JP2008014679A - Facility diagnostic method, facility diagnostic system, and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a highly accurate facility diagnostic method and a facility diagnostic system without requiring to provide a number of sensors for every facility. <P>SOLUTION: The facility diagnostic system includes a data acquisition part 21 which acquires time-series data which is a variable showing a condition of a facility of a diagnostic target and fluctuates with time; a translational deviation operation part 24 which calculates a value showing determinism to be an index of whether the time-series data in the facility is deterministic or stochastic; and a failure determining part 26 which determines that the facility is in a condition to be warned when the value showing determinism calculated in the translational deviation operation part 24 changes exceeding a prescribed threshold. The value showing determinism is either a translational deviation or a permutation entropy, and a permutation entropy operation part is equipped when the permutation entropy is used as a value to show determinism. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a facility diagnosis method, a facility diagnosis system, and a computer program for diagnosing the state of a facility.

  In large-scale factories and the like, many mechanical facilities such as manufacturing facilities and air conditioning facilities are installed. In order to find these abnormalities, skilled maintenance personnel patrol and listen to the sound with their ears, or visually or palpate the state of vibration to diagnose the state of the equipment.

  Therefore, for example, in order to accurately detect various abnormalities of an air conditioning fan or pump and prevent a stop due to a failure, a non-contact type equipment diagnosis method using an acoustic method has been proposed (Patent Document 1). In the technique of Patent Document 1, when detecting an abnormal signal by comparing a normal sound pressure signal measured in advance with a sound pressure signal at the time of measurement, the signal is converted into a low frequency region and a member corresponding to the rotation frequency. After separating into a high frequency range corresponding to the natural frequency of the sound, using a filter based on the AR model (Auto-Regressive (auto-regressive) model) to which the linear prediction method is applied, the sound pressure signal at the time of measurement is normal Abnormalities of the fan and the pump are detected based on the value obtained by removing the characteristic of the pressure signal.

  On the other hand, there is a method for analyzing time series data based on nonlinear dynamic theory using recent chaotic time series analysis. For example, Patent Document 2 aims to provide a system suitable for multifaceted and comprehensive analysis of various types of time series data such as power consumption, gas consumption, and physical flow, such as short-term prediction. The technology of Patent Document 2 is based on orbit data in the state space reconstructed by time series data embedding processing, orbit display analysis, dimension analysis, Lyapunov spectrum analysis, entropy analysis, deterministic nonlinear prediction analysis, etc. The present invention relates to a time-series data non-linear analysis system that performs multi-faceted analysis and comprehensively processes each result using each other.

In addition, a technique of an ultrasonic diagnostic apparatus that outputs diagnostic information that accurately reflects the state of a complex and delicate tissue such as a living body has been proposed (Patent Document 3). In the technique of Patent Literature 3, a position vector of a reference point randomly selected on an attractor is set as a target vector, and K neighboring points of the target vector are searched to set those position vectors. For each of the reference point and the K neighboring points, a position vector after T steps (after the elapse of T time) is detected. Then, for each of the reference point and the K neighboring points, a translation vector indicating how much it has moved after T steps is calculated, and a translation error (vector) is calculated from the translation vector of the reference point and the translation vectors of the K neighboring points. Variance).
JP-A-10-133740 JP-A-6-96055 JP 2005-95327 A

  In the conventional equipment diagnosis method, since diagnosis is performed by a maintenance staff patrol, it is difficult to automate maintenance and save labor. Moreover, it is necessary to provide many sensors for every installation for vibration detection. In many cases, regular maintenance is performed in which the maintenance time is not predicted in advance or detected in accordance with the state of the equipment, but is determined at a period recommended by the equipment manufacturer.

  In the technique of Patent Document 1, it is necessary to adjust the frequency and detection level of abnormal sound according to the difference in characteristics for each facility. In addition, when there is no characteristic difference between the spectral distributions of the normal sound and the abnormal sound, it is difficult to detect the abnormality.

  The technique of Patent Document 2 relates to a time-series data nonlinear analysis system, but it is not clear how to perform short-term prediction from time-series data. Moreover, it does not describe how to diagnose the equipment. Patent Document 3 is a technique of an ultrasonic diagnostic apparatus such as a living body, but does not describe the relationship between time-series data of equipment and the state of equipment.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a highly accurate equipment diagnosis method and equipment diagnosis system that do not require many sensors for each equipment.

The facility diagnosis method according to the first aspect of the present invention includes:
A data acquisition step for acquiring time-series data that is a variable indicating the state of equipment to be diagnosed and fluctuates with time;
A deterministic calculation step for calculating a value representing determinism that serves as an indicator of whether the time series data in the facility is deterministic or probabilistic from the time series data acquired in the data acquisition step. When,
A warning determination step for determining that the equipment should be in a warning state when a value representing the determinism calculated in the deterministic calculation step changes beyond a predetermined threshold;
It is characterized by providing.

In particular, the value representing the determinism is a translation error calculated from the time series data,
The deterministic calculation step includes
An embedding step for calculating an embedding vector of a certain dimension from the time series data;
A nearest neighbor vector extracting step of extracting a predetermined number of closest vectors for a certain embedded vector among the embedded vectors calculated in the embedding step;
A translation error calculating step for calculating a translation error that is a variance of a predetermined number of the closest vectors extracted in the nearest vector extraction step;
Including
The warning determination step includes:
When the translation error in the normal state of the equipment is larger than the predetermined threshold, the equipment warns when the translation error calculated in the translation error calculation step becomes smaller than the predetermined threshold. When the translation error in the normal state of the equipment is smaller than the predetermined threshold value, the translation error calculated in the translation error calculation step is greater than the predetermined threshold value. When it becomes larger, it is determined that the equipment should be warned.
It is characterized by that.

Or, the value representing the determinism is permutation entropy calculated from the time series data,
The deterministic calculation step includes
An embedding step for calculating an embedding vector of a certain dimension from the time series data;
For all the embedding vectors calculated from the time-series data at the predetermined time calculated in the embedding step, the frequencies in the same order as the order of the magnitude relationship of the elements of the embedding vector are accumulated, and the predetermined time A realization frequency calculating step for calculating as a relative realization frequency with respect to the number of all the embedded vectors calculated from the time series data in
A permutation entropy calculating step for calculating permutation entropy, which is an entropy having a probability distribution of the relative realization frequencies calculated in the realization frequency calculation step, with all permutations consisting of the order of the number of dimensions of the embedded vector as random variables; ,
Including
The warning determination step includes:
When the permutation entropy calculated in the permutation entropy calculating step is smaller than the predetermined threshold when the permutation entropy in the normal state of the equipment is larger than the predetermined threshold, the equipment warns When the permutation entropy in the normal state of the equipment is smaller than the predetermined threshold value, the permutation entropy calculated in the permutation entropy calculation step is greater than the predetermined threshold value. When it becomes larger, it is determined that the equipment should be warned.
It is characterized by that.

