JPH10258974A - Unsteady signal analyzer and medium recording unsteady signal analysis program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To appropriately diagnose an abnormal state of a monitoring object by non-linearly convert a time coordinate of a wavelet spectrum data, which is found from an unsteady signal by an inverse function of a state quantity change function expressing the time change in the specific state quantity of the monitoring object, into a specific state quantity coordinate. SOLUTION: Unsteady signal data x(t) obtained by a response data measurement means 1 is sent to a wavelet conversion calculating means 2 so as to perform a wavelet conversion, the calculated wavelet spectrum data Wt(a, b) is sent to a time coordinate non-linear conversion means 3, and the time coordinate of the wavelet spectrum data wt(a, b) is converted into the non-linear coordinate in relation to the specific state quantity of a monitoring object. Secondly, the expanded wavelet spectrum wt(a, z) found by a time coordinate non-linear conversion means 3 is sent to a display means 7 so that a two-dimensional function wt (ω, z) for a frequency coordinate ω=a-1 and the state quantity coordinates z is displayed as the expanded wavelet spectrum data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an unsteady signal analyzer for analyzing unsteady signals generated from various mechanical systems, processes, and the like to be monitored.
In particular, the present invention relates to a non-stationary signal analyzer for analyzing a non-stationary signal generated from an elevator.

[0002] The present invention also relates to a medium recording a program for analyzing, by a computer, an unsteady signal generated from an object to be monitored.

[0003]

2. Description of the Related Art Conventionally, a signal generated from a mechanical system, a process, or the like to be monitored is measured by a measuring device, and signal data obtained by the measurement is analyzed to detect an abnormality of the monitoring target, and an operator ( Various diagnostic systems have been proposed as warning systems for displaying an abnormal state to an operator or a user.

[0004] In these diagnostic systems, generally, a method of monitoring a spectrum by Fourier transforming signal data obtained from a monitoring target, or estimating a characteristic model from input / output data of the monitoring target by a system identification method is mainly used. It is.

[0005] However, for an unsteady signal measured in an unsteady state in which the operating state of the monitored object changes abruptly, a spectrum that changes every moment is obtained by Fourier transform, or a characteristic model that fluctuates by system identification. I couldn't ask.

Therefore, as a technique for analyzing an unsteady signal to detect an abnormality of a monitoring target, a wavelet analysis technique using a wavelet transform has been attracting attention. Hereinafter, this wavelet analysis method will be described.

The signal data measured from the object to be monitored is x
(T), the conventional Fourier transform is given by the following equation (1)

[0008]

(Equation 3)

Is represented by On the other hand, the wavelet transform is expressed by the following equation (2)

[0010]

(Equation 4)

Is represented by Here, φ (•) is a basis function for transformation called a mother wavelet.

The Fourier transform corresponds to a case where the basis functions in the wavelet transform are φ (t) = e ^−jt , b = 0 and a = ω ^−1, and the basis functions are shown in FIG. It is a function that extends from the infinite past to the future. Therefore, the spectrum obtained by the Fourier transform is
For example, as shown in FIG. 2B, it is a one-dimensional function only on the frequency axis, and it is not possible to determine the time dependence of which part of the observation data is a feature.

On the other hand, in the wavelet transform, for example, the following equation (3)

[0014]

(Equation 5)

By using a function called a Gabor function shown as a basis function, the basis function becomes a function localized in time as shown in FIG.

For this reason, the spectrum obtained by the wavelet transform becomes a two-dimensional function with respect to the frequency axis and the time axis as shown in FIG.
The time dependency of each frequency component in the signal can be determined based on the dimensional function.

For details regarding the wavelet transform theory, see, for example, Susumu Sakakibara: Mathematical Science Seminar “Wavelet Beginner's Guide”, Tokyo Denki University Press (199)
It is introduced in 5).

As described above, since the wavelet transform can extract the spectrum distribution of the observed data at each time, it is effective as an analyzing means for an unsteady signal. Therefore, the operating state of the monitored object changes every moment. It is said that it is also effective when converting.

[0019]

However, since the conventional diagnostic system described above simply performs a wavelet transform on an unsteady signal obtained from a monitored object, the analysis result is based on the frequency. It merely shows the time dependence of the spectrum, and was insufficient as an analysis method for diagnosing abnormalities in the analysis target.

For example, even if the analysis result of the conventional diagnostic system as shown in FIG. 3B is viewed, it is not possible to read how the analysis result relates to the change in the state of the monitoring target.

Accordingly, the present invention provides an unsteady signal analyzing apparatus which solves the above-mentioned problems and which can accurately diagnose an abnormal state of a monitored object by analyzing an unsteady signal obtained from the monitored object. The purpose is to provide.

[0022]

According to a first aspect of the present invention, there is provided an unsteady signal analyzing apparatus for analyzing an unsteady signal generated from an object to be monitored. A wavelet transform calculating means for generating wavelet spectrum data, a state quantity change function setting means for setting a state quantity change function representing a time change of a specific state quantity in the monitoring target, and an inverse of the state quantity change function. A time coordinate non-linear conversion means for non-linearly converting the time coordinate of the wavelet spectrum data into the coordinate of the specific state quantity by a function,
It is characterized by having.

According to a second aspect of the present invention, there is provided a non-stationary signal analyzer for monitoring an elevator, and analyzing an acceleration signal which is a non-stationary signal measured in a car of the elevator. In the device,
A wavelet transform calculating means for performing wavelet transform on the acceleration signal to generate wavelet spectrum data, and a state quantity for setting a state quantity change function representing a time change of an ascending / descending position or an ascending / descending speed which is a specific state quantity of the car. Change function setting means, and time coordinate nonlinear conversion means for nonlinearly converting the time coordinate of the wavelet spectrum data into the coordinate of the elevation position or the elevation speed by an inverse function of the state quantity change function, I do.

According to a third aspect of the present invention, in the non-stationary signal analyzing apparatus, the time coordinate non-linear transforming means includes an extended wavelet transform type.

[0025]

(Equation 6)

The spectral data obtained by nonlinearly converting the time coordinate of the wavelet spectral data into the coordinate of the specific state quantity is calculated.

According to a fourth aspect of the present invention, in the non-stationary signal analyzing apparatus, the time coordinate nonlinear transformation means divides the wavelet spectrum data for each time, and stores a relationship between time and the specific state quantity. Based on the data table or the state quantity change function, the divided data is rearranged in the order of the state quantity, and interpolation or smoothing processing is performed between the respective data, so that the time coordinate of the wavelet spectrum data is set to the specific state quantity. It is characterized by calculating spectrum data that has been nonlinearly transformed into coordinates.

