JP2017058273A - Fatigue tester - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a fatigue tester capable of precisely correcting a driving waveform even for a nonlinear response.SOLUTION: A correction part 60 comprises: a wavelet transformation part 61 which performs wavelet transformation on an input waveform; a correction calculation part 62 which calculates a wavelet transformation result of a next-period driving waveform from a result of wavelet transformation on the driving waveform, a result of wavelet transformation on a response waveform, and a result of wavelet transformation on a target waveform; a reverse wavelet transformation part 73 which performs reverse wavelet transformation on the wavelet transformation result of the next-period driving waveform found by the calculation so as to obtain a next-period driving waveform of time-series data; and a determination part 64 which determines whether the next-period driving waveform obtained by the reverse wavelet transformation part 73 reaches a correction end standard so as to determine whether to further make a correction repeatedly.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a fatigue testing machine that performs a fatigue / endurance test by applying vibration to a specimen.

  In the fatigue / endurance test, vibrations received in an environment where parts and materials are actually used are mechanically reproduced and applied to the specimen. In order to execute a fatigue / endurance test with a waveform close to a target waveform corresponding to an actual vibration waveform, a fatigue testing machine in which a drive waveform for operating the apparatus is corrected is known.

  In this type of fatigue testing machine, the transfer function of the system is identified from the ratio of the random waveform and its response waveform, and the drive waveform is generated from the target waveform using the inverse transfer function that is the inverse of the transfer function. . Then, the target waveform and response waveform are converted from the time domain to the frequency domain by Fourier transform, and the drive waveform is corrected by updating the transfer function calculated from the ratio or difference between the target waveform and the response waveform in the frequency domain. (See Patent Document 1 and Patent Document 2).

JP 2014-13176 A JP 2010-266398 A

  The correction of the drive waveform using the Fourier transform and its inverse transform is also referred to as transfer function correction. When performing the transfer function correction, time information is lost if the entire waveform is Fourier transformed. For this reason, if the response ratio to the input is different from the other at a certain location (time) of the waveform, the entire waveform is averaged and corrected in the frequency range, so that it is difficult to correct the drive waveform at that specific location. In addition, if there are places where the response waveform does not change accordingly even if the input waveform is changed, such as an impact load due to mechanical displacement, the drive waveform will be overcorrected, and the correction accuracy of the entire waveform will be improved. It was falling.

  Furthermore, instead of Fourier transforming the entire waveform, assuming that the response in a short time is linear, performing a short-time Fourier transform that performs Fourier transform by separating the signal using a window function of a certain size In this case, the time information is discrete for each section. If there is a portion where the response is nonlinear with respect to the input, it is difficult to correct the low frequency component if the section (window) is made fine in order to separate such a portion from other portions. In addition, when a drive waveform is generated by connecting sections, a frequency that does not exist in the actual waveform may be mixed in at the connection between sections. Since the window is of a certain size, if there is a portion in the response waveform that does not change according to the input waveform, such as impact load due to mechanical deviation in the section, as in the case where the entire waveform described above is Fourier transformed, A drive waveform in which a non-responsive portion has been corrected is input to the drive system. For this reason, even if the correction is repeated, the response waveform may be far from the target waveform.

  The present invention has been made to solve the above problems, and an object of the present invention is to provide a fatigue testing machine capable of accurately correcting a drive waveform even with respect to a non-linear response.

  The invention according to claim 1 is a fatigue testing machine that applies a load to a specimen by driving a load actuator with a drive signal based on a target waveform, wherein the drive waveform for driving the load actuator is based on the target waveform. A drive waveform generator for generating, a receiver for receiving a response waveform from the specimen side when the load actuator is driven by the drive waveform, and a wavelet transform for the drive waveform, the target waveform, and the response waveform, respectively And a correction unit that corrects the drive waveform using the result obtained.

  The invention according to claim 2 is the fatigue testing machine according to claim 1, wherein the correction unit is a value obtained by multiplying a difference between a wavelet transformation result of the target waveform and a wavelet transformation result of the response waveform by a correction coefficient. Is used to correct the wavelet transform result of the drive waveform.

  According to a third aspect of the present invention, in the fatigue testing machine according to the second aspect, the correction calculation unit uses the frequency correction coefficient as a wavelet transform result of the drive waveform that is an input and the response waveform that is a response thereof. It calculates using the input response ratio showing the correlation of the input and response which are obtained from the wavelet transformation result, and the safety factor showing the correction strength.

