Methods for analyzing the condition of mechanical elements

Abstract

Analyze the state of mechanical elements in a short processing time. [Solution] The method for analyzing the state of a mechanical element involves preparing an analysis model that models the mechanical element, has internal parameters, and outputs output values ​​in response to input values. Using the analysis model, a sensitivity function is prepared for both the low-frequency and high-frequency regions to define the relationship between the fluctuations of the internal parameters and input values ​​and the fluctuations of the output values ​​through simulation. Operating data of the mechanical element corresponding to the output values ​​is acquired. Using the output values ​​corresponding to the normal operation of the mechanical element as normal data, the fluctuations of the output values ​​are acquired for both the low-frequency and high-frequency regions based on the difference between the operating data and the normal data. For both the low-frequency and high-frequency regions, the fluctuations of the internal parameters corresponding to the fluctuations of the output values ​​are acquired using the sensitivity function. The portion of the mechanical element corresponding to the internal parameters with large fluctuations is estimated as the fluctuation-responding portion.

Description

Technical Field

[0001] The present disclosure relates to a state analysis method for mechanical elements.

Background Art

[0002] Conventionally, systems for estimating quality parameters for constituent elements of equipment have been known for purposes such as predicting the remaining service life of the equipment. Patent Document 1 discloses an estimation system of this type.

[0003] The estimation system of Patent Document 1 is configured to estimate the quality of a plurality of engine constituent elements by, for example, comparing actual sensor values from a plurality of sensors of a plurality of engine constituent elements of equipment (specifically, a gas turbine engine) with virtual sensor values calculated by simulation using an engine model, thereby generating quality parameters for the engine constituent elements.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of Invention

Problems to be Solved by the Invention

[0005] The configuration of the above Patent Document 1 cannot perform quality estimation at locations where no sensor is installed, and also requires a long processing time due to the large scale of the simulation. It is also conceivable to analyze the state of locations without sensors by improving the simulation method, but in that case, the scale of the simulation becomes even larger, and the time required for processing becomes even longer.

[0006] The present disclosure has been made in view of the above circumstances, and an object thereof is to analyze the state of mechanical elements with short-time processing.

Means for Solving the Problems

[0007] The problems that this disclosure aims to solve are as described above, and next we will explain the means and effects of solving these problems.

[0008] From the perspective of this disclosure, the following method for analyzing the state of a mechanical element is provided. Specifically, this method for analyzing the state of a mechanical element involves preparing an analysis model that models the mechanical element and has internal parameters, and outputs an output value in response to an input value. By simulation using the analysis model, a sensitivity function that defines the fluctuation of the internal parameters and the relationship between the fluctuation of the input value and the fluctuation of the output value is prepared for each of the low-frequency region including 0 Hz and the high-frequency region higher than the low-frequency region. Operating data of the mechanical element corresponding to the output value is acquired. The output value corresponding to the normal operation of the mechanical element is used as normal data, and the fluctuation of the output value is acquired for each of the low-frequency region and the high-frequency region based on the difference between the operating data and the normal data. For each of the low-frequency region and the high-frequency region, the fluctuation of the internal parameter corresponding to the fluctuation of the output value is acquired using the sensitivity function. The portion of the mechanical element corresponding to the internal parameter with large fluctuations is estimated as the fluctuation-responding portion.

[0009] This allows the fluctuation in output values ​​to be obtained for both the low-frequency and high-frequency regions based on the difference between operating data and normal data for the mechanical element. Then, using the sensitivity function obtained from simulations using the analysis model, the fluctuation in internal parameters in response to the fluctuation in output values ​​is obtained, and the part of the mechanical element corresponding to the internal parameter with the largest fluctuation is estimated as the fluctuation-responding part. Since this series of estimations is performed separately for the low-frequency and high-frequency regions, the computational load is reduced compared to when the two regions are not separated, and thus the state of the mechanical element can be estimated or analyzed in a short amount of time. [Effects of the Invention]

[0010] According to this disclosure, the state of mechanical elements can be analyzed in a short amount of time. [Brief explanation of the drawing]

[0011] [Figure 1] A flowchart illustrating the procedure for analyzing the state of a gas turbine engine using a mechanical element state analysis method according to one embodiment of the present disclosure. [Figure 2] A block diagram showing an example of a gas turbine engine state analysis model. [Figure 3] A diagram conceptually illustrating the sensitivity function of the submodel. [Figure 4] Flowchart of a method for analyzing the state of mechanical elements according to the first modified example. [Figure 5] Flowchart of a method for analyzing the state of mechanical elements related to the second modified example. [Modes for carrying out the invention]

[0012] Next, embodiments of the present disclosure will be described with reference to the drawings. Figure 1 is a flowchart showing the procedure for analyzing the state of a gas turbine engine 10 using the mechanical element state analysis method according to the present embodiment. Figure 2 is a block diagram showing an example of a state analysis model of the gas turbine engine 10. Figure 3 is a diagram conceptually showing the sensitivity function of a submodel.

[0013] As shown in Figure 1, the method for analyzing the state of mechanical elements according to this embodiment uses a gas turbine engine 10, which is an example of equipment, as the object of analysis. The gas turbine engine 10 comprises a compressor 11, a fuel control valve 12a, a combustor 13, a turbine 14, and a rotating shaft 15. Each of these components is an example of a mechanical element. Thus, in this embodiment, each mechanical element constitutes a part of the equipment.