According to a fifth aspect of the present invention, there is provided an unsteady signal analyzing apparatus further comprising response data measuring means for measuring the unsteady signal.

According to a sixth aspect of the present invention, in the non-stationary signal analyzer, the state quantity change function setting means estimates the state quantity change function based on measurement data relating to a state quantity other than the specific state quantity. It is characterized by the following.

An unsteady signal analyzing apparatus according to a seventh aspect of the present invention is characterized in that measurement data relating to a state quantity other than the specific state quantity is measurement data relating to the unsteady signal.

According to an eighth aspect of the present invention, in the non-stationary signal analyzer, the state quantity change function setting means uses a state observer or a Kalman filter based on a dynamic characteristic model of the monitored object to calculate the specific state quantity. The state quantity change function is estimated by estimating a time change based on measurement data of a state quantity other than the specific state quantity.

According to a ninth aspect of the present invention, in the non-stationary signal analyzer, the state quantity change function setting means obtains the state quantity change function based on measurement data of the specific state quantity. And

An unsteady signal analyzing apparatus according to a tenth aspect is characterized in that the state quantity change function setting means uses the state quantity change function obtained in advance.

An unsteady signal analyzing apparatus according to an eleventh aspect of the present invention further comprises a display means for displaying the analysis result of the time coordinate non-linear conversion means in a coordinate system having at least the coordinates of the specific state quantity and the coordinates of the frequency. It is characterized by having.

An unsteady signal analyzing apparatus according to a twelfth aspect of the present invention is characterized by further comprising an abnormality determining means for determining whether or not an abnormality has occurred in the monitored object based on an analysis result of the time coordinate nonlinear conversion means. And

According to a thirteenth aspect of the present invention, in the non-stationary signal analyzing apparatus, a specific region in the entire display is defined with respect to a display by the display means for spectrum data which is an analysis result of the time coordinate non-linear conversion means. It is characterized by further comprising an area designating means for designating, and data extracting means for extracting spectrum data corresponding to the area designated by the area designating means and sending it to the abnormality determining means.

According to a fourteenth aspect of the present invention, there is provided the non-stationary signal analyzer, wherein a result of the judgment by the abnormality judging means is displayed on the display means.

According to a fifteenth aspect of the present invention, there is provided the non-stationary signal analyzing apparatus, further comprising abnormality display means for displaying a determination result by the abnormality determination means.

The medium storing the non-stationary signal analysis program according to the present invention is a medium storing a program for analyzing a non-stationary signal generated from an object to be monitored by a computer. The program has a computer that performs a wavelet transform on the non-stationary signal to generate wavelet spectrum data, and a state quantity change function that sets a state quantity change function representing a time change of a specific state quantity in the monitoring target. It is characterized by realizing a function setting function and a time coordinate non-linear conversion function for non-linearly converting the time coordinate of the wavelet spectrum data into the coordinate of the specific state quantity by an inverse function of the state quantity change function.

The medium on which the non-stationary signal analysis program according to the present invention is recorded, wherein the monitored object is an elevator, the unsteady signal is an acceleration signal measured in a car of the elevator, The state quantity is a vertical position or vertical speed of the car.

The medium on which the non-stationary signal analysis program according to claim 18 is recorded, wherein the time-coordinate non-linear transformation means is an extended wavelet transform type.

[0042]

(Equation 7)

The spectral data obtained by nonlinearly transforming the time coordinate of the wavelet spectral data is calculated.

The medium on which the non-stationary signal analysis program according to the invention of claim 19 is recorded, wherein the time coordinate non-linear transformation means divides the wavelet spectrum data for each time and calculates a time and the specific state quantity. Based on the data table storing the relations or the state quantity change function, the divided data is rearranged in the order of the state quantity, and interpolation or processing is performed between the respective data, so that the time coordinate is nonlinear to the coordinate of the specific state quantity. It is characterized by calculating the converted spectrum data.

[0045]

DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment A first embodiment of the non-stationary signal analyzer according to the present invention will be described below with reference to FIGS.

FIG. 1 shows a schematic configuration of an entire non-stationary signal analyzer according to the present embodiment. This non-stationary signal analyzer is a response data measuring means for measuring an unsteady signal generated from an object to be analyzed. 1 is provided. The response data measuring means 1 is composed of a sensor, an A / D converter, various noise removing filters, and the like.

The unsteady signal data (response data time series) x (t) obtained by the response data measuring means 1 is sent to the wavelet transform calculating means 2.

The wavelet transform calculating means 2 is provided, for example, by the above-described wavelet transform equation (2)

[0049]

(Equation 8)

Is held. Here, a is the reciprocal of the frequency ω, and b is the time t.

Then, in the wavelet transform calculation means 2, the non-stationary signal data x (t) is calculated using the above equation (2).
Is subjected to wavelet transform, and wavelet spectrum data (wavelet transform data) wt (a, b) as shown in FIG. 3B is calculated.

The wavelet spectrum data wt (a, b) obtained by the wavelet transform calculation means 2
Is sent to the time coordinate nonlinear transformation means 3. The time-coordinate non-linear transforming means 3 outputs the wavelet spectrum data wt obtained by the wavelet transform calculating means 2.
This is means for performing non-linear coordinate transformation of the time coordinate of (a, b) with respect to a specific state quantity (physical quantity) of the monitoring target.

Here, the specific state quantity is, for example, a speed or a position, if the unsteady signal measured by the response data measuring means 1 is a signal relating to acceleration. It is. In this regard,
In the first and second embodiments described below, an elevator and a railway train will be described in detail as examples.

The unsteady signal analyzer according to the present embodiment provides state quantity change function data ｛z (t ₁ ), z (t ₂ )... Z (t _N ) representing the relationship between time and a specific state quantity. ) A time-state quantity conversion table 4 for writing｝ is provided. The time-state amount conversion table 4 constitutes a state amount change function setting unit that sets a state amount change function representing a time change of a specific state amount in the monitoring target together with the state amount estimation unit 6 described later.

Then, the time-state quantity conversion table 4
State change function data {z (t ₁ ), z
There are several methods for obtaining (t ₂ )... Z (t _N )}, but in the present embodiment, state quantity change function data {z (t ₁ ) obtained in advance. , Z
(T ₂ )... Z (t _N )} is changed by the input means 5 to time
It is written in the state quantity conversion table 4.

