Evaluation method, evaluation program, and evaluation device

Abstract

To evaluate a vortex state by a simple procedure as compared with a conventional technique.SOLUTION: A calculation step of calculating, based on an input parameter related to a range including a suction port of a pump, an output parameter that can be associated with a physical variable including at least one of a physical quantity related to the range and an amount of calculation calculated from the physical quantity, and a time variable indicating a change over time of the physical variable, and when an explanatory variable that can be associated with the physical variable is input, giving an input based on the output parameter calculated in the calculation step to a learned model that outputs an evaluation related to a state of a vortex in the range; And an output step of outputting a time-series evaluation result based on the evaluation relating to the temporal change in the state of the vortex in the evaluation step and the time variable included in the output parameter.SELECTED DRAWING: None

Description

[Technical Field]

[0001] The present invention relates to an evaluation method, an evaluation program, and an evaluation device for evaluating the state of a vortex at the suction port of a pump. [Background technology]

[0002] At the suction port of a pump operating in a pump intake sump, vortices may occur depending on the pump structure, operating conditions, etc. Accurate evaluation of vortex generation is necessary to optimize the pump structure, sump shape, flow pattern, and operating conditions.

[0003] For example, Patent Document 1 and Non-Patent Document 1 disclose a method for observing a fluid, characterized in that a camera is used to photograph the background of an observation area through the fluid, and the photographed image data is recorded and then analyzed. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 11-237397 [Non-patent literature]

[0005] [Non-Patent Document 1] Turbomachinery, Vol. 26, No. 3 (March 1998), pp. 28-32 Summary of the Invention [Problem to be solved by the invention]

[0006] The observation methods described in Patent Document 1 and Non-Patent Document 1 require that an image when no vortex is generated be recorded in advance as a reference image. Furthermore, several processing steps are required to obtain an image that can be used to determine the state of the vortex, such as a subtraction process between the observed image and the reference image and a binarization process of the subtracted image. Furthermore, in order to improve the accuracy of determining the state of the vortex, various conditions may need to be considered when capturing the image.

[0007] Therefore, there is a need to realize an evaluation method, evaluation program, and evaluation device that can evaluate the state of vortices using simpler procedures than those of conventional techniques. [Means for solving the problem]

[0008] The evaluation method of the present invention is characterized by comprising: a calculation step of calculating, based on input parameters relating to a range including at least the pump inlet, output parameters that can be associated with physical variables including at least one of a physical quantity relating to the range and a calculation amount calculated from the physical quantity, and a time variable that indicates changes in the physical variables over time; an evaluation step of providing input based on the output parameters calculated in the calculation step to a trained model that outputs an evaluation of the state of a vortex in the range when explanatory variables that can be associated with physical variables including at least one of a physical quantity relating to the range including at least the pump inlet and a calculation amount calculated from the physical quantity, thereby obtaining an evaluation of the change in the state of the vortex over time that corresponds to the output parameters; and an output step of outputting a time-series evaluation result based on the evaluation of the change in the state of the vortex over time in the evaluation step and the time variable included in the output parameters.

[0009] The evaluation program of the present invention, when executed on a computer, is characterized by realizing the following: a calculation function that calculates, based on input parameters relating to a range including at least the pump inlet, output parameters that can be associated with physical variables including at least one of a physical quantity relating to the range and a calculation quantity calculated from the physical quantity, and a time variable that indicates changes in the physical variables over time; an evaluation function that, when an explanatory variable that can be associated with physical variables including at least one of a physical quantity relating to the range including at least the pump inlet and a calculation quantity calculated from the physical quantity is input, provides input based on the output parameters calculated by the calculation function to a trained model that obtains an evaluation of the state of a vortex in the range, thereby obtaining an evaluation of the change in the state of the vortex over time that corresponds to the output parameters; and an output function that outputs time-series evaluation results based on the evaluation of the change in the state of the vortex over time in the evaluation function and the time variable included in the output parameters.

[0010] The evaluation device of the present invention is characterized in that it comprises a computer capable of realizing: a calculation function that calculates, based on input parameters relating to a range including at least the pump inlet, output parameters that can be associated with physical variables including at least one of a physical quantity relating to the range and a calculation quantity calculated from the physical quantity, and a time variable that indicates changes in the physical variables over time; an evaluation function that, when an explanatory variable that can be associated with physical variables including at least one of a physical quantity relating to the range including at least the pump inlet and a calculation quantity calculated from the physical quantity is input, provides input based on the output parameters calculated by the calculation function to a trained model that obtains an evaluation of the state of a vortex in the range, thereby obtaining an evaluation of the change in the state of the vortex over time that corresponds to the output parameters; and an output function that outputs time-series evaluation results based on the evaluation of the change in the state of the vortex over time in the evaluation function and the time variable included in the output parameters.