[0014] The compressor 11 compresses and discharges air. The compressor 11 has a fixed blade and a rotating blade that is rotatable relative to the fixed blade. The rotating blade is connected to a rotating shaft 15 and rotates relative to the fixed blade as the rotating shaft 15 rotates. As a result, the compressor 11 compresses and discharges air drawn in from the atmosphere. The compressed air discharged by the compressor 11 is supplied to the combustor 13.

[0015] The fuel control valve 12a adjusts the flow rate of fuel supplied to the combustor 13. The fuel control valve 12a is installed in the fuel supply pipe 12 connected to the combustor 13, and its opening degree is adjusted by a control device (not shown) to adjust the flow rate of fuel flowing through the fuel supply pipe 12. The opening degree of the fuel control valve 12a is adjusted based on the rotational speed of the rotating shaft 15, etc.

[0016] The combustor 13 receives compressed air discharged by the compressor 11 and also receives fuel from the fuel supply pipe 12. The combustor 13 then mixes the compressed air and fuel and burns them. The high-temperature, high-pressure combustion gas produced by this combustion is supplied to the turbine 14.

[0017] The turbine 14 converts the thermal energy of the combustion gas into rotational kinetic energy of the rotating shaft 15. The turbine 14 has a fixed blade and a rotating blade that is rotatable relative to the fixed blade. The turbine 14 rotates the rotating blade using the combustion gas supplied from the combustor 13. The rotating blade is connected to the rotating shaft 15, and its rotation causes the rotating shaft 15 to rotate as well.

[0018] The rotating shaft 15 is connected to the rotating blades of the compressor 11 and the turbine 14, respectively. As described above, the rotating shaft 15 is rotated by the turbine 14, and this rotation is transmitted to the compressor 11 and a load not shown (e.g., a generator). In this way, the thermal energy of the combustion gas produced in the combustor 13 is used for the compression operation of the compressor 11 and for the predetermined operation of the load.

[0019] Further, a large number of sensors 21 are attached to the gas turbine engine 10. The sensors 21 detect the state of each part of the gas turbine engine 10. The number of the sensors 21 is arbitrary, and may be, for example, several tens. The state quantities detected by each of the sensors 21 are arbitrary, and include, for example, temperature, pressure, flow rate, vibration, electric current, electric energy, rotational speed, and the like.

[0020] Each of the sensors 21 is connected to the above-mentioned control device by wire or wirelessly. The control device is a known computer including a CPU, a ROM, a RAM, and the like. The data values acquired by the sensors 21 are stored in the control device together with the time of acquisition. Note that the machine element state analysis method described below may be executed using this control device, or may be executed using a separate computer.

[0021] As shown in Figure 1, in the machine element state analysis method according to the present embodiment, as preliminary preparation in the previous step of analyzing the state of the gas turbine engine 10, preparation of a model that mathematically represents the above-described gas turbine engine 10 and preparation of sensitivity functions for a plurality of sub-models included in the model are performed (S1). Hereinafter, the model corresponding to the gas turbine engine 10 may be referred to as a state analysis model. Here, "preparing X" may refer to either newly creating X, or making an existing X available for use. In the following, a case where the state analysis model and the sensitivity functions are newly created will be described as an example.

[0022] First, regarding the creation of the state analysis model (S1A), the gas turbine engine 10 described above can be represented by the state analysis model shown in the block diagram in Figure 2. In the example in the same figure, the compressor 11, fuel control valve 12a, combustor 13, turbine 14, and rotating shaft 15 are each modeled as submodels (analysis models of mechanical elements) that constitute a part of the state analysis model of the gas turbine engine 10. Each submodel of a component has internal parameters and outputs output values ​​in response to input values. Each submodel can be constructed, for example, as a mathematical model or a finite element model, but is not limited to these. Internal parameters are parameters that represent the physical characteristics of the mechanical element (or component). For example, in the case of the gas turbine engine 10, adiabatic efficiency, mechanical efficiency, or cross-sectional area can be used as internal parameters.

[0023] The submodel of compressor 11 has an adiabatic efficiency η and a mechanical efficiency ηm as internal parameters. The adiabatic efficiency η is highly correlated with the degree of deterioration of each blade and seal of compressor 11. The mechanical efficiency ηm is highly correlated with the degree of deterioration of each blade and bearing of compressor 11. The submodel of compressor 11 receives the atmospheric pressure Po and temperature To, the pressure P1 on the inlet side of combustor 13, and the rotational speed Nm of the rotating shaft 15 as input values. The submodel of compressor 11 outputs the enthalpy H1 and flow rate G1 of the discharge fluid (compressed air), and the compressor power Wc as output values.

[0024] In the submodel of compressor 11, atmospheric pressure Po and temperature To are given as analysis conditions. The inlet pressure P1 of combustor 13 is input from the submodel of combustor 13. The rotational speed Nm of the rotating shaft 15 is input from the submodel of rotating shaft 15. The enthalpy H1 and flow rate G1 of the discharge fluid (compressed air) are input to the submodel of combustor 13. The compressor power Wc is input to the submodel of rotating shaft 15.

