JP7306916B2 - Anomaly detection system for inspection equipment - Google Patents

Description

The present invention relates to an abnormality detection system for inspection equipment.

The inspection apparatus includes a controller and an output unit that provides an output to a sample according to an input from the controller, and inspects the performance and durability of the sample. The controller feeds back the output of the output unit, controls the inspection device according to inspection conditions predetermined for the sample, and gives the output to the sample.

As such an inspection device, for example, there is a vibration inspection device. The vibration inspection device has a vibration exciter as an output unit, and vibrates a specimen such as a mechanical part or a damper with the vibration exciter. In this case, the output of the vibration inspection apparatus is physical quantities such as the load, velocity, and displacement applied to the specimen. Displacement is given (see, for example, Patent Document 1).

JP 2019-032261 A

The inspection of a specimen using such an inspection apparatus is performed on the premise that the inspection apparatus outputs in accordance with the output indicated by the input in response to the input. may compromise the reliability of the test results obtained.

As a method of detecting an abnormality in an inspection device, there is a method of determining an abnormality when the difference between the maximum output indicated by the given input and the actual output of the inspection device exceeds a threshold value. In this case, it is merely a comparison between the output indicated by the input and the actual output, and it is difficult to accurately judge the abnormality of the inspection device.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide an abnormality detection method for an inspection apparatus and an abnormality detection system for an inspection apparatus that can accurately detect an abnormality in the inspection apparatus.

In order to achieve the above object, an abnormality detection system for an inspection apparatus according to the present invention includes a sensor for detecting the output of an inspection apparatus when inspecting a specimen or without load, and a processor for processing the output of the inspection apparatus. The average value of the feature values of the output when the processing device inspected the reference sample by the normal inspection device in the past, and the output feature when the current inspection device inspected the reference sample or based on the difference between the average value of the output feature value when inspected by a normal inspection device in the past without load and the difference between the output feature value when inspected by the current inspection device without load a distortion factor calculation unit that detects an abnormality of the inspection apparatus by using a distortion factor calculation unit that determines the distortion factor based on the ideal output of the inspection apparatus for the input and the output of the inspection apparatus for the input; and the output of the inspection apparatus that is repeatedly changed. a variation degree calculation unit that obtains the degree of variation in the output obtained when the measurement is performed, a correlation coefficient calculation unit that obtains the correlation coefficient between the average value of the past outputs of the inspection device and the current output of the inspection device, and the distortion factor and a judgment unit for judging whether or not there is an abnormality in the inspection apparatus based on the difference between the average value of the past feature amount and the current feature amount, and the correlation coefficient, using the variation degree as the feature amount.

According to the inspection device abnormality detection method and the inspection device abnormality detection system configured in this way, since the average value of the feature amount of the output of the past normal inspection device is used, the normal inspection device can It is possible to determine whether the current inspection apparatus is abnormal by using the obtained feature amount as an index.

Further, in an abnormality detection system for an inspection device, the processing device includes an ideal output, which is an ideal output of the inspection device with respect to the input, and a distortion factor calculation unit that obtains a distortion factor based on the output of the inspection device with respect to the input. A variation degree calculator that calculates the degree of variation in the output obtained when the output is repeatedly changed, and a correlation coefficient that determines the correlation coefficient between the average value of the output of the past normal inspection device and the current output of the inspection device. a calculation unit, and a judgment unit that judges whether or not there is an abnormality in the inspection apparatus based on the difference between the average value of the past feature amount and the current feature amount, and the correlation coefficient, using the distortion rate and the degree of variation as the feature amounts. You may prepare. According to the abnormality detection system for an inspection apparatus configured in this way, the presence or absence of an abnormality in the inspection apparatus is determined using the distortion factor, the degree of variation, and the correlation coefficient. Since the presence or absence of abnormality in the inspection device can be grasped from the viewpoint of the degree of matching, the presence or absence of abnormality in the inspection device can be determined with high accuracy.

In addition, in the anomaly detection system for an inspection device, the distortion factor calculation unit includes an integrated value of the power spectrum density of an ideal waveform, which is an ideal output waveform of the inspection device for an input that changes the output with a sine wave of a predetermined period, and The distortion component may be the difference between the integrated value of the power spectral density of the output waveform, which is the output waveform of the inspection apparatus, and the distortion factor may be obtained based on the integrated value of the power spectral density of the output waveform and the distortion component. According to the anomaly detection system for the inspection apparatus configured as described above, not only the distortion of the ideal waveform and the output waveform with respect to the maximum load, but also the distortion factor that takes into account the difference between the two over the entire frequency range can be obtained. It is possible to more accurately grasp the actual dynamic behavior and judge the abnormality of the inspection device.

Furthermore, in the anomaly detection system of the inspection apparatus, the distortion factor calculator may remove the offset component of the output waveform to obtain the power spectral density of the output waveform. According to the anomaly detection system for an inspection apparatus configured in this way, since the power spectral density is calculated after removing the offset component from the output waveform, low-frequency noise can be removed from the power spectral density, and the distortion component is reduced from the actual value. is prevented from becoming large, and the distortion factor can be obtained with high accuracy. Therefore, according to the abnormality detection system for the inspection device, it is possible to more accurately detect the abnormality of the inspection device.

Further, in the anomaly detection system of the inspection apparatus, when the initial value and the final value of the output waveform are not the same value in the distortion factor calculation unit, the output waveform has the same value as the initial value and the initial value is an integer multiple of a predetermined period from the initial value. The power spectral density of the output waveform may be obtained after cutting off the terminal end of the output waveform at a point near the point of . According to the anomaly detection system for an inspection apparatus configured in this manner, high-frequency noise can be removed from the power spectral density, the distortion component can be prevented from becoming larger than it actually is, and the distortion factor can be obtained with high accuracy. Therefore, according to the abnormality detection system for the inspection device, it is possible to more accurately detect the abnormality of the inspection device.

Furthermore, the abnormality detection system of the inspection device has, in the degree-of-variation calculation unit, an output waveform that is the waveform of the output of the inspection device when an input that changes the output of the inspection device with a sine wave of a predetermined cycle is given a plurality of times. The degree of dispersion may be obtained based on the standard deviation or variance of . According to the anomaly detection system for an inspection device configured in this way, it is possible to obtain the degree of variation that indicates the magnitude of dynamic load fluctuations when the inspection device is actually used for a long period of time. It is possible to grasp the magnitude of the dynamic output fluctuation of the inspection device. Therefore, according to the abnormality detection system for the inspection device, it is possible to more accurately detect the abnormality of the inspection device.

In the anomaly detection system of the inspection device, the correlation coefficient calculation unit may obtain the correlation coefficient between the average value of the past outputs of the inspection device and the frequency distribution of the current output of the inspection device. According to the anomaly detection system for an inspection apparatus configured in this way, even if there is a phase shift between the past average waveform and the current output waveform of the inspection apparatus for which the correlation coefficient is to be obtained, the correct correlation coefficient between the two can be obtained. can ask. Therefore, according to the abnormality detection system for the inspection device, it is possible to more accurately detect the abnormality of the inspection device.

Further, the abnormality detection system for the inspection device further includes an output offset amount in the feature amount, the processing device has an offset amount calculation unit for obtaining the offset amount of the output of the inspection device, and the determination unit further includes the past inspection device output offset amount. The presence or absence of an abnormality in the inspection device may be determined based on the difference between the average value of the normal offset amounts and the current offset amount of the inspection device. According to the abnormality detection system for an inspection apparatus configured as described above, an abnormality determination process is performed using the offset amount, so an abnormality in the output of the inspection apparatus and an abnormality in the sensor can be detected.

Furthermore, the abnormality detection system of the inspection device is a vibration inspection device in which the inspection device vibrates the specimen, the rigidity of the output is further included in the feature amount, and the sensor detects the load of the inspection device and the displacement. The processor includes a stroke sensor, the processor has a stiffness calculator that obtains the stiffness from the load and displacement of the inspection device, and the determination unit further calculates the average value of the normal stiffness of the past inspection device and the stiffness of the current inspection device. The presence or absence of an abnormality in the inspection apparatus may be determined based on the difference between the . According to the abnormality detection system for an inspection apparatus configured in this way, the abnormality determination process is performed using the rigidity, so it is possible to detect an abnormality or deterioration of parts in the inspection apparatus and an abnormality of the sensor.

Furthermore, in the anomaly detection system of the inspection device, when the difference between the average value of the past feature quantity and the current feature quantity becomes equal to or greater than a threshold value set based on the standard deviation or variance of the past feature quantity, the inspection device may be judged to be abnormal. According to the anomaly detection system for an inspection device configured in this manner, an appropriate threshold value can be set to determine the range of values that can be taken by the feature quantity of a normal inspection device, so an anomaly in the inspection device can be accurately detected.

Then, when the inspection device is determined to be normal, the abnormality detection system of the inspection device calculates the average value of the feature amounts when the processing device has inspected normal samples in the past with the inspection device, and and a specimen abnormality judgment unit for judging whether or not there is an abnormality in the specimen based on the difference between a normal specimen and a specimen of the same product. According to the abnormality detection system for an inspection apparatus configured in this way, it is possible to detect an abnormality in a sample by performing the same process as that for determining whether or not there is an abnormality in not only the inspection apparatus but also the sample apparatus.

Then, in the anomaly detection system of the inspection apparatus, the specimen anomaly determination unit determines that the difference between the average value of the past feature amount and the current feature amount is equal to or greater than a threshold set based on the standard deviation or variance of the past feature amount. Then, the specimen may be judged as abnormal. According to the abnormality detection system for the inspection device T configured in this way, an appropriate threshold value can be set to determine the range of values that the characteristic quantity of a normal specimen can take, so that an abnormality of the inspection device can be accurately detected.

According to the abnormality detection system for an inspection apparatus of the present invention, an abnormality in an inspection apparatus can be accurately detected.

1 is a configuration diagram of an abnormality detection system for an inspection device according to an embodiment; FIG. It is a side view of an inspection device. It is the graph which showed the ideal load and the load waveform. It is a block diagram of a processing apparatus. It is the graph which showed the power spectral density of an ideal load, and the power spectral density of a load waveform. It is a figure explaining a distortion component. 7 is a graph for explaining the error between the power spectral density of the load waveform and the base frequency estimated from the power spectral density of the load waveform; 4 is a graph showing a load waveform containing offset components; 4 is a graph showing the power spectral density of a load waveform containing offset components; It is the graph which showed the load waveform from which an initial value and a final value differ. FIG. 4 is a graph showing the power spectral density of a load waveform containing high frequency noise; FIG. It is a figure explaining the process in a variation degree calculation part. It is the figure which showed the average waveform of the past load waveform of a normal inspection apparatus, and a load waveform. It is the graph which set the division with respect to the load waveform. It is the graph which showed the appearance frequency of load. It is a figure explaining the offset amount of the load waveform of an inspection apparatus. It is the figure which showed the characteristic of the load of an inspection apparatus, and a displacement. 4 is a flow chart showing an example of a procedure of abnormality determination processing of an inspection device in a processing device; 4 is a flow chart showing an example of the procedure of specimen abnormality determination processing of an inspection device in a processing device.

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described below based on embodiments shown in the drawings. As shown in FIG. 1, an abnormality detection system 1 for an inspection device T in one embodiment includes a sensor 2 for detecting an output of the inspection device T in response to an input when inspecting a damper D as a specimen, and a sensor 2 for detecting an output of the inspection device T. A processing device 3 for processing the output is provided, and an abnormality of the inspection device T is detected.

Each part of the abnormality detection system 1 of the inspection device T will be described in detail below. In this embodiment, the inspection device T is a vibration inspection device that vibrates a damper D as a specimen. The abnormality detection system 1 detects an abnormality of the inspection apparatus T by using the output of the inspection apparatus T detected when a vibration is applied to a damper D serving as a reference as a specimen. Further, in the anomaly detection system 1, when judging an anomaly of the inspection device T, the data of the output when the same inspection device T was judged to be normal in the past is used, and the same damper D as the past output is used. The presence or absence of an abnormality in the inspection device T is detected based on the output detected this time when the same inspection device T is used for inspection. That is, in the abnormality detection system 1, the output obtained when the damper D was inspected by the normal inspection apparatus T in the past by using the damper D as a sample for judging whether or not there is an abnormality in the inspection apparatus T. , the presence or absence of an abnormality in the inspection device T is detected based on the output obtained when the inspection device T inspects the damper D in order to newly detect an abnormality this time. The characteristics of the damping force exerted with respect to the expansion/contraction speed of the damper D are grasped in advance, and when the same vibration is input from the inspection device T, the damper D expands/contracts at the same speed based on the characteristics and exerts the same damping force.

A damper D, which is a sample in the present embodiment, is a telescopic damper that includes a cylinder 5 and a rod 6 that moves in and out of the cylinder 5. The rod 6 is axially displaced with respect to the cylinder 5. Demonstrates damping force when expanding and contracting.

On the other hand, the inspection device T has a controller C and a vibrator E, as shown in FIG. The vibration exciter E includes a pedestal 10, a holding portion 11 which is provided on the pedestal 10 and is movable in the horizontal direction in FIG. and an actuator 13 connected to the damper D to vibrate the damper D.

The actuator 13 includes a cylinder 13a, a piston (not shown) which is movably inserted into the cylinder 13a and partitions the inside of the cylinder 13a into an expansion side chamber and a compression side chamber (not shown), and a piston which is movably inserted into the cylinder 13a. It is provided with a rod 13b connected to a piston and a servo valve 13c for selectively feeding pressurized oil supplied from a pump (not shown) to the expansion side chamber and the compression side chamber.

Although not shown in detail, the servo valve 13c includes a hollow housing, a spool that is movably inserted into the housing, a solenoid that drives the spool, a spring that positions the spool at a neutral position, and a spring that positions the spool at a neutral position. and a drive circuit that receives an input and drives the solenoid. The solenoid can change the thrust applied to the spool according to the amount of current supplied from the drive circuit, and can adjust the position of the spool. Depending on the position of the spool, the servo valve 13c has a position that supplies pressure oil to the expansion side chamber, a position that supplies pressure oil to the compression side chamber, and a position that cuts off the supply of pressure oil to both. , and the flow rate is adjusted according to the amount of current supplied to the solenoid at the position where pressure oil is supplied to the expansion side chamber or the compression side chamber.

In this embodiment, when the servo valve 13c receives a current command I as an input, it adjusts the thrust of the solenoid, adjusts the position of the spool with respect to the housing, and either the expansion side chamber or the compression side chamber is instructed by the input. The chamber is supplied with pressurized oil at the flow indicated by the input. The actuator 13 expands the chamber to which the pressure oil is supplied from the servo valve 13c and shrinks the chamber to which the pressure oil is not supplied to extend and contract. In this way, the vibration exciter E, upon receiving an input from the controller C, causes the actuator 13 to expand and contract, vibrates one end of the damper D, and vibrates the damper D. As shown in FIG. The drive circuit has a current sensor that detects current flowing through the solenoid, feeds back the current detected by the current sensor, and supplies current to the solenoid according to the current command I input from the controller C. The drive circuit may be included in the controller C instead of the servo valve 13c side.