[0011] These configurations make it possible to evaluate the state of a vortex using simpler procedures than conventional techniques, without requiring consideration of image data acquisition and preprocessing.

[0012] Preferred embodiments of the present invention will be described below, but the scope of the present invention is not limited to the preferred embodiments described below.

[0013] In one aspect of the evaluation method of the present invention, it is preferable that the output parameters and the computational quantities related to the trained model include at least one selected from the group consisting of vorticity, Laplacian of pressure, helicity of pressure, and vortex axis.

[0014] This configuration makes it easy to understand the relationship between the output parameters and the state of the vortex.

[0015] In one aspect of the evaluation method of the present invention, the output parameters include an image generated based on the physical variables and showing the state of the range, and the explanatory variables of the trained model include an image generated based on physical variables including at least one of a physical quantity related to a range including at least the suction port of the pump and a calculation quantity calculated from the physical quantity, and in the evaluation process, it is preferable to provide the trained model with the image generated in the calculation process to obtain an evaluation of the state of the vortex corresponding to the image.

[0016] According to this configuration, the output parameters are handled in a form including images that are easy for humans to understand, making it easy to verify the validity of the evaluation results.

[0017] In one aspect of the evaluation method according to the present invention, the output step preferably includes outputting a graph representing the time-series evaluation results.

[0018] According to this configuration, the change in the evaluation points over time is visualized, making it easier to understand the extent of the vortex.

[0019] In one aspect of the evaluation method according to the present invention, the output step preferably includes outputting a graph representing a moving average of the time-series evaluation results.

[0020] According to this configuration, by taking the evaluation point at each time point as a moving average, it is possible to perform an evaluation that takes into account the continuity of the physical phenomenon when a vortex is generated.

[0021] In one aspect of the evaluation method of the present invention, in the output process, the graph is displayed on a plane having an axis representing time and an axis representing the time-series evaluation results or a moving average of the time-series evaluation results, and a curve representing the progress of the evaluation results or the moving average is preferably displayed, and the time-series evaluation results of the vortex state are indexed based on the relationship between the curve and an arbitrary straight line parallel to the axis representing time.

[0022] According to this configuration, the time-series evaluation results of the vortex state can be indexed, making it easy to compare and evaluate the results of multiple analyses performed under different conditions.

[0023] In the evaluation method of the present invention, in the output process, it is preferable that the time-series evaluation results of the vortex state are indexed based on the area of ​​the region surrounded by an arbitrary straight line parallel to the axis representing time and the curve.

[0024] This configuration allows evaluation to be performed taking into account both the strength and duration of the vortex.

[0025] The evaluation method of the present invention preferably further includes a correction step of correcting the evaluation of the vortex state based on a specific output parameter using an evaluation of the vortex state based on a time-series output parameter different from the specific output parameter, and in the output step, the time-series evaluation result is output based on the evaluation of the vortex state corrected in the correction step.

[0026] This configuration can improve the accuracy of the evaluation.

[0027] In the evaluation method according to the present invention, it is preferable that the state of the vortex is classified into a plurality of levels depending on the degree of vortex formation, and a predetermined evaluation score is set for each of the classified levels.

[0028] According to this configuration, the state of the vortex can be indexed using the evaluation score, making it easy to understand the extent of the vortex.

[0029] Further features and advantages of the present invention will become more apparent from the following description of exemplary and non-limiting embodiments, which is given with reference to the drawings. [Brief explanation of the drawings]

[0030] [Figure 1] FIG. 2 is a schematic diagram of an area to be evaluated in the evaluation method according to the embodiment. [Figure 2] FIG. 10 is a diagram showing an example of an image output in a calculation step of the evaluation method according to the embodiment. [Figure 3] FIG. 10 is a diagram showing an example of an image output in a calculation step of the evaluation method according to the embodiment. [Figure 4] FIG. 10 is a schematic diagram showing a state in which a depression vortex is generated. [Figure 5] FIG. 10 is a schematic diagram showing a state in which an intermittent vortex is generated. [Figure 6] FIG. 10 is a schematic diagram showing a state in which a continuous vortex is generated. [Figure 7] FIG. 10 is a diagram showing an example of a graph output by the observation device according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0031] An evaluation method, an evaluation program, and an evaluation device according to an embodiment of the present invention will be described below with reference to the drawings. In the following, the present invention will be described using as an example a case in which the observation target is an area including the suction port 11 of a pump installed in a test tank 10 (FIG. 1).