  Preferably, the variable indicating the state of the equipment to be diagnosed is a temperature sensor, a temperature / humidity sensor, a wind speed sensor, a liquid flow sensor, an optical fiber sensor, a pressure sensor, a rotation sensor, an acceleration sensor, an acoustic sensor, or a vibration sensor. It includes at least one of time-series data.

The equipment to be diagnosed includes a rotating machine,
The data acquisition step acquires an acoustic signal as time-series data from the equipment with an acoustic sensor,
The failure determination step determines a state of the rotating machine that should be warned.
This may be a feature.

The facility diagnosis system according to the second aspect of the present invention is:
A data acquisition means for acquiring time-series data that is a variable indicating a state of equipment to be diagnosed and fluctuates with time;
Deterministicity calculating means for calculating a value representing determinism as an index of whether the time series data in the facility is deterministic or probabilistic from the time series data acquired by the data acquisition means When,
Warning determination means for determining that the equipment should be warned when a value representing determinism calculated by the deterministic calculation means changes beyond a predetermined threshold;
It is characterized by providing.

In particular, the value representing the determinism is a translation error calculated from the time series data,
The deterministic calculation means is
Embedding means for calculating an embedding vector of a certain dimension from the time series data;
Of the embedded vectors calculated by the embedding means, a nearest neighbor vector extracting means for extracting a predetermined number of closest vectors for a certain embedded vector;
A translation error calculating means for calculating a translation error which is a variance of a predetermined number of closest vectors extracted by the nearest neighbor vector extracting means;
Including
The warning determination means includes
When the translation error in the normal state of the equipment is larger than the predetermined threshold, the equipment warns when the translation error calculated by the translation error calculation means becomes smaller than the predetermined threshold. When the translation error in the normal state of the equipment is smaller than the predetermined threshold value, the translation error calculated by the translation error calculation means is greater than the predetermined threshold value. When it becomes larger, it is determined that the equipment should be warned.
It is characterized by that.

Or, the value representing the determinism is permutation entropy calculated from the time series data,
The deterministic calculation means is
Embedding means for calculating an embedding vector of a certain dimension from the time series data;
For all the embedding vectors calculated from the time-series data at the predetermined time calculated by the embedding means, the frequencies in the same order as the order of the magnitude relation of the elements of the embedding vector are accumulated, and the predetermined time Realization frequency calculation means for calculating as relative realization frequency for the number of all the embedded vectors calculated from the time series data in
Permutation entropy calculating means for calculating permutation entropy, which is entropy with all the permutations consisting of the order of the number of dimensions of the embedded vector as random variables and the relative realization frequency calculated by the realization frequency calculating means as a probability distribution; ,
Including
The warning determination means includes
When the permutation entropy calculated by the permutation entropy calculating means is smaller than the predetermined threshold when the permutation entropy in the normal state of the equipment is larger than the predetermined threshold, the equipment warns Or when the permutation entropy in the normal state of the equipment is smaller than the predetermined threshold, the permutation entropy calculated by the permutation entropy calculating means is greater than the predetermined threshold. When it becomes larger, it is determined that the equipment should be warned.
It is characterized by that.

  Preferably, the variable indicating the state of the equipment to be diagnosed is a temperature sensor, a temperature / humidity sensor, a wind speed sensor, a liquid flow sensor, an optical fiber sensor, a pressure sensor, a rotation sensor, an acceleration sensor, an acoustic sensor, or a vibration sensor. It includes at least one of time-series data.

The equipment to be diagnosed includes a rotating machine,
The data acquisition means acquires an acoustic signal as time series data from the equipment with an acoustic sensor,
The failure determination means determines a state of the rotating machine to be warned;
This may be a feature.

A computer program according to the third aspect of the present invention provides:
A data acquisition means for acquiring time-series data that is a variable indicating a state of equipment to be diagnosed and fluctuates with time;
Deterministicity calculating means for calculating a value representing determinism of the time series data in the equipment from the time series data acquired by the data acquisition means,
Warning determination means for determining that the equipment should be in a warning state when a value representing determinism calculated by the deterministic calculation means has changed beyond a predetermined threshold;
It is made to function as.

  In the present invention, the equipment should be warned is in a state where the equipment needs inspection / maintenance, a state where it is time to replace parts or units for preventive maintenance, or some trouble has occurred. Or the state that is out of order.

  According to the equipment diagnosis method and the equipment diagnosis system of the present invention, even if the power spectrum is the same for different time series data at first glance, the deterministic structure behind the time series data is quantified, and the time delay By reconstructing the trajectory in the phase space from the time series data by coordinate transformation, the difference between the time series data at the normal time and the time of the failure can be quantitatively expressed. Further, it is possible to perform fault diagnosis with high accuracy even with a relatively small number of data. Moreover, it is not necessary to install many sensors for every installation.

(Embodiment 1)
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals, and description thereof will not be repeated. An explanation will be given below by taking as an example a facility diagnosis system that targets a compressor such as a semiconductor manufacturing apparatus.

  FIG. 1 is a block diagram showing a logical configuration of equipment diagnosis system 1 according to Embodiment 1 of the present invention. The equipment diagnosis system 1 includes a data acquisition unit 21, an embedded vector generation unit 22, a proximity vector extraction unit 23, a translation error calculation unit 24, a median value calculation / average processing unit 25, a data holding unit 5, a failure determination unit 26, a display The processing unit 27, the display device 7, the printer device 8, and the like are included. The data holding unit 5 stores and holds time-series data representing the state of the equipment that is the target of failure diagnosis, embedded vectors, nearest neighbor vectors, and translation errors.

  Here, the translation error and the fact that it becomes an index representing determinism will be described. Using the time series analysis algorithm of Wayland et al. (R. Wayland, D. Bromley, D. Pickett, and A. Passamante, Physical Review Letters, Vol. 70, pp. 580-582, 1993.) The degree of deterministic aspect can be quantitatively evaluated.

From the time series data {r (ti)} (i = 0,..., N−1), the embedded vector r (ti) = {r (ti), r (ti−Δt),. . . , R (ti- (n-1) [Delta] t)} ^T
Is generated. The superscript ^{T on the} right shoulder represents a transposed matrix. n is an appropriate embedding dimension. Δt is, for example, an appropriate time difference selected from the mutual information amount.

From the set of embedding vectors, K nearest vectors of a certain embedding vector r (t ₀ ) are extracted. The distance between vectors is measured by the Euclidean distance. Write the K nearest neighbor vectors as r (tj) (j = 0, ..., K). For each r (tj), the vector after elapse of time TΔt is r (tj + TΔt). At this time, the change in the trajectory of the embedded vector over time is
v (tj) = r (tj + T.DELTA.t) -r (tj) (1)
Is approximately given by If the time evolution looks deterministic, the close part of each trajectory group of the proximity vector will be moved to the close part after TΔt. Therefore, the variance in the direction of v (tj) is an index for quantitatively evaluating how deterministic the observed time evolution looks. The variance in the direction of v (tj) is given by

Etrans is called a translation error. In order to suppress the error of Etrans arising from the selection of r (t _0), the operation of finding the median of Etrans relates randomly selected the M r (t ₀₎ repeated Q times, and the average of Q median Etrans is evaluated by value. Etrans → 0 as the deterministic aspect of time series data increases. If the time-series data is white noise, the difference vector v (tj) is uniformly distributed, so Etrans as the median value is close to 1. If the time-series data is a stochastic process having a strong linear correlation, the direction of the adjacent trajectory groups is aligned to some extent due to autocorrelation, so Etrans takes a value smaller than 1.