The state quantity change function data ｛z
As another method for obtaining (t ₁ ), z (t ₂ )... Z (t _N )｝, for example, a specific state quantity z (for example, speed) is directly measured to obtain a state quantity measurement value. By
A method of directly acquiring the time change of the specific state quantity z, or estimating the time change of the specific state quantity z based on measurement data on a state quantity (eg, acceleration) other than the specific state quantity z (eg, speed) And the like.

The latter method of estimating the time change of the specific state quantity z from the state quantities other than the specific state quantity is a method using the state quantity estimating means 6 shown in FIG. A modification of the embodiment will be described later.

The time coordinate nonlinear conversion means 3 reads out the state quantity change function data {z (t ₁ ), z (t ₂ )... Z (t _N )} written in the time-state quantity conversion table 4.
The time coordinate b of the wavelet spectrum data wt (a, b) is converted into the coordinate of the state quantity z based on the read state quantity change function data. Specifically, the time coordinate nonlinear transformation means 3 obtains an inverse function t (z) of a function (state quantity change function) z (t) with respect to time t of a specific state quantity z,
Based on the inverse function t (z), the above-described wavelet transform equation (2) is variable-converted from the time t to a specific state quantity z.

The wavelet transform equation (2) thus transformed into a specific state quantity z is as shown in the following equation (4).

[0060]

(Equation 9)

In the following, the transform expressed by the above equation (4) will be referred to as an extended wavelet transform for convenience. By converting the time coordinate b of the wavelet spectrum data wt (a, b) into the coordinate of the state quantity z by the extended wavelet transform equation (4), the extended wavelet spectrum wt ( a, z) are obtained.

Note that the above wavelet spectrum w
In order to emphasize the meaning of t (a, b) as a function of the frequency ω, ω = a ^-1 is hereinafter referred to as wt (ω, b), wt (a, b).
^-1 , b) will also be used. In the following, the conventional wavelet spectrum is denoted by wt (ω, t), and the extended wavelet spectrum is denoted by wt (ω, z).

The extended wavelet spectrum wt
As another way of calculating (ω, z), wavelet spectrum data wt (ω, t) obtained by the conventional wavelet transform is converted into data {wt (ω, ω) at each time {t ₁ , t ₂ ... T _n }. t ₁ ), wt (ω, t ₂ ) ... wt (ω,
t _n )} and a state quantity change function z (t) indicating the relationship between time and a specific state quantity or a data table {z (t ₁ ), z (t ₂ ). t _n )｝, the divided data is rearranged in the order of the state quantity z, and interpolation or smoothing processing is performed between the data, so that the extended wavelet spectrum wt (ω, z ).

Next, the extended wavelet spectrum wt (a, z) obtained by the time coordinate nonlinear transformation means 3 is sent to the display means 7. The display means 7 converts the time coordinate axis into a variable wavelet spectrum data (wavelet analysis data) based on the extended wavelet spectrum wt (a, z) to obtain the frequency coordinate ω.
= A ^-1 (or its reciprocal a) and 2 for the state quantity z
The dimensional function wt (ω, z) is displayed.

[0065]

(Equation 10)

Further, the unsteady signal analyzing apparatus according to the present embodiment automatically determines whether or not an abnormality has occurred in a monitoring target based on the extended wavelet spectrum data calculated by the time coordinate nonlinear transformation means 3. 8 is provided. The abnormality judging means 8 automatically judges whether the monitoring target is normal or abnormal by using a predetermined abnormality diagnosis method, sends alarm information, abnormal mode information, etc., which are the judgment results, to the display means 7 to display a warning to the operator. It is.

Here, as the predetermined abnormality diagnosis method, for example, the extended wavelet spectrum data w
Based on the power spectrum value of a specific portion of t (ω, z) {| wt (ω ₁ , z ₁ ) |,..., | wt (ω _m , z _m ) |}, the following equation (5) if ( | Wt (ω _i , z _i ) |> ε _i ) then threshold judgment of abnormal i (5) and the like and judgment by various complex means are performed.

The determination result by the abnormality determining means 8 is as follows:
In addition to displaying on the display means 7, an alarm can be displayed on the abnormality display means 9 provided separately from the display means 7.

The user designates a specific area from the whole display by a pointing device or the like with respect to the display on the display means 7 for the extended wavelet spectrum data which is the analysis result of the time coordinate nonlinear transformation means 3,
It is also possible to take out only the extended wavelet spectrum data corresponding to the specific area and send it to the abnormality judging means 8, and the abnormality judging means 8 judges the occurrence of an abnormality in the monitoring target using only the data.

Thus, the user can determine and take out a portion having a different characteristic from the extended wavelet spectrum once displayed on the display means 7 and analyze only that portion to obtain another portion. The analysis work can be performed directly without being affected by noise, disturbance, and other factors included in the data, and as a result, the accuracy of abnormality determination can be improved.

As described above, according to the non-stationary signal analyzer of the present embodiment, the non-stationary signal data measured from the monitored object is subjected to the wavelet transform to obtain the wavelet spectrum data. Since the time coordinate axis is coordinate-transformed to the coordinate axis of a specific state quantity (physical quantity), not only the time change of the frequency spectrum but also the specific state quantity (for example, position, velocity, acceleration, etc. in a mechanical system) and the frequency spectrum Can be easily grasped.

For this reason, when an abnormal phenomenon occurs in the monitored object, an analysis result / display that makes it easy to understand the abnormal phenomenon from the viewpoint of the laws of physics is obtained, and for example, the location where the abnormal phenomenon occurs can be easily specified. .

Further, according to the non-stationary signal analyzer of the present embodiment, in the non-stationary state where the operating state and the internal state of the monitored object frequently change, the spectrum distribution can be analyzed for each changing state. Therefore, it is extremely effective for analyzing non-stationary signals, and as a result, it is possible to perform effective analysis even for short fragmentary data.

Modification Next, a modification of the above-described first embodiment will be described with reference to FIG.

The non-stationary signal analyzer according to the present modified example has state quantity change function data {z (t ₁ ), z (t ₂ )... Z (t _N )} for writing in the time-state quantity conversion table 4.
Is estimated and created by the state quantity estimating means 6 based on measurement data on state quantities other than the specific state quantity z. Therefore, it is extremely effective when the specific state quantity z cannot be directly measured.

The state quantity estimating means 6 in the present modified example estimates the time change of a specific state quantity z from measured data in real time using state estimating means based on a dynamic characteristic model of a monitored object, thereby obtaining a state quantity. Quantity change function data ｛z
(T ₁ ), z (t ₂ )... Z (t _N )}.