[0032] [Configuration of evaluation device] The evaluation device according to this embodiment includes a computer on which the evaluation program according to this embodiment is installed. The computer constituting the evaluation device may be a known computer. Furthermore, the number of computers constituting the evaluation device is not limited, and a single computer may realize all functions, or multiple computers may share and realize each function. When the evaluation device includes multiple computers, each computer may be capable of communication. The communication path may be either wired or wireless. Furthermore, each computer may be capable of communication via a network such as the Internet.

[0033] [Evaluation method structure] Next, an evaluation method according to this embodiment will be described. The evaluation method according to this embodiment is realized through the functions of the evaluation device according to this embodiment, and each function of the computer of the evaluation device is realized by executing an evaluation program, which is one embodiment of the evaluation program according to the present invention. Therefore, the following description will not only describe each step of the evaluation method according to this embodiment, but also each function related to the evaluation program and evaluation device according to this embodiment.

[0034] (1) Arithmetic process (arithmetic function) The calculation step is a step of calculating, based on input parameters relating to a range, output parameters that can be associated with physical variables including at least one of a physical quantity relating to the range and a calculation quantity calculated from the physical quantity, and a time variable that indicates a change in the physical variable over time. Furthermore, in the calculation step, an image showing the state of the range may be generated based on the obtained output parameters.

[0035] The input parameters are a group of parameters that characterize the range to be evaluated. Specific parameters included in the input parameters include, but are not limited to, the shape and dimensions of the suction port 11, the shape and dimensions of the test tank 10, the shape and dimensions of the pump impellers, the pump rotation speed and discharge rate, the water temperature in the test tank 10, and the water flow rate in the test tank 10.

[0036] The output parameters calculated based on the input parameters can be associated with physical variables including at least one of a physical quantity related to a range and a calculation quantity calculated from the physical quantity, and a time variable indicating a change in the physical variables over time. As an example, the output parameters are a set of one or more physical variables and a time variable indicating a change in the physical variables over time. The physical variables and the time variable are associated with each portion included in the range. For example, each portion of the range is identified in a three-dimensional coordinate system, and the physical variables and the time variable are associated with the coordinates of each portion.

[0037] Examples of physical quantities related to a range include, but are not limited to, pressure, mass, density, velocity, acceleration, viscosity, etc. at each coordinate of the range. Examples of calculation quantities calculated from physical quantities include, but are not limited to, vorticity (a calculation quantity calculated from velocity), Laplacian of pressure (a calculation quantity calculated from pressure), helicity of pressure (a calculation quantity calculated from pressure), and vortex axis (a calculation quantity calculated from velocity). The number and combination of physical quantities and calculation quantities included in a physical variable are not limited. Therefore, a physical variable may be a single physical quantity, a single calculation quantity, a combination of multiple physical quantities, a combination of multiple calculation quantities, or a combination of a single or multiple physical quantities and a single or multiple calculation quantities.

[0038] A time variable is a variable that indicates the change in a physical variable over time. If the calculation method for calculating output parameters based on input parameters is unsteady analysis, the output parameters essentially contain a time element, and therefore can be a time variable. If the calculation method is steady-state analysis, the output parameters do not contain a time element, but it is possible to consider changes over time in a pseudo-time step method or the like, and thus identify a pseudo-time variable. Note that some physical variables contain a time element, and in this case, the physical variable in question is also a time variable. For example, speed, which is an example of a physical quantity, is a physical quantity that contains a time dimension, so a physical variable that contains speed is also a time variable.

[0039] The calculation method for calculating the output parameters based on the input parameters may be, for example, a known method of computational fluid dynamics analysis (CFD analysis). Such computational fluid dynamics analysis can be performed using software such as Ansys (registered trademark) Fluent (manufactured by Ansys, Inc.). The computational fluid dynamics analysis may be performed on the entire range or only on a part of the range. It is possible to perform computational fluid dynamics analysis on only the liquid phase portion of the range (excluding the gas phase portion).

