Building member estimation device and building member estimation method

Abstract

To provide a building member estimation device and a building member estimation method for improving, by means of numerical analysis, the estimation accuracy of a physical property value of a building member used for a building structure.SOLUTION: An estimation device 100 includes: an analysis unit 3 that, by using a physical property value determined for a building member of a building structure constructed in a predetermined construction state, performs numerical analysis on a model for reproducing the building structure; an evaluation unit 4 that evaluates an analysis result of the numerical analysis, based on a measurement result obtained by measuring the building structure; and a change unit 5 that changes the physical property value, based on an evaluation result when the analysis result is evaluated. The physical property value may be changed by Bayesian optimization or a genetic algorithm.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a construction component estimation device and a construction component estimation method. [Background technology]

[0002] In recent years, technological developments related to optimizing architectural structures and building materials through numerical analysis have been progressing, and related inventions have been made public. For example, Patent Document 1 discloses an estimation device that estimates physical properties of the earth in underground exploration without conducting large-scale tests such as boring inspections. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2022-157558 Summary of the Invention [Problem to be solved by the invention]

[0004] When performing numerical analysis, in order to make the analysis results match the actual phenomenon (measurement results), it is necessary to adjust the parameters of the numerical analysis model, that is, to match the model. Parameters include, for example, the geometric conditions of the model (e.g., size, length, modeling range), boundary conditions (e.g., fixed support, simple support, impedance boundary), physical properties (e.g., elastic modulus, Young's modulus, loss factor, Poisson's ratio), and parameters of the numerical analysis model itself (e.g., mesh size, time discretization width). Analysts often set parameters based on experiments and empirical rules, and build models that are in line with reality. Also, if the analysis results of the constructed model do not match the actual phenomenon, analysts may adjust the parameters appropriately to make them match.

[0005] Here, since the physical property values ​​as parameters cannot be obtained as measurement results, there are cases where they have to be estimated by experimenting in the construction state of the experimental system. For example, the resonance method is known as a method for determining the elastic modulus, but when determining the elastic modulus of a sealant used to waterproof window glass, a dedicated experimental system must be set up in consideration of the strong viscosity of the sealant. Specifically, it is necessary to set up an experimental system to observe the resonance frequency generated when vibration is applied to a hard concrete placed on the sealant. It is possible to estimate the elastic modulus from the observed resonance frequency. However, since the construction state of the experimental system is different from the actual construction state of the window frame in which the sealant is used, the estimated elastic modulus may differ from the (true) elastic modulus exhibited in the actual construction state. More specifically, the sealant in the experimental system is crushed by the weight of the concrete, so the strain-stress relationship is in a nonlinear region, the crushed sealant has room to stretch in a plane, so the vibration energy escapes horizontally, and the setting of the Poisson's ratio needs to be changed, etc., so there is a limit to the accuracy of estimating the elastic modulus. Even if a numerical analysis is performed using a model that reproduces the actual construction conditions based on such estimated results, there is a possibility that the analysis results will not match the actual phenomenon (measurement results).

[0006] Furthermore, when the parameters are multivariable, it is usually difficult to fit a model through human trial and error, and even if fitting is possible, there are inconveniences such as the need for a great deal of time and the tendency for the results of fitting the model to vary depending on the analyst. Also, the sealant serves to connect the glass to the window frame, but if you want to know how the physical properties of the sealant have deteriorated over time, you need to remove the sealant and conduct an experiment, which is time-consuming. From this perspective, an objective of the present invention is to provide a building component estimation device and a building component estimation method that improve the estimation accuracy of physical property values ​​of building components used in architectural structures through numerical analysis. [Means for solving the problem]

[0007] The present invention, which solves the above-mentioned problems, provides an estimation device for a building component, comprising an analysis unit that performs numerical analysis on a model that reproduces an architectural structure using physical property values ​​determined for architectural components of the architectural structure constructed in a specified construction state, an evaluation unit that evaluates the analysis results of the numerical analysis based on measurement results obtained by measuring the architectural structure, and a modification unit that modifies the physical property values ​​based on the evaluation results when the analysis results are evaluated. The present invention also provides a method for estimating building components, comprising the steps of: performing a numerical analysis on a model that reproduces an architectural structure using physical property values ​​determined for architectural components of the architectural structure constructed in a specified construction state; evaluating an analysis result of the numerical analysis based on measurement results obtained by measuring the architectural structure; and changing the physical property values ​​based on an evaluation result of the analysis result. According to this configuration, it is possible to calculate the physical properties of the building components in the actual construction state, not in the experimental construction state. Furthermore, the calculated physical properties can be repeatedly changed so that the analysis results match the measurement results. The finally obtained physical properties are highly accurate values.