In detecting an abnormality in the inspection apparatus T, the controller C repeatedly provides the actuator 13 with a current command I that causes the actuator 13 in the vibration exciter E to expand and contract with a sine wave of a predetermined cycle. In this manner, the inspection apparatus T drives the actuator 13 to repeatedly apply sinusoidal vibration to the damper D, which is the specimen. In this embodiment, the output of the inspection apparatus T is the specimen. A load F, a velocity V, and a displacement X are given to the damper D. Therefore, the waveforms of the ideal load, the ideal velocity, and the ideal displacement as the ideal output (ideal output) of the vibrator E indicated by the current command I, that is, the ideal waveform Wa, as shown in FIG. It becomes a waveform that changes with a sine wave of a period. The current command I for changing the output of the inspection device T with such a sine wave of a predetermined period may be stored in advance in the controller C or may be stored in the controller C at the time of inspection. Note that the predetermined period can be set arbitrarily. Although the output of the inspection device T is the load F, the velocity V and the displacement X in this embodiment, the output is not limited to these. Either one or two may be used.

Subsequently, the sensor 2 detects the output of the inspection device T when inspecting the damper D according to the current command I as an input from the controller C. FIG. In the present embodiment, the output of the inspection device T is the load F, the velocity V and the displacement X. Therefore, the sensor 2 is installed between the actuator 13 and the damper D and detects the load exerted by the actuator 13. A load cell 2a for detection, a speed sensor 2b attached to the rod 13b of the actuator 13 to detect the speed of the rod 13b, and a stroke sensor 2c for detecting the expansion and contraction displacement of the actuator 13 are provided. The sensor 2 detects the load F, the velocity V and the displacement X as actual outputs of the inspection device T and inputs them to the processing device 3 . The velocity sensor 2b detects the acceleration of the rod 13b and integrates the detected acceleration to detect the velocity V of the inspection apparatus T. The velocity V is detected by differentiating the displacement X detected by the stroke sensor 2c. You may make it Further, in detecting the velocity, the acceleration of the rod 13b detected by the acceleration sensor may be input to the processing device 3, and the acceleration may be integrated by the processing device 3 to detect the velocity V of the rod 13b.

As shown in FIG. 4, the processing unit 3 is a computer system, and stores an arithmetic processing unit 3a and a program necessary for control and processing of the processing unit 3, and the arithmetic processing unit 3a is necessary for executing the program. an interface 3c that receives signals from the load cell 2a, the speed sensor 2b, and the stroke sensor 2c; an input device 3d such as a keyboard and a mouse; a display device 3e; an auxiliary storage device 3f; It has a printer 3g as a printing device and a bus 3h for communicably connecting these devices.

The arithmetic processing device 3a is a CPU or the like that performs arithmetic processing, and controls each part of the processing device 3 by executing an operating system and other programs. Based on X, processing is performed to obtain four values of the distortion factor ε, the degree of variation Sd, the offset amount O, and the stiffness K and five indices of the correlation coefficient R as the characteristic quantities of the inspection apparatus T. The feature quantity is an index capable of grasping the static or dynamic characteristics of the inspection apparatus T, and is composed of the four indices of the distortion rate ε, the degree of variation Sd, the offset amount O, and the stiffness K. . In addition to the values described above, the special quantity includes the maximum value of the output with respect to the input and the followability of the output with respect to the input. The storage device 3b has a hard disk in addition to ROM and RAM. The interface 3c converts analog signals input from the load cell 2a, speed sensor 2b, and stroke sensor 2c into digital signals readable by the arithmetic processing unit 3a. The display device 3e has a screen for displaying data processed by the arithmetic processing device 3a, and is, for example, a liquid crystal display. The auxiliary storage device 3f is composed of a computer-readable recording medium and a storage medium drive device, and the storage medium is a magnetic disk, a magneto-optical disk, an optical disk, a semiconductor memory, or the like. In addition, the processing device 3 generates a graph plotting the values and values of the load F, the speed V and the displacement X detected by the load cell 2a, the speed sensor 2b and the stroke sensor 2c, the distortion factor ε, the degree of variation Sd and the correlation coefficient R. An application program for obtaining the distortion factor ε, the degree of variation Sd, and the correlation coefficient R is stored in the storage device 3b.

In the present embodiment, the arithmetic processing device 3a of the processing device 3 executes a program for obtaining the distortion factor ε, the degree of variation Sd, the offset amount O, the stiffness K, and the correlation coefficient R, which are characteristic quantities. By doing so, a distortion rate calculation unit 3a1 for obtaining the distortion rate ε, a variation degree calculation unit 3a2 for obtaining the degree of dispersion Sd, a correlation coefficient calculation unit 3a3 for obtaining the correlation coefficient R, and an offset amount of the output of the inspection apparatus T An offset amount calculation unit 3a4 for obtaining O and a rigidity calculation unit 3a5 for obtaining the rigidity of the inspection device T are implemented. Further, the arithmetic processing unit 3a of the processing unit 3 executes a program for judging whether or not there is an abnormality in the inspection device T based on the distortion factor ε, the degree of variation Sd, and the correlation coefficient R, and realizes the judging unit 3a6. there is The processing device 3 obtains the load F of the inspection device T, the strain rate ε associated with the velocity V and the displacement X, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K of the inspection device T, which is the object of abnormality detection. , the obtained distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the rigidity K, it is determined whether or not the inspection device T is abnormal.

Calculation of the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K will be described in detail below. The distortion factor ε, the degree of variation Sd, the correlation coefficient R, and the offset amount O are obtained for the load, velocity, and displacement in a plurality of past outputs of the normal inspection device T, and for detecting an abnormality this time, The load F, the velocity V and the displacement X in the output detected by the sensor 2 by giving an input to the inspection device T are also obtained.

The past output data of the normal inspection device T is the output data of the normal inspection device T obtained in the past. In the anomaly detection system 1, in addition to current output data of the inspection apparatus T detected by giving an input used for anomaly detection to the inspection apparatus T which is about to detect an anomaly, the normal past data of the same inspection apparatus T is detected. Since it is also necessary to use the output data of , the past output data are stored in the storage device 3b or the auxiliary storage device 3f. In addition, in order to calculate the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K, the same input as the input given to the inspection device T for abnormality detection was input to the same inspection device T in the past. You need the output you get when you give it. The input that the controller C gives to the inspection device T for detecting an abnormality is a command to increase or decrease the load, speed, and displacement of the inspection device T in a sine wave. Increments and decrements waveforms containing unit waves of minutes. A large number of past normal outputs detected by the sensor 2 when the above input was given to a normal inspection apparatus T are prepared in advance for abnormality detection. Sd, offset amount O and stiffness K are obtained one by one. When an input is given to the inspection apparatus T for abnormality detection and it is determined that the inspection apparatus T is normal for the current input, the output of the current inspection apparatus T is used as the output of the next inspection apparatus T. In determining an abnormality, past normal outputs may be handled, and past output data may be accumulated.

The process of calculating the strain rate ε of the load F, velocity V and displacement X is the same, and the process of calculating the degree of variation Sd of the load F, velocity V and displacement X is the same. The calculation process of the correlation coefficient R of is also the same. The calculation process of the offset amount O of the load F, velocity V and displacement X is also the same. Therefore, in the description of the calculation of the distortion factor ε, the degree of variation Sd, the correlation coefficient R, and the offset amount O, the calculation of the distortion factor ε, the degree of variation Sd, the correlation coefficient R, and the offset amount O with respect to the load F will be described. , the distortion rate ε, the degree of variation Sd, the correlation coefficient R, and the offset amount O for the velocity V and the displacement X, detailed description of which will be omitted.

Furthermore, the process of calculating the distortion factor ε, the degree of variation Sd, the offset amount O, and the stiffness K for the past output of the inspection device T, and the distortion factor ε and the degree of variation Sd for the current output of the detection device T to be detected as an abnormality detection target. , the offset amount O and the stiffness K are the same. In this way, the distortion factor calculator 3a1 obtains the distortion factor ε from a plurality of past outputs of the inspection device T, and similarly obtains the current output of the inspection device T when executing abnormality detection. The variation degree calculator 3a2 obtains the variation degree Sd from a plurality of past outputs of the inspection device T, and similarly obtains the current output of the inspection device T when executing the abnormality detection. The offset amount calculator 3a4 obtains the offset amount O from a plurality of past outputs of the inspection device T, and similarly obtains the current output of the inspection device T when executing abnormality detection. The stiffness calculator 3a5 obtains the stiffness K from a plurality of past outputs of the inspection device T, and also from the current output of the inspection device T when executing the abnormality detection. The correlation coefficient calculation unit 3a3 obtains a normal waveform from a plurality of past outputs of the inspection device T, and determines the correlation between the normal waveform and the waveform of the current output of the inspection device T when executing abnormality detection. Find the number R. Further, the stiffness calculator 3a5 obtains a plurality of stiffnesses K from a plurality of loads F and displacements X of the inspection device T in the past. , the distortion factor ε, the degree of variation Sd, the offset amount O, and the stiffness K are calculated in advance and stored as data in the storage device 3b or the auxiliary storage device 3f of the processing device 3 instead of being obtained each time an abnormality is detected. , can be used for anomaly detection.

First, the distortion factor calculator 3a1 will be described in detail. As shown in FIG. 3, a load waveform Wb ( dashed line). The same applies to the velocity waveform drawn by the velocity V and the displacement waveform drawn by the displacement X. Differences occur between the velocity waveform drawn by the velocity and the ideal waveform, and between the displacement waveform and the ideal waveform. The load waveform Wb is a waveform obtained by plotting the values of the load F detected at a predetermined sampling period in chronological order, with the load on the vertical axis and the time on the horizontal axis. , and time on the horizontal axis. There are values of the load waveform Wb corresponding to the values. As can be understood from the graph of FIG. 3, when the controller C inputs a current command I that increases or decreases the output in a sine wave of a predetermined cycle to the inspection device T, the ideal output of the inspection device T indicated by the current command I is an ideal output. A difference occurs between the load and the load F that the inspection apparatus T actually outputs in response to the current command I being input.

Therefore, the distortion factor calculator 3a1 obtains the distortion factor ε based on the difference between the ideal load and the load F. The distortion factor ε is a value that serves as a measure for grasping how much the load waveform Wb is distorted with respect to the ideal waveform Wa. The distortion factor calculation unit 3a1 calculates the distortion factor ε related to the load F of the normal inspection apparatus T obtained in the past, and calculates the load waveform Wb formed of the data group of the load F of the normal inspection apparatus T obtained in the past. In addition, in calculating the distortion factor ε related to the current load F of the inspection device T newly obtained at the time of abnormality detection, the load waveform Wb consisting of the data group of the load F of the inspection device T this time is used as use. The larger the distortion factor ε, the more the dynamic behavior of the inspection apparatus T is disturbed with respect to the load indicated by the current command I, and the larger the load F fluctuates. is one index for evaluating Therefore, if the distortion factor ε is selected as the feature quantity of the inspection apparatus T for detecting an abnormality, the abnormality can be detected from the viewpoint of the dynamic characteristics of the inspection apparatus T.

Since the ideal load is the load indicated by the current command I, the processing device 3 stores in advance an ideal waveform Wa consisting of a data group of the load indicated by the current command I given to the inspection device T by the controller C, The distortion factor calculation part 3a1 may use this, or the current command I is received from the controller C each time an abnormality of the inspection apparatus T is detected, and the distortion factor calculation part 3a1 converts the current command I into an ideal waveform Wa. You can use it by Further, the load F of the inspection device T detected by the load cell 2a is obtained at a predetermined sampling period and input to the processing device 3, so that the load waveform Wb consisting of the data group of the load F is obtained from the load cell 2a, and the strain It is used in the ratio calculator 3a1, the variation degree calculator 3a2, the correlation coefficient calculator 3a3, the offset amount calculator 3a4, and the stiffness calculator 3a5. Similarly to the load F, the velocity V and the displacement X detected by the velocity sensor 2b and the stroke sensor 2c are also calculated by a distortion factor calculator 3a1, a variation degree calculator 3a2, a correlation coefficient calculator 3a3, an offset amount calculator 3a4, and a stiffness calculator. 3a5, the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K related to the velocity V and the displacement X are obtained.

In the present embodiment, as shown in FIG. 5, the distortion factor calculator 3a1 obtains the power spectral density Pa (solid line in FIG. 5) of the ideal waveform Wa by Fourier transforming the ideal waveform Wa, and A power spectral density Pb (broken line in FIG. 5) of the load waveform Wb is obtained by Fourier transforming the load waveform Wb detected when a current command I that changes the load with a sine wave of a predetermined cycle is given. The power spectral density Pa of the ideal waveform Wa is such that the current command I indicates the load of the sinusoidal wave with a predetermined cycle, so as shown in FIG. It becomes a standing waveform. On the other hand, the power spectral density Pb of the load waveform Wb is, as shown in FIG. Since the resulting vibration and noise-induced vibration are superimposed on the load waveform Wb, these vibration components appear on the higher frequency side than the predetermined frequency.

As shown in FIG. 6, the distortion factor calculator 3a1 calculates the ideal waveform Wa The integrated value of the power spectral density Pa obtained by integrating the power spectral density Pa on the frequency axis is subtracted to obtain the distortion component Pab (the area of the hatched portion in FIG. 6), which is the difference between the two.

Then, the distortion factor calculation unit 3a1 divides the distortion component Pab by the integrated value Zb, which is the area of the hatched portion in FIG. It is obtained as the rate ε. That is, the distortion factor calculator 3a1 obtains the distortion factor ε by calculating ε=Pab/Zb. Thus, the distortion factor ε is obtained by obtaining the distortion component Pab, which is the difference between the integrated value of the power spectral density Pa of the ideal load (ideal output) and the integrated value Zb of the power spectral density Pb of the load waveform Wb. It is obtained by dividing by the integrated value Zb of the power spectral density Pb of the waveform Wb.

The distortion factor calculation unit 3a1 calculates the distortion factor using the distortion component Pab, which includes the vibration component over all frequencies in which the load F of the inspection apparatus T deviates from the ideal load when the load is changed by a sine wave of a predetermined cycle. Find ε. Therefore, it is possible to objectively determine how much the load F of the inspection device T deviates from the ideal load from the distortion factor ε, and to grasp the actual behavior of the inspection apparatus T from the distortion factor ε.

The base frequency is estimated from the power spectral density Pb of the load waveform Wb of the inspection apparatus T using the zero-crossing method, the autocorrelation method, the period histogram, etc., and the power spectral density Pa of the ideal waveform Wa is calculated from the power spectral density Pb. It may be estimated, but in this way, as shown in FIG. 7, the base frequency of the estimated power spectral density Pb indicated by the solid line in FIG. 7 and the load waveform Wb indicated by the broken line in FIG. An error λ occurs in the frequency, and the value of the distortion component Pab increases. Since noise and offset components are superimposed on the load waveform Wb, when the base frequency is estimated, it inevitably deviates from the frequency of the ideal waveform Wa, causing such a phenomenon. On the other hand, the distortion factor calculation unit 3a1 in the processing device 3 of the present embodiment obtains the power spectral density Pa with the waveform of the load indicated by the current command I as the ideal waveform Wa, and calculates the power spectral density Pa of the ideal waveform Wa. The distortion component Pab is obtained by subtracting the integrated value of Pa from the integrated value of the power spectral density Pb of the load waveform Wb. In this way, the frequency at which the power spectral density Pa obtained from the ideal waveform Wa reaches its peak and the frequency at which the power spectral density Pb obtained from the load waveform Wb reaches its peak can be matched with high accuracy, and the distortion component Pab can be prevented from becoming excessively larger than it actually is.