FIG. 4 shows a schematic configuration of the state quantity estimating means 6 in this modification. In the state quantity estimating means 6, the difference between the output predicted value y hat (t) when the input signal u (t) of the monitoring target 10 is input to the output signal prediction model 11 and the actual output signal y (t) is obtained. Based on the estimated error signal e (t), the estimated state amount correcting means 12 sequentially corrects the state amount estimation value in the output signal prediction model 11 to obtain a specific state amount z that cannot be directly measured.
(T) can be estimated in real time.

In a configuration using a Kalman filter or a state observer as an example of the state estimating means, the output signal prediction model is expressed by the following equations (6) and (7), and the estimated state quantity correcting means is expressed by the following equation (8).

[0079]

[Equation 11]

Here, A, B, and C are coefficient matrices related to the dynamic characteristic model to be monitored, and K is a Kalman gain (or state observer gain).

By this successive calculation, the internal state quantity vector z (k | k) of the monitoring target can be estimated from the observation data sequence of the input signal u (k) and the output signal y (k) of the monitoring target. Removed some elements of state vector in this manner is estimated as a particular state variable z, time series _{{z (t 1), z} (t 2) ...... z (t N)} from the time- A state quantity conversion table 4 is created.

The above-mentioned estimation of the state quantity includes a method of performing offline processing in advance and a method of performing real-time processing while observing data.

As described above, according to this modification, even when the specific state quantity z cannot be directly measured, the state quantity estimating means 6 compares the specific state quantity z with the observed data spectrum. You can understand the relationship. In addition, it becomes easy to combine another analysis technique with the wavelet analysis technique.

First Example Next, as a first example of the unsteady signal analyzer according to the first embodiment, a case where an object to be monitored is an elevator which is a mechanical system will be described with reference to FIGS. . In this embodiment, the measurement signal (unsteady signal)
Is the acceleration signal measured in the elevator car, and the specific state quantity used for the non-linear conversion is the elevator position or the elevator speed of the elevator car.

As shown in FIG. 5, the elevators to be monitored include a motor 51, sheaves (pulleys) 52a, 52
b, 52c, 52d, a car frame 53, a car 54, a guide roller 55, a guide rail 56, and a counterweight 57.

The unsteady signal analyzing apparatus according to the present embodiment includes the acceleration sensor 20 arranged in the car 54 as shown in FIG. This acceleration sensor 2
The acceleration signal measured by the A / D converter 21
After the data is converted into the data, it is taken into an analysis / display device (for example, a personal computer) 22. Acceleration sensor 20 and A
The / D converter 21 constitutes the response data measuring means 1 shown in FIG.

In the analysis / display device 22, extended wavelet spectrum data is calculated by the processing shown in FIG. 1, and the calculated extended wavelet spectrum data is displayed on the screen of the analysis / display device 22. Also,
The analysis result or abnormality diagnosis result is sent to a remote monitoring center via the modems 23 and 23 and a public line, displayed on the central monitoring terminal 24 of the monitoring center, and an alarm is generated according to the abnormal state.

FIG. 7 shows a specific flowchart in the analysis / display device 22.

First, when the diagnosis is started, the acceleration sensor 2
The acceleration signal x (t) in the car 54 is measured by 0 (step 1). Next, the wavelet spectrum data wt (a, b) is calculated based on the measured acceleration signal x (t) using the above-described equations (2) and (3) (step 2).

Then, the wavelet spectrum data wt (a, b) or wt (ω, b) which is the analysis result is obtained.
Is displayed on the display terminal with respect to the time axis and the frequency axis (step 3). Here, ω = a ⁻¹ is the frequency of the wavelet spectrum. Next, the user selects either the car speed or the car position as the specific state quantity to be analyzed (step 4). In addition, about this step 4, it is also possible to automatically process both procedures of the car speed and the car position.

Next, when the car position is selected in the selection of the specific state quantity in step 4, the acceleration signal x
By integrating the second order of (t), the car position signal p
(T) is generated (step 5). And step 5
Car position signal data {p (t ₁ ), p
Based on (t ₂ )... P (t _N )}, a function table of time t and position p is created (step 6).

Next, the time coordinates of the wavelet spectrum data calculated in step 2 are converted into the coordinates of the car position p based on the function table created in step 6, and the expanded wavelet spectrum data wt (ω,
p) is obtained (step 7). Then, the extended wavelet spectrum data wt (ω, p) as the analysis result is displayed on the display (step 8).

Further, as shown in the following equation (9), wt
The change rate with respect to the position p is calculated by (ω, p), and whether or not the change rate exceeds a predetermined threshold is determined by a determination formula, and whether or not there is a sudden change with respect to the position of the power spectrum is determined (step 9). .

[0094]

(Equation 12)

If it is determined in step 9 that there is a sudden change, the sudden change point p (t _i ) is detected and displayed on the display as an abnormality in the elevator system rail or rope (step 10). On the other hand, if it is determined that there is no sudden change, "no abnormality" is displayed on the display (step 11), and the diagnosis is terminated or the operation waits until the next operation cycle.

If the car speed is selected as a specific state quantity in step 4, the acceleration signal x (t)
Is first-order integrated to generate a car speed signal v (t) (step 12). Then, the car speed signal data {v (t ₁ ), v (t ₂ ).
Based on (t _N )}, a function table of time t and speed v is created (step 13).

Next, the time coordinates of the wavelet spectrum data wt (a, b) calculated in step 2 are converted into the coordinates of the car speed v based on the function table created in step 13, and the expanded wavelet spectrum data wt (Ω, v) is obtained (step 14). And
Extended wavelet spectrum data w that is the analysis result
Display t (ω, v) on the display (Step 1)
5).

Further, as shown in the following equation (10), | w
t (ω, v) | of the threshold determination _{| wt (ω i, v i} ) |> by .epsilon.v (10), data of a portion having the power spectrum above the threshold (peak spectrum) {wt (omega ₁ , V ₁ ), wt
(Ω ₂ , v ₂ )... Wt (ω _m , v _m )} is detected. In addition, the proportional relationship between the frequency ω and the car speed _{v (11) v i = rω} i + e i (11) assuming, coefficient sum of squares Σe _i ² of the error e _i is the minimum r
Least-squares solution of

[0099]

(Equation 13)

Is obtained.

At this time, each data point ｛(ω ₁ , v ₁ ),
(Ω ₂ , v ₃ )... (Ω _m , v _m )｝ The proportionality equation (11) determines the variance of the distance d from the straight line.

[0102]

[Equation 14]

If the above holds, it is determined that there is a strong correlation (proportional relationship) between the speed v and the frequency ω (step 16).