[0008] It is also preferable that the building component is a viscoelastic material, the physical property value is at least one of Young's modulus, loss coefficient, elastic modulus, and Poisson's ratio, the measurement result is an actual compliance obtained by applying vibration to a hard body bonded to the viscoelastic material, and the analysis result is an analyzed compliance obtained by applying vibration as an acting factor to a hard body bonded to the viscoelastic material in the model. This allows easy implementation of tools for optimizing the acoustics of building components. It is also preferable that the evaluation unit evaluates the analysis result using an error function expressed using a difference between the actual compliance and the analyzed compliance. This makes it easy to evaluate the validity of the analysis results. It is also preferable to express the error function using coherence obtained as the measurement result. This makes it possible to adjust the error function so that it is not strongly influenced by the low-reliability measured compliance, which contributes to improving the estimation accuracy of the physical property values. It is also preferred that the discrete frequencies at which the measured compliance is obtained are interpolated from linear spacing to logarithmic spacing. This makes it possible to make the evaluation of the measured compliance and the analyzed compliance in the high frequency region equivalent to the evaluation of the measured compliance and the analyzed compliance in the low frequency region, thereby contributing to improving the estimation accuracy of the physical property values. It is also preferable that the physical property values ​​are changed by Bayesian optimization or a genetic algorithm. This makes it possible to efficiently optimize the physical properties of building components. Effect of the Invention

[0009] According to the present invention, it is possible to improve the accuracy of estimating the physical property values ​​of materials used in building components by using numerical analysis. [Brief description of the drawings]

[0010] [Figure 1] FIG. 2 is a functional configuration diagram of the estimation device according to the present embodiment. [Diagram 2] 3 is a flowchart showing an estimation method according to the present embodiment. [Diagram 3] FIG. 4 is a cross-sectional view of a window portion of this example. [Figure 4] FIG. 2 is an explanatory diagram of excitation points and receiving points. [Diagram 5] 13 is an example of a graph of the measurement results. [Figure 6] 13 is an example of a graph showing analysis results and measurement results. [Figure 7] FIG. 1 is an explanatory diagram regarding coherence. [Figure 8] 1 is an example of a graph in which measurement results are plotted at linear intervals. [Figure 9] 1 is an example of a graph in which measurement results are plotted at logarithmic intervals. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, the embodiment of the present invention will be described in detail with reference to the drawings. Each drawing is merely a schematic illustration to allow a sufficient understanding of the present invention. Therefore, the present invention is not limited to the illustrated examples. In each drawing, common or similar components are denoted by the same reference numerals, and duplicated explanations thereof will be omitted.

[0012] [composition] FIG. 1 is a functional configuration diagram of the estimation device of this embodiment. The estimation device 100 is a computer that estimates the physical property values ​​of architectural members that constitute an architectural structure constructed in a specified construction state. The estimation device 100 includes hardware such as an input unit, an output unit, a control unit, and a storage unit. For example, when the control unit is configured with a CPU (Central Processing Unit), information processing by a computer including the control unit is realized by program execution processing by the CPU. In addition, the storage unit included in the computer stores various programs for realizing the functions of the computer by instructions from the CPU. This realizes collaboration between software and hardware. The programs can be provided by recording them on a recording medium or via a network. The estimation device 100 may be communicably connected to a console (not shown) used by a user. The console can display the processing results of the estimation device 100 and input requests to the estimation device 100.