  According to a fourth aspect of the present invention, in the fatigue testing machine according to the third aspect, the correction calculation unit obtains the input response ratio based on a frequency response transfer function that is a ratio of an input and a response for each frequency. To do.

  According to a fifth aspect of the present invention, in the fatigue testing machine according to the third aspect, the correction unit is configured to increase or decrease an input based on a change in value at each point of the wavelet transform result of the drive waveform that is an input, A safety factor calculation unit is provided that calculates the safety factor by utilizing a correlation with a response increase / decrease based on a change in value at each point of the wavelet transform result of the response waveform as a response.

  According to the first to fifth aspects of the invention, by using the wavelet transform for correcting the drive waveform, it is possible to use not only information in the frequency domain but also information for each time for correction. For this reason, it is possible to easily set the range of the area in the waveform that can be corrected and to change the correction intensity, prevent excessive correction and overshoot, and speed up the response waveform to reach the target waveform. Therefore, it is possible to correct the drive waveform with high accuracy even for a specimen having a non-linearity.

  According to the invention described in claim 5, by providing the safety factor calculation unit, it is possible to have a correction coefficient that varies depending on time for each frequency by automatic calculation according to the transfer characteristics of the system of the fatigue tester. Become. As a result, the drive waveform can be corrected more efficiently.

1 is a schematic diagram of a fatigue testing machine according to the present invention. 3 is a functional block diagram of a control device 50. FIG. 3 is a functional block diagram of a correction unit 60. FIG. It is a flowchart explaining the correction procedure of a drive waveform. It is a functional block diagram showing a modification of control device 50. FIG. 10 is a functional block diagram showing a modification of the correction unit 60. It is a graph which shows the wavelet transformation result of a target waveform. It is a graph which shows the wavelet transformation result of a response waveform. It is a graph which shows a correction result. It is a graph explaining the difference | error of the target waveform and response waveform before and behind correction | amendment of a drive waveform. It is a graph explaining the difference | error of the target waveform and response waveform before and behind correction | amendment of a drive waveform. It is a conceptual diagram of the approximation coefficient A and the detailed coefficient D of discrete wavelet transform. It is a graph explaining the correction | amendment of the drive waveform using discrete wavelet transform. It is a graph explaining the correction | amendment of the drive waveform using discrete wavelet transform. It is a graph explaining the correction | amendment of the drive waveform using discrete wavelet transform. It is a graph explaining the correction | amendment of the drive waveform using discrete wavelet transform.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a schematic diagram of a fatigue testing machine according to the present invention.

  This fatigue testing machine performs a fatigue / endurance test by applying a load vibration to a test piece TP as a specimen by an actuator 31. A cross head 25 is installed on a pair of support posts 23 erected on the table 24, and an actuator 31 is attached to the cross head 25. The test piece TP is gripped at both ends by an upper grip 21 disposed on the cross head 25 side and a lower grip 22 disposed on the table 24 side. The upper gripping tool 21 is connected to the piston rod of the actuator 31. Therefore, when the upper gripper 21 moves up and down in conjunction with the movement of the piston rod, a test force is repeatedly applied to the test piece TP.

  A drive signal for the actuator 31 is transmitted from the control device 50. The digital signal generated by the control device 50 is converted into an analog signal by the D / A converter 41, amplified by the amplifier 35, and input to the actuator 31. In this embodiment, an electromagnetic actuator is used as the actuator 31, but when a hydraulic actuator is used instead of the electromagnetic actuator, a drive signal is input to the servo valve.

  The displacement amount of the test piece TP is detected by a displacement meter 32 disposed in the cross head 25. The displacement signal is amplified by the amplifier 37, converted from an analog signal to a digital signal by the A / D converter 43, and then input to the control device 50.

  The test force applied to the test piece TP is detected by a load meter 33 disposed between the table 24 and the lower gripping tool 22. The load signal is amplified by the amplifier 36, converted from an analog signal to a digital signal by the A / D converter 42, and then input to the control device 50.

  The control device 50 is composed of a computer or sequencer including a CPU that executes logical operations, a ROM that stores a program necessary for controlling the device, a RAM that temporarily stores data during control, and the like. Control overall operation. A target waveform generator 46 that generates a target waveform corresponding to the actual vibration waveform applied to the test piece TP is connected to the control device 50.