  According to numerical experiments, Etrans> 0.5. In the range of 0.1 <Etrans <0.5, the process may be a stochastic process or may be a deterministic time series contaminated with observation noise. If Etrans <0.1, the stochastic process cannot be explained, and the deterministic aspect is fully recognized.

  In the present embodiment, the failure of the facility is diagnosed by utilizing the fact that the translation error Etrans becomes an index representing the determinism of the time-series data indicating the state of the facility to be diagnosed. If the time series data in the normal state of the equipment is stochastic, that is, the translation error is a large value, the equipment will fail when the translation error becomes a small value exceeding a predetermined threshold. Judge that For example, in a facility that controls to keep temperature and rotation speed constant, the temperature and rotation speed, which are time series data, vary irregularly due to various disturbances and are considered probabilistic. When the deterministic aspect of temperature and rotational speed fluctuations increases, it can be determined that some fixed factor (failure) has occurred.

  On the contrary, when the time series data in the normal state of the equipment is deterministic, that is, when the translation error is a small value, the translation error exceeds a predetermined threshold value and becomes a large value (approaching 1). It is determined that the equipment has failed. For example, when the processing stage is moved up and down, time series data such as sound, vibration, and displacement are governed by the lift control operation, and thus are considered deterministic. Therefore, when the stochastic aspect increases in time series data such as sound, vibration, and displacement, it can be determined that an external factor (failure) other than control has occurred.

  Since the translation error is a continuously changing value, set the threshold appropriately and the equipment to be diagnosed needs to be inspected / maintained, for example, cleaning or lubrication. It is also possible to distinguish a state in which wear or the like has progressed and it is better to replace parts or the like in a preventive maintenance manner or a state in which a failure has occurred, although the failure has not been reached. In the present invention, these states are collectively referred to as a state to be warned. The actual warning condition (how to set the threshold value) is selected in consideration of the use of the equipment to be diagnosed, the influence of the failure, or economic characteristics.

  Returning to FIG. 1, the operation of each part of the equipment diagnosis system 1 will be described. The data acquisition unit 21 collects time-series data indicating the state of the equipment to be diagnosed and stores the collected time-series data 51 in the data holding unit 5. Examples of the time series data include temperature, flow rate, liquid level, pressure, rotational speed, displacement, sound, and vibration. Time-series data is sampled and input from various sensors (not shown) at an appropriate period.

  The embedded vector generation unit 22 generates the aforementioned embedded vector r (ti) from the collected time series data 51. The dimension n and the time difference Δt of the embedded vector are set in advance according to the characteristics of the time series data. The embedded vector generating unit 22 stores the generated set of embedded vectors in the data holding unit 5 (embedded vector data 52).

The proximity vector extraction unit 23 selects an arbitrary embedding vector r (t ₀ ) from the set of embedding vectors, and K embedding vectors (k ₀ ) closest to the embedding vector r (t ₀ ) selected from the embedding vector set. The closest vector) r (tj) (j = 0, 1,..., K) is extracted. For the M embedding vectors selected at random, the nearest vector is extracted and stored in the data holding unit 5 (nearest vector data 53). The number of adjacent vectors K and the number of selections M are set in advance according to the properties of the time-series data in order to suppress statistical errors in translation error calculation. Further, random selection of M embedding vectors and their closest vector extraction are repeated Q times.

  The translation error calculation unit 24 calculates a translation error Etrans, which is a variance in those directions, from the set of closest vectors. The translation error Etrans is calculated for the nearest vector of M embedding vectors selected at random. Further, a translation error Etrans is calculated for a set of Q nearest M vector embedded vectors, and these values are stored in the data holding unit 5 (translation error data 54).

  The median value calculation / average processing unit 25 obtains the median value of M translation errors for each time, and calculates the average of the median values for each of Q repetitions. The average value of the translation error is input to the failure determination unit 26. The failure determination unit 26 compares the predetermined threshold value with the average value of the translation error, and determines that the target equipment is in a state to warn when the average value of the translation error changes beyond the threshold value. .

  As described above, when the translation error in the normal state is probabilistic, it is determined that the state should be warned when the average value of the translation error becomes smaller than the threshold value. When the translation error in the normal state is deterministic, it is determined that the state should be warned when the average value of the translation error becomes larger than the threshold value.

  The display processing unit 27 displays, for example, the transition of the average value of translation errors and the failure determination result on the display device 7. At the same time, the data is output to the printer 8. The display processing unit 27 may display the time series data and the translation error together. When it is determined that a failure has occurred, an alarm display such as blinking a light or sounding a buzzer may be performed.

  FIG. 2 is a block diagram illustrating an example of a physical configuration of the equipment diagnosis system 1. As shown in FIG. 2, the facility diagnosis system 1 of the present invention shown in FIG. 1 includes a control unit 11, a main storage unit 12, an external storage unit 13, an operation unit 14, a screen display unit 15, and a print output unit. 16, a transmission / reception unit 17, a display device 7 and a printer device 8. The control unit 11, main storage unit 12, external storage unit 13, operation unit 14, screen display unit 15, print output unit 16, and transmission / reception unit 17 are all connected to the control unit 11 via the internal bus 10.

  The control unit 11 includes a CPU (Central Processing Unit) and the like, and according to a program stored in the external storage unit 13, a data acquisition unit 21, an embedded vector generation unit 22, a proximity vector extraction unit 23, and a translation error calculation unit 24. The median calculation / average processing unit 25, the failure determination unit 26, and the display processing unit 27 are executed. The data acquisition unit 21, the embedded vector generation unit 22, the proximity vector extraction unit 23, the translation error calculation unit 24, the median value calculation / average processing unit 25, the failure determination unit 26, and the display processing unit 27 are connected to the control unit 11 and above Realized by a program executed in

  The main storage unit 12 includes a RAM (Random-Access Memory) and the like, and is used as a work area for the control unit 11. The data holding unit 5 is stored and held in a part of the main storage unit 12 as a storage area structure.

  The external storage unit 13 includes a nonvolatile memory such as a flash memory, a hard disk, a DVD (Digital Versatile Disc), a DVD-RAM (Digital Versatile Disc Random-Access Memory), a DVD-RW (Digital Versatile Disc ReWritable), and the like. A program for causing the control unit 11 to perform the above process is stored in advance, and data of the program is supplied to the control unit 11 according to an instruction from the control unit 11, and the data supplied from the control unit 11 is stored. For example, the time series data may be stored in the external storage unit 13.