[0013] The estimation device 100 includes a construction unit 1, a setting unit 2, an analysis unit 3, an evaluation unit 4, and a change unit 5. The construction unit 1 constructs a model that reproduces an architectural structure constructed in a specified construction state. The model is a model for numerical analysis, and it is possible to set geometric conditions (e.g., size, length, modeling range), boundary conditions (e.g., fixed support, simple support, impedance boundary), physical properties (e.g., elastic modulus, Young's modulus, loss coefficient, Poisson's ratio), and parameters of the numerical analysis model itself (e.g., mesh size, time discretization width). The setting unit 2 sets rules for executing the calculation of the numerical analysis. For example, the rules include, but are not limited to, constraint conditions for physical properties (e.g., possible ranges of physical properties), termination conditions for terminating evaluation of the analysis results of the numerical analysis, and a method for changing physical properties. The analysis unit 3 performs a numerical analysis on the constructed model using the physical property values ​​determined for the building components of the architectural structure. The numerical analysis is, for example, but not limited to, FEM (Finite Element Method). The evaluation unit 4 evaluates the analysis results of the numerical analysis based on the measurement results obtained by measuring the architectural structure. The change unit 5 changes the physical property value based on the evaluation result obtained when the analysis result is evaluated. For example, the method of changing the physical property value is the method of changing the physical property value set by the setting unit 2.

[0014] [process] The estimation method of this embodiment will be described. FIG. 2 is a flowchart showing the estimation method of this embodiment. Note that the setting unit 2 may set various rules in advance or at an appropriate timing during the processing of FIG. 2. First, the construction unit 1 constructs a model that reproduces an architectural structure constructed in a predetermined construction state (step S1). The construction unit 1 has a function of automatically generating a mesh, for example. Next, the estimation device 100 determines the initial value of the physical property value of the estimation target (step S2). Next, the analysis unit 3 performs a numerical analysis on the constructed model using the determined physical property value (initial value or physical property value changed by the change unit 5) (step S3). Next, the evaluation unit 4 compares the measurement result obtained by measuring the architectural structure with the analysis result of the numerical analysis (step S4). Next, the evaluation unit 4 determines whether the comparison result satisfies the end condition (step S5). If the end condition is satisfied (Yes in step S5), the processing of FIG. 2 is terminated. On the other hand, if the end condition is not satisfied (No in step S5), the change unit 5 changes the physical property value (step S6). After that, the process of step S2 is performed on the changed physical property value.

[0015] [Specific example] (Acquisition of measurement results) Each step of the estimation method of this embodiment will be described in detail. First, the measurement of an actual architectural structure will be described. FIG. 3 is a cross-sectional view of a window of this specific example. The window 200 of FIG. 3 is an architectural structure including architectural components such as a base member 11, a plate-shaped glass 12, window frames 13, 13, and sealing materials 14, 14. The window 200 has a construction state in which the window frames 13, 13 are arranged on both sides of the peripheral portion of the glass 12 standing on the base member 11, and the glass 12 and the window frames 13, 13 are joined by the sealing materials 14, 14. The unit of the numerical values ​​indicating the dimensions shown in FIG. 3 is mm. This specific example will be described for the case where the Young's modulus and loss factor of the sealing materials 14, 14, which are viscoelastic bodies, are estimated as physical property values.

[0016] Vibration measurement (a type of resonance method) is performed on the window portion 200 in FIG. 3 using a hammer. Note that vibration measurement may be performed in a state close to the actual state shown in FIG. 3. In this vibration measurement, instead of vibrating the sealing materials 14, 14, which are viscoelastic bodies, using a hammer, the glass 12, which is a hard body, is vibrated with a hammer, and the vibration displacement caused by the vibration is recorded by a vibration sensor. Compliance is calculated as a transfer function based on the recorded vibration displacement. Compliance is vibration displacement per unit excitation force. Velocity (vibration speed per unit excitation force) or accelerance (vibration acceleration per unit excitation force) may be calculated as a transfer function instead of compliance. Furthermore, compliance is calculated for multiple pairs, with each pair being a combination of an excitation point and a receiving point on a predetermined plane of the glass 12.