  FIG. 2 is a functional block diagram of the control device 50.

  As a functional configuration, the control device 50 includes an initial drive waveform generation unit 52 that creates an initial drive waveform input to the actuator 31, a correction unit 60 that corrects the drive waveform during a test, a displacement signal, and a load signal. A detection signal selection unit 54 that receives and receives which signal is input to the correction unit 60 as a response waveform, and a drive signal output unit 55 that outputs a drive waveform supplied to the actuator 31 as a digital signal are provided.

  FIG. 3 is a functional block diagram of the correction unit 60.

  The correction unit 60 performs the next drive waveform from the wavelet transform unit 61 that performs wavelet transform on the input waveform, the wavelet transform result of the drive waveform, the wavelet transform result of the response waveform, and the wavelet transform result of the target waveform. A correction calculation unit 62 that calculates the wavelet transform result of the current waveform, a reverse wavelet transform unit 63 that performs a reverse wavelet transform on the wavelet transform result of the next drive waveform obtained by the calculation to obtain the next drive waveform of time series data, and an inverse wavelet And a determination unit 64 that determines whether or not to repeatedly perform correction by determining whether or not the next drive waveform obtained in the conversion unit 63 has reached the correction end reference.

  The correction of the drive waveform in the fatigue testing machine of the present invention having the above configuration will be further described. FIG. 4 is a flowchart for explaining a drive waveform correction procedure.

  The target waveform generated from the target waveform generator 46 is input to the initial drive waveform generation unit 52 and the correction unit 60. Then, the initial drive waveform generator 52 generates an initial drive waveform_0 (t) based on the target waveform (step S1).

  The actuator 31 is vibrated by an initial drive waveform — 0 (t) that is a signal of time series data, and the load signal detected by the load meter 33 and the displacement signal detected by the displacement meter 32 are input to the control device 50. . The detection signal selection unit 54 selects which signal of the load signal or the displacement signal is input to the correction unit 60, and the selected signal is input to the correction unit 60 as a response waveform_1 (t) (step S2). ). Then, i = 1st correction (step S3) is executed.

  The initial drive waveform_0 (t) in step S1 may use a target waveform when it can be assumed that the response and input are close to 1: 1. In this case, the target waveform input to the initial drive waveform generation unit 52 is input to the drive signal output unit 55 (see FIG. 2). When the target waveform is not used for the initial drive waveform_0 (t), it may be obtained using a frequency response transfer function as follows.

  FIG. 5 is a functional block diagram of a modified example of the control device 50.

  In this modification, the control device 50 includes a transfer function calculation unit 51 that identifies a transfer function of a system including the actuator 31 and the test piece TP. When an arbitrary waveform is given to the system as a dynamic load, a transfer function calculation unit 51 calculates a frequency response transfer function based on the input arbitrary waveform and its response. The arbitrary waveform here is a waveform having a frequency component necessary for obtaining a transfer function for each frequency, and, for example, a PSD random waveform is adopted as the arbitrary waveform. The PSD random waveform is a waveform (white noise) having a constant PSD (Power Spectrum Density) indicating the power spectrum value for each frequency, and is selected as a composite waveform including various frequency components or a model waveform. It is created based on the actual vibration waveform. The arbitrary waveform is supplied from a waveform generator connected to the control device 50 separately from the target waveform generator 46.

  When the target waveform is not used for the initial drive waveform — 0 (t), an arbitrary waveform is input to the transfer function calculation unit 51 and the drive signal output unit 55. The actuator 31 is driven by an arbitrary waveform input to the drive signal output unit 55, and the response waveform is input to the transfer function calculation unit 51 via the detection signal selection unit 54. In the transfer function calculation unit 51, the transfer function of the system including the actuator 31 and the test piece TP is identified from the arbitrary waveform and the response waveform. The initial drive waveform generation unit 52 generates a drive waveform — 0 (t), which is an initial drive waveform, from the inverse transfer function that is the reciprocal of the transfer function identified by the transfer function calculation unit 51 and the target waveform.