  The operation unit 14 includes a pointing device such as a key switch, a jog dial, a keyboard and a mouse, and an interface device for connecting them to the internal bus 10 in order for the operator to give a command to the equipment diagnosis system 1. Commands such as failure determination condition input, time-series data input, failure diagnosis start, and the like are input via the operation unit 14 and supplied to the control unit 11. In addition, the embedding dimension n, the time difference Δt, the number K of adjacent vectors, the number M of embedding vectors, the number Q of repetition operations, and the like are input from the operation unit 14 and set.

  The display device 7 is composed of a CRT (Cathode Ray Tube) or LCD (Liquid Crystal Display), etc., and in accordance with a command from the control unit 11 in accordance with a command input from the operation unit 14, time series data, translation error transition, Display fault diagnosis results. The screen display unit 15 converts screen data to be displayed on the display device 7 into a signal for driving the display device 7.

  The print output unit 16 includes a serial interface, a parallel interface, or a LAN (Local Area Network) interface connected to the printer device 8. The control unit 11 outputs display data to be printed to the printer device 8 via the print output unit 16.

  The transmission / reception unit 17 includes a modem or network termination device, and a serial interface or LAN (Local Area Network) interface connected thereto. The control unit 11 inputs time series data from each sensor via the transmission / reception unit 17. The time series data may be stored in another server (not shown). In that case, the control unit 11 receives time-series data from a server or the like via the transmission / reception unit 17 via a network (not shown).

  FIG. 3 is a diagram illustrating an example of the configuration of the equipment diagnosis system 1 when the compressor 3 is a diagnosis target. A microphone 2 for inputting the sound of the compressor 3 is connected to the equipment diagnosis system 1. The equipment diagnosis system 1 samples and collects the acoustic signals input from the microphone 2 at a certain period, and sets the time series data.

  If the compressor 3 is in steady operation, the time series data of the acoustic signal is considered to be stochastic when it is normal. Therefore, it can be determined that the compressor 3 should be warned when the translation error calculated from the time-series data of the acoustic signal becomes smaller than a certain threshold value.

  Next, the operation of the equipment diagnosis system 1 will be described. As described above, the operation of the equipment diagnosis system 1 is performed by the control unit 11 in cooperation with the main storage unit 12, the external storage unit 13, the operation unit 14, the screen display unit 15, the print output unit 16, and the transmission / reception unit 17. Do it.

  FIG. 4 is a flowchart illustrating an example of an operation of facility diagnosis using a translation error. First, time-series data indicating the state of equipment to be diagnosed is input (step A1). In the example of the compressor 3 in FIG. 3, the acoustic signal input from the microphone 2 is sampled and collected at a certain period to obtain time series data.

From the collected time series data, an embedding vector r (ti) is generated with a set embedding dimension n and a time difference Δt (step A2). An arbitrary embedding vector r (t ₀ ) is selected from the set of embedding vectors, and the nearest vector r (t j) (j = 0,..., K) is extracted (step A3). The closest vector is extracted for M randomly selected embedding vectors, and the random selection of M embedding vectors and their closest vector extraction are repeated Q times.

  A translation error Etrans is calculated for the nearest vector of M embedding vectors selected at random, and a translation error is calculated for a set of M nearest embedding vectors of M times (step A4). . Then, the median value of the translation error Etrans is obtained for the nearest vector of the M embedding vectors, and the Q average value is obtained for the median value of the set of M translation errors (step A5).

  A failure is determined from the average value of translation errors (step A6). FIG. 5 is a flowchart illustrating an example of a failure determination operation. The criterion for determining the failure is changed depending on whether the judgment value of the normal state (in this case, the translation error) is probabilistic (that is, larger than the threshold value) or deterministic (that is, smaller than the threshold value) (step) B1).

  If the judgment value in the normal state is larger than the threshold value (step B1;> threshold value), if the current judgment value (average value of translation error) is less than the threshold value (step B2; Yes), a failure (Step B3). When the determination value is larger than the threshold value (step B2; No), it is not determined as a failure.

  If the judgment value in the normal state is smaller than the threshold value (step B1; <threshold value), if the current judgment value (average value of translation error) is equal to or greater than the threshold value (step B4; Yes), a failure occurs. (Step B5). When the determination value is smaller than the threshold value (step B4; No), it is not determined as a failure.

  In steps B2, B3, B4, and B5, as described above, a plurality of threshold values are set, and a state where inspection / maintenance is necessary, a state where parts replacement is necessary for preventive maintenance, or a state where a failure occurs You may make it distinguish.

  Returning to the flowchart of FIG. 4, when it is determined as a failure as a result of the failure determination (step A7; Yes), it is displayed on the display device 7 that the facility is failed (step A8). At the same time, it may be printed out to the printer 8. If it is not determined that there is a failure (step A7; No), it is not displayed as a failure. When the determination value tends to approach the failure determination threshold value from a normal value, an alarm prompting inspection / maintenance may be displayed.

(Example 1)
FIG. 6 is a graph showing an example of time-series data of the acoustic signal of the compressor 3. FIG. 6A is an acoustic signal before the bearing replacement in the failure mode, and FIG. 6B is an acoustic signal in a normal state after the bearing replacement. There is no characteristic difference between the two power spectra, for example, the spectrum at a certain frequency is large or small. In power spectrum analysis, it is difficult to clearly distinguish between failure and normality.

  FIG. 7 is a graph showing an example of the translation error of the sound data of the compressor 3. 6 shows the result of calculating the translation error for the acoustic signals before and after the bearing replacement in FIG. 6 by changing the embedding dimension with the embedding dimension on the horizontal axis and the translation error on the vertical axis. In any case, the time difference Δt is 10 sampling periods. 7A shows a translation error for an acoustic signal after bearing replacement (After), and B shows a translation error for an acoustic signal before bearing replacement (Before).

  In the range where the embedding dimension is small, the difference between A and B is small, but when the embedding dimension is 5 or more, the difference between A and B is clear. An appropriate threshold value is set, and it can be determined that a normal operation is obtained when the translation error is larger than the threshold value, and a failure is found when the translation error is less than the threshold value. In this example, it is desirable that the embedding dimension is 6 or more.

  According to the facility diagnosis method and the facility diagnosis system 1 of the present invention, the deterministic structure behind the time-series data is quantified even in the case where the power spectrum is the same for different time-series data at first glance. By reconstructing the trajectory in the phase space from the time series data by the delayed coordinate transformation, the difference between the time series data at the normal time and the time of the failure can be quantitatively expressed. Further, it is possible to perform fault diagnosis with high accuracy even with a relatively small number of data.

(Embodiment 2)
Next, the equipment diagnosis system 1 when permutation entropy is used as a value representing determinism will be described. FIG. 8 is a block diagram showing a logical configuration of the equipment diagnosis system 1 according to Embodiment 2 of the present invention.