[0025] The submodel of the fuel control valve 12a has a friction coefficient μ as an internal parameter. The friction coefficient μ is highly correlated with the degree of deterioration of the fuel control valve 12a. The submodel of the fuel control valve 12a receives the rotational speed Nm of the rotating shaft 15 as an input value. The submodel of the fuel control valve 12a outputs the flow rate of fuel passing through the fuel control valve 12a, i.e., the fuel flow rate Gf, as an output value.

[0026] In the submodel of the fuel control valve 12a, the rotational speed Nm of the rotating shaft 15 is input from the submodel of the rotating shaft 15. The fuel flow rate Gf is input to the submodel of the combustor 13.

[0027] The submodel of the combustor 13 has a cross-sectional area S and a pressure drop coefficient Ag as internal parameters. The cross-sectional area S is highly correlated with the degree of fouling of the combustor. The pressure drop coefficient Ag is highly correlated with the degree of clogging and fouling of the nozzles of the combustor 13. The submodel of the combustor 13 receives the enthalpy H1 and flow rate G1 of the discharge fluid from the compressor 11, the fuel flow rate Gf of the fuel control valve 12a, and the flow rate G2 of the combustion gas in the turbine 14 as input values. The submodel of the combustor 13 outputs the pressure P1 on the inlet side of the combustor 13, and the enthalpy H2 and pressure P2 on the outlet side of the combustor 13 as output values.

[0028] In the submodel of the combustor 13, the enthalpy H1 and flow rate G1 of the discharge fluid from the compressor 11 are input from the submodel of the compressor 11. The fuel flow rate Gf of the fuel control valve 12a is input from the submodel of the fuel control valve 12a. The flow rate G2 of the combustion gas in the turbine 14 is input from the submodel of the turbine 14. The pressure P1 on the inlet side of the combustor 13 is input to the submodel of the compressor 11. The enthalpy H2 and pressure P2 on the outlet side of the combustor 13 are input to the submodel of the turbine 14.

[0029] The submodel of turbine 14 has adiabatic efficiency η and mechanical efficiency ηm as internal parameters. The adiabatic efficiency η is highly correlated with the degree of deterioration of each blade and seal of turbine 14. The mechanical efficiency ηm is highly correlated with the degree of deterioration of each blade and bearing of turbine 14. The submodel of turbine 14 receives the enthalpy H2 and pressure P2 on the outlet side of combustor 13, the pressure P3 on the outlet side of turbine 14, and the rotational speed Nm of the rotating shaft 15 as input values. The submodel of turbine 14 outputs the enthalpy H3 of the discharged fluid and the turbine power Wt as output values.

[0030] In the submodel of turbine 14, the enthalpy H2 and pressure P2 on the outlet side of combustor 13 are input from the submodel of combustor 13. The pressure P3 on the outlet side of turbine 14 is given as an analysis condition. The rotational speed Nm of the rotating shaft 15 is input from the submodel of rotating shaft 15. The enthalpy H3 of the discharged fluid is output as part of the analysis results. The turbine power Wt is input into the submodel of rotating shaft 15.

[0031] The submodel of the rotating shaft 15 has a rotational resistance R as an internal parameter. The rotational resistance R is highly correlated with the degree of deterioration of the bearing that rotatably supports the rotating shaft 15. The submodel of the rotating shaft 15 receives compressor power Wc and turbine power Wt as input values. The submodel of the rotating shaft 15 outputs rotational speed Nm as an output value.

[0032] In the submodel of the rotating shaft 15, the compressor power Wc is input from the submodel of the compressor 11. The turbine power Wt is input from the submodel of the turbine 14. The rotational speed Nm is input to the submodel of the compressor 11, the submodel of the fuel control valve 12a, and the submodel of the turbine 14, respectively.

[0033] Thus, some of the submodels are related such that the input value of one becomes the output value of the other. For example, the submodel of the compressor 11 and the submodel of the combustor 13 are related such that the input value of the former (i.e., pressure P1) becomes the output value of the latter, and the input value of the latter (i.e., enthalpy H1 and flow rate G1) becomes the output value of the former. As can be seen from Figure 2, such interrelationships also exist between the submodel of the compressor 11 and the submodel of the rotating shaft 15, between the submodel of the combustor 13 and the submodel of the turbine 14, and between the submodel of the turbine 14 and the submodel of the rotating shaft 15.

[0034] After creating submodels for each component, the next step in the preparation process is to create a sensitivity function for each submodel (S1B). Here, the sensitivity function is a function that defines the relationship between the fluctuations of internal parameters and input values ​​and the fluctuations of output values ​​in each submodel. For example, the sensitivity function of the submodel of compressor 11 is a function that defines the relationship between the fluctuations of the adiabatic efficiency η and mechanical efficiency ηm of compressor 11, the fluctuations of atmospheric pressure Po and temperature To, the fluctuations of the inlet pressure P1 of combustor 13, and the fluctuations of the rotational speed Nm of the rotating shaft 15, the fluctuations of the enthalpy H1 and flow rate G1 of the discharge fluid, and the fluctuations of the compressor power Wc, and can be expressed as a matrix. The sensitivity function of each submodel can be created by performing simulations in which the internal parameters and input values ​​are varied in the submodel, determining the fluctuations of output values ​​in response to these fluctuations, and then creating the function based on the relationship between the two fluctuations.