  The correction procedure in the correction unit 60 will be further described with reference to FIG. In the wavelet transform unit 61 of the correction unit 60, the wavelet transform is performed on the drive waveform_0 (t) and the response waveform_1 (t), which are input waveforms, by a known wavelet transform formula, respectively, and the drive waveform_0 (a, b ), Response waveform_1 (a, b) is acquired (step S4). Note that (t) indicates time-series data, and (a, b) are wavelet transform results. a is a scale parameter and corresponds to the frequency. b is a shift parameter and corresponds to time.

  The wavelet transform results can be shown as a graph in which the x-axis is time (seconds), the y-axis is frequency (Hz), and the magnitude of the frequency component is represented by a color difference (the upper part of FIGS. 7 and 8 described later). See graph). That is, the wavelet scale “a” can be converted into a frequency (Hz) by a known relational expression between the scale and the frequency, so that the y-axis can be expressed in a graph with the frequency (Hz).

  In the correction calculation unit 62, the wavelet transform coefficient of the next drive waveform is calculated using the following equation (1) (step S5).

  i indicates the i-th correction. The coefficient (a, b) is a correction coefficient, and may be a uniform value, or may have a different value for each scale a and shift b, that is, for each (a, b) point. . Furthermore, the coefficient (a, b) can be determined by the following equation (2).

  The input response ratios (a, b) may all be uniform values, or may be response ratios (so-called transfer functions) with respect to inputs for each frequency. In other words, the theoretical value may be input by the operator via the input means, or a value obtained by actual measurement may be input. As a method of obtaining a value by actual measurement, a static load is actually given to the system, a value is obtained from the result of measuring the relationship between input and response, and a dynamic load such as the random load described above is given to the system. And a method of obtaining a frequency response transfer function. In this case, the correction calculation unit 62 can obtain the input response ratio (a, b) by inputting the frequency response transfer function to the correction unit 60 from the transfer function calculation unit 51 shown in FIG. .

  The safety factor (a, b) represents the strength of the correction, that is, the correction strength, and may be a uniform value, and may be a value different for each frequency (scale a) and time (shift b). May be. That is, an operator may input a single value via the input means and apply it to all frequencies. A large value (for example, maximum value = 1) is used for a frequency that is unlikely to overshoot. A small value may be input to a frequency that is likely to be overshot.

  When the safety factor (a, b) is changed due to the possibility of overshoot, the target waveform can be reached quickly at the frequency at which a large value is input, and the drive waveform at the frequency at which a small value is input. Can be corrected excessively, and control can be prevented from becoming unstable. In addition, even at frequencies where control becomes unstable near the resonance frequency or at high frequencies where response is difficult as a test system, lowering the safety factor prevents control from becoming unstable due to excessive correction of the drive waveform. be able to.

  The drive waveform_1 (a, b) calculated by using the equation (1) in the correction calculation unit 62 is subjected to inverse wavelet transform by the inverse wavelet transform unit 63, and drive waveform_1 (t) that is time-series data of the next drive waveform. ) Is obtained (step S6). The drive waveform_1 (t) is input from the correction unit 60 to the drive signal output unit 55, and a drive signal based on the drive waveform after the first correction is given to the actuator 31. When the response waveform_2 (t) after the test piece TP is vibrated with the drive waveform after the first correction is input to the correction unit 60 via the detection signal selection unit 54 (step S7), the determination unit 64 It is determined whether or not the drive waveform after the first correction has reached the correction end reference, and it is determined whether the correction is repeatedly performed or the correction is ended (step S8).

  As correction end criteria, the area error rate of <1> target waveform (t) and response waveform_i + 1 (t) is below a certain value, <2> target waveform (a, b) and response waveform_i + 1 (a, b) In the determination unit 64, three criteria are set in advance: <3> a predetermined number of repetitive corrections is completed. In this embodiment, when any of these conditions is met, the repeated correction of the drive waveform is completed.

  When the drive waveform has not reached the correction end reference and the correction is continued repeatedly, 1 is added to i indicating the number of corrections (step S9), and steps S4 to S8 are repeated again. Note that the reason for adding <3> a predetermined number of repeated corrections to the correction end criterion as a condition is that a load is applied to the test piece TP even in the preliminary test stage for correcting the drive waveform. Therefore, from the viewpoint of the strictness of the relationship between the load and the fatigue of the test piece TP in this test, it is necessary to avoid a long time for correction.

  FIG. 6 is a functional block diagram illustrating a modified example of the correction unit 60.