  The equipment diagnosis system 1 includes a data acquisition unit 21, an embedded vector generation unit 22, a realization frequency calculation unit 28, a permutation entropy calculation unit 29, an average processing unit 30, a data holding unit 5, a failure determination unit 26, a display processing unit 27, The display device 7 and the printer device 8 are configured. The data holding unit 5 stores and holds collected time-series data 51, embedded vector data 52, permutation realization frequency data 55, and permutation entropy data 56 representing the state of the equipment that is the target of failure diagnosis.

  Here, the permutation entropy and the fact that it becomes an index representing determinism will be described. Permutation entropy (C. Bandt and B. Pompe, Physical Review Letters, Vol. 88, pp. 174102-1-174102-4, 2002.) introduced by Bandt and Pompe is the Kolmogorov-Sinai Although it is an asymptotically equivalent quantity to entropy, permutation entropy is defined as:

  All embedded vectors r (ti) of a certain dimension n are generated from time-series data at a predetermined time. For each embedding vector, the elements are numbered in ascending or descending order according to the magnitude relationship of the embedding vector elements. The array of numbers in ascending or descending order of elements is a permutation of the number of elements. For all the embedding vectors at a predetermined time, the number of embedding vectors having the same ascending or descending permutation is counted. The total number is the realization frequency of the permutation having the order. The realization frequency with respect to the number of all embedding vectors is a relative realization frequency. The sum of the relative realization frequencies is 1. The time difference Δt of the embedding vector may be 1. When Δt = 1, the embedding vector r (ti) is composed of n consecutive elements of time series data.

The order of the magnitude relation of the elements of the embedding vector is a permutation composed of the order of the number of dimensions of the embedding vector. For the number n of dimensions of the embedding vector, a set of n permutations from 1 to n is denoted by Π, and an element (permutation) of the permutation set is denoted by π. When the number of time-series data at a predetermined time is N, the embedding dimension is n, and the time difference is Δt, the number of embedding vectors generated from the time-series data at a predetermined time is N− (n−1) Δt. When the realization frequency of a certain permutation is written as m (π), the relative realization frequency p (π) of a certain permutation π is expressed by the following equation (4).

The relative realization frequency p (π) is equivalent to coarsely classifying the complexity of time variation into patterns. Considering the relative realization frequency p (π) as the realization probability of the permutation π and calculating the information entropy, the complexity (determinism) of the original time series can be quantitatively evaluated. An entropy having a permutation π of the number of dimensions of the embedded vector as a random variable and a relative realization frequency p (π) as a probability distribution is referred to as a permutation entropy. The permutation entropy is defined by the following equation (5).


However, the term of p (π) = 0 is not counted.

  Permutation entropy can quantitatively evaluate the complexity (determinism) of the original time series. The simplest behavior is a monotonic process. In a monotonic increasing process or a monotonic decreasing process, the permutation entropy is minimized. On the other hand, the most complicated behavior is a completely random process. In this case, all possible patterns are realized, so the permutation entropy is maximized.

π is a permutation of the embedding dimension n, and the permutation set Π is n! Since it includes a number of elements (permutations), 0 ≦ H (n) ≦ log ₂ n! It is. The lower limit corresponds to a monotone increasing process or a monotonic decreasing process. The upper limit represents a completely random process.

Bandt and Pompe paid attention to the fact that H (n) increases linearly with respect to n, and introduced the amount defined by the following equation (6).

h (n) is log ₂ n! It is convenient to use the entropy defined by the following equation (7).


0 ≦ h ^* (n) ≦ 1 holds. As the deterministic aspect of time series data increases, h ^* (n) → 0. If the time series data is white noise, h ^* (n) is a value close to 1.

In this embodiment, permutation entropy H (n) or h ^* (n) obtained by normalizing H (n) is used as an index representing the determinism of time-series data indicating the state of equipment to be diagnosed. Then, the failure of the equipment is diagnosed. If the time series data in the normal state of the equipment is stochastic, i.e. the permutation entropy is a large value, the equipment will fail when the permutation entropy falls below a predetermined threshold. Judge that For example, in a facility that controls to keep temperature and rotation speed constant, the temperature and rotation speed, which are time series data, vary irregularly due to various disturbances and are considered probabilistic. When the deterministic aspect of fluctuations in temperature and rotational speed increases (permutation entropy decreases), it can be determined that some fixed factor (failure) has occurred.

On the other hand, when the time series data in the normal state of the equipment is deterministic, that is, the permutation entropy is a small value, the permutation entropy exceeds the predetermined threshold value (h ^* (n) It is determined that the equipment has failed. For example, when the processing stage is moved up and down, time series data such as sound, vibration, and displacement are governed by the lift control operation, and thus are considered deterministic. Therefore, when the stochastic aspect increases (time permutation entropy increases) in time series data such as sound, vibration, and displacement, it can be determined that an external cause (failure) other than control has occurred.

  Since the permutation entropy is a continuously changing value, like the translation error in the first embodiment, the threshold value is set appropriately, and the equipment to be diagnosed requires inspection and maintenance. It is also possible to distinguish a state in which it is better to replace parts or the like in a preventive maintenance manner, or a state in which a failure has occurred. The actual warning condition (how to set the permutation entropy threshold) depends on the use of the equipment to be diagnosed, the impact of the failure, or economic characteristics. select.

  Returning to FIG. 8, the operation of each part of the equipment diagnosis system 1 will be described. The data acquisition unit 21 collects time-series data indicating the state of the equipment to be diagnosed and stores the collected time-series data 51 in the data holding unit 5. Examples of the time series data include temperature, flow rate, liquid level, pressure, rotational speed, displacement, sound, and vibration. Time-series data is sampled and input from various sensors (not shown) at an appropriate period.

  The embedded vector generation unit 22 generates the above-described embedded vector r (ti) from the time series data. The dimension n and the time difference Δt of the embedded vector are set in advance according to the characteristics of the time series data. The embedded vector generating unit 22 stores the generated set of embedded vectors in the data holding unit 5 (embedded vector data 52).

  The realization frequency calculation unit 28 orders the elements according to the magnitude relation of the elements of the embedding vector, and totals the number of embedding vectors having the same order for all the embedding vectors for a predetermined time. Totalized permutation realization frequencies are stored in the data holding unit 5 (permutation realization frequency data 55).

The permutation entropy calculation unit 29 converts the permutation realization frequency 55 into a relative realization frequency and calculates the permutation entropy. Further, h ^* (n) in equation (7) is calculated. The permutation entropy H (n) or h ^* (n) is stored in the data holding unit 5 (permutation entropy data 56).

The average processing unit 30 averages the permutation entropy H (n) or h ^* (n) calculated from a plurality of time-series data at a predetermined time. It may be a moving average of permutation entropy H (n) or h ^* (n). If the time-series data for a predetermined time is sufficiently long, the averaging process may not be performed.