[0035] Figure 3 conceptually illustrates that by inputting the variation dP of the internal parameter P and / or the variation dX of the input value X for any submodel, the variation dY of the output value Y is obtained, and a sensitivity function S is created based on the relationship between the two variations. Here, equation (1), with the subscript o in each term, represents the creation of the sensitivity function So of the submodel in the low-frequency region, and equation (2), with the subscript v in each term, represents the creation of the sensitivity function Sv of the submodel in the high-frequency region.

[0036] A distinctive feature of this disclosure is that the sensitivity function of each submodel is created for both the low-frequency region, which includes 0 Hz, and the high-frequency region, which is higher than the low-frequency region. In this disclosure, 0 Hz is included in the low-frequency region. The frequency domains of the low-frequency region and the high-frequency region can be appropriately selected according to the operating characteristics of the instrument (or mechanical element) being analyzed. In this embodiment, the low-frequency region is only 0 Hz, and the high-frequency region includes multiple frequency domains (i.e., multiple non-contiguous frequency domains). In this example, the number of frequency domains included in the high-frequency region is two, but it is not limited to this.

[0037] In this embodiment, the low-frequency region is limited to 0 Hz. Accordingly, the sensitivity function for the low-frequency region can be created by varying the internal parameters of each submodel so as to offset them from their design values, and / or by varying the input values ​​of each submodel so as to offset them from their normal values, determining the corresponding changes in output values, and then creating the function based on the relationship between the two changes. On the other hand, the sensitivity function for the high-frequency region can be created by varying the internal parameters and / or input values ​​of each submodel so as to oscillate at the frequency corresponding to that high-frequency region, determining the corresponding changes in output values, and then creating the function based on the relationship between the two changes.

[0038] In the mechanical element state analysis method according to this embodiment, fault detection is performed when any fault occurs in the gas turbine engine 10 after the preliminary preparations have been completed. Here, fault detection in this embodiment includes both outputting the location where the fault is estimated and acquiring information about the fault, but one of them may be omitted. Outputting the location where the fault is estimated can be achieved, for example, by displaying it on the display of the computer on which the fault detection is performed, but is not limited to this.

[0039] As shown in Figure 1, in fault detection, first, operating data for each component corresponding to the input value and output value (i.e., the compressor 11, fuel control valve 12a, combustor 13, turbine 14, and rotating shaft 15) is acquired (S2 and S3). The operating data can be acquired using the sensor 21 described above, but is not limited to this. For example, the operating data could be time-series data going back for an arbitrary period from the time the fault occurred.

[0040] Next, the operating data of each component corresponding to the input value is input to the submodel corresponding to that component, and the corresponding output value is obtained (S4 and S5). The output value obtained at this time is obtained from the submodel of each component, that is, from a model that simulates each component in a normal state without failure, and is therefore called normal data. Note that the method of obtaining normal data is not limited to this, for example, the operating data of the component corresponding to the output value obtained by the sensor 21 when each component is operating normally may be used as normal data, or normal data may be obtained by utilizing the interrelationships between submodels. To explain an example of the acquisition method using the interrelationships between submodels, normal data corresponding to one submodel can be input to another submodel that can take this as an input value, and normal data corresponding to that other submodel can be obtained using the sensitivity function of this other submodel. Normal data obtained by the three methods described in this paragraph are all obtained as output values ​​corresponding to the normal operation of each component of the gas turbine engine 10. Note that the method of obtaining normal data is not limited to these three methods.

[0041] Next, for both the low-frequency and high-frequency regions, the fluctuation in the output value is obtained based on the difference between the operating data corresponding to the output value and the normal data described above (S6). This acquisition of the output value fluctuation is performed for each component of the gas turbine engine 10. The output value fluctuation in the low-frequency region can be obtained as the average value of the difference between the operating data and the normal data. The output value fluctuation in the high-frequency region can be obtained as the RMS value of the remaining component after subtracting the output value fluctuation in the low-frequency region from the difference between the operating data and the normal data. In addition, the output value fluctuation in the high-frequency region can also be obtained as the power spectrum of a specific frequency region when the difference between the operating data and the normal data is Fourier transformed.

[0042] Next, a sensitivity function is used to determine the internal parameters that have a high impact on failure (S7). In this embodiment, first, for both the low-frequency and high-frequency regions, the sensitivity function is used to obtain the internal parameters corresponding to the fluctuations in the output value and the fluctuations in the input value. Then, based on the obtained results, the internal parameters that have a high impact on failure are determined, and the locations related to those internal parameters are obtained as locations where failures are estimated (S8). The acquisition of fluctuations in internal parameters and input values ​​is performed for each component of the gas turbine engine 10 (or for each submodel). Two specific methods can be exemplified below.