When the load waveform Wb of the load F is offset by low-frequency noise, the distortion factor calculator 3a1 calculates the power spectral density Pb after removing the offset component from the load waveform Wb. Specifically, the distortion factor calculator 3a1 shifts the load waveform Wb up and down in FIG. 8 to search for an offset component that makes the positive side integral value and the negative side integral value of the load waveform Wb equal. That is, if the broken line indicating the offset amount in FIG. 8 is taken as the horizontal axis, the integral value on the positive side and the integral value on the negative side of the load waveform Wb match. In this way, the distortion factor calculator 3a1 obtains the offset component, subtracts the offset component from the value of the load waveform Wb, obtains a new waveform, and then obtains the power spectral density Pb. As shown in FIG. 8, for example, when the load waveform Wb is offset by the superimposition of a low-frequency offset component, if the power spectral density Pb is obtained without removing the offset component from the load waveform Wb, the result shown in FIG. As shown, low-frequency noise appears in the obtained power spectral density Pb, and if the distortion component Pab is obtained as it is, the value of the distortion component Pab becomes larger than the actual value, resulting in an increase in the distortion factor ε. On the other hand, when the load waveform Wb of the load F is offset by low-frequency noise, the distortion factor calculation unit 3a1 calculates the power spectrum density Pb after removing the offset component from the load waveform Wb. Low-frequency noise can be removed from the density Pb, and an accurate distortion factor ε can be obtained. Since the distortion factor calculation unit 3a1 processes the load waveform Wb in this way, in the abnormality detection system 1 of the inspection apparatus T of the present embodiment, the distortion component Pab is prevented from becoming larger than it actually is, and the distortion factor ε can be obtained with high accuracy.

In order to remove the offset component, in addition to the processing described above, for example, the median value of the maximum value and the minimum value is obtained for each cycle of the load waveform Wb, and the average value of these median values is regarded as the drift component. Alternatively, the average value of the median values may be subtracted from the load waveform Wb, or the average value of the values of the load F may be subtracted from the load waveform Wb as an offset component.

Further, as shown in FIG. 10, when the initial value and the final value of the load waveform Wb observed at the load cell 2a are different, when the load waveform Wb is Fourier-transformed as it is, the power spectrum density Pb shows high-frequency noise as shown in FIG. appears, the distortion component Pab becomes larger than it actually is, and the distortion factor ε becomes large. Therefore, when the initial value and the final value of the observed load waveform Wb are not the same, the distortion factor calculation unit 3a1 of the present embodiment calculates the same value as the initial value W1 in the load waveform Wb as shown in FIG. Then, the power spectrum density Pb of the load waveform Wb is obtained after cutting off the end side of the load waveform Wb at a point W2 near a point that is an integral multiple of the predetermined period from the initial value W1. The point may be a point with a number obtained by adding an integral multiple of the number of data contained in the unit waveform from the first data, which is the initial value W1, and the closest point before and after this point is F2. do it. The current command I given to the inspection device T changes the load with a sine wave of a predetermined cycle. , the initial value W1 of the load waveform Wb thus extracted and the value at the point W2 are the same. Note that there are several points in the load waveform Wb that have the same value as the initial value W1, but if a point that appears at the end of the load waveform Wb among the points that satisfy the above conditions is selected as F2, the load waveform Wb can be obtained by selecting the point that appears at the end of the load waveform Wb as F2. This is preferable because it increases the number of waveforms included. If there is no value that completely matches the initial value W1, the point that becomes the same value by the same rounding as the value obtained by performing the rounding of the initial value W1 is set as F2, or the effective value of the initial value W1 is set. A point having a value that matches the number may be designated as F2, and this may be treated as a point having the same value as the initial value W1. If a point W2 having the same value as the initial value W1 cannot be found even after performing such a process, one of the values falling within a predetermined threshold value from the initial value W1 going back from the final data of the load waveform Wb is set to the initial value W1. The point with the closest value should be chosen for F2.

In this way, when the distortion factor calculation unit 3a1 processes the load waveform Wb and then Fourier-transforms the load waveform Wb to obtain the power spectral density Pb, the above-described high-frequency noise can be removed from the obtained power spectral density Pb. Since the distortion factor calculation unit 3a1 processes the load waveform Wb in this way, in the abnormality detection system 1 of the inspection apparatus T of the present embodiment, the distortion component Pab is prevented from becoming larger than it actually is, and the distortion factor ε can be obtained with high accuracy.

When obtaining the distortion factor ε, for example, the difference between the maximum value of the ideal waveform Wa and the maximum value of the load waveform Wb is divided by the maximum value of the load waveform Wb. The value may be obtained by calculating the difference between the value of the ideal waveform Wa and the value of the load waveform Wb corresponding to this value, and dividing the average value of the difference obtained by integration by the maximum amplitude of the load waveform Wb. On the other hand, as described above, the integrated value of the power spectral density Pa of the ideal waveform Wa and the integrated value of the power spectral density Pb of the load waveform Wb are subtracted to obtain the distortion component Pab. When the distortion factor ε is obtained by dividing Wb by the integrated value of the power spectral density Pb, the distortion factor ε taking into account not only the distortion of the ideal waveform Wa and the load waveform Wb with respect to the maximum load but also the difference between the two over the entire frequency range can be obtained. Therefore, the actual dynamic behavior of the inspection device T can be grasped more accurately.

The distortion factor calculation unit 3a1 processes the load waveform Wb obtained from the load F detected by the load cell 2a and the ideal waveform Wa which is the waveform of the ideal load, which is the ideal output indicated by the current command I, and calculates the past of the inspection device T. The distortion factor ε related to the load is obtained for each of the plurality of outputs of and the current output. Similarly, the distortion factor calculator 3a1 calculates the velocity waveform obtained from the velocity V detected by the velocity sensor 2b and the ideal waveform, which is the waveform of the ideal velocity, which is the ideal output indicated by the current command I, for the aforementioned load F. By performing the same processing as the processing, the distortion factor ε related to the speed is obtained for each of the past outputs and the current output of the inspection device T. FIG. Further, the distortion factor calculation unit 3a1 calculates the displacement waveform obtained from the displacement X detected by the stroke sensor 2c and the ideal waveform, which is the waveform of the ideal displacement, which is the ideal output indicated by the current command I. By performing the same processing as in , the distortion factor ε related to the displacement is obtained for each of the past outputs and the current output of the inspection apparatus T. Therefore, the distortion factor calculator 3a1 obtains the distortion factor ε from a plurality of past data of the normal inspection apparatus T, and also obtains the distortion factor ε from the output of the inspection apparatus T this time.

Next, the variation degree calculator 3a2 will be described in detail. When the current command I that changes the load with a sine wave of a predetermined cycle is repeatedly given from the controller C to the inspection device T, the characteristics such as the overlap characteristic, the underlap characteristic, or the bridge characteristic due to the structure of the servo valve 13c A spike waveform or vibration may be randomly superimposed on the load waveform Wb to disturb the waveform. Such waveform disturbance may appear at different points in one cycle of the waveform (unit waveform) included in the load waveform Wb, and may not appear at a specific point in each unit waveform. Moreover, since the disturbance of the load of the inspection device T does not necessarily occur at the maximum load, the dynamic characteristics of the inspection device T cannot be grasped only by focusing on the variation of the maximum load.

Therefore, the current command I is repeatedly given a sufficient number of times to calculate the degree of variation Sd of the inspection device T, and the load F is sequentially detected by the load cell 2a. A value dispersion degree Sd is obtained. The variation degree calculation unit 3a2 uses the load F of the normal inspection device T obtained in the past in calculating the degree of variation Sd related to the load F of the normal inspection device T obtained in the past, and also uses the load F of the normal inspection device T obtained in the past. In calculating the degree of variation Sd related to the load F of the inspection device T this time, which is newly obtained, the load F of the inspection device T this time is used.

The degree of variation Sd is a value used as a scale for grasping whether or not the load of the inspection device T repeatedly matches the load instructed by the current command I when the load as the output of the inspection device T is repeatedly increased and decreased. be. That is, the degree of variation Sd is an index for evaluating the dynamic characteristics of the inspection device T from the viewpoint of repeatability, which is whether or not the load can be repeatedly output according to the current command I when the inspection device T repeats the load fluctuation. A larger value indicates that the behavior of the inspection device T is disturbed and the repeatability deteriorates. Therefore, if the degree of variation Sd is selected as the feature quantity of the inspection apparatus T for detecting anomalies, anomalies can be detected from the viewpoint of the repeatability of the inspection apparatus T. FIG.

To obtain the degree of variation Sd, the degree-of-variation calculator 3a2 specifically extracts values of the same phase for each unit waveform of the load waveform Wb of the inspection apparatus T to obtain the variance of the values of the same phase. The load waveform Wb is a waveform obtained by plotting the values of the load F captured by the processing device 3 at a predetermined sampling period in time series, with the load on the vertical axis and the time on the horizontal axis. The variance is determined for the number of values contained in . For example, when the predetermined period is 1 second, the current command I is given to the inspection device T 100 times, and the sampling period is 0.01 second, the number of data included in the unit waveform of the load waveform Wb is is 100, and the number of unit waveforms included in the load waveform Wb is 100.

As shown in FIG. 12, the variation degree calculator 3a2 extracts the values of the same phase from the 100 unit waveforms, obtains the values of the same phase in each of the 100 unit waveforms, and calculates the values of the 100 unit waveforms. Find the variance σ ² of the values. Let _Fn be the value of the n-th (n=1, 2, 3, . When obtaining σ _N ² , extract the N-th data in all unit waveforms, obtain the average value from the extracted 100 values, and sum the square values of the values obtained by subtracting the average value from each data value. to obtain the variance σ _N ² . The degree-of ^{-variation calculator 3a2 obtains a total of 100 variances σ n 2} _{( n} ⁼ 1, 2, 3, .

The variance σ _n ² is the variance of values between the same phases in each unit waveform in the load waveform Wb. Then, the degree-of-variation calculator 3a2 finds the average value of the variances σ _n ² and multiplies it by three to obtain the degree of variation Sd. That is, the variation degree calculator 3a2 obtains the variation degree Sd by calculating D=(3/100)×(σ ₁ ² +σ ₂ ² +σ ₃ ² + . . . +σ ₁₀₀ ² ).

Generalizing this, when the number of unit waveforms included in the load waveform Wb is M, and the number of data in the waveform for one cycle is N, the value of the nth data in the mth unit waveform is expressed as _Fmn . First, when obtaining the variance for the N-th data in the unit waveform, the variation degree calculator 3a2 extracts only the N-th data from the M unit waveforms, obtains the average value _μn of _{the n} -th data, The variance σ _N ² is obtained by calculating the following equation (1). The variation degree calculator 3a2 sequentially calculates the equation (1) for the N pieces of data included in each unit waveform to obtain a total of N pieces of variance σ _n ² (n=1, 2, 3 . . . N). ).

Subsequently, the variation degree calculator 3a2 obtains the variation degree Sd by multiplying the average value of the N variances by 3, as shown in the following equation (2).

Here, the dispersion calculated by the variation degree calculating unit 3a2 and the average value calculated for obtaining the dispersion are obtained from the data in the load waveform Wb. The data in the load waveform Wb can be regarded as a sample if the data obtained in 1 is taken as the population. Assuming that the population follows a normal distribution, if the number of samples is sufficiently large, the variance σ _n ² obtained as described above can be regarded as the variance of the population. Note that an unbiased variance may be obtained instead of the variance σ _n ² described above, and this unbiased variance may be treated as the variance of the population. Therefore, the variation degree calculator 3a2 calculates the unbiased variance S _n 2 by calculating the following equation ( ³ ), substitutes the unbiased variance S _n ² into the equation (2) instead of the variance σ _n ² , and calculates the variation A degree Sd may be obtained.

When the population follows a _normal distribution, the standard deviation of the population is _σn , and the average value of the population is _μn . The degree of variation Sd is a value obtained by multiplying the average value of the sum of N variances σ _n ² by three. This value takes into consideration not only the maximum load but also the variation in the entire unit waveform. In this way, the variation degree Sd indicates the magnitude of dynamic load fluctuation when the inspection device T is actually used for a long period of time. It is possible to grasp the magnitude of fluctuation. If it is assumed that the population follows other probability distributions such as the chi-square distribution, the dispersion degree Sd can be obtained by obtaining the variance in the probability distribution that is assumed to follow.

Also, instead of the variance σ _n ² , the standard deviation σ _n may be obtained, and the average value of the sum of 3σ _n may be used as the degree of dispersion Sd. Since the standard deviation σ _n is a scale indicating the variation of the load as well as the variance σ _n ² , the magnitude of the dynamic load variation of the inspection apparatus T is similar to the variation degree Sd obtained by using the variance σ _n ² . can be grasped. However, since the variance σ _n ² is peakier than the standard deviation σ _n , it is better to use the variance σ _n ² to obtain the degree of variation Sd. Since it appears prominently in the value of Sd, it becomes easier to grasp the repeatability of the inspection apparatus T. FIG.

The variation degree calculator 3a2 processes the load waveform Wb obtained from the load F detected by the load cell 2a, and obtains the load variation degree Sd for each of the past outputs and the current output of the inspection device T. . Similarly, the degree-of-variation calculator 3a2 performs the same processing as that for the load F described above on the velocity waveform obtained from the velocity V detected by the velocity sensor 2b, and compares the past outputs of the inspection device T with the velocity waveform. The degree of variation Sd related to the speed V is obtained for each of the current outputs. Furthermore, the degree-of-variation calculator 3a2 performs the same processing as that performed for the load F described above on the displacement waveform obtained from the displacement X detected by the stroke sensor 2c, and outputs a plurality of past outputs of the inspection device T and the current output. , and the degree of variation Sd related to the displacement X is obtained. Therefore, the variation degree calculator 3a2 obtains the variation degree Sd from a plurality of past data of the normal inspection device T, and also obtains the variation degree Sd from the output of the inspection device T this time.

Next, the correlation coefficient calculator 3a3 will be described in detail. If the characteristics change due to aged deterioration or abnormality due to long-term use of the inspection device T, even if the same input and the same output are given to the same damper D, the load given to the damper D by the normal inspection device T and the load given to the damper D by the abnormal inspection device T will be different. A difference occurs between the load F applied to D and the load F applied to D. This is the same even if the same damper D is given the velocity V or the displacement X as the output of the inspection device T. difference occurs. If there is a large difference between the current output of the inspection device T and the output of the previous normal inspection device T when detecting an abnormality, the inspection device T cannot normally give an output to the damper D. A target load F, velocity V, or displacement X cannot be applied to the damper D. On the other hand, if a certain correlation is recognized between the output of the inspection apparatus T that was normal in the past and the output of the inspection apparatus T this time, it is said that the inspection apparatus T can give the output to the damper D as instructed by the input. good.

The correlation coefficient calculation unit 3a3 calculates the average value of the past outputs of the normal inspection device T and the load F as the current output of the inspection device T obtained by vibrating the damper D for detecting an abnormality. A correlation coefficient R is obtained. The correlation coefficient R is a value that indicates whether the average value of the past outputs of the normal inspection device T and the load F as the current output of the inspection device T are correlated with each other, and the value is 1. It shows that the correlation between the current load F of the inspection device T and the load F of the past normal inspection device T is higher as it approaches. Therefore, if the correlation coefficient R is obtained for abnormality detection, the abnormality of the inspection apparatus T can be detected from the viewpoint of the degree of waveform matching.

