Device performance evaluation device, device performance evaluation method, and device performance evaluation program

Abstract

A performance evaluation device for a device comprises: a total performance function acquisition unit configured to obtain a total performance function representing a temporal change in a performance indicator of the device to be evaluated based on data acquired during the operation of the device to be evaluated; an individual performance function definition unit configured to define a plurality of individual performance functions representing a temporal change in the performance indicator, each caused by a plurality of change factors of the performance indicator; and a model creation unit configured to create a performance estimation model of the device to be evaluated by performing a superposition of the plurality of individual performance functions. The model creation unit is configured to determine the coefficients of each of the plurality of individual performance functions in the superposition so that the superposition of the plurality of individual performance functions approaches the total performance function.

Description

Technology Field

[0001] The present disclosure relates to a device for evaluating the performance of a device, a method for evaluating the performance of a device, and a program for evaluating the performance of a device.

[0002] The present application claims priority based on patent application No. 2021-153785 filed with the Japan Patent Office on September 22, 2021, and incorporates the contents thereof herein by reference. Background Technology

[0003] In plants such as power generation plants, there are cases where the operation of the plant or equipment is managed by using predicted results regarding the performance of the plant or its constituent equipment.

[0004] Patent Document 1 discloses a method for predicting the performance of a power generation plant using a hybrid prediction model comprising a static physical base model and a corrector model that corrects the prediction based on the physical base model based on data collected from the plant. In the method of Patent Document 1, the corrector model is trained using the latest plant operation data so that performance changes accompanying equipment deterioration or the updating of control mechanisms are reflected in the performance prediction. Prior art literature

[0005] Japanese Patent Publication No. 2012-079304 The problem to be solved

[0006] However, there are often multiple factors contributing to performance changes (such as performance degradation) in plants or their constituent equipment. Yet, conventional prediction methods predict performance changes as a whole for the plant or equipment, and therefore cannot estimate the specific factors causing the performance change.

[0007] Taking into account the circumstances described above, at least one embodiment of the present invention aims to provide a device performance evaluation device, a device performance evaluation method, and a device performance evaluation program that can quantitatively estimate performance changes over time with respect to a device subject to evaluation, and can also estimate the factors of performance changes. means of solving the problem

[0008] A performance evaluation device for a device according to at least one embodiment of the present invention is,

[0009] A total performance function acquisition unit configured to obtain a total performance function representing the temporal change of a performance indicator of the device under evaluation based on data acquired during the operation of the device under evaluation, and

[0010] An individual performance function definition unit configured to define a plurality of individual performance functions each representing a temporal change of the performance indicator caused by a plurality of change factors of the performance indicator, and

[0011] A model generation unit configured to generate a performance estimation model of the device to be evaluated by performing the superposition of the plurality of individual performance functions, and

[0012] The above model creation unit is configured to determine each coefficient of the plurality of individual performance functions in the overlap such that the overlap of the plurality of individual performance functions approximates the overall performance function.

[0013] In addition, a method for evaluating the performance of a device according to at least one embodiment of the present invention is,

[0014] A step of obtaining an overall performance function representing the temporal change of a performance indicator of a device under evaluation based on data acquired during the operation of the device under evaluation, and

[0015] A step of defining a plurality of individual performance functions each representing a temporal change of the performance indicator attributable to a plurality of change factors of the performance indicator, and

[0016] The method comprises a step of creating a performance estimation model of the device to be evaluated by performing the superposition of the plurality of individual performance functions.

[0017] In the step of creating the above performance estimation model, the coefficients of each of the plurality of individual performance functions in the overlap are determined such that the overlap of the plurality of individual performance functions approximates the overall performance function.

[0018] In addition, a performance evaluation program of a device according to at least one embodiment of the present invention is,

[0019] on the computer,

[0020] A procedure for obtaining an overall performance function representing the temporal change of a performance indicator of a device under evaluation based on data acquired during the operation of the device under evaluation, and

[0021] A procedure for defining a plurality of individual performance functions, each representing a temporal change of the performance indicator attributable to a plurality of change factors of the performance indicator, and

[0022] A procedure to create a performance estimation model of the device to be evaluated by performing the superposition of the above plurality of individual performance functions is executed, and

[0023] In the procedure for constructing the above performance estimation model, the coefficients of each of the plurality of individual performance functions in the overlap are determined such that the overlap of the plurality of individual performance functions approximates the overall performance function. Effects of the invention

[0024] According to at least one embodiment of the present invention, a device performance evaluation device, a device performance evaluation method, and a device performance evaluation program are provided, which can quantitatively estimate performance changes over time for a device subject to evaluation and estimate the factors of performance changes. Brief explanation of the drawing