[0043] The first method involves obtaining variations in internal parameters and input values ​​by utilizing the interrelationships between submodels. Specifically, in this method, for a submodel of a given component, a sensitivity function is used to obtain the variation in input values ​​corresponding to the variation in output values ​​in both the low-frequency and high-frequency regions. This is equivalent to solving the inverse function in that submodel. Known methods such as the least squares method or the Moore-Penrose pseudoinverse can be used to solve the inverse function. Next, the magnitude of the input value variation is determined. A threshold value set for each input value can be used for this determination. If the input value variation is large, this large variation is used as the variation in the output value of another related submodel, where the input value becomes the output value, and the variation in internal parameters and input values ​​is obtained in that other submodel using the sensitivity function. This process is repeated until a submodel with small input value variation is reached. On the other hand, if the aforementioned input value variation is small (including when a submodel with small input value variation is reached), the magnitude of the variation in the internal parameters of the submodel of that component is determined. Here, if the output value variation of the submodel is large and the input value variation is small, the variation in internal parameters is usually large. If there is a submodel with large fluctuations in its internal parameters, it suggests that the component corresponding to this submodel is the faulty part (the part that handles the fluctuations). Note that there are no particular limitations on the method for determining whether internal parameter fluctuations are large; depending on the method used, it may be possible to omit the determination of the magnitude of the internal parameter fluctuations. Similarly, the determination of the magnitude of input value fluctuations may also be omitted in some cases.

[0044] The second method involves individually obtaining the fluctuations in internal parameters and input values ​​for each submodel using a sensitivity function. Specifically, for each submodel of a component, the fluctuations in internal parameters and input values ​​corresponding to the fluctuations in output values ​​are obtained using a sensitivity function for both the low-frequency and high-frequency regions. This is equivalent to solving the inverse function for that submodel, as in the first method. Next, the magnitude of the fluctuations in the internal parameters is determined for each submodel. A threshold value set for each internal parameter can be used for this determination. If a submodel exhibits large fluctuations in its internal parameters, it suggests that the component corresponding to this submodel is the faulty part (the part corresponding to the fluctuation). Note that the method for determining whether the fluctuations in internal parameters are large is not particularly limited, and depending on the method used, it may be possible to omit the determination of the magnitude of the fluctuations in internal parameters.

[0045] In this embodiment, the first method described above is adopted, but of course, the second method can also be used. Furthermore, it is possible to combine the first and second methods as needed. For example, the first method could be used for some submodels, while the second method could be used for the remaining submodels.

[0046] Next, a simulation is performed on a submodel that has internal parameters that have a high impact on failure, by changing those internal parameters (S9). In other words, a simulation is performed on only the components that include the failure location (variable part) that the submodel suggests failure to, by changing the internal parameters of the submodel, thereby obtaining information about the failure. The information about the failure is obtained by reproducing the behavior of the component including the variable part when it fails, and can be used as a clue to investigate the failure location and the physical phenomena leading to the failure (S10). If the behavior can be reproduced in the simulation, it is thought that the difference between the operating data corresponding to the input and / or output values ​​and the input and / or output values ​​calculated by the simulation will be sufficiently small, and this can be used to confirm the reproduction of the behavior.

[0047] As described above, in this embodiment, the state analysis of the mechanical elements is performed as follows. Specifically, an analysis model (submodel) is prepared that models the components of the gas turbine engine 10 and has internal parameters, and outputs an output value in response to an input value. Through simulation using the analysis model, a sensitivity function that defines the fluctuation of internal parameters and the relationship between the fluctuation of input values ​​and the fluctuation of output values ​​is prepared for both the low-frequency region including 0 Hz and the high-frequency region higher than the low-frequency region. Operating data of the components corresponding to the output values ​​is acquired. Using the output value corresponding to the normal operation of the components as normal data, the fluctuation of the output value is acquired for both the low-frequency region and the high-frequency region based on the difference between the operating data and the normal data. For both the low-frequency region and the high-frequency region, the fluctuation of internal parameters corresponding to the fluctuation of the output value is acquired using the sensitivity function. The portion of the components corresponding to internal parameters with large fluctuations is estimated as the fluctuation-responding portion.

[0048] As a result, for both the low-frequency and high-frequency regions, the fluctuation in output value is obtained based on the difference between the operating data and the normal data of the components of the gas turbine engine 10. Then, using the sensitivity function obtained from the simulation using the analysis model, the fluctuation in the internal parameters in response to the fluctuation in output value is obtained, and the part of the components corresponding to the internal parameters with large fluctuations is estimated as the fluctuation-responding part. Since this series of estimations is performed separately for the low-frequency and high-frequency regions, the computational load is reduced compared to when the two regions are not separated, and therefore the state of the components can be estimated or analyzed in a short amount of time.

[0049] Furthermore, in the mechanical element state analysis method of this embodiment, the low-frequency region is limited to 0 Hz only.

[0050] This allows for state analysis to be performed in a frequency range suitable for the components of the gas turbine engine 10.

[0051] Furthermore, in the mechanical element state analysis method of this embodiment, the high-frequency region includes multiple frequency regions.

[0052] This allows for state analysis to be performed in a frequency range suitable for the components of the gas turbine engine 10.

[0053] Furthermore, in the mechanical element state analysis method of this embodiment, there are multiple components, and these multiple components constitute at least a part of the gas turbine engine 10. For each of the multiple components, the following steps are taken: preparation of an analysis model, preparation of a sensitivity function, acquisition of operating data, acquisition of output value fluctuations, acquisition of internal parameter fluctuations, and estimation of the fluctuation-responding portion.

[0054] This reduces the computational load and shortens the computation time by performing each process on multiple components that are divisions of at least a portion of the gas turbine engine 10, rather than performing each process on the entire gas turbine engine 10.