[0017] Figure 4 is an explanatory diagram of excitation points and receiving points. For example, for a rectangular glass 12 measuring 2,000 mm x 2,900 mm, two receiving points (●: receiving point) and six excitation points (■: driving point) can be prepared, and an experimental system can be introduced to calculate a total of 12 sets of compliance. Note that it is preferable to prepare excitation points and receiving points for a certain mode so as to avoid a state in which only nodes are excited or received. Also, according to Maxwell's reciprocity theorem, if it is preferable for measurement to prepare a relatively large number of receiving points, the excitation points and receiving points in Figure 4 may be interchanged. Fig. 5 is an example of a graph of the measurement results. The vertical axis of the graph in Fig. 5 is compliance (solid line) and coherence (dashed line), and the horizontal axis is frequency. For convenience of illustration, Fig. 5 shows the compliance and coherence calculated for a total of six pairs of the six excitation points and point A as the vibration receiving point in Fig. 4 as the measurement results. The estimation device 100 uses the measurement results shown in Fig. 5. The calculated compliance can be calculated based on the vibration displacement that is generated from the excitation point of the glass 12 and propagated through the sealing materials 14, 14, extracted from the vibration displacement recorded at the receiving point.

[0018] (Model Construction) The construction unit 1 constructs a model that reproduces the window glass 200 in Fig. 3. At this time, the construction unit 1 sets, for the constructed model, geometric conditions (e.g., size, length, modeling range), boundary conditions (e.g., fixed support, simple support, impedance boundary), physical property values ​​(e.g., elastic coefficient, Young's modulus, loss coefficient, Poisson's ratio), and parameters of the numerical analysis model itself (e.g., mesh size, time discretization width).

[0019] (Determination of initial values ​​of physical properties) Regardless of the optimization method used to estimate the physical property values, it is necessary to set the range (constraint conditions) in which the physical property values ​​can be taken. For example, since the Young's modulus of materials used as glass sealants is often about 10e4 to 10e8 (Mpa), it is recommended to set the lower limit of Young's modulus to a value slightly smaller than 10e4 (Mpa) and the upper limit to a value slightly larger than 10e8 (Mpa). For the loss factor, it is recommended to set the lower limit to 0 and the upper limit to about 1. There are various methods for determining the initial value of the physical property values. For example, there is a method in which a computer simply randomly selects one point from the constraint conditions. In addition, in optimization methods that require the construction of a model such as Bayesian optimization, several sampling points can be selected using the Latin Hypercube method, or any point may be selected. The initial value of the physical property value determined in this way is x0. Note that x0 is a vector notation. For example, when estimating Young's modulus and loss factor, x0 is a vector consisting of two components, the initial value of Young's modulus and the initial value of the loss factor. The setting unit 2 can set the range of possible physical property values ​​and the method of determining the initial values ​​of the physical property values.

[0020] (Numerical Analysis) The analysis unit 3 performs a numerical analysis on the model constructed by the construction unit 1 using the given physical property value x0. Here, the measurement conditions when the compliance was measured for the window portion 200 shown in Fig. 3, specifically, the measurement conditions equivalent to the measurement conditions using the pair of excitation point-receiving point shown in Fig. 4, are introduced into the model. Note that the positions of the excitation point and the receiving point in the measurement conditions for the numerical analysis introduced into the model may be reversed to those in the measurement conditions in Fig. 4. FIG. 6 is an example of a graph of analysis results and measurement results. The vertical axis of the graph in FIG. 6 is compliance, and the horizontal axis is frequency. The solid line in the graph is the analysis result ("Analysis" in FIG. 6), and the dashed line is the measurement result ("Measured value" in FIG. 6). The pair of excitation point-receiving point referred to when obtaining the analysis result in FIG. 6 is the same as the pair of excitation point-receiving point referred to when obtaining the measurement result in FIG. 6. In addition, when obtaining the graph in FIG. 6, the Young's modulus of the physical property value x0, which is a two-component vector, was set to 1.0×10e8 (Mpa), and the loss factor was set to 0.5. According to FIG. 6, the frequency and sharpness of the peaks in the graph are different between the analysis result and the measurement result, and the analysis cannot correctly simulate the actual measurement.