[0025] FIG. 1 is a schematic diagram of an evaluation target device (turbine) in several embodiments. FIG. 2 is a schematic diagram of a performance evaluation device for a device according to one embodiment. FIG. 3 is a flowchart of a method for evaluating the performance of a device according to one embodiment. Figure 4 is a graph showing an example of the change in internal turbine efficiency over time. Figure 5 is a graph showing an example of the overall performance function F(t). Figure 6 is a graph schematically showing an example of an individual performance function. Figure 7 is a diagram schematically illustrating an example of the first correlation. FIG. 8 is a diagram schematically illustrating an example of a second correlation. Figure 9 is a diagram illustrating a method for determining the second correlation. Figure 10 is a diagram illustrating a method for determining the second correlation. Figure 11 is a schematic graph visually illustrating an example of a performance estimation model. Figure 12 is a schematic graph visually illustrating an example of a performance estimation model. Specific details for implementing the invention

[0026] Hereinafter, several embodiments of the present invention will be described with reference to the attached drawings. However, the dimensions, materials, shapes, and relative arrangements of the components described as embodiments or illustrated in the drawings are not intended to limit the scope of the present invention and are merely illustrative examples.

[0027] (Configuration of the performance evaluation device)

[0028] FIG. 1 is a schematic diagram of a turbine (2), which is an example of a device to be evaluated by a performance evaluation device according to several embodiments. FIG. 2 is a schematic diagram of a performance evaluation device according to one embodiment.

[0029] The device subject to evaluation by the performance evaluation device regarding several embodiments (the device subject to evaluation) may be the whole or part of a plant, or a device constituting the plant or a part thereof.

[0030] The turbine (2) illustrated in FIG. 1 is configured to be driven by a working fluid. The working fluid is introduced into the turbine (2) through the turbine inlet. Additionally, the working fluid, after completing work in the turbine (2), is discharged from the turbine (2) through the turbine outlet. A generator may be connected to the rotating shaft of the turbine (2).

[0031] The turbine (2) may be a steam turbine configured to be driven by steam. Alternatively, the turbine (2) may be a gas turbine configured to be driven by gas produced by the combustion of fuel.

[0032] The turbine (2) illustrated in FIG. 1 includes an upstream step section (2a), which is an upstream portion in the direction of flow of the working fluid, and a downstream step section (2b), which is a downstream portion of the upstream step section (2a), among the multiple stages of the turbine blades. In one embodiment, the turbine (2) may be configured to extract the working fluid from the upstream step section (2a) and the downstream step section (2b).

[0033] In one embodiment, a plant including a turbine (2) may be the equipment to be evaluated. In one embodiment, the turbine (2) may be the equipment to be evaluated. In one embodiment, the upstream stepped portion (2a) or downstream stepped portion (2b) of the turbine (2) may be the equipment to be evaluated.

[0034] The performance evaluation device (20) illustrated in FIG. 2 is configured to evaluate the performance of a device to be evaluated by processing information obtained from a measurement unit (12) and / or a memory unit (14).

[0035] The measurement unit (12) is configured to measure parameters regarding the performance indicators of the device to be evaluated.

[0036] As a performance indicator of the turbine (2) as the device to be evaluated, for example, the internal efficiency of the turbine may be adopted. In this case, the measuring unit (12) may include a plurality of sensors for measuring the pressure P1 and temperature T1 at the turbine inlet, and the pressure P2 and temperature T2 at the turbine outlet, respectively. The internal efficiency of the turbine can be calculated from the measured values ​​of these parameters.

[0037] When the device to be evaluated is the upstream step section (2a) of the turbine (2), the internal turbine efficiency of the upstream step section (2a) may be used as a performance indicator. In this case, the measuring unit (12) may include a plurality of sensors for measuring the pressure P1 and temperature T1 at the turbine inlet, and the pressure Pi and temperature Ti between the upstream step section (2a) and the downstream step section (2b).

[0038] When the device to be evaluated is the downstream step section (2b) of the turbine (2), the internal efficiency of the turbine in the downstream step section (2b) may be used as a performance indicator. In this case, the measuring unit (12) may include a plurality of sensors for measuring the pressure Pi and temperature Ti between the upstream step section (2a) and the downstream step section (2b), and the pressure P2 and temperature T2 at the turbine outlet, respectively.

[0039] If the equipment subject to evaluation is the entire plant, generator output or heat consumption rate may be used as performance indicators for the equipment subject to evaluation.

[0040] The performance evaluation device (20) is configured to receive a signal from the measurement unit (12) indicating a measured value of a parameter regarding a performance indicator. The performance evaluation device (20) may also be configured to receive a signal indicating a measured value from the measurement unit (12) at a specified sampling period. Additionally, the performance evaluation device (20) is configured to process the signal received from the measurement unit (12) to evaluate the performance of the device to be evaluated. The evaluation result by the performance evaluation device (20) may be displayed on a display unit (16) (display, etc.).

[0041] As illustrated in FIG. 2, a performance evaluation device (20) according to one embodiment includes an overall performance function acquisition unit (22), an individual performance function definition unit (24), a model creation unit (26), and an evaluation unit (28).