[0055] Furthermore, in the mechanical element state analysis method of this embodiment, at least a portion of the multiple analysis models are related such that the input value of one becomes the output value of the other. For one analysis model, the fluctuation of the input value corresponding to the fluctuation of the output value is obtained using a sensitivity function for both the low-frequency and high-frequency regions. If the fluctuation of the input value is large, that large fluctuation is taken as the fluctuation of the output value of the other analysis model, and the fluctuation of the internal parameters is obtained using a sensitivity function in the other analysis model.

[0056] This allows us to leverage the relationships between analysis models to obtain not only the fluctuations in the output values ​​of a given analysis model, but also the fluctuations in the internal parameters of related analysis models, based on the fluctuations in the output values ​​of that particular analysis model.

[0057] Furthermore, in the mechanical element state analysis method of this embodiment, if at least a part of the gas turbine engine 10 fails, the variable-response part is output as the location where the failure is presumed.

[0058] This allows operators and others to identify the location of the malfunction early on.

[0059] Furthermore, the mechanical element state analysis method of this embodiment obtains information about the failure by performing a simulation in which the internal parameters of the analysis model are varied only for the components including the variable-response parts when at least a part of the gas turbine engine 10 fails.

[0060] This allows for obtaining fault information in a shorter time compared to performing parameter variation simulations on the entirety of at least a portion of the gas turbine engine 10.

[0061] Next, a first modified example of the above embodiment will be described. Figure 4 is a flowchart of the condition analysis method for mechanical elements according to this modified example. In the description of this modified example, the same or similar components as those in the above embodiment will be denoted by the same reference numerals in the drawings, and their descriptions may be omitted.

[0062] As shown in Figure 4, in this modified example, multiple submodels (analysis models) corresponding to multiple physical phenomena are prepared for each component (i.e., the compressor 11, fuel control valve 12a, combustor 13, turbine 14, and rotating shaft 15) (S21). Examples of physical phenomena include heat distribution within the component, vibration of the component, flow of air and fuel within the component, and structural strength of the component. For example, a thermal analysis model can be prepared as a submodel corresponding to the heat distribution. Subsequently, for each of the multiple submodels, a sensitivity function is prepared (S22), operating data is acquired (S23), output value fluctuations are acquired (S24), internal parameter fluctuations are acquired (S25), and the fluctuation-corresponding part is estimated (or the fault location is suggested) (S26).

[0063] To briefly explain each step, in the preparation of the sensitivity function (S22), for each submodel, a sensitivity function is created for both the low-frequency and high-frequency regions that defines the relationship between the variation of the internal parameters and the variation of the input value and the variation of the output value. For example, a submodel for thermal analysis may have thermal conductivity, heat transfer coefficient, and specific heat as internal parameters, and temperature as the input and output values.

[0064] In acquiring operational data (S23), for each submodel, operational data corresponding to the input and output values ​​can be acquired using sensor 21 as time-series data going back for an arbitrary period from the time of failure. For example, for a submodel for thermal analysis, time-series data of the temperature at the sensor placement location of the corresponding component can be acquired as operational data.

[0065] In acquiring the output value fluctuations (S24), the output value corresponding to the normal operation of each component is used as the normal data, and the output value fluctuations are acquired for both the low-frequency and high-frequency regions based on the difference between the operating data and the normal data. For example, for the compressor 11, the temperature data acquired by the sensor 21 during normal operation can be input as an input value to the corresponding submodel for thermal analysis, and the temperature data output by the submodel can be used as the normal data. In this case, the output value fluctuations (for example, temperature fluctuations in both the low-frequency and high-frequency regions) can be acquired based on the difference between the normal data and the operating data.

[0066] In acquiring the fluctuations of internal parameters (S25), the fluctuations of internal parameters corresponding to the fluctuations of the output value are acquired for both the low-frequency and high-frequency regions using a sensitivity function. For example, for the compressor 11, it is conceivable to acquire the fluctuations of thermal conductivity (internal parameter) corresponding to the fluctuations of temperature (output value) using a sensitivity function.

[0067] In estimating the variable-response parts (or suggesting fault locations) (S26), the magnitude of the fluctuations in the internal parameters is determined, and then the parts of the components corresponding to the internal parameters with large fluctuations are estimated as the variable-response parts. For example, in the compressor 11, if the fluctuation of the internal parameter thermal conductivity exceeds a predetermined threshold, it may be suggested that there is an abnormality in the thermal insulation structure of the compressor 11 (for example, that each blade or seal part is deteriorating). Note that the method for determining whether the fluctuation of the internal parameters is large is not particularly limited, and depending on the method adopted, it may be possible to omit the determination of the magnitude of the fluctuations in the internal parameters.

[0068] In this embodiment, the simulation results of one submodel may be used in the simulation of another submodel among multiple submodels. For example, data on the temperature distribution within a component obtained from thermal analysis can be input into a submodel for structural analysis. In this case, it becomes possible to perform structural analysis that takes into account the deformation and strength changes of each member due to temperature changes.

[0069] As described above, the method for analyzing the state of mechanical elements in this modified example involves preparing multiple analysis models corresponding to multiple physical phenomena in the components of the gas turbine engine 10 during the preparation of the analysis models. For each of the multiple analysis models, the following steps are taken: preparation of a sensitivity function, acquisition of operating data, acquisition of output value fluctuations, acquisition of internal parameter fluctuations, and estimation of the fluctuation-corresponding portion.