In this case, it is determined that there is an abnormality in any of the elevator rotation systems (motors, sheaves, bearings for each shaft, guide rollers, etc.). This is because the frequency of torque unevenness due to eccentricity of the rotating system is proportional to the rotation speed, and the car speed is also proportional to the rotation speed.

Further, it is determined which rotational system is based on the proportional coefficient r, and the result of the determination is displayed on the display (step 1).
7). For example, if (r / 2π) coincides with the radius of the sheave (pulley), car speed = 2π (sheave radius) x (sheave rotation frequency). You. On the other hand, if it is determined in step 16 that there is no correlation, "no abnormality" is displayed on the display (step 11), and the diagnosis is terminated or the operation waits until the next operation cycle.

Next, step 6 or step 1 corresponding to the processing in the time coordinate nonlinear transformation means 3 shown in FIG.
The specific procedure of coordinate conversion in 3 will be described.

First, the car position signal p (t) or the car speed signal v (t) will be referred to as a state quantity signal z (t) here, and the data string ｛z (t ₁ ), z (t ₂ ). z
It is assumed that a function table (time to state quantity conversion table 4) consisting of (t _N )｝ has been obtained. Then, data obtained by ordinary wavelet transform wt (a, b) = {wt (a _i , b _j ) | i = 1,..., N1, j = 1,. ), The relationship of frequency ω = a ⁻¹ is substituted, and wt (ω, b) = ｛wt (ω _i , b _j ) | ω _i = a _i ⁻¹ , i = 1, ..., n1, j = 1,..., n2｝ (16)

[0108] Next, with respect to time coordinate b _j of each data element, finding a _{t k ≦ b j ≦ t k} + 1 (17) comprising t _k from _{{t 1, t 2 ... t} N}, the linear interpolation formula

[0109]

(Equation 15)

By calculating the corresponding state quantity z (b _j ), the expanded wavelet spectrum data wt (ω, z) = ｛wt (ω _i , z (b _j )) | ω _i = a _i ^-1 , I = 1, .., n1, j = 1, .., n2 １９ (19).

As another means, a function z (t) is estimated from the time-state quantity conversion table 4. For _{example, z (t) = z 0} + z 1 t + ... z p t p (20) becomes a polynomial assuming, coefficient z _0, z ₁ ... z _p data {z
(T ₁ ), z (t ₂ )... Z (t _N )}. Next, the inverse function t (z) is obtained.

Finally, for the observation data x (t),

[0113]

(Equation 16)

The extended wavelet spectrum data can also be obtained by directly numerically integrating.

FIGS. 8 to 15 show the results of analyzing the acceleration data of the elevator car 54 using the unsteady signal analyzer according to the present embodiment.

FIGS. 8 to 12 show the motor 5 of the elevator.
An example in which abnormal vibration occurs in the car 54 due to uneven rotation torque due to the eccentricity of one shaft is shown. FIG. 8A shows a torque command value of the motor 51, and FIG.
(B) is the car speed signal v (t) estimated from the integral calculation of the car acceleration signal x (t) at that time, and FIG. 8 (c) is similarly estimated from the second-order integral calculation of the car acceleration signal x (t). This is the car position signal p (t).

At this time, the acceleration data x (t) observed inside the car 54 is an unsteady signal whose frequency characteristic changes with speed as shown in FIG. 9A. The reason is that the frequency of the torque unevenness due to the eccentricity of the motor shaft changes in proportion to the car speed as shown in FIG. 9B. Therefore, even if the acceleration data x (t) is simply Fourier-transformed, only the entire spectral distribution is known as shown in FIG. 9C, and the dependence on the speed signal cannot be determined.

FIGS. 10, 11 (a), (b), and 12 show the results of the conventional wavelet transform of the acceleration signal x (t), the results of the extended wavelet transform based on the velocity signal v (t), and the position, respectively. The result of the extended wavelet transform based on the signal p (t) is shown.

For example, looking at the wavelet spectrum data based on the velocity signals shown in FIGS. 11A and 11B, the peak of the spectrum (peak spectrum) shows the velocity signal v (t).
And the frequency ω = a ⁻¹ can be determined to be aligned on a straight line representing a proportional relationship. From this result, it is determined that the rotation system of the elevator is abnormal, and further, it is determined from the proportional relationship between the speed and the frequency that the axis of the motor 51 is abnormal.

FIGS. 13 to 15 show the results of analysis when the guide rail 56 of the elevator has an abnormality. (A), (b), and (c) of FIG. 13 show a motor torque signal, a car speed signal v (t), and a car position signal p, respectively.
(T) is shown. FIGS. 14A and 14B show the in-car acceleration signal x (t) and its Fourier transform result, respectively.

While the car 54 is rising, there is a step due to the seam of the guide rail 56 at a height of about 10.7 m, and the car 54 receives an impulse-like external force due to the step and starts to vibrate. . However, FIG.
From the result of the Fourier transform shown in (b), it cannot be determined at which portion the external force has been applied.

Next, the position signal p of the acceleration signal x (t)
FIG. 15 shows the result of the extended wavelet transform for (t).
Shown in As can be seen from FIG. 15, a sharp change in the spectrum is observed at the position of p = 10.7 m with respect to the car position p-axis direction. Therefore, it is determined that the abnormality of the guide rail 56 has occurred at the portion of p = 10.7 m.

As described above, according to this embodiment, the acceleration signal of the elevator car 54 is subjected to the extended wavelet transform with respect to the car speed, and the peak spectrum frequency and the car speed in the obtained extended wavelet spectrum data are obtained. It can be determined from the proportional relationship that the torque fluctuation of the rotating system has occurred, and the radius of the rotating system that causes the fluctuation can be specified from the proportional coefficient.

Similarly, the presence or absence of a flaw such as a rail or a rope can be determined from the change in the spectrum position of the extended wavelet spectrum data with respect to the car position of the acceleration signal.
The position can be specified.

Therefore, it is possible to greatly improve the efficiency of elevator abnormality diagnosis, maintenance work, and the like. In addition, even when the moving distance of the elevator is short and the observation data length is short, highly accurate analysis and abnormality diagnosis can be performed.

The abnormality determining means of this embodiment stores the acceleration signal in the car of the elevator once,
Abnormal analysis in offline processing, data measurement in real time, estimation of car speed / acceleration as a specific state quantity,
In some cases, the extended wavelet calculation and abnormality determination are all performed in real time.