[0042] The performance evaluation device (20) includes a calculator equipped with a processor (CPU, etc.), a memory device (memory device; RAM, etc.), an auxiliary memory unit, and an interface. The performance evaluation device (20) is configured to receive a signal indicating the measured value of a parameter regarding the performance indicator of the device to be evaluated from the measurement unit (12) through the interface. The processor is configured to process the signal received in this manner. Additionally, the processor is configured to process a program deployed in the memory unit. By doing so, the functions of each of the above-described functional units (the overall performance function acquisition unit (22), etc.) are realized.

[0043] The processing content in the performance evaluation device (20) is implemented as a program executed by a processor. The program may be stored in an auxiliary memory. When the program is executed, these programs are deployed in the memory. The processor reads the program from the memory and executes the instructions included in the program.

[0044] The overall performance function acquisition unit (22) is configured to obtain an overall performance function that represents the change in performance indicators of the device being evaluated over time, based on data acquired during the operation of the device being evaluated (data acquired by the measurement unit (12)).

[0045] The individual performance function definition unit (24) is configured to define a plurality of individual performance functions that represent a change over time of a performance indicator caused by a plurality of change factors of the performance indicator, respectively.

[0046] The model creation unit (26) is configured to create a performance estimation model of the device to be evaluated by performing a superposition of a plurality of individual performance functions defined in the individual performance function definition unit (24). Additionally, the model creation unit (26) is configured to determine the coefficients of each of the plurality of individual performance functions in the superposition described above so that the superposition of the plurality of individual performance functions approximates the overall performance function obtained by the overall performance function acquisition unit (22).

[0047] The evaluation unit (28) is configured to estimate the performance change over time of the device to be evaluated based on the performance estimation model created by the model creation unit (26). Alternatively, the evaluation unit (28) is configured to estimate the factors of the performance change over time of the device to be evaluated.

[0048] (Flow of device performance evaluation)

[0049] Below, a method for evaluating the performance of a device in several embodiments is described. Furthermore, below, the method for evaluating the performance of a device in one embodiment is described using the performance evaluation device (20) described above; however, in several embodiments, the method for evaluating the performance of a device may be executed using a different device. In the following description, the device to be evaluated is a turbine (2). Furthermore, below, the method for evaluating the performance deterioration of the device to be evaluated is described.

[0050] FIG. 3 is a flowchart of a method for evaluating the performance of a device according to several embodiments. FIGS. 4 to 12 are drawings for explaining a method for evaluating the performance of a device according to several embodiments.

[0051] As illustrated in FIG. 3, in some embodiments, first, data regarding the performance indicators of the turbine (2) is acquired during the operation of the turbine (2) which is the device to be evaluated (S2).

[0052] In step S2, first, measurement data of parameters regarding the performance indicator of the turbine (2) (device to be evaluated) is acquired by the measurement unit (12). Here, the turbine internal efficiency is used as the performance indicator of the turbine (2), and measurement data of pressure P1 and temperature T1 at the turbine inlet and pressure P2 and temperature T2 at the turbine outlet are acquired respectively as parameters regarding the turbine internal efficiency. The measurement values ​​of these parameters are acquired repeatedly over time.

[0053] Then, based on the measurement data of the parameters described above, the internal efficiency of the turbine as a performance indicator is calculated. FIG. 4 is a graph showing an example of the change over time (performance deterioration) of the internal efficiency of the turbine (2) obtained in this way.

[0054] Next, the overall performance function acquisition unit (22) obtains an overall performance function representing the temporal change of the turbine internal efficiency (performance indicator) of the turbine (2) based on the data obtained in step S2. The overall performance function can be obtained by applying a general time series model to the performance indicator data obtained in step S2 (see FIG. 4). As a method for applying a general time series model to the performance indicator data, the Holt-Winsters method, ARIMA method, SARIMA method, Gaussian process regression, or Prophet can be used. FIG. 5 is a graph showing an example of the overall performance function F(t) obtained in this way for the turbine internal efficiency data obtained in step S2.

[0055] Next, the individual performance function definition unit (24) defines a plurality of individual performance functions each representing a change over time (e.g., performance degradation) of the corresponding performance indicator caused by a plurality of change factors of the performance indicator (here, turbine internal efficiency) described above (S6).

[0056] Regarding multiple factors of change in performance indicators, those assumed in advance are set. Factors of performance deterioration of the turbine (2) include widening of the clearance between the turbine blade and the casing, deterioration of the surface roughness of the turbine blade surface, widening of the erosion of the turbine blade, increased leakage of the working fluid (steam, etc.), and reduction of the flow path cross-sectional area due to scale deposition.