The correlation coefficient calculator 3a3 calculates a plurality of past load waveforms Wb of the load F detected when the inspection device T determined to be normal increases or decreases the load applied to the damper D by the input of the current command I. Calculate the average value from the data of Specifically, the average value of the load F in a plurality of past load waveforms Wb of the inspection device T judged to be normal is calculated and obtained, and the average waveform of the average value of the past load F is obtained. Since the load F is detected at a predetermined sampling period, each waveform is cut out so that the phases of a plurality of past load waveforms Wb and the number of unit waveforms included in the load waveform Wb are the same, and a plurality of waveforms are obtained. Find the average of the values in the same order. For example, when the average value of the load F in the three past load waveforms Wb1, Wb2, and Wb3 of the normal inspection apparatus T is taken to obtain the average waveform, the average waveform WB shown in FIG. 13 is obtained. The average waveform WB, which is the average value of a plurality of loads F of past normal inspection devices T, is composed of the average values of the normal values that the output of the normal inspection device T can take. Therefore, if the correlation between the average waveform WB and the load waveform Wb that is the load F of the current inspection device T is high, it can be estimated that the current inspection device T is normal.

From the above, the correlation coefficient R between the load F of the average waveform WB and the load F of the inspection apparatus T at this time can be obtained. Here, if the average waveform WB and the load waveform Wb of the load F detected by the inspection apparatus T this time are out of phase, even if both waveforms match except for the phase, the value of the correlation coefficient R becomes smaller and the correlation becomes weaker.

Therefore, in the present embodiment, the correlation coefficient calculation unit 3a3 increases or decreases the load applied to the damper D by the inspection device T according to the frequency distribution of the load F in the average waveform WB described above and the input of the current command I. A correlation coefficient R between the frequency distribution of the load F in the detected current load waveform Wb is obtained. In calculating the correlation coefficient R, the controller C gives the current command I to the repetitive inspection device T to obtain the load waveform Wb including a sufficient number of unit waveforms.

With the load on the vertical axis and the time on the horizontal axis, a section with a predetermined width is determined with respect to the load, and the number of data belonging to each section of the load F of the load waveform Wb of the current inspection apparatus T is counted, Get the frequency distribution. For example, the correlation coefficient calculator 3a3 determines 20 sections of the same width for the load, and counts the number of data to which the data of the load waveform Wb belongs in each of the 20 sections, as shown in FIG. . Then, as shown in FIG. 15, for each inspection device T, a frequency distribution, which is the appearance frequency of the load F of the inspection device T in each section, is obtained. The frequency of each interval is treated as the number of data with the median value of that interval. For example, if the median value in the third interval in FIG. It is treated as having 1250 data of 75 kN. Further, the correlation coefficient calculator 3a3 also performs the same processing as the above-described load waveform Wb on the load F of the average waveform WB of the past normal inspection apparatus T, and calculates the number of data belonging to the same interval. are counted to obtain the frequency distribution of the average value of the load F of the past normal inspection apparatus T. FIG.

Then, the correlation coefficient calculator 3a3 obtains the correlation coefficient R between the frequency distribution of the average value of the load F of the inspection device T that was normal in the past and the frequency distribution of the load F of the current inspection device T. FIG. Specifically, in the frequency distribution of the average value of the load F of the normal inspection apparatus T in the past, the standard deviation of the data group P, which is a set of all sections of the data consisting of the median values of the number of frequencies for each section, is σ _P and On the other hand, in the frequency distribution of the load F of the inspection apparatus T this time, the standard deviation of the data group Q, which is a set of all sections of the data consisting of the median value of the number of frequencies for each section, is set to σ _Q , and the data group P and the data group Assuming that the covariance of Q is S _PQ , the correlation coefficient calculator 3a3 calculates the correlation coefficient R by calculating R=S _PQ /(σ _P ×σ _Q ). In order to obtain the covariance, the number of data in the data group P and the data group Q must be the same.

As described above, when obtaining the correlation coefficient R between the value of the average waveform WB and the load F of the current inspection device T, the phase of the average waveform WB and the load waveform Wb of the load F detected by the current inspection device T is out of phase. In this case, even if both waveforms match except for the phase, the value of the correlation coefficient R becomes small and the correlation becomes weak.

On the other hand, when the load is divided into categories and the frequency distribution of the load F is obtained, the phase is not reflected in the frequency distribution, but the average value of the load F of the past normal inspection apparatus T is When the average waveform WB and the load waveform Wb of the load F detected by the inspection device T this time match, the same frequency distribution is obtained. Therefore, when the correlation coefficient calculator 3a3 obtains the correlation coefficient R as described above, the average waveform WB, which is the average value of the load F of the normal inspection device T in the past, and the load F of the inspection device T are detected. Even if there is a phase shift with the load waveform Wb obtained this time, as long as the average waveform WB and the load waveform Wb have similar shapes, the value of the correlation coefficient R will take a value close to 1. . Therefore, the correlation coefficient calculation unit 3a3 obtains the frequency distribution of the average value of the load F of the past normal inspection device T and the frequency distribution of the load F detected this time by the inspection device T, as described above. If the correlation coefficient R of is obtained, the correct correlation coefficient R of both can be obtained ignoring the phase.

Further, the correlation coefficient calculation unit 3a3 processes the load F detected by the load cell 2a to obtain the correlation coefficient R related to the load, and calculates the speed V detected by the speed sensor 2b as the processing performed for the load F described above. A correlation coefficient R between the inspection devices T related to the speed is obtained by performing the same processing. Further, the correlation coefficient calculator 3a3 performs the same processing as that performed on the load F described above for the displacement X detected by the stroke sensor 2c, and obtains the correlation coefficient R between the inspection devices T related to the displacement.

Next, the offset amount calculator 3a4 will be described in detail. When the current command I that changes the load with a sine wave of a predetermined cycle is repeatedly applied from the controller C to the inspection device T and the load F is detected by the load cell 2a, the sensor bias and the disturbance of the vibration of the inspection device T and the damper D are detected. Due to the influence of non-linear damping force characteristics, as shown in FIG. 16, the load waveform Wb may be entirely offset in the positive load direction or the negative load direction. Then, when the offset amount O from the horizontal axis of the load waveform Wb, which is the output waveform, becomes large, it is considered that the inspection apparatus T cannot load the damper D as the specimen with a normal output, or that the sensor 2 has an abnormality. be done. Therefore, the processing device 3 is provided with an offset amount calculator 3a4 in order to obtain the offset amount O. As shown in FIG.

In the present embodiment, the offset amount calculator 3a4 shifts the load waveform Wb of the load F vertically in FIG. Look for O. That is, if the broken line L indicating the offset amount O in FIG. 16 is taken as the horizontal axis, the integral value on the positive side and the integral value on the negative side of the load waveform Wb match. The processing of the offset amount calculator 3a4 is the same as the processing of obtaining the offset component of the load waveform Wb by the distortion factor calculator 3a1. can be used by the distortion factor calculator 3a1. Alternatively, the median value of the maximum value and the minimum value may be obtained for each cycle of the load waveform Wb, and the average value of these median values may be used as the offset amount O.

The offset amount calculator 3a4 processes the load F of the past normal inspection apparatus T to obtain the offset amount O of the load waveform Wb of the past normal inspection apparatus T, and also processes the current load F of the inspection apparatus T. Then, the offset amount O of the load waveform Wb of the inspection apparatus T this time is obtained. Similarly, the offset amount calculation unit 3a4 processes the velocity V of the past normal inspection apparatus T to obtain the offset amount O of the velocity waveform of the past normal inspection apparatus T, and the current velocity V of the inspection apparatus T. is processed to obtain the offset amount O of the velocity waveform of the inspection apparatus T this time. Further, the offset amount calculator 3a4 processes the displacement X of the past normal inspection apparatus T to obtain the offset amount O of the displacement waveform of the past normal inspection apparatus T, and calculates the current displacement X of the inspection apparatus T. After processing, the offset amount O of the displacement waveform of the inspection apparatus T this time is obtained.

The offset amount O is a value that serves as a measure for the inspection device T to grasp changes and deviations in the output. In other words, the offset amount O is an index for grasping an abnormality in the output of the inspection device T or an abnormality in the sensor 2. As the value increases, the deviation in the output of the inspection device T becomes large and normal output cannot be exhibited. . Therefore, if the offset amount O is selected as the feature amount of the inspection device T for detecting an abnormality, an abnormality in the output of the inspection device T or an abnormality in the sensor 2 can be detected.

Next, the stiffness calculator 3a5 will be described in detail. When the actuator 13 is driven to apply a load to the damper D as a specimen, the load F and the displacement X of the inspection device T are detected, and from the load F and the displacement X simultaneously detected by the load cell 2a and the stroke sensor 2c, The stiffness K of the inspection device T can be obtained. Specifically, when a current command I that changes the load with a sine wave of a predetermined cycle is repeatedly given from the controller C to the inspection device T, and a load F is applied to the damper D, the load is taken on the vertical axis and the displacement is taken on the horizontal axis. When the load F and the displacement X detected simultaneously are plotted, a graph in which the trajectory of the intersection of the load F and the displacement X draws an ellipse as shown in FIG. 17 is obtained because the damper D has hysteresis.

When the slope of the straight line L connecting the point F _MAX of the maximum value of the load F and the point F _MIN of the minimum value of the load F in this graph is obtained, the slope is the difference between the maximum value and the minimum value of the load F at the point F _MAX is a value obtained by dividing by the difference between the displacement of the horizontal axis of F MIN and the displacement of the horizontal axis of the point F _MIN . Therefore, the stiffness calculator 3a5 calculates the difference ΔF between the maximum value of the load F and the difference between the displacement X when the load F reaches its maximum value and the displacement X when the load F reaches its minimum value. ΔX is obtained, and the stiffness K is obtained by calculating K=ΔF/ΔX.

The stiffness calculator 3a5 processes the load F and the displacement X of the past normal inspection device T to obtain the stiffness K of the past normal inspection device T, and obtains the stiffness K of the current inspection device T.

Factors that cause the stiffness K to change include deterioration of the inspection device T, occurrence of slack or backlash in parts of the inspection device T, and mounting abnormality or deterioration of the sensor 2 . Therefore, the stiffness K is an index that serves as a scale for grasping the abnormality and deterioration of the parts of the inspection device T and the abnormality and deterioration of the sensor 2 . Therefore, if the stiffness K is selected as the feature quantity of the inspection device T for detecting an abnormality, it is possible to detect the abnormality or deterioration of the parts in the inspection device T and the abnormality of the sensor 2 .

In calculating the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K, the data of the load F, the velocity V, and the displacement X, which are referred to as outputs, may be the same data. The data of the load F, the velocity V, and the displacement X need not be collected separately for calculating the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount, and the stiffness K. Therefore, the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K are calculated from the load F, velocity V, and displacement X of each inspection device T obtained by repeatedly giving the current command I from the controller C. do it.

Next, the determination section 3a6 will be described in detail. The distortion factor calculation unit 3a1 obtains the distortion factor ε, the variation degree calculation unit 3a2 obtains the variation degree Sd, the correlation coefficient calculation unit 3a3 obtains the correlation coefficient R, the offset amount calculation unit 3a4 obtains the offset amount O, Furthermore, when the stiffness calculation unit 3a5 obtains the stiffness K, the determination unit 3a6 uses the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K to determine whether or not there is an abnormality in the inspection device T. do.

First, the abnormality determination processing of the inspection device T using the distortion factor ε will be described. Let _εG be the distortion factor obtained from the past output of the normal inspection device T, and let _εT be the distortion factor obtained from the current output of the inspection device T when an input for detecting an abnormality is given. The determination unit 3a6 obtains an average value ε _GA and a standard deviation σε _G of a plurality of strain factors ε _G obtained from a plurality of outputs of the past normal inspection apparatus T obtained by the strain calculation unit 3a1. Then, the determination unit 3a6 sets a value three times the standard deviation σε _G as the threshold 3σε _G , and subtracts the distortion factor ε _T obtained this time from the average value ε _GA to determine the absolute value |Δε| of the value Δε as the threshold 3σε _G If the strain rate ε is less than or equal to the strain rate _ε , the inspection apparatus _T is judged to be normal, and if the absolute value | _Δε | The inspection device T is determined to be abnormal with respect to ε. Here, the distortion rate ε _G can be regarded as a sample extracted from a population of distortion rate data of normal inspection devices T. Assuming that the population follows a normal distribution, if the number of samples is sufficiently large, the standard deviation σε _G obtained as described above can be regarded as the standard deviation of the population.

If the population follows a normal distribution, then there is a 99.73% probability that the population data lies within the range ε _GA ±3σε _G. As described above, the strain factor ε _G of a normal inspection apparatus T is a value within the range of ε _GA ±3σε _G even if there is variation. The rate ε _T should likewise take a value within the range of ε _GA ±3σε _G with a probability of 99.73%. The distortion factor _εG was originally obtained from the output of a normal inspection apparatus T, and all of _them are normal values even if there are _{variations .} If it is within the range, it may be considered that the inspection apparatus T is normal in the current abnormality detection process. Therefore, in the present embodiment, the threshold value is set to a positive value 3σε _G , and the determination unit 3a6 subtracts the distortion factor ε _T obtained this time from the average value ε _GA to determine the absolute value |Δε| exceeds the threshold value _3σεG , the inspection device T judges whether or not there is an abnormality in the distortion factor ε. In this way, the abnormality detection system 1 can accurately determine whether the inspection apparatus T has an abnormality. Since the distortion factor ε is obtained not only for the load F but also for the velocity V and the displacement X, the judging section 3a6 also performs the same processing as described above for the distortion factor ε of the velocity V and the displacement X to judge abnormality. When the determination unit 3a6 performs abnormality determination processing using the distortion factor ε in this manner, an abnormality can be detected from the viewpoint of the dynamic characteristics of the inspection apparatus T. FIG.

Note that the judgment unit 3a6 obtains the variance σε _G ² of a plurality of distortion factors ε _G and sets the threshold value 3σε _G ² as the square value Δε obtained by subtracting the distortion factor ε _T obtained this time from the average value ε _GA ^. is equal to or less than the threshold value 3σε _G ² , the inspection device T may be determined to be normal with respect to the distortion factor ε _T , and when the squared value Δε ² exceeds the threshold value 3σε _G ² , the inspection device T may be determined to be abnormal with respect to the distortion factor ε. .

Note that the threshold used for abnormality determination is set to a value that is three times the standard deviation _σεG of the past distortion factor _εG of the output of the inspection apparatus T that was normal in the past. , the threshold may be set to any value less than _3σεG . For example, if the threshold is set to a value twice the standard deviation _σεG , if the inspection apparatus T is normal, the distortion factor _εT obtained this time is also within the range of _εGA ± _2σεG with a probability of 95.45%. Since it should be a value in between, the threshold may be set in this manner.

Next, abnormality determination processing of the inspection device T using the degree of variation Sd, the offset amount O, and the stiffness K will be described. The judging unit 3a6 performs the same processing as the abnormality judgment of the inspection device T for the distortion factor ε for the abnormality judgment of the inspection device T using the variation degree Sd, the offset amount O, and the stiffness K. That is, the determination unit 3a6 obtains the average value and standard deviation of the variation degree Sd, the offset amount O, and the stiffness K obtained from the output of the normal inspection apparatus T in the past, and uses a value three times the standard deviation as a threshold value. , the degree of variation Sd, the offset amount O and the stiffness K obtained from the current output of the inspection device T, and the absolute values of the differences from the corresponding average values are equal to or less than the corresponding threshold values, the inspection device T is determined to be normal. , the inspection apparatus T is determined to be abnormal when the absolute value exceeds the threshold.