[0040] Non-limiting examples of analysis conditions for computational fluid dynamics analysis are shown below, but the following analysis conditions are merely examples. Analysis type: Steady-state analysis Number of phases: Single phase (liquid phase only) Gravity: Not considered Solver: Pressure-based SIMPLE method Turbulence model: RANS Standard k-ε model Discretization method: quadratic precision Convergence test: Residual 10 -5 less than Fluid properties: Water density 998.2 kg / m 3 ,Viscosity 0.001Pa·s

[0041] The calculation step may optionally generate an image showing the state of the area based on the obtained physical variables. Because this image is based on the physical variables, it is one form of output parameter that can be associated with the physical variables. For example, by determining the representation format (color, shape, etc.) of each coordinate of the area based on the physical variables of that coordinate, and then depicting each coordinate according to that representation format, an image visualizing the distribution of the physical variables in the area can be obtained.

[0042] Figures 2 and 3 are example images that visualize the distribution of vorticity in a range, with coordinates where the vorticity is above a predetermined threshold (air-entraining vortices 23 / s or higher, underwater vortices 37 / s or higher) colored. In Figures 2 and 3, the colored areas are marked with the symbol V. In Figure 2, coordinates with high vorticity are present locally around the inlet 11 but are almost absent near the water surface, suggesting that no vortices will occur. In Figure 3, coordinates with high vorticity are present continuously between the inlet 11 and the water surface, suggesting that vortices will occur. In this way, by generating an image showing the state of a range based on physical variables, it becomes easier to visually determine the presence and extent of vortices in the state represented by the physical variables.

[0043] Furthermore, in the calculation step, an image representing a range state estimated from the input parameters may be output using a surrogate model or the like. In this case, the physical variables themselves are not output. However, since the range state can be associated with the physical variables, the image representing the range state can be associated with the physical variables. Therefore, an image output using a surrogate model or the like is also one form of output parameters that can be associated with physical variables. As can be understood from this example, as long as the output parameters can be associated with the physical variables and the time variables, the output parameters do not need to include one or both of the physical variables and the time variables themselves.

[0044] (2) Evaluation process (evaluation function) The evaluation process is a process of using the trained model to output an evaluation of the state of the vortex corresponding to the output parameter from the output parameter.

[0045] (Pre-trained model) The trained model used in the evaluation process may be generated by the evaluation device itself or by another device. The trained model outputs an evaluation of the state of vortices in the range of the evaluation target when explanatory variables that can be associated with physical variables including at least one of physical quantities related to the range of the evaluation target and computational quantities calculated from the physical quantities, and a time variable representing changes in the physical variables over time, are input. Here, the state of vortices is classified, for example, according to TSJ S 002 and selected from four states (examples of multiple levels): (i) a state without vortices (FIG. 1), (ii) a state with a dip vortex (FIG. 4), (iii) a state with an intermittent vortex (FIG. 5), and (iv) a state with a continuous vortex (FIG. 6).

[0046] The trained model is generated through supervised learning. The method for creating training data for supervised learning is described below.

[0047] First, using a method similar to that used in the calculation step, output parameters are calculated based on input parameters for a range including at least the pump inlet port, and the output parameters are associated with physical variables including at least one of the physical quantities related to the range and the calculation quantities calculated from the physical quantities. An appropriate number of sets (e.g., 200 or more sets) of input parameters are prepared as the number of data points constituting the training data, and a group of output parameters corresponding to each input parameter is obtained. For example, if the output parameters include vorticity as a calculation quantity, the output parameters constituting the training data also include vorticity.

[0048] Next, an image is generated based on each output parameter in the same manner as in the calculation process. As described above, generating an image makes it easier to visually determine the presence or absence and degree of a vortex, and therefore makes it easier to artificially determine the presence or absence and degree of a vortex in the state represented by the output parameters.

[0049] Next, for each of the generated images, the vortex state is manually determined as one of the above (i) to (iv). Through this procedure, a dataset is obtained in which each image is labeled with a vortex state. However, each of the multiple images used in this determination corresponds one-to-one with each of the output parameter groups. Therefore, through the procedure of labeling each image with a vortex state, a state in which not only the image but also the output parameters are labeled with a vortex state is achieved. Therefore, the training data in supervised learning may be a dataset related to pairs of images and vortex states, or a dataset related to pairs of output parameters and vortex states.

[0050] When a dataset relating to a pair of an image and a state of a vortex is used as training data, a trained model is obtained in which the image is an explanatory variable and the state of a vortex is a target variable. When a dataset relating to a pair of an output parameter and a state of a vortex is used as training data, a trained model is obtained in which the output parameter is an explanatory variable and the state of a vortex is a target variable. In the following, an example will be described in which a dataset relating to a pair of an image and a state of a vortex is used as training data.