The average value of the permutation entropy H (n) or h ^* (n) is input to the failure determination unit 26. The malfunction determining unit 26, a predetermined threshold value and permutations compares the entropy average value of H (n) or h ^* (n), the mean value of the permutation entropy H (n) or h ^* (n) is the threshold It is determined that the target equipment is in a state to be warned when it has changed beyond.

As described above, when the permutation entropy in the normal state is probabilistic, it is a state to be warned when the average value of the permutation entropy H (n) or h ^* (n) becomes smaller than the threshold value. Is determined. Further, when the permutation entropy in the normal state is deterministic, it is determined that the state should be warned when the average value of the permutation entropy H (n) or h ^* (n) becomes larger than the threshold value. To do.

The display processing unit 27 displays, for example, the transition of the average value of the permutation entropy H (n) or h ^* (n) and the failure determination result on the display device 7. At the same time, the data is output to the printer 8. The display processing unit 27 may display time series data, permutation entropy H (n) or h ^* (n) together. When it is determined that the state should be warned, an alarm display such as blinking a light or sounding a buzzer may be performed.

  Also in the second embodiment, the physical configuration of the facility diagnosis system 1 is the same as that of the first embodiment. The facility diagnosis system 1 is shown, for example, in the block diagram of the physical configuration in FIG. The data acquisition unit 21, the embedded vector generation unit 22, the realization frequency calculation unit 28, the permutation entropy calculation unit 29, the average processing unit 30, the failure determination unit 26, and the display processing unit 27 are executed on the control unit 11 and above. Realized programmatically.

  Regarding the facility diagnosis system 1 of the second embodiment, for example, the configuration of the facility diagnosis system 1 is shown as a case where the compressor 3 of FIG. A microphone 2 for inputting the sound of the compressor 3 is connected to the equipment diagnosis system 1. The equipment diagnosis system 1 samples and collects the acoustic signals input from the microphone 2 at a certain period, and sets the time series data.

  If the compressor 3 is in steady operation, the time series data of the acoustic signal is considered to be stochastic when it is normal. Therefore, when the column entropy calculated from the time-series data of the acoustic signal becomes smaller than a certain threshold value, it can be determined that the compressor 3 should be warned, for example, has failed.

  Next, the operation of the equipment diagnosis system 1 when permutation entropy is used as a value representing determinism will be described. As described above, the operation of the equipment diagnosis system 1 is performed by the control unit 11 in cooperation with the main storage unit 12, the external storage unit 13, the operation unit 14, the screen display unit 15, the print output unit 16, and the transmission / reception unit 17. Do it.

  FIG. 9 is a flowchart illustrating an example of operation of equipment diagnosis using permutation entropy. First, time-series data indicating the state of equipment to be diagnosed is input (step C1). In the example of the compressor 3 in FIG. 3, the acoustic signal input from the microphone 2 is sampled and collected at a certain period to obtain time series data.

  From the collected time series data, an embedding vector r (t) is generated with a set embedding dimension n and a time difference Δt (step C2). The elements are ordered according to the size relationship of the elements of the embedding vector, and the number of embedding vectors having the same order is totalized as the realization frequency for all embedding vectors for a predetermined time (step C3).

The permutation realization frequency is converted to the relative realization frequency, and the permutation entropy is calculated (step C4). Further, h ^* (n) in equation (7) is calculated. Then, the permutation entropy H (n) or h ^* (n) calculated from a plurality of time-series data at a predetermined time is averaged (step C5). As described above, the averaging process may be omitted.

A failure is determined from the average value of the permutation entropy H (n) or h ^* (n) (step C6). The operation for failure determination is the same as that in the first embodiment, and FIG. 5 shows an example of a flowchart. In the second embodiment, only the permutation entropy H (n) or h ^* (n) is used in place of the translation error, and the description of the failure determination operation is omitted.

  Returning to the flowchart of FIG. 9, if it is determined as a failure as a result of the failure determination (step C7; Yes), it is displayed on the display device 7 that the facility is failed (step C8). At the same time, it may be printed out to the printer 8. When it is not determined that there is a failure (step C7; No), it is not displayed as a failure. When the determination value tends to approach the failure determination threshold value from a normal value, an alarm for prompting maintenance may be displayed.

(Example 2)
FIG. 10 is a graph showing an example of permutation entropy of the sound data of the compressor 3. 6 shows the result of calculating permutation entropy for acoustic signals before and after bearing replacement in FIG. 6 by changing the embedding dimension with the embedding dimension on the horizontal axis and the normalized permutation entropy h ^* (n) on the vertical axis. . In any case, the time difference Δt is 5 sampling periods. FIG. 7A shows the normalized permutation entropy h ^* (n) for the acoustic signal after the bearing replacement (After), and B shows the normalized permutation entropy h ^* (for the acoustic signal before the bearing replacement (Before). n).

In the range where the embedding dimension is small, the difference between A and B is small, but in the range where the embedding dimension is 6 to 8, the difference between A and B is clear. When an appropriate threshold value is set and the permutation entropy H (n) or h ^* (n) is greater than the threshold value, normal, and the permutation entropy H (n) or h ^* (n) is less than or equal to the threshold value Can be determined as a failure. In this example, it is desirable to set the embedding dimension to 6 or 7.

  In FIG. 10, the difference between A and B is small when the embedding dimension is 9 or more, and both are small values because the embedding dimension is large with respect to time-series data for a predetermined time, This is because the number is not enough. When the embedding dimension or the time difference is increased, the number of embedding vectors generated is relatively small, and the number of permutations that can be generated (n!) Is increased. The number needs to be large enough.

  According to the facility diagnosis method and the facility diagnosis system 1 of the present invention, the deterministic structure behind the time-series data is quantified even in the case where the power spectrum is the same for different time-series data at first glance. By reconstructing the trajectory in the phase space from the time series data by the delayed coordinate transformation, the difference between the time series data at the normal time and the time of the failure can be quantitatively expressed. A highly accurate equipment diagnosis can be performed even with a relatively small number of data.

(Application of the embodiment)
FIG. 11 shows an example of the configuration of the facility diagnosis system 1 when a semiconductor manufacturing apparatus is the object of diagnosis. The semiconductor manufacturing apparatus of FIG. 11 is a plating processing apparatus 4 including a semiconductor substrate cleaning apparatus.

  The plating apparatus 4 includes a cassette station 41 and a processing station 42. The cassette station 41 carries wafers supplied from the outside in units of wafer cassettes from the cassette 43 to the plating apparatus 4, or carries out the plated wafers from the plating apparatus 4 to the cassette 43.

  The cassette station 41 is provided with a cassette mounting table 44, and a wafer cassette 43 containing a wafer to be plated is supplied from the outside. In the cassette mounting table 44, the plated wafer is stored in the cassette 43 for carrying out.