That is, it processes as follows. Regarding the variation degree Sd, the variation degree obtained from the past output of the normal inspection device T is Sd _G , the average value and standard deviation of the variation degree Sd _G are Sd _GA and σSd _G respectively, and the current output of the inspection device T is Let _SdT be the degree of dispersion obtained from . _Determining unit 3a6 sets a value three times the standard deviation σSd _G as threshold 3σSd _G , and _determines whether the absolute value |ΔSd _| When the absolute value |ΔSd _| of the value ΔSd obtained by subtracting the degree of variation Sd T obtained this time from the average value Sd _GA exceeds the threshold value 3σSd _G , the degree of variation Sd The inspection device T is determined to be abnormal. When the determination unit 3a6 performs abnormality determination processing using the degree of variation Sd in this manner, an abnormality can be detected from the viewpoint of the repeatability of the inspection apparatus T. FIG.

Regarding the offset amount O, the offset amount obtained from the past output of the normal inspection apparatus T is _OG , the average value and standard deviation of the offset amount _OG are _OGA and _σOG , respectively, and the current output of the inspection apparatus T is Let the offset amount obtained from be _OT . Determining unit 3a6 determines that the absolute value |ΔO| of the value ΔO obtained by subtracting the offset _amount OT obtained this time from the average value _OGA is equal to or less than the threshold _3σOG , with a threshold _3σOG that is three times the standard deviation _σOG If there is, the inspection apparatus _T is judged to be normal with respect to the offset amount _O , and if the absolute _value |ΔO| Judge the inspection device T as abnormal with respect to the quantity O. When the determination unit 3a6 performs abnormality determination processing using the offset amount O in this manner, an abnormality in the output of the inspection device T or an abnormality in the sensor 2 can be detected.

Regarding the stiffness K, the stiffness obtained from the past output of the normal inspection device T is K _G , the average value and standard deviation of the stiffness K _G are K _GA and σK _G , respectively, and the stiffness obtained from the current output of the inspection device T is Let _KT be the stiffness obtained. Determination unit 3a6 _sets a value three times the standard deviation σK _G as a _threshold 3σK _G , and the absolute value | _ΔK | If it is less than or equal to the stiffness K, the inspection device _T is _judged to be normal, and the stiffness _K is inspected when the absolute value |ΔK| Device T is judged to be abnormal. When the determination unit 3a6 performs abnormality determination processing using the rigidity K in this way, an abnormality or deterioration of the parts in the inspection device T and an abnormality of the sensor 2 can be detected.

Since the degree of variation Sd, the offset amount O, and the stiffness K are obtained for the velocity V and the displacement X in addition to the load F, the determination unit 3a6 also determines the degree of variation Sd of the velocity V and the displacement X, the offset amount O, and the stiffness K. Anomaly determination is performed by performing the same processing as described above.

Note that the determination unit 3a6 sets the threshold value to three times the variance in the abnormality determination process of the inspection apparatus T using the variation degree Sd, the offset amount O, and the stiffness K, similarly to the abnormality determination process using the strain rate ε. Whether or not the inspection device T is abnormal may be determined based on whether or not the square values ΔSd ² , ΔO ² , and ΔK ² exceed the threshold values. The thresholds used in the abnormality determination for the degree of variation Sd, the offset amount O, and the stiffness K may be set to values smaller than three times the standard deviation, similarly to the setting of the thresholds for the distortion rate ε. In addition, in the abnormality determination process of the inspection device T using the distortion rate ε, the degree of variation Sd, the offset amount O, and the stiffness K, when it is assumed that the population follows another probability distribution such as a chi-square distribution, if The threshold may be determined by obtaining the standard deviation and variance in the assumed probability distribution.

Finally, the abnormality determination process of the inspection device T using the correlation coefficient R will be described. The correlation coefficient R takes a value from -1 to 1. If the average waveform WB, which is the average value of the output of the normal inspection device T in the past, and the current output waveform of the inspection device T match well, The higher it is, the closer it is to 1. When the correlation coefficient R is 1, the average waveform WB and the output waveform completely match, and when the correlation coefficient R is 0, there is no correlation between the average waveform WB and the output waveform, When the correlation coefficient R is -1, the average waveform WB and the output waveform are oriented in completely opposite directions, resulting in a negative correlation between the two. When the divergence between the average waveform WB and the output waveform becomes large, it can be determined that there is an abnormality in the inspection apparatus T and the output instructed by the input cannot be produced.

Therefore, the determination unit 3a6 compares the correlation coefficient R with the threshold value α, determines that the inspection device T is normal when the correlation coefficient R is equal to or less than the threshold value α, and determines that the correlation coefficient R exceeds the threshold value α. and the inspection device T is determined to be abnormal. If the determination unit 3a6 performs abnormality determination processing using the correlation coefficient R in this way, an abnormality in the inspection device T can be detected from the viewpoint of the degree of waveform matching.

The threshold value α set for the correlation coefficient R is such that the correlation between the average output of the past normal inspection device T and the current output of the inspection device T is weak, and if the value of the correlation coefficient R becomes smaller than this, the average It is set to a value that is considered to affect the inspection result of the damper D, which is the specimen, due to the difference between the waveform WB and the output waveform of the inspection apparatus T this time.

The correlation coefficient R is for determining the correlation between the average value of the past output of the inspection device T and the newly obtained current output of the inspection device T, and the correlation coefficient R also indicates the characteristics of the inspection device T. is an indicator. In this embodiment, since only one correlation coefficient R can be obtained, the judgment unit 3a6 judges whether or not the value of the correlation coefficient R is a normal value as described above. If the correlation coefficient R is to be treated as a feature amount, the correlation coefficient R between the output and the ideal waveform Wa is obtained for each of a plurality of past outputs of the inspection device T. At the same time, the additive coefficient R between the current output of the inspection apparatus T and the ideal waveform Wa is obtained, and using these correlation coefficients R as feature amounts, the average value of the correlation coefficients R related to the past outputs and the current output are calculated. If the absolute value of the difference between the correlation coefficient R and the correlation coefficient R is equal to or less than a threshold value, the inspection device T may be determined to be normal, and if not, it may be determined to be abnormal. In this case, the threshold may be set in the same manner as the thresholds for other feature quantities using the standard deviation or variance of the correlation coefficient R relating to past outputs.

As described above, in this embodiment, when detecting an abnormality in the inspection apparatus T, the average value and the standard deviation or variance of a plurality of past outputs of the normal inspection apparatus T are obtained. The accuracy of the abnormality detection of the inspection apparatus T improves as the number increases. Since the average value of the past outputs is required to calculate the correlation coefficient R, even in the abnormality detection using the correlation coefficient R, the larger the number of samples of the past normal outputs, the more accurately the inspection apparatus can be used. The accuracy of T anomaly detection is improved. In order to obtain the past normal output, the abnormality detection process should be performed for the required number of samples in advance on the normal inspection apparatus T, and the output should be detected each time to secure the number of output samples. In addition, even if the number of past output samples is not secured, if the output data collected when anomaly detection is performed and judged to be normal can be accumulated as past normal output data, anomalies can be generated. The accuracy of detection will improve.

The determination unit 3a6 performs the abnormality detection process as described above, and determines the presence or absence of an abnormality in the inspection device T using the five indexes of the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the rigidity K. are independently determined, and if any one of the determinations using the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K is abnormal, the other indicators are normal. Even if it is determined, the inspection apparatus T is determined to be abnormal.

Then, the processing device 3 displays the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K on the display device 3e. Which index among the degree Sd, the correlation coefficient R, the offset amount O, and the stiffness K is determined to be abnormal is displayed on the display device 3e in a manner that the operator of the processing device 3 can visually recognize. For example, if the distortion factor ε indicates an abnormality in the inspection apparatus T, the processing device 3 displays the distortion factor ε by displaying "the distortion factor ε is an abnormal value" on the display device 3e or by displaying the distortion factor ε in a color that attracts the operator's attention. is displayed.

In response to a request from the operator or triggered by the completion of calculation of the distortion rate ε, the degree of variation Sd and the correlation coefficient R, the processing device 3 outputs the distortion rate ε, the degree of variation Sd and the correlation coefficient from the printer 3g to the paper medium. The value of the relational coefficient R may be printed. At that time, data observed by the load cell 2a, the speed sensor 2b, and the stroke sensor 2c to obtain the distortion rate ε, the degree of variation Sd, and the correlation coefficient R, a graph plotting these data, and further, the distortion rate ε, the variation The values obtained in the calculation process of the degree Sd, the correlation coefficient R, the offset amount O, and the stiffness K may also be printed. In addition, data observed by the load cell 2a, the speed sensor 2b, and the stroke sensor 2c to obtain the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K, a graph plotting these data, and The values obtained in the calculation process of the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K may be displayed on the display device 3e.

The abnormality determination processing of the abnormality detection system 1 of the inspection apparatus T described above will be described with reference to the flowchart shown in FIG. A damper D, which is a sample, is attached to the inspection device T, and a current command I is repeatedly input from the controller C to apply a load, velocity, and displacement to the damper D to drive the actuator 13, and the load as the output of the inspection device T, The speed and displacement are changed (step ST1), the load F, speed V and displacement X output from the inspection device T are detected by the load cell 2a, the speed sensor 2b and the stroke sensor 2c (step ST2). is input to the arithmetic processing unit 3a.

The processing device 3 obtains the distortion factor ε from the data of the past load F, velocity V and displacement X of the normal inspection device T and the data of the current load F, velocity V and displacement X of the same inspection device T (step ST3). . Further, the processing device 3 obtains the variation degree Sd from the data of the past load F, velocity V and displacement X of the normal inspection device T and the data of the current load F, velocity V and displacement X of the same inspection device T (step ST4). Furthermore, the processing device 3 obtains the correlation coefficient R from the data of the past load F, velocity V and displacement X of the normal inspection device T and the data of the current load F, velocity V and displacement X of the same inspection device T ( step ST5). Subsequently, the processing device 3 obtains the offset amount O from the data of the past load F, velocity V and displacement X of the normal inspection device T and the data of the current load F, velocity V and displacement X of the same inspection device T ( step ST6). Further, the processing device 3 obtains the stiffness K from the data of the past load F, velocity V and displacement X of the normal inspection device T and the data of the current load F, velocity V and displacement X of the same inspection device T (step ST7 ). The processing device 3 can arbitrarily change the order in which the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K are obtained.

After obtaining the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K, the processing device 3 calculates the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness Based on K, the process of judging whether or not there is an abnormality in the inspection apparatus T is performed as described above (step FT8).

After performing the process of step ST8, the processing device 3 displays the obtained distortion factor ε, degree of variation Sd, correlation coefficient R, offset amount O, and stiffness K on the screen of the display device 3e (step ST9). At that time, if the processing device 3 determines that the inspection device T is abnormal, the processing device 3 provides information specifying the values determined to be abnormal among the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K. It is displayed on the display device 3e.

Since the abnormality detection system 1 of the inspection apparatus T operates as described above and detects an abnormality of the inspection apparatus T, the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K are used as indices. determines whether or not there is an abnormality in the inspection apparatus T, displays these indicators on the display device 3e and, if there is an abnormality, information specifying the indicator that is abnormal among these indicators, or The printer 3g prints on a paper medium so that the operator can visually recognize these indices and the like.

In the above-described embodiment, the output of the inspection apparatus T is the load, velocity, and displacement. , correlation coefficient R, and offset amount O may be obtained. In the present embodiment, since the output of the inspection device T is used as a plurality of parameters of load, speed, and displacement, the presence or absence of an abnormality in the inspection device T can be determined more accurately. Further, in the present embodiment, since the inspection device T is a vibration inspection device, acceleration may be added to the output.

Furthermore, in the above-described embodiment, the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K are obtained from the output of the inspection apparatus T with the same input when the damper D serving as a reference is inspected as a specimen. However, for the distortion factor ε, the degree of variation Sd, the correlation coefficient R, and the offset amount O, the same input is applied to the inspection device T without a sample attached and without a load. It may be obtained from the output of the inspection device T when given. In this case, the load cannot be detected because the specimen is not attached to the inspection device T, but since the actuator 13 expands and contracts, the velocity and displacement can be detected. Further, in the inspection apparatus T according to the present embodiment, the specimen used as a reference for abnormality detection may be an elastic body or the like in addition to the damper D.

The abnormality detection system 1 of the inspection apparatus T determines whether or not there is an abnormality in the inspection apparatus T as described above. .

The inspection apparatus T is provided for inspecting samples, and inspects a large number of samples such that when the inspection of one sample is completed, the next sample is inspected. If the inspection device T is normal, it can output according to the conditions specified by the inspection conditions and apply a load to the specimen. The test result of the sample by the test apparatus T is provided to the test operator who determines whether the sample is acceptable. For example, if the specimen is a damper, the value of the damping coefficient and the maximum load obtained from the damper's load and speed may be used as a pass/fail decision. There is no guarantee that there will be products with disordered characteristics and internal abnormalities.

Therefore, in the abnormality detection system 1 of the present embodiment, the abnormality detection processing of the inspection apparatus T is performed for the output during the inspection of the sample performed by the inspection apparatus T judged to be normal so that the presence or absence of an abnormality in the sample can be determined. A sample abnormality determination process is executed to determine an abnormality of the sample by performing the same processing as in . In this embodiment, the specimen is a damper, and the abnormality detection system 1 determines abnormality of the damper.

Specifically, the abnormality detection system 1 gives the same input as in the abnormality detection process to the inspection apparatus T in the same manner as the abnormality detection of the inspection apparatus T, and outputs the output of the inspection apparatus T during inspection of the sample as a load F and a speed As V and displacement X, these are detected by the sensor 2, and the distortion factor ε, the degree of variation Sd, the offset amount O and the stiffness K are obtained as feature quantities from the output obtained during the inspection, and the correlation coefficient R is obtained. In addition, the abnormality detection system 1 detects the distortion factor ε, the degree of variation Sd, the offset amount O, and the stiffness from a plurality of outputs of the inspection apparatus T in the past when a normal sample that is the same product as the newly inspected sample is inspected. K is obtained as a feature amount, and a correlation coefficient R is obtained. In this embodiment, the output of the inspection device T detected by the sensor 2 is the load F, the velocity V and the displacement X. F, velocity V and displacement X are determined respectively.

Therefore, the distortion factor calculation unit 3a1, the variation degree calculation unit 3a2, the correlation degree calculation unit 3a3, the offset amount calculation unit 3a4, and the stiffness calculation unit 3a5 in the processing device 3 each perform the above-described processing, and From the output of the inspection device T this time, the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O and the stiffness K are obtained. When the same product is used as a sample, the average value and standard deviation of multiple outputs obtained in the past inspection of the same product sample were determined to be normal, and the result of the current sample inspection is If the absolute value of the difference between the obtained output and the average value is three times the standard deviation or less, it can be determined that the sample is an acceptable product without any abnormalities. In other words, in the sample abnormality determination processing, the data obtained when the sample of the product whose output from the inspection device T in the past and the output obtained from the current sample inspection of the inspection device T are the same are used, Furthermore, as for the past output of the inspection device T, only the data in which the sample was determined to be normal is used.

Further, if the input given by the controller C to the inspection apparatus T in the inspection is a command to increase or decrease the load F, velocity V, and displacement X of the inspection apparatus T, the distortion factor ε, the degree of variation Sd, the correlation coefficient R, The offset amount O and stiffness K can be determined. For this reason, it is not necessary to provide the same input as that used for the abnormality detection of the inspection apparatus T. Instead, it is necessary to provide the input used for the inspection of the sample to be inspected, and the above-mentioned feature quantity and A correlation coefficient R can be obtained.