[0070] This allows for a reduction in computational load and shorter computation time by creating an analysis model for each physical phenomenon and performing each step separately, rather than grouping multiple physical phenomena together and performing each step separately.

[0071] Next, a second modified example of the above embodiment will be described. Figure 5 is a flowchart of the condition analysis method for mechanical elements according to this modified example. In the description of this modified example, the same or similar reference numerals are used in the drawings for components that are the same as or similar to those in the above embodiment, and their descriptions may be omitted.

[0072] As shown in Figure 5, in this modified example, for each component (i.e., the compressor 11, fuel control valve 12a, combustor 13, turbine 14, and rotating shaft 15), a submodel is created as a preliminary step (S31), and sensitivity functions for the low-frequency and high-frequency regions are created for each submodel (S32). Subsequently, the acquisition of operating data (S33), the acquisition of output value fluctuations (S34), and the acquisition of internal parameter fluctuations (S35) are performed multiple times (S36). The manner in which each acquisition is performed multiple times is not particularly limited. For example, each acquisition may be performed multiple times periodically or multiple times irregularly. The length of the period between multiple acquisitions is also not particularly limited. For example, each acquisition may be repeated on an hourly, daily, weekly, or monthly basis. The specific method of each acquisition is the same as in the above embodiment, so a detailed explanation is omitted.

[0073] Next, based on the changes in internal parameters acquired over multiple cycles, failure of each component is determined and / or predicted (S37). For example, if the change in a certain internal parameter changes abruptly at a certain point in time, it can be determined that the component corresponding to that internal parameter has failed. As another example, if a certain internal parameter changes by a certain amount compared to its design value, in other words, if the difference between the current value and the design value of a certain internal parameter exceeds a predetermined threshold, failure of the component corresponding to that internal parameter can be predicted. In both the case of failure determination and failure prediction, it is desirable to inform the operator and others accordingly to encourage necessary maintenance.

[0074] As described above, the mechanical element state analysis method of this modified example involves acquiring operating data, output value fluctuations, and internal parameter fluctuations multiple times. Based on the changes in internal parameter fluctuations, failures of the components of the gas turbine engine 10 are determined and / or predicted.

[0075] This makes it possible to perform fault detection and fault prediction on the components of the gas turbine engine 10 while reducing the computational load.

[0076] Preferred embodiments and modifications of the present disclosure have been described above, but the above configuration can be modified as follows, for example. Modifications may be made individually or in any combination of multiple modifications.

[0077] The low-frequency range may include not only 0Hz but also frequencies above 0Hz.

[0078] The high-frequency region may consist of a single frequency region.

[0079] A single machine element may constitute a single device. In this case, multiple analytical models (submodels) corresponding to multiple physical phenomena may be prepared for the device, but this is not the only option.

[0080] The types of equipment are not limited to gas turbine engines 10. For example, hydraulic equipment such as crushers, or trolleys for various vehicles, are also examples of equipment.

[0081] [Aspect] The embodiments described above are specific examples of the following embodiments. (Aspect 1) Prepare an analysis model that models machine elements and has internal parameters, and outputs output values ​​according to the input values. Through simulations using the aforementioned analysis model, a sensitivity function defining the relationship between the fluctuations of the internal parameters and the fluctuations of the input value and the fluctuations of the output value is prepared for both the low-frequency region including 0 Hz and the high-frequency region higher than the low-frequency region. The operating data of the machine element corresponding to the output value is acquired. The output value corresponding to the normal operation of the mechanical element is taken as normal data, and the fluctuation of the output value is acquired for each of the low-frequency and high-frequency regions based on the difference between the operating data and the normal data. For each of the low-frequency region and the high-frequency region, the change in the internal parameter corresponding to the change in the output value is obtained using the sensitivity function. A method for analyzing the state of a mechanical element, wherein the portion of the mechanical element corresponding to the internal parameter with large fluctuations is estimated as the fluctuation-responding portion. (Aspect 2) A method for analyzing the state of a mechanical element as described in Embodiment 1, A method for analyzing the state of a mechanical element, wherein the low-frequency region is limited to 0 Hz. (Aspect 3) A method for analyzing the condition of a mechanical element according to embodiment 1 or 2, The aforementioned high-frequency region includes multiple frequency regions, and is a method for analyzing the state of a mechanical element. (Aspect 4) A method for analyzing the state of a mechanical element as described in any one of embodiments 1 to 3, There are multiple such mechanical elements, and these multiple mechanical elements constitute at least a part of a single device. A method for analyzing the state of a machine element, comprising, for each of the plurality of machine elements, preparing the analysis model, preparing the sensitivity function, acquiring operating data, acquiring the fluctuation of the output value, acquiring the fluctuation of the internal parameter, and estimating the fluctuation-responding portion. (Appendix 5) A method for analyzing the state of a mechanical element as described in Embodiment 4, At least some of the aforementioned multiple analysis models are related such that the input value of one becomes the output value of the other, For the aforementioned analysis model, the variation in the input value corresponding to the variation in the output value is obtained for each of the low-frequency and high-frequency regions using the sensitivity function. A method for analyzing the state of a mechanical element, wherein, when the fluctuation of the input value is large, the large fluctuation is used as the fluctuation of the output value of the other analysis model, and the fluctuation of the internal parameter is obtained in the other analysis model using the sensitivity function. (Aspect 6) A method for analyzing the condition of a mechanical element according to embodiment 4 or 5, A method for analyzing the condition of a mechanical element, wherein, when at least a part of the equipment malfunctions, the variable-response portion is output as a location where a malfunction is presumed. (Aspect 7) A method for analyzing the state of a mechanical element as described in any one of embodiments 4 to 6, A method for analyzing the state of a mechanical element, wherein, when at least a part of the equipment fails, a simulation is performed to vary the internal parameters of the analysis model for only the mechanical element, including the variable-response portion, thereby obtaining information about the failure. (Pattern 8) A method for analyzing the state of a mechanical element as described in any one of embodiments 1 to 7, In preparing the analysis model, a plurality of analysis models corresponding to a plurality of physical phenomena in the machine element are prepared. A method for analyzing the state of a mechanical element, comprising preparing the sensitivity function, acquiring operating data, acquiring fluctuations in the output value, acquiring fluctuations in the internal parameters, and estimating the fluctuation-responding portion for each of the multiple analysis models. (Aspect 9) A method for analyzing the state of a mechanical element as described in any one of embodiments 1 to 8, The acquisition of the aforementioned operating data, the acquisition of the aforementioned changes in output values, and the acquisition of the aforementioned changes in internal parameters are performed multiple times. A method for analyzing the state of a mechanical element, which determines and / or predicts a failure of the mechanical element based on the changes in the internal parameters. [Explanation of Symbols]

