identified critical locations and outputs may be updated or modified based on the received operational information.

[0102] Figure 6 shows an example of historical data at different power levels. This historical data graphs data related to the operation of the physical structure at 20%, 50%, 80%, 90%, and 100% power levels. The historical data can be summarized in terms of the relationship between time and stress. Stress is a measure of force, temperature change, or other form of stress acting on the physical structure.

[0103] Figure 7 shows examples of simulated stresses at different power levels. The stresses of a physical structure operating at a specific power level can be predicted based on historical data. For example, predictions are possible at lower power levels such as 85%, 70%, 55%, and 40%. This prediction may be performed by interpreting and interpolating historical data from multiple power levels using an autoregressive time series model. For example, stress levels at 80% and 90% can be used to predict the stress at 85%.

[0104] Figure 8 shows an example of rainflow counts and detected cycles. Stress at a given location can be measured as pressure (Pascals) over several seconds. Cycles are identified by boxes on the pressure graph. Cycles can be identified, for example, by identifying four turning points (A, B, C, D) in the signal (where the sign of the gradient changes, i.e., where the stress trend changes from downward to upward or vice versa), considering period A-D as a rainflow cycle, and processing the rest of the signal.

[0105] Figure 9 illustrates an example of quantifying uncertainty for identifying critical locations for crack initiation. The graph shows the number of cycles required for crack initiation at a given location at different angles on the pipe. Lower values ​​indicate fewer cycles required for crack initiation (i.e., a more dangerous situation). The zero-degree location can be selected as the location of the modeled high-stress region. Observed data are shown with asterisks, average data with a solid line, and confidence levels with shaded areas.

[0106] Figure 10 shows an example of crack propagation over a time cycle. This graph shows the relationship between the number of cycles and crack length. As shown in the figure, the crack length increases with increasing cycle number. The average crack propagation curve (solid line) and the uncertainty zone around it (shaded area around the solid line) are also shown. For these uncertainty zones, the crack initiation stage (N i In addition to the uncertainty arising from the uncertainty of the initial crack size, material properties, and noise term (see Figure 9), quantification is performed by considering the uncertainty of the initial crack size. Using a particle forward scattering approach, a sample is extracted from the probability distribution of the uncertain parameters and propagated during the crack propagation phase. As a result, crack propagation curves with multiple velocities / slopes are obtained. The confidence / uncertainty band shown in Figure 10 represents the standard deviation of these multiple crack propagation curves.

[0107] As mentioned above, due to the high frequency and complex nature of turbulence, it can be difficult to evaluate damage caused by flow-induced thermal fatigue using in-plant measuring instruments. Therefore, the high-fidelity digital twin in the exemplary embodiment, by using STRUCT as the basis for the high-fidelity digital twin, enables the development of predictive and diagnostic maintenance approaches. The exemplary embodiment can use STRUCT to perform multiscale, multiphysics simulations and evaluate thermal mixing. By integrating the high-fidelity virtual model with diagnostic and predictive models, the initiation and propagation of cracks in components can be calculated. This allows for the evaluation and monitoring of crack propagation at all locations and angles of the component under predetermined operating conditions, and this information can be used to make decisions regarding operation and maintenance activities.

[0108] While this application discloses several exemplary embodiments, it should be understood that other modifications are possible. Such modifications should not be considered departures from the spirit and scope of the concept of the invention, and all changes that are obvious to those skilled in the art are included in the appended claims. Furthermore, while this application discloses a process, it should be understood that the elements described in the process may be carried out using different orders, different selections of elements, or combinations thereof. For example, some exemplary embodiments of the disclosed process may be carried out using fewer or more elements than the illustrated and described process.

Claims

1. A method for identifying at least one important location on a physical structure, The process involves receiving operating information, the operating information including, as information relating to the operation of the physical structure, different points in time and different operating levels of the physical structure, the operating level relating to the output level from the physical structure, and at least a portion of the operating information being received from a sensor that senses the state of the physical structure. A step in which damage to the physical structure is predicted based on the operating information, the predictive operation of the physical structure at at least one of the different operating levels, and at least one model of the physical structure, and the occurrence of damage at multiple locations of the physical structure is predicted regardless of the proximity of the sensor to each of the multiple locations, A method comprising the step of identifying at least one important location on the physical structure based on the predicted damage.

2. The method according to claim 1, further comprising the step of generating a work order or alarm based on the at least one important location and the at least one model.

3. The method according to claim 1, wherein the at least one model includes one or more machine learning regression models or machine learning time series models.

4. The aforementioned operating information includes available time-series data indicating stress or material temperature within the physical structure, and unavailable time-series data where the stress or material temperature within the physical structure is unknown. The step of predicting damage to the physical structure is: Interpolating the aforementioned operating information, additional operating information is predicted for the same location or angle on the physical structure across different operating levels of the physical structure associated with the unavailable time-series data. By linking the aforementioned driving information with the aforementioned additional driving information over a specified forecast period, and generating complete driving information, The method according to claim 1, comprising the step of predicting initial damage to the physical structure.