In particular, regarding the application to the abnormality diagnosis of the elevator system, the correlation of the vibration spectrum included in the acceleration signal with the car position and the car speed can be clearly grasped.
Diagnosis and identification of abnormal parts can be easily performed.

Second Example Next, as a second example of the unsteady signal analyzer according to the first embodiment, a case where a monitoring target is a train will be described with reference to FIGS.

Trains and trains, such as railways, may emit abnormal vibrations or noises due to wear of wheels, distortion or deformation of rails, etc., which may cause a decrease in ride comfort, discomfort to passengers, and a train accident. Become.

Therefore, a case where the non-stationary signal analyzer is used as a train abnormality diagnosis system will be described below.

FIG. 16 is a block diagram schematically showing an unsteady signal analyzing apparatus according to this embodiment. FIG. 17 is an explanatory diagram for explaining a state where the unsteady signal analyzing apparatus according to this embodiment is mounted on a train. It is.

As shown in FIG. 16, the unsteady signal analyzing apparatus according to the present embodiment includes an acceleration sensor 30 and an acoustic sensor 31 which constitute the response data measuring means 1.
These sensors 30, 31 are attached to the train 32 as shown in FIG. The detection signal of the response data measuring means 1 comprising the acceleration sensor 30 and the acoustic sensor 31 is sent to the wavelet transform calculating means 2 where it is converted into a wavelet spectrum.

As shown in FIG. 17, a train 32 is conventionally provided with a position sensor 33 and an encoder 34. The position sensor 33 recognizes a ground identifier (mark) 35 installed on the ground. On the other hand, the encoder 34 is attached to the wheel shaft of the train 32 and detects the rotation of the wheel shaft.

Further, the unsteady signal analyzing apparatus according to the present embodiment is provided with speed / position detecting means 36 as shown in FIG. 16, and this speed / position detecting means 36 is provided with the above-described position sensor 33 and encoder. 34, based on the signal from 34, the speed and position of the train, which is a specific state quantity,
Create time-position data or time-speed data.
These time-position data or time-speed data are sent to the time-state quantity conversion table 4 and stored therein.

The wavelet spectrum calculated by the wavelet transform calculating means 2 is sent to the time coordinate non-linear transform means 3, and this time coordinate non-linear transform means 3
Based on the time-position data or the time-speed data from the time-state quantity conversion table 4, the time coordinates of the wavelet spectrum are converted into train position coordinates or train speed coordinates to generate an extended wavelet spectrum.

The conversion result of the time coordinate non-linear conversion means 3 is sent to the abnormality determination means 8, and the abnormality determination means 8 determines whether it is normal or abnormal. Here, as a method of abnormality determination, first, the extended wavelet spectrum for the position to frequency is compared with past normal spectrum data, and when the difference is equal to or larger than the threshold value, it is determined that the rail is abnormal in the line, Further, the location of the rail where the abnormality has occurred is specified.

Further, as another abnormality determination method, speed
The extended wavelet spectrum for the frequency is compared with the past normal spectrum data, and when the difference is equal to or larger than the threshold value, it is determined that the train is abnormal, and the abnormal wheel is specified.

[0138] The normal data used at the time of the abnormality judgment is separately prepared based on the data of the normal operation or the data of the train test, and is prepared in advance.

The result of the judgment by the abnormality judging means 8 is the train 32
It is sent to a display / warning device (abnormality display means) 9 provided therein, and in the case of an abnormality, an alarm is issued to the operator.
The determination result of the abnormality determining means 8 is transmitted to the receiving means 38 in the train control center by a wired or wireless communication means 37.
And sent to the display means 7 in the train control center to be displayed there.

In the above embodiment, it is assumed that the processing is executed in real time, but it is also possible to execute the processing offline instead of in real time.

Further, in the above embodiment, the unsteady signal analyzing device, which is an abnormality diagnosis system, is installed inside the train. However, the unsteady signal analyzing device is installed outside the train, and the acceleration sensor 30 and the By attaching the acoustic sensor 31 to the track side, the same function as described above can be realized.

Third Example Next, as a third example of the unsteady signal analyzer according to the first embodiment, an unsteady signal analysis tool which is a general-purpose abnormality diagnosis tool for analyzing an abnormality of an unspecified monitoring target is described. The apparatus will be described with reference to FIGS. The non-stationary signal analyzer according to the present embodiment is a portable analyzer or an abnormality diagnosis device in which a sensor, an arithmetic function, a display function, and the like are integrated.

FIG. 18 is a perspective view showing an external appearance of an unsteady signal analyzing apparatus (general-purpose abnormality diagnosis apparatus) 40 according to this embodiment. FIG. 19 is a view showing the internal system configuration of the unsteady signal analyzing apparatus. FIG.

As shown in FIGS. 18 and 19, the non-stationary signal analyzer 40 has a display unit 41 for displaying the extended wavelet spectrum data as the analysis result. The pointing device 42 allows the user to specify a specific area on the screen. As a pointing device, a mouse can be provided instead of the electronic pen.

Further, the unsteady signal analyzer 40 according to the present embodiment includes an acceleration sensor 43 serving as response data measuring means.
And an external signal input terminal 44 for taking in a state quantity signal to be monitored. Detection signal from acceleration sensor 43 and external signal input terminal 44
Is sent to a central processing unit (CPU) 45 provided inside the non-stationary signal analyzer 40. A memory 46 is connected to the CPU 45, and sensor information is stored in the memory 46,
45 performs the extended wavelet transform calculation.

FIG. 20 shows an example of a state in which the extended wavelet spectrum data obtained by the CPU 45 is displayed on the display unit 41. The horizontal axis on the screen is a specific state quantity (position, speed, etc., of the monitoring target). , The vertical axis indicates frequency, and the intensity of spectrum power is indicated by contour lines. In addition, it is also possible to display the power of the spectrum by color instead of the contour line.

Then, the user operates a pointing device 42 such as an electronic pen or a mouse to specify a specific area for which abnormality diagnosis is desired on the screen of the display unit 41 by using an area specifying line. Note that the shape of the designated area is not limited to a rectangle, but may be any shape.

When a specific area is specified on the screen of the display unit 41 in this way, only the extended wavelet data corresponding to the specified area is cut out, and the abnormality diagnosis processing is continuously performed on only the cut out data. Will be implemented. Further, after the data of the portion other than the designated area is regarded as 0, the subsequent abnormality diagnosis processing may be performed.

Next, with reference to FIG. 21, a description will be given of the abnormality diagnosis processing after area designation. In FIG.
Is an area specifying means including the pointing device 42. When the area to be diagnosed is specified by the area specifying means 47, the extended wavelet spectrum data corresponding to the specified area is extracted by the data extracting means 48, or , Data in a portion other than the designated area is set to 0.