[0021] (Comparison of measurement results and analysis results) In order to enable the evaluation by the evaluation unit 4 to be performed automatically without human intervention, a quantitative index that indicates the deviation between the analysis result and the measurement result is introduced. For example, an error function can be used as the index. The error function is an example of the evaluation result of the evaluation by the evaluation unit 4. For example, the error function F can be expressed by equation (1).

[0022]

number

[0023] Here, G m,j (f i ) is the discrete frequency f i The actual compliance (measurement result) is shown. F,j (f i ) is the discrete frequency f i The analytical compliance (analysis result) at C is shown. Note that both the measured compliance and the analytical compliance are values ​​that include not only the absolute value but also the phase, that is, they are complex numbers. m,j (f i ) indicates the coherence (measurement result). K is a function that takes the coherence as an argument.

[0024] The first half of equation (1) is a simple error ratio. By dividing by the measured compliance (or its squared value) instead of the difference between the measured compliance and the analytical compliance (or its squared value), it is possible to prevent only frequencies with large compliance or measurement points (pairs of excitation point and receiving point) with large compliance from being evaluated as an error function. As shown in Figure 6, for a given discrete frequency f i For this, the errors e1% to e6% expressed in the first half of equation (1) can be calculated. The error e6%, which shows a relatively large value, is divided by the corresponding measured compliance, so its influence on the evaluation function is not much different from the other errors e1% to e5%.

[0025] <Handling coherence> On the other hand, the coherence C in the latter part of equation (1) m,j (f i ) is a value indicating the degree of interference of multiple waves, but has the meaning of a correlation function between the excitation force at the excitation point and the vibration displacement at the receiving point, and takes a value in the range of 0 to 1. The closer it is to 1, the stronger the correlation between the excitation force and the vibration displacement, and the more the vibration displacement is caused by the excitation force. In other words, the vibration displacement generated at the excitation point is recorded without fail by the vibration sensor at the receiving point without being converted into thermal energy or the like and dissipated to the surroundings. Experimental values ​​such as actually measured compliance often have a small S / N ratio and contain large errors at points that do not vibrate much, such as near nodes. For this reason, if such unreliable experimental values ​​are used to compare the analysis results, the analysis results will be matched to the experimental values ​​that contain a lot of errors, and as a result, there is an increased possibility of errors in the estimated values ​​(physical properties).

[0026] In view of such circumstances, in this specific example, a measure is taken by making the coherence work as shown in formula (1). In other words, the experimental value of reliability can be adjusted so that even if the difference from the analysis result becomes large, it is not reflected much in the error function F. FIG. 7 is an explanatory diagram regarding coherence. FIG. 7 corresponds to the graph of FIG. 6 to which a graph of coherence as a measurement result is added. According to the graph of FIG. 7, in the frequency band of 1.5×10e1 to 2.3×10e1 (Hz), the coherence is almost 1, and the reliability of the measured compliance is high. Therefore, the error in this frequency band (the first half of formula (1)) is reflected in the error function F. On the other hand, in the frequency band of frequencies of 1.5×10e1 or less, the coherence is below 1, and the reliability of the measured compliance is low. Therefore, the error in this frequency band (the first half of formula (1)) is not reflected much in the error function F.

[0027] In addition, the function K in equation (1) is the coherence C m,j (f i ) to be reflected in the error function F. For example, m,j (f i ))={C m,j (f i )} 2 In addition, if you want to incorporate a frequency band with low coherence to some extent into the optimization results of this specific example, you can use a simple linear function K(C m,j (f i ))=C m,j (f i ) can also be used.