[0057] In this embodiment, four factors are established as factors for performance degradation of the internal efficiency of the turbine: widening of the clearance between the turbine blade and the casing (hereinafter, clearance), deterioration of the surface roughness of the turbine blade surface (hereinafter, blade roughness), degree of erosion of the turbine blade (hereinafter, erosion), and other factors (hereinafter, other factors). Additionally, an individual performance function representing performance degradation caused by clearance is denoted as A(t), an individual performance function representing performance degradation caused by clearance is denoted as B(t), an individual performance function representing performance degradation caused by erosion is denoted as C(t), and an individual performance function representing performance degradation caused by other factors is denoted as etc(t).

[0058] Here, with reference to FIGS. 6 to 10, a method for defining individual performance functions according to several embodiments is described. Here, as an example, a method for defining an individual performance function A(t) representing a temporal change in performance (performance degradation) caused by the expansion of clearance as a performance degradation factor is described.

[0059] In some embodiments, at step S6, for a change factor in the performance indicator (performance degradation factor; here, expansion of clearance), an individual performance function is defined based on the theoretical or measured value of a parameter (e.g., clearance value) related to the change factor. FIG. 6 is a graph schematically illustrating an example of an individual performance function A(t) for a change (performance degradation) in turbine internal efficiency (performance indicator) caused by the expansion of clearance (performance degradation factor) obtained at step S6.

[0060] More specifically, in step S6, an individual performance function is defined based on a first correlation between the value of the aforementioned parameter and the magnitude of the performance change of the device under evaluation, and a second correlation between the value of the aforementioned parameter and time. The aforementioned first correlation and second correlation can be obtained, for example, by the procedure described below. In step S6, by combining the first correlation and second correlation obtained in this way, an individual performance function representing the time change (performance deterioration) of the performance indicator regarding the parameter can be obtained. Additionally, the first correlation and second correlation may be stored in advance in the memory unit (14).

[0061] Here, FIG. 7 is a diagram schematically illustrating an example of a first correlation. The first correlation illustrated in FIG. 7 is the correlation between a clearance value (horizontal axis) as the value of the parameter described above and a loss (vertical axis) as the magnitude of the performance change described above. That is, the graph in FIG. 7 is a curve (loss increase curve) representing the amount of loss increasing with the expansion of clearance. The first correlation can be obtained, for example, from design data (theoretical value).

[0062] FIG. 8 is a diagram schematically illustrating an example of a second correlation. The second correlation shown in FIG. 8 is the correlation between the clearance value (vertical axis) as the value of the parameter described above and time (horizontal axis). That is, the graph in FIG. 8 is a curve (time-dependent change in physical quantity curve) showing the change (increase) of clearance over time.

[0063] The aforementioned second correlation can be obtained based on actual values ​​from the device under evaluation or through a literature review. For example, in one embodiment, the second correlation is obtained by fitting a base curve having the shape of the correlation between the value of the parameter and time to the actual values ​​of the aforementioned parameter (clearance, etc.). As the base curve, a curve of a general function (such as an exponential function) may be used.

[0064] FIGS. 9 and FIGS. 10 are drawings for illustrating an example of a method for obtaining a second correlation, respectively.

[0065] In the example illustrated in FIG. 9, a second correlation (102) is obtained by fitting a base curve (100) to the measured values ​​M1 to M3 of the above-described parameters (here, clearance) in the device (turbine (2)) to be evaluated. The measured values ​​M1 to M3 of the parameters in the device to be evaluated can be obtained, for example, during a periodic inspection. Additionally, the fitting of the base curve (100) to the measured values ​​M1 to M3 may be performed by determining the coefficients of a function representing the base curve (i.e., by the least squares method) so that the total sum of the distances between the measured values ​​M1 to M3 and the second correlation (102) based on the base curve (100) is minimized.

[0066] In the example illustrated in FIG. 10, curves (e.g., Q1 to Q3) representing the relationship between the measured values ​​of the above-described parameters (here, clearance) and time in a plurality of plants (other plants similar to the plant containing the equipment to be evaluated) are obtained. Additionally, these curves (Q1 to Q3) may be obtained by the method described with reference to FIG. 9. Furthermore, the curves Q1 to Q3 based on the measured values ​​and the base curve (100) may be fitted by determining the coefficients of a function representing the base curve so that the curve of the second correlation relationship (104) based on the base curve (100) approximates these curves (Q1 to Q3).

[0067] In addition, the measured values ​​of the parameters (e.g., P1 to P3) or the curves (Q1 to Q3) based on the measured values ​​may be stored in advance in the memory unit (14).

[0068] By following the above procedure, individual performance functions A(t), B(t), C(t) and etc(t) representing performance degradation caused by each of the above-described multiple performance degradation factors (clearance, wing roughness, erosion, and other factors) can be defined.

[0069] In addition, regarding the individual performance function etc(t) corresponding to other factors, provided that other individual performance functions A(t), B(t), and C(t) have already been obtained, it may be defined as follows. Here, FIGS. 11 and FIGS. 12 are schematic graphs that visually represent an example of a performance estimation model. In FIGS. 11 and FIGS. 12, F(t) represents the overall performance function obtained in step S4, and G(t) represents the sum of the individual performance functions A(t), B(t), and C(t) (i.e., G(t) = A(t) + B(t) + C(t)).