The data processed in this way is sent to the abnormality judging means 8, and based on these data, the presence / absence of an abnormality is judged by, for example, the procedure shown in the above-mentioned equations (12) and (13). Is done. The result of the determination by the abnormality determining means 8 is sent to the display means 7 and displayed here. Note that the display unit 41 can also be used as the display unit 7.

As described above, in this embodiment, a part of the entire spectrum data displayed on the display unit 41 can be specified by the area specifying unit 47 and the data extracting unit 48. The user can determine and extract a portion exhibiting an unusual characteristic from the entire spectrum data displayed at 41, and analyze only the extracted portion by the abnormality determining means 8. For this reason, at the time of abnormality diagnosis analysis, the accuracy of abnormality determination can be improved without being affected by noise, disturbance, and other factors included in portions other than the designated area.

Second Embodiment Next, a medium on which an unsteady signal analysis program according to a second embodiment of the present invention is recorded will be described with reference to FIGS.

The medium on which the unsteady signal analysis program according to the present embodiment is recorded includes the wavelet transform calculating means 2, the time coordinate non-linear transforming means 3, and the state quantity change function setting means (time to state quantity) in the first embodiment. It is a machine-readable or computer-readable recording medium (storage medium) that stores an unsteady signal analysis program for causing a computer to realize the functions of the conversion table 4 and the state quantity estimating means 6).

Further, a program relating to the function of the abnormality determination means 8 in the first embodiment described above can be added to the non-stationary signal analysis program.

The analysis procedure by the non-stationary signal analysis program in this embodiment is the same as the analysis procedure described in the first embodiment and its modified examples, or the first to third examples of the first embodiment.

FIG. 22 is a perspective view showing a computer system for reading a program for recording an unsteady signal analysis program according to the present embodiment from a medium on which the program is recorded. The data is read out by the mounted recording medium driving device and used for analyzing an unsteady signal.

As shown in FIG. 22, the computer system 50 includes a computer main body 51 housed in a housing such as a mini tower and a display device such as a cathode ray tube (CRT), a plasma display, a liquid crystal display (LCD), and the like. 52
, A printer 53 as a recording output device, a keyboard 54a and a mouse 54b as input devices, a flexible disk drive device 56, and a CD-ROM drive device 57.

FIG. 23 is a block diagram showing a computer system for reading a non-stationary signal analysis program according to the present embodiment from a medium recording the program. As shown in FIG. 23, an internal memory 5 such as a RAM is provided in a housing in which the computer main body 51 is stored.
And an external memory such as a hard disk drive unit 58.

As shown in FIG. 22, the flexible disk 61 on which the non-stationary signal analysis program is recorded can be inserted into a slot of the flexible disk drive 56 and can be read out based on a predetermined application program. The medium on which the program is recorded is not limited to the flexible disk 61,
The D-ROM 62 may be used. The recording medium is a MO (MagnetoOptical) disk (not shown), an optical disk,
It may be a DVD (Digital Versatile Disk), a card memory, a magnetic tape, or the like.

[0160]

As described above, according to the non-stationary signal analyzer according to the present invention, it is possible to obtain the dependence of the frequency change on the specific state quantity of the monitored object and the correlation, so that the monitored object is abnormal. The condition can be diagnosed accurately.

[Brief description of the drawings]

FIG. 1 is a configuration diagram schematically showing an unsteady signal analyzer according to a first embodiment of the present invention.

2A is a graph showing a basis function of Fourier transform, and FIG. 2B is a graph showing a power spectrum by Fourier transform.

3A is a graph showing a basis function of a wavelet transform, and FIG. 3B is a graph showing a wavelet power spectrum by a wavelet transform.

FIG. 4 is a configuration diagram schematically showing a state quantity estimating unit according to a modification of the first embodiment of the present invention.

FIG. 5 is a configuration diagram schematically showing an elevator to be analyzed in a first example of the first embodiment of the present invention.

FIG. 6 is a configuration diagram showing a hardware configuration of a first example of the first embodiment of the present invention.

FIG. 7 is a flowchart showing an elevator abnormality diagnosis algorithm based on the extended wavelet transform in the first example of the first embodiment of the present invention.

FIGS. 8A, 8B, and 8C are graphs showing motor torque, car speed, and car position of the elevator when the motor shaft eccentricity is abnormal.

9 (a), (b), and (c) are graphs showing Fourier transform results of the in-car acceleration, the rotational torque fluctuation, and the in-car acceleration of the elevator when the motor shaft eccentricity is abnormal.

FIG. 10 is a graph showing a result of analyzing acceleration data in a car of an elevator when a motor shaft eccentricity is abnormal by a conventional wavelet transform method.

FIGS. 11A and 11B are graphs showing the results of an extended wavelet transform of the in-car acceleration data of the elevator when the motor shaft eccentricity is abnormal with respect to the car speed.

FIG. 12 is a graph showing a result of an extended wavelet transform of the acceleration data in the car of the elevator when the motor shaft eccentricity is abnormal with respect to the car position.

13 (a), (b), and (c) are graphs showing motor torque, car speed, and car position of an elevator when a guide rail is abnormal, respectively.

14 (a) and (b) are graphs respectively showing the in-car acceleration of the elevator and the Fourier transform result of the in-car acceleration when the guide rail is abnormal.

FIG. 15 is a graph showing an extended wavelet transform result with respect to the car position of the elevator acceleration in the car when the guide rail is abnormal.

FIG. 16 is a configuration diagram schematically showing a non-stationary signal analyzer according to a second example of the first embodiment of the present invention.

FIG. 17 is an explanatory diagram for explaining a state where the unsteady signal analyzer according to the second example of the first embodiment of the present invention is mounted on a train.

FIG. 18 is a perspective view showing an appearance of a non-stationary signal analyzer according to a third example of the first embodiment of the present invention.

FIG. 19 is a configuration diagram showing an internal system configuration of a non-stationary signal analyzer according to a third example of the first embodiment of the present invention.

FIG. 20 is a diagram showing an example of a display state of a display unit of the non-stationary signal analyzer according to the third example of the first embodiment of the present invention.

FIG. 21 is a configuration diagram schematically showing a non-stationary signal analyzer according to a third example of the first embodiment of the present invention.