  In this modified example, a safety factor calculation unit 65 is further provided as a functional configuration of the correction unit 60, and the safety factor of the equation (2) is executed in the middle of performing the correction in which steps S4 to S8 of the procedure shown in FIG. 4 are repeated. (A, b) is automatically calculated and applied. The calculation of the safety factor (a, b) is performed by repeatedly changing the value at each (a, b) point of the wavelet transformation result of the drive waveform that is the input, and each of the wavelet transformation results of the response waveform that is the response. This is performed using the relationship between the input increase / decrease and the response increase / decrease based on the change in the value at point (a, b). That is, a correlation coefficient that is a correlation of a ratio of response increase / decrease to input increase / decrease is calculated from the change of input value and the response value, and this correlation coefficient is applied to equation (2) as a safety factor (a, b). To do. In this case, the correction strength of the low correlation between the drive waveform before correction and its response waveform is lowered, thus preventing the drive waveform from diverging and preventing the control from becoming unstable. Can do.

  Furthermore, a modified example of the automatic calculation of the safety factor of Expression (2) will be described. Unlike the Fourier transform, the wavelet transform has time information. For this reason, as described above, the result of the wavelet transform has the values of the scale a and the shift b as fluctuation elements. On the other hand, in this modified example, the safety factor (a, _) is used as the safety factor of Expression (2). That is, the safety factor (a, _) sets the same safety factor for the same frequency (scale a) regardless of the shift b, that is, time. A frequency transfer function calculated in advance is used for the safety factor (a, _). The frequency transfer function may be acquired by inputting the value calculated in the transfer function calculation unit 51 shown in FIG. In this case, the correction strength of the frequency having a high input response ratio compared to the low frequency stable region is lowered, and overshoot due to excessive correction can be prevented.

  Next, the target waveform in which two signs having different frequencies are superimposed is the drive waveform_0 (t), and the correction result when the fatigue tester vibrates the test piece TP that responds nonlinearly is shown. explain. In the example described below, the detection signal selection unit 54 selects the load signal measured by the load meter 33 as the response waveform input to the correction unit 60.

  FIG. 7 is a graph showing the wavelet transformation result of the target waveform, and FIG. 8 is a graph showing the wavelet transformation result of the response waveform. FIG. 9 is a graph showing the correction result. 7, 8, and 9 are graphs of waveforms in which the horizontal axis represents time (seconds) and the vertical axis represents the test force (kN: kilonewton) applied to the test piece TP. 7 and 8 show the result of wavelet transform of the lower waveform, and the upper part of FIG. 9 shows the wavelet transform result of the next drive waveform obtained by the equation (1). 7, 8 and 9, the horizontal axis is the same time (seconds) as the waveform graph, and the vertical axis is the frequency (Hz). In the actual display of the wavelet transform result on a display device or the like, the frequency components are expressed by coloring red, yellow, green, light blue, and blue in descending order of frequency components, for example, FIG. 7, FIG. 8, and FIG. In the upper graph of 9, the magnitude of the frequency component is indicated by different hatching. Note that the wavelet transform examples in FIGS. 7, 8, and 9 use Morlet wavelets as mother wavelets.

  When a test force of 8 kN to 10 kN is applied to the test piece TP exhibiting a non-linear vibration response, as shown by the symbol e in the lower graph of FIG. 8, compared with the amplitude of the target waveform (see the lower graph of FIG. 7). The response waveform has a large superimposed amplitude. On the other hand, when a test force of 4 kN is applied, the response waveform superimposed amplitude is reversed to the amplitude of the target waveform (shown in FIG. (See the lower graph).

  The drive waveform in which the response waveform exhibits such behavior is corrected using Equation (1). The wavelet transformation result of the next drive waveform obtained by substituting the result of the continuous wavelet transformation shown in the upper part of FIG. 7 and FIG. 8 into the equation (1) is as shown in the upper part of FIG. The next drive waveform of the time series data obtained in this way is as shown in the lower graph of FIG. In the next drive waveform in the lower part of FIG. 9, the amplitude when a test force of 8 kN to 10 kN is loaded is smaller than the target waveform (see the lower graph in FIG. 7) as the initial drive waveform before correction, and 4 kN. The amplitude when the test force is applied is larger than before the correction. That is, in this correction, since the frequency and time at which the amplitude of the response waveform with respect to the drive waveform is increased and the frequency and time at which the amplitude of the response waveform with respect to the drive waveform is decreased are provided with different frequency correction coefficients, As a whole, the response waveform is closer to the target waveform.