[0070] As illustrated in FIG. 11, when the minimum value of (F(t)-G(t)) at all times is greater than or equal to zero (when min(F(t)-G(t))≥0), the individual performance function etc(t) can be defined as the difference between F(t) and G(t) (Equation (B) below).

[0071] etc(t)=F(t)-G(t) … (B)

[0072] Meanwhile, as illustrated in FIG. 12, when there exists a time when the minimum value of (F(t)-G(t)) is less than zero (when min(F(t)-G(t))<0), the individual performance function etc(t) can be defined by the following equation (C).

[0073] etc(t)=F(t)-α×G(t) … (C)

[0074] Here, α in the above equation (C) is minimized when (F(t)-G(t)) is minimized (t=t min It is the ratio of F(t) and G(t) of ). That is, α can be expressed by the following equation (D).

[0075] α=F(tmin ) / G(t min ) … (D)

[0076] Next, the model creation unit (26) creates a performance estimation model of the turbine (2) (device to be evaluated) by performing a superposition of the plurality of individual performance functions A(t), B(t), C(t) and etc(t) defined in step S6 (S8).

[0077] The superposition of multiple individual performance functions in Step S8 may be performed by a statistical superposition method. As a statistical superposition method, a generalized linear model or a generalized additive model may be used.

[0078] The performance estimation model described above can be expressed, for example, as a linear combination of multiple individual performance functions as Equation (A) below. In Equation (A) below, n1 to n4 are coefficients of multiple individual performance functions (A(t), B(t), C(t), etc(t)).

[0079] F'(t)=n1×A(t)+n2×B(t)+n3×C(t)+n4×etc(t) … (A)

[0080] In the above-described step S8, the model creation unit (26) determines each coefficient of a plurality of individual performance functions in the above-described overlap (e.g., coefficients n1 to n4 in the above-described equation (A)) so that the overlap of a plurality of individual performance functions approaches the overall performance function.

[0081] The determination of each coefficient of a plurality of individual performance functions in Step S8 may be performed using methods such as multiple regression analysis, Bayes estimation, or neural networks (such as LSTM (Long Short-Term Memory)).

[0082] Next, based on the performance estimation model created in step S8, the performance degradation of the turbine (2) (device to be evaluated) is quantitatively estimated, or the factors of performance degradation are estimated (S10). From the performance estimation model, the trend of performance degradation based on each of the multiple factors of performance degradation, or the contribution of each factor to the overall trend of performance degradation, can be determined. Based on this, the performance degradation of the turbine (2) (device to be evaluated) can be quantitatively estimated, or the factors of performance degradation can be estimated.

[0083] According to the above-described embodiment, coefficients (n1 to n4) in the superposition of individual performance functions (A(t), B(t), C(t), etc(t)) representing performance changes attributable to each of the multiple change factors (performance change factors) of the performance indicator are determined (i.e., weights are assigned) so as to approximate the overall performance function (F(t)) representing the time-dependent change of the performance indicator (e.g., internal efficiency of the turbine) of the turbine (2) (device to be evaluated). By doing so, a performance estimation model (F'(t)) including the respective weights of the multiple performance change factors can be created. Accordingly, for the turbine (2) (device to be evaluated), it is possible to quantitatively estimate the time-dependent performance change and to estimate the factors of the performance change.

[0084] In addition, as described above, in Step S6, for each of the plurality of performance change factors, an individual performance function may be determined based on the theoretical or measured value of the parameter related to the corresponding change factor. In this case, the accuracy of the performance change estimation of the device under evaluation is improved.

[0085] In addition, as described above, in Step S6, individual performance functions may be determined based on a first correlation between the value of a parameter related to a performance change factor and the magnitude of the performance change, and a second correlation between the value of the parameter and time. By doing so, individual performance functions for each performance change factor can be appropriately defined. Consequently, the accuracy of the performance change estimation of the device under evaluation is improved.

[0086] In addition, as described above, in Step S6, a second correlation may be obtained by fitting a base curve having a predetermined shape to the measured values ​​of parameters related to the performance change factors. By doing so, a second correlation based on the measured values ​​can be obtained. Accordingly, the performance change factors of the device under evaluation can be estimated with high precision.

[0087] In some embodiments, in step S8 described above, the model creation unit (26) may determine each coefficient (n1 to n4) of a plurality of individual performance functions (A(t), B(t), C(t), etc(t)) using a Bayes estimation method.

[0088] Here, the procedure for determining coefficients n1 to n4 in the above-described performance estimation model y=F'(t)=n1×A(t)+n2×B(t)+n3×C(t)+n4×etc(t) using the Bayesian estimation method is briefly explained.