FIG. 22 is an exemplary perspective view showing a computer system for reading a non-stationary signal analysis program from a medium on which the program is recorded according to a second embodiment of the present invention;

FIG. 23 is a block diagram showing a computer system for reading out a non-stationary signal analysis program according to a second embodiment of the present invention from a medium on which the program is recorded.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Response data measurement means 2 Wavelet transformation calculation means 3 Time coordinate non-linear transformation means 4 Time to state quantity conversion table 5 Input means 6 State quantity estimation means 7 Display means 8 Abnormality judgment means 10 Monitoring target 11 Output signal prediction model 12 Estimated state quantity Correction means 20, 30, 43 Acceleration sensor 21 A / D converter 22 Analysis / display device 31 Acoustic sensor 32 Train 36 Speed / position detection means 40 Unsteady signal analysis device (general-purpose abnormality diagnosis device) 41 Display unit 42 Pointing device 45 Central processing unit (CPU) 47 Area designation means 48 Data extraction means 54 Riding car 61 Flexible disk 62 CD-ROM

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification code FI G06F 17/14 G06F 15/332 S

Claims (19)

[Claims] 1. An unsteady signal analyzing apparatus for analyzing an unsteady signal generated from a monitored object, a wavelet transform calculating means for performing a wavelet transform on the unsteady signal to generate wavelet spectrum data, State quantity change function setting means for setting a state quantity change function representing a time change of a specific state quantity; and a time coordinate of the wavelet spectrum data to a coordinate of the specific state quantity by an inverse function of the state quantity change function. A non-stationary signal analyzing device, comprising: a time coordinate non-linear conversion means for performing non-linear conversion. 2. A non-stationary signal analyzer for analyzing an acceleration signal which is a non-stationary signal measured in a car of the elevator, wherein the acceleration signal is subjected to a wavelet transform. Wavelet transform calculation means for creating spectrum data; state quantity change function setting means for setting a state quantity change function representing a time change of a vertical position or a vertical speed as a specific state quantity of the car; and the state quantity. A non-stationary signal analysis device, comprising: a time coordinate non-linear conversion unit that non-linearly converts the time coordinate of the wavelet spectrum data into the coordinate of the elevation position or the elevation speed using an inverse function of a change function. 3. The time coordinate non-linear transform means includes an extended wavelet transform formula 3. The non-stationary signal analyzer according to claim 1, wherein spectral data obtained by nonlinearly transforming the time coordinate of the wavelet spectrum data into the coordinate of the specific state quantity is calculated. 4. 4. The time coordinate non-linear transformation means divides the wavelet spectrum data for each time, and based on a data table storing a relationship between time and the specific state quantity or the state quantity change function, By rearranging the divided data in order of the state quantity and performing interpolation or smoothing processing between the data, it is possible to calculate spectrum data obtained by nonlinearly transforming the time coordinate of the wavelet spectrum data into the coordinate of the specific state quantity. The non-stationary signal analyzer according to claim 1 or 2, wherein: 5. The non-stationary signal analysis device according to claim 1, further comprising a response data measuring unit for measuring the non-stationary signal. 6. The apparatus according to claim 1, wherein said state quantity change function setting means estimates the state quantity change function based on measurement data relating to a state quantity other than the specific state quantity. Item 6. The non-stationary signal analyzer according to any one of items 5. 7. The non-stationary signal analyzer according to claim 6, wherein the measurement data relating to a state quantity other than the specific state quantity is measurement data relating to the non-stationary signal. 8. The state quantity change function setting means uses a state observer or a Kalman filter based on a dynamic characteristic model of the monitored object to calculate a time change of the specific state quantity over a state quantity other than the specific state quantity. The non-stationary signal analysis device according to claim 6 or 7, wherein the state quantity change function is estimated by estimating based on the measurement data. 9. The state quantity change function setting means according to claim 1, wherein said state quantity change function is obtained based on measurement data of said specific state quantity. The non-stationary signal analyzer according to claim 1. 10. The apparatus according to claim 1, wherein the state quantity change function setting means uses the state quantity change function obtained in advance. Transient signal analyzer. 11. The apparatus according to claim 1, further comprising display means for displaying an analysis result of said time-coordinate non-linear conversion means in a coordinate system having at least coordinates of said specific state quantity and coordinates of frequency. The non-stationary signal analyzer according to any one of claims 10 to 10. 12. An unsteady signal analyzing apparatus according to claim 11, further comprising an abnormality judging means for judging whether or not an abnormality has occurred in said monitoring target based on an analysis result of said time coordinate nonlinear conversion means. . 13. An area designating means for designating a specific area in the entire display with respect to a display by the display means for spectrum data as an analysis result of the time coordinate non-linear transformation means, 13. The non-stationary signal analyzer according to claim 12, further comprising: data extracting means for extracting spectrum data corresponding to the area designated by the means and sending the data to the abnormality determining means. 14. The non-stationary signal analyzer according to claim 12, wherein a result of the judgment by said abnormality judging means is displayed on said display means. 15. The non-stationary signal analyzer according to claim 12, further comprising an abnormality display unit that displays a result of the determination by the abnormality determination unit. 16. A medium on which a program for analyzing a non-stationary signal generated from a monitoring target by a computer is recorded. The non-stationary signal analysis program stores the non-stationary signal in a wavelet spectrum by performing a wavelet transform on the non-stationary signal. A wavelet transform calculation function for creating data; a state quantity change function setting function for setting a state quantity change function representing a time change of a specific state quantity in the monitoring target; and the wavelet by an inverse function of the state quantity change function. A non-stationary signal analysis program for realizing a time coordinate non-linear conversion function of non-linearly converting a time coordinate of spectrum data into a coordinate of the specific state quantity. 17. The monitoring target is an elevator, the unsteady signal is an acceleration signal measured in a car of the elevator, and the specific state quantity is an ascending position or a descending speed of the car. 17. A medium on which the non-stationary signal analysis program according to claim 16 is recorded. 18. The method according to claim 17, wherein the time-coordinate non-linear transformation means uses an extended wavelet transformation formula. 18. The medium storing the non-stationary signal analysis program according to claim 16 or 17, wherein spectrum data obtained by nonlinearly transforming the time coordinate of the wavelet spectrum data is calculated. 19. The time coordinate non-linear transformation means divides the wavelet spectrum data for each time, and based on a data table storing a relationship between time and the specific state quantity or based on the state quantity change function. The spectrum data obtained by non-linearly converting a time coordinate into the coordinate of the specific state quantity is calculated by rearranging the divided data in order of a state quantity and performing interpolation or smoothing processing between the respective data. A medium on which the non-stationary signal analysis program according to claim 16 or 17 is recorded.