[0051] The algorithm used to generate a trained model from training data is not particularly limited. Examples include, but are not limited to, support vector machines (regression, classification), decision trees, random forests, gradient boosting, logistic regression, neural networks (simple perceptron, multilayer perceptron), Gaussian process regression, Bayesian networks, k-nearest neighbors, lasso regression, multiple regression analysis, ridge regression, elastic nets, and partial least squares regression. Among these, it is preferable to use a neural network that is highly suitable for image recognition, and it is particularly preferable to use a convolutional neural network. Note that the validity of the generated trained model may be verified by using part of the training data and the rest as test data.

[0052] (Implementation of evaluation) When the image obtained in the calculation step is input to the trained model, the state of the vortex selected from (i) to (iv) above is output as the evaluation result. The state of the vortex selected from (i) to (iv) is an example of the evaluation of the state of the vortex in the evaluation step. In addition, the reliability of the evaluation result is output together with the evaluation result. The reliability is expressed, for example, as a percentage, and indicates the accuracy of the evaluation result. Note that, since the input parameters include a time variable as described above, the evaluation result and reliability output in the evaluation step can be associated with the time variable.

[0053] Furthermore, to facilitate subsequent calculations, the evaluation results are quantified based on the evaluation scores assigned to each of the states (i) to (iv). The evaluation scores are an example of the evaluation of the vortex state in the evaluation process. The magnitude of the evaluation scores assigned to each state can be changed depending on the purpose of the observation. For example, if continuous vortices, intermittent vortices, and recessed vortices are all of interest, allocating evaluation scores equally to each state as in the example above makes it easier to recognize changes in the state. On the other hand, if the purpose of the observation is to identify a structure for the suction port 11 that can eliminate continuous vortices and intermittent vortices and recessed vortices are not of interest, assigning a relatively large evaluation score to "(iv) a state in which a continuous vortex is generated" makes it easier to grasp the occurrence of a continuous vortex. In the following, as an example, the evaluation score for "(i) a state in which no vortex is present" is 0.25, the evaluation score for "(ii) a state in which a recessed vortex is generated" is 0.50, the evaluation score for "(iii) a state in which an intermittent vortex is generated" is 0.75, and the evaluation score for "(iv) a state in which a continuous vortex is generated" is 1.0.

[0054] Using the evaluation score, the evaluation result for a certain time series (certain image) is obtained as a pair of evaluation score and reliability. Furthermore, multiplying the evaluation score and reliability results in a numerical value that takes into account both the evaluation of the vortex state and the reliability of that evaluation. Hereinafter, this numerical value will be referred to as the "estimated value." The estimated value is an example of the evaluation of the vortex state in the evaluation process.

[0055] The above evaluation is performed for each output parameter for each time series. Since each output parameter has a time variable, an estimated value is output for each time series.

[0056] Note that when a data set relating to a pair of output parameters and vortex states is used as training data, it is not necessary to generate an image in the calculation step. In this case, when the output parameters calculated in the calculation step are input to the trained model, a vortex state selected from (i) to (iv) above is output as the evaluation result. In this case, when creating training data to be used to generate the trained model, it is not prohibited to generate an image based on the output parameters as a means for artificially identifying the vortex state corresponding to the output parameters.

[0057] Regardless of which trained model is used, the input provided to the trained model in the evaluation process is based on the output parameters (the output parameters themselves, and images based on the output parameters).

[0058] (3) Correction process (correction function) The correction step is a step of correcting the evaluation of the state of the vortex based on a specific output parameter using an evaluation of the state of the vortex based on an output parameter of a time series different from the specific output parameter.

[0059] As mentioned above, the vortex state can be selected from (i) a state where there is no vortex (Fig. 1), (ii) a state where a dip vortex is generated (Fig. 4), (iii) a state where an intermittent vortex is generated (Fig. 5), and (iv) a state where a continuous vortex is generated (Fig. 6), and these four states are arranged in order of the degree of vortex from weak to strong. In reality, changes in the vortex state are a continuous phenomenon, so it is unlikely that the state will change to one that deviates significantly from the previous state.

[0060] For example, if the assessment of the vortex state based on a certain output parameter is "(iv) A state in which a continuous vortex is occurring," and the assessment of the vortex state based on the output parameters of the immediately preceding time series is "(iii) A state in which an intermittent vortex is occurring," this analysis result can be understood as an analysis result that shows a continuous phenomenon in which an intermittent vortex increases in strength and becomes a continuous vortex, and corresponds to a phenomenon that may actually occur. In other words, there is a certain degree of certainty that the assessment of the vortex state based on a certain output parameter is "(iv) A state in which a continuous vortex is occurring." Hereinafter, this example will be referred to as "Example 1."