A large number of past outputs of the inspection apparatus T are prepared in advance for the determination of specimen abnormality, and the distortion factor ε, the degree of variation Sd, the offset amount O, and the stiffness K are obtained for each of these large numbers of outputs. If an input is given to the inspection apparatus T for sample abnormality determination, and the sample inspected this time is determined to be normal, the current output of the inspection apparatus T is used as the past output in the next sample abnormality determination process. , and past output data may be accumulated.

In the sample abnormality determination process described above, the data to be handled is the output of the inspection device T detected in the inspection of the sample by the inspection device T in the past, and the output judged to be normal for the sample and the sample of the same product It is different from the abnormality detection process of the inspection device T in that the output of the inspection device T detected when the is newly inspected is handled. The calculation process for obtaining the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K is exactly the same as the abnormality detection processing of the inspection apparatus T, so the sample abnormality determination processing will be briefly described below. .

In the sample abnormality determination process, the distortion factor calculation unit 3a1 sets the waveform of the output of the inspection apparatus T instructed by the same input as in the abnormality detection process as an ideal waveform, and Fourier-transforms this ideal waveform to obtain the power spectral density of the ideal waveform. At the same time, the power spectral density of the output waveform is obtained by Fourier transforming the output waveform, which is the waveform of the output actually output by the inspection apparatus T, in response to the same input as in the abnormality detection process. Then, the distortion factor calculation unit 3a1 obtains the power spectrum by integrating the power spectrum density of the ideal waveform on the frequency axis from the integrated value of the power spectrum density obtained by integrating the power spectrum density of the output waveform on the frequency axis. By subtracting the integrated value of the density, the distortion component which is the difference between the two is obtained, and the value obtained by dividing this distortion component by the integrated value of the power spectrum density of the output waveform is obtained as the distortion factor ε. In the sample abnormality determination process, the distortion factor ε is obtained for each of the past outputs of the inspection apparatus T obtained when inspecting a normal sample of the same product, and the inspection detected when the sample of the same product is newly inspected. A distortion factor ε is obtained for the current output of the device T. In this embodiment, the output of the inspection apparatus T is the load F, the velocity V, and the displacement X, and the distortion factor ε is obtained for the load F, the velocity V, and the displacement X, respectively. Further, in calculating the distortion factor ε, the distortion factor calculating section 3a1 performs a process of removing the drift component of the output waveform, and a process of cutting off the end side when the initial value and the final value of the observed output waveform are different. you can go As described above, the distortion factor ε is a value that serves as a measure for grasping how much the output waveform is distorted with respect to the ideal waveform. The larger the distortion factor ε, the more disturbed the behavior of the specimen when inspecting the specimen and the more the output fluctuates. It is one index. Therefore, if the distortion factor ε is selected as the feature quantity of the inspection apparatus T for detecting specimen abnormality, abnormality can be detected from the viewpoint of the dynamic characteristics of the specimen.

In the sample abnormality determination process, the degree-of-variation calculator 3a2 obtains the variance of the same-phase values of each unit waveform in the output waveform of the inspection device T for the number of data included in the unit waveform, and calculates the average value of these variances by 3. The multiplied value is obtained as the degree of variation Sd. In the sample abnormality determination process, the degree of variation Sd is obtained for each of the past outputs of the inspection device T obtained when inspecting a normal sample of the same product, and the inspection detected when the sample of the same product is newly inspected. A variation degree Sd is obtained for the output of the apparatus T this time. In this embodiment, the output of the inspection device T is the load F, the velocity V and the displacement X, and the variation degree Sd is obtained for the load F, the velocity V and the displacement X respectively.

When the output of the inspection device T is repeatedly increased and decreased during inspection of the sample, the load of the inspection device T repeatedly increases and decreases, and the degree of variation Sd corresponds to the load indicated by the current command I. It is a value that serves as a measure for grasping whether or not In other words, the degree of variation Sd is an index for evaluating the dynamic characteristics of a specimen from the viewpoint of repeatability, i.e., whether the specimen repeats the same reaction when the output of the inspection device T is repeatedly increased and decreased. It shows that the behavior of the specimen is disturbed and the repeatability is deteriorated. Therefore, if the degree of variation Sd is selected as the characteristic amount of the inspection apparatus T for detecting specimen abnormality, abnormality can be detected from the viewpoint of the repeatability of the specimen.

In the specimen abnormality determination process, the correlation coefficient calculation unit 3a3 calculates the average value of a plurality of past outputs of the inspection apparatus T obtained when inspecting normal specimens of the same product, and newly inspecting the specimen of the same product. A correlation coefficient R between the current output of the inspection device T detected at the time of detection is obtained. More specifically, similar to the abnormality detection process, the correlation coefficient calculator 3a3 calculates the frequency distribution of the load in the average waveform, which is the data group of the average values, and the frequency distribution of the load in the current output waveform. Then, the covariance between the frequency distributions is obtained, and this covariance is defined as the correlation coefficient R. In calculating the correlation coefficient R, the controller C repeatedly gives the current command I to the inspection apparatus T as an input similar to that used during the abnormality detection process, and obtains an output waveform containing a sufficient number of unit waveforms. In this embodiment, the output of the inspection device T is the load F, the velocity V, and the displacement X, and the correlation coefficient R is obtained for the load F, the velocity V, and the displacement X, respectively.

The correlation coefficient R is a scale indicating whether the average value of the past outputs of the normal inspection apparatus T that inspected normal samples and the current output of the inspection apparatus T are correlated with each other in the sample abnormality determination process. value, and the closer the value is to 1, the higher the correlation between the average value of past outputs of the normal inspection apparatus T that inspected normal specimens and the current output of the inspection apparatus T. Therefore, if the correlation coefficient R is obtained for specimen abnormality detection, the specimen abnormality can be detected from the viewpoint of the degree of waveform matching.

Furthermore, in the sample abnormality determination process, the offset amount calculation unit 3a4 newly inspects the sample of the same product with the output waveforms of the past outputs of the inspection apparatus T obtained when inspecting the normal sample of the same product. An offset amount O is obtained for each of the output waveforms in the current output of the inspection device T detected at the time. In this embodiment, the output of the inspection device T is the load F, the velocity V, and the displacement X, and the offset amount O is obtained for the load F, the velocity V, and the displacement X, respectively. The offset amount O is a value that serves as a measure for grasping the bias of the damping force of the damper as an inspection. In other words, the offset amount O is an index capable of grasping an abnormality in the damping force of the damper as an inspection in the sample abnormality determination process. Therefore, if the offset amount O is selected as the feature amount of the inspection apparatus T for detecting specimen abnormality, it is possible to detect specimen abnormality.

Then, in the sample abnormality determination process, the stiffness calculator 3a5 calculates a plurality of stiffnesses K from a plurality of loads and a plurality of displacements corresponding to the past loads of the inspection apparatus T obtained during inspection of normal samples of the same product. demand. Further, the stiffness calculator 3a5 obtains the stiffness K from the current load of the inspection apparatus T detected when the sample of the same product is newly inspected and the displacement corresponding to the load. The stiffness K assumes a value different from the normal value when there is a problem with the attachment of the damper as the specimen to the inspection apparatus T or when there is some abnormality in the damper as the specimen in the specimen abnormality determination process. Therefore, if the stiffness K is selected as the characteristic amount of the inspection apparatus T for the specimen abnormality determination process, it is possible to detect the mounting failure of the specimen to the inspection apparatus or the abnormality of the specimen.

In the sample abnormality determination process, as in the abnormality determination process of the inspection apparatus T, the strain factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K are calculated. The data of the velocity V and the displacement X may be the same data. Each data need not be collected separately. Therefore, the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K are calculated from the load F, velocity V, and displacement X of each inspection device T obtained by repeatedly giving the current command I from the controller C. do it.

Next, the judgment section 3a6 will be explained. The distortion factor calculation unit 3a1 obtains the distortion factor ε, the variation degree calculation unit 3a2 obtains the variation degree Sd, the correlation coefficient calculation unit 3a3 obtains the correlation coefficient R, the offset amount calculation unit 3a4 obtains the offset amount, and When the stiffness calculation unit 3a5 obtains the stiffness K, the determination unit 3a6 uses the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K to determine whether or not there is an abnormality in the inspection device T. A process similar to the abnormality determination process is performed to determine whether or not there is an abnormality in the specimen.

First, the specimen abnormality determination processing using the distortion factor ε will be described. The determination unit 3a6 obtains the average value ε _GA and the standard deviation σε _G of the distortion factor ε obtained from the output of inspecting a normal specimen of the same product with a normal inspection apparatus T in the past, and obtains a value that is three times the standard deviation. is the _{threshold 3σεG} , and the absolute value |Δε| If _{the absolute value |Δε| exceeds the threshold value 3σεG ,} the sample is determined to be abnormal. The distortion factor ε obtained from the output when a normal sample is tested by a normal testing device T has a probability of 99.73% even if there is variation, and is within a range of three times the standard deviation centered on the average value. , it can be considered that the sample is normal if the distortion factor ε obtained from the output this time falls within the above range. _Therefore , as described above, the determination _unit 3a6 _determines whether or not the absolute value |Δε| The presence or absence of abnormalities in specimens is determined for In this way, the abnormality detection system 1 can accurately determine the presence or absence of abnormality in the sample. Since the strain rate ε can be obtained not only for the load F but also for the velocity V and the displacement X, the judging section 3a6 performs the same processing as described above for the strain rate ε of the velocity V and the displacement X, and judges the abnormality of the specimen. . In addition, by using the strain factor ε to determine anomaly of the specimen in this manner, the anomaly detection system 1 can detect an anomaly from the viewpoint of the dynamic characteristics of the specimen.

Note that the judgment unit 3a6 obtains the variance σε _G ² of a plurality of distortion factors ε _G and sets the threshold value 3σε _G ² as the square value Δε obtained by subtracting the distortion factor ε _T obtained this time from the average value ε _GA ^. is less than or equal to the threshold value 3σε _G ² , the specimen may be determined to be normal with respect to the distortion factor ε _T , and when the squared value Δε ² exceeds the threshold value 3σε _G ² , the specimen may be determined to be abnormal with respect to the distortion factor ε. By setting the threshold value in this manner, the range of possible values of the feature quantity of a normal sample can be defined appropriately, so that the abnormality of the inspection device T can be detected accurately. In addition, as described above, the threshold used for specimen abnormality determination is a value three times the standard deviation σε _G of the output distortion factor ε _G when a normal specimen was inspected by a normal inspection apparatus T in the past. Although it is set, the threshold may be set to any value smaller than _3σεG to make the abnormality determination more severe. For example, if the threshold is set to a value twice the standard deviation σε _G , then if the specimen is normal, the distortion factor ε _T obtained this time is also within the range of ε _GA ±2σε _G with a probability of 95.45%. Since it should be a certain value, the threshold may be set in this manner.

Next, abnormality determination processing of the inspection device T using the degree of variation Sd, the offset amount O, and the stiffness K will be described. The judging unit 3a6 performs the same processing as the abnormality judgment of the specimen with respect to the strain factor ε for the abnormality judgment of the inspection apparatus T using the variation degree Sd, the offset amount O, and the stiffness K. In other words, the determination unit 3a6 obtains the average value and standard deviation of the degree of variation Sd, the offset amount O, and the stiffness K obtained from the output when the normal test device T tested a normal sample in the past, and the standard deviation Using a value three times as a threshold, the variation degree Sd, the offset amount O, and the stiffness K obtained from the current output of a normal inspection apparatus T detected when a new sample of the same product is inspected correspond to If the absolute value of the difference from the average value is equal to or less than the corresponding threshold value, the sample is determined to be normal, and if the absolute value exceeds the threshold value, the sample is determined to be abnormal.

That is, it processes as follows. Regarding the degree of variation Sd, the determination unit 3a6 obtains the average value Sd _GA and the standard deviation σSd _G of the degree of variation Sd obtained from the outputs of past inspections of normal samples of the same product by a normal inspection device T, and calculates the standard deviation. 3 times the threshold 3σSd _G , and the absolute value of the difference ΔSd between the degree of variation Sd _T obtained from the current output of the inspection device T detected when the sample of the same product is newly inspected and the average value Sd _GA If the value |ΔSd| is equal to or less than the threshold _3σSdG , the specimen is determined to be normal, and if the absolute value |ΔSd| exceeds the threshold _3σSdG , the specimen is determined to be abnormal. In addition, by using the degree of variation Sd to determine abnormality of a sample in this way, the abnormality detection system 1 can detect an abnormality from the viewpoint of the repeatability of the sample.

Regarding the offset amount O, the determination unit 3a6 obtains the average value O _GA and the standard deviation σO _G of the offset amount O obtained from the output of inspecting the same normal product specimen with the normal inspection apparatus T in the past, and calculates the standard deviation , and the absolute _{value of the difference ΔO between the offset amount OT and} the average value _OGA obtained from the current output of the inspection device T detected when the sample of the same product is newly inspected. If the value |ΔO| is equal to or less than the threshold _3σOG , the sample is determined to be normal, and if the absolute value |ΔO| exceeds the threshold _3σOG , the sample is determined to be abnormal. In addition, by using the offset amount O to determine anomaly of the sample in this way, the anomaly detection system 1 can detect an anomaly of the sample.

Regarding the stiffness K, the judging unit 3a6 obtains the average value K _GA and the standard deviation σK _G of the stiffness K obtained from the output of inspecting the same normal product specimen with the normal inspection device T in the past, and calculates the standard deviation 3 The absolute value of the difference ΔK between the stiffness _KT obtained from the current output of the inspection device _T and the average value K _GA |ΔK is equal to or less than the threshold 3σK _G , the sample is determined to be normal, and when the absolute value |ΔO| exceeds the threshold 3σK _G , the sample is determined to be abnormal. In addition, when the abnormality of the specimen is determined by using the stiffness K in this way, the abnormality detection system 1 can detect the mounting failure of the specimen to the inspection apparatus or the abnormality of the specimen.

Since the degree of variation Sd, the offset amount O, and the stiffness K are obtained for the velocity V and the displacement X in addition to the load F, the determination unit 3a6 also determines the degree of variation Sd of the velocity V and the displacement X, the offset amount O, and the stiffness K. The same processing as described above is performed to determine whether the sample is abnormal.

Note that the determination unit 3a6 sets the threshold value to a value three times the variance in the sample abnormality determination processing using the variation degree Sd, the offset amount O, and the stiffness K, as in the sample abnormality determination processing using the strain rate ε. , the presence or absence of an abnormality in the specimen may be determined based on whether or not the squared values ΔSd ² , ΔO ² , and ΔK ² exceed the threshold values. The thresholds used in the sample abnormality determination for the degree of variation Sd, the offset amount O, and the stiffness K may be set to values smaller than three times the standard deviation, similarly to the setting of the thresholds for the strain factor ε. In addition, in the abnormality determination process of the inspection device T using the distortion rate ε, the degree of variation Sd, the offset amount O, and the stiffness K, when it is assumed that the population follows another probability distribution such as a chi-square distribution, if The threshold may be determined by obtaining the standard deviation and variance in the assumed probability distribution.

Regarding the correlation coefficient R, the determination unit 3a6 compares the correlation coefficient R with a threshold value β, and determines that the sample is normal when the correlation coefficient R is equal to or less than the threshold value β. If the threshold value β is exceeded, the sample is determined to be abnormal. When the determination unit 3a6 performs abnormality determination processing using the correlation coefficient R in this manner, an abnormality in the specimen can be detected from the viewpoint of the degree of coincidence of waveforms.