  The wafer is transferred on the cassette mounting table 44 described above by the transfer mechanism 45. The transfer mechanism 45 can move in the x-axis direction (direction perpendicular to the paper surface) and can be moved up and down in the z-axis direction so that a plurality of wafer cassettes 43 placed on the cassette mounting table 44 can be accessed. It is. Further, the wafer can be rotated around the z axis so that the wafer can be transferred from the processing station 42 to the cassette mounting table 44. The transport mechanism 45 includes a displacement sensor and an acceleration sensor that detect displacement in the x-axis direction and the z-axis direction. In addition, a rotation sensor and an acceleration sensor for detecting a rotation angle around the z axis are provided.

  The cassette station 41 and the processing station 42 keep the internal atmosphere clean by downflow of clean air. The plating apparatus 4 includes a temperature / humidity sensor and a wind speed sensor in order to keep the atmosphere state constant.

  The processing station 42 includes a plurality of plating processing units 46 that perform plating processing on a wafer one by one and a plurality of cleaning / drying units 47 that perform cleaning and drying after the plating processing at predetermined positions.

  In the plating unit 46, the wafer on which the seed layer is formed is subjected to a plating process, for example, a Cu thin film is formed on the wafer. The plating apparatus 4 includes a temperature sensor for detecting the temperature of the solution and a liquid flow rate sensor for controlling the flow rate of the chemical solution. In addition, an optical fiber sensor for detecting the liquid level of the solution, a pressure sensor for the chemical pump, or an acceleration sensor is provided.

In the cleaning / drying unit 47, the front surface, back surface, and peripheral edge of the plated wafer are cleaned with a cleaning liquid such as a chemical solution or pure water, and after cleaning, the wafer is rotated at high speed under N ₂ purge to dry the wafer. The cleaning / drying unit 47 includes a temperature sensor for adjusting the temperature of the chemical solution, a liquid flow rate sensor, an optical fiber sensor for detecting the liquid level, a pressure sensor for the chemical solution pump, and an encoder as a rotation sensor for detecting the rotation speed of the wafer. .

  The displacement sensor, the rotation sensor, the acceleration sensor, the temperature / humidity sensor, the wind speed sensor, the temperature sensor, the liquid flow sensor, the optical fiber sensor, and the pressure sensor are controlled in order to control the plating apparatus 4 as a semiconductor manufacturing apparatus. Connected to the device. At the same time, the outputs of these sensors are branched and input to the equipment diagnosis system 1.

  You may connect to the equipment diagnostic system 1 via a network using the apparatus which transmits the output of each sensor. By connecting the facility diagnosis system 1 via a network, a single facility diagnosis system 1 can perform failure diagnosis of a plurality of facilities. In addition, it is possible to perform equipment failure diagnosis from a remote location.

  The equipment diagnosis system 1 calculates the translation error or permutation entropy, which is a value representing determinism, using the signals input from each sensor as time series data, and when these values change beyond a threshold value, It is determined that the target unit should be warned.

  For example, a temperature sensor, a temperature / humidity sensor, a wind speed sensor, an optical fiber sensor, a rotation sensor for high-speed rotation of a wafer are controlled so as to maintain a constant value, and thus fluctuations in output are probabilistic. Conceivable. When the time series data becomes deterministic, that is, when the translation error or permutation entropy falls below a threshold value, it can be determined that the state should be warned.

  Further, for example, a displacement sensor, a rotation sensor and an acceleration sensor of a transport mechanism, a pressure sensor for a chemical liquid pump, and the like detect a movement during operation, so that time series data in that case is considered deterministic. When the time series data has a probabilistic tendency, that is, when the translation error or the permutation entropy exceeds a threshold value, it can be determined that the transport mechanism or the chemical pump should be warned.

  As described above, according to the facility diagnosis system 1 of the present invention, it is not necessary to provide a special sensor for failure diagnosis like the semiconductor manufacturing apparatus of FIG. Equipment diagnosis can be performed by inputting and analyzing time-series data from the output of the sensor. Moreover, it can be diagnosed whether it is a state which should warn in the operating state of an installation.

  In addition, the above-described hardware configuration is an example, and can be arbitrarily changed and modified.

  The central part that performs equipment diagnosis processing composed of the control unit 11, the main storage unit 12, the external storage unit 13, the operation unit 14, the screen display unit 15, the print output unit 16, the transmission / reception unit 17, the internal bus 10, etc. It can be realized using a normal computer system regardless of a dedicated system. For example, a computer program for executing the above operation is stored and distributed in a computer-readable recording medium (flexible disk, CD-ROM, DVD-ROM, etc.), and the computer program is installed in the computer. Thus, the facility diagnosis system 1 that executes the above-described processing may be configured. Alternatively, the facility diagnosis system 1 may be configured by storing the computer program in a storage device included in a server device on a communication network such as the Internet and downloading the computer program from a normal computer system.

  In addition, when the functions of the facility diagnosis system 1 are realized by sharing an OS (operating system) and an application program, or by cooperation between the OS and the application program, only the application program portion is stored in a recording medium or a storage device. May be.

  It is also possible to superimpose a computer program on a carrier wave and distribute it via a communication network. For example, the computer program may be posted on a bulletin board (BBS, Bulletin Board System) on a communication network, and the computer program distributed via the network. The computer program may be started and executed in the same manner as other application programs under the control of the OS, so that the above-described processing may be executed.

It is a block diagram which shows the logical structure of the equipment diagnostic system which concerns on Embodiment 1 of this invention. It is a block diagram which shows an example of a physical structure of an equipment diagnosis system. It is a figure which shows the example of a structure of the equipment diagnostic system in case a compressor is made into the object of a diagnosis. It is a flowchart which shows an example of the operation | movement of the equipment diagnosis using a translation error. It is a flowchart which shows an example of operation | movement of a failure determination. It is a graph which shows the example of the time series data of the acoustic signal of a compressor. It is a graph which shows the example of the translation error of the sound data of a compressor. It is a block diagram which shows the logical structure of the equipment diagnostic system which concerns on Embodiment 2 of this invention. It is a flowchart which shows an example of operation | movement of the equipment diagnosis using permutation entropy. It is a graph which shows the example of the permutation entropy of the sound data of a compressor. It is a figure which shows the example of a structure of the equipment diagnostic system in case a semiconductor manufacturing apparatus is made into the object of a diagnosis.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Equipment diagnostic system 2 Microphone 3 Compressor 4 Plating processing apparatus (semiconductor manufacturing apparatus)
DESCRIPTION OF SYMBOLS 5 Data holding part 7 Display apparatus 8 Printer apparatus 10 Internal bus 11 Control part 12 Main memory part 13 External storage part 14 Operation part 15 Screen display part 16 Print output part 17 Transmission / reception part 21 Data acquisition part 22 Embedded vector generation part 23 Proximity Vector extraction unit 24 Translation error calculation unit 25 Median value calculation / average processing unit 26 Failure determination unit 27 Display processing unit 28 Realization frequency calculation unit 29 Permutation entropy calculation unit 30 Average processing unit 51 Collected time series data 52 Embedded vector data 53 Nearest neighbor Vector data 54 Translation error data 55 Permutation realization frequency data 56 Permutation entropy data