[0089] (a) First, the prior distribution (range, shape, etc.) of the linear parameters (coefficients) N=(n1, n2, n3, n4) of the performance estimation model y=F'(t) is set based on empirical rules, etc. Here, if there is little prior information, the conjugate prior distribution or the uniform distribution may be used as the prior distribution.

[0090] (b) The linear parameter N=(n1, n2, n3, n4) of the performance estimation model y=F'(t) is varied to calculate the likelihood function representing the degree of agreement with the data Y=F(t). The likelihood function is the product of probabilities P(F(t)|F'(t), N) which are determined by the deviation of the sample's observed data Y=F(t) from the y-coordinate corresponding to the t-coordinate of each sample on the curve (y=F'(t)) for the linear parameter N at the current calculation step (i.e., P(F(t1)|F'(t1), N) × P(F(t2)|F'(t2), N) × … × P(F(t1) k )|F'(t k It is calculated as ), N). By organizing these calculated values ​​on the horizontal axis n1 to n4 respectively, the likelihood P(Y|N) is obtained.

[0091] (c) For each of the linear parameters n1 to n4, random sampling is performed using a random number generation algorithm based on the prior distribution. As a random number generation algorithm, for example, Markov chain Monte Carlo methods (MCMC methods) can be used. Additionally, as an algorithm for generating MCMC samples in MCMC methods, Hamilton Monte Carlo, Gibbs sampler, or Metropolis methods can be used.

[0092] (d) For each sampling point of linear parameters n1 to n4 sampled in (c) above, the likelihood is calculated from (b).

[0093] (e) By multiplying the prior distribution and the likelihood, the probability density distribution P(Y|N)P(N) is obtained for each of the linear parameters n1 to n4.

[0094] (f) For each of the linear parameters n1 to n4, the probability density distribution P(Y|N)P(N) obtained in (e) is divided by its area (normalized) to obtain the probability density distribution P(Y|N) with an area of ​​1 as the posterior distribution.

[0095] (g) For each of the linear parameters n1 to n4, the expected value of the posterior distribution P(Y|N) is taken as the estimate of the corresponding linear parameter.

[0096] In the above-described embodiment, when determining the coefficients (n1 to n4) in the superposition of individual performance functions (A(t), B(t), C(t), etc(t)) representing performance changes attributable to each of multiple performance change factors, a probabilistic model based on the Bayesian estimation method is used, thereby allowing the subjective information of the model creator to be reflected. Consequently, the accuracy of the performance change estimation of the device under evaluation is improved. Furthermore, by using the Bayesian estimation method, there is an effect of being robust against outliers.

[0097] In addition, when calculating the posterior distribution of each coefficient in Bayesian estimation, using a random number generation algorithm makes it possible to calculate the posterior distribution of the Bayesian estimation using a calculator.

[0098] Furthermore, by determining the range or shape of the prior distribution of coefficients n1 to n4 in Bayesian estimation based on previously acquired data or physical assumptions, physical assumptions or empirical rules can be reflected in the prior distribution in Bayesian estimation. For this reason, even when there is insufficient data to determine individual performance functions for each performance change factor, the accuracy of the performance change estimation of the device under evaluation is improved.

[0099] The contents described in each of the above embodiments are understood as, for example, as follows.

[0100] (1) A performance evaluation device (20) for a device according to at least one embodiment of the present invention,

[0101] A total performance function acquisition unit (22) configured to obtain a total performance function (e.g., F(t) described above) representing a change in performance indicators over time of the device being evaluated (e.g., the turbine (2) described above) based on data acquired during the operation of the device being evaluated, and

[0102] An individual performance function definition unit (24) configured to define a plurality of individual performance functions (e.g., A(t), B(t), C(t), etc(t) described above) each representing a temporal change of the performance indicator caused by a plurality of change factors of the performance indicator, and

[0103] The present invention provides a model creation unit (26) configured to create a performance estimation model of the device to be evaluated (e.g., the above-described F'(t)) by performing the superposition of the above-described plurality of individual performance functions, and

[0104] The above model creation unit is configured to determine each coefficient (e.g., n1 to n4 described above) of the plurality of individual performance functions in the superposition so that the superposition of the plurality of individual performance functions approximates the overall performance function.

[0105] According to the configuration of (1) above, coefficients are determined (i.e., weights are assigned) in the superposition of individual performance functions representing performance changes attributable to each of the multiple change factors (performance change factors) of the performance indicators so as to approximate the overall performance function representing the temporal change of the performance indicators of the device under evaluation. By doing so, a performance estimation model including the respective weights of the multiple performance change factors can be created. Therefore, for the device under evaluation, it is possible to quantitatively estimate the temporal performance change and also estimate the factors of the performance change.

[0106] (2) In some embodiments, in the configuration of (1) above,

[0107] The above model creation unit is configured to determine each coefficient of the plurality of individual performance functions using a Bayes estimation method.