[0061] On the other hand, if the assessment of the vortex state based on a certain output parameter is "(iv) A state in which a continuous vortex is occurring," and the assessment of the vortex state based on the output parameters of the time series immediately preceding that is "(i) A state in which there is no vortex," then it can be said that the analysis results show a discontinuous phenomenon in which the state changes suddenly from a state in which there is no vortex to a state in which there is a continuous vortex. However, such a sudden change is unlikely to occur in reality. Therefore, there is doubt about the accuracy of the assessment of the vortex state based on a certain output parameter being "(iv) A state in which a continuous vortex is occurring." This example is called "Example 2."

[0062] Therefore, in the correction process, the evaluation of the state of the vortex to be corrected (the state of the vortex based on a specific output parameter) is corrected using the evaluation of the state of the vortex based on an output parameter that precedes the output parameter in the time series. The correction method is not limited, but one example is a method in which the average of an estimated value (the product of the evaluation score corresponding to the estimated state of the vortex and the reliability of the estimation) calculated based on a specific output parameter and the estimated values ​​of a predetermined number of output parameters (e.g., three) that precede the output parameter in the time series, i.e., the moving average of the estimated values, is used as the corrected value of the estimated value calculated based on the specific output parameter (hereinafter referred to as the "corrected estimated value"). As mentioned above, since the output parameter includes a time variable, the time series of the output parameter can be identified based on the time variable.

[0063] In the above "Example 1," the moving average of the evaluation points corresponding to the output parameter to be evaluated is 0.81 ((1.0 + 0.75 × 3) ÷ 4 = 0.8125), a value that is valid as the point at which the state of the vortex based on the output parameter has transitioned from "(iii) a state in which intermittent vortices are occurring" to "(iv) a state in which continuous vortices are occurring," and the state of the vortex based on the output parameter is expressed by this value. On the other hand, in the above "Example 2," the moving average of the evaluation points corresponding to the output parameter to be evaluated is 0.44 ((1.0 + 0.25 × 3) ÷ 4 = 0.4375), a value that expresses a certain degree of doubt about the estimation result that the state of the vortex based on the output parameter is "(iv) a state in which continuous vortices are occurring."

[0064] (4) Output process (output function) The output step is a step of outputting a time-series evaluation result based on the evaluation of the vortex state in the evaluation step and the time variable included in the output parameter. Here, being based on the evaluation of the vortex state in the evaluation step includes not only using the evaluation of the vortex state in the evaluation step as is, but also using the evaluation of the vortex state corrected in, for example, the correction step. In other words, it is sufficient that the evaluation of the vortex state in the evaluation step is used directly or indirectly.

[0065] In this embodiment, the corrected estimated value determined in the correction process (which is an example of an evaluation of the state of the vortex corrected in the correction process) is plotted in chronological order. That is, a graph is output with the horizontal axis representing time and the vertical axis representing the corrected estimated value (Figure 7). In this graph, areas where the corrected estimated value is large represent time periods in which the strength of the vortex is estimated to be relatively strong (intermittent or continuous vortices are occurring). In this way, by looking at the graph output in the output process, it is possible to easily understand the changes in the state of the vortex over time.

[0066] Furthermore, as an overall evaluation of the input parameters to be evaluated, the area of ​​the region enclosed by the curve connecting the plotted corrected estimated value in the graph with the corrected estimated values ​​plotted at the preceding and following times (hereinafter referred to as the evaluation curve) and the horizontal axis (time axis) may be calculated, and this may be used as an index (score) for evaluation. This index is referred to as the first score value. Since the longer the time period during which the strength of the vortex is strong, the larger the first score value, by comparing the first score values ​​of multiple evaluations, the amount and degree of the vortex can be accurately compared. In other words, the input parameters can be evaluated based on the first score value, either manually or by an evaluation device.

[0067] Alternatively, a line parallel to the horizontal axis (time axis) may be drawn at any corrected estimated value, and the area enclosed by the line and the evaluation curve may be calculated and used as an index (score) for evaluation. This index is referred to as the second score value. For example, if a line parallel to the horizontal axis is drawn at a corrected estimated value of 0.60 and the area enclosed by the line and the evaluation curve in the region above the line (where the corrected estimated value is large) is calculated, the area will be positive for periods when intermittent vortices or continuous vortices occur, and zero for periods when depression vortices occur or when no vortices occur. Comparing the first score value and the second score value, when the first score value is used, the first score value will be larger if the period of occurrence of depression vortices is long, whereas when the second score value is used, depression vortices do not affect the second score value. Therefore, using the second score value allows evaluation to be performed with emphasis on the length of the period when intermittent vortices or continuous vortices occur and the strength of the vortices. For example, this method is effective when it is necessary to find input parameters that meet the condition that sags are allowed but sags larger than intermittent sags are not allowed.