Claims (11)

A data acquisition step for acquiring time-series data that is a variable indicating the state of equipment to be diagnosed and fluctuates with time;
A deterministic calculation step for calculating a value representing determinism that serves as an indicator of whether the time series data in the facility is deterministic or probabilistic from the time series data acquired in the data acquisition step. When,
A warning determination step for determining that the equipment should be in a warning state when a value representing the determinism calculated in the deterministic calculation step changes beyond a predetermined threshold;
A facility diagnosis method comprising: The value representing the determinism is a translation error calculated from the time series data,
The deterministic calculation step includes
An embedding step for calculating an embedding vector of a certain dimension from the time series data;
A nearest neighbor vector extracting step of extracting a predetermined number of closest vectors for a certain embedded vector among the embedded vectors calculated in the embedding step;
A translation error calculating step for calculating a translation error that is a variance of a predetermined number of the closest vectors extracted in the nearest vector extraction step;
Including
The warning determination step includes:
When the translation error in the normal state of the equipment is larger than the predetermined threshold, the equipment warns when the translation error calculated in the translation error calculation step becomes smaller than the predetermined threshold. When the translation error in the normal state of the equipment is smaller than the predetermined threshold value, the translation error calculated in the translation error calculation step is greater than the predetermined threshold value. When it becomes larger, it is determined that the equipment should be warned.
The equipment diagnosis method according to claim 1, wherein: The value representing the determinism is permutation entropy calculated from the time series data,
The deterministic calculation step includes
An embedding step for calculating an embedding vector of a certain dimension from the time series data;
For all the embedding vectors calculated from the time-series data at the predetermined time calculated in the embedding step, the frequencies in the same order as the order of the magnitude relationship of the elements of the embedding vector are accumulated, and the predetermined time A realization frequency calculating step for calculating as a relative realization frequency with respect to the number of all the embedded vectors calculated from the time series data in
A permutation entropy calculating step for calculating permutation entropy, which is an entropy having a probability distribution of the relative realization frequencies calculated in the realization frequency calculation step, with all permutations consisting of the order of the number of dimensions of the embedded vector as random variables; ,
Including
The warning determination step includes:
When the permutation entropy calculated in the permutation entropy calculating step is smaller than the predetermined threshold when the permutation entropy in the normal state of the equipment is larger than the predetermined threshold, the equipment warns When the permutation entropy in the normal state of the equipment is smaller than the predetermined threshold value, the permutation entropy calculated in the permutation entropy calculation step is greater than the predetermined threshold value. When it becomes larger, it is determined that the equipment should be warned.
The equipment diagnosis method according to claim 1, wherein:   The variable indicating the state of the equipment to be diagnosed is a time sensor of temperature sensor, temperature / humidity sensor, wind speed sensor, liquid flow sensor, optical fiber sensor, pressure sensor, rotation sensor, acceleration sensor, acoustic sensor or vibration sensor. 4. The facility diagnosis method according to claim 1, wherein at least one of them is included. 5. The equipment to be diagnosed comprises a rotating machine,
The data acquisition step acquires an acoustic signal as time-series data from the equipment with an acoustic sensor,
The warning determination step determines a state of the rotating machine that should be warned.
The facility diagnosis method according to any one of claims 1 to 3, wherein A data acquisition means for acquiring time-series data that is a variable indicating a state of equipment to be diagnosed and fluctuates with time;
Deterministicity calculating means for calculating a value representing determinism as an index of whether the time series data in the facility is deterministic or probabilistic from the time series data acquired by the data acquisition means When,
Warning determination means for determining that the equipment should be warned when a value representing determinism calculated by the deterministic calculation means changes beyond a predetermined threshold;
An equipment diagnosis system comprising: The value representing the determinism is a translation error calculated from the time series data,
The deterministic calculation means is
Embedding means for calculating an embedding vector of a certain dimension from the time series data;
Of the embedded vectors calculated by the embedding means, a nearest neighbor vector extracting means for extracting a predetermined number of closest vectors for a certain embedded vector;
A translation error calculating means for calculating a translation error which is a variance of a predetermined number of closest vectors extracted by the nearest neighbor vector extracting means;
Including
The warning determination means includes
When the translation error in the normal state of the equipment is larger than the predetermined threshold, the equipment warns when the translation error calculated by the translation error calculation means becomes smaller than the predetermined threshold. When the translation error in the normal state of the equipment is smaller than the predetermined threshold value, the translation error calculated by the translation error calculation means is greater than the predetermined threshold value. When it becomes larger, it is determined that the equipment should be warned.
The equipment diagnosis system according to claim 6. The value representing the determinism is permutation entropy calculated from the time series data,
The deterministic calculation means is
Embedding means for calculating an embedding vector of a certain dimension from the time series data;
For all the embedding vectors calculated from the time-series data at the predetermined time calculated by the embedding means, the frequencies in the same order as the order of the magnitude relation of the elements of the embedding vector are accumulated, and the predetermined time Realization frequency calculation means for calculating as relative realization frequency for the number of all the embedded vectors calculated from the time series data in
Permutation entropy calculating means for calculating permutation entropy, which is entropy with all the permutations consisting of the order of the number of dimensions of the embedded vector as random variables and the relative realization frequency calculated by the realization frequency calculating means as a probability distribution; ,
Including
The warning determination means includes
When the permutation entropy calculated by the permutation entropy calculating means is smaller than the predetermined threshold when the permutation entropy in the normal state of the equipment is larger than the predetermined threshold, the equipment warns Or when the permutation entropy in the normal state of the equipment is smaller than the predetermined threshold, the permutation entropy calculated by the permutation entropy calculating means is greater than the predetermined threshold. When it becomes larger, it is determined that the equipment should be warned.
The equipment diagnosis system according to claim 6.   The variable indicating the state of the equipment to be diagnosed is a time sensor of temperature sensor, temperature / humidity sensor, wind speed sensor, liquid flow sensor, optical fiber sensor, pressure sensor, rotation sensor, acceleration sensor, acoustic sensor or vibration sensor. The equipment diagnosis system according to any one of claims 6 to 8, wherein at least one of them is included. The equipment to be diagnosed comprises a rotating machine,
The data acquisition means acquires an acoustic signal as time series data from the equipment with an acoustic sensor,
The warning determination means determines a state of the rotating machine to be warned;
The equipment diagnosis system according to any one of claims 6 to 8, wherein To diagnose the condition of the equipment,
A data acquisition means for acquiring time-series data that is a variable indicating a state of equipment to be diagnosed and fluctuates with time;
Deterministicity calculating means for calculating a value representing determinism of the time series data in the equipment from the time series data acquired by the data acquisition means,
Warning determination means for determining that the equipment should be in a warning state when a value representing determinism calculated by the deterministic calculation means has changed beyond a predetermined threshold;
A computer program that functions as a computer program.