[0108] According to the configuration of (2) above, in determining the coefficients in the superposition of individual performance functions representing performance changes caused by each of the multiple performance change factors, a probabilistic model based on the Bayesian estimation method is used, so subjective information of the model creator can be reflected, and the accuracy of the performance change estimation of the device under evaluation is improved.

[0109] (3) In some embodiments, in the configuration of (2) above,

[0110] The above model creation unit is configured to obtain the posterior distribution of the coefficients in the Bayes estimation using a random number generation algorithm.

[0111] According to the configuration of (3) above, when calculating the posterior distribution in Bayesian estimation, a random number generation algorithm is used, so it is possible to calculate the posterior distribution of Bayesian estimation using a calculator.

[0112] (4) In some embodiments, in the configuration of (2) or (3) above,

[0113] The above model creation unit is configured to determine the range or shape of the prior distribution of the coefficients in the Bayes estimation based on previously acquired data or physical assumptions.

[0114] According to the configuration of (4) above, physical assumptions or empirical rules can be reflected in the prior distribution in Bayesian estimation. For this reason, even when there is little data to determine individual performance functions for each performance change factor, the accuracy of the performance change estimation of the device under evaluation is good.

[0115] (5) In some embodiments, in the configuration of any of (1) to (4) above,

[0116] The individual performance function definition unit is configured to define the individual performance function for each of the plurality of change factors based on the theoretical or measured value of a parameter related to the change factor.

[0117] According to the configuration of (5) above, for each of the multiple performance change factors, an individual performance function can be determined based on the theoretical or measured value of the parameter related to the corresponding change factor. Because of this, the accuracy of the performance change estimation of the device under evaluation is improved.

[0118] (6) In some embodiments, in the configuration of (5) above,

[0119] The individual performance function definition unit is configured to define the individual performance function based on a first correlation between the value of the parameter and the magnitude of the performance change of the device under evaluation, and a second correlation between the value of the parameter and time.

[0120] According to the configuration of (6) above, individual performance functions are determined based on a first correlation between the value of a parameter related to a performance change factor and the magnitude of the performance change, and a second correlation between the value of the parameter and time. Therefore, individual performance functions for each performance change factor can be appropriately defined, and thus the accuracy of the performance change estimation of the device under evaluation is improved.

[0121] (7) In some embodiments, in the configuration of (6) above,

[0122] The individual performance function definition unit is configured to obtain the second correlation by fitting a base curve having the shape of the correlation between the value of the parameter and time to the actual value of the parameter.

[0123] According to the configuration of (7) above, a base curve having a predetermined shape is fitted to the measured value of a parameter related to the performance change factor, so a second correlation based on the measured value can be obtained. Therefore, the performance change factor of the device to be evaluated can be estimated with high precision.

[0124] (8) A method for evaluating the performance of a device according to at least one embodiment of the present invention,

[0125] A step (S4) of obtaining an overall performance function representing the temporal change of the performance indicator of the device to be evaluated based on data acquired during the operation of the device to be evaluated, and

[0126] A step (S6) for defining a plurality of individual performance functions each representing a temporal change of the performance indicator caused by a plurality of change factors of the performance indicator, and

[0127] The method comprises a step (S8) of creating a performance estimation model of the device to be evaluated by performing the superposition of the plurality of individual performance functions.

[0128] In the step of creating the above performance estimation model, the coefficients of each of the plurality of individual performance functions in the overlap are determined such that the overlap of the plurality of individual performance functions approximates the overall performance function.

[0129] According to the method of (8) above, coefficients are determined (i.e., weights are assigned) in the superposition of individual performance functions representing performance changes attributable to each of the multiple change factors (performance change factors) of the performance indicator so as to approximate the overall performance function representing the temporal change of the performance indicator of the device being evaluated. By doing so, a performance estimation model can be created that includes the respective weights of the multiple performance change factors. Therefore, for the device being evaluated, it is possible to quantitatively estimate the temporal performance change and to estimate the factors of the performance change.

[0130] (9) A performance evaluation program for a device according to at least one embodiment of the present invention,

[0131] on the computer,

[0132] A procedure for obtaining an overall performance function representing the temporal change of a performance indicator of a device under evaluation based on data acquired during the operation of the device under evaluation, and

[0133] A procedure for defining a plurality of individual performance functions, each representing a temporal change of the performance indicator attributable to a plurality of change factors of the performance indicator, and

[0134] A procedure to create a performance estimation model of the device to be evaluated by performing the superposition of the above plurality of individual performance functions is executed, and

[0135] In the procedure for constructing the above performance estimation model, the coefficients of each of the plurality of individual performance functions in the overlap are determined such that the overlap of the plurality of individual performance functions approximates the overall performance function.

[0136] According to the program of (9) above, coefficients are determined (i.e., weights are assigned) in the superposition of individual performance functions representing performance changes caused by each of the multiple change factors (performance change factors) of the performance indicator so as to approximate the overall performance function representing the temporal change of the performance indicator of the device being evaluated. By doing so, a performance estimation model can be created that includes the respective weights of the multiple performance change factors. Therefore, for the device being evaluated, it is possible to quantitatively estimate the temporal performance change and to estimate the factors of the performance change.