The threshold value α set for the correlation coefficient R is such that the correlation between the average value of the output of the normal inspection apparatus T when normal specimens were inspected in the past and the current output of the inspection apparatus T is weak, and the correlation When the value of the number R becomes small, it is set to a value at which it can be determined that the specimen is abnormal.

As described above, in the present embodiment, the average value and the standard deviation or the variance of the output of the normal inspection apparatus T during the inspection of the normal sample in the past are obtained in detecting the abnormality of the sample. The greater the number, the higher the accuracy of specimen anomaly detection. Since the average value of the past outputs is required to calculate the correlation coefficient R, even in the abnormality detection using the correlation coefficient R, the larger the number of samples of the past normal outputs, the more accurately the inspection apparatus can be used. The accuracy of T anomaly detection is improved. In order to obtain a past normal output, the specimen abnormality detection process is performed for the required number of times in advance by the normal inspection apparatus T to inspect normal specimens, and the output is detected each time to determine the number of output samples. should be secured. In addition, even if the number of past output samples is not secured, if the output data collected when abnormal detection is performed and judged to be normal is accumulated as past normal output data, the sample The accuracy of anomaly detection will improve.

The determination unit 3a6 performs the sample abnormality detection process as described above, and uses the five indices of the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K to determine whether there is an abnormality in the sample. Judgments are made independently, and if any abnormality is found in the judgment using the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K, the other indices are judged to be normal. The specimen is judged to be abnormal even if it is In this way, the determination unit 3a6 performs sample listing detection processing and also functions as a sample abnormality determination unit.

Then, the processing device 3 causes the display device 3e to display the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the amount of offset O, and the stiffness K. , the correlation coefficient R, the offset amount O, and the stiffness K is displayed on the display device 3e in a manner that the operator of the processing device 3 can visually recognize. For example, when the distortion factor ε indicates an abnormality in the specimen, the processing device 3 displays the distortion factor ε in a color that attracts the operator's attention, such as "the distortion factor ε is an abnormal value" on the display device 3e. and so on.

In response to a request from the operator or triggered by the completion of calculation of the distortion rate ε, the degree of variation Sd and the correlation coefficient R, the processing device 3 outputs the distortion rate ε, the degree of variation Sd and the correlation coefficient from the printer 3g to the paper medium. The value of the relational coefficient R may be printed. At that time, data observed by the load cell 2a, the speed sensor 2b, and the stroke sensor 2c to obtain the distortion rate ε, the degree of variation Sd, and the correlation coefficient R, a graph plotting these data, and further, the distortion rate ε, the variation The values obtained in the calculation process of the degree Sd, the correlation coefficient R, the offset amount O, and the stiffness K may also be printed. In addition, data observed by the load cell 2a, the speed sensor 2b, and the stroke sensor 2c to obtain the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K, a graph plotting these data, and The values obtained in the calculation process of the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K may be displayed on the display device 3e.

The sample abnormality determination processing of the abnormality detection system 1 of the inspection apparatus T described above will be described with reference to the flowchart shown in FIG. A sample is attached to the inspection device T, and a current command I is repeatedly input from the controller C to apply load, speed and displacement to the sample to drive the actuator 13, and the load, speed and displacement as the output of the inspection device T are changed. (step ST11), the load F, velocity V and displacement X output from the inspection device T are detected by the load cell 2a, the velocity sensor 2b and the stroke sensor 2c, respectively (step ST12). Enter to

The processing device 3 stores the data of the past load F, velocity V, and displacement X of the inspection device T obtained when the normal same product specimen is inspected by the normal inspection device T, and the data of the same product specimen newly. A strain rate ε is obtained from data of the current load F, velocity V, and displacement X of the inspection apparatus T at the time of inspection (step ST13). In addition, the processing device 3 combines the data of the past load F, velocity V, and displacement X of the inspection device T obtained when the normal same product specimen was inspected by the normal inspection device T, and the same product specimen. The degree of variation Sd is obtained from the data of the current load F, velocity V, and displacement X of the inspection apparatus T at the time of new inspection (step ST14). Further, the processing device 3 combines the data of the past load F, velocity V, and displacement X of the inspection device T obtained when the normal specimen of the same product was inspected by the normal inspection device T, and the specimen of the same product. A correlation coefficient R is obtained from the data of the current load F, velocity V and displacement X of the inspection apparatus T at the time of new inspection (step ST15). Subsequently, the processing device 3 collects the data of the past load F, velocity V, and displacement X of the inspection device T obtained when the normal specimen of the same product was inspected by the normal inspection device T, and the data of the specimen of the same product. The amount of offset O is obtained from the data of the current load F, velocity V and displacement X of the inspection device T when inspecting is newly performed (step ST16). In addition, the processing device 3 combines the data of the past load F, velocity V, and displacement X of the inspection device T obtained when the normal same product specimen was inspected by the normal inspection device T, and the same product specimen. Rigidity K is obtained from data of current load F, velocity V, and displacement X of inspection apparatus T at the time of new inspection (step ST17). The processing device 3 can arbitrarily change the order in which the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K are obtained.

After obtaining the distortion rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K, the processing device 3 calculates the strain rate ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness Based on K, the process of judging whether or not there is an abnormality in the specimen is performed as described above (step FT18).

After performing the process of step ST8, the processing device 3 displays the calculated distortion factor ε, degree of variation Sd, correlation coefficient R, offset amount O, and stiffness K on the screen of the display device 3e (step ST19). At that time, when the processing device 3 determines that the sample is abnormal, the processing device 3 displays information specifying the values determined to be abnormal among the distortion factor ε, the degree of variation Sd, the correlation coefficient R, the offset amount O, and the stiffness K. 3e.

The anomaly detection system 1 of the inspection apparatus T operates as described above to detect an anomaly of the sample. , determine whether or not there is an abnormality in the specimen, display these indicators on the display device 3e and, if there is an abnormality, information specifying the abnormal indicator among these indicators, or print on paper with the printer 3g By printing on a medium, an operator can visually recognize these indicators and the like.

In the above-described embodiment, the output of the inspection apparatus T is the load, velocity, and displacement. , correlation coefficient R, and offset amount O may be obtained. In this embodiment, since the output of the inspection apparatus T is used as a plurality of parameters of load, velocity, and displacement, it is possible to more accurately determine the presence or absence of abnormalities in the specimen. Further, in the present embodiment, since the inspection device T is a vibration inspection device, acceleration may be added to the output.

As described above, the abnormality detection system 1 of the inspection apparatus T of the present embodiment can detect the load (output) F, velocity (output) V, and displacement of the inspection apparatus T when inspecting the damper (specimen) D with or without load A sensor 2 for detecting (output) X, and a processing device 3 for processing the load (output) F, velocity (output) V and displacement (output) X of the inspection device T, and the processing device 3 detects the past inspection The average value of the feature values of the normal load (output) F, velocity (output) V, and displacement (output) X of the device T and the load (output) F, velocity (output) V, and displacement (output) of the inspection device T this time ) An abnormality in the inspection apparatus T is detected based on the difference from the feature amount of X.

In addition, the abnormality detection method of the inspection apparatus T of the present embodiment includes the load (output) F, velocity (output) V, and displacement (output ) X, the average value of the feature values of the normal load (output) F, velocity (output) V, and displacement (output) X of the past inspection device T, and the current load (output) F of the inspection device T , and a detection process for detecting an abnormality of the inspection apparatus T based on the difference between the velocity (output) V and the displacement (output) X feature values.

According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured as described above, since the average value of the feature amount of the output of the past normal inspection apparatus T is used, the normal Since it is possible to determine whether or not the current inspection device T is abnormal by using the feature quantity that the inspection device T can take as an index, it is possible to accurately determine whether or not the inspection device T is abnormal. In the above description, the four indices of the distortion rate ε, the degree of variation Sd, the offset amount O, and the stiffness K are used as the feature quantity, but an arbitrarily selected index from among these indices may be used as the feature quantity. The feature amount is not limited to the above-described index, as long as it is an index that can grasp the static characteristics and dynamic characteristics of the inspection apparatus T. Further, in this embodiment, since the inspection apparatus T is a vibration inspection apparatus, the stiffness K can be obtained, but if it is a pressure inspection apparatus, the elastic modulus may be used as the feature amount. Further, the inspection device T may be a temperature inspection device that applies temperature to a sample, or other inspection devices other than the vibration inspection device and the pressure inspection device.

Further, in the abnormality detection system 1 of the inspection device T of the present embodiment, the processing device 3 generates an ideal output, which is an ideal output of the inspection device T with respect to the input, the load (output) F of the inspection device T with respect to the input, the speed Based on the (output) V and the displacement (output) X, the distortion factor calculator 3a1 obtains the output distortion factor ε, and the load (output) F and velocity (output ) A variation degree calculator 3a2 that obtains a variation degree Sd of V and displacement (output) X, and an average value of load (output) F, velocity (output) V, and displacement (output) X of a normal inspection apparatus T in the past. A correlation coefficient calculator 3a3 for obtaining a correlation coefficient between the current load (output) F, velocity (output) V, and displacement (output) X of the inspection device T, and the distortion factor ε and the degree of variation Sd as feature quantities, A judging section 3a6 judges whether or not there is an abnormality in the inspection device T based on the difference between the average value of the past feature amount and the current feature amount and the correlation coefficient R. Furthermore, in the method for detecting an abnormality in an inspection apparatus according to the present embodiment, the detection process includes an ideal output, which is an ideal output of the inspection apparatus T with respect to an input, and a load (output) F and a speed (output) of the inspection apparatus T with respect to an input. V and the displacement (output) X based on the distortion rate calculation process of obtaining the distortion rate ε, and the load (output) F, velocity (output) V and displacement ( The process of calculating the variation degree Sd of the output) X, the average value of the past normal load (output) F, velocity (output) V, and displacement (output) X of the inspection device T and the current load of the inspection device T (Output) F, Velocity (output) V, and Displacement (output) X and the correlation coefficient calculation process for obtaining the correlation coefficient R, and the average value of the past feature amounts with the distortion rate ε and the degree of variation Sd as the feature amounts and a judgment process for judging whether or not there is an abnormality in the inspection apparatus T based on the difference between the current feature amount and the correlation coefficient R.

According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured as described above, the presence or absence of an abnormality in the inspection apparatus T is detected using the distortion factor ε, the degree of variation Sd, and the correlation coefficient R. Since the presence or absence of an abnormality in the inspection apparatus T can be grasped from the viewpoint of the dynamic characteristics, repeatability, and degree of waveform matching of the inspection apparatus T, the presence or absence of an abnormality in the inspection apparatus T can be determined with high accuracy.

In addition, in the abnormality detection system 1 of the inspection device T of the present embodiment, the distortion factor calculation unit 3a1 changes the output by a sine wave with a predetermined cycle. ), and the integrated value of the power spectral density Pb of the output waveform Wb, which is the waveform of the output of the inspection apparatus T with respect to the current command (input) I, is distorted. The distortion factor ε may be obtained based on the integrated value of the power spectral density Pb of the load waveform (output waveform) Wb and the distortion component Pab. Then, in the process of calculating the distortion factor, the method of detecting an abnormality in the inspection apparatus T according to the present embodiment determines the ideal load (ideal output) of the inspection apparatus T with respect to the current command (input) I that changes the output in a sine wave of a predetermined cycle. The integrated value of the power spectral density Pa of the ideal waveform Wa, which is a waveform, and the integration of the power spectral density Pb of the load waveform (output waveform) Wb, which is the waveform of the load (output) F of the inspection device T with respect to the current command (input) I The distortion factor ε may be obtained based on the integrated value of the power spectral density Pb of the load waveform (output waveform) Wb and the distortion component Pab. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured as described above, not only the distortion of the ideal waveform Wa and the load waveform Wb with respect to the maximum load but also the difference between the two over the entire frequency range can be detected. Since the distortion factor ε with consideration is obtained, the actual dynamic behavior of the inspection device T can be grasped more accurately, and the abnormality of the inspection device T can be determined.

Furthermore, in the abnormality detection system 1 of the inspection apparatus T of the present embodiment, the distortion factor calculation unit 3a1 removes the drift component of the load waveform (output waveform) Wb, and obtains the power spectrum density Pb of the load waveform (output waveform) Wb. may In addition, the abnormality detection method of the inspection apparatus T of the present embodiment removes the drift component of the load waveform (output waveform) Wb in the distortion factor calculation process, and obtains the power spectrum density Pb of the load waveform (output waveform) Wb. good. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured as described above, the drift component is removed from the load waveform (output waveform) Wb, and then the power spectral density Pb is calculated. Low-frequency noise can be removed from the power spectral density Pb, the distortion component Pab can be prevented from becoming larger than it actually is, and the distortion factor ε can be obtained with high accuracy. Therefore, according to the abnormality detection system 1 for the inspection device T and the abnormality detection method for the inspection device T, the abnormality of the inspection device T can be detected more accurately.

Further, in the anomaly detection system 1 of the inspection apparatus T of the present embodiment, when the initial value W1 and the final value of the load waveform (output waveform) Wb are not the same in the distortion factor calculation unit 3a1, the load waveform (output waveform ) in Wb, the end of the load waveform (output waveform) Wb is cut off at a point W2 that has the same value as the initial value W1 and is in the vicinity of a point that is an integral multiple of the predetermined period from the initial value W1, and then the load waveform ( The power spectral density Pb of the output waveform) Wb may be obtained. Then, in the method of detecting abnormality of the inspection apparatus T of the present embodiment, in the distortion factor calculation process, if the initial value W1 and the final value of the load waveform (output waveform) Wb are not the same value, the load waveform (output waveform) Wb At point W2, which has the same value as the initial value W1 and is in the vicinity of a point that is an integral multiple of the predetermined period from the initial value W1, the end side of the load waveform (output waveform) Wb is cut off, and then the load waveform (output waveform ) may determine the power spectral density Pb of Wb. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured in this way, high-frequency noise can be removed from the power spectrum density Pb, and the distortion component Pab can be prevented from becoming larger than it actually is. can obtain the distortion factor ε with high accuracy. Therefore, according to the abnormality detection system 1 for the inspection device T and the abnormality detection method for the inspection device T, the abnormality of the inspection device T can be detected more accurately.

Furthermore, in the abnormality detection system 1 of the inspection apparatus T of the present embodiment, the variation degree calculation unit 3a2 outputs a current command (input) I that changes the load (output) of the inspection apparatus T with a sine wave of a predetermined cycle a plurality of times. The degree of variation Sd may be obtained based on the standard deviation σ _n or variance σ _n ² for each phase of the load waveform (output waveform) Wb, which is the waveform of the load (output) F of the inspection apparatus T when applied. Then, in the abnormality detection method of the inspection device T of the present embodiment, when the current command (input) I for changing the load (output) of the inspection device T with a sine wave of a predetermined cycle is given a plurality of times in the process of calculating the degree of variation, The degree of variation Sd may be obtained based on the standard deviation σ _n or the variance σ _n ² for each phase of the load waveform (output waveform) Wb, which is the waveform of the load (output) F of the inspection apparatus T. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured in this way, when the inspection apparatus T is actually used for a long period of time, the magnitude of the dynamic load fluctuation can be obtained, and the magnitude of the dynamic output fluctuation of the inspection apparatus T can be grasped. Therefore, according to the abnormality detection system 1 for the inspection device T and the abnormality detection method for the inspection device T, the abnormality of the inspection device T can be detected more accurately.