[0082] 10. Gas turbine engine (equipment) 11. Compressor (mechanical element) 12 Fuel supply pipe 12a Fuel control valve (mechanical element) 13. Combustor (mechanical element) 14. Turbine (mechanical element) 15. Rotating shaft (mechanical element) 21 sensors

Claims

1. Prepare an analysis model that models machine elements and has internal parameters, and outputs output values ​​according to the input values. Through simulations using the aforementioned analysis model, a sensitivity function defining the relationship between the fluctuations of the internal parameters and the fluctuations of the input value and the fluctuations of the output value is prepared for both the low-frequency region including 0 Hz and the high-frequency region higher than the low-frequency region. The operating data of the machine element corresponding to the output value is acquired. The output value corresponding to the normal operation of the mechanical element is taken as normal data, and the fluctuation of the output value is acquired for each of the low-frequency and high-frequency regions based on the difference between the operating data and the normal data. For each of the low-frequency region and the high-frequency region, the change in the internal parameter corresponding to the change in the output value is obtained using the sensitivity function. A method for analyzing the state of a mechanical element, wherein the portion of the mechanical element corresponding to the internal parameter with large fluctuations is estimated as the fluctuation-responding portion.

2. A method for analyzing the state of a mechanical element according to claim 1, A method for analyzing the state of a mechanical element, wherein the low-frequency region is limited to 0 Hz.

3. A method for analyzing the state of a mechanical element according to claim 1 or 2, The aforementioned high-frequency region includes multiple frequency regions, and is a method for analyzing the state of a mechanical element.

4. A method for analyzing the state of a mechanical element according to claim 1 or 2, There are multiple such mechanical elements, and these multiple mechanical elements constitute at least a part of a single device. A method for analyzing the state of a machine element, comprising, for each of the plurality of machine elements, preparing the analysis model, preparing the sensitivity function, acquiring operating data, acquiring the fluctuation of the output value, acquiring the fluctuation of the internal parameter, and estimating the fluctuation-responding portion.

5. A method for analyzing the state of a mechanical element according to claim 4, At least some of the aforementioned multiple analysis models are related such that the input value of one becomes the output value of the other, For the aforementioned analysis model, the variation in the input value corresponding to the variation in the output value is obtained for each of the low-frequency and high-frequency regions using the sensitivity function. A method for analyzing the state of a mechanical element, wherein, when the fluctuation of the input value is large, the large fluctuation is used as the fluctuation of the output value of the other analysis model, and the fluctuation of the internal parameter is obtained in the other analysis model using the sensitivity function.

6. A method for analyzing the state of a mechanical element according to claim 4, A method for analyzing the condition of a mechanical element, wherein, when at least a part of the equipment malfunctions, the variable-response portion is output as a location where a malfunction is presumed.

7. A method for analyzing the state of a mechanical element according to claim 4, A method for analyzing the state of a mechanical element, wherein, when at least a part of the equipment fails, a simulation is performed to vary the internal parameters of the analysis model for only the mechanical element, including the variable-response portion, thereby obtaining information about the failure.

8. A method for analyzing the state of a mechanical element according to claim 1 or 2, In preparing the analysis model, a plurality of analysis models corresponding to a plurality of physical phenomena in the machine element are prepared. A method for analyzing the state of a mechanical element, comprising preparing the sensitivity function, acquiring operating data, acquiring fluctuations in the output value, acquiring fluctuations in the internal parameters, and estimating the fluctuation-responding portion for each of the multiple analysis models.

9. A method for analyzing the state of a mechanical element according to claim 1 or 2, The acquisition of the aforementioned operating data, the acquisition of the aforementioned changes in output values, and the acquisition of the aforementioned changes in internal parameters are performed multiple times. A method for analyzing the state of a mechanical element, which determines and / or predicts a failure of the mechanical element based on the changes in the internal parameters.

Images