[0137] Although embodiments of the present invention have been described above, the present invention is not limited to the embodiments described above and includes forms in which modifications are made to the embodiments described above or forms in which the forms are appropriately combined.

[0138] In this specification, expressions indicating relative or absolute arrangements such as “in a certain direction,” “along a certain direction,” “parallel,” “orthogonal,” “center,” “concentric,” or “coaxial” are used not only to strictly indicate such arrangements, but also to indicate a state of relative displacement with a tolerance or an angle or distance such that the same function is obtained.

[0139] For example, expressions indicating that things are in an equivalent state, such as "identical," "equivalent," and "homogeneous," are meant to indicate not only a strictly equivalent state, but also a state in which tolerances or differences in the degree to which the same function is obtained exist.

[0140] In addition, in this specification, expressions indicating shapes such as square shapes or cylindrical shapes represent not only shapes such as square shapes or cylindrical shapes in a geometrically strict sense, but also shapes including uneven parts or chamfered parts within the scope where the same effect is obtained.

[0141] Furthermore, in this specification, expressions such as “comprising,” “including,” or “having” one component are not exclusive expressions excluding the existence of other components. Explanation of the symbols

[0142] 2: Turbine 2a: Upstream step 2b: Downstream step 12: Measurement section 14: Memory Department 16: Display section 20: Performance evaluation device 22: Overall Performance Function Acquisition Unit 24: Individual Performance Function Definition Section 26: Model Creation Section 28: Evaluation Department 100: Base curve 102: Secondary Correlation 104: Secondary Correlation

Claims

Claim 1 A performance evaluation device for a device comprising: a total performance function acquisition unit configured to obtain a total performance function representing a temporal change in a performance indicator of the device to be evaluated based on data acquired during the operation of the device to be evaluated; an individual performance function definition unit configured to define a plurality of individual performance functions representing a temporal change in the performance indicator, each caused by a plurality of change factors of the performance indicator; and a model creation unit configured to create a performance estimation model of the device to be evaluated by performing a superposition of the plurality of individual performance functions, wherein the model creation unit is configured to determine each coefficient of the plurality of individual performance functions in the superposition so that the approximation of the superposition of the plurality of individual performance functions to the total performance function increases. Claim 2 In claim 1, the model creation unit is configured to determine each coefficient of the plurality of individual performance functions using a Bayes estimation method, a performance evaluation device of a device. Claim 3 In paragraph 2, the above model creation unit is configured to obtain the posterior distribution of the coefficients in the Bayes estimation using a random number generation algorithm, the performance evaluation device of the device. Claim 4 A performance evaluation device of a device, wherein, in paragraph 2 or 3, the model creation unit is configured to determine the range or shape of the prior distribution of the coefficients in the Bayes estimation based on previously acquired data or physical assumptions. Claim 5 A performance evaluation device of a device, wherein, in any one of claims 1 to 3, the individual performance function definition unit is configured to define the individual performance function for each of the plurality of change factors based on the theoretical value or measured value of a parameter related to the change factor. Claim 6 In claim 5, the individual performance function definition unit is configured to define the individual performance function based on a first correlation between the value of the parameter and the magnitude of the performance change of the device to be evaluated, and a second correlation between the value of the parameter and time, in a performance evaluation device. Claim 7 In claim 6, the individual performance function definition unit is configured to obtain the second correlation by fitting a base curve having the shape of the correlation between the value of the parameter and time to the actual value of the parameter, thereby forming a performance evaluation device of the device. Claim 8 A method for evaluating the performance of a device, comprising: a step of obtaining an overall performance function representing a temporal change in a performance indicator of the device to be evaluated based on data acquired during the operation of the device to be evaluated; a step of defining a plurality of individual performance functions each representing a temporal change in the performance indicator caused by a plurality of change factors of the performance indicator; and a step of creating a performance estimation model of the device to be evaluated by performing a superposition of the plurality of individual performance functions, wherein in the step of creating the performance estimation model, the coefficients of each of the plurality of individual performance functions in the superposition are determined such that the approximation of the superposition of the plurality of individual performance functions to the overall performance function increases. Claim 9 A performance evaluation program for a device stored in a memory device, wherein the computer executes a step of obtaining an overall performance function representing a temporal change in a performance indicator of the device to be evaluated based on data acquired during the operation of the device to be evaluated, a step of defining a plurality of individual performance functions each representing a temporal change in the performance indicator caused by a plurality of change factors of the performance indicator, and a step of creating a performance estimation model of the device to be evaluated by performing a superposition of the plurality of individual performance functions, wherein in the step of creating the performance estimation model, the coefficients of each of the plurality of individual performance functions in the superposition are determined such that the approximation of the superposition of the plurality of individual performance functions to the overall performance function increases.

Images