Further, in the abnormality detection system 1 of the inspection apparatus T of the present embodiment, a value three times the average value of the standard deviation σ _n or the variance σ _n ² for each phase of the load waveform (output waveform) Wb is used as the degree of variation Sd. you may ask. In the abnormality detection method of the inspection apparatus T of the present embodiment, a value three times the average value of the standard deviation _σn or variance _σn ² for each phase of the load waveform (output waveform) Wb is calculated in the process of calculating the degree of variation. It may be obtained as the degree of variation Sd. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured in this way, the inspection apparatus T is actually used for a long period of time. Since the variation degree Sd that objectively indicates the dynamic output fluctuation can be obtained, the magnitude of the dynamic output fluctuation of the inspection device T can be grasped more accurately. Therefore, according to the abnormality detection system 1 for the inspection device T and the abnormality detection method for the inspection device T, the abnormality of the inspection device T can be detected more accurately.

In the abnormality detection system 1 of the inspection device T of the present embodiment, the correlation coefficient calculation unit 3a3 calculates the average value of the past load (output) F of the inspection device T and the current load (output) of the inspection device T. A correlation coefficient R between frequency distributions may be obtained. In addition, the abnormality detection method of the inspection apparatus T of the present embodiment is such that the average value of the past load (output) F of the inspection apparatus T and the frequency distribution of the current load (output) of the inspection apparatus T in the process of calculating the correlation coefficient A correlation coefficient R between them may be obtained. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured as described above, the past average waveforms and current output waveforms of the two inspection apparatuses T for which the correlation coefficient R is to be obtained are Even if there is a phase shift between the two, the correct correlation coefficient R between the two can be obtained. Therefore, according to the abnormality detection system 1 for the inspection device T and the abnormality detection method for the inspection device T, the abnormality of the inspection device T can be detected more accurately.

Furthermore, in the abnormality detection system 1 of the inspection apparatus T of the present embodiment, the feature amount further includes the offset amount O of the load (output) F, and the processing apparatus 3 detects the offset amount O of the load (output) F of the inspection apparatus T. and the determination unit 3a6 further determines the difference ΔO between the average value _OGA of the normal offset amounts O of the past inspection apparatus T and the current offset amount _OT of the inspection apparatus T. The presence or absence of an abnormality in the inspection apparatus T may be determined by Further, in the abnormality detection method for the inspection apparatus T of the present embodiment, the feature amount further includes the offset amount O of the load (output) F, and the offset amount calculation for obtaining the offset amount O of the load (output) F of the inspection apparatus T In the determination process, the presence or absence of an abnormality in the inspection apparatus T is further determined based on the difference ΔO between the average value of the normal offset amount O of the inspection apparatus T in the past and the offset amount O of the current inspection apparatus T. You can judge. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured in this way, since the abnormality determination process is performed using the offset amount O, the output abnormality of the inspection apparatus T and the sensor 2 abnormalities can be detected.

Furthermore, the abnormality detection system 1 of the inspection apparatus T of the present embodiment is a vibration inspection apparatus in which the inspection apparatus T vibrates the damper (specimen) D, and the stiffness K of the load (output) F is added to the feature amount. the sensor 2 includes a load cell 2a for detecting the load F of the inspection device T and the stroke sensor 2c for detecting the displacement X; the processing device 3 calculates the stiffness K from the load F and the displacement X of the inspection device T 3a5, and a determination unit 3a6 further determines whether the inspection device T is abnormal based on the difference ΔK between the average value _KGA of the past normal stiffness _KG of the inspection device T and the current stiffness _KT of the inspection device T. You can judge whether or not Further, the abnormality detection method of the inspection apparatus T of the present embodiment is a vibration inspection apparatus in which the inspection apparatus T vibrates the damper (specimen) D, the feature amount further includes the rigidity K of the inspection apparatus T, and the inspection A stiffness calculation process is provided to obtain the stiffness K from _the load F and the displacement _X of the device T. The presence or absence of an abnormality in the inspection apparatus T may be determined based on the difference _ΔK from KT. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured as described above, since the abnormality determination process is performed using the stiffness K, the abnormality or deterioration of the parts in the inspection apparatus T, the sensor 2 abnormalities can be detected.

Furthermore, in the anomaly detection system 1 of the inspection apparatus T of the present embodiment, the processing device 3 includes the past distortion factor (feature amount) ε, the variation degree (feature amount) Sd, the offset amount (feature amount) O, and the stiffness (feature amount) K average value ε _GA , Sd _GA , O _GA , K _GA and current distortion rate (feature quantity) ε _T , variation degree (feature quantity) Sd _T , offset amount (feature quantity) _OT and stiffness (feature amount) _KT differences Δε, ΔSd, ΔO, ΔK are the standard deviations of the past distortion rate (feature amount) ε, degree of variation (feature amount) Sd, offset amount (feature amount) O, and stiffness (feature amount) K The inspection _apparatus ^T may be judged to be abnormal when it exceeds a threshold set based on σε _G , σSd _G , σO _G , σK _G or variances σε _G ² , σSd _G ² , σO _G ² , and σK G 2 . In addition, the abnormality detection method for the inspection apparatus T of the present embodiment includes, in the determination process, the past distortion rate (feature amount) ε, the degree of variation (feature amount) Sd, the offset amount (feature amount) O, and the stiffness (feature amount ) K average value ε _GA , Sd _GA , O _GA , K _GA , current distortion rate (feature quantity) ε _T , degree of variation (feature quantity) Sd _T , offset quantity (feature quantity) _OT and stiffness (feature quantity ) _KT differences Δε, ΔSd, ΔO, and ΔK are the standard deviation σε of the past distortion rate (feature amount) ε, variation degree (feature amount) Sd, offset amount (feature amount) O, and stiffness (feature amount) K. _The inspection apparatus T may be determined to be abnormal when it exceeds a threshold value set based on _G , σSd _{G ,} σO _G , σK _G or variances σε G ² , σSd _G ² , σO _G ² , and σK G ² . According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured in this way, an appropriate threshold value can be set for determining the range of values that the characteristic amount of the inspection apparatus T can take. , the abnormality of the inspection device T can be accurately detected.

Then, the abnormality detection system 1 for the inspection apparatus T of the present embodiment detects the distortion caused when the processing apparatus 3 inspects a normal sample with the inspection apparatus T in the past when the inspection apparatus T is determined to be normal. ratio (feature quantity) ε, variation degree (feature quantity) Sd, offset quantity (feature quantity) O, and stiffness (feature quantity) K average values ε _GA , Sd _GA , O _GA , K _GA , and the inspection apparatus T Distortion rate (feature quantity) ε _T , degree of variation (feature quantity) Sd _T , offset quantity (feature quantity) _OT and stiffness (feature quantity) _KT , ΔSd, ΔO, and ΔK, and a specimen abnormality judgment unit for judging whether or not there is an abnormality in the specimen. Further, in the abnormality detection method of the inspection apparatus T of the present embodiment, when the inspection apparatus T is determined to be normal, the distortion rate (feature amount) when the inspection apparatus T has inspected a normal sample in the past is Average values ε _GA , Sd _GA , O _GA , and K _GA of ε, degree of variation (feature value) Sd, offset amount (feature value) O, and stiffness (feature value) K, and normal specimen and The difference Δε between the strain rate (feature value) ε _T , the degree of variation (feature value) Sd _T , the offset value (feature value) _OT , and the stiffness (feature value) _KT when inspecting samples of the same product. , .DELTA.Sd, .DELTA.O, .DELTA.K. According to the abnormality detection system 1 for the inspection device T and the abnormality detection method for the inspection device T configured in this way, the same processing as the processing for determining whether or not there is an abnormality in the specimen device T as well as in the inspection device T is performed. Abnormalities in specimens can also be detected.

In the abnormality detection system 1 of the inspection apparatus T of the present embodiment, the determination unit (specimen abnormality determination unit) 3a6 has the past distortion rate (feature amount) ε, the degree of variation (feature amount) Sd, the offset amount (feature amount ) O and the average values ε _GA , Sd _GA , O _GA , and K _GA of the stiffness (feature quantity) K, the current distortion factor (feature quantity) ε _T , the degree of variation (feature quantity) Sd _T , and the offset amount (feature quantity) Differences _Δε , ΔSd, ΔO, and ΔK from OT and stiffness (feature quantity) _KT are the past distortion rate (feature quantity) ε, degree of variation (feature quantity) Sd, offset amount (feature quantity) O, and stiffness ( Feature quantity) If the standard deviation σε _G , σSd _G , σO _G , σK _G or the variance σε _G ² , σSd _G ² , σO _G ² , σK _G ² of K is equal to or greater than a threshold set, the sample is judged to be abnormal. You may In addition, the abnormality detection method of the inspection apparatus T of the present embodiment includes the past distortion rate (feature amount) ε, the variation degree (feature amount) Sd, the offset amount (feature amount) O, and the stiffness ( The average value ε _GA , Sd _GA , O _GA , and K _GA of the feature quantity) K, the current distortion rate (feature quantity) ε _T , the degree of variation (feature quantity) Sd _T , the offset quantity (feature quantity) _OT and the stiffness ( The differences Δε, ΔSd, ΔO, and ΔK from the feature amount) _KT are the past distortion rate (feature amount) ε, the degree of variation (feature amount) Sd, the offset amount (feature amount) O, and the stiffness (feature amount) K When the standard deviation σε _G , σSd _G , σO _G , σK _G or the variance σε _G ² , σSd _G ² , σO _G ² , σK _G ² is equal to or greater than a threshold value set based on the above, the inspection apparatus is judged to be abnormal. good too. According to the abnormality detection system 1 for the inspection apparatus T and the abnormality detection method for the inspection apparatus T configured as described above, an appropriate threshold value can be set to determine the range of values that the characteristic amount of a normal specimen can take. Abnormality of the inspection device T can be detected immediately.

In the above-described embodiment, the output of the inspection device T is the load, the velocity, and the displacement. Feature quantities such as the rate ε, the degree of variation Sd, the offset amount O and the stiffness K, and the correlation coefficient R may be obtained. In this embodiment, since the output of the inspection device T is a plurality of parameters of the load F, the velocity V, and the displacement X, it is possible to more accurately determine whether or not each inspection device T has an abnormality. Further, in the present embodiment, since the inspection device T is a vibration inspection device, acceleration may be added to the output.

Further, in the above-described embodiment, the inspection apparatus T is a vibration inspection apparatus using the damper D as the specimen. not something. Therefore, the inspection apparatus T may be an inspection apparatus that applies load, speed, displacement, pressure, impact, temperature, etc. to the specimen as an output, and the specimen can be appropriately changed according to the contents of the examination.

Although preferred embodiments of the invention have been described in detail above, modifications, variations, and changes are possible without departing from the scope of the claims.

1 ... Abnormality detection system 2 ... Sensor 2a ... Load cell 2c ... Stroke sensor 3 ... Processing device 3a1 ... Distortion rate calculation unit 3a2 ... Variation degree calculation Section 3a3 Correlation coefficient calculation section 3a4 Offset amount calculation section 3a5 Stiffness calculation section 3a6 Judgment section (specimen abnormality judgment section)

Claims (11)

An anomaly detection system for detecting an anomaly in an inspection device,
a sensor for detecting the output of the inspection device when inspecting a specimen or inspecting it without load;
A processing device for processing the output of the inspection device detected by the sensor,
The processing device calculates the average value of the feature amount of the output when the reference sample was inspected by the normal inspection device in the past, and the output feature when the reference sample is inspected by the inspection device this time. or the difference between the average value of the output feature amount when inspected by the normal inspection device in the past without load and the difference between the output feature amount when inspected by the inspection device this time without load Detecting an abnormality in the inspection device based on
a distortion factor calculation unit that obtains a distortion factor based on an ideal output, which is an ideal output of the inspection device with respect to the input, and an output of the inspection device with respect to the input;
a variation degree calculation unit for obtaining a variation degree of the output obtained when the output of the inspection device is repeatedly changed;
a correlation coefficient calculation unit that calculates a correlation coefficient between an average value of past outputs of the inspection device and a current output of the inspection device;
a judgment unit for judging whether or not there is an abnormality in the inspection device based on the difference between the average value of past feature amounts and the current feature amount, and the correlation coefficient, with the distortion factor and the degree of variation as the feature amounts; An abnormality detection system for an inspection device, comprising: The distortion factor calculation unit calculates an integrated value of a power spectral density of an ideal waveform, which is a waveform of an ideal output of the inspection device with respect to an input that changes the output with a sine wave of a predetermined cycle, and the output of the inspection device with respect to the input. A difference between the integrated value of the power spectral density of the output waveform, which is the waveform of the output waveform and the integrated value of the power spectral density, is used as a distortion component, and the distortion factor is obtained based on the integrated value of the power spectral density of the output waveform and the distortion component. An abnormality detection system for an inspection apparatus according to claim 1 . 3. The anomaly detection system for an inspection apparatus according to claim 2, wherein the distortion factor calculator removes a drift component from the output waveform and obtains the power spectral density of the output waveform. When the initial value and the final value of the output waveform are not the same value, the distortion factor calculation unit assumes the same value as the initial value in the output waveform and the vicinity of a point that is an integral multiple of the predetermined period from the initial value. 4. The anomaly detection system for an inspection apparatus according to claim 2, wherein the power spectrum density of the output waveform is obtained after cutting off the terminal end of the output waveform at a certain point. The degree-of-variation calculator calculates the standard deviation or variance for each phase of the output waveform, which is the waveform of the actual output of the inspection device when an input that changes the output with a sine wave of a predetermined cycle is given a plurality of times. The abnormality detection system for an inspection apparatus according to any one of claims 1 to 4 , wherein the degree of variation is obtained. 6. The correlation coefficient calculator according to any one of claims 1 to 5, wherein the correlation coefficient calculation unit calculates a correlation coefficient between an average value of the past outputs of the inspection device and a frequency distribution of the current output of the inspection device. 1. An abnormality detection system for the inspection device according to item 1. The feature amount further includes an offset amount of the output,
The processing device has an offset amount calculation unit that calculates an offset amount of the output of the inspection device,
The determination unit further determines whether or not there is an abnormality in the inspection device based on a difference between an average value of the normal offset amounts of the inspection device in the past and the current offset amount of the inspection device. 7. The abnormality detection system for inspection equipment according to claim 1. The inspection device is a vibration inspection device that vibrates the specimen,
The sensor includes a load cell that detects the load of the inspection device and a stroke sensor that detects displacement,
The feature quantity further includes the stiffness of the inspection device,
The processing device has a stiffness calculator that calculates stiffness from the load and displacement of the inspection device,
The determination unit further determines whether or not there is an abnormality in the inspection device based on a difference between an average value of the normal stiffness of the inspection device in the past and the stiffness of the current inspection device. An abnormality detection system for an inspection apparatus according to any one of claims 1 to 7. The processing device is
The inspection device is determined to be abnormal when a difference between the average value of the past feature amount and the current feature amount is equal to or greater than a threshold value set based on the standard deviation or variance of the past feature amount. An abnormality detection system for an inspection apparatus according to any one of claims 1 to 8. When the inspection device is determined to be normal,
The processing device is
The average value of the feature values obtained when normal specimens were tested by the testing device in the past, and the feature values obtained when testing the same product samples as the normal samples by the testing device this time. 10. The abnormality detection system for an inspection apparatus according to any one of claims 1 to 9, further comprising a specimen abnormality judgment unit that judges whether or not there is an abnormality in the specimen based on the difference between the two. The specimen abnormality determination unit
The specimen is determined to be abnormal when the difference between the average value of the past feature amount and the current feature amount is equal to or greater than a threshold set based on the standard deviation or variance of the past feature amount. An abnormality detection system for an inspection apparatus according to claim 10.

Images