[0028] <Logarithmic and linear spacing> FIG. 8 is an example of a graph in which the measurement results are plotted at linear intervals. FIG. 9 is an example of a graph in which the measurement results are plotted at logarithmic intervals. In FIGS. 8 and 9, like FIG. 6, the vertical axis of the graph is compliance and the horizontal axis is frequency (Hz). The solid lines in the graphs are the analysis results ("Analysis" in FIGS. 8 and 9), and the dashed lines are the measurement results ("Measured Values" in FIGS. 8 and 9). In obtaining the measurement results, the discrete frequency f i That is, it is preferable to set the sampling frequency at logarithmic intervals when obtaining the actual compliance. Usually, the actual compliance is obtained by analyzing acceleration data detected by a vibration sensor, and more specifically, it is obtained by FFT (Fast Fourier Transformation) of the excitation force and vibration displacement on the time axis. Therefore, the actual measurement result is linearly spaced. For example, the actual compliance value for each frequency is output in increments of 0.625 Hz, and the measurement result plotted at every 0.625 Hz is obtained as shown in FIG. 8. However, in this case, as shown in FIG. 8, the plots become dense in the high frequency range (14 Hz or more), and there is a possibility that the measurement result in the high frequency range is overestimated. In view of this situation, it is preferable to interpolate the measurement result from the linearly spaced result to the logarithmic spaced result, and compare the logarithmically spaced measurement result with the analysis result. In other words, it is preferable to interpolate the discrete frequency at which the actual measurement result is obtained from the linear spaced to the logarithmic spaced. FIG. 9 shows the measurement result interpolated to the logarithmic spaced, and the density of the plots in the high frequency range is eliminated. In addition, in Figs. 8 and 9, the analysis results are both plotted in logarithmic intervals.

[0029] (Ending conditions) For example, the termination condition set by the setting unit 2 may be that the error function F is smaller than a predetermined threshold. Alternatively, the termination condition may be that the repeated calculations of the numerical analysis (No → S6 → S2 in steps S2 → S3 → S4 → S5 in FIG. 2) exceed a predetermined number of times because the error function F does not become smaller than the predetermined threshold. The evaluation unit 4 may output the physical property value that minimizes the error function F at the time when the termination condition is satisfied as the estimated physical property value. Depending on the optimization method, the termination condition may be that the amount of change (amount of improvement) between the error function calculated at the end of the repeated calculations and the error function calculated in the previous iteration of the repeated calculations is smaller than a predetermined threshold.

[0030] (Change (update) physical properties) In order to enable the change unit 5 to automatically change the physical property values ​​without manual intervention, an optimization method is introduced. The optimization method may be, but is not limited to, Bayesian optimization or a genetic algorithm. In this specific example, Bayesian optimization is used. By using the optimization method, the aforementioned numerical analysis is repeated to change the physical property values. The physical property values ​​used in the n-th repeated calculation are denoted by x n (n=0,1,2,…). Not limited to the physical property value x0 already explained, n The previous explanation applies to this.

[0031] When using Bayesian optimization, the procedure for changing the nth physical property is as follows. First, the n physical property values ​​x0 to x1 determined by the nth change are n-1 A Gaussian process regression model is constructed from the sample. The Gaussian process regression model shows the relationship between the physical property value x, which is expressed as a vector consisting of two components, Young's modulus and loss factor, and the error function F. Next, by using the Gaussian process regression model, the predicted value μ(x) of F for x and its variance σ 2 (x). Next, the predicted value μ(x) and variance σ 2 The property value for which the acquisition function with (x) as an argument is maximized or minimized is the next property value x n+1 Let us assume that.

[0032] (Estimated value confirmed) For example, the evaluation unit 4 can select one physical property value from the n + 1 physical property values x determined by the above procedure n (n = 0, 1, 2,...) and output it as an estimated value. That is, it is possible to select the physical property value with the highest probability of having a better correspondence than the physical property value that has had the best correspondence with the experiments so far (that is, the error function is minimized).

[0033] Note that in Bayesian optimization and genetic algorithms, a method of normalizing the physical property value vector x to 0 to 1 for calculation may be used. Each component x of one vector i can be normalized by a separate method, but it is preferable to linearly correspond the normalized physical property value component x i (0 < x i < 1) and the actual physical property value component X i (a < X i < b) as shown in equation (2). X i = (b - a)x i + a (Equation (2)) Here, a and b are the minimum and maximum values of the actual physical property value X, respectively. On the other hand, when the range of the constraint conditions is large, such as the Young's modulus, in the method of equation (2), the actually selected physical property values tend to be biased towards the larger side. In this case, it is preferable to correspond logarithmically as in equation (3). X i = a(a / b)^x i (Equation (3)) The inventor performed numerical analysis using the finally selected physical property values, and confirmed that, in contrast to FIG. 6, the analysis could correctly simulate the actual measurement.