  10 and 11 are graphs for explaining an error between the target waveform and the response waveform before and after correction of the drive waveform. FIG. 10 shows the error between the target waveform and the response waveform before correction, and FIG. 11 shows the error between the target waveform and the response waveform after repeated correction in the correction unit 60. 10 and FIG. 11, the horizontal axis represents time (seconds), and the vertical axis represents test force (kN). Further, the graph of FIG. 10A shows the broken target waveform and the solid response waveform superimposed, and the graph of FIG. 11A shows the response waveform after repeated correction. The graphs of FIG. 10B and FIG. 11B show an error waveform obtained by dividing the response waveform from the target waveform. The target waveform is a superposition of two sine waves having different frequencies, as in FIGS.

  Before the correction, as shown in FIG. 10, there is a large deviation of the response waveform with respect to the target waveform, and there is an error of about plus or minus 3 kN at maximum for the entire waveform. On the other hand, after the repeated correction, as shown in FIG. 11, the responsiveness is improved until the response waveform becomes almost the same as the target waveform, and the error is also reduced to 1/10 or less before the correction.

  In the correction of the drive waveform in the correction unit 60, the waveform can be decomposed while leaving the time information by wavelet transform, so that the possibility of correction on the time axis and the degree of freedom in changing the correction strength are improved, and the target waveform is improved. Therefore, the response waveform can be quickly brought close to the target waveform without excessive correction in a region having a good correlation between the response waveform and the response waveform.

  In the correction examples with reference to FIGS. 7 to 11 described above, the case where the continuous wavelet transform is performed on the waveform has been described. However, in this continuous wavelet transform, the wavelet function satisfies a condition called an admissibility condition. If not, reverse conversion cannot be performed. When the wavelet function does not satisfy the admissible condition, or when it is desired to reduce the amount of calculation, correction may be performed by discrete wavelet transform. Hereinafter, correction of the drive waveform when the wavelet transform unit 61 and the inverse wavelet transform unit 63 in the correction unit 60 perform discrete wavelet transform and its inverse transform will be described.

  FIG. 12 is a conceptual diagram of the approximation coefficient A and the detailed coefficient D of the discrete wavelet transform. The discrete wavelet transform is also a kind of subband coding that uses a plurality of filters to decompose an input signal into a plurality of frequency domain signals. FIG. 12 shows an example in which a set of low-pass filter 72 and high-pass filter 71 is configured in three stages.

  In the discrete wavelet transform, as shown in FIG. 12A, the signal x is decomposed by using a filter set of a low-pass filter 72 whose impulse response is g and a high-pass filter 71 whose impulse response is h. A large scale obtained through the low-pass filter 72 and the down sampler 73 that thins out the signal every other sample point is called an approximation coefficient, and a small scale obtained through the high-pass filter 71 and the down sampler 73 is a detailed coefficient (detail coefficient). Called. In the first stage decomposition, the approximate coefficient A1 and the detailed coefficient D1 are obtained. By performing the second stage decomposition on the approximate coefficient A1, the approximate coefficient A2 and the detailed coefficient D2 are obtained. When A3 is decomposed at the third stage, an approximate coefficient A3 and a detailed coefficient D3 are obtained. In this way, at each stage of the filter set, it is decomposed into an approximate coefficient A that is a low frequency component of the signal and a detailed coefficient D that is a high frequency component of the signal.

  The approximate coefficient A3 and the detailed coefficients D1 to D3 obtained as output signals by the decomposition can be synthesized into time series data equivalent to the original input signal by inverse transformation, that is, signal processing reverse to that at the time of decomposition. That is, as shown in FIG. 12B, the signal x is synthesized through an upsampler 74, a low-pass filter 72, a filter set of a high-pass filter 71 and an adder 75 that insert a value of 0 between each sample point. Can do.

  FIG. 13 to FIG. 16 are graphs for explaining the correction of the drive waveform using the discrete wavelet transform. FIGS. 13 to 16 perform discrete wavelet transform using a filter set having a six-stage configuration, and among the detailed coefficients D1 to D6 and the approximate coefficient A6 obtained thereby, the formulas ( It is a graph which performs the correction | amendment using 1), and also shows a result as time series data by reverse wavelet transformation. 13 shows the correction result of the detailed coefficient D4, FIG. 14 shows the correction result of the detailed coefficient D5, FIG. 15 shows the correction result of the detailed coefficient D6, and FIG. 16 shows the correction result of the approximate coefficient A6. 13 to 16, the response waveform is indicated by a broken line, and the error waveform indicating the difference between the target waveform and the response waveform is indicated by a solid line.