[0068] Furthermore, the area of ​​the region enclosed by each of the multiple straight lines and the evaluation curve may be calculated, and the evaluation may be performed by comprehensively considering these. For example, by calculating both the first score value and the second score value, it is possible to find input parameters that increase the frequency of vortex generation but decrease the strength, and input parameters that decrease the frequency of vortex generation but increase the strength of vortexes, etc. In other words, it becomes easier to evaluate input parameters along multiple evaluation axes.

[0069] Alternatively, instead of calculating the area, a straight line parallel to the horizontal axis (time axis) may be drawn at any corrected estimated value, and the ratio of the length of time the evaluation curve is above the straight line (when the corrected estimated value is large) to the length of time the evaluation curve is below the straight line (when the corrected estimated value is small) may be calculated, and this may be used as an index (score) to be used for evaluation. This index is referred to as a third score value. Note that in this case as well, multiple straight lines may be drawn to calculate multiple indices, and the input parameters may be evaluated by combining these indices.

[0070] The input parameters are a group of parameters that characterize the range to be evaluated. Therefore, the evaluation of the input parameters obtained in each of the above examples can be understood as an evaluation of the range, and in particular, an evaluation of the pump. Therefore, through the evaluation of the input parameters, conditions that provide a suitable pump can be found.

[0071] Other Embodiments Finally, other embodiments of the evaluation method, evaluation program, and evaluation device according to the present invention will be described. Note that the configurations disclosed in the following embodiments can be applied in combination with the configurations disclosed in other embodiments, as long as no contradiction occurs.

[0072] In the above embodiment, an evaluation method including a correction step has been described as an example. However, in the present invention, the presence or absence of the correction step is optional. Examples of a configuration that does not implement the correction step include a configuration in which the estimated value (estimated evaluation point) itself is plotted in the output step, or a configuration in which the product of the estimated value and the reliability is plotted. Note that regardless of the form of the output step, it is possible to index the time series of the vortex state, as exemplified by the first score value, second score value, and third score value in the above embodiment.

[0073] In the above embodiment, an example has been described in which the correction process corrects the evaluation of the vortex state to be corrected (the evaluation of the vortex state based on a specific output parameter) using an evaluation of the vortex state based on an output parameter that precedes the output parameter in the time series. However, the gist of the correction process is to correct the evaluation of the vortex state based on a specific output parameter using an evaluation of the vortex state based on an output parameter that is different from the specific output parameter. Therefore, the chronological relationship between the output parameter on which the evaluation of the vortex state to be corrected is based and the output parameter on which the evaluation of the vortex state used for correction is based is not limited.

[0074] In the above embodiment, an example has been described in which the state of a vortex is classified in accordance with TSJ S 002. However, when classifying the state of a vortex in the present invention, the manner of classification is not limited.

[0075] In the above embodiment, an example has been described in which a trained model is generated using training data based on output data calculated in the same manner as the calculation process. However, the present invention is not limited to a method for generating a trained model. As an example, training data created by an experimental method may be used. In this case, for example, first, a plurality of still image data are captured of an area including the air inlet, and the vortex state of each of the still image data is manually determined. This makes it possible to create a data set in which each still image data is labeled with the vortex state. This data set may be used as training data. Therefore, as one aspect, the present invention can be an evaluation method having: a calculation step of calculating, based on input parameters relating to a range including at least the pump inlet, output parameters including physical variables including at least one of a physical quantity relating to the range and a calculation quantity calculated from the physical quantity, a time variable indicating a change in the physical variable over time, and an image generated based on the physical variables and indicating the state of the range; an evaluation step of inputting the image included in the output parameters calculated in the calculation step into a trained model that evaluates the state of a vortex included in an image of a range including at least the pump inlet when the image is input, to obtain an evaluation of the state of the vortex corresponding to the output parameters; and an output step of outputting a time-series evaluation result based on the evaluation of the state of the vortex in the evaluation step and the time variable included in the output parameters.

[0076] Regarding other configurations, it should be understood that the embodiments disclosed in this specification are illustrative in all respects and that the scope of the present invention is not limited thereby. Those skilled in the art will easily understand that appropriate modifications are possible without departing from the spirit of the present invention. Therefore, other embodiments modified without departing from the spirit of the present invention are naturally included in the scope of the present invention. [Industrial Applicability]

[0077] The present invention can be used, for example, to evaluate pumps.