5. The step of predicting initial damage to the physical structure is: The steps include: performing a rainflow count (RC) using a simplified RC algorithm to count the number of cycles in the complete operating information; A step of estimating the number of cycles until damage occurs at the location or angle of the physical structure where the aforementioned operating information or the aforementioned additional operating information is available, The steps include: predicting the number of cycles until damage occurs at different locations or angles where complete operational information is unavailable by using a machine learning model to quantify the level of uncertainty in the number of cycles until damage occurs; The damage rate in each cycle of the aforementioned number of cycles in the complete operating information is The steps include, "D" is the damage rate, "k" is the number of stress levels, and "n" is the number of stress levels. ｉ " is the cumulative number of cycles, "N ｉ The method according to claim 4, wherein '' is the number of cycles until damage occurs at the i-th stress.

6. The step of predicting damage to the physical structure is further based on manipulating a digital twin of the physical structure, the digital twin being an electronically generated model of the physical structure. The method according to claim 4, wherein the initial damage comprises at least one of fatigue damage, creep damage, oxidation damage, or wear damage of the physical structure, and the fatigue damage comprises a surface crack or subsurface crack of the physical structure.

7. The step of predicting damage to the physical structure further includes predicting the growth of initial damage to the physical structure over time based on the following formula: Here, "a ｎ " is the crack length in cycle n, and includes the uncertainty level regarding the number of cycles until damage occurs, and "c" and "m" are material coefficients that include the uncertainty level regarding the physical structure, and "K ｅｆｆ " is the stress level (σ ｎ The effective stress intensity factor is calculated using the following formula based on the crack shape and the crack geometry. Here, "σ ｎ，ｍｉｎ " is the minimum stress, "σ ｎ，ｍａｘ " is the maximum stress, The method according to claim 4, wherein is the shape of the initial damage to the physical structure and the stress intensity factor that changes based on the physical structure.

8. The physical structure includes a nuclear power plant further comprising a reactor, the operating level is the power level of the reactor, the operating information includes details of the repair and installation of the reactor, and the reactor state includes temperature data and flow rate data. The above method further, The steps include updating the state of the reactor at at least one important location, The method according to claim 1, comprising the step of updating the at least one model based on the updated state of the reactor at the at least one important location.

9. The method according to claim 1, further comprising the step of controlling the apparatus to change the state in the physical structure based on the at least one important location and the at least one model.

10. A device configured to identify at least one important location on a physical structure, Includes processing circuitry, The aforementioned processing circuit is Operating information is received, and the operating information includes, as information relating to the operation of the physical structure, different time points in time and different operating levels of the physical structure, the operating level relating to the output level from the physical structure, and at least a portion of the operating information is received from a sensor that senses the state of the physical structure. Based on the aforementioned operating information, the predictive operation of the physical structure at at least one of the different operating levels, and at least one model of the physical structure, damage to the physical structure is predicted, and the occurrence of damage at multiple locations of the physical structure is predicted regardless of the proximity of the sensor to each of the multiple locations. A device configured to identify at least one important location on the physical structure based on the predicted damage.

11. The apparatus according to claim 10, wherein the processing circuit is further configured to generate work instructions or alarms based on the at least one important location and the at least one model.

12. The apparatus according to claim 10, wherein the at least one model includes one or more machine learning regression models or machine learning time series models.

13. The aforementioned operating information includes available time-series data indicating stress or material temperature within the physical structure, and unavailable time-series data where the stress or material temperature within the physical structure is unknown. The aforementioned processing circuit is Interpolating the aforementioned operating information, additional operating information is predicted for the same location or angle on the physical structure across different operating levels of the physical structure associated with the unavailable time-series data. By linking the aforementioned driving information with the aforementioned additional driving information over a specified forecast period, and generating complete driving information, The apparatus according to claim 10, configured to predict at least initial damage to the physical structure and to predict damage to the physical structure.

14. The processing circuit, in predicting initial damage to the physical structure, further, A simplified RC algorithm is used to perform a rainflow count (RC) and count the number of cycles in the complete operating information. The number of cycles until damage occurs at the location or angle of the physical structure where the aforementioned operating information or the aforementioned additional operating information is available is estimated. The number of cycles until damage occurs at different locations or angles where complete operational information is unavailable is predicted by quantifying the level of uncertainty in the number of cycles until damage occurs using a machine learning model. The damage rate in each cycle of the aforementioned number of cycles in the complete operating information is It is configured to calculate by, "D" is the damage rate, "k" is the number of stress levels, "n" ｉ " is the number of cumulative cycles, "N" ｉ " is the number of cycles until damage occurs at the i-th stress, the apparatus according to claim 13.

15. The prediction of damage to the physical structure is further based on manipulating a digital twin of the physical structure, the digital twin being an electronically generated model of the physical structure. The apparatus according to claim 13, wherein the initial damage includes at least one of fatigue damage, creep damage, oxidation damage, or wear damage of the physical structure, and the fatigue damage includes a surface crack or subsurface crack of the physical structure.

Images