  In this embodiment, among the detailed coefficients D1 to D6 and the approximate coefficient A6 obtained by performing discrete wavelet transform with a filter set having a six-stage configuration, the detailed coefficients D1 to D3 that are high frequency components are the safety in the expression (2). Correction was not performed by setting the rate (a, b) = 0, but the detailed coefficients D4 to D6 were corrected repeatedly. This is to prevent the control waveform from becoming unstable due to excessive correction of the drive waveform at a high frequency at which the test system does not respond.

  When iterative correction is performed on the low frequency components (detail coefficients D4 to D6, approximate coefficient A6), it is understood that the error between the target waveform and the response waveform at the low frequency is corrected as shown in FIGS. The

  The drive waveform decomposed by the discrete wavelet transform is corrected by the equation (1), and then subjected to the inverse wavelet transform, as shown in FIG. The drive waveform is transmitted as an output waveform from the correction unit 60 to the drive signal output unit. The response waveform resulting from the excitation by the corrected drive waveform is the same as that in FIG. 11A, and the error between the target waveform and the response waveform is converged to be equal to the error waveform in FIG. Is done.

  In the discrete wavelet transform, as shown in FIG. 12A, the detailed coefficient D and the approximate coefficient A are obtained via the filter set and the downsampler 73, so that the time resolution is the original as the number of stages of the filter set increases. The signal is halved (the number of sample points is halved) and the frequency resolution is doubled (see response waveforms in FIGS. 13 to 16). As described above, in the discrete wavelet transform, wavelet coefficients are output by passing through a filter. Therefore, compared with the continuous wavelet transform, the calculation time is short and the calculation load of the apparatus can be reduced.

DESCRIPTION OF SYMBOLS 21 Upper gripper 22 Lower gripper 23 Support | pillar 24 Table 25 Cross head 31 Actuator 32 Displacement meter 33 Load meter 36 Amplifier 37 Amplifier 41 D / A converter 42 A / D converter 43 A / D converter 46 Target waveform generator DESCRIPTION OF SYMBOLS 50 Control apparatus 51 Transfer function calculating part 52 Detection signal selection part 60 Correction | amendment part 61 Wavelet transformation part 62 Correction calculation part 63 Inverse wavelet transformation part 64 Safety factor calculation part 71 High pass filter 72 Low pass filter 73 Down sampler 74 Up sampler 75 Adder

Claims (5)

A fatigue testing machine that drives a load actuator with a drive signal based on a target waveform and applies a load to a specimen,
A drive waveform generator for generating a drive waveform for driving the load actuator based on the target waveform;
A receiver for receiving a response waveform from the specimen side when the load actuator is driven by the drive waveform;
A correction unit that corrects the drive waveform using the result of wavelet transform of the drive waveform, the target waveform, and the response waveform, and
A fatigue testing machine comprising: In the fatigue testing machine according to claim 1,
The correction unit is
A fatigue testing machine comprising a correction calculation unit that corrects a wavelet transformation result of the drive waveform using a value obtained by multiplying a difference between the wavelet transformation result of the target waveform and the wavelet transformation result of the response waveform by a correction coefficient. In the fatigue testing machine according to claim 2,
The correction calculation unit
The correction coefficient includes an input response ratio representing a correlation between an input and a response obtained from a wavelet transform result of the drive waveform as an input and a wavelet transform result of the response waveform as a response, and a safety factor representing a correction strength. Fatigue testing machine that calculates using In the fatigue testing machine according to claim 3,
The correction calculation unit
A fatigue testing machine that acquires the input response ratio based on a frequency response transfer function that is a ratio of an input and a response for each frequency. In the fatigue testing machine according to claim 3,
The correction unit is
Correlation between an input increase / decrease based on a change in value at each point of the wavelet transform result of the drive waveform as an input and a response increase / decrease based on a change in value at each point of the wavelet transform result of the response waveform as a response A fatigue testing machine provided with a safety factor calculation unit that calculates the safety factor using

Images