Claims

1. a calculation step of calculating, based on input parameters relating to a range including at least the suction port of the pump, output parameters that can be associated with physical variables including at least one of a physical quantity relating to the range and a calculation quantity calculated from the physical quantity, and a time variable that indicates a change in the physical variable over time; an evaluation step of providing an input based on the output parameters calculated in the calculation step to a trained model that outputs an evaluation of the state of a vortex in a range when explanatory variables that can be associated with physical variables including at least one of a physical quantity related to a range including at least the suction port of the pump and a calculation quantity calculated from the physical quantity are input, thereby obtaining an evaluation of the change over time in the state of a vortex corresponding to the output parameters; An evaluation method comprising an output process for outputting a time-series evaluation result based on the evaluation of the change in the state of the vortex over time in the evaluation process and the time variable included in the output parameters.

2. The evaluation method according to claim 1, wherein the output parameters and the computational quantities related to the trained model include at least one selected from the group consisting of vorticity, Laplacian of pressure, helicity of pressure, and vortex axis.

3. the output parameters include an image that is generated based on the physical variables and indicates a state of the range; the explanatory variables of the trained model include an image showing a state of the range generated based on physical variables including at least one of a physical quantity related to the range including at least a suction port of a pump and a calculation quantity calculated from the physical quantity, The evaluation method according to claim 1, wherein in the evaluation step, the image generated in the calculation step is provided to the trained model to obtain an evaluation of the state of the vortex corresponding to the image.

4. The evaluation method according to claim 1 , wherein the output step outputs a graph representing the time-series evaluation results.

5. The evaluation method according to claim 1 , wherein the output step outputs a graph showing a moving average of the time-series evaluation results.

6. In the output step, the graph is a plane having an axis representing time and an axis representing the time-series evaluation results or a moving average of the time-series evaluation results, on which a curve representing the transition of the evaluation results or the moving average is expressed; 6. The evaluation method according to claim 4, wherein the evaluation results of the time series of the state of the vortex are indexed based on the relationship between an arbitrary straight line parallel to an axis representing time and the curve.

7. The evaluation method according to claim 6, wherein in the output step, the time series evaluation results of the vortex state are indexed based on the area of ​​the region enclosed by an arbitrary straight line parallel to the axis representing time and the curve.

8. The method further includes a correction step of correcting the evaluation of the state of the vortex based on the specific output parameter using an evaluation of the state of the vortex based on an output parameter of a time series different from the specific output parameter, 2. The evaluation method according to claim 1, wherein the output step outputs a time-series evaluation result based on the evaluation of the vortex state corrected in the correction step.

9. 2. The evaluation method according to claim 1, wherein the state of the vortex is classified into a plurality of levels according to the degree of vortex formation, and a predetermined evaluation score is assigned to each of the classified levels.

10. When executed on a computer, a calculation function that calculates, based on input parameters relating to a range including at least the suction port of the pump, output parameters that can be associated with physical variables including at least one of a physical quantity relating to the range and a calculation quantity calculated from the physical quantity, and a time variable that indicates a change in the physical variable over time; an evaluation function that, when an explanatory variable that can be associated with a physical variable including at least one of a physical quantity relating to a range including at least the pump inlet and a calculation quantity calculated from the physical quantity is input, provides an input based on the output parameter calculated by the calculation function to a trained model that obtains an evaluation of the state of the vortex in the range, and obtains an evaluation of the change over time in the state of the vortex corresponding to the output parameter; An evaluation program that realizes an output function that outputs time-series evaluation results based on an evaluation of the change in the state of the vortex over time in the evaluation function and the time variable included in the output parameters.

11. a calculation function that calculates, based on input parameters relating to a range including at least the suction port of the pump, output parameters that can be associated with physical variables including at least one of a physical quantity relating to the range and a calculation quantity calculated from the physical quantity, and a time variable that indicates a change in the physical variable over time; an evaluation function that, when an explanatory variable that can be associated with a physical variable including at least one of a physical quantity relating to a range including at least the pump inlet and a calculation quantity calculated from the physical quantity is input, provides an input based on the output parameter calculated by the calculation function to a trained model that obtains an evaluation of the state of the vortex in the range, and obtains an evaluation of the change over time in the state of the vortex corresponding to the output parameter; An evaluation device comprising a computer capable of implementing an evaluation function relating to the change in the state of the vortex over time in the evaluation function and an output function that outputs time-series evaluation results based on the time variable included in the output parameters.

Images