[0034] [Effect] According to this embodiment, it is possible to calculate the physical property values ​​of a building component in an actual construction state, not in an experimental construction state. Furthermore, the calculated physical property values ​​can be repeatedly modified so that the analysis results match the measurement results. The finally obtained physical property values ​​are values ​​with high estimation accuracy. Therefore, it is possible to improve the estimation accuracy of the physical property values ​​of materials used in building components by using numerical analysis. In addition, the estimation of physical properties is automatically processed by algorithms, without the need for analyst intervention, which reduces the labor required. In addition, the analysis can be performed without depending on the analyst's working hours, which shortens the estimation time.

[0035] Additionally, tools for optimizing the acoustics of building components can be easily implemented. In addition, the validity of the analysis results can be easily evaluated by using the error function. In addition, by using the coherence, it is possible to adjust the error function so that the unreliable measured compliance is not strongly reflected in the error function, which contributes to improving the estimation accuracy of the physical property value. In addition, by performing logarithmic interpolation, the evaluation of the measured compliance and the analyzed compliance in the high frequency region can be made equivalent to the evaluation of the measured compliance and the analyzed compliance in the low frequency region, which contributes to improving the estimation accuracy of the physical property values. Furthermore, the physical properties of building components can be efficiently optimized using Bayesian optimization or a genetic algorithm.

[0036] [Variations] (a): In this embodiment, a case has been described in which a numerical analysis is performed using a model constructed by the construction unit 1. However, a model constructed in advance by a computer other than the estimation device 100 may be read and numerical analysis may be performed. Furthermore, even if the computer is the same as the estimation device 100, a model constructed in advance may be read and numerical analysis may be performed by executing an application for model construction other than an application for numerical analysis. (b) The mesh for the numerical analysis may be generated, for example, when constructing the model, or when setting the setting unit 2. (c) It is also possible to realize a technology that appropriately combines the various technologies described in this embodiment. (d) The software described in the present embodiment can be realized as hardware, and vice versa. (e) In addition, the hardware, software, flow charts, etc. may be modified as appropriate without departing from the spirit and scope of the present invention. [Explanation of symbols]

[0037] 100 Estimator 1. Construction Department 2. Settings 3 Analysis section 4 Evaluation section 5. Changes

Claims

1. An analysis unit that performs a numerical analysis on a model that reproduces the architectural structure by using physical property values ​​determined for architectural components of the architectural structure constructed in a predetermined construction state; an evaluation unit that evaluates an analysis result of the numerical analysis based on a measurement result obtained by measuring the architectural structure; A modification unit that modifies the physical property value based on an evaluation result obtained when the analysis result is evaluated.

2. The building component is a viscoelastic body, The physical property value is at least one of Young's modulus, loss coefficient, elastic modulus, and Poisson's ratio, the measurement result is an actual compliance obtained by applying vibration to a hard body coupled to the viscoelastic body, 2. The device for estimating a construction component according to claim 1, wherein the analysis result is an analytical compliance obtained by applying vibration as an acting factor to a hard body coupled to a viscoelastic body in the model.

3. The estimation device for a building component according to claim 2 , wherein the evaluation unit evaluates the analysis result using an error function expressed using a difference between the actual compliance and the analyzed compliance.

4. The device for estimating a construction component according to claim 3 , wherein the error function is expressed using coherence obtained as the measurement result.

5. The estimator of claim 2 , wherein the discrete frequencies at which the measured compliance is obtained are interpolated from linear to logarithmic spacing.

6. The estimation device for a building component according to claim 1 , wherein the change in the physical property value is performed by Bayesian optimization or a genetic algorithm.

7. A step of performing a numerical analysis on a model that reproduces the architectural structure using physical property values ​​determined for the architectural components of the architectural structure constructed in a predetermined construction state; evaluating an analysis result of the numerical analysis based on a measurement result obtained by measuring the architectural structure; and changing the physical property value based on an evaluation result obtained by evaluating the